UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A knowledge-based framework for construction methods selection Al-Hammad, Ibrahim A. 1991

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1991_A1 A42.pdf [ 17.14MB ]
Metadata
JSON: 831-1.0050498.json
JSON-LD: 831-1.0050498-ld.json
RDF/XML (Pretty): 831-1.0050498-rdf.xml
RDF/JSON: 831-1.0050498-rdf.json
Turtle: 831-1.0050498-turtle.txt
N-Triples: 831-1.0050498-rdf-ntriples.txt
Original Record: 831-1.0050498-source.json
Full Text
831-1.0050498-fulltext.txt
Citation
831-1.0050498.ris

Full Text

A K N O W L E D G E - B A S E D F R A M E W O R K FOR C O N S T R U C T I O N M E T H O D S S E L E C T I O N by Ibrahim A. Al-Hammad B.A.Sc, King Fahad University of Petroleum and Minerals, Saudia Arabia, 1981 M.A.Sc, The University of Colorado, U.S.A., 1985 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF D O C T O R OF PHILOSOPHY in T H E FACULTY OF GRADUATE STUDIES DEPARTMENT OF CIVIL ENGINNERING We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH COLUMBIA April 1991 © Ibrahim A. Al-Hammad, 1991 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. The University of British Columbia Vancouver, Canada Department of DE-6 (2/88) Abstract The objectives of this thesis are to investigate, formulate, and structure the problem of methods selection, and apply a Knowledge-Based Expert System (KBES) approach. A complete, conceptual KBES framework for the methods selection problem is proposed and selected aspects of i t were implemented using NExpert Object. Defined hierarchically, a conceptual method frame consists of the following attributes: design element, construction strategy, construction resources, and construction process model. The roles of the KBES control strategy are to f i r s t specify a method and then rank i t versus others. In so doing, the control strategy i s applied at two levels: a preliminary f e a s i b i l i t y level, and a detailed f e a s i b i l i t y level. The former i s used to reduce the number of available methods and rank them for processing by the latter. The preliminary f e a s i b i l i t y part constitutes declarative knowledge with high level premises. The detailed f e a s i b i l i t y level, develops the attributes of the method. This component contains empirical, analytical, and procedural knowledge that draws on the c i v i l engineering knowledge domains of design, analysis and construction. Because the notion of a frame i s a useful way of identifying the attributes of a construction method, a conceptual frame i s used throughout to demonstrate the build-up of the method attributes through preliminary, then detailed f e a s i b i l i t y . An expert system called CMSA (Construction Methods Selection Assistant) was developed to implement a subset of the proposed solution approach with Cut-and-Cover tunnelling as the problem domain. CMSA, as designed, constitutes a methods selection shell that can be applied to other domains. It entails a solution paradigm of Suggest, Design, Predict, and Analyze operators. CMSA incorporates previous experience (shallow knowledge) as well as algorithmic procedures (deep knowledge). Key elements central to CMSA knowledge base include risk, design technical f e a s i b i l i t y , resources compatibility, cost and time performance measures, and regulatory constraints. Allowance i s made for modelling project context variables. A range of geotechnical conditions were treated for the example problem domain. The KBES framework proposed for the methods selection problem shows promise for tackling this ill-structured problem, helping to organize site experience, and contributing to productivity improvement. i i i Contents Abstract i i Contents iv Figures v i i Tables v i i i Screens ix Listings ix Acknowledgement x Acronyms xi 1. Introduction 1 1.1 Background 1 1.2 Research Objectives and Methodology 2 1.3 Problem Domain 4 1.4 Organization of the Thesis 5 2. Literature Survey for Methods Selection Problem 9 2.1 Introduction 9 2.2 Construction Methods 9 2.2.1 Definition of Construction Methods 12 2.2.2 Terminology Used in the Thesis 16 2.3 Decision Making Model for Method Selection 20 2.3.1 Background 2 0 2.3.2 Simulation Techniques 20 2.3.3 Decision Analysis 22 2.3.4 Decision Support Systems (DSS) 23 2.4 Knowledge-Based Expert Systems 28 2.4.1 KBES Components 29 2.4.2 Expert Systems for Construction Management 33 2.4.3 KBES for Construction Methods Selection 35 3. Cut-and-Cover Methods in Soft Ground 45 3.1 Introduction 45 3.2 Tunnelling Background 46 3.3 Cut-and-Cover Tunnelling Alternatives 48 3.3.1. Background 4 8 3.3.2 Traditional Cut-and-Cover Tunnelling 50 3.3.3 Milano Cut-and-Cover Tunnelling 50 3.3.4 Major Operations Common to Cut-and-Cover Tunnelling 51 3.4 GWSS Alternatives 54 3.4.1 Common Types of GWSSs 57 3.5 Excavation Operations 67 3.6 Factors Affecting Methods Selection and Design 70 3.7 Cut and Cover Tunnelling Project Example 73 3.7.1 Background 73 3.7.2 Lagging and Excavation Construction Cycle 77 iv Contents 4. A KBES Framework for Methods Selection and Design 81 4.1 Introduction 81 4.2 A KBES framework for Method Selection 84 4.2.1 General 84 4.2.2 Methods Selection Defined 84 4.2.3 Methods Shell 89 4.2.4 Sketch of System Features and Operation 92 4.3 CMSA Development 106 4.3.1 Overview 106 4.3.2 Context Modelling 110 4.3.3 Preliminary Feasibility 115 4.3.4 Detailed Feasibility Level 130 4.4 CMSA Risk Component Development and Evaluation 149 5. CMSA Implementation 154 5.1 Introduction 154 5.2 NExpert Object Overview 156 5.2.1 Major NExpert Object Modules 156 5.2.2 NExpert Primitives 160 5.2.3 Viewing Knowledge Structure 167 5.2.4 The Inference Process 168 5.3 CMSA Implementation 174 5.3.1. CMSA Overview 174 5.3.2 Solution Paradigm and Knowledge Base 176 5.3.3 Knowledge Representation 187 5.3.4 Technical Feasibility Part 207 5.3.5 CMSA Chaining and Reasoning (Control Strategy) 214 6. The Prototype Example 219 6.1 Introduction 219 6.2 Example Problem Description 219 6.2.1 Session Start 220 6.2.2 Problem Context Specification 224 6.2.3 Modified Example 238 6.3 Risk Component Assessment Implemented 239 6.3.1 Introduction 239 6.3.2 NExpert Risk Implementation 6.3.3 Risk Routine 246 7. Conclusions and Recommendations for Further 250 7.1 Summary 250 7.2 Contribution of The Thesis 251 7.3 Further Research 252 Bibliography 255 v Contents Appendix A: Pressures and Moments Computation 262 A.l Introduction 262 A.2 Lateral Pressure Calculations 262 A. 3 Design Principles for Structural Members 267 Appendix B: Pile Driving Production Rate Derivation 272 B. l Introduction 272 B.2 Soil/Pile Friction Calculations 272 B. 3 Pile Driving Production Rate Estimation 278 Appendix C: Interviews 297 C. l Introduction 297 C.2 Minutes of Meeting with Dillingham Contractors 297 C.3 Minutes of Meeting with Quadra Construction 306 C. 4 Project Site V i s i t 308 Appendix D: CMSA Partial Listing and Miscellany 310 D. l Introduction 310 D.2 Partial Listing of CMSA Knowledge Base 311 D.3 Vibratory Hammer Selection Knowledge 331 D.4 Unit Cost Quotations 333 D.5 Sample Data Base Files 335 v i Figures Figure 2.1 Design and Construction Interaction 11 Figure 2.2 Construction Model Process 13 Figure 2.3 Overview of Classification System for Construction Technology 14 Figure 2.4 Example of Element, Attribute, 14 Figure 2.5 Suggested Data Structure for Selected Technology 28 Figure 2.6 Sample Element Activity Frame 38 Figure 2.7 Example of Knowledge Source 40 Figure 2.8 Labor Component Frame 42 Figure 2.9 Equipment Component Frame 43 Figure 2.10 Process Component Frame 43 Figure 3.1 Cost Comparison for Tunnelling Alternative versus Cut-and-Cover Alternative 48 Figure 3.2 Cut and Cover Tunnel Project Isometric 75 Figure 3.3 Barchart and Time-Space Diagram for the Seattle Project 76 Figure 4.1 Hierarchy of Construction Method Frame Attributes 85 Figure 4.2 Construction Methods Selection Shell 90 Figure 4.3 Construction Methods Selection System Process 94 Figure 4.4 Detailed Feasibility (Phase 2) 95 Figure 4.5 Steel Sheet Pile (SSP) Method Frame 107 Figure 4.6 Prototype Model 109 Figure 4.7 Soil Profile Scenarios 112 Figure 4.8 GWSS Frame Synthesis 117 Figure 4.9 CMSA Rule Execution Loop 119 Figure 4.10 Drive.c Routine Interface with CMSA 145 Figure 4.11 States of Nature for Methods Selection 150 Figure 4.12 Risk Assessment Tree Diagram 153 Figure 5.1 NExpert Object Open Al Environment Framework 159 Figure 5.2 NExpert Rule Construct 162 Figure 5.3 The Class and Object Hierarchy 166 Figure 5.4 Rules Perpendicular to Frames 169 Figure 5.5 Backward Chaining for Inference 171 Figure 5.6 NExpert Inference Framework 173 Figure 5.7 Knowledge Base Organization and Control Strategy 177 Figure 5.8 Implementation Solution Paradigm 178 Figure 5.9 Design Element Class Hierarchy 190 Figure 5.10 Design Element Instance Frame 191 Figure 5.11 Steel Sheet Pile Class in NExpert 192 Figure 5.12 Steel Sheet Pile Selection Rule 193 Figure 5.13 Steel Sheet Piles Database (SSP.NXP) 195 v i i Figures Figure 5.14 Construction Resource Class Hierarchy 196 Figure 5-15 Impact Hammer Element 197 Figure 5.16 Vibratory Pile Driver Element 198 Figure 5.17 Double Acting Hammer Database (DAAH.NXP) 199 Figure 5.18 Vibratory Hammer Database (VIBRO.NXP) 199 Figure 5.19 Impact Hammer Class in NExpert 200 Figure 5.20 Hammer Selection Rule in NExpert 201 Figure 5.21 Construction Strategy Class Hierarchy 205 Figure 5.22 Construction Process Model Class Hierarchy 206 Figure 5.23 Technical Feasibility Rule 208 Figure 5.24 Method Technical Feasibility i s True 212 Figure 5.25 Technical Feasibility Diagnostic Rule 213 Figure 5.26 CMSA Model of Chaining and Reasoning 217 Figure 6.1 Instantiation Tree 237 Figure 6.2 Risk Assessment Decision Tree for Steel Sheet Pile 242 Figure 6.3 Risk Framework Assessment Flow Chart 248 Figure A.l Pressure and Moments Envelopes 264 Figure A.2 Soil Profile for Two Soils Scenario 267 Figure B.l Hammer Blow Count versus Soil Resistance 281 Figure B.2 Hammer Blow Count versus Driving Depth 285 Figure B.3 Drive.c Routine Flow Chart 287 Figure B.4 Drive.c Routine Development Flow Chart 288 Tables Table 3.1 Partial Space of Design/Construction Elements for Cut-and-Cover Tunnel 53 Table 3.2 Seattle Project General Information 74 Table 4.1 Methods Selection Space for GWSS 87 Table 4.2 GWSS Project Context Data 97 Table 4.3 Soil Profile Input-Format 1 113 Table 4.4 Soil Profile Input-Format 2 114 Table 4.6 Hammers for Different Soils 137 Table 5.1 Truth Matrix for NExpert 164 Table 5.2 Knowledge Base Statistics 186 Table 6.1 Risk Assessment Data Input Summary 247 Table 6.2 "SSP_Risk.nxp" F i l e for SSP Alternative 249 Table A.l Soil Types Properties Employed in CMSA 263 Table A.2 Lagging Members 271 Table B.l Values for Angle of Internal Friction 274 Table B.2 Ultimate Skin Friction for Sands 274 Table B.3 "Out.out" Sample Output 284 Table C.l Sample Pile Driving Resources Unit Cost 307 v i i i Tables Table D.l Vibratory Pile Drivers Sizing 332 Table D.2 Crane Selection Format 333 Table D.3 Representative Resources Unit Costs 334 Screens Screen 6.1 Knowcess Hypothesis Command Menu 221 Screen 6.2 CMSA Overview Rule Network Window 221 Screen 6.3 CMSA Rule Network Window 222 Screen 6.4 GWSS Feasible Alternatives 225 Screen 6.5 Soil Profile Specification (1) 225 Screen 6.6 Soil Profile Specification (2) 225 Screen 6.7 Soil Profile Specification (3) 226 Screen 6.8 Soil Profile Specification (4) 226 Screen 6.9 Water Table Level Input 226 Screen 6.10 Hypothesis l lSelect_Suitable_Sheet_Pile" 228 Screen 6.11 Hypothesis nSelect_Suitable__HammerM 230 Screen 6.12 Selected_Hammers Class and i t s Dynamic Objects 230 Screen 6.13 Hammer Efficiency Input 232 Screen 6.14 "Drive.txt" Explanatory F i l e 232 Screen 6.15 "Resultsl.nxp" F i l e 235 Listings Listing B.l Drive.c Routine for Pile Driving 289 Listing C.l Extracted Rules of Thumb 3 04 Listing C.l Extracted Rules of Thumb (continued) 305 Listing D.l Partial Listing of CMSA 311 Listing D.2 Steel Sheet Data Base "SSP.nxp" 336 Listing D.3 Soldier Piles Data Base "HP_Pile.nxp" 337 Listing D.4 Struts Data Base "Strut.nxp" 338 Listing D.5 Lagging Data Base "Lag.nxp" 339 Listing D.6 Impact Hammer Data Base "Hammer.nxp" 340 Listing D.7 Vibratory Hammers Sample Data Base 341 ix Acknowledgements I am greatly indebted in large part of this thesis to my advisor Professor Alan Russell for his invaluable and incisive input and constructive criticism. Professor Russell provided priceless guidance throughout this research. My gratitude extends to my thesis supervisory committee members of Dr. Caselton, Dr. Sassani, and Dr. S.O. Russell. Special recognition extends to the Saudi Arabian Educational Mission and British Columbia Science Council for purchasing NExpert Object program and IBM 386 to conduct the research work. Special thanks to Stuart Brown of Dillingham Co. and John Simonett of Quadra Inc. for granting me interviews and exposing their experience. Reflecting back at my residence in campus, I have benefited greatly from a number of people. I would like to thank Ronald Yaworsky and Leon Phem for their assistance. Along with other friends, I extend my appreciation to Saad Al-Mubiyedh, Mohammad Al-Robesh, Abdul-Aziz A l - J a l l a l , Tariq Al-Faris, Bernardo de Castello, Rachid Nakeeb, and Malik Ranasinghe — for their support, encouragement, and friendship that I value. Everything I have achieved can be attributed to the love i and caring that my mother and family have provided. x Acronyms BC Backward Chaining CMSA Construction Methods Selection Assistant CS Construction Strategy CPM Construction Process Model CR Construction Resource DA Design Alternative DMM Decision Making Models DSS Decisions Support Systems PC Forward Chaining GWSS Ground Wall Support System IE Inference Engine KB Knowledge Base KBES Knowledge Based Expert Systems, lbc Lower Bound Cost for a Project or Activity lbpr Lower Bound Production Rate for a Project or Activity PZ Heavy Steel Sheet Piles SPL Soldier Piles and Lagging SPT Standard Penetration Test SSP Steel Sheet Pile STW Slurry Trench Wall ubc Upper Bound Cost for a Project or Activitye ubpr Upper Bound Production Rate for a Project or Activity. x i 1. Introduction 1.1 Background Construction methods selection i s a challenging problem. In general, there are numerous alternative methods for performing each major activity in a project. The number of methods available for each activity, and the potential for interaction among them, makes methods selection a complex process. In practice, decision makers rely on past experience from similar projects to provide solutions to the current ones. Given the selection of a method, traditional techniques, such as network analysis, simulation or decision analysis, can be used to predict time and cost performance. However, these techniques are evaluative, not generative, and do not incorporate heuristic knowledge ex p l i c i t l y . The strength of quantitative modelling techniques l i e s in their lower level prediction and optimization. A combination of both descriptive and procedural knowledge is essential to effective methods selection. Consequently, a computerized decision-making tool which embodies a Knowledge Based Expert System (KBES) i s worthy of investigation. Such a tool should integrate quantitative and qualitative assessments in order to produce and analyze acceptable solutions. Chapter 1. Introduction 2 The focus of this thesis i s on developing a conceptual framework for describing the methods selection problem and on identifying the roles that can be played by knowledge based systems. 1.2 Research Objectives and Methodology An extensive review of the literature revealed that no general statement and formulation of the methods selection problem has been developed. In fact, few researchers ever address the issue, preferring to focus on specific problems. In treating specific problems, research has been directed at applying operations research and systems analysis tools. L i t t l e attempt has been made to treat the problem as a design simile as opposed to an analysis one and incorporate construction knowledge in the cognitive process. Complicating the problem is the dimensionality of a construction method in terms of a large number of quantitative and qualitative attributes, the combinatorial problem of combining methods and the multiple c r i t e r i a used for evaluating a method. The main goals of this thesis are twofold: to develop a generalized statement and structure for the methods selection problem, and to demonstrate the applicability of a knowledge based approach to this problem. For the latter, a prototype expert system called CMSA (Construction Methods Selection Assistant) has been developed. Chapter 1. Introduction 3 This thesis addresses the methods selection problem in broad terms of organizing, structuring and formulating i t , and to propose and implement a KBES framework approach. An appropriate domain example of Cut-and-Cover tunnelling w i l l be used. A central research objective i s to contribute toward the representation of a construction methods design environment that can handle a wide variety of methods/strategy selection problems. The main function of the environment is to provide the construction user with a subset of feasible methods, given a project context, including preliminary values for design parameters of the short l i s t e d methods. Specific research objectives are as follows: 1. Develop a generalized description and structure for the construction methods selection problem; 2. Identify specific roles that a knowledge based system can f u l f i l l ; 3. Develop a detailed representation structure for describing individual methods; 4. Formulate a process of traversing alternate methods and pruning alternatives; . 5. Treat multiple decision c r i t e r i a , including time, cost, and risk; 6. Consider both heuristic and procedural knowledge; and 7. Develop a prototype system, using the context of Cut-and-Cover tunneling, to demonstrate and partially validate findings from objectives 1 through 6. Chapter 1. Introduction 4 The research methodology employed consists of a number of parts. F i r s t , an extensive review of literature was conducted to identify the state-of-the-art and to identify useful approaches to each of the objectives. Second, a specific project context was selected (Cut-and-Cover tunnelling) to bring specificity to the process and to provide a base which could be further generalized by examining other methods selection problems. Specific research methodologies are as follows: 1. Extensive literature review of previous approaches to the problem as well as conducting f i e l d interviews; 2. Specific project context selection (sheet p i l i n g GWSS alternative for the Cut-and-Cover tunnelling); 3. Devising a wholistic definition for the construction method attributes which serves as a basis for structuring the problem; 4. Employing KBES Techniques, including knowledge acquisition, knowledge representation schemes, and control strategy design; and 5. U t i l i z i n g an Expert Systems shell (NExpert Object) as a basis for a working prototype to demonstrate v i a b i l i t y of suggested approach. 13 Problem Domain Cut-and-Cover tunnelling has been chosen as a vehicle to explore and structure the methods selection problem. Its richness i s in terms of the number of design and construction alternatives available, the high interaction among construction a c t i v i t i e s , and i t s distinctive, Chapter 1. Introduction 5 discrete, repetitive construction a c t i v i t i e s . Further, the experiential knowledge associated with i t spans the geotechnical, structural and construction fields, which makes i t a good candidate for exploring a general problem solving framework. The poorly structured nature of the problem domain makes i t amendable to a KBES approach. Problem domain solving, within the thesis context, requires knowledge about soil-structure interaction to design the ground wall support system (GWSS); Cut-and-Cover construction techniques; sequencing and scheduling; optimizing resources; and general project management. 1.4 Organization of the Thesis The remainder of this thesis i s organized as follows. Chapter 2 examines previous work on construction methods selection. A start i s made toward setting out a general definition of the methods selection problem. Conventional approaches (decision analysis, operations research, simulation and so forth), and knowledge based approaches applied to methods selection problem modelling, are reviewed. The goal of this chapter i s to provide a comprehensive view of the state-of-art of previous attempts at modelling the construction methods selection problem. Chapter 3 examines Cut-and-Cover tunnelling, the selected problem domain, and related construction methods, with emphasis placed on the ground wall support system (GWSS) Chapter 1. Introduction 6 design and installation alternatives. Analytical and heuristic design and construction procedures are discussed. Cut-and-Cover tunnelling is a knowledge-rich problem domain that i s used as a vehicle to explore the methods selection problem through hypothetical examples. These examples are used to expose how excavation operations are interwoven with GWSS retaining system design attributes. Chapter 4 sets out a framework for structuring the methods selection problem. Design and construction tasks are analyzed for methods selection; a comprehensive method frame definition i s introduced; and a conceptual comprehensive KBES approach for method selection i s prescribed including system features and control strategy description. The control strategy ranks and synthesizes method alternatives by operating at two levels: a high level, or preliminary f e a s i b i l i t y that constitutes declarative knowledge; and detailed f e a s i b i l i t y that constitutes procedural knowledge. Evaluation c r i t e r i a based on costs are used to rank feasible alternatives at both levels. Key elements that make a method accepted or rejected at the two levels are decomposed into procedural and declarative knowledge categories of design, risk, compatibility, performance measures, and regulatory. A simplified approach to risk assessment is presented. Chapter 1. Introduction 7 Treatment of project context variables (state variables), particularly of s o i l conditions, is also discussed. Implementation of the model features described in the last part of chapter 4 in the form of an expert system, is described in chapter 5. The shell used i s NExpert Object. The main features of this shell are bri e f l y highlighted. Knowledge structure and control strategy for CMSA prototype development follow the conceptual model. Frames are used to represent method attributes of design element, construction strategy, construction related resources and construction process models. Rules are used for the control strategy, and to screen alternatives and represent s o i l profiles. Samples of knowledge constructs u t i l i z e d in CMSA are provided. Chapter 6 presents a detailed example of the CMSA prototype synthesizes a Cut-and-Cover tunnelling alternative of steel sheet piles. Conclusions and recommendations for further research are contained in chapter 7. A number of Appendices contain details of the algorithms and CMSA knowledge base. Appendix A contains formulas for pressure and moment calculations for the design of the ground wall support system (GWSS). Appendix B covers the mathematical derivations for s o i l resistance for impact pile driving hammers. They serve as an algorithm to predict the duration of pile driving for a single p i l e . This Chapter l . Introduction 8 information i s used in turn by the CMSA to predict a total production rate. Appendix C provides insights into the knowledge acquisition process followed. Covered are the formal and informal interviews conducted and site v i s i t s that contributed to the knowledge acquisition process. Appendix D supplies a l i s t i n g of the CMSA knowledge base and the data bases used. 2. Literature Survey for Methods Selection Problem 2.1 Introduction The literature review in this chapter i s divided into three parts. First, previous instances of methods selection problem attributes and definitions are summarized. Subsequently, the attributes and definition for a formal general methods selection problem are set out for this thesis. Second, operations research and systems analysis approaches to specific methods selection problems are examined. Third, recent work involving the application of expert systems to this problem i s explored. For the latter approach, i t i s assumed that the reader i s familiar with KBES concepts and terminology. 2.2 Construction Methods To r e s t r i c t the scope of the thesis problem, we assume that the design of the permanent f a c i l i t i e s i s fixed. Clearly, the kind of methods selection tool investigated in this thesis could be used by designers to better link the design and construction processes, as well as by contractors who seek to optimize their decisions given a design. Gray (1986) has argued the case for the former where he asserts that the concept of buildability suggests i t i s advisable to involve the contractor in the early stages of Chapter 2. Method Selection Problem and i t s Approaches 10 the design process. In essence, i t means that a deeper understanding of the methods used by contractors to analyze the problems and risks inherent in a design has to be achieved. Few construction contracts allow early contractor involvement in the planning and design stages. Traditionally, most contracts inhibit this in order to f a c i l i t a t e more competition among bidders. Occasionally, project contract documents may specify the construction methods to be used with provisions that permit the contractor to submit his methods for approval. More frequently, methods are seen as the purview of the contractor and contract documents are silent on the methods to be employed. In a conventional system, the contractor makes preliminary decisions regarding construction methods based on the information available during bid preparation. If the contractor i s successful in being awarded the job, previous decisions are reviewed and further decisions are taken in light of more complete information. In an idealized system, the contractor i s allowed to bring his construction methods knowledge to the early design stage. Figure 2.1 contrasts conventional versus idealized systems (Gray 1985). This figure illustrates interaction among relevant decision categories. For the conventional Chapter 2. Method Selection Problem and i t s Approaches 11 system, the arrows show a unidirectional influence of previous decisions on more recent ones with l i t t l e modification. For the idealized case, the arrows are b i -directional between the three decision categories. Approach Parties Conventional Idealized Planning and Design Construction Planning Construction Execution Feasibility Study Master Plan Design Alternatives Work Volume Contract Preparation Resource Allocation Productivity Planning Methods Selection Detailed Scheduling Procurement Subcontract Evaluation Different Site Conditions Change Orders Actual Productivity Owner, A/E, Creditor, CM, Contractor* CM, Contractors, Subconts., Sureties, Material Suppliers, Traders,.. CM, Contractor, Owner, A/E, Subcontractors, Sureties, Material Suppliers, Trades * involved in idealized system Figure 2.1 Design and Construction Interaction [Adapted from Gray 1985] Chapter 2. Method Selection Problem and i t s Approaches 12 The decision making process of methods selection tends to be evolutionary in nature, in which each decision i s limited by decisions made at earlier stages, as in a traditional approach, or dynamically, as in an i d e a l i s t i c approach. 2.2.1 Definition of Construction Methods (Previous Work) No universal definition of the construction methods selection problem has emerged. Most definitions given in the literature are context sensitive. In this section, the terminology commonly found in the literature i s reviewed. We incorporate i t in our general definitions on the methods selection problem as appropriate. Construction technology can be defined as the science of construction involving the judicious use of available materials, methods, and equipment including the necessary planning, preparation, and execution (Merritt 1976). Tatum described a construction technology classification system that includes a hierarchy of four parts: components, elements, attributes, and values (Tatum 1987 and 1988). Figure 2.2 presents a sketch of a proposed model for the hierarchical construction process as defined by Tatum (1987). Further, figure 2.3 shows the four major components: material and equipment resources, construction applied resources, project requirements and constraints, and construction processes (Tatum 1988). A synthesis of the Chapter 2. Method Selection Problem and i t s Approaches 13 f i r s t three components, Construction process, represents a model by which performance measures under different scenarios are derived. The attributes of a construction method can be further elaborated upon as shown in figure 2.4. ( CapLl ") Place Tools People ^ Owiwr ^ -^Coratniotor^" /ArehltocV ~N \jnflln««r J I Supply Resource* (Elements) Foreman Crew J Method Product Information Energy Materials Figure 2.2 Construction Model Process [from Tatum 1987] Chapter 2. Method Selection Problem and i t s Approaches 14 Construction -Applied Resources + Information + Skills + Equipment + Tools + General Conditions + Space + Energy + Time Materials and Permanent Equipment Resources Construction Processes + Methods + Tasks Project Requirements and Constraints Constructed Product Legend: • component of technology + element of technology Figure 2.3 Overview of Classification system for Construction Technology [from Tatum 1988] 1 Construction Methods + primary location + degree of automation'3 + degree of complexity3 + experience available3 + degree of interdependancy^ + point of origination + fundamental process + basic type + degree of uncertainty Possible Text Values of Attributes: fab shop, offsite, staging area, workface, yard crew, designer, planner, superintendent, vendor batch flow, Job shop, worker-paced assembly line hand tools, heavy equipment manual, precision Notes 1. element of the construction process component 2. text values for attributes that do not allow quantitative values 3. attribute has quantitative value ranging from 1 (minor) to 6 (extreme) + Indicates an attribute of the element Figure 2.4 Example of Element, Attribute, and Value [from Tatum 1988] Chapter 2. Method Selection Problem and i t s Approaches 15 Construction method has been defined as the manner in which resources on site are used to achieve specified forms of construction (Mansero 1987). In this reference, major elements identified as being part of the description of a construction method include: 1. a precise sequence of operations; 2. the relative pace of composite parts of operations; 3. interaction patterns with other operations; 4. construction plant; 5. expendable material and temporary works; 6. temporary services; and 7. craft s k i l l s . Although many factors must be considered in plant selection, there are basic principles that can be used to confine this problem. For example, in building construction, a crane i s considered to be a key resource for materials handling for major operations such as transporting forms, concrete, and other materials. For instance, Gray (1986) developed a systematic approach for the selection of a suitable crane for a high-rise building that embraces several factors. Among these factors are number and type of tower cranes versus mobile cranes, work load, required productivity rate, and crane work space and reach, given a f i n i t e number of available cranes. This provides a tangible Chapter 2. Method Selection Problem and i t s Approaches 16 example of structuring the knowledge pertaining to a complex problem of multi-task resource selection. 2.2.2 Terminology Used in the Thesis This section presents the definitions of a number of terms that form part of the overall construction methods selection problem. They are elaborated upon in subsequent chapters. We take a restrictive view that the design of the permanent f a c i l i t y i s fixed. Later, we acknowledge that contract provisions may consider other design alternatives or value engineering proposals. In addition, the requirements for temporary works, such as shoring and formwork, may become a significant design component. Design Approach: Includes dimensioning the structure and i t s elements, and specifying the types of materials in response to loads, required functions, site features, regulations, and so on. Construction Plan/Strategy: Represents the high level abstraction of major aggregate ac t i v i t i e s that are sequenced in a pre-determined logical manner to realize a design. The estimate of an activity duration i s dependant on the selected construction method and process. Chapter 2. Method Selection Problem and i t s Approaches 17 An example of a construction strategy in a high-rise building i s to proceed excavation activity downward, simultaneously with constructing the super-structure upward, in contrast to the traditional bottom up construction. Construction Process Model: Involves the interaction among equipment, labor and material, under physical and other project context constraints which characterize a major operation or construction cycle that i s used in analyzing construction process. For instance, in a reinforced concrete high-rise building, concrete placement for the superstructure can be modelled using CYCLONE (construction process simulation program, see Halpin 1976) to measure i t s progress rate which indirectly indicates the whole project progress pace. The construction process model combines design elements and construction resources according to a set of c r i t e r i a . Thus, i t could be used as a decision-making tool to rank, and accept or reject alternatives. Quantifiable measures for the model may include progress rate (productivity of the system), total cost, and duration. Qualitative variables may include the quality of work and safety during construction. Chapter 2. Method Selection Problem and i t s Approaches 18 Construction Method: Consists of a hierarchical assembly of design, strategy, resource, and process components that characterize a specific operation. For instance, a slurry trench wall (STW) method for a GWSS encompasses the following components. The design elements include a specified slurry type and density, and concrete mix design (temporary and/or permanent f a c i l i t y ) . The construction strategy i s bottom-up. The resources employed include a l l those involved in each operation, such as excavation, concrete placement, and retaining system installment. The process includes the arrangement of resource interaction, which constitutes a construction cycle. The construction cycle specifies a sequencing of the a c t i v i t i e s based on physical and resource interactions as well as other requirements imposed by the project context. Construction Methods Selection: A wholistic paradigm of methods selection consists of: Suggesting a preliminary method which consists of design elements, construction strategy, and construction resources, and construction process; Designing a the method in terms of specifying i t s attributes; Synthesizing a method by means of a model to predict i t s performance measures for the suggested method; Analyzing the performance measures versus expected goals, thereby accepting or rejecting a method alternative. Chapter 2. Method Selection Problem and i t s Approaches 19 "Recommendation" for a change in the method attribute values to alter a method that was rejected or to enhance method design effectiveness, i s included in the Analyze operator. In essence, each element of the method i s instantiated, and the resultant attributes values define the most suitable method based on project context variables. Chapter 2. Method Selection Problem and i t s Approaches 2 0 23 Decision Making Model for Method Selection 2.3.1 Background Considerable effort has been expended in applying and adapting various operations research and systems analysis tools to the problem of construction methods selection. In general, the approach has been to show how a specific problem might be analyzed using optimization (Gates and Scarpa 1980, and 1984), queuing theory (Ringwald 1987), dynamic programming (Gaarslev 1977, Selinger 1980), simulation, (Halpin 1976, Ashley 1980), and so forth: i.e. given a construction related problem, here i s how i t might be analyzed. Generalized approaches and definitions of methods selection which can be applied to a large class of problems, are the exception, not the rule. A notable exception i s the work of Halpin (Halpin and Woodhead 1976, Halpin and Bernold 1986), in which he has attempted to develop a simulation approach for treating a broad range of problems. In this section we review previous work directed at improving one's a b i l i t y to model and refine construction methods selection. 2.3.2 Simulation Techniques Simulation can be used in planning and scheduling highly repetitive cycles in a construction project. A construction oriented simulation called CYCLONE was developed by Halpin Chapter 2. Method Selection Problem and i t s Approaches 21 (Halpin and Woodhead 1976), and later refined (Riggs 1980, Halpin and Bernold 1986), extended and integrated with other systems such as INSIGHT (Paulson et a l . 1981 and 1987). CYCLONE is used for modelling at the construction operation level and i s especially useful for predicting the behavior of a construction cycle design, given the required input data of durations and methods (Halpin 1976). Sensitivity analysis can be used to show the impact of each major variable on productivity. Simulation applications in tunnelling are mostly used to predict the tunnelling advance rate, and cost breakdowns for major equipment items. Both deterministic and stochastic approaches have been used (Miller 1987 and Touran 1987). Performance measures for those methods are in terms of tunnel advance rate (feet/day) and total costs in general. Knowledge Based Expert Systems (KBES) and simulation can be combined to form a computer aided decision making tool for modelling processes. OxKeefe (1986) has explored the Expert Systems-simulation taxonomy and i t s application areas. Suitable applications include intelligent front ends for simulation packages which provide advice on how best to formulate and interpret the results from a simulation model. For example, Bernold (1987) showed how heuristic rules may be used in conjunction with CYCLONE to evaluate construction process scenarios. Chapter 2. Method Selection Problem and i t s Approaches 22 2.3.3 Decision Analysis Decision analysis i s a term used to describe a body of knowledge and professional practice for the logical illumination of decision problems (Howard 1983). The decision modelling procedure follows three sequential phases: deterministic, stochastic, and informational. These three phases are repeated until the value of additional information is less than the cost of obtaining i t . Decision analysis has been applied to methods selection (Gaarslev 1977, Ashley et a l . 1979 and 1983). It i s basically an analysis, as opposed to design tool. Ashley (1983) proposed c r i s i s decision analysis as a tool to aid the manager to select the most appropriate alternative relying primarily on his experience and intuition. This methodology i s bui l t on two bases: a decision tree and p o l i t i c a l conflict resolution. This model is supposed to be u t i l i z e d during construction at the strategic planning level. The tool, using project profit as the decision criterion, has been applied to select an alternative construction method for a hypothetical sewer tunnel during a c r i s i s dealing with encountered surface settlement. Ayyub and Haider (1985) proposed a decision analysis framework which considers information on relative risk, cost, benefits, and consequences of each construction strategy alternative. The decision criterion, safety of the Chapter 2. Method Selection Problem and i t s Approaches 2 3 construction operations as a function of a construction strategy, i s measured in terms of the completed structure's consequent probability of damage. The main factors that affect safety are identified as being qualitative. The states of these factors are quantified using fuzzy set theory to estimate the risk of construction failures. The best construction alternative i s the one with the minimum cost, with cost including i n i t i a l cost of the structure under construction and expected cost of structure failure. The strength of this approach l i e s in treating lin g u i s t i c terms. However, i t s weakness l i e s in i t s dependance on a single objective function (cost) based on safety factors only. Besides a decision analysis approach, decision making can employ other relatively new techniques including computer aided tools, such as decision support system (DSS). Since there i s some overlap between DSS and ES, this subject w i l l be investigated next. 2.3.4 Decision Support Systems (DSS) DSS can be defined as a computer-based decision aid that provides convenient access to decision models dealing with production, distribution, financial analysis, and so forth (Blanning 1984) . Both DSS and ES incorporate features of management informations systems (MIS) and operations research (OR) techniques. Chapter 2. Method Selection Problem and i t s Approaches 24 Turban and Watkins (1985) contrasted differences in approach between DSS and ES, and examined the possible connections between the two systems and the benefits of their integration. A major difference i s that an ES makes, rather than supports, decisions and contains a judgmental model rather than a causal model. Furthermore, an ES offers conclusions along with supporting j u s t i f i c a t i o n or an explanation using transferred expertise. A DSS helps a user evaluate and choose among alternatives based on the system model, mainly using the user's judgement and discretion. One can loosely think of a DSS as a quantitative (causal) modeling approach to a problem, whereas an ES i s regarded as a qualitative (judgmental) and quantitative modeling approach. DSS f o r Methods Selection: Within the construction industry, Mansero and Chapman (1987) argued that DSS provides the best means for methods selection for reinforced concrete structures rather than an ES, the output from which would be too prescriptive. They proposed a DSS system to model alternative ways of providing in-situ reinforced concrete framed buildings, with particular emphasis on selecting suitable formwork methods from a construction planning perspective. Kim (1984) proposed and implemented a DSS to select an adaptable tunnelling method (which adapts construction Chapter 2. Method Selection Problem and i t s Approaches 25 methods to encountered geological conditions) to optimize the design/construction methods selection for tunnels in rock. In his work, he deals with construction methods for excavation and wall support, including selection of equipment. He proposes a framework for generating decision support information in adaptable tunnelling and then derives analytical methodologies for the proposed DSS, employing stochastic dynamic programming (DP) algorithms. A relative confidence level i s used as a measure for ranking competing alternatives. Decision variables considered are excavation and support methods. State variables are geological conditions, and cost/time factors (productivity in terms of advance rate, equipment and material costs, and overhead costs). The objective function selected i s expressed in terms of stochastic variables describing geological conditions. Cost/time factors are treated as deterministic with constants derived from previous tunnelling projects. The objective of the optimization framework, in adaptable tunnelling, i s to identify the most cost-effective chain of excavation and support methods, each of which is technically feasible. At the same time, each must be economically feasible for the anticipated geological conditions in i t s subsequent tunnel segments (i.e. the tunnel i s divided into equal segments to reflect the changing geological conditions along the tunnel). Chapter 2. Method Selection Problem and i t s Approaches 26 The objective function (total construction cost), i s subject to two constraints. The f i r s t , technical f e a s i b i l i t y , refers to the a b i l i t y to employ an excavation method suited to the encountered ground conditions and a support method that satisfies structural and functional c r i t e r i a . The second constraint, economic optimization, refers to employing the most cost-effective combination of different methods along the tunnel, given geological variations. The features of Kim's work that are of direct relevance to the thesis work described herein are: 1. Planning decisions are divided into a pre-construction and a construction phase. 2. Output decision variables, for the pre-construction and construction phases, include feasible types of combined construction methods, level of confidence in selecting a method (the major criterion for ranking alternatives), expected loss for each method as an upper limit for the additional geotechnical exploration expense, and total cost and time for each alternative. 3. Input state variables encompass geological conditions and cost/time data for each method. Costs are defined as cost of equipment, material, labor, and mobilization/demobilization charges for methods changes. Time data include the productivity of each method versus a ground class, method change duration, and lead time as pre-determined data. 4. Variations of a construction method are treated as discrete methods. For instance, the model has been applied to select combined methods among five method variations of d r i l l and blast excavation/support alternatives, as opposed to Chapter 2. Method Selection Problem and i t s Approaches 27 competing with other major methods such as tunnel boring machine, or partial face boring. 5. The model does not consider important project attributes such as physical constraints, material handling, ground subsidence, and ground water control. Law (1987) proposed a conceptual DSS for the detailed design of construction act i v i t i e s associated with projects characterized by significant repetition. High-rise building was selected as the problem domain. A Work Breakdown Structure (WBS) was used to decompose ac t i v i t i e s into smaller work operations in order to select appropriate construction technologies. The attributes of construction alternatives consist of productivity, equipment and material unit costs, and crew makeup. The suggested data structure, see figure 2.5, does not incorporate some important attributes such as the technical f e a s i b i l i t y of those methods in terms of project physical constraints, f l e x i b i l i t y of the equipment used, and the level of confidence associated with each method. Law indicated, in the problem recognition section, that in the design of the construction a c t i v i t i e s , f i e l d engineers have to draw on their experience with similar previous projects to apply their knowledge of construction method (CM) selections. Such knowledge resides with a few key personnel and is rarely documented. It i s obvious that the DSS approach for methods selection addresses the quantitative aspects of the problem. Law stressed the Chapter 2. Method Selection Problem and i t s Approaches 28 importance of modelling the qualitative or judgmental part of construction methods assessment and selection. EFCO Floating Slab Formwork System Work Tasks: Literature File No.: SlbFm-012 Slab Formwork Dimension: 20 M by 10 M Bay ** INSTALL AND DISMANTLE FORM * " <all time in minutes > 01 Set roller support bracket ahead 02 Lower support brackets & form 03 Strip slab edge hand rail (1 end) 04 Attach tugger winch 05 Position rolling scaffold 06 Roll slab form out to pick-points 07 Crane hookup 4 points 08 Slab swing-out and reset 09 Clean and oil form 10 Install filler panel No of Units '84sf Unit Time Total Time 40 15 30 30 15 30 5 20 80 6/100sf 46 90 180 320 120 30 30 15 30 20 Comments 4 Men and Crane 2 Men Figure 2.5 Suggested Data Structure for Selected Technology [from Law 1987] 2.4 Knowledge-Based Expert Systems In this section, we identify the main components of a Knowledge Based Expert System and identify relevant work in the construction domain, especially methods selection, that relates to each component. Modelling of uncertainty in knowledge has not been explored — existing uncertainty methods include probabilistic methods (Bonissone 1985, Duda et a l . 1979), Chapter 2. Method Selection Problem and i t s Approaches 29 confirmation theory (Buchanan and Shortliff 1985), fuzzy set theory (Zadeh 1975) and Dempster-Shafer theory (Shafer 1976). While relevant to construction management problems, i t i s outside the scope of this thesis. However, the uncertainty in terms of outcomes of site conditions (especially of geological conditions) was treated. Gashing (1985) defines KBES as an interactive computer program incorporating judgement, experience, rules of thumb, intuition, and other expertise, to provide knowledgeable advice about a variety of tasks. 2.4.1 KBES Components A typical KBES has four major components; a Knowledge Base, consisting of Knowledge Representation and Acquisition; an Inference Engine; a Context; and an Explanation F a c i l i t y . 1. Knowledge Base (KB) The knowledge base (KB) contains a symbolic representation of expert rules of judgement and experience in a form that enables the inference engine to perform logical deductions upon i t . Such facts and rules are specific to the domain of the problem. D i f f i c u l t i e s in developing a KBES are attributed to knowledge representation and knowledge acquisition. Knowledge Representation : Knowledge Representation (KR) i s the set of syntactic and semantic conventions used to encode Chapter 2. Method Selection Problem and i t s Approaches 30 the facts and relationships that constitute knowledge in a knowledge based system (Winston 1986). Experience in developing KBES has shown that a robust, yet precise knowledge representation i s often the key to avoiding superficiality or shallowness in the solution of r e a l i s t i c problems (Jackson 1986). Selection of a KR technique i s a fundamental step in the application of ES to a problem. The KR process i s concerned with the problem of encoding the knowledge so that i t can be easily manipulated by the computer. In general the following elements must be represented: domain terms which deal with the language or jargon used by the expert in the f i e l d ; structural relationships which treat the interconnections of compound entities; and causal relationships which deal with cause-effect relations between components. KR techniques may include production rules (Buchanan and Shortliffe 1976), predicate logic (Clocksin and Mellish 1981), semantic nets (Minsky 1968, Duda et a l . 1978), frames (Minsky 1975), and object oriented programming (Bobrow and Stefik 1983, Goldberg 1981). Knowledge Acquisition : Knowledge Acquisition (KA) is the transfer, and the transformation, of problem-solving techniques from some knowledge source to a program (Buchanan and Shortliffe 1985). The major bottleneck in building an Chapter 2. Method Selection Problem and i t s Approaches 31 ES i s the scarcity of knowledge engineering s k i l l s to interact with one or more human experts. Several methods are used in the KA process. They include the use of unstructured interviews, structured interviews, prototype system development, rule induction, observation, and f i n a l l y machine learning of rules (Gruber and Cohen 1987). These methods embody theories and knowledge from computer science, psychology, linguistics, and sociology, in addition to technological expertise. 2. Inference Engine The inference engine (IE) i s the part of a KBES that contains the general problem-solving knowledge and i s characterized by strategies which draw inferences and control the reasoning process (Mikroudis 1986). The IE guides the development of a solution using the facts and rules stored in i t s KB and the information i t acquires from the user. Thus, the IE i s used to derive new facts from known facts and to regulate the order in which reasoning occurs. IE strategies used to make inferences include, but are not limited to, modus ponens, resolution, and inheritance. Control strategies include backward chaining, forward chaining, agenda control, mixed, and others (Charniak and McDermott 1985). Chapter 2. Method Selection Problem and i t s Approaches 32 3. Context The context i s a temporary data storage in which known and deduced facts are stored during a consultation session. The context builds up dynamically during the solution process of a particular problem. It i s used by the inference engine to determine the next step in the process. Data may come from, or go to, an extended data base, analysis/design programs, or even data acquisition devices. The inference tree provides a further mechanism for representing hierarchical relationships and for assigning values to object slots by instantiating. 4. Explanation F a c i l i t y and Others The explanation f a c i l i t y (EF) component serves to partially trace the ES reasoning process in order to ju s t i f y the conclusions made during a consultation. The two widely used commands are HOW and WHY. For a network of goals, rules, and hypotheses, HOW asks what rules were involved in solving the problem. WHY asks for the reasons some information i s requested by the system. In a goal driven ES, the HOW rule propagation direction i s from goals to the i n i t i a l states (backward chaining). The WHY direction i s vice versa (forward chaining). Chapter 2. Method Selection Problem and i t s Approaches 33 2.4.2 Expert Systems for Construction Management Construction engineering and management involves many complex decision-making problems, such as resource allocation, planning and scheduling, safety, analysis of construction processes, and productivity measurement and improvement. The solution of these are highly dependant on engineering and trade judgement, rules of thumb, and subjective evaluations. As stated in the previous section, construction management decision-making tools have traditionally employed quantifiable models (networks, OR techniques, etc.). Their strength l i e s in their rigorous analysis of the available data culminating in an optimal, or near optimal, solution to the problem. Their main weakness is the total dependance on the quantitative data necessary to represent the various relationships that describe the problem, many of which are imperfectly understood (Warszwaski 1986). When construction management decisions involve more qualitative information or relationships (experience, judgement, and intuition), or when multiple decision c r i t e r i a are present, the traditional approach i s of limited use, being more at the tactical than strategic level. Such limitations can be overcome to a certain degree by incorporating the experience, heuristics, and judgement of acknowledged experts into an ES. Chapter 2. Method Selection Problem and i t s Approaches 34 Since 1984 several papers, articles, and conference proceedings have been published that provide an overview of ES applications in Construction Management (Wager 1985, CIB-86 1987, Levitt 1987, Mohan 1990), and suggest further applications in this f i e l d (Chin 1987, Mohan 1990). Reviews of current ES applications have been reported recently (Wager 1986, Levitt 1987). Applications identified include: 1. construction project organization design (Zurich, Switzerland); 2. time estimating systems (Civil & Civic Australian construction firm); 3. repetitive construction risk analysis (University of Texas); 4. decision making and risk analysis (Georgia Institute of Technology); 5. intelligent construction risk identification systems (University of Texas); 6. layout of temporary construction f a c i l i t i e s (Stanford University); 7. evaluation of project personnel based on progress data from project time/cost monitoring systems (MIT); 8. vertical construction planning/scheduling (University of I l l i n o i s , and CMU); 9. project planning and control (Stanford University, Levitt et al 1988); 10. construction project monitoring (CMU); 11. maintenance advisor (PTY Ltd, Australian elevator construction and maintenance contractor); Chapter 2. Method Selection Problem and i t s Approaches 3 5 12. equipment and plant selection (University of Technology, Loughborough, U.K.). Other applications have been suggested by Trimble (1987), Warszwaski (1985), Chin (1987), Mohan (1990) as follows: 1. design synthesis and interpretation of building code regulations; 2. estimating procedures and cost control; 3. the analysis and evaluation of construction scheduling; 4. selection of appropriate plant and equipment; 5. site planning; 6. construction financing; 7. design and construction planning of prefabricated buildings; 8. quality control; 9. safety practices; 10. contractual claims analysis; and 11. evaluation of alternative construction methods at early design stages. 2.4.3 KBES for Construction Methods Selection Research work on the use of expert systems for methods selection has been tailored to specific applications. Generally, such systems can be classified as rule based, or frame and/or object oriented. Within the f i r s t category, some applications are directed at selecting a key resource, particularly equipment for a specific job. Gray and L i t t l e (1985) examined the influence of craneage resources required to l i f t large units in a Chapter 2. Method Selection Problem and i t s Approaches 36 high-rise building and the effect of multiple crane resources on the activity duration calculations. Subsequently, an expert system for craneage resource analysis was developed to select and locate the most desirable crane on the site for a high-rise and low-rise j building construction. Later, this expert system was imported into a more comprehensive planning rule-based expert system directed at determining a l l work ac t i v i t i e s implicit in the design of a high-rise building (Gray 1986). Activities were defined according to type of work (resource labelled: material, trade, plant), operationally significant function (direction of movement: vertical or horizontal), and operationally significant location (grouping ac t i v i t i e s of different sequence and size). Components of an activity's duration are work volume and resource level. The work volume is set by the design. Resource level i s variable. Resources are set in f a i r l y coarse groups, either gang or piece of plant, at the minimum level consistent with normal practice. Rules and heuristics were used to select an activity, link a c t i v i t i e s (precedence, and time links), and perform problem solution processes. This application shows how expert systems can be used to encode algorithms (network analysis), and heuristics for activity selections. Chapter 2. Method Selection Problem and i t s Approaches 37 In the same category, other applications include selecting a crane type (tower crane versus mobile crane) and size for high-rise building construction (Harris and Wijestundera 1987), and scraper equipment for road earthmoving, given the specific project conditions (Harris 1988). The former authors concluded that selecting construction equipment i s largely based on uncertain and intuitive knowledge, allowing only broad rules of thumb to be formulated. Moreover, they suggested that further applications should include output data and production information pertaining to plant and labor resource evaluation. In both examples, the knowledge base is largely heuristic in nature, and knowledge acquisition was essential for deriving the inference for problem solution. Also, in those applications, methods were represented as equipment for capital intensive projects. The second application category, frame based expert systems, includes work done by Logcher and Nay (1985), Kunz et a l . (1986), and Hendrickson et a l . (1988). Chief among those applications i s Construction Planex (Hendrickson et a l . 1988). Planex, a KBES for construction planning, i s used to plan modular high-rise buildings, including excavation, foundation, and structure construction. Planex starts with a design alternative as input which consists of several design elements (a footing, Chapter 2. Method Selection Problem and i t s Approaches 3 8 column, and beam for a modular reinforced-concrete building). The design element generates construction element activity frames (i.e excavation, pouring concrete, etc). Figure 2.6 shows a sample element activity frame created to describe the excavation activity required for a footing. The element activity i s identified by a code number using the extended MASTERFORMAT code. The f i r s t six slots define i t s designation and relevant parents of design elements and project a c t i v i t i e s . These are followed by slots for amount of work, unit-of-measure, crew, material-package, duration, and successor element a c t i v i t i e s . The crew attribute has been evaluated to excavation-foundation-05. Element-Activity p01 -•OO-bOO-1 O0-ca-02-220-10-01 SLOT VALUE is-a ea ea-name axcavatlon-column-fbotIng-01 ea-code 01-220-1^ 01 ea-of-DE p01 -sOO-MO-fOO-de-60-01 -01 parent-EA p01-sCO-b00-f00-ea-01-220-10 ea-of-PA p01-200-b00-(00-pa-10-60 amount-of-work 24.0 unit-of-measure cu-yd crew excavation-fbundation-05 material-package none duration 16 hours successors p01-s00-b00-f00-ea-02-220-10-02 Figure 2.6 Sample Element Activity Frame [from Hendrickson et al 1988] After element ac t i v i t i e s have been created, they are grouped into project a c t i v i t i e s based on selected Chapter 2. Method Selection Problem and i t s Approaches 39 technologies. The activ i t i e s are then sequenced and their duration estimated in order to develop the schedule. The knowledge base i s organized into a set of knowledge sources that represents rules, heuristics, and calculation functions. Decisions and computations undertaken during the planning process can be stored in any of the frames in the Planex hierarchy of frames and can be inherited upward and downward between design element, project activity, and activity element frames. When frames are created by the operator modules, relevant knowledge sources in the knowledge base w i l l be evaluated. Within a sequence of operations applied to create a construction plan, the selected technology operator uses heuristics related to s o i l and site information, resource productivity information and other factors, by activating relevant knowledge sources (KS) designated as KS-technology-xx-xx, to group element and project a c t i v i t i e s under an auxiliary group object that i s used to store the common technology choice. Figure 2.7 shows an example of a knowledge source, namely KS-Technology, for selecting excavating equipment pictured as a decision table, whereas i t is actually encoded as frames and production rules. This KS contains two project context conditions, three rules, and three actions. The second condition of "KS-water-level" i s an embedded knowledge source that has to be evaluated f i r s t . The third rule Chapter 2. Method Selection Problem and i t s Approaches 40 indicates that i f none of the previous two rules were fired, the appropriate technology i s "special machine". KS-Technology-Example Object Slot Op Value RULES soil-characterlstlcs soil-type Is hard true false notfireo KS-water-level Is wet false ture notfirea • then T then T then power-shovel selected clamshell selected special-machine selected Figure 2.7 Example of Knowledge Source [from Hendrickson et a l . 1988] According to the inference strategy implemented in Planex, element and project a c t i v i t i e s for the excavation are supposedly selected and assembled based on the plant selected from the KS-technology instantiating shown in figure 2.7. Task durations are estimated from decision tables and calculating rules in a manner similar to that used in MASON system (Hendrickson et a l . 1987). Precedences among element ac t i v i t i e s are also determined and recorded in slots of the element activity frames. These precedences can be of two types: physical or resource-related. Chapter 2. Method Selection Problem and i t s Approaches 41 For the Planex system, the authors indicated that determining the equipment to be used, the number of crews or pieces of equipment, inter-task precedence and task duration, involve diagnosis and prediction as contrasted to synthesis, in activity definition. This example shows that the way in which a method, mimicked as equipment selected based on project context, i s used to combine element and project a c t i v i t i e s among numerous alternatives, is essential in constructing the planning schedule. Logcher and Nay (1985) described a conceptual expert system for analyzing construction project risks. Hierarchical frames were used to represent project tasks, work packages, risks, resources, and additional s i t e -relevant information. Figures 2.8, 2.9, and 2.10 show a sample of the labor, equipment, and process frames that are to be created during a session. Their values w i l l be inherited by Work Package and Review Data frames for further manipulation of the project risk analysis. The frames presented show how construction related concepts could be represented by describing and detailing their attributes. In this application, frames were found to be a knowledge representation strategy capable of capturing relevant problem characteristics, while rule-directed inference was used to associate project risks with work packages. Chapter 2. Method Selection Problem and i t s Approaches 4 2 Labor type: <inclusion: Work Package, Review Data> union: <value> cost: <value> quality: <value> ( s k i l l and manpower required by work package) Productivity: <inclusion: work Package, Review Data> quantity: <value> productivity level: <value> (output/unit time) schedule: <value list> (regular and overtime hours/week) production rate: <value> (output/unit input) morale: <value> <exclusion: Work Package> Safety: <exclusion: Work Package> accidents: <value list> shutdowns: <value list> Figure 2.8 Labor Component Frame [from Logcher and Nay 1985] The foregoing expert systems sample applications show the diversity of the methods, selection literature. From the thesis viewpoint, each example addresses specific instances of methods selection. There i s a lack of a wholistic scheme for specifying and analyzing a method. The literature has showed how some expert systems have been incorporated and/or evolved into a larger intelligent system for planning, site layout, etc. In the same vein, i t i s conceivable to envision a generic tool that consists of a series of ESs, small and big, that are tied together to select and specify the most suitable method. Chapter 2. Method Selection Problem and i t s Approaches 43 Equipment type: general specification: <value list> <inclusion: Work Package, Review Data> description: date information supplied: <value> supplied by: <value> equip name: <value> rated capacity: <value> alternative equip type: <value> operating hrs. un t i l mainten.: <value> source: supplier: <value> <inclusion: Work Package, Review Data> <domain: Owned, Leased, Owned by Subcontractor cost: <value> <inclusion: Work Package, Review Data> quality: <value> <inclusion: Review Data> Productivity: <inclusion: Work Package, Review Data> production rate: <value> Rel i a b i l i t y : <value> <inclusion: Review Data> Figure 2.9 Equipment Component Frame [from Logcher and Nay 1985] Process <inclusion: Work Package> work package complexity: <value> company experience: <value> contractor: <value> experience: <value> Figure 2.10 Process Component Frame [from Logcher and Nay 1985] Chapter 2. Method Selection Problem and i t s Approaches 44 As a guide to general topics for KBES based method selection, areas of research interest include configuration of crews; choice of construction methods; man-machine tradeoffs; choice of transportation modes for the movement of materials, personnel and equipment; selection of optimum sizes,, configurations, and methods for joining various components in modular construction; and deep-excavation problems (Mohan 1990). Of particular value, i s the work done in the f i e l d of industrial engineering, specifically in manufacturing, which provides useful ideas for the definition of the methods attributes outlined in chapter 4. Those include a KBES approach for modelling a plant production plan (Reddy and Fox 1982), specification and design of flexible manufacturing systems (Mellichamp and Wahab 1987), creation of intelligent environments embodying simulation and a r t i f i c i a l intelligence, applied to simulation (Shannon 1986 and 1987), and automation of the model construction phase of the simulation l i f e cycle (Murray and Sheppard 1988). 45 3. Cut-and-Cover Methods in Soft Ground 3.1 Introduction A Cut-and-Cover tunnelling project has been selected as the problem domain in which to explore the methods selection problem because of i t s accessible knowledge base and established alternative methods. The goals of this chapter are to give the reader an example of Cut-and-Cover tunnelling, expose empirical and procedural knowledge for the design and construction of related ground wall support systems, and identify the kinds of decisions relevant to the methods selection problem. A general introduction to the tunnelling domain i s given in section 3.2. Section 3.3 covers the Cut-and-Cover tunnelling alternative in more detail including i t s characteristics. Section 3.4 reviews GWSS alternative design and construction procedures; in particular, steel sheet piles which i s used as the actual project context for the CMSA development. Section 3.5 covers the excavation operation briefly, since i t i s incidental to GWSS retaining system design variables. Section 3.6 examines the factors influencing the selection of Cut-and-Cover tunnelling alternatives. Section 3.7 provides a summary of a hypothetical soldier p i l e and lagging Cut-and-Cover tunnelling example, based on Chapter 3. Cut-and-Cover Methods 46 a real project, highlighting project context goals, constraints, and construction cycle decision variables. This example serves as a primary source for subsequent variations of Cut-and-Cover tunnelling throughout the thesis that are applicable to traditional (top-down) Cut-and-Cover tunnelling. 3.2 Tunnelling Background Tunnelling can be defined as the construction, by any method, of a covered cavity of pre-designed geometry whose fin a l location and use i s under the surface (Bickel and Kuesel 1982). Costs of tunnel construction include fixed and variable costs. Fixed costs are a function of design features, such as location, grade, types of wall support and fin a l tunnel structure, and the cost of special equipment, such as tunnel boring machines, including mobilization. Variable costs are a function of time, labor, u t i l i t i e s , supervision, and equipment. The two major approaches for constructing tunnels are underground tunnelling, such as using a tunnel boring machine (TBM), New Austrian Method, d r i l l and blast, and Cut-and-Cover tunnelling. Both approaches employ different design elements, construction strategies, construction resources, and construction process model. The decision to Chapter 3. Cut-and-Cover Methods 47 choose one approach over the other depends on costs including environmental and social costs, as well as project parameters such as duration and geotechnical factors. Figure 3.1 shows a comparison between costs of both construction approaches. This figure highlights the wide variance of both tunnelling alternatives costs, which is context sensitive especially because of the uncertainty of the geological conditions. It may appear from the figure that the cost of boring tunnelling excavation i s more than Cut-and-Cover excavation; however, their total costs might be the same. For shallow tunnels, the direct cost of this Cut-and-Cover method is l i k e l y to be much less than the cost of the tunnelling method. However, environmental and social costs can change the balance completely. With increasing depth, the direct cost of the trench excavation and temporary wall support system increases rapidly. Chapter 3. Cut-and-Cover Methods 48 Largest Dimension of Finished Cross-Section, ft Figure 3.1 Cost Comparison for Tunnelling Alternative versus Cut-and-Cover Alternative [from B ideal and Kuesel 1982] The dimensions of the tunnel depend on tunnel use and ground conditions. For instance, the width of the underground structure for a highway i s normally 35-65 feet, while for a rapid transit structure, i t i s 35 feet. The depth below ground varies from 10 feet to 100 feet. 33 Cut-and-Cover Tunnelling Alternatives 3.3.1 Background Based on the excavation method employed, Cut-and-Cover tunnelling can be categorized into traditional and Milano approaches. Each one embraces different method attributes Chapter 3. Cut-and-Cover Methods 4 9 of design elements, construction strategies, resources, and so forth. Before we describe these alternatives, we c l a r i f y the meaning of some of the terminology. Fi r s t , a construction strategy i s a function of physical structure (design elements) and project context (design features, site features, contractual provisions, regulatory environment, prevailing economic conditions and so forth) which entails the logical sequence of a set of ac t i v i t i e s . Each activity has several attributes: a work scope (amount of material to be added or removed), a construction technology, a duration, etc. An activity could be executed by means of operations which break i t down into smaller tasks that could be achieved by one or multiple resources, i.e., an operation designates a grouping of resources u t i l i z e d directly or indirectly to construct the design elements. A variation of task/resources make-up may produce several alternative operations. For instance, a s o i l stabilization activity is not required for the slurry trench wall design element while i t i s required for the soldier piles and lagging design element under a given project context. For the latter case, a s o i l stabilization alternative operation could include s o i l freezing, micro-piling, cement grouting, etc. Note that a construction strategy label could be a misnomer. For the two types of Cut-and-Cover tunnelling, Chapter 3. Cut-and-Cover Methods 50 the top-down and bottom-up (Milano) construction strategies refer mainly to the excavation progression that distinguish them. Next, we br i e f l y describe the two approaches. 3.3.2 Traditional Cut-and-Cover Tunnelling (Top-Down) Excavation i s the major activity to distinguish this approach from the other. A top-down construction strategy is used to designate this approach since i t implies that excavation proceeds from the ground surface to the bottom of the tunnel. For the traditional Cut-and-Cover tunnelling strategy, the operations involved and their sequencing are excavate and support the trench simultaneously, construct the tunnel structure, b a c k f i l l the trench and restore the surface. 3.3.3 Milano Cut-and-Cover Tunnelling (Upside-Down) A bottom-up or upside-down construction strategy refers to the fact that the excavation proceeds from the bottom of the tunnel to the ground surface. This method i s relatively new and associated only with the ground wall support system design element alternative of slurry trench wall (STW), although i t could be done by top-down excavation as well. STW and other GWSS techniques w i l l be reviewed later in section 3.4.1. Chapter 3. Cut-and-Cover Methods 51 The upside down or, so called, Milano strategy provides an alternative construction strategy which speeds up t r a f f i c restoration. Instead of excavating to the f u l l depth using temporary bracing, permanent steel may be installed as the excavation progresses. The major ac t i v i t i e s include: 1. construct STW, 2. excavate to the roof slab elevation, 3. construct the roof and use i t as permanent bracing, 4. start b a c k f i l l , and/or 5. excavate below the roof, and/or 6. construct base slab (invert) and complete the structure. 3.3.4 Major Operations Common to Cut-and-Cover Tunnelling Major operations during the construction phases Cut-and-Cover tunnelling are decomposed into the following categories: 1. stabilizing surrounding structures (underpinning), 2. ground sta b i l i t y , 3. ground water control, 4. ground wall support installation, 5. excavation, 6. permanent structure installation, Chapter 3. Cut-and-Cover Methods 5 2 7. u t i l i t y relocations, backfilling, and compaction. Each of the above operations employs different construction procedures (resources and construction process model) subject to the interactions between them. For example, the pace of excavating both horizontally and verti c a l l y i s bounded by the shoring retaining system spacing that provides structural s t a b i l i t y for the GWSS during construction. To give a flavor of the size and complexity of construction method attributes, consider table 3.1. It provides a subset of method attributes including variable design elements for the temporary GWSS, fixed design elements of the permanent tunnel structure, and alternate operations. Table 3.1 is set up so that each operation is independent and has several alternatives. Therefore, each combination of those attributes, vertically in the table, may represent a feasible alternative to a method embracing design/construction attributes. Hypothetical c r i t e r i a are provided to rank feasible alternatives based on one criterion or a mix of c r i t e r i a . For instance, consideration of multiple c r i t e r i a could be used to select Alternative (3) as the most favorable. Chapter 3. Cut-and-Cover Methods 53 Alternatives (1) (2) (3) . . 1. Design Components 1.1. GWSS Design Soil Nailing STW SP/Lagging 1.2. Tun. St. Des. Pre. Conc/RC STW Steel Ring 2. Activities Possible Construction Methods 2.1. Underpinning Shoring Grouting Micropil. 2.2. Gr.Water Cont. Dewatering Grouting Comp. Air 2.3. Ground Treat. Freezing Grouting Earth Ren. 2.4. GWSS Soil Nailing STW Sheet Pil. 2.5. Perm. Structure Prec. Cone. RC Steel Ring 2.6. Excavation Power Shovel Backhoe Clamshell 2.7. Backfill and Front Shovel Scraper Bulldozer Compaction and Vib. Tamp. and Sheep and Tamp.--foot -foot foot 3. Evaluation Criteria 3.1. Tun. Prog.Rate (ft.run/day) 6.5 8 10 3.2. Cost (CAN $/ft. run) 65K 80K 100K 3.3. Safety Excellent Good Average 3.4. Total Evaluation Criteria 5.5 7 9 Table 3.1 Partial Space of Design/Construction Elements for Cut-and-Cover Tunnel It i s useful to analyze the cost breakdown of the structural aspects of a Cut-and-Cover tunnel project in order to identify the most important cost components in the construction process. Shen (1981a) estimated that approximately 60% of the cost i s directly associated with the cost of the temporary ground wall support system structure and excavation. 80% to 90% of the cost of building the permanent tunnel structure is influenced significantly by decisions about the type of ground wall support system, including shoring system. Thus, any significant cost saving in the GWSS w i l l result in a Chapter 3. Cut-and-Cover Methods 54 substantial lowering of total project cost. It is therefore important to examine GWSS design and installation, and excavation operations more closely. The former w i l l be treated as a separable construction process with i t s associated resources, and the latter w i l l be treated as a method that embraces design elements and construction resources, while construction strategies and processes are implicitly discussed. Next, we examine common GWSS alternatives and treat their design features and construction procedures. Such procedures exhibit the analytic and heuristic knowledge required for construction method attributes synthesis. Some of those procedures w i l l be ut i l i z e d in the CMSA prototype development in chapters 4 and 5. 3.4 GWSS Alternatives The term GWSS, ground wall support system, i s taken to mean temporary and/or permanent support walls and the shoring schemes used to retain them. Alternatives for temporary (removable) and permanent support systems are as follows: 1. Temporary support systems 1. Soldier Piles with Lagging; 2. Sheet Piling (steel, timber, or concrete) ; or 3. Soil Nailing. 2. Temporary/Permanent support systems Chapter 3. Cut-and-Cover Methods 55 1. Slurry Trench Wall (cast in place, precast); 2. Secant piles (cast in place, precast); or 3. Any combination of the above. Alternative retaining systems include: 1. Wales and struts or rakers; 2. Compression rings (when excavation i s relatively small in plan); or 3. Tieback anchorages. In deep but narrow excavations, the preferred retaining system i s generally struts and wales, whereas for wide excavation areas, cross bracing and rakers, or similar shoring systems, are employed. Tie backs are popular in deep and wide cuts to provide obstruction free excavations (Winterkorn and Fang 1975). Other GWSS methods such as freezing and grouting might be used as the wall support system; however, they are not used frequently due to their high cost. They may be required, nevertheless, for localized water control or treatment. GWSS methods combined with stabilization methods are used to control s o i l and ground water movement during or after construction. Such movements may often cause settlement and damage to excavation support and adjacent installations. The design of the GWSS depends upon the ground conditions and the importance of preventing surface settlement. Settlement i s caused by movement of the s o i l surrounding the Chapter 3. Cut-and-Cover Methods 5 6 excavation as a result of bottom heave, inward movement of the GWSS or loss of ground through the GWSS. To evaluate the need for underpinning adjacent structures, the amount of settlement must be estimated for different types of GWSS methods. Cost trade-offs between cheaper, more flexible, GWSS and underpinning, should be assessed against more ri g i d but more expensive GWSS' which require a minimum of underpinning. The amount of settlement depends upon s o i l type, size of excavation, construction methods and quality of workmanship. The prediction of the amount of settlement i s uncertain and thus experience and observational data are needed as guides to judgment. Earth pressure envelopes which influence the design of the GWSS, represent the distribution of the cut pressure that i s exerted on the sides of the GWSS. They do not necessarily indicate the actual pressure distribution, but rather represent design pressure values that, i f used, can be expected to provide a safe and serviceable structure. The pressure envelopes assume that the installation of struts and bracing w i l l proceed as soon as the excavation reaches the level where they are to be placed. The method of construction is an important factor in determining the actual distribution of pressure on the GWSS. Recommended design pressure envelopes for use in calculating working lateral loads in sand and clay, are found in Peck Chapter 3. Cut-and-Cover Methods 57 (1984). These distributions are used later to determine the type of GWSS, retaining system, and installation method. 3.4.1 Common Types of GWSSs The sheeting part of the outside GWSS usually follows the physical dimensions of the tunnel, allowing just enough space to permit construction of the permanent structure. Common types of GWSS, retaining systems, and their installation methods aire described below. Emphasis i s placed on GWSS design alternatives of steel sheet piles and soldier piles and lagging. Specific examples of the aforementioned GWSS are used for analyzing Cut-and-Cover tunnelling operations and are used for CMSA implementation as revealed in chapters 3,4, and 5. 1. Sheet Piling The function of sheet pi l i n g (SP) i s to retain s o i l and ground water. SP may be used to minimize the settlement, though i t can never eliminate i t . It is used where there i s a danger of bottom heave in soft clay soils and sand liquefaction. It is also used to avoid dewatering when there i s a risk that lowering the water table may cause subsidence in the surrounding ground. Settlement of this type i s not usually a problem i f the s o i l i s dense sand or hard/stiff clay strata; however, in loose sand or soft clay, Chapter 3. Cut-and-Cover Methods 58 substantial settlement can occur i f the water table i s lowered. SP i s not satisfactory in hard clay, very dense sand or rock bearing strata. The SP i s normally designed as continuous sheeting with supports of struts or anchors installed as excavation progresses. Pile types may be defined by the effect of installation on the s o i l . Three major types of piles are in common use: displaced, small displaced, and non-displaced piles (Palmer and Tomlinson 1981). The f i r s t and second are normally driven, the third i s formed by boring or excavation (predrilled p i l e ) . SP f a l l s in the second category. It provides lateral resistance against s o i l pressure but has low bearing capacity. Sheet Piles can be further classi f i e d in terms of their material composition: steel, wood, or concrete (cast in place or pre-cast). Attention w i l l be focused on the f i r s t type, steel sheet piles, since i t s use is predominant in the f i e l d of construction. 1.1 Steel Sheet Piles: Structural shapes have been developed for steel sheet piles (SSP) and are in widespread use. They are stronger, more durable, less subject to damage during handling and easier to drive than wood or concrete sheet piles. In addition, they can be made relatively watertight with minimum effort and expertise. Chapter 3. Cut-and-Cover Methods 59 SSP i s reasonably flexible, causing some loss of ground and settlement, which may exceed allowable limits. Another possible source of settlement i s compaction of loose granular s o i l resulting from vibration. From an environmental standpoint, pile driving i s a noisy operation, although some relatively quiet methods have been developed. SSP provides an effective temporary barrier to water except where piles get disturbed or when workmanship i s poor. Water tightness of SSP is rarely adequate for permanent tunnel requirements (Megaw and Bartlett 1981). SSP may be l e f t in place after installing the tunnel structure or removed for reuse in other projects. SSP i s available in different forms, sizes, and stiffness. Each form has i t s own inter-lock system to provide a continuous GWSS. 1.1.2 Steel Sheet Pile Section Types: Forms of SSP can be clas s i f i e d into the following categories: 1. Z-sections; 2. arch web (U-section); 3. straight web; 4. special Y and T sections. U and Z sections are normally used. Typically, the range of weight of normal sections is about 15 to 60 lbs/square foot of wall area, with corresponding section modulus of 10 Chapter 3. Cut-and-Cover Methods 60 to 80 cubic inches per foot run of wall. Thickness varies from 1/4 to 7/8 inches, and width varies from 12 to 30 inches. Normal sections are used to resist larger horizontal forces because of their great stiffness, which in turn reduces the bracing required. Straight web sheets do not use much space, are comparatively flexible, and hence suitable only for relatively light lateral pressure. They are designed to have a high transverse inter-lock strength in tension, and possess negligible bending strength. They are applicable only for cellular construction. The notion of efficient sections governs the design process, where efficiency i s defined as the ratio of section modulus per foot run of wall to the cost (weight) of pil i n g per square foot of wall. Improved efficiency results from the use of a wider section with some reduction in thickness (Winterkorn and Fang 1975). Three features give normal sheet pile sections their advantage. These are interlock, shape of the pi l e cross section, and steel quality. Greater penetration can be obtained with SSP than with concrete or timber SP's which have, comparatively, a much larger cross section. In addition, for a given bending strength, a steel SP i s lighter in weight than other materials. Transport and handling into position for driving is simpler and less Chapter 3. Cut-and-Cover Methods 61 costly. A lighter hammer can be used to achieve the same rate of penetration for SSP than for other materials. 1.1.3 SSP Design: The main forces on SSP are from s o i l pressure and the support system of struts or anchorages. Several methods can be used for the computation of lateral pressure distribution. Some of the better known ones are attributed to Tschzaboterioff (1951) , Terzaghi and Peck (1967) and Peck (1969) for different types of s o i l . The design process for strutted excavation i s now regarded as completely empirical for estimating the bracing loads, and subject to construction procedures such as welding, and dewatering methods (Ratay 1984). A general design procedure for steel sheet p i l e i s as follows (Winterkorn 1981): 1. compute bending moments from the earth pressure distribution and strut reaction on the wall; 2. compute the section modulus for the sheets and consult the allowable stress; and 3. from standard AISC 1 for steel sheet piles, select a corresponding section (grade, strength, size, and other properties). The same design methods are applicable to the design of soldier piles in the Soldier Piles and Lagging alternative. The above procedure represents the conventional design approach for a single wall braced structure of sheet piles. •'•American Iron and Steel Association. Chapter 3. Cut-and-Cover Methods 62 See Bowles (1977) for more modern design procedures based on the f i n i t e element method. Appendix A presents the design computation method adapted for the Construction Methods Selection Assessment prototype (CMSA) described in chapters 4 and 5. 1.1.4 Construction Resources for SSP installation (Pile drivers): Construction resources applied for SSP installation include cranes (fixed and/or mobile), pile driver (vibratory and/or hammers), and crews required to drive SSP to their refusal depth. Among those resources, pi l e drivers are treated as variables and the rest are treated as fixed in the actual project context for CMSA development in chapter 4. The function of a pile hammer i s to furnish the energy to drive a p i l e . Pile drivers are designated by type and size. Hammer types are commonly classified as: 1. drop hammer; 2. single acting hammer (steam or air or diesel) ; 3. double acting hammer (steam or a i r or diesel); 4. differential acting hammer; and 5. hydraulic hammer. When noise i s a problem, the following can be used: 1. TW pile master; and 2. vibratory pil e driver. Chapter 3. Cut-and-Cover Methods 63 Operating principles, descriptive data, and relative advantages and disadvantages of each of the above p i l e drivers can be found in Jones (1963) and Peurifoy (1970). It i s important to note that the driving capability of hammers is generally expressed in i t s theoretical foot-pound (ft-lb) of energy delivered per blow. Selecting a Pile Hammer: Factors that influence hammer selection include the size and type of piles, the number of piles, the s o i l condition, the location of the project, topography of the site, and type of r i g available. Given the different sheet pil e and pile driver types available, and the range of s o i l types and job conditions encountered, selecting the best pil e type and hammer i s a non-trivial task (Jones 1963). The contractor i s usually concerned with selecting a hammer that w i l l drive piles for his project at the lowest cost. Manufacturers' recommendations are seldom reliable. There are some rules of thumb that help in selecting the size of the hammer. As an example, the maximum ratio of pile weight to ram weight should be four to one as suggested by Vulcan Workers, Inc. (Jones 1963). A more rational selection process is the one suggested by Peurifoy (1970). He presented, in tabular form, data for selecting p i l e hammers for driving different types of piles in a variety of s o i l conditions. Data from this reference Chapter 3. Cut-and-Cover Methods 64 were used in the CMSA prototype developed as part of this thesis for the two s o i l types previously described. 2. Soldier Pile and Lagging Steel soldier piles in the form of H steel sections are placed prior to the start of excavation by driving or d r i l l i n g at typically 8 to 10 foot centers. Horizontal lagging i s placed between the piles as the excavation progresses. The minimum thickness of lagging should be 1/24 of the span width (Ratay 1984). A variant is where a length of wall i s concreted in a slurry f i l l e d trench excavated between the King piles (concrete placed around the H steel sections after placing the sections in a pre-drilled hole). The piles can be designed to carry a temporary road deck, under which construction can proceed. Lagging can be eliminated i f the s o i l i s highly cohesive and capable of arching. It is inappropriate to use this method in perfectly cohesionless s o i l in which sheet piles must be used (Schroeder 1980) . Possible settlement of adjacent structures must be considered, and headroom for setting the piles has to be ensured. Chapter 3. Cut-and-Cover Methods 6 5 3. I n s i t u Reinforcement Techniques — S o i l N a i l i n g Recently, new in-situ earth reinforcement and lateral support systems for deep excavations have been introduced. Unlike conventional systems that serve to retain s o i l behind a positive support system, the use of these alternative systems i s based on the concept of s o i l reinforcement. That i s , native s o i l adjacent to the excavation i s strengthened so that i t can stand unsupported at depths which would normally require the installation of conventional GWSS (e.g. Soldier Piles, Sheet Piling). Three main categories of insitu reinforcement exist: s o i l nailing, reticulated micro-piling, and dowelling. In s o i l nailing, the reinforcement i s installed horizontally immediately after excavation so that i t improves the shearing resistance of s o i l by acting in tension. Soil nailing works by reinforcing the ground in situ with relatively small, fully-bonded inclusions, usually steel bars (Bruce 1987) . They are introduced to the s o i l mass, the face of which i s locally stabilized by sprayed concrete, and act to provide a zone of reinforced ground. Although the s o i l nailing system is intended to serve as temporary support only, i t has been shown that i t has a great deal of long term s t a b i l i t y (Shen 1981b). Reticulated micro-piles are steeply inclined into s o i l at various angles both perpendicular and parallel to the face. Chapter 3. Cut-and-Cover Methods 6 6 The overall aim i s similar to s o i l nailing: to provide a stable block of reinforced s o i l acting like a retaining structure. The reinforcement acts to resist bending and shearing forces. Soil dowelling is applied to reduce or slow down slope movements on a well defined shear surface. The reinforcement provides resistance to shear forces. The p i l e diameter for s o i l dowelling i s generally far greater than that for s o i l nailing. 4. Concrete Bored Piles These are large diameter bored piles at close centers that can be used as. a wall. They might be contiguous or may overlap to form what is called Secant Piles. With good workmanship, substantial water tightness i s possible, particularly with secant piles. 5. Slurry Trench Wall The slurry trench wall (STW) technique, also called Diaphragm Concrete Walls, is relatively new. STW can be applied to construct two types of walls. These are Concrete Diaphragm wall and Soldier Pile Tremie Concrete Wall. 5.1 Concrete Diaphragm Wall: Constructed as a normal cast in place wall with supports, a concrete diaphragm wall provides a watertight barrier which can be installed with a minimum of back slope subsidence. The wall becomes part of the permanent structure. Because this wall i s impervious, Chapter 3. Cut-and-Cover Methods 6 7 dewatering of granular soils i s often not required. Diaphragm walls are considered to be semi-rigid and walls of shallow depth in moderate ground conditions are sometimes l e f t unsupported. In a deep excavation, support i s required to r e s t r i c t lateral deflections. Skin f r i c t i o n along the wall tends to reduce settlement of adjacent soils below an amount which i s common for soldier piles or steel sheet piles. It i s reported that vertical settlement appears to be in the order of 0.25% of excavation depth for diaphragm walls versus generally less than 0.40% for other support systems (Ratay 1984). 5.2 Soldier Pile Tremie Concrete Wall (SPTC): This i s a modification of the standard STW techniques to increase lateral resistance. It consists of a tremie concrete wall in which steel soldier piles are embedded for reinforcement. For deep excavations in soft ground, these walls may help alleviate potential subsidence problems, but they s t i l l require some retaining system. 3.5, Excavation Operations The excavation activity can be divided into a primary excavation of most of the tunnel space bulk, and a secondary one that i s prior to GWSS installation. Examples of the latter include pre-drilling and/or excavation for secant piles installation, and slurry trench wall. Excavation can Chapter 3. Cut-and-Cover Methods 68 proceed simultaneously both vertically and horizontally. The pace of excavation i s a function of s o i l cohesion, tunnel dimensions, the GWSS, and the muck removal system selected. Within the methods definition in this thesis, the excavation activity represents a tact i c a l construction operation that could be separable and distinct yet integral to the GWSS type and alternative, and can be described in terms of i t s resources (the reader i s referred to the interviews in Appendix C). Section 3.8, a specific Cut-and-Cover tunnelling example, examines and structures pertinent excavation roles in balancing a construction cycle. Excavation, in Cut-and-Cover tunnelling, i s substantially different from underground tunnelling in terms of operations, construction methods and scheduling. For a given tunnel, i t can proceed in two directions: horizontally, parallel to the tunnel, and vertically, toward the bottom of the tunnel. The pace of excavation in either direction i s subject to s o i l s t r a t i f i c a t i o n , tunnel dimensions, retaining system, compatibility of muck removal system, and so forth. The excavation progress, interwoven with other a c t i v i t i e s , eventually yields a construction cycle. Excavation cycle optimization has a substantial impact on total productivity and production rates. Chapter 3. Cut-and-Cover Methods 69 For shallow cuts, the excavation i s usually made from the surface, using a backhoe, and dumping the muck directly into trucks or into a stock pil e along the side of the excavation. For deep cuts, the excavation may be carried out using trucks with haul ramps, or clamshell hoists, or by using cranes to l i f t the excavated material from the bottom cut into the stock pil e or waiting trucks. A wide range of equipment may be used, such as backhoes, front end loaders, bulldozers, trucks, draglines, scrapers, clamshells and cranes. Each type of equipment comes in different capacities and i s preferred depending on job conditions. Some variables that affect the excavator productivity are (Caterpillar 1982): 1. Soil type and condition 1. conduit type and size; 2. number of cross lines; 3. trench dimension; 4. spoil pile dimensions; 5 . truck loading requirements; 6. operating a b i l i t y ; and 7. j ob management. 2. Machine related factors 1. proper attachment for the job; 2. size of digging and spoiling envelope; Chapter 3. Cut-and-Cover Methods 7 0 3. payload capability (bucket size); 4. cycle time (breakout force and hydraulic speed); and 5. l i f t i n g capability. After describing various Cut-and-Cover tunnelling alternatives, one needs to consider the major factors that influence alternative choices. These are discussed next. 3.6 Factors Affecting Methods Selection and Design Cut-and-Cover tunnelling construction methods embrace a wide range of alternatives. Influencing factors may interact and/or override others under different project conditions. It i s not always immediately obvious what i s the best solution, due to the complexity of the interaction among operations, and given the project context variables, such as ground conditions and access to the work area. It i s useful to explore some of these factors in more detail to demonstrate the diversity of problems that have to be treated, the range of methods, etc. Where appropriate, reference i s made to simplifying assumptions adopted for i l l u s t r a t i o n purposes in the thesis. 1. Ground Conditions Among various influencing factors, ground conditions dominate the choice of construction methods. Ground conditions do not always f a l l into convenient and clearly defined groups. Soil has relatively low strength and high Chapter 3. Cut-and-Cover Methods 71 deformability. These adverse mechanical characteristics have a direct influence on excavation, s t a b i l i t y , type of permanent structure, and compaction methods. Ground conditions generally can be divided into five broad groups: 1. Cohesive s o i l ; 2. Noncohesive s o i l ; 3. Soft rock; 4. Hard rock; and 5. Organic s o i l . To reduce the size and dimensionality of the problem at hand, only a small subset of s o i l types w i l l be considered in this study. Mutually exclusive s o i l types included in this study are: loose sand, and dense sand; and soft clay, and hard or s t i f f clay. The maximum number of layers i s assumed to be two. These limitations were set so that the number of s o i l strata scenarios (e.g. loose sand on top of hard clay i s one out of 12 perturbations of s o i l layers order) w i l l be a relatively small size which simplifies the choice of a method. 2. Ground water conditions The presence of water, i t s level, and i t s chemical composition may dictate compatible control methods that minimize ground subsidence. This could be done by lowering the ground water table, by grouting, freezing, compressed air, or a combination of the foregoing. Further, water Chapter 3. Cut-and-Cover Methods 7 2 exerts pore pressure onto the sides of the excavation in addition to the s o i l pressure which creates extra pressure on the ground wall support system. 3. Size of the proposed excavation The physical dimensions of the trench have a substantial impact on almost every method used during construction. The dimensions, along with geotechnical conditions, dictate the proper equipment for the excavation and material handling strategies. For instance, a shallow depth may suggest the use of a backhoe, a front shovel, or a scraper, while deeper trenches require the use of a clamshell or a crane with a loader or a dragline. 4. Working space access Some Cut & Cover techniques may require more working space and access than other competing alternatives. This may be of great importance especially where allowable work space i s small and/or congested. 5. Diversion of services Removal of underground u t i l i t i e s or diversion of buried services before excavation i s the ideal for Cut-and-Cover tunnelling. However, this is rarely possible for a l l services. Temporary diversion, or the cutting-off of some services may be unacceptable, especially in urban areas. Those which need to cross the trench must be supported, Chapter 3. Cut-and-Cover Methods 73 protected, and maintained during the entire construction process. Such situations may result in undesirable delays. 6. Environmental requirements Construction operations may result in t r a f f i c re-routing, closing segments of roads to pedestrians, destabilizing of adjacent structures, and producing hazards (e.g. noise, water contamination). Environmental considerations may eliminate potential methods, res t r i c t daily working time, impose a geotechnical monitoring program, and demand more safety procedures. Next, a traditional Cut-and-Cover tunnelling project w i l l be described to give a flavor to the ac t i v i t i e s breakdown, planning techniques used, constraints imposed on operations, GWSS installation and excavation interaction, a l l within a specific project context. 3.7 Cut and Cover Tunnelling Project Example 3.7.1 Background The objective of this example is multifold: highlight some Cut-and-Cover tunnelling characteristics and spatial constraints; and show how decisions (production rate, resource selection, etc) pertaining to the excavation construction cycle, and GWSS retaining system design elements (spacing of retaining system), are interwoven with excavation. This example serves as a reference for Chapter 3. Cut-and-Cover Methods 7 4 subsequent variations of the Cut-and-Cover tunnelling projects throughout the thesis. Further, a Cut-and-Cover project in downtown Seattle (Turnham 1988), i s adapted to exploit, characterize, and implement the ideas and solutions for methods selection. Table 3.2 displays information extracted from the project with some modification. Location: Downtown Seattle Type of Project: Metro Tunnel Tunnel Dimensions: 38 feet wide, 60 feet deep, and 800 feet long. Type of GWSS: Soldier Piling and Lagging Ground Condition: Soft Ground (tough dry clay) Progress Rate: 30 feet per day (for excavation and lagging) Ground Work Control: Pumping Ground Treatment: Not Available Underpinning and Utilities: None Design Requirements: Excavation does not exceed 10 feet of depth without lagging and bracing. Project Duration: May 1 to November 1,1987 Selection of Methods: Not Available Type of Retaining System: Struts (and/or Tie-backs) Table 3.2 Seattle Project General information [Adapted from Turnham 1988] Figure 3.2 i s an ill u s t r a t i o n of the aforementioned reported project during construction. Chapter 3. Cut-and-Cover Methods 75 Fa2?J? 3 * l , C U t a n d C o v e r T u n n e l Project Isometric view for Soldier Piles and Lagging with struts for Seattle Project [Turnham 1988] Chapter 3. Cut-and-Cover Methods 76 O U < Submittal — Soldier piles Utilities Excavate — Lagging Bracing struts Working floor Base slab Wall Struts Backfill Streets Roof 10 15 20 Project Duration (weeks) (a) Project Barchart 25 Location (50' Segments) 2 3 4 Duration (Weeks) (b) Project Time-Space Diagram Figure 3.3 Barchart and Time-Space Diagram for the Seattle Project [from Turnham 1988] Chapter 3. Cut-and-Cover Methods 77 The site sketch of figure 3.2 presents, from right to l e f t , the following flow of work: primary excavation at the top done by a backhoe; secondary excavation at the bottom of the trench done by a bulldozer; pre-installation of soldier piles as GWSS; installation of struts as retaining system; construction of permanent concrete tunnel floor; and placement of sides and roof; and, f i n a l l y , backfilling. For this Cut-and-Cover example, the bar chart and space time diagram for major act i v i t i e s , shown in figure 3.3 (a,b), serve as a planning tool. 3.7.2 Lagging and Excavation Construction Cycle A construction cycle can be defined as a sequenced repetition of construction operations that constructs a pre-specified modular unit of a f a c i l i t y . For instance, a floor in a high-rise building or tunnel excavation unit step, or a house unit in a residential housing project. A construction cycle i s mostly applicable to linear projects. For this example, the project engineer f i r s t creates barchart and space-time diagrams to show major ac t i v i t i e s , their logical sequence, and their inter-relationship. These planning tools set the pace of production for construction operations, and sets the total duration for the project a c t i v i t i e s . The contract project target duration of six months sets the pace for the required production rate for each successive activity. Chapter 3. Cut-and-Cover Methods 7 8 The excavation activity w i l l be the lead one that triggers lagging and other following a c t i v i t i e s . Lagging, on the other hand, could delay excavation i f i t s installation pace in the vertical direction i s slowed down to less than the excavation production rate. Therefore, in order to optimize the cycle production, the rate of lagging production should equal, or be slightly less than the excavation rate. One can picture, physically and geometrically by dividing the tunnel into segments, the sequence and the pace of excavation and lagging per construction cycle. First , excavation proceeds horizontally at the top layer, and after finishing the second segment, lagging starts at the f i r s t segment and at the same rate of the former operation. Thus time and space lags are specified. Figures 3.4 a,b,c, and d show the routine successive progress of excavation and lagging for the soldier piles and lagging alternative. To summarize the aforementioned description, the identified major construction cycle decision variables are: 1. Segment dimensions; 2. Depth of excavation; 3. Direction mode for Excavation; 4. Time lag between ac t i v i t i e s ; and 5 . Space lag between crews (in number of segments). Chapter 3. Cut-and-Cover Methods 79 3.7.2.1 Constraints for Construction Operations Central to the construction cycle for lagging installation and excavation operations, are two categories of constraints. 1. Engineering Design Constraints Fi r s t , the type of GWSS (Soldier Piles and Lagging) dictates a top-down excavation strategy. Design specifications prohibit excavations to proceed more than 10 feet below lagging nor deeper than 12 feet below bracing. Therefore, excavation and lagging were constrained to follow a leap-frog sequencing to reach the required 60 foot depth. Figure 3.4 shows the time-space diagram for these two a c t i v i t i e s , where the rate of production for both excavation and lagging, and the time lag between them, is suited to complete one segment of 50 feet of permanent tunnel structure. 2. Geometric Constraints The tunnel structure forces equipment and formwork (shoring) to move in the direction of the tunnel. Typically, for wide and deep tunnels, excavation proceeds in angled slides (slopes) rather than vertical slides (shallow tunnels) to allow excavating and hauling equipment easy excess. In addition, the slope stabilizes i t s e l f with no support. The size of the tunnel and project location allows several combinations of methods^ For , example, for the method attribute of construction resource, a backhoe at the Chapter 3. Cut-and-Cover Methods 8 0 top half of the tunnel i s excavating and hauling, whereas a bulldozer excavates the bottom-half and pushes i t to the upper half within the reach of the backhoe, as shown in figure 3.2. An alternative for the previous combination may include a front end loader for excavation and a crane for hauling the muck into trucks. The dimensions of the tunnel may allow smaller equipment to work in parallel, or alternatively, one large machine. Careful selection of methods for major operations requires defining the tasks that a method performs and allocating resources for these methods. Some resources may perform multiple tasks, such as a backhoe and a crane, whereas others perform specific tasks, such as front end loader (hauling muck). Therefore, geometric constraints allow only some scenarios of construction methods combinations. Our intention here was to hint at the low level construction process variables. Chapter 4 shows how these decision variables are incorporated into a method frame. 4. A K B E S Framework for Methods Selection and Design 4.1 Introduction The purpose of this chapter i s to structure the methods selection problem by describing the problem, proposing a KBES framework approach, and prescribing the attributes for a control strategy shell. This chapter builds, step-by-step, a vision of a systematic knowledge based approach for the methods selection problem. Set out in section 4.2 i s a KBES framework for this problem. This framework reflects the hierarchical structure of the decision making process. A method i s described by i t s attributes which include design element, construction strategy, construction resources, and construction method process. A conceptual KBES that integrates procedural, factual, and judgmental knowledge approaches for methods selection i s then described. This i s followed by a description of system features and a pseudo rule-based control strategy. The goal of the system i s to eliminate undesirable alternatives, rank the remaining ones and suggest values for the attributes of feasible methods. An example for Cut-and-Cover tunnelling i s presented to demonstrate how the KBES framework i s intended to work. Chapter 4. A KBES Framework for Methods Selection 82 Steel sheet p i l i n g is selected as the example context for this implementation. Described in section 4.3.2 are the decisions variables and project context variables treated for this example. In section 4.3.4, a conceptual control strategy i s described. The control strategy treats methods selection in a two step process: preliminary f e a s i b i l i t y , which screens out undesirable alternatives; and detailed f e a s i b i l i t y , which specifies the remaining alternatives. Key c r i t e r i a which make a method accepted or rej ected at the two levels in the control strategy are identified, and include: design, risk, resource compatibility, performance measures, and regulatory considerations. At the preliminary f e a s i b i l i t y level, high level rules based on empirical knowledge are used to reduce alternatives. Detailed f e a s i b i l i t y analysis involves a multi-stage process. It attempts to synthesize design attributes of the methods that pass through preliminary f e a s i b i l i t y . Once a detailed synthesis i s successfully completed for a method, i t i s evaluated and ranked against other feasible alternatives. Evaluation c r i t e r i a based on costs are used to rank alternatives at both levels. At the preliminary level, alternatives w i l l be ranked according to their Chapter 4. A KBES Framework for Methods Selection 8 3 preliminary unit costs. At the detailed level, costs include an assessment of risk in terms of money. The risk assessment framework adopted i s elaborated in section 4.4. Risks have been cla s s i f i e d into normal (acceptable) and catastrophic (unacceptable) risks. The former includes risk categories of equipment, material, productivity, subsurface subsidence, seasonal, and other losses. A decision tree incorporating three states of nature for encountered subsurface conditions, i s used to treat risk. These are: encountered conditions are more favorable than expected, as expected, and worse than expected. From the latter, three further states of nature are treated : no damage, element damage, and system damage. If the probability of system damage exceeds a threshold value, catastrophic damage i s assumed to occur and the method alternative i s rejected, otherwise, i t w i l l be accepted. The aforementioned risk categories are applied at the decision tree terminals as possible outcomes. Chapter 4. A KBES Framework for Methods Selection 84 4.2 A KBES framework for Method Selection 4.2.1 General In this section, a general KBES framework for method selection i s presented. The method definition introduced in section 2.2.1 is further refined to provide this framework. For discussion purposes, we treat time as the main criterion at this stage, while cost, quality, safety, and others, are treated as secondary goals that could be violated within a range. A pseudo rule-based control strategy i s used to il l u s t r a t e operations of the framework. 4.2.2 Methods Selection Defined Based on the discussion in section 2.2.1, the term "construction method" i s used in a wholistic sense. That i s , i t embraces design, strategy, resource and process considerations. Specifically, a construction method i s described in terms of a conceptual frame. Attributes of this frame are structured in a hierarchical fashion. They include design concept for temporary f a c i l i t i e s , construction strategy, resource requirements and construction process. Not a l l levels are meaningful for every methods selection problem. To i l l u s t r a t e the use of the definition, consider figure 4.1 and table 4.1. Chapter 4. A KBES Framework for Methods selection 85 Design Element J L Construction Strategy Required Const-ruction Resources FT Construction Process Model CPM 1 CPM 2 (a) (b) Figure 4.1 Hierarchy of Construction Method Frame Attributes Figure 4.1 (a) depicts a hierarchy of the method a t t r i b u t e s , while figure 4.1 (b) exhi b i t s various a t t r i b u t e a l t e r n a t i v e s at each l e v e l , and the bold l i n e represents the most preferred a l t e r n a t i v e at each l e v e l , which are taken together to define a construction method. Table 4.1, corresponds to figure 4.1. I t r e f l e c t s the kinds of alt e r n a t i v e s a v a i l a b l e and decisions to be made at each l e v e l f o r the Cut-and-Cover tunnel problem. Design a l t e r n a t i v e s r e f e r to the GWSS temporary structure as decision variables, whereas the permanent tunnel structure i s treated as fixe d . Each GWSS al t e r n a t i v e has i t s own design and construction procedures as b r i e f l y described i n chapter 3. Lower l e v e l decisions pertinent to t h i s l e v e l Chapter 4. A KBES Framework for Methods Selection 8 6 imply specifying structural members for the GWSS and the retaining system, including design c r i t e r i a selection (moments, deflection, etc.) and type of materials employed. For instance, specifying a GWSS of steel sheet piles includes their members sections selection and sizing based on moment criterion, where steel grade, section modulus, moment of interia, and other properties characterize a member. At the construction strategy phase, two levels are identified. The f i r s t one is a high level decision, whether an upside-top (Milano Method) should be ut i l i z e d as opposed to the traditional top down, which dictates the sequence of construction operations and scheduling. Lower level decisions include tactical (operational) decisions such as selecting operations variable values, and retaining system. For example, the operations variables of a GWSS installation for pil e driving include piles driven in waves, in singles or in doubles. For the latter, retaining systems include two choices: struts and tie-backs, which provide obstruction free access for construction operations and thus shorten the construction duration. Chapter 4. A KBES Framework for Methods Selection 8 7 LEVEL ALTERNATIVES Design Alternative For Temporary Facilities Ground Wall Support System (GWSS) Temporary/Permanent GWSS Soldier Piles & Lagging Steel Sheet Piling Slurry Trench Wall Soil Nailing Shotcrete Secant Piling Construction Strategy High Level Strategy Upside-Top, Top-Down Pile Driving Patterns Waves, Doubles, Singles Retaining System Wales and Struts, Tiebacks Construction Resource Requirements GWSS Installation Pile Drivers Drilled Piles Lagging Installation Ret. System Installation Crew_B 20 Construction Processes Model • Cut-and-Cover Tun. Process Network Models GWSS Installation & Simulation, CPM, . . Excavation Are Separable Process (GWSS) + Process (EXC) + Process (..) + .. GWSS installation & Process {GWSS + EXC} Excavation Are Intertwined Table 4.1 Methods Selection Space for GWSS The third attribute, construction resource requirement, takes in a pool of available resources to construct the GWSS f a c i l i t y . For instance, i f steel sheet piles was considered feasible as a design alternative, and a top-down construction strategy i s selected, then the ava i l a b i l i t y of the materials and resources for the GWSS and the retaining Chapter 4. A KBES Framework for Methods Selection 88 system — sheet piles, struts/tie-backs, wales, p i l e drivers and crews are c r i t i c a l to the acceptance of the GWSS alternative. The construction process model embraces c r i t e r i a for measuring the performances of alternatives at several levels: project, activity, and operation levels. For instance, the progress rate of the two prominent a c t i v i t i e s of GWSS installation and excavation may set the project pace. Those ac t i v i t i e s may be separate — e.g. Process (GWSS) + Process (EXCAVATION) — that i s a process per activity as shown in Table 4.1, in the case of the steel sheet piles alternative; or act i v i t i e s may be intertwined— e.g. Process (GWSS + EXCAVATION} — that i s a process for more than one activity, as for the s o i l nailing alternative. Activities broken down into smaller tasks with allocated resources are emulated via a predictive/analytical process that results in performance measure values. Three things ought to be verified for prescribing a process: the performance measures required (time, cost, safety, quality) and the process to be applied (simulation for excavation, dynamic formulas for pile driving), and construction resources (pile drivers, cranes), given that existing resources are compatible (pile driver size dictates crane size). For instance, at the project level, the c r i t i c a l path method could be used as a predictive and analytical Chapter 4. A KBES Framework for Methods Selection 8 9 tool for project total duration. For lower level operations, such as the pil e driving activity, dynamic hammer formulas are used to predict p i l e driving rate for a given soil/hammer/pile scenario, while "automated" rule based knowledge is used for diagnosis. 4.2.3 Methods Shell The basic structure corresponds to a methods selection shell in which we use expert systems to control the selection process as well as to encode knowledge pertaining to various project contexts displayed in figure 4.2. Relevant functions, task descriptions, data and knowledge bases, and analysis tools, are shown for each component, as appropriate. This notion of a shell i s reflected in the various applications modules such as Cut-and-Cover tunneling, High-rise structural work, Pier construction for bridges, etc. Chapter 4. A KBES Framework for Methods Selection Numerical Routines Earth Pressure and Moments Risk Assessment Evaluation Criteria Pile Driving Production Prediction User Interface Tunnel Dimensions Soil Profile Input Risk Threshold " Knowledge Base x Frames for DE, CS, CR, and CPM Rules for Preliminary Feasibility Rules for Soil Representation Rules for Sheet Piles Selection Rules for Hammer Selection Rules for Hammer/Pile Compatibility Rules for Risk Assessment y Context Design Parameters User Interference Analysis Results Data base Structural Members Construction Resources Soil Properties Risk Categories Inference Engine . NExpert Object Programming Language | FC for SSP Design i FC and BC for Hammer Selection ' FC and BC for Technical Feasibility | Highway Methods Selector Tunnel Boring Machine Selector Bridge Methods Selector Hlghrlse Building Method Selector Figure 4.2 Construction Methods Selection Shell Chapter 4. A KBES Framework for Methods Selection 91 Major components of this shell include: 1. User interface to f a c i l i t a t e user input, intervention, and system shell interrogation. For instance, the user may override the system ranking of preliminary feasible GWSS alternatives and answer queries with uncertainty; 2. Knowledge base to contain procedural and declarative knowledge required to specify and marry method attributes; 3. Context reflects the current problem environment during a session. Derived context variables of lateral pressure, moments, progress rate, method attributes evaluation, and so forth, are retained in the context f i l e structure; 4. Inference engine performs the methods selection solution and controls the system session, screening and ranking methods alternatives, instructing the knowledge base to find attribute parameter values, interacting with the user, and so forth; 5. Numeric routines include procedural and analysis knowledge that i s encoded in traditional programming languages, e.g., design procedures based on strength criterion, predictive modules based on simulation, etc; and 6. Data base cl a s s i f i e s and stores methods in terms of their attributes, structural members, construction resources, s o i l properties, etc., that are required for method synthesis and accessed by the knowledge base. In section 4.2, emphasis is placed on items 2 to 4 while section 4.3 and section 5.3 focus on items 2 to 6. Chapter 4. A KBES Framework for Methods Selection 92 4.2.4 Sketch of System Features and Operation A two phase system i s conceived for ranking and synthesizing methods alternatives: identification and elimination of preliminary feasible alternatives (phase 1) , and detailed specification for the feasible methods (phase 2) . Figure 4.3. shows a sketch of the components of two phases. The objective of phase 1 i s to identify and eliminate the number of preliminary feasible alternatives, thus, reducing them to a few candidates. Phase 2 further carries out a f e a s i b i l i t y analysis and methods specification at a detailed level. In relation to figure 4.2, the following discussion basically covers the role of the inference engine (control strategy), and describes the specific knowledge for the methods selection problem solution. Phase 1 considers a l l alternatives options for methods under a scenario of project context and goals. Preliminary project context information may include site layout and access, s o i l profile and conditions, water table conditions, site location, etc. Project goal includes duration, cost, safety, etc. Preliminary screening follows after possible methods have been identified. Some w i l l be eliminated through considerations of one or more c r i t e r i a (e.g. cost, time, dimensionless number). Section, 4.3.3 presents the kind of knowledge central to evaluating preliminary f e a s i b i l i t y . Chapter 4. A KBES Framework for Methods Selection 93 Assignment of values to the attributes, which define a construction method, takes place at the detailed level, i.e. phase 2, which constitutes the body of the thesis. The process envisaged, as shown in figure 4.3, i s further elaborated on in figure 4.4. Phase 2 consists of three main parts: low-level method specification of the preliminary feasible candidates including specifying method attributes, process modelling, and method analysis. Also illustrated in figure 4.4 are system functions and example specifications for the Cut-and-Cover tunneling problem example. What follows is a step by step description of the application of phase 2 of the model shown in figure 4.3. In giving this description, important issues to be resolved are identified. Step 1 The process starts with the user inputting information relevant to the project context. This information can be structured under several headings, as illustrated in Table 4.2. Figure 4.3 Construction Methods Selection System Process Chapter 4. A KBES Framework for Methods Selection Stag* -1-/Prellminary Candidates Phase -1-Select Construction Strategy Select Construction Resources Select Construction Process Model Figure 4.4 Detailed Feasibility (Phase 2) Chapter 4. A KBES Framework for Methods Selection Evaluate Alternative Cost Estimate Unit Cost Including Risk Method Attributes 1 Stage -3-Design Element 1.1. Specify Sheet Pile (Type, Size, and Quantity) 1.2. Specify Tie-Backs 2. Construction Strategy (Top-Down) 3. Required Resources 3.1. Use Vulcan Single Acting Air Hammer (Hammer Models OR) 3.2. Use Crane, 15 ton 3.3 4. Construction Process Model (CPM(i)) 5. Performance Measures 5.1. Duration 5.2. Cost 5.3. Safety 5.4. ... Specify Method Attributes (Terminate Figure 4.4 Detailed F e a s i b i l i t y (Phase 2) (continued) Chapter 4. A KBES Framework for Methods Selection 97 Permanent Structure Purpose: Utilities Tunnel. Dimensions: L = 1200 ft; W = 20 ft; D = 20 ft. Site Conditions Location: Downtown urban area. U / G Utilities: None Tiebacks: Easement available Soil Conditions Soil Type: Soft clay Water Table: 15 feet below surface elevation Design Constraints Retaining System: < 12 feet below retaining system. Displacement: No more than 3 inches. Project Goals GWSS Target production rate (tpr): 30 feet/day Target cost (tc): $ 15.00/fr2 Excavation Target production rate (tpr): 30 feet/day Target cost (tc ): $20.00/cy Table 4 .2 GWSS Project Context Data This information is . used to help identify and screen possible methods. At the outset, high level information is sought. Once a detailed examination of methods i s performed, more low level, detailed information i s requested from the user. The dialogue i s meant to identify the information required and i t s influence on the decision process. The context information i s requested in the form of rules. Chapter 4. A KBES Framework for Methods Selection 98 Step 2 The selection process starts with a classification of the methods alternatives known to the system, (i.e. stored in i t s data base), which provides a representation or categorization scheme for a l l methods such that other desired or new methods could be entered by the user. A methods data base thus contains a library of existing methods characterized by a set .of dimensions, including method attributes and pertinent factors such as level of risk, safety, resources compatibility, level of expertise, and so forth: i.e. a method associated with specific knowledge. Specifications for a method database i s not part of this thesis. We come out of the preliminary screening with a set of possible design alternatives: DA = (DA(1), DA(2), .. DA(n)} and a set of possible construction strategies: CS = (CS(1), CS(2), .. CS(n)} — where links between the DA's and the CS's may exist. Allowance i s made for the user to augment the l i s t of design alternatives and add to the knowledge base. User context input is applied to eliminate alternatives and rank those that are feasible. This process uses knowledge in the form of rules for this task. The standard Chapter 4. A KBES Framework for Methods Selection 99 form adopted for the formulation of a design alternative (DA) f e a s i b i l i t y rule is as follows: RULE Design Alternative (i) Is Technically Feasible IF [ DA(i) Technical Feasibility Conditions Are Satisfied ] AND [ DA(i) Time/Cost Performance Conditions Are Satisfied ] THEN [ DA(i) is Feasible ] Technical f e a s i b i l i t y refers to the a b i l i t y of the design alternative (or construction strategy or resource assignment) to satisfy project specifications, regulatory constraints, work space constraints, etc. Each of the elements of this rule format may correspond to the evaluation of a set of rules. An example of a simplified rule for f i l t e r i n g GWSS alternatives, for the steel sheet pile alternative to be feasible at this stage, is as follows: RULE SSP (DA(1)) Is Feasible IF Tunnel Depth < 20 feet AND Soil Condition = Soft Clay AND No U / G Utilities = True AND lbprl < tpr < ubprl AND lbcl < tc < ubcl THEN DA(1) = Steel Sheet Pile AND Steel Sheet Pile is Feasible " l b p r l " , "ubprl", " l b c l " , "ubcl" are the lower and upper bounds of production rate and cost ranges for the methods design alternative level, given a specific project context. Chapter 4. A KBES Framework for Methods Selection 1 0 0 Step 3 Construction strategy for the GWSS could be discussed at two levels: high , or project, level of Top-Down versus Bottom-Up; and low, or activity, level of retaining system type of tie-backs versus wales and struts. The f i r s t drastically alters the make up and sequencing of major tunnelling ac t i v i t i e s , and subsequently the method attributes. The second i s central to excavation activity where the tie-backs retaining system speeds up excavation where this activity i s a major one for Cut-and-Cover tunnelling. For our example, we started with a GWSS design alternative of steel sheet piles which implicitly embraces a top-down construction strategy. However, when a GWSS slurry trench wall i s considered, and ground surface restoration is regarded as a primary concern, then of the two construction strategies, the bottom-up w i l l be preferred. Its choice eliminates the SSP alternative and other traditional GWSS alternatives. Further low level construction strategies for pile driving patterns, i.e., in waves, doubles, or singles, etc., may be considered. The standard form for the rule that determines f e a s i b i l i t y of a construction strategy i s as follows. RULE [Construction Strategy (j) + Design Alternative (i)] Is Feasible IF [ CS(j) COMPATIBLE with DA(i)] AND [ CS(j) Technical Feasibility conditions Are Satisfied ] AND [ CS(j) Time/Cost Performance conditions Are Satisfied ] THEN [ CSG) + DA(i) Is Feasible] Chapter 4. A KBES Framework for Methods Selection 101 The rule defines compatibility as the a b i l i t y to combine a specific design alternative with a specific construction strategy. Excavation could be a bottleneck activity, therefore we have two feasible choices, CS(1) and CS(2), of tie-back and anchorage retaining system, that are feasible for DA(1) of steel sheet p i l e , i.e. either CS(1) + DA(1) or CS(2) + DA(1). Since minimizing time i s the prime goal ( a hypothesis made for discussion purposes), the f i r s t choice w i l l be favored, given an easement permitting tie-backs is possible, allowing more freedom of resource movement and feasible resource combinations. -/ Should a l l strategies f a i l to be feasible, then the control system selects the next preferred design alternative. A new set of construction strategies appropriate for the selected design alternative i s examined. Determination of a feasible strategy at this stage passes control on to the next stage. Step 4 Key resources are selected from a data base of resource alternatives. The guiding principle in assigning resources is that the largest capacity equipment that satisfies space and/or av a i l a b i l i t y constraints i s selected, in order to maximize the production rate, and reduce production risks. Chapter 4. A KBES Framework for Methods Selection 102 If no feasible resource combination exists then the plan f a i l s and the system backtracks to previous stages to alter the plan. The standard form of the rule for resource assignment parallels that for the design alternative level. An example rule which reflects the interdependencies between work packages i s as follows. RULE Select Resource Class for Excavation IF Activity = Excavation AND Geometry = Class 1 AND DA[GWSS] = DA(1) {Steel Sheet Pile} AND CS[GWSS] = CS(1) {Tiebacks} AND lpr{Excavation} < Required Excavation Rate < upr{Excavation} THEN Excavation Resource = <backhoe class > AND Set <backhoe class >= Cat 245 For our example, a shallow tunnel, this corresponds to selecting the largest piece of equipment which satisfies the technical f e a s i b i l i t y constraints, for example, a Caterpillar Hydraulic Backhoe 245 from the set of 245, 235, 225, 215 models (Caterpillar 1982). Similarly, the steel sheet p i l e installation activity requires specifying a pile driver and a crane resources. At the end of phase 1, the attributes of one or more feasible methods are passed to phase 2, detailed f e a s i b i l i t y , for further manipulation. Step 5 This step i s directed at specifying the representation and analysis methods, given specification of the design Chapter 4. A KBES Framework for Methods Selection 103 alternative, construction strategy and resource assignment at the detailed level. The design alternative i s further specified in terms of i t s structural members and i t s retaining system such as soldier piles, sheet piles, trench wall, shotcrete, wales, anchorages, struts, etc. Resources are sized. For example, a resource for excavation such as the Cat 245 backhoe class suggested by phase 1 (see previous rule) i s further specified as having a 30 f t , one-piece boom with a 14 f t 6 inch stick, 2 1/2 cu bucket, etc. , attributes that are pertinent to i t s production. Construction strategy decisions include specifying vertical and horizontal spacings, and other construction cycle decisions (time and space lags of a c t i v i t i e s ) , and construction model processes. Construction Process Model (CPM) refers to a feasible combination of construction resources under a project context (including geometry) that can be modelled via simulation, dynamic programming, and other OR techniques, or mimicked by previous experience, to predict their performance measures. Representation formats include networks, line of balance charts, and simulation models. Other knowledge incorporated in this stage deals with the degree of definition appropriate to the modeling task, for example number of acti v i t i e s , time/space lags, and so on. Chapter 4. A KBES Framework for Methods Selection 104 At stage 1, an attempt i s made to marry method attributes variables under an appropriate construction model process scheme or scenario. If GWSS DA(i) installation and excavation (the retaining system spacing i s implicitly incorporated in the construction cycle design) are separable, then each construction process model i s evaluated separately, and the project progress w i l l be controlled by the slowest pace of either. If the GWSS DA(i) i s soldier p i l e and lagging, then the lagging and retaining system installation are intertwined with excavation. For intertwined a c t i v i t i e s , one construction process model i s used to collectively predict their performance measures. For our example of steel sheet p i l e , the GWSS installation and excavation are intertwined i f they share a common multi-task resource such as the crane, which can be used to load/unload muck and construction materials, and hold/move the pi l e driver. If separated, then steel sheet pi l e installation process w i l l be modeled by the hammer energy formula which predicts the productivity rate of pile driving based on the soil/pile/hammer scenario. The retaining system installation is always intertwined with excavation as discussed in section 3.7.2, and thus implicitly incorporated. Chapter 4. A KBES Framework for Methods Selection 105 A simplified rule which i s more applicable for soldier piles and lagging (not applicable (N/A) for steel sheet piles since the GWSS installation and excavation are separate) i s as follows: RULE Select/Assign Process Model for DA(1) and CS(1) IF GWSS Installation Activity AND Excavation Activity = Intertwined AND Excavation Resources = Backhoe_Cat_245 AND Retaining System Resources = Crew_B_50 THEN Assign CPM = Process.GWSS{DA(i), CSQ) + Process Excavation {DA(m), CS(n)} AND Find Performance Measures The goal here is to guide the user in conducting an analysis through the use of rules. Standard models would be predefined so as to automate as much of the construction model process formulation as possible. Step 6 Control i s then transferred to the process models which are executed and the results interpreted (stage 2 and stage 3 ) . If the predicted production rate and cost are below the required rates, possible changes to the process model or resource assignment may be suggested to the user so that model changes can be made. If satisfactory time performance cannot be obtained, then the control system attempts to direct the search back to the f i r s t or second stage search processes. Given satisfactory time performance, the process i s repeated for cost performance. The total process is Chapter 4. A KBES Framework for Methods Selection 106 completed by assigning attribute values to the methods frame (stage 4). The control strategy for the above generic system was described using pseudo rules for i l l u s t r a t i v e purposes. They are not rigorous rules designed for implementation. 43 CMSA Development 4.3.1 Overview The scope of the prototype system i s described in this section. Central to the development of a methods selection system are representation issues as they relate to context, design alternatives, construction strategies, resource and process models, development of suitable control strategies and treatment of f e a s i b i l i t y c r i t e r i a . Effective approaches for dealing with these topics constitute the basic building blocks of the system. Thus, attention has been restricted to showing how a typical element at each level in the decision making hierarchy can be modelled, and how both the declarative and procedural knowledge required to generate the attributes/decision demanded at each level can be represented. In particular, a very restricted set of context, design, construction and resource alternatives have been selected for examination for the ground wall support system design element for Cut-and-Cover tunnelling. A description of two s o i l geotechnical contexts provides insights into the Chapter 4. A KBES Framework for Methods Selection 107 representation of project context.. Steel sheet p i l i n g has been chosen for the design alternative. Driving in singles or pairs has been selected for the construction strategies. The selection of a pile hammer provides the focus for resources, and determination of pi l e driving rate constitutes the process model. Design Element o Construction Strategy O Required Const-ruction Resources X I Construction Process Model SSP GWSS Alternatives Retaining-System_DE I $SP DE Pile Driving ln_Slngles CR 4 \ C r a n e CPM 1 Pile Driving InJPaIrs Construction Strategy Specification Pile Driver Resources Selection CPMJt Performance Prediction (Apply Dynamic Formula) (a) <b) Figure 4.5 Steel Sheet Pile (SSP) Method Frame Chapter 4. A KBES Framework for Methods Selection 108 This path through the decision alternatives at each decision level i s illustrated in figure 4.5. As described in section 4.2, a two stage process has been adopted as the basic decision model. This process i s depicted in figure 4.6. Preliminary f e a s i b i l i t y i s charged with the task of providing a short l i s t of potentially feasible alternatives, while detailed f e a s i b i l i t y has the task of confirming f e a s i b i l i t y and assigning the attributes of the design, construction strategy and resource alternatives selected. At the preliminary f e a s i b i l i t y level, broad brush rules are used to select and prune alternatives, based on an assessment of various c r i t e r i a which are encoded in a set of rules. This set corresponds to knowledge base 1 (KB-1) in figure 4.6, where knowledge bases at both levels are partitioned according to their corresponding c r i t e r i a : risk, design, construction resource compatibility, construction performance, and regulatory. While i t s features are described in section 4.3.3, i t has not been implemented in the current work. Instead, implementation efforts have concentrated on defining and implementing the knowledge bases at the detailed f e a s i b i l i t y level shown in figure 4.6. Chapter 4. A KBES Framework for Methods Selection 109 Figure 4.6 Prototype Model Chapter 4. A KBES Framework for Methods Selection 110 This section i s organized as follows. In section 4.3.2, representation of the project context in terms of s o i l conditions i s examined. In section 4.3.3, a descriptive model of the preliminary f e a s i b i l i t y stage i s presented. In section 4.3.4, the detailed f e a s i b i l i t y model is described for the alternatives previously identified. The presentation i s organized around subheadings of context, c r i t e r i a , control strategy and knowledge representation. Because the notion of a frame i s a useful way of identifying the attributes of a construction method, a conceptual frame is used throughout to demonstrate the build-up of the method attributes through preliminary, then detailed, f e a s i b i l i t y . 4.3.2 Context Modelling Context modelling, or representation, consists of two parts. The f i r s t deals with the representation of the i n i t i a l site conditions, contractual conditions, and so forth. The volume of information that has to be e l i c i t e d from the user can be substantial, leading to design issues regarding the user interface. For example, should information be requested only i f needed by a rule, or should a comprehensive description of project conditions be input f i r s t (a potentially onerous burden for the user). For the prototype work, the latter approach has been adopted. Chapter 4. A KBES Framework for Methods Selection 111 The second part of the context modelling deals with the derivation of additional context information through the application of engineering knowledge (e.g. lateral pressure) and exploration of tentative decisions regarding construction method attributes (e.g. space requirement for heavy equipment). Both of these aspects are examined below by way of the modelling of s o i l conditions, which have a significant affect on decisions relating to the selection of the ground wall support system and pi l e driving productivity. To have a small subset of s o i l types for the prototype system, four s o i l types were treated: soft clay, hard clay, loose sand, and dense sand. For each type, a data base of properties (dry unit weight, submerged unit weight, unconfined compressive strength, angle of shearing strength, etc) was developed. A maximum of two s o i l layers was treated (see figure 4.7). To gain a sense of how many s o i l profile scenarios could be generated by these four s o i l types and two layer model, consider this: for a single layer model, four options exist: soft clay, loose sand, etc; while for a two layer model, 12 combinations of conditions exist: "soft clay on s t i f f clay", "soft clay on Loose Sand", etc. Chapter 4. A KBES Framework for Methods Selection 112 Steel Sheet Pile Pile Refusal Depth Excavation Depth Ground Level Soil Layer -1-~Zi Water Table Loose to Medium Sand (Choices: Loose, Medium, Dense) Soil Layer - 2 -Sttff Clay (Choices: Soft, Medium, Stiff) (a) Soil Profile with Properties of Each Layer Entered by the User Pile Refusal Depth Excavation depth Steel Sheet Pile Standart Penetration Test (SPT Profile) - V_ Ground Level Soil Layer -1 — Water Table CSandJ" Soil Layer - 2 -(Clay) (b) Soil Profile with SPT Profile for Sand and Clay Figure 4.7 s o i l Profile Scenarios Thus s o i l profile combinations grow exponentially in size with the number of s o i l types and s o i l layers. The context model developed was rich enough to demonstrate the dependance of methods selection decisions on context Chapter 4. A KBES Framework for Methods Selection 113 variables. For example, the order and depth of each s o i l layer significantly influences the selection of the hammer as shown in section 4.3.4. The s o i l profile also affects the pace of pi l e driving as incorporated in the hammer energy formulas. The two approaches explored for modelling s o i l properties are described below. 1. Layer by Layer Property Specification The user i s queried as to the properties of each layer and the placement of the tunnel structure within the layers. Example data required by the prototype CMSA system are shown in table 4.3. Soil Type 1 for Layer 1 = Loose Sand Depth of Layer 1, = 40 ft Unit Weight of SoilTypel = 100 pcf Wet Unit Weight of Soiltypel = 120 pcf Submerged Weight for Soil Typel = 38 pcf Angle of Shearing Strength for Soil Type 1 = 30 Unconfined Comp. Strength for Soil Typel = NA. Depth of Water Table, = 20 ft Depth of The Tunnel or the Excavation = 60 ft Soil_Type_2 for Layer 2 = Soft to Med. Clay Depth of Layer 2 = 17 ft Unit Weight of Soil_Type_2 = 111 pcf Submerged Weight for SoilType 2 = 52 pcf Angle of Shearing Strength for Soil Type 2 = NA. Unconfined Comp. Strength for Soil Type 2 = 120 pcf Table 4.3 Soil Profile Input-Format 1 2. Soil Penetration Test (SPT) Model Figure 4.7 (b) shows two s o i l layers identified as sand and clay. Indirect relationships exist among the Standard Chapter 4. A KBES Framework for Methods selection 114 Penetration Test (SPT) profile, s o i l resistance and s o i l properties for each type of s o i l layer. Using these relationships, CMSA infers s o i l properties through a data base of SPT profile versus s o i l properties for a s o i l type. A sample of s o i l information required by CMSA as provided by the user for the SPT model i s shown in table 4.4. It i s elaborated upon in table A.l in Appendix A. Once a s o i l model i s selected and parameter values specified by the user, additional context information essential to the selection of design alternatives, construction strategies and resources can be generated. Most of this information can be derived using procedural knowledge. The routines used for determining lateral earth 2Typically, a standard penetration test (SPT) readings are taken at 3 f t or 5 f t intervals to the refusal depth of the pil e . Here, we approximate the readings interval as every 10 f t . SoilTypel of Layerl Depth of Layerl Soil_Type_2 of Layer_2 Depth of the Tunnel Depth of Water Table "LooseSand" Sand 40 ft "StiffClay" Clay SPT at depth 10 ft SPT 2 at depth 20 ft SPT at depth 30 ft SPT at depth 40 ft SPT at depth 50 ft SPT at depth 60 ft SPT at depth 70 ft 65 ft 17 ft 30 20 21 11 15 30 33 Table 4.4 Soil Profile Input-Format 2 Chapter 4. A KBES Framework for Methods Selection 1 1 5 pressure and moments for the ground wall support systems are described in Appendix A. The output from these routines forms an important part of the technical f e a s i b i l i t y check at the design alternative level. Appendix B contains the derivation and the code for a prediction algorithm to forecast the pil e driving productivity rate and check i f p i l e refusal depth has been achieved. The value of this approach was confirmed through interviews with personnel from a pile driving company (see Appendix C). 4.3.3 Preliminary Feasibility Figure 4.8 presents a way of prescribing a GWSS construction method frame throughout f e a s i b i l i t y synthesis over time: a frame builds up incrementally, while slots are instantiated and added over time. This figure alludes to the two levels of f e a s i b i l i t y where primary and secondary variables are evolved. GWSS methods w i l l be retrieved from a data base characterized by universal attributes (slots) corresponding to method elements and knowledge attached to them. A GWSS attribute contains technical f e a s i b i l i t y , c r i t e r i a , knowledge, context, and control strategy, and i s subjected to the f e a s i b i l i t y checks of risk, design, compatibility, construction performance and regulatory environment. Not Chapter 4. A KBES Framework for Methods Selection 116 a l l of these factors are applicable for a GWSS attribute and/or project scenario. The preliminary knowledge base, KB-1 in figure 4.6, identifies, screens, and ranks, feasible alternatives at this level, and then passes control to the detailed f e a s i b i l i t y level for further method specification. The goal i s to reduce the solution set as quickly as possible by treating various f e a s i b i l i t y c r i t e r i a in the form of high level rules. The knowledge here is declarative, based mainly on experiential and common sense engineering knowledge. This knowledge, in the form of rules, i s attached to each method alternative in the data base. Condition variables in these rules are assigned values based on context parameter values, and then the rules are evaluated. GWSS Gene/a/Frame Method Attribute (I) -Technical Feasibility - Criteria - Control Strategy - Context Feasibility Checks -Risk - Design - Compatibility - Construction Performance - Regulatory 1 Steel Sheet Pile Frame -Criteria Rule of "Thumb - Control Strategy Eliminate SSP if Attribute (i) Is Not Feasible Design Element - Technical Feasibility SSP strength. Soil Permeability,.. -Context Project Location, SPT Profile, Construction Strategy - Technical Feasibility SSP strength,.. -Context SPT Profile, Target Cost and Production Rate,.. Required Construction Resources - Technical Feasibility Capacity, Space,. -Context SPT Profile, Target Cost and Production Rate,.. SSP Strength (Z section), Construction Process Uodel -Context CPM, Simulation, Queuing Theory,.. Feasibility Checks -Risk -Design -Compatibility - Construction Performance - Regulatory Soldier Piles and Lagging Slurry Trench Wa> GWSSfl) Steal Sheet Pile Frame - Control Strategy Eliminate SSP if Attribute (i) Is Not Feasible Design Element (PZ_27) • Criteria (Strength), Deflection, Buckling. - Context SPT Profile, Earth Pressure SSP properties,.. Construction Strategy (Pile Driving ln_Pairs) - Criteria Trial and Error • Context SPT Profile, SSP Type, Hammer Type, ..-Required Construction Resources (Hammer SAAH01, Crew_BS. 30 T Crane,..) - Context SPT Profile. Target Cost and Production Bate,.. Available Resources, Resources Compatibility,. Construction Process Model - Criteria Dynamic Formulas, WEAP,. Feasibility Checks -Risk -Design - Compatibility • Construction Performance -Regulatory Soldier Pile* and Lagging Slurry Trench Wall GWSS (I) Chapter 4. A KBES Framework for Methods Selection 118 The Control Strategy at this level may pass control v e r t i c a l l y to a lower level component or horizontally to another rule category when necessary. For instance, for a given shoring alternative and project context, the risk component may be processed f i r s t , followed by design f e a s i b i l i t y , construction resources f e a s i b i l i t y , performance f e a s i b i l i t y , and so forth. Another alternative with the same or different project context may dictate checking the design f i r s t , followed by performance measures test, and others. In other words, the search strategy i s sensitive to project context and user intervention . For CMSA, KB-1 initiates the solution process by testing the preliminary f e a s i b i l i t y of the risk component. After satisfying the risk requirement, control passes to the design component. If the user wants to override this, he instructs the system to pursue further low level risk analysis. Then design, construction resource compatibility, construction performance, and regulatory f e a s i b i l i t y are checked, subject to user/project context intervention to re-direct control. High level rules are used to represent c r i t e r i a knowledge for each alternative. A rule consists of a RULE heading to describe the rule, IF conditions that represent high level premises (which may embody a another layer of rules), and THEN consequences or actions that take place i f this rule is Chapter 4. A KBES Framework for Methods Selection 119 fired. Control Strategy clauses are labelled to indicate passing of control. This highlights passing the control in either direction: horizontally within one level, and/or ve r t i c a l l y between the two levels of f e a s i b i l i t y as shown in figure 4.6. I Rule Condition CMSA Default Value 1 Assign Object Attribute Value Execute the Rule Control Row Figure 4.9 CMSA Rule Execution Loop Figure 4.9 shows a typical CMSA rule execution order. Each condition (or premise) must have a value, where some Chapter 4. A KBES Framework for Methods Selection 120 parameters' values are provided by the system, subject to user override, while others must be provided by the user. Once a rule i s fired, then, some conclusion clauses f i r e to perform another operation and/or pass control to another hypothesis. For instance, i f a rule in the preliminary f e a s i b i l i t y failed, then CMSA is supposed to stop the session, however, the user may override this and command the system to pursue the solution. We i l l u s t r a t e and describe each f e a s i b i l i t y check for each rule category for our Steel Sheet Pile, GWSS design alternative. The SSP alternative i s presumed to exist in the methods database with pertinent knowledge for each f e a s i b i l i t y check. A risk factor i s f i r s t applied before further refined information i s required from the user for other factors. The system evaluates the SSP alternative risk-wise by checking variable conditions such as the state of water washout. These conditions are either volunteered by the user, or provided by the system as a system default, database, context, etc. A typical rule for risk f e a s i b i l i t y i s shown below. Chapter 4. A KBES Framework for Methods Selection 121 1. Risk Is Feasible RULE RisklsFeasible for A GWSS IF GWSS is SteelSheetPiles AND WaterJLevel/TunnelDepth Is High AND Water Washout.Risk Is Low OR Local_Expertise Is Medium AND Hypothesis Expected_SSP.Risk = < Catastrophic.Risk THEN Steel_Sheet_Pile.Risk Is Acceptable ELSE Steel_Sheet_Pile.Risk Is Not_Acceptable Source Database/User User Context, User User {Control Clauses if Acceptable} AND/OR Check DesignFeasibility AND/OR Check DetailedDetailed for Risk Analysis AND/OR Check Performance Measures Are Feasible {Control Clauses if NotAcceptable} AND Eliminate SSP AND Go to Next GWSS Alternative The premises of this rule, and other rules to follow, are used to perform the f e a s i b i l i t y check. The premise conditions are high primitives and expect a boolean answer of YES or NO. If any one of these conditions f a i l s , then the rule f a i l s , and ELSE of the control (strategy) clause decides what to do next. For instance, the above rule reads that WaterJWashout .Risk_Is_Low implies that the user has answered YES to the question "Is there a Low_Risk for a Water_Washout ?". If the answer i s YES for this clause and other IF clauses, then Steel_Sheet_Pile.Risk i s Acceptable. The system keeps the boolean value of this rule as TRUE. If the same clause of the Risk.level was not low, then the rule Chapter 4. A KBES Framework for Methods Selection 1 2 2 f a i l s and i t i s the task of the control strategy to decide what to do next. Multiple values for the state of water washout such as MediumRisk,or HighRisk, where i t jeopardizes GWSS fe a s i b i l i t y , may be considered by evoking risk rules which contain these values in their premises. The LocalExpertise state i s treated in the same manner. This i s a high level primitive clause of great importance in assessing the risk. Interviews with a contractor (Appendix C) indicated that local practice is very important when assessing risk. Even i f a method is proven and well known elsewhere, i f local expertise i s not available, the method is perceived as having high, and potentially unacceptable, levels of risk. This argues strongly for using methods familiar to a firm's practices. This i s evident in the interview section of Appendix C, in which a contractor indicated Shot-Crete is a common shoring method in Vancouver, whereas, in the Toronto area, H piles and lagging are more prevalent. Although this clause, Local_Expertise conditions, may deserve a component by i t s e l f , i t was combined together with the risk component for knowledge organization. It i s not the intent of this thesis to identify and analyze i t s attributes. If the risks associated with the method are feasible, then Control i s transferred either horizontally, to check Chapter 4. A KBES Framework for Methods Selection 123 other components, or vertically, to further check detailed level risk assessment i f the user chooses to. If the alternative i s found to be unacceptable (infeasible), then SSP i s eliminated and control picks another candidate. The design f e a s i b i l i t y at this level i s basically technical. That i s , given project and site conditions, i s the design feasible with regard to leaving adjacent structures undamaged, testing i t s constructability, etc. A representative rule for design f e a s i b i l i t y is exemplified next. 2. DesignAlternative IsFeasible RULE DesignlsFeasible Source IF Adjacent Structures Settlement Are Tolerable User AND Pile Driving Condition Are Not Hard Context, User AND Water_Tightness Is_Not Necessary User, Context AND THEN GWSS of Steel_Sheet_Piles Design Is Acceptable ELSE GWSS of Steel Sheet Piles Design Is NotAcceptable {Control Clauses if Acceptable} AND/OR Check ResourcesCompatibilitylsFeasible AND/OR Check PerformanceMeasuresFeasibility AND/OR Check DetailedFeasibility for Design {Control Clauses if NotAcceptable} AND Eliminate SSP AND Go to Next GWSS Alternative The f i r s t premise examines whether settlement of the adjacent structures to the shoring GWSS is permissible, i.e. tolerable. The user i s assumed to volunteer the answer. Chapter 4. A KBES Framework for Methods Selection 124 The p i l e driving conditions are essential since SSP are installed by this technique. Therefore, i t i s desirable to satisfy this condition. The third premise tests the importance of water tightness which depends on the water table level. For some projects i t i s important to have water tightness, while others do not stipulate i t . In this case, this condition i s phrased such that water-tightness has to be satisfied. The source of this information could be either the user or the context, whereby the system presumably assesses the water table elevation versus project conditions and returns an answer. The above rule checks the f e a s i b i l i t y of the sheet piles design (or other GWSS) within a given project context. Further, i f this rule is fired successfully, control then may be transferred to check horizontal components or proceed to a lower level of detailed design. Next, control passes to check the resource compatibility for the given design alternative, where available equipment and crew spreads are identified and balanced for smooth production. The following rule demonstrates a compatibility check. Chapter 4. A KBES Framework for Methods Selection 12 5 3. Resource Compatibility Is Feasible RULE ResourceCompatibilitylsFeasible Source IF Soil Profile Is Cohesive Context, User AND Required Hammer Energy > 33,000 lb. ft. Context, User AND SSPDesignation Is PZ27 Context, User THEN (Hammer Size + SSP Size + Soil Profile) Is Technically Feasible User, Context AND HammerandSSP Are Compatible {Control Clauses if Accepted} AND Check DetailedFeasibility for ResourceCompatibility AND Check Performance_Measures_Feasibility {Control Clauses if NotAcceptable} AND Eliminate SSP AND Go to Next GWSS Alternative Resources, including design materials, must be compatible. For instance, some materials are only constructed with a subset of existing tools and/or equipment, or vice versa, where some materials have to be substituted to suit existing resources, such as crew expertise and available equipment. In this vein, the f i r s t premise in the above rule pronounces the s o i l context condition to be cohesive s o i l . The second premise dictates a lower bound for the hammer power. This i s provided by the context which draws from a lower level rule that relates the likelihood of SPT bring greater than a threshold and minimum hammer energy, thus defining a feasible hammer subset. The third premise Chapter 4. A KBES Framework for Methods Selection 126 identifies the type of SSP section as PZ sections, which are heavy and have a high section modulus. This i s based on the context or user volunteered information. For the former, an implied relationship between the pil e driving conditions, and/or hammer size, versus SSP minimum size, i s established, reducing the set of feasible SSP. The object i s to reduce set of available l i s t s of hammers and SSPs into reduced feasible subsets that satisfy expected goals, e.g. cost and time. Control passes horizontally to check the performance measures f e a s i b i l i t y . If the horizontal inference carries on, the next performance f e a s i b i l i t y w i l l be checked at this stage. This component establishes upper and lower bounds for production and costs from previous projects, and/or unit cost data manuals, given for a project context scenario including s o i l profile and equipment/material spread. The following rule mirrors this component function. Chapter 4. A KBES Framework for Methods Selection 127 4. Performance Measures Feasibility RULE Performance Measures Are Feasible Source AND IF Lower_Prod._Bounds < ProductionRate < Upper_Prod._Bounds LowerCostBounds < ProductionCost < Upper_Cost_Bounds User User AND THEN Steel Sheet Piles.Performance Measures Are Feasible {Control Clauses if Accepted} AND AND/OR Check Detailed_Feasibility for Other Components Check Detailed_Feasibility for PerformanceMeasures {Control Clauses if NotAcceptable} AND AND Eliminate SSP Go to Next GWSS Alternative These upper and lower bounds are goals (stated in the contract as unit cost and construction duration) , and the user i s supposed to estimate the Production_Rate and Production_Cost for a given design alternative, and construction resources and processes. At this level, the estimates are conceptual, a rule of thumb based on experience from a project context, whereas further precision of estimates could be done at the lower level. Control then passes to the regulatory component to check compliance with regulatory and safety requirements. An example i s presented in the next rule. Chapter 4. A KBES Framework for Methods Selection 128 5. Regulations Are Satisfied RULE Regulation Is Satisfied Source AND AND AND IF OSHA and Local Safety Regulations Are Satisfied Environmental Hazards Are Acceptable Pile Driving Level of Noise Is Acceptable Hammer Vibrations Are Acceptable User User User, Context Context, User THEN Regulatory Conditions Are Satisfied {Control Clauses if Accepted} AND/OR AND/OR AND/OR Evaluate Another GWSS Alternative Rank Preliminary Feasible Alternatives Check Detailed_Feasibility Regulations {Control Clauses if NotAcceptable} AND AND Eliminate SSP Go to Next GWSS Alternative The f i r s t premise ensures that safety regulations relevant to working conditions for labor are satisfied. This means that proper labor allocation, labor protection, and so forth, have to be met for the major operations (excavation, p i l e driving, muck removal, etc) of a proposed method, where some methods require more consideration than others. The user has to ensure the validity of this clause since there are numerous provisions to meet. The second premise veri f i e s acceptance for the environmental hazards of a method. For example, i f diesel material disposal i s not possible in some areas, the diesel hammers class w i l l be dismissed. This condition must be verified by the user. Chapter 4. A KBES Framework for Methods Selection 129 The third premise sets a lower bound for an accepted noise emanating from p i l e driving, e.g. noise from construction i s being severely limited with an objective of not more than 85 dB at 50 f t created by compressors for hammers and vibros (Hunt 1979). This limitation could be checked by the derived context where properties of pile driver components are retrieved from i t s data base and confirmed by the system. The user may exercise his judgment as to whether further restriction are warranted. The fourth premise ensures that vibrations emanating from p i l e driving do not destabilize surrounding structures, where, in some cases, surrounding structures are braced or other precautions are warranted. Similar to the previous premise, user or derived context could be used to approve this condition. Once this rule category i s satisfied, and by implication, others as well, then a GWSS method alternative is considered to be preliminarily feasible. In this example, steel sheet pile , i s regarded as preliminarily feasible. After an SSP is successfully selected, another GWSS alternative in the l i s t of the available alternatives w i l l be tested for preliminary f e a s i b i l i t y by repeating the same cycle shown above. The foregoing control structure may allow solution synthesis to continue even i f a high level condition failed Chapter 4. A KBES Framework for Methods Selection 130 (say regulatory condition) , or was violated, within limits. The system may allow the user to further pursue and explore partial solutions for the components that remain feasible. Therefore, a tentative l i s t of preliminarily feasible alternatives i s available for further synthesis at the detailed level. 4.3.4 Detailed Feasibility Level The detailed f e a s i b i l i t y level contains detailed design and analysis knowledge which is organized into several knowledge bases as shown in figure 4.6. The objectives of this level are to confirm the f e a s i b i l i t y of alternatives produced at the preliminary f e a s i b i l i t y stage, and to complete the frame description of each method which survives a l l f e a s i b i l i t y checks. The control strategy guides this process beginning with the design component knowledge base (KB 3,4, and 5). At f i r s t , technical f e a s i b i l i t y for an attribute i s sought through simplified analysis and design procedures (KB-3). For instance, i f the available sheet piles, represented by a data base, do not contain the required section, then a message is sent to the user advising of the technical i n f e a s i b i l i t y of the attribute. Control i s then passed to the next preliminary GWSS alternative. Assuming the steel sheet pile (propped sheet piles) design element was feasibly sized, control i s passed to the Chapter 4. A KBES Framework for Methods Selection 131 p i l e driver selection and sizing knowledge base (KB-4). Based on the s o i l profile and conditions, and specified design element of the steel sheet p i l e attributes, the system attempts to pick the most suitable and productive hammer type and size from a hammer data base. The hammer selection must satisfy the technical f e a s i b i l i t y conditions. On the other hand, i f the hammer type and size i s selected f i r s t , because of availability, this may dictate the size of the sheet p i l e . Thus, steel sheet and hammer sizing could be reversed. For CMSA, the former knowledge processing approach design element sizing to hammer sizing i s adopted. Included in the technical f e a s i b i l i t y test for the hammer type and sizing, i s the pile driving strategy (KB-5), i.e., drive piles in singles or in pairs. (Other strategies may include other wave patterns). The choice of either strategy is dependant on the s o i l conditions, p i l e driving energy, and so forth. After a steel sheet pil e , hammer, and strategy are specified, a prediction module i s applied to predict the performance measures (time and cost) for the candidate alternative (KB-6). After the cost estimate for the alternative i s attained, control i s passed to the risk assessment routine (KB-2), which uses the risk assessment framework described in section 4.4. It is used to determine whether or not the risks associated with the GWSS alternative, when priced out, exceeds some maximum threshold Chapter 4. A KBES Framework for Methods Selection 132 value. If they do, the alternative i s deemed to be infeasible and the control strategy moves to the next candidate. If the alternative's risk costs are acceptable, then i t passes control to the diagnosis or analysis component (KB-7 ) . If the method synthesis (technical feasibility) i s not satisfied, then KB-7 recommends a change in either construction strategy, hammer type or size, or steel sheet pi l e type, size, and grade. Currently, only the f i r s t is automated, with the alternatives being to drive piles in singles or in pairs. If the recommendations s t i l l yield an infeasible SSP GWSS alternative, then the alternative i s declared to be infeasible. If i t i s technically feasible, then production rate and unit cost are determined using the prediction module of KB-6. At the end, an evaluation criterion w i l l be chosen to rate the feasible SSP alternative in order to rank i t with other successful candidates. Section 4 . 5 elaborates on several c r i t e r i a schemes. Other issues relevant to this section concern what to do when a solution f a i l s within the system — what kind of remedy to explore, what attributes values have to be remembered, where to track back to, and when the system stops and declares an alternative i s not feasible within the system. Chapter 4. A KBES Framework for Methods Selection 133 In the discussion that follows, a step-by-step approach is presented, along with i l l u s t r a t i v e rules to demonstrate the approach just outlined. First, we examine the design element synthesis. 1. Design Element Synthesis (KB-3) Attention i s directed at sizing the design element (in this case the steel sheet piling) to satisfy the technical f e a s i b i l i t y requirement. For this example, technical f e a s i b i l i t y means that given default attributes for a support system, piles can be sized so that stress and deflection c r i t e r i a (not implemented) are satisfied. A three part process i s adopted. First, a rule-based approach is used to determine pressures and moments. Second, a search i s made for a sheet pile that satisfies the allowable stress c r i t e r i a . Third, i f a pi l e can't be found, the retaining strategy i s altered (spacing of struts and/or wales i s involved) and control i s passed back to the second step, with iterations occurring u n t i l either a feasible steel sheet p i l e design i s confirmed or no feasible solution exits. This third step has not been implemented in the prototype. Example rules for the f i r s t two parts of this process are now described. 1.1 Pressure and Moment Calculations (KB-3-1) Rules are used to retrieve the required properties corresponding to the s o i l strata scenario (context Chapter 4. A KBES Framework for Methods Selection 134 information) , to assign the default horizontal and vertical spacing for the retaining spacing (struts spacing horizontally 12 f t and vertically 15 f t ) , and to perform the calculations of pressure and moments. Details of the calculations are given in Appendix A. An example of the rule format follows. Rule: Compute Earth Pressure and Moments for the two Soil Layers Scenario Source IF Soil_Layers.Number is Two User AND Soil_layer_l is Loose_Sand User AND Soil_Layer_2 is Stiff_Clay User THEN Maximum_Lateral_Pressure = (Loose Sand.Unitweight * LooseSand.Depth * (K(a) for LS) + StiffClay Unit Weight * Stiff_Clay_Layer.Depth) * (K(a) for SC) AND Moments for Steel Sheet Piles = Maximum_lateral_Pressure * L (vertical spacing) ~2 / 8 {Control Strategy} AND Specify the Steel Sheet Pile Section The f i r s t three premises query the user for the s o i l p r o f i l e . The system retrieves relevant s o i l properties to compute the lateral earth pressure — unit weight, K(a) , etc. In the concluding part, pressure and moments are computed (See Appendix A) , and control i s passed to specify the sheet p i l e . 1.2 Selecting and Sizing Design Elements [KB-3-2] Moment information i s passed from the f i r s t step to the second, and combined with an allowable stress condition to determine the section modulus required. If the modulus required exceeds the maximum size available in the data Chapter 4. A KBES Framework for Methods Selection 135 base, the current design i s technically infeasible. Feasibility may be achieved by modifying the spacing of the retaining system. The KB-3 design component contains mainly factual rules applied for sheet piles, soldier piles, lagging, wales, and struts. It currently uses a single criterion based on strength. Other design c r i t e r i a such as deflection, allowable settlement, and so forth, have not been treated. A typical rule employed i s : RULE IF AND THEN AND AND Select a Steel Sheet Pile SSP.Section_Modulus = Maximum_Moments / (Fb * Fs) 38.3 in~3 < SteelPileSection Modules < = 46.8 in^ SSP.designation Is PZ_27 Retrieve PZ27 Properties Database Calculate the Quantity Take-Off for Sheet Piles Source Context (step 1) Context {Control Strategy} AND Select a Hammer Class This rule i s internal, where derived context from the previous rule i s used to specify a SSP of PZ_27. Control i s passed to hammer selection next. K B - 3 - 3 , although not implemented, would permit changing the retaining system in an attempt to find a feasible steel sheet p i l e size. Chapter 4. A KBES Framework for Methods Selection 136 2. Specifying A Construction Resource (Resource Level) 2.1 Selecting a Hammer (KB-4-1) The main types of hammers include: 1. Drop hammers; 2. Steam hammers (Single and Double Acting), 3. Air hammers (Single and Double Acting); 4. Diesel Hammers (Single, Double, and Differential Acting hammers; 5. Hydraulic hammers; 6. Vibratory Hammers; and 7. others. Selecting the most suitable p i l e hammer for a given project involves the consideration of several factors, such as size and piles types, number of piles, characteristics of the s o i l , location of the project, topography of the site, type of r i g available, and the types and sizes of hammer owned by contractor. A p i l e driving contractor usually i s concerned with selecting the hammer that w i l l drive the piles for a project at the lowest cost within the required production rate. Broad brush rules found in the literature are similar to the one shown in table 4.6, which recommend the hammer most suitable for different homogeneous s o i l classifications. Such a table i s convenient for a One_Soil_Layer representation; however, the selection process i s more complicated when two or more s o i l layers are present. The Chapter 4. A KBES Framework for Methods Selection 137 selection depends on the ordering-of the s o i l layers, the depth Of each layer, and the SPT profile. For example, i f the top layer i s soft s o i l and the lower layer i s dense s o i l , a vibratory, light impact hammer i s used to drive them to the dense layer, and then another, heavier hammer i s ut i l i z e d to drive them to refusal. 8AND8 (N0N-C0HE8IVE 80IL8) Wood Pipe Open Pipe Ctoeet H-Beam 8heet Pile Cone-orvte Very LOOM DA vgjB, VjNB) VgjB, VgjB) DA \ QQM9 DA V£NB) DA V(NB) v r , DA Medium SA VjNB) DA VjNB) VgjB) SA Dent* SA V£A' SA VgjB) VgjB) SA Very Dense SA SA SA SA VgjB, SA (a) CLAYS (COHESIVE 8OILS) Wood Pipe Open Pipe Cloeec H-Beam Sheet Pile Cone-crete Very8oft DA VgjB) DA VgjB) V SA Medium DA V^ NB) SA V(NB) DA V DA SA Stiff SA DA SA DA DA SA VeryStltf SA SA SA SA SA SA Hard SA SA SA SA SA SA W DA • Double Acting (Diem or Mr/Steam) SA - Slngl* Acting (Oleeri or Air/Steam) V-Vlbratory NB • No Bearing Formula Required Table 4.6 Hammers for Different Soils [from Barber 1987] Chapter 4. A KBES Framework for Methods Selection 138 A typical rule format for the single s o i l layer case i s : RULE Select A Single Acting Air Hammer Source IF Soil_Layers.Number = One Soil Layer User AND Type of Soil_Profile = Cohesive Soil3 User, Context THEN Select An Impact Hammer Context (Experiential) AND Choose A Single Acting Air Hammer OR Choose Double Acting Air Hammer Control Strategy AND Specify a Hammer Size The f i r s t and second premises query the user for the s o i l p r o f i l e . The latter inquires about the cohesiveness of the layer although the system is already aware of the s o i l layer properties. The reason for this i s to allow the user to use his judgement in determining this quality since this clause, i f true, excludes the vibratory class from consideration and thus focuses on the impact hammer class. For two s o i l layers, there could be several feasible hammer alternatives — e.g. use a single hammer (impact or vibratory), use a combination of hammer types, or a range of sizes of the same hammer (use a lighter one to drive the top, soft layers and the heavier one to drive piles to their refusal depth). Vibratory pil e driver use i s not recommended for a s o i l p r o f i l e with a sizable cohesive layer (e.g. clay). Therefore, this premise eliminates the vibratory p i l e driver subset and examines impact hammers only. Chapter 4. A KBES Framework for Methods Selection 139 A typical rule format for the two s o i l layers case i s : RULE Select A Single Acting Air Hammer Source IF Top_Soil_Layer is LooseSand Context AND Lower Soil Layer Is Stiff Clay Context AND Stiff_Clay_Layer.Depth > 15 ft Context THEN Choose A Single Acting Air Hammer AND Size the Hammer {Control Strategy} AND Specify a Hammer Size The above rule checks the cohesive s o i l layer thickness to a threshold which excludes vibratory hammers. If the clay depth to the sand layer ratio i s very high then a vibratory p i l e driver could be favorable. Thus, for this rule, the impact hammer, single acting a i r hammer, i s selected with reference to table 4 . 5 . 2.2 Sizing The Hammer [KB-4-2] The sizing process starts by selecting the highest hammer theoretical energy. This i s consistent with the "greedy" algorithm, described previously, in which maximum production rates and reserve capacity are sought. Two additional conditions are examined however. First, damage to the sheet pi l e must be avoided. Second, the hammer must be capable of driving the pi l e to refusal depth 4. If either of these 4Refusal i s the depth to which piles have to be driven, to attain their designed resistance strength through skin and end bearing. For non-displacement piles (see figure 4 . 7 ) , sheet piles and soldier piles, their depth extends below the tunnel bottom (excavation depth) by 5 f t to 15 f t (Winterkorn and Fang 1975) . Chapter 4. A KBES Framework for Methods Selection 140 conditions cannot be met, then either the hammer energy i s decreased and/or the sheet p i l e size i s increased. If no satisfactory solution can be found, the GWSS alternative of SSP i s considered to be infeasible. A rule format for determining Hammer Size according to SSP size, using empirical knowledge, i s shown below. RULE Size the Hammer ' Source IF SSPCrossSectionArea is Ap Context (SSP Database) AND Hammer_Type Is SingleActingAirHammer Context THEN HammerSize (Hammer_Rated_Energy) Context = < 3,000 * Ap (lb-ft) AND Single_Acting_Air_Hammer.Size > (DAAHDatabase), Context = RequiredHammer.Size {Control Strategy} AND Do Pile_Driving_Strategy The above rule, based on contractors' experience, determines the maximum magnitude of the hammer energy by multiplying 3,000 lb/in A2 by SSP_Cross_Section_Area to prevent p i l e damage. After executing this rule successfully, control i s passed to selecting the pile driving strategy. 3. Pile Driving Strategy fKB-5") Soil conditions and hammer power dictate the pi l e driving strategy in terms of driving in singles or pairs. Secondary factors relevant to this strategy are SSP size and length of Chapter 4. A KBES Framework for Methods Selection 141 pi l e segment. To prevent buckling, the maximum allowable driving depth with respect to adjacent piles i s < = 13 f t . As a greedy approach i s favoured, and as other contractors have suggested that trying to drive piles in pairs i s a preferred approach, the pi l e driving strategy In_Pairs i s selected f i r s t . Driving conditions are restricted to two states: soft and hard driving Conditions. They are inferred from the s o i l conditions using simplified rules. For the former, soft driving conditions, the driving strategy i s In_Pairs; while the latter driving strategy i s InSingles. Other pertinent factors such as hammer type and size, type and grade of sheet piles, number of p i l e drivers and complexity of the project were not considered e x p l i c i t l y . The following rules exemplify the conditions for pile driving In_Singles and In_Pairs. Pile driving In_Singles rule (experiential) i s : RULE Drive in Singles Source IF PileDriving.Conditions Are Hard Context, User AND Refusal.depth > 30 ft Context AND Hammer.size < 22,000 Ib-ft Context THEN Drive Piles In_Singles {Control Strategy} AND Do Performance Measures Chapter 4. A KBES Framework for Methods Selection 142 Pile driving In_Pairs rule (experiential) i s : RULE Drive Piles in Pairs Source IF Pile Driving.Conditions Are Soft Context, User AND RefusaLdepth < 90 ft Context AND Hammer.size > 10,000 Ib-ft Context THEN Drive Piles In Pairs {Control Strategy} AND Do Performance_Measures Predict (KB-6) The steps involved are: 1. Determine production rate and cost. 2. During p i l e driving production analysis, a pile damage check i s involved (blow counts) If blow count exceeded, then backtrack and change one or more choices Otherwise determine production + cost. If not acceptable, then . . In choosing between pile driving ln_Pairs or In_Singles, consideration must be given to both set up time and driving time. Production Time and Cost Performance Given a feasible SSP size, hammer type and size, and driving strategy, then i t remains to determine time and cost performance, and perform other checks, such as regulatory considerations on noise l e v e l 5 . 5Safety and regulation factors are not implemented at this level. The noise level i s implicitly satisfied as affirmed by the preliminary regulation f e a s i b i l i t y check. Chapter 4. A KBES Framework for Methods Selection 143 Using a hammer dynamic energy (modified Engineering News formula 6) production routine written in C (Drive.c), control i s passed to this routine for purposes of computing a p i l e driving production rate excluding fixed set up time. Output information i s passed back to the control strategy which in turns interprets the routine results. A constraint i s included dealing with the maximum number of blows per foot, beyond which p i l e damage is l i k e l y . In the event that this blow count i s reached, the routine stops summing up the incremental production time and returns a message to CMSA indicating this violation and where i t happened — i.e. the depth of p i l e where i t interrupted driving. The hammer blow count i s empirical and can vary with different soil/pile/hammer where technical f e a s i b i l i t y i s monitored by observation. For CMSA, the system takes action based on a theoretical blow count from the model derived in Appendix B. bThis formula, among numerous of hammer energy formulas, i s applied only to some types of impact hammers. For vibratory p i l e drivers, rule of thumbs are u t i l i z e d to estimate their productivity. Chapter 4. A KBES Framework for Methods Selection 144 The rule format for running the "Drive.c" routine i s : RULE Do Performance Measures Source IF Selected_SSP_Type Is PZ27 Context AND Selected Hammer = Single Acting Air_Hammer (SAAH) Context AND SAAH.Rated_Delivered_Energy = 15000 Ib-ft Context AND Pile Driving Conditions Is Soft Context AND PilesDrivingStrategy Is "InPairs" Context {Control Structure if Accepted} THEN Compute ProductionRate AND DoSSP.Risk Assessment {Control Structure if Unaccepted} ELSE 8 Analyze Technical Feasibility The above rule pools specified method attributes — sheet pi l e , hammer and pi l e driving strategy, and sends them as input parameters to the Drive.c routine. Figure 4.10 shows input parameters for the Drive.c routine passed by CMSA via an input text f i l e which then i s processed by the routine. After executing the numerical routine, output parameters are passed back to the control strategy for interpretation and further manipulation. Output variables include incremental and cumulative values for each of skin f r i c t i o n , end bearing s o i l 'Blow count satisfaction is implicit in this clause. If the threshold blow count per foot i s violated, then control is passed to the technical f e a s i b i l i t y analysis to investigate a remedy. Technical f e a s i b i l i t y at this stage refers to whether a method attributes combination (SSP + Pile Driver + Construction Strategy) achieves i t s goals of production, cost, damage free driving, and so forth. Chapter 4. A KBES Framework for Methods Selection 145 resistance, blow count, and production progress (relationships and runs are detailed in Appendix B). Drlve.c Input File /Git •/Soil Profile SSP properties Hammer Properties. Compute: Skin Friction End Bearing Friction Hammer Blows Rate Incremental Pile Penetration Rate NO / Is Average _ Set Satisfied? YES Compute Pile Driving Cycle Duration Figure 4.10 Drive.c Routine Interface with CMSA A criterion of maximum acceptable hammer blow count of 150 blows/ft (a bench-mark from f i e l d engineers to interrupt i f refusal i s reached) i s set as the threshold. If the blow count exceeds this limit, then the Drive.c routine stops Chapter 4. A KBES Framework for Methods Selection 146 computation, flags the depth where i t happened and sends a message to the control strategy which in turn interprets that the method combination (technical feasibility) i s not feasible, or else, i t i s feasible and the production rate i s passed by the routine to the control strategy. After the SSP alternative passed this test successfully, the quantity take-off and cost estimate computation for the whole project follow. Once detailed method cost i s known, a risk assessment i s ensued subsequently. 5. Risk Assessment (KB-2) The risk factor has a significant impact on the choice of construction methods in general and on shoring method selection in particular. The considerable emphasis placed on an informal risk assessment, particularly with respect to the likelihood of catastrophic risk, was highlighted in an interview with a seasoned construction engineer (see Appendix C). Based on discussions with construction personnel, a review of the literature and an analysis of the amount of data l i k e l y to be available when making decisions about construction methods, a simplified analytical CMSA risk assessment framework was developed as described in section 4.4. What i s important i s the role of risk in the control strategy. If the potential for catastrophic or unacceptable Chapter 4. A KBES Framework for Methods Selection 147 risks are high, then a method w i l l be dropped, not withstanding i t s appeal because of time or unit cost performance. As described in section 4.4, the risk model involves three states of nature of geological conditions, given a design alternative. The three states correspond to better than, equal to, and worse than expected conditions. Several cost categories describe the outcome for each state of nature. An example rule for the case when the state of nature i s AsExpected, i s as follows: RULE Compute Risk for "As-Expected" State of Nature Source IF GWSS = Steel_Sheet_Pile (SSP) Preliminary Feasibility AND SoilConditions Is AsExpected Context AND As_Expected_Conditions.Likelihood = Pp User AND Consequence Costs (Dp) = Sum (Dp(i)) User, Context THEN Compute AsExpected RiskCostComponent AND AsExpectedConditions.RiskComponent = (Pp * Dp) Control Strategy AND Compute Other Risk Component Costs The f i r s t premise identifies the candidate in order to evoke risk relevant slots from the preliminary level. The second premise affirms the state of nature by context. The third premise queries the user about the individual cost item estimates (Dp(i)), while context adds them up. In the conclusion, the As_Expected risk element w i l l be evaluated as the product of i t s likelihood by the sum of i t s cost Chapter 4. A KBES Framework for Methods Selection 148 items. Next, control i s passed to compute other states of nature risk components — i.e. better than expected and worse than expected. Next, we examine the SSP method analysis component. 6. Analyze (KB-7) The following control strategy rule examples are used to direct the search for changes to the construction method in order to achieve f e a s i b i l i t y . RULE (1) Change Pile Driving Strategy from "In_Pairs" to "InjSingies" Source IF PueDrivingStrategy Is "In_Pairs" Context, User AND TechnicalJ'easibility.State is False Context (blow count < = threshold) {Control Strategy} THEN Change PileDrivingStrategy to InSingles AND Do Performance Measures RULE (2) Increase Hammer Energy Source IF Pile_Driving_Strategy Is InSingles Context AND TechnicalFeasibility Is False Context {Control Strategy} THEN Increase the Hammer Delivered Energy AND Do Performance Measures Appendix B contains the logic and relevant dynamic formulas derivation for the technical f e a s i b i l i t y test under pile/hammer/soil scenarios. Chapter 4. A KBES Framework for Methods Selection 149 RULE (3) Change GWSS Alternative Source IF AND Driving_Strategy Is In_Singles Hammer/Pile Are Not_Compatible Context Context {Control Strategy} THEN AND Change SSPAlternative to SPLAlternative Do Design SPLAlternative 4.4 CMSA Risk Component Development and Evaluation Interviews and discussions with contractors' personnel regarding the decision making process dealing with methods selection have highlighted their concern, early on in the process, with risks. Particular emphasis is placed on the potential for large/catastrophic risks which often accompany underground work, work in water, etc. A method which is more l i k e l y to be subject to such risks tends to be shunned even i f there are significant cost/time benefits associated with i t . Contractors tend to seek a solution that actively, as opposed to passively, controls risk (e.g. H piles and lagging with struts rather than shotcrete). Based on the above, the use of a risk criterion to screen alternatives early on i s important. The risk assessment framework should be simple to use and not overly data intensive. For example, a complex set of states of nature may r e a l i s t i c a l l y portray the operating environment for a method, e.g. see figure 4.11. The effort and data required to specify each risk type and corresponding states simply are not available. So risk assessment for Cut-and-Cover tunnelling alternatives for Chapter 4. A KBES Framework for Methods Selection 150 this thesis, must be simplified to treat only those conditions that could lead to unacceptable risks. In this section, the context of GWSS is used to i l l u s t r a t e the approach. Two risk categories are considered: Figure 4.11 States of Nature for Methods Selection 1. Normal Risks: Normal risks deals with those site conditions (access, weather, ground, management, etc.) that influence productivity and other variables, thus creating uncertainty in time and cost estimates. 2. Large/Catastrophic Risks: These are treated e x p l i c i t l y in the decision making process through a simplified decision tree shown in figure 4.12. The basis of this decision tree Chapter 4. A KBES Framework for Methods Selection 150 this thesis, must be simplified to treat only those conditions that could lead to unacceptable risks. In this section, the context of GWSS is used to i l l u s t r a t e the approach. Two risk categories are considered: Figure 4.11 States of Nature for Methods Selection 1. Normal Risks: Normal risks deals with those site conditions (access, weather, ground, management, etc.) that influence productivity and other variables, thus creating uncertainty in time and cost estimates. 2. Large/Catastrophic Risks: These are treated e x p l i c i t l y in the decision making process through a simplified decision tree shown in figure 4.12. The basis of this decision tree Chapter 4. A KBES Framework for Methods Selection 151 i s as follows. First, since we are using risk to preserve alternatives, we examine alternatives indirectly. Second, three basic state of nature are treated: 1. conditions better than expected; 2. conditions as expected; and 3. conditions worse than expected. Conditions here refers to that condition most li k e l y to lead to unacceptably large or catastrophic risks. Then, for each state of nature treat three more conditions — no failure, minor failure, major failure; or no damage, element damage, system damage, where: 1. minor failure refers to partial damages for the wall structures and/or the surroundings; and 2. major failure refers to GWSS collapse, or retaining system collapse, and/or other major surrounding damages. The user i s required to assign probabilities to each branch in the decision tree. At the end of each path i s a vector of incremental costs (positive/negative). This vector of costs i s : Labor Equipment Materials Loss of Life Loss of Reputation Subsurface Subsidence Season Loss The user i s asked to estimate the costs associated with this vector as a fraction or percentage of total estimate, Chapter 4. A KBES Framework for Methods Selection 152 given the states of nature. This corresponds to the way contractors estimate. In terms of computation, the incremental costs are summed, and then discounted by the probabilities. If risks are deemed to be unacceptable, the user i s required to assign an i n f i n i t e cost to a cost category. This effectively eliminates the alternative. Then, Total cost of alternative = base cost + expected value of incremental costs For further development of the expert system, the challenge i s to identify, for each method alternative, the governing risk considerations (e.g. ground condition variable, flood potential, etc.) We also need to allow the f a c i l i t y for the user to describe and record the thought process that leads to the specification of the decision tree. Shoring Alternatives Event Chances Consquences Costs (Outcomes) Encountered Geological Conditions More Favorable Than Expected (PI) CD- Steel Sheel Pile Catastrophic Damagex (Not Acceptable) Risk Category Risk Cost Equipment Loss %ot Total Cost Labor Loss %of Total Cost Material Loss %of Total Cost Life Loss %ot Total Cost Subsurface Sub-sidence Loss %of Total Cost Season Loss %of Total Cost Other Losses %of Total Cost 154 5. C M S A Implementation 5.1 Introduction The primary objective of this chapter i s to explore the issues involved in the implementation of a prototype CMSA system. The f i r s t part covers an overview of NExpert Object, an expert system used to implement CMSA. The objective i s to familiarize the reader with the NExpert environment: knowledge constructs, syntax, operators, inference mechanism, and so forth. The second part covers selected details of the CMSA prototype development. The heuristic problem solution paradigm (figure 4 . 6 ) consists of Suggest, Design, Predict, and Analyze operators. The f i r s t operator, Suggest, maps the preliminary knowledge that suggests a preliminary feasible GWSS alternative for further detailing by the low level f e a s i b i l i t y part. The other three operators correspond to the low level f e a s i b i l i t y component: Design specifies the design element, construction resources, and construction strategy; Predict applies the hammer dynamic formula to predict i t s performance; and Analyze diagnoses the synthesized method to test i t s f e a s i b i l i t y . If the assembled method meets i t s goals, then i t accepts the GWSS alternative (SSP) . If i t does not meet i t s goal, then i t suggests recommendations for re-designing the method. Chapter 5. CMSA Implementation 155 The construction method attributes of the design elements, and construction resources are represented as frames. Rules are used to represent other specific knowledge and control strategy. These include engineering design knowledge for structural members (sheet piles), resource selection (pile drivers selection), construction strategy (pile driving pattern), and construction process model evaluation (dynamic formulas), analyzing the method synthesis (technical f e a s i b i l i t y ) , and ranking the GWSS alternatives. Examples of frames, rules, and databases written in NExpert Object w i l l be provided. Rules of thumb and algorithms as presented in the literature review in chapters 2,3, and 4 and in the appendices were extracted from texts, journals, and interviews. Knowledge acquisition for methods selection and analysis i s the domain of a number of experts, engineers, superintendents, foremen, and others. Each contributes to the method selection process at different times, and to varying degrees. Knowledge acquisition from site and office experts from one project was undertaken. Appendix C describes a site v i s i t and provides the results of two interviews. Such interviews contributed to improving and validating the problem solving knowledge base. Chapter 5. CMSA Implementation 156 5.2 NExpert Object Overview An Al toolkit i s ideal for problems that need a mix of specialized tools. It i s desirable that such a toolkit should offer a hybrid rule system, object oriented programming, and access to a general purpose language. Additionally, the toolkit should interface with conventional software such as databases, spreadsheets, graphics packages, and word processors. Neuron Data's NExpert Object exhibits several of these features. NExpert i s a powerful, hybrid, rule and object based expert system shell that speeds up the prototyping process for expert systems for non-programmers. It i s mainly a rule based system f i t t e d with object oriented features, such as f i r i n g a routine from a premise i f certain conditions have changed. The version used by the author suffered from a lack of good documentation and examples to explain and demonstrate a l l of NExpert's features. 5.2.1 Major NExpert Object Modules NExpert Object consists of the following modules: 1. NExpert Development Package, i s the core of the system. It contains pop-up windows for editing text, database, rules, objects, classes, etc; visual display for rules and objects networks; a reasoning kernel providing inheritance control, backward chaining, forward chaining, pattern Chapter 5. CMSA Implementation 157 matching c a l l s to external routines, and others; and trace f a c i l i t i e s (Transcript, Encyclopedia, Reports). 2. The Callable Interface, i s a library of C routines and function kernels used to embed NExpert within a conventional programming language. It consists of C functions that make up the NExpert Object Development and Runtime Environments. It can be used to establish communications between external application programs and NExpert applications by taking advantage of structures and functions that NExpert uses internally. There are at least three ways to use the callable interface. NExpert functions can be called directly by external code as exported C functions; external code can "trap" specific NExpert functions and declare their own replacement, or a standard message can be passed between other windows applications and NExpert using Microsoft's dynamic exchange protocol (a system for setting up standard messages). The callable interface includes C functions to i n i t i a l i z e , start, stop, and resume sessions; query or change knowledge structures; and change the l i s t of hypotheses or agendas. In order to i n s t a l l function handlers in NExpert, or embed NExpert within another application, there are additional software requirements such as the MS Windows Chapter 5. CMSA Implementation 158 Software Development Kit, version 2.03 or later, and a Microsoft C Compiler version 5.0 or later. 3. NExpert Object Runtime, i s a run time package that i s used to run an application without access to the knowledge base (a development package stripped of i t s debugging mechanisms). The developer can define, in detail, how an application w i l l run and design an interface that i s v i s i b l e to the client. An overview of the NExpert development framework i s shown in figure 5.1. 4. Hardware and Software Requirements — Minimum hardware and software requirements for NExpert Object under MS/Windows are: IBM PC or compatible with 64OK of conventional memory, plus 1 MByte of expanded memory on a 286 machine 1 MByte of expanded or extended memory on a 386 machine (2M i f you are using Windows 386) Enhanced Graphics Adapter (EGA) or VGA. Color EGA requires a video board with at least 64K of memory. - 1.44 M floppy disk drive hard disk with at least 3 MByte available Microsoft Windows Runtime 286 or 386, version 2.03 or later. Mouse (compatible with MS-Windows: Bus or Serial, Microsoft, LogiMouse, Mouse Systems, etc.), Parallel port. Chapter 5. CMSA Implementation 159 Recommended hardware and software in addition to the above: 2 additional MBytes of Extended memory. This allows you to store NExpert Object in RAM drive, 386-based machine (for development) External devices/computers multi-process real-time data Input Into NEXPERT event-driven architecture Multi-tasking Networking Ethernet, DecNet. Inter-process Communication Vax Calling Conventions Dynamic Data Exchagne DataBases DB III. Lotus 123, ORACLE, SOL, RDP, EXECL... data storage read/write N E X P E R T ! o Retrieve • Show dynamic access to knowledgebases Load/Unload graphics text Knowledge Bases explanations graphics text focus of attention conclusions reasoning trace active values reports what If Figure 5.1 NExpert Object Open Al Environment Framework [from Neuron Data 1989] Chapter 5. CMSA Implementation 160 5.2.2 NExpert Primitives - Building Blocks and Operations The basic building blocks of NExpert Object are described in this subsection. 1. Rules Rules are the preferred way to process objects in either forward or backward chaining. In NExpert, rules are expressed in the following form (see figure 5.2 for NExpert Rule Syntax) IP Conditionl AND Condition2 AND ConditionM THEN Conclusion (Hypothesis) AND Actionl AND Action2 AND ActionM The l e f t hand side (LHS) of a rule tests the value of a property for some object or class. Mixed qualifiers such as existential operator (test the condition: i s there any instance ...) and universal operator (test the condition: are a l l instances ...) can be used to lend expressive power to the LHS conditions. The LHS of the rule is composed of one or more antecedent (if) clauses which are called conditions. Operators and their purposes on this side include: Chapter 5. CMSA Implementation 161 Yes and No test Boolean Variables; [>/ </ =/ <>/ <=/ =>] Is and IsNot Name Reset Equal and NotEqual Show to test an expression or variable against a numerical value; test String Keywords; Type Data versus reserved Evaluate an expression and/or variable and assign i t to another variable; reset the value of an attribute to the original Unknown, as i f the value has not been asked; / compare an expression to a slot; used to show a text f i l e or a graph during running an application; Member and NotMember test whether a particular object belongs to a given l i s t of objects obtained by pattern matching or; verify that an object does not belong to a l i s t ; Execute Retrieve c a l l a subroutine written in a high level language such as C or FORTRAN; allows the system to read a f i l e of values stored on disk and/or to query a database. A right hand side (RHS) of a rule has a conclusion or hypothesis and optional actions. The hypothesis defines the rule's major topic and, as such, can be used to control backward and forward chaining. If two rules share the same RHS hypothesis, then both become candidates in backward chaining on that hypothesis. A LHS condition in one rule can also be the RHS hypothesis in another. Chapter 5. CMSA Implementation 162 If... conditions I t and do... actions Figure 5.2 NExpert Rule Construct [from Neuron Data 1989] The right-hand side (RHS) of the rule consists of a consequence called an "Hypothesis" in NExpert which can be True, False, or Notknown, and optional actions that occur when the hypothesis i s true. The RHS action can be used to control rule processing as an application runs. For example, consider the case in which two rules share a RHS hypothesis: Rl: LHS1 Then Hypothesis and Actionl R2: LHS2 Then Hypothesis and Action2 Suppose that i f Rl fire s , we want any subsequent backward chaining to be exhaustive; namely, we want the system to backward chain on every rule that has the goal hypothesis, regardless of whether or not the backward chaining f a i l s on a fired rule. If R2 fire s , however, we want to backward chain only u n t i l we succeed once; at that point, the ^ .then..., hypotnesii Chapter 5. CMSA Implementation 163 backward chaining halts. Actionl and Action2 would implement the different strategies with a Strategy action. NExpert allows similar dynamic, or runtime, control over inheritance. Such control adds to problem solving f l e x i b i l i t y . Operators on the RHS consist of: DO Let Reset assigns a value to one or a l i s t of variables. The value can be that of a single atom or of an expression; assigns a string value to any variable of the same type; sets the value of i t s argument to the "Unknown" state; Strategy allows the user to alter control mechanisms. It opens the Strategy Dialog window for setting the inference engine and inheritance mechanism's default parameters (or i n i t i a l parameters since the rules can change them). Control propagation involves constructs as contexts, backward exhaustive evaluation, RHS actions, semantic gates, inheritability, inheritance strategy, sources application, and "If Change" operator actions. CreateObject / DeleteObject CreateObject i s used to create objects and/or add them to class or classes. DeleteObject deletes a link between a l i s t of objects (class) and/or deletes an object from a Knowledge Base. These operators are part of NExpert's hybrid environment. Execute, Show same as in (LHS) Retrieve and/or Write allows the user to read and write data from f i l e s stored on disk. LoadKB makes i t possible to load a knowledge base incrementally, subsequent to the f i r i n g of a Chapter 5. CMSA Implementation 164 rule. Thus, a l l rules do not have to be loaded at the start session. UnloadKB unloads a knowledge base incrementally, subsequent to the f i r i n g of a rule or a meta-slot action. An NExpert rule i s conjunctive in nature in that a hypothesis i s true i f , and only i f , a l l the conditions are true. The truth table 5.1, shown below, provides a matrix for the state of the hypothesis given the added NExpert variable NotKnown, for the union "AND" and intersection "OR" operators. AND OR T NK F T NK F T T NK F T T T T NK NK NK F NK T NK NK P F F F T T NK F Key: T: True, NK: Notknown, F: False. Table 5.1 Truth Matrix for NExpert NotKnown i s rarely used. In some other expert system shells i t stands for the degree of uncertainty, a logical string which implies that the user does not know the truthfulness or precision of a premise, or has no information for a given data item. It adds an extra primitive for the state of the hypothesis truthfulness. NExpert can express disjunctive (OR) conditions with two rules pointing to the same hypothesis. When a l l conditions Chapter 5. CMSA Implementation 165 are true, the hypothesis i s true and the actions are enabled. As with a l l shells, the developer must design the NExpert rule base with inference in mind. Each rule must have at least one hypothesis that i s a condition in other rules, and some rules must include conditions that are hypotheses in other rules. 2. Classes and Objects NExpert's object-oriented component is derived from a generic a r t i f i c i a l intelligence tool known as a frame-based system. Frames can represent arbitrary entities, or concepts, and collections of entities. A frame can have an arbitrary number of slots (properties), each with a value. A slot may have a procedure or function as i t s value. Frames can occur in an inheritance network that propagates properties and values from parents to children, and vice versa, as well as multiple inheritance. An object or class can inherit properties and values from more than one source. There are two types of frames: classes, which are collections, and objects, which typically belong to one or more classes. In NExpert, slots are called "Properties". A class or an object can have many properties, and the inheritance network of objects and classes can be a r b i t r a r i l y complex. Figure 5.3 p a r t i a l l y shows the Chapter 5. CMSA Implementation 166 relationship between a NExpert class and object in which circle s represent classes, and triangles represent objects. Figure 5.3 The Class and Object Hierarchy [from Neuron Data 1989] Chapter 5. CMSA Implementation 167 3. Properties Every property of an object or class has a value, which defaults to Unknown. A property also has properties of i t s own called "Meta-Slot" that determine i t s behavior, inheritance, i n i t i a l i z a t i o n , and modifications. A meta slot can be used to implement a demon, which i s a piece of code that monitors a property and reacts appropriately to changes in i t s value. Demons are a type of procedural attachment that can take us outside NExpert into a conventional language such as C. Through the "If Change" meta-slot one can have NExpert invoke a function whenever a property value is updated. A function could be used to issue a warning, return a value, or even create a new NExpert object. 5.2.3 Viewing Knowledge Structure A knowledge base consists mainly of rules (LHS, RHS, and Hypothesis), frames (classes, objects, properties), text and database f i l e s , and procedures, which are accessed using "Notebooks" and "Editors". NExpert offers graphics tools, called Networks Inspector for viewing the knowledge base (rules, classes, and objects) and focusing on a piece of knowledge for analysis during development. If the viewed structure requires editing, one can zoom to i t directly on the Network and use a pop-up window to change i t . Chapter 5. CMSA Implementation 168 A full-screen window with a pull-down bar menu across the top allows the developer to navigate through the shell using the following: Edit, to access editing windows for knowledge elements such as rules; Expert, to load a knowledge base, start the inference process, or change the inference strategy; Inspector, to display the knowledge base graphically; and Report, to access information about the status of inference and knowledge structures. The Encyclopedia option retrieves l i s t s of existing rules, data, hypotheses, objects, classes, and properties, in the form of alphabetically indexed pages. The Windows option allows the developer to control the display of windows. 5.2.4 The Inference Process Unlike most other shells, NExpert can reason using a number of strategies and has several options for beginning the inference process. Each knowledge base has a default strategy, but the developer can change this and choose between forward (event-driven) processing, backward (goal-driven) processing, or a combination of the two. Additional inference pathways are identified by keeping a l i s t of agendas which are hypotheses of successfully fired rules. Figure 5.4 reflects that at any given time, NExpert i s focusing attention on a point that l i e s at the intersection of two orthogonal, but always intersecting, dimensions: Chapter 5. CMSA Implementation 169 representation and reasoning. When the system i s evaluating a rule, i t i s necessarily concerned with structures, objects, and classes that are part of the description of the world about which i t i s reasoning. Movement the representation plane Figure 5.4 Rules Perpendicular to Frames [from Neuron Data 1989] The user starts the inference process by selecting the MKnowcessM operator from the Expert menu, which displays the Session control window. NExpert creates a l i s t of i n i t i a l hypotheses and begins to process the rules and objects. At the same time, the system overlays several other windows on the screen to display information such as hypotheses Chapter 5. CMSA Implementation 170 currently under investigation (the hypothesis window), transcripts of any values the system has obtained or altered (the transcript window), the rules currently under investigation through backward chaining (the Rule window), and any concluded hypotheses (the Conclusions). The user can access these windows throughout the inference process. The user can interrupt the inference engine to "Volunteer" (input) information to the NExpert system regarding whether new data are relevant to the rules and hypothesis currently under investigation. The user can start a forward-chaining process, for example, by entering a l l information about a problem and then letting the system f i r e every possible rule to develop hypotheses. The user can suggest one or more hypotheses for the system to use to trigger the inference process. After studying the i n i t i a l or suggested hypotheses, the system then identifies other hypotheses worth examining and may propagate in a backward or forward mode of chaining. For instance, figure 5.5 shows Reset operator sets a hypothesis into "Unknown" and eventually resets a l l preceding hypotheses to "Unknown" in a backward mode of chaining. This operator was used to reset the state of technical f e a s i b i l i t y to "Unknown" when a combination of soil/hammer/pile f a i l s (see figure 5.25). Chapter 5. CMSA Implementation 171 Reset Figure 5.5 Reset Operator for Inference [from Neuron Data 1989] A developer can use NExpert's Strategy menu, for example, to f i r e only the rules concerned with confirmed hypotheses, examine only data that would reject an hypothesis, or examine a l l data regardless of whether the associated hypothesis i s under investigation. A rule can be forced to be evaluated and an object can confirm an hypothesis under investigation. With dynamic control over the inference process, one can combine a l l strategies to create highly complex reasoning pathways. After the system finishes i t s user interrogation, the results of the inference are displayed in the conclusions window, and a trace of the system operations can be recorded on the Transcript window for debugging. Each hypothesis Chapter 5. CMSA Implementation 172 investigated i s either rejected, confirmed, or already known to be true. Rules can be grouped together and linked into related sets called knowledge islands, which can store knowledge about a large, complex domain in classes. These are templates that provide a general description of a category of items in the domain. Specific instances of each item are called objects. NExpert uses rules to reason about objects and classes. Thus, the two representation methods can be integrated and customized for the inference process for specific application needs. Figure 5.6 depicts how rules manipulate and interact with other NExpert constructs — objects, meta-slots, methods, etc. Chapter 5. CMSA Implementation forward / backward Inheritance multiple inheritance dynamic variable objects methods if change then do user defined inheritance and inference strategies procedural attach em errta selection operators embedded lists Figure 5.6 NExpert Inference Framework [from Neuron Data 1989] Chapter 5. CMSA Implementation 174 53 CMSA Implementation 5.3.1 CMSA Overview The prototype system, CMSA, i s an implementation of the conceptual system described in chapter 4. The CMSA control strategy and i t s knowledge bases are described in section 5.3.2. The latter includes procedural knowledge: pile design and pile driving production; analytical: risk analysis and evaluation c r i t e r i a ; heuristic knowledge: pile driver selection and sizing; and factual knowledge: sheet pi l e selection. Knowledge representation employs frames for method attributes (section 5.3.3); rules for the control strategy and s o i l context description; data bases used for structural members and construction resources; and numerical routines for design, moment, and pi l e driving progress prediction procedures. The implementation of technical f e a s i b i l i t y checks is described in section 5.3.4. Chaining and reasoning procedures used in the prototype system are described in section 5.3.5. The CMSA prototype was implemented on an IBM PS2/386 personal computer equipped with 4 Mbytes of memory using the NExpert Object expert system shell. The CMSA inference engine, or control strategy, acts as a methods selection shell, or skeleton, that i s applicable to other methods Chapter 5. CMSA Implementation 175 domains besides the GWSS method selection for Cut-and-Cover Tunnelling. The user interface and explanation f a c i l i t i e s for the prototype are not provided. If desired, the Screen Builder Module for NExpert could be used to enhance the interface. Text f i l e s are used extensively in CMSA to explain to the user what data i s required, why i t i s required, and what is expected next. The prototype implementation deals with the detailed f e a s i b i l i t y part of section 4.3.4. The control strategy was meant to be implemented as an independent knowledge base. However, in NExpert Object, this required loading and f i r i n g other knowledge bases which in turn overloaded the memory and caused a crash of NExpert Object. To work around this problem, the rules from each KB were grouped into one knowledge base resulting in the control strategy rules being interwoven with specific knowledge rules. The risk component, due to i t s size, i s treated separately and results are then input to the previous knowledge base. Chapter 5. CMSA Implementation 176 5.3.2 Solution Paradigm and Knowledge Base In this section, a solution paradigm of Suggest, Design, Predict, and Analyze, as described by Clancey (1985), i s superimposed on the augmented knowledge base to aid in the exposition. Figure 5.7 i s a restatement of figure 4.6. The solid lines correspond to the control structure implemented. The dotted lines show suggested future developments. The knowledge bases are represented by rectangles, while the enclosed dotted shapes correspond to the Suggest, Design, Predict, and Analyze operators. Figure 5.8 depicts the abstract CMSA system operations cycle which i s elaborated upon later. Chapter 5. CMSA Implementation 177 Figure 5.7 Knowledge Base Organisation and Control Strategy Chapter 5. CMSA Implementation 178 SUGGEST DESIGN PREDICT ANALYZE Figure 5.8 Implementation Solution Paradigm The foregoing operators and their corresponding knowledge bases are described below. 1. Suggest (KB-1) Because no prototype was implemented for the preliminary f e a s i b i l i t y aspect of the system, the Suggest operator simply consists of a l i s t of alternatives from which the user can select one or more candidates for the detailed f e a s i b i l i t y analysis. The Suggest operator triggers the session f i r s t by f i r i n g the design component knowledge base. Chapter 5. CMSA Implementation 179 2. Design (KB-3, KB-4, KB-5) 2.1 Specify Design Elements of GWSS (KB-3) This component consists of two procedural knowledge components. The f i r s t one, KB-3-1, calculates the earth lateral pressure, given a s o i l profile, and the resulting moments exerted on the GWSS wall. Default retaining system (back-ties or struts) spacings are required in order to perform these computations. Rules of thumb, as found in the literature (Tomlinson 1975), suggests a 15 f t vertical spacing and 12 f t horizontal spacing between the retaining system members. After the s o i l profile i s described by the user, the CMSA retrieves relevant s o i l layer properties from a s o i l data base. Appendix A contains the formulas u t i l i z e d for calculating pressures and moments envelopes based on the approach suggested by Peck and Terzaghi (Winterkorn and Fang 1975). The second procedural knowledge component, KB-3-2, specifies a steel sheet pil e (KB-3-2-1) and soldier piles (KB-3-2-2) from their respective data bases (SSP.nxp and SPL.nxp) using the section modulus for the sheet piles and soldier piles derived from KB-3-1 moments. Currently, a strength criterion is checked. Deflection and buckling considerations are not considered. Chapter 5. CMSA Implementation 180 In KB-3-2, structural member selection i s based on aggregating sections prudently into groups defined by their upper and lower bounds. This i s a commonly used design approach in developing expert systems for engineering design (Adeli and Paek 1986). The design of retaining system members (struts, wales, lagging) for the GWSS i s also treated in KB-3-2. Strut and wale design knowledge i s based on moment c r i t e r i a . Lagging design i s based on Canadian Standards Association c r i t e r i a (1976). Conversely, retaining system structural members are considered to be fixed variables for a l l GWSS. In the future, KB-3-3, could be ut i l i z e d to investigate the trade-off between the retaining system cost and the cost of the GWSS piles. Once KB-3 i s successful in specifying sheet and soldier p i l e sections, control moves to selection of a hammer (construction resource). 2.2 Select Construction Resources (KB-4) Resources employed for sheet pi l i n g include: a pi l e driver, a crane, a pi l e frame, and pile driving crew. Crane size, in tons, i s proportional to the weight of the hammer and i t s attachments. Chapter 5. CMSA Implementation 181 For labelling purposes, note that KB-4-1 and KB-4-2 of figure 4.6 are combined into one component of KB-4 for specifying impact hammers. For knowledge base organization purposes (see table 5.2), KB-4-1, KB-4-2, and KB-4-3 refer to impact hammers, vibratory p i l e drivers, and cranes knowledge base components respectively. 2.2.1 Impact Hammers Specification (KB-4-1) This component of the knowledge base contains factual and experiential knowledge in order to specify a hammer. Several types (Single and Double Acting Air Hammers) and sizes of hammers are stored in separate data bases (e.g. DAAH.nxp for Double Acting Air Hammer). The following hammer specification heuristics are employed. 1. Determine the class or type of hammer based on the s o i l profile (see section 4.3.2). 2. Use the experiential rule of proportioning the hammer size equal to or less than the steel p i l e cross section area multiplied by a factor of 3 000. This i s to provide an upper bound on the hammer size in order not to damage the pile , and corresponds to a commonly used rule of thumb in the industry. If the class of hammers determined in step 1 does not satisfy the condition in step 2, then another class should be selected and tested. Chapter 5. CMSA Implementation 182 2.2.2 Vibratory Hammers Specification KB (KB-4-2) The vibratory hammer differs substantially from the impact hammer in i t s operating principles, therefore, the selection criterion i s different. A vibratory p i l e driver can be specified in terms of i t s dynamic force and amplitude. Under favorable conditions, a vibratory hammer i s 4 to 8 times faster than an impact hammer; however, i t i s more expensive to rent and operate. In the prototype system, emphasis has been placed on sizing the impact hammer, since a predictive model exists in the form of a dynamic formula for a subclass of i t s hammers. This model allows prediction of production, for a given soil/hammer/pile scenario, and indicates whether or not a pil e can be driven to i t s refusal. A similar predictive model for the vibratory hammer does not exist. Crane attributes are dependant upon selected hammer size, p i l e length, etc (see Appendix C). Currently, i t i s treated as a fixed variable, although experiential knowledge pertaining to crane selection i s included in KB-4-3. 2.3 Construction Strategy Selection (KB-5) For driving of sheet piles, construction strategy refers to the pattern of pi l e driving. For the prototype, this is simplified into driving sheets either as singles or in doubles. Chapter 5. CMSA Implementation 183 The selection of a construction strategy involves trade-offs among several factors: s o i l s t r a t i f i c a t i o n and conditions; p i l e driver size; p i l e capacity and size; pi l e segment length; pil e driving fixed time (positioning, splicing); and variable driving time. Soil s t r a t i f i c a t i o n and conditions are the dominant factors. If the s o i l s t r a t i f i c a t i o n and conditions state can be characterized as soft, CMSA attempts to drive piles in doubles. If the doubles configuration can not be driven to refusal, then CMSA retracts this construction strategy and explores driving in singles. To predict the f e a s i b i l i t y of whether each strategy, and other method attributes, are most suitably synthesized, a test i s required for the foregoing stated conditions. This is elaborated on in the technical f e a s i b i l i t y section. 3. Predict (KB-6) This operator basically decides the appropriate construction process model to be used. For CMSA, the dynamic formula derived in Appendix B i s applied only after a sheet pile section and hammer have already been selected. For a given scenario of soil/hammer/pile, the rate of pile penetration decreases with pile depth. A routine, named "Drive.c", described in section 4.3.4, i s used to predict the incremental velocity of pile driving and to check i f the refusal depth i s achieved. If the latter i s achieved, then Chapter 5. CMSA Implementation 184 this routine returns a "True" value for the technical f e a s i b i l i t y state and the total penetration time for a single p i l e . From the latter, CMSA computes total p i l e driving production, including fixed time, and thus production cost. Total cost computation including quantity take-off for the design elements, p i l e driver and crane, and labor costs i s then done. Risk assessment (KB-2) involves application of a rule based analytical risk assessment based on the approach described in section 4.4. This component has been programmed as a separate routine since i t requires substantial user input (over 3 0 data input parameters for the user to volunteer). This component returns the expected risk cost for each alterative which i s then incorporated in the total cost computation for the alternative. 4. Analyze (KB-7) The current method being evaluated by CMSA i s then analyzed to see i f i t satisfies the target project time/costs. If i t does, then i t i s accepted. Otherwise, recommendations to alter method attributes to achieve f e a s i b i l i t y are suggested by the system. CMSA may carry out some of those recommendations, by backtracking and changing the state value of a method attribute to permit re-evaluation (re-Chapter 5. CMSA Implementation 185 iterate) by either the Suggest or Design processors, followed by the Predict processor. 5. Control Strategy (KB-8) The control strategy i s the skeleton of the system, and i s responsible for managing the problem solution strategy. In the CMSA, the control strategy i s intertwined with other knowledge bases in one overall knowledge base. In designing the CMSA implementation design strategy separate KBs were developed so that they could be run independently of each other. The reasons for this strategy include: 1. The user may choose to conduct a risk appraisal for an alternative without going through the interrogation. For instance, i f one wants to assess the risk status of a Secant Pile alternative, the risk f e a s i b i l i t y component (KB-2) can be independently fired. 2. Some KBs are incomplete in their development. For instance, the vibratory hammer KB (KB-4-2) incorporates experiential knowledge for hammer selection, but no process for predicting a production rate because no algorithm exists. 3. Integration of independent knowledge bases, including the control strategy meta-rules, into one knowledge base proved to be cumbersome. With NExpert Object, such integration requires the control strategy to execute operators "Load KB" and "Unload KB" to f i r e other KB's. Invariably, this creates a system crash because memory is overloaded. So the strategy adopted was: develop independent KBs for testing separately; then merge into one overall KB for purposes of developing a working prototype CMSA. Chapter 5. CMSA Implementation 186 4." From experience in developing the prototype system, i t is easier to develop and edit each knowledge base independently, which permits an incremental approach to development of a system. Table 5.2 shows summary st a t i s t i c s for the CMSA knowledge bases shown in figure 5.7. The last column indicates the number of rules extracted from the individual knowledge bases for purpose of assembling the prototype. KB Size 10 Knowledge Base Number and Name Number of Rules Number of Rules for each KB in CMSA (KB-1) Preliminary Feasibility 10 2 (KB-2) Risk Component Feasibility 18 4 Wall Design Elements (KB-3-1) Earth Pressures and Moments 5 3 Structural Members (KB-3-2-1) Steel Sheet Piles 18 18 (KB-3-2-2) Soldier Piles 18 Hammer/Crane Selection (KB-4-1) Impact Hammer Selection 16 16 (KB-4-2) Vibratory Hammer Select 15 (KB-4-3) Crane Selection 6 (KB-5) Construction Strategy Selection 15 15 (KB-6, KB-7) Predict and Analyze 10 10 (KB-8) Control and others 10 Total Rules 78 Table 5.2 Knowledge Base Statistics -•-"Number of rules i s an indirect measure of knowledge base size. The number of rules i s a function of system objectives, control strategy required, ES shell representation schemes, knowledge organization, and the way rules are collapsed on each other. The rule construct in NExpert Object is a high level operator that in comparison requires several rules to emulate in other purely rule based system. Chapter 5. CMSA Implementation 187 5.3.3 Knowledge Representation The prime representation constructs in CMSA are frames for method attributes and rules for s o i l profile and conditions, performance c r i t e r i a evaluation, (quantity take-off and cost estimate) and the control strategy. Analytical routines are used for p i l e driving progress rate prediction, and pressures and moment computation. Data bases are used to store the properties of structural members (piles, wales, struts), hammers, and s o i l types. When the CMSA retrieves data from a data base, i t assigns a corresponding class and object hierarchy. An object corresponds to a record, and object attributes correspond to a record f i e l d in NExpert. 1. Soil Profile As shown in figure 4.7, a maximum of two s o i l layers and four s o i l types: sand types (Loose_Sand, Dense_Sand) and/or clay (soft clay, s t i f f clay) were treated. Soil types are represented as database records. Each record corresponds to a s o i l type object, and the record fields correspond to object attributes. In section 4.3.2, two forms of entering s o i l information were identified. The f i r s t i s used when the user i s asked to identify one s o i l type per layer: e.g. dense sand for the top layer and s t i f f clay for the bottom layer. Given this input, the system then retrieves the Dense_Sand record Chapter 5. CMSA Implementation 188 from the s o i l database to obtain the saturated unit weight, angle of internal f r i c t i o n , and other properties. For the second mode of input, i f the user has a standard penetration test log (SPT) for the s o i l profile, then CMSA identifies the s o i l layers against SPT values (e.g. i f SPT is high and s o i l i s sand, then this i s a Dense_Sand). The f i r s t s o i l input format i s used for calculating pressures and moments in order to size the pi l i n g . The second s o i l input format is used for predicting the pile driving rate using the Drive.c routine. 2. Method Attributes Frames, called objects in NExpert Object, were used to represent method attributes, with emphasis being placed on design elements and construction resources. Construction strategy and construction model processes have been treated as predicates of object-attributes-values in a rule format. In future work, i t is suggested that they be treated using a frame representation. 2.1 Design Element A design element in CMSA corresponds to a member of the ground wall support system: i.e. steel sheet pil e , soldier p i l e , wale, strut, and lagging. Figure 5.9 displays an overall hierarchy of design element classes and objects. Chapter 5. CMSA Implementation 189 A generic class of design element, a parent, has universal slots or attributes that can be inherited by i t s children (descendants). Attribute values are either inherited, defaulted, or evaluated via a procedure. Descendants of this global class are a design element subclass of either the GWSS subclass and/or the retaining system subclass. Further branching from the GWSS includes subclasses for the GWSS of the type of structural member types. To i l l u s t r a t e the foregoing, consider the steel piling alternative. Sheet piles can be either steel, wood or concrete, each of which have a set of material properties, such as modulus of elasticity, yield strength, bending stress, unit costs, and so forth. For example, steel sheet p i l e properties include moment of inertia, allowable bending stress, cross section area, type of alloy material and designation. In the CMSA, some of these properties are required in the approximate design routines to establish construction resource compatibility, and for quantity take-off purposes. Chapter 5. CMSA Implementation 190 Design Element Class Slot Type of Element: Structural Members: Members Shape: Retaining System Spacing: Material Type: Material Quantity Take-off: Material Unit Cost: Wall, Retaining System,.. Column, Beam, Lag., Roof,.. I, Z, H,.. sections for SSP 15ftHor., 12 ft vert. Steel, Timber, Concrete GWSS H-Piles Subclass (SP, Beams, Unlt_Welght SectionJModulua Moment_ofJnterla O Cross_Seetion_Area SelectedSSP:YES Surface Area O Class A Object • Slot Figure 5.9 Design Element Class Hierarchy Chapter 5. CMSA Implementation 191 A subclass of steel sheet pil e consists of several steel p i l e objects (SSP_1, SSP_2, SSP_n) whose attribute values include designation, unit weight, section modulus, cross section area, and surface area. Unit weight i s used to estimate the total cost (unit cost in $/ton); section modulus i s used to satisfy a strength criterion; cross section area i s used in heuristic rules to assess the damage potential from driving; and surface area is used in the skin f r i c t i o n resistance calculation to estimate p i l e driving speed. Figure 5.10 shows a steel sheet pil e frame format. Class : Steel Sheet Piles Object : SSP3 SubObjects : None Slots Designation UnitWeight SectionModulus CrossSectionArea SurfaceArea Driving Width ASTM A36 "PZ36" "57" lb/ft "46.77" in~4 "16.77" in^ "5.52" in~2/in "18" in Figure 5.10 Design Element Instance Frame Values in key fields for steel sheet p i l e records, stored in the sheet p i l i n g data base, are retrieved by the CMSA and inserted into the appropriate slots of the design element frame. This representation scheme is equally applicable to other design elements such as H piles, sheets, wales and lagging. A partial design element data base i s presented in Appendix D. Chapter 5. CMSA Implementation 192 An example of the steel sheet pile class representation implementation using NExpert i s shown in figure 5.11 for the selected SSP. (@CLASS= SelectedSSP (©PROPERTIES= CrossSection Area designation driving_width SectionModulus SurfaceArea Weight_per_Foot ) ) Figure 5.11 Steel Sheet Pile Class in NExpert A l l sheet piles in the database are f i r s t retrieved and loaded into the Selected_SSP class using the rule shown in figure 5.12. This i s a factual rule, based on judicious section modulus range partitions, which results in an SSP from the l i s t of Selected_SSP being chosen as a Matched_SSP. CMSA, thereafter, fires such a rule to pick the sheet pile that satisfies the design strength c r i t e r i a based on the section modulus, which i s then saved in the Matched_SSP Class for later treatment. Steel sheet pile sections are stored in groups, with each grouping having a minimum and a maximum section modulus. A preliminary sizing is used to f i r e a rule so that NExpert retrieves members of the group. Rules are then fired in order to select a specific member of the group that satisfies the strength criterion. Chapter 5. CMSA Implementation 193 The NExpert text rule syntax for selecting a steel sheet pi l e i s shown in figure 5.12 which i s fired after steel sheet piles are retrieved and stored in the Selected_SSP. (@RULE= Select_SSP_of_PSA_32 ©COMMENTS = "Select SSP designation of PSA_32, SectionModulus is in in/v3";@WHY="This is a factual rule used to size a sheet pile where its properties in turn contributes to hammer sizing and pile penetration rate ."; (@LHS = (>= (SectionModulus) (1.9)) (< (Section_Modulus) (2.4)) (Is (< |Selected_SSP| > .designation) ("PSA32")) ) (@HYPO= SelectSSP) (@RHS = (Do ( < | Selected_SSP | > .designation) (Selected_Steel_Pile)) (CreateObject ( < | Selected_SSP | > ) (| Matched_SSP |)) (Do (SelectPileDriver) (SelectPileDriver)) ~ ) Figure 5.12 Steel Sheet Pile Selection Rule This rule i s interpreted clause by clause as follows. 1. Left hand Side The f i r s t and second clause say that i f the required section modulus (which i s calculated by a prior rule) for a sheet pi l e i s between 1.9 i n 3 and 2.4 i n 3 , and i f the third clause says there i s a sheet pile object in the Selected_SSP that satisfies those section modulus limits, then "PSA_32" of A36 ASTM is selected. The <|Selected_SSP|> notation implies a pattern matching operator for sheet piles which are treated as objects in a l i s t in order to match whatever property i s specified in the LHS. Chapter 5. CMSA Implementation 194 2. Hypothesis The hypothesis ("Hypo") name called Select_SSP i s invoked by "backward chaining" via the control strategy. If the Left Hand Side conditions were satisfied, then this hypothesis is evaluated as TRUE: otherwise i t i s FALSE or NOTKNOWN i f one of the conditions was false or not known respectively. 3. Right Hand side The "Do" operator assigns the designation of a steel sheet pi l e to the variable "Selected_Steel_Pile", and then uses the "CreateObject" operator to link the selected sheet pile (Selected_SSP) to a new class or l i s t of "Matched_SSP", to separate i t from the rest of the sheet piles for subsequent use in hammer selection and technical f e a s i b i l i t y . The Matched_SSP object inherits i t s attributes and values from the Selected_SSP class (pattern matched sheet p i l e ) . The last clause of the "Do" operator transfers control in order to invoke the "Select_Pile_Driver" hypothesis using backward chaining. This hypothesis i s then used to select a hammer that satisfies s o i l and pile conditions including the Matched_SSP properties which are treated as constraints. NExpert text database Format (*.nxp), f l a t database, for the steel sheet piles i s shown in figure 5.13. Chapter 5. CMSA Implementation 195 \SSP_l.Designation\ = "PZ38" \SSP-l.Weight_per_foot\="57.00" \SSP_l.Cross section_area\="16.77" \SSP l.Driving_width\ = "18" \SSP_l.Surface_area\="5.52" \SSP_l.Section_modulus\="46.8" \SSP_2.Designation\="PZ32" \SSP-2.Weight_per foot\="56.00" \SSP_2.Cross section_area\="16.47" \SSP 2.Driviifg_width\="21" \SSP_2.Surface_area\="5.52" \SSP_2.Section_modulus\="383" Figure 5.13 Steel Sheet Files Database (SSP.NXP) 2.2 Construction Resource Class Hierarchy Figure 5.14 shows a construction resource hierarchy which divides Cut-and-Cover tunnelling capital intensive resources into classes and objects. A resource class has universal slots for activity and task type and identification, unit cost, productivity, etc. The subclasses of bulldozers, cranes, and hammers have further slots to characterize them in terms of their functionality, size, operational properties, etc. For instance, for a GWSS of steel sheet piles, bulldozers are used for "clearing" and excavation a c t i v i t i e s ; and cranes are used for handling and hoisting materials, muck removal, and carrying a hammer for the pile driving activity. Chapter 5. CMSA Implementation 196 Bulldozer Subclass Construction Resource Class Slots Activity: Excavation, Pile Driving,.. Task: Hoisting, Pile Driving, Excavation,.. Resource Type: L a D O r Intensive, Capital Intensive. Hydraulic Hammer Subclass Vibratory Subclass Matched_Hammer Subclass Drop Hammer Vlbro Slots Hammer_Mode! Theortlcal_Energy -| | Stroke_per_MIn Length_of_Stroke Ram_Welght Manufacturer Selected SAAH: YES ( ^ ) Class, Subclass A Ob|ect • Slot Figure 5.14 construction Resource class Hierarchy Chapter 5. CMSA Implementation 197 The hammer subclass was divided into impact hammers, vibratory hammers and hydraulic hammers. Impact hammers were further classif i e d in accordance with their operating mode as single acting air hammer, double acting a i r hammer, diesel hammer, differential acting hammer and drop hammer. Each hammer type i s represented by a frame, an example of which i s shown in figure 5.15. Class Single Acting Air Hammer Objects SAAH2 SubObjects None Slots Hammer Model "S 20" Ram Weight "20,000", lb Strokes Per Minute "60" Length of Stroke "36", in Theoretical Energy "60,000", lb-ft HammerManufacturer Vulcan" Efficiency "87%" Figure 5.15 Impact Hammer Element Suppose that the s o i l context variables and resource av a i l a b i l i t y suggest use of a single acting a i r hammer (SAAH). Then, a database for the SAAH subclass would be retrieved and required record attributes would be mapped into the corresponding object and properties in NExpert. The construction knowledge for hammer selection and performance was encoded in a rule, which involves retrieving s o i l s t r a t i f i c a t i o n information and pile properties from the Chapter 5. CMSA Implementation 198 pi l e frame via a message which sends the selected p i l e cross section to a heuristic rule which limits the maximum size of the hammer. In addition to developing a data base for single acting ai r hammers, data bases were created for double acting a i r hammers as well as vibratory ones. This permits one to explore the efficiency of the prototype system to select the most appropriate resource as a function of s o i l context. An example frame for a vibratory hammer i s shown in figure 5.16. Knowledge for selecting and sizing a vibratory hammer is described in Appendix D. The prediction of production rates requires heuristic knowledge, and i s done by mapping vibratory hammer type and size versus s o i l s t r a t i f i c a t i o n . Class Objects SubObjects Slots Vibratory Model DynamicForce Horse_Power Frequency Amplitude MaximumPull SuspendedWeight ShippingWeight VibratoryManuf. Efficiency Vibratory Vibrol None "1412" "20,000", lb "650", HP "400", Vibration per min "1.5", in "80", Tons "1020", lb "20.5", lb "ICE" "93%" Figure 5.16 vibratory Pile Driver Element Chapter 5. CMSA Implementation 199 The data bases developed for the double acting a i r hammer types and vibratory hammer types are shown in figures 5.17 and 5.18 respectively. \Hammer_01.Hammer_Model\="2" \Hammer_01.Ram_Weight\="3000" \Hammer_01.Strokes_per_Min\ = "70" \Hammer_01.Length_of_Stroke\="29" \Hammer_01.Thero_Energy\ = "7260" \Hammer_02.Hammer_Model\="1" \Hammer_02.Ram_Weight\="5000" \Hammer_02.Strokes_per_Min\ = "60" \Hammer_02.Length_of_Stroke\ = "36" \Hammer_02.Theor_Energy\="15000" Figure 5.17 Double Acting Hammer Database (DAAH.NXP) \Vibratory_l.Dynamic_Force\ = "204" \Vibratory_l.Model \ = "1412" \Vibratory_l.Manufacturer\ = "ICE" \Vibratory_l.Frequency\ = "1200" \Vibratory_lAmplitude\ = "1" \Vibratory_l.Horse_Power \ = "650" \Vibratory_l.Max_Pull_Extract\ = "80" \Vibratory_l.Pile_Clamp_Force \ = "250" \Vibratory_l.Suspended_Weight\ = "10.20" \Vibratory_l.Shipping_Weight\ = "20.5" \Vibratory_l.Dynamic_Force\ = "204" \Vibratory_l.Model \ = "1412" \Vibratory_l.Manufacturer\ = "ICE" \Vibratory_l.Frequency\ = "400" \Vibratory_lAmplitude\ = "1.50" \Vibratory_l.Horse_Power \ = "650" \Vibratory_l.Max_Pull_Extract\ = "80" \Vibratory_l.Pile_Clamp_Force \ = "250" \Vibratory_l.Suspended_Weight\ = "10.20" \Vibratory_l.Shipping_Weight\ = "20.5" Figure 5.18 Vibratory Hammer Database (VIBRO.NXP) Chapter 5. CMSA Implementation 200 The NExpert text format for the selected impact hammer class i s shown in figure 5.19. (@CLASS= Selected Hammer (©PROPERTIES = HammerModel Length_of_Stroke Ram_Weight Strokes_per_Min Theor Energy ) ) Figure 5.19 Impact Hammer Class in NExpert Shown in figure 5.20 i s an impact hammer selection rule. This rule i s fired i f previous rules have indicated that a single acting a i r hammer may be suitable. The hammer selected ( i f one i s feasible) i s linked to Matched_Hammer Class for later use. The MatchedJHammer inherits the same attributes and values of the "pattern" <|Selected_Hammer|>. Chapter 5. CMSA Implementation 201 (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_LS_ON_STCL ©COMMENTS = "Select a hammer type based on soil profile and conditions (Using Hunt 1979 Table)";@WHY="Inference category for this rule is set to 1 since double acting air hammer (DAAH-Hammer) overrides the vibratory selection under this rule condition (assumption)"; (@LHS = (Is (Soil_Profile_Scenario) ("Loose_Sand_ON_Stiff_Clay")) (Retrieve ("daah.nxp") (@TYPE = NXP;@FILL=ADD;@CREATE= | SelectedHammer | ;\ )) (< = (< | SelectedHammer | > .Theor_Energy (|Matched_SSP|.Cross_Section_Area * 3000) (Name (MAX(< |Selected_Hammer| > .Theor_Energy)) (Max_Energy)) (= (< | Selected_Hammer | > .Theor_Energy-Max Energy) (0)) ) (@HYPO= SelectPileDriver) (@RHS = (Let (Hammer_Type) ("Double_Acting_Air_Hammer")) (CreateObject ( < | Selected_Hammer | > ) (| Matched_Hammer |)) (DeleteObject ( < | Selected_Hammer | >) (| Selected_Hammer |)) (Do (< | Matched_Hammer | > .Hammer_Model) (Hammer_Model)) (Do (SPProduction) (SPProduction)) ) ) Figure 5.20 Hammer Selection Rule in NExpert The rule in figure 5.20 can be interpreted clause by clause as follows: 1. Left Hand Side 1.1 The f i r s t condition checks the s o i l profile scenario that dictates the type and/or class of hammers that should be used (Impact versus . Vibratory, Double_Acting_Air_Hammer (DAAH), Diesel_Hammer, etc. Suppose, based on experience, that a s o i l profile of loose sand on s t i f f clay i s best handled using an impact hammer, and within this subclass, a double acting air hammer is the preferred choice. Chapter 5. CMSA Implementation 202 1.2 If the f i r s t condition i s successfully satisfied (returns True), the second condition assumes that the DAAH subclass best matches the s o i l profile. Therefore the database of DAAH w i l l be retrieved and attached to Selected_Hammer Class. 1.3 The third condition reduces the l i s t of Selected_Hammers by retaining only those hammers which have a theoretical energy (lb-ft) equal to or less than 3,000 times the cross sectional area of the Matched_SSP. This rule of thumb i s used in the f i e l d by p i l e driving contractors to prevent steel sheet pil e damage caused by over-sized p i l e driving equipment. The driving strategy state i s assumed here to be 11 In_Singles". However, i f i t is "In_Pairs", then the upper bound i s doubled. There i s a tradeoff between the driving strategy and hammer size. Furthermore, the hammer weight may dictate the weight of the pile segment (length and weight), especially for the case of a vibratory p i l e driver. 1.4 One hammer selection criterion i s to pick the most productive, or largest, hammer in terms of i t s theoretical energy (cost i s not considered at this-level) assuming those hammers have the same hammer efficiency. Different models within same hammer subclass have different hammer efficiencies for Chapter 5. CMSA Implementation 203 pile/soil/hammer scenarios. Therefore, condition four picks the largest hammer from the Selected_Hammer l i s t using NExpert operator "MAX". 2. The hypothesis named Select_Pile_Driver i s triggered through "Forward chaining" via the control strategy. If the Left Hand Side conditions were satisfied, then this hypothesis i s evaluated as "True"; otherwise i t i s "False" or "Notknown" when one of the conditions was Notknown. 3. Right Hand Side The "Let" operator assigns the hammer subclass type as a string, Double_Acting_Air_Hammer, to the Hammer_Type variable. The "CreateObject" operator links the hammer which satisfied the previous conditions to the Matched_Hammer class. The "DeleteObject" operator deletes the Matched_Hammer from Selected_Hammer i f this Matched_Hammer failed to produce the required production rate, as inferred from the technical f e a s i b i l i t y diagnosis. The "Do" operator then assigns the hammer model designation to a variable of Hammer_Model for subsequent treatment. The last "Do" operator transfers control to the SP_Production hypothesis where the Chapter 5. CMSA Implementation 204 pile/soil/hammer combination i s examined for technical f e a s i b i l i t y . 3. Construction Strategy Class Hierarchy For the Cut-and-Coyer tunnelling problem, the construction strategy class can be described in terms of a hierarchy (see figure 5.21). At the higher levels, overall construction approaches, such as top-down or bottom-up (e.g. the Milano method), are treated. These high level strategies greatly influence lower level strategies, a c t i v i t i e s and their sequencing. For instance, for a top-down strategy using a steel sheet pi l e GWSS, strategies at the pil e driving activity level are drive "In_Singles 1 1 or "In_Pairs". CMSA deals only with the lower level strategy, whereas the higher level strategies are assumed to be fixed. 4. Construction Process Model Class The construction process model (CPM) draws on selected design and resource frames in a process aimed at satisfying certain constraints and quantifying construction method performance measures. Chapter 5. CMSA Implementation 205 GWSS Class Figure 5.21 Construction Strategy Class Hierarchy Figure 5.22 shows, at the top level, a CPM element with representative slots. Class subobjects (model building elements) include crews, equipment, layout, etc. One, or a construction of more than one, of these entities emulates a systematic representation of operations, which in turn yields quantitative/qualitative performance measures. For CMSA, i t i s assumed that qualitative variables are dealt with prior to using the process model to determine quantifiable performance. Thus, a subset of resource attributes need to be accessible to the process modelling phase. For instance, at the pil e driving activity level, Chapter 5. CMSA Implementation 206 three object instances, from two resource classes (hammers and piles) , and the s o i l profile, have to be bound in the process model. SSP Subclass Soil Profile Subclass' SAAH Hammers Soll.Pro (Proflle-1 Pilejnstance (SSP-3) (Pile Driving Process Subob|ects) Construction Process Model Class Slots Subobjects: Crews, Equipments, Layout,. Model Criteria: Analytical, Approximate., Heuristics,.. Activities Involved: Pile Driving, Excavation,.. Model Performance Attributes: Time, Cost,.. Safety, Pile Driving Process Model Slots • Pile Driving Strategy: ln_Pairs, ln_Slngles Model Criteria: Dynamic Formula, or WEAP. Technical Feasibility. True, False Q Class, Subclass / \ Object, Subobject • Slot Figure 5.22 Construction Process Model Class Hierarchy At the process model level, the slots treat p i l e driving strategy, process model type and technical f e a s i b i l i t y . Chapter 5. CMSA Implementation 207 Process model type deals with the type of analytic or numerical algorithm used. In CMSA, for pile driving, the solution model used i s based on dynamic formulas which combine the hammer and s o i l properties to provide an approximate solution. The third slot i s of the technical f e a s i b i l i t y state which has a boolean value of either true or false. 5.3.4 Technical Feasibility Fart Given the selection of a pile type, a hammer and a pile driving strategy, the next step i s to check the f e a s i b i l i t y of the combination using the construction process model. This check involves assessing technical f e a s i b i l i t y and predicting time and cost performance. Technical f e a s i b i l i t y involves determining i f the pile can be driven to refusal without damaging i t and whether or not the rate of driving can satisfy production rate constraints or targets. The discussion below i s given in the context of the PREDICT and ANALYZE operators of section 4.3.4. 1. Predict: The "Predict" operator selects the appropriate construction process model for predicting the method performance after "Design" has been done. The rule shown in figure 5.23 combines the pre-selected matched sheet pi l e section and matched hammer properties in order to Chapter 5. CMSA Implementation 208 predict p i l e driving progress rate based on the dynamic formula, as derived in Appendix B. (@RULE = Single_Pile_Variable_Production_Time @INFCAT=5; ©COMMENTS = "This rule computes the variable component of driving a single pile based on Dynamic Formula. Note that effective energy consideration is not included -- e.g. hammer efficiency as function of the type of the hammer, and pile group effect on pile driving are not treated."; (@LHS = (Name (Tunnel.depth + 5) (L)) (Show ("Drive.txt") (@KEEP = FALSE; ©WAIT = TRUE;)) (Is (Driving_Conditions) ("Soft")) (Name("In_Pairs") (PilesDrivingPattern)) (Name ( < | Matched_Hammer | >. Strokes_per_Min) (F)) (Name( < | Matched_Hammer | > .Theor_Energy*Hammer. Total_Efficiency) (E)) (Name( < | Matched_SSP | > .Surface_Area*2) (SA)) (Name(< | Matched_SSP | > .Cross_Section_Area * 2) (Ap)) ) (@HYPO= SPProduction) (@RHS = (Write ("Hammer.nxp") (©TYPE = NXP; ©FILL=NEW; ©ATOMS = L,F,E,SA,Hammer_Type;\ )) (Write ("soil.nxp") (©TYPE = NXP; ©FILL = NEW; ©ATOMS = Soil_Type_l,\Start_l,Finish_l,Soil_Type_2, Start_2,Finish_2;\)) (Execute("Drive.exe") (©TYPE=EXE;@WAIT=TRUE;)) (Retrieve("out.nxp") (©TYPE = NXP; ©FILL=ADD; ©FWRD = TRUE; ©CREATE = |Var_Time|; \@ATOMS= Variable_Time.amount;)) (Retrieve ("out.nxp") (@TYPE=NXP; @FTLL=ADD; ©FWRD = TRUE; ©CREATE = | Feasibility | ;\@ATOMS=TechnicaI_Feasibility.State;)) (Do (CheckFeasibility) (CheckFeasibility)))) Figure 5.23 Technical Feasibility Rule 1. Left Hand Side The rule shown in figure 5.23 starts with the depth of the tunnel, or excavation, from the context information (tunnel depth) and adds 5 feet to i t , as a default, to determine the Chapter 5. CMSA Implementation 209 p i l e length. The extra length i s to minimize, i f not prevent, s o i l boiling or heave. The second condition displays a text f i l e , "Drive.txt", which briefs the user on the questions to be asked by the system, what this rule w i l l do and what is expected to happen after f i r i n g this rule. The third condition checks that the driving conditions correspond to Soft, as inferred from the strategy component based on s o i l conditions. The next condition assigns a string value of , ,In_Pairs , , for the Pile_Driving_Pattern strategy variable. Conditions four and five, assign the "Matched_Hammer" frequency and theoretical energy to variables F and E. Condition six assigns the Matched_SSP surface and cross section areas to variables SA and Ap. These variables are used as part of the input for dynamic formula routine. The hypothesis, SP_Production, refers to whether or not the sheet pil e production rate i s acceptable (True or False). 2. Right Hand Side The input required for pile driving routine i s written in two f i l e s (*.nxp format) using the "Write" operator. The f i r s t one i s "hammer.nxp", which includes hammer and pile data, and the second i s "soil.nxp", which includes the s o i l p r o f i l e input. Chapter 5. CMSA Implementation 210 The "Execute" operator cal l s the executable f i l e , "Drive.exe", and runs i t in the DOS environment. "Drive.exe" i s a compiled program for predicting the speed of p i l e driving given a soil/hammer/pile combination. It f a i l s i f a constraint such as the allowable number of blows per foot run i s violated. The "Retrieve" operator retrieves the database f i l e "Out.nxp", which contains a summary of running "Drive.exe" results. More detailed output is contained in the f i l e "Out.out". The "Retrieve" operator returns two values. The f i r s t is the variable time for pile driving, which indicates the speed of driving, and which can be tested against the pile driving productivity. The second variable i s the state of technical f e a s i b i l i t y , which is True i f the pi l e driving operation i s successful, otherwise i t i s False i f the pile i s not driven to i t s refusal or the damageability bound has been violated. The next hypothesis, "Do Check Feasibility", involves diagnosing the cause of a False response for technical f e a s i b i l i t y . For example, a driven "In_Pairs" construction strategy i s infeasible because the production rate i s too low, or the maximum number of blows before damage w i l l occur is exceeded. Chapter 5. CMSA Implementation 211 2. Analyze (Diagnose): Here, we examine how the CMSA can diagnose a failure of a technical f e a s i b i l i t y test and suggest a remedy. Consider the hypothetical example of a s o i l profile of loose sand on s t i f f clay (given a high loose sand/Stiff clay depth ratio) which implies soft driving conditions which in turn suggests driving in pairs. Suppose that the state of technical f e a s i b i l i t y returns the value False. Thus: The "In_Pairs" p i l e driving strategy may be infeasible and p i l e driving should be done In_Singles, even though the s o i l scenario suggested soft driving conditions. (Note that the pile driving strategy also affects the rate of production as well as the fixed time for pile driving.) The rationale behind this i s that when piles are driven "In_Singles", s o i l resistance, mainly skin f r i c t i o n and secondary end bearing, w i l l be decreased by half. The number of blows per foot exceeds the allowable limit which indicates high s o i l resistance, or insufficient hammer energy. One. way to remedy this situation i s to pick a larger hammer. However, the hammer selection criterion already considered picking the largest hammer that satisfied the steel sheet p i l e constraints. Therefore, i f a bigger hammer i s to be chosen, then the strength of the sheet p i l e must be increased. In other words, a heavier steel sheet pile section i s required. The production rate did meet the required progress rate variable plus fixed time. As a result driving "In_Singles", and/or use of a larger hammer and pile section may be suggested. Chapter 5. CMSA Implementation 212 A text f i l e of "Tech_Fea.txt" i s displayed to the user to explain this rule, as shown in figure 5.24, along with the expected actions. If the technical f e a s i b i l i t y condition is true, then the "Do" operator invokes the hypothesis of Compute_Production_of_SSP in order to calculate p i l e driving costs. (@RULE= Technically Feasible Alternative ©COMMENTS = "If pile driving conditions state (based on soil profile and conditions) is soft, then the pile driving pattern will be In_Pairs, else will be In_Singles. The selection of either driving strategy will be reflected in pile driving rate where fixed and variable time computation will be different for each.; (@LHS = (Yes (Technical_Feasibility.State)) (Is (DrivingConditions) ("Hard")) (Is (PilesDrivingPattern) ("In_Pairs")) (Show ("Tech_Fea.txt") (@KEEP = FALSE;@WAIT = TRUE;)) ) (@HYPO= CheckFeasibility) (@RHS = (Do (Compute_Production_of_SSP) (ComputeProductionofSSP)) ) Figure 5.24 Method Technical Feasibility i s True A rule directed at reversing the technical f e a s i b i l i t y condition from False to True by changing the pile driving strategy from "In_Pairs" to "In_Singles" i s shown in figure Chapter 5. CMSA Implementation 213 (@TRUE= TechnicallyFeasible @INFCAT=3; ©COMMENTS = "If the driving conditions is soft, pile driving pattern was In_pairs, and combination is not technically feasible then reset the pile driving pattern status into In_Singles. This requires setting the hypothesis SPProduction to Unknown to reevaluate the technical feasibility with the new strategy; (@LHS = (No (Technical Feasibility.State)) (Is (DrivingConditions) ("Soft")) (Is (Piles_Driving_Pattern) ("InPairs")) (Show ("Tech_Fea.txt") (©KEEP=FALSE;@WAIT=TRUE;)) ) (@HYPO= CheckFeasibility) (@RHS = (DO ("InSingles") (Piles_Driving_Pattern)) (Reset (SP_Production)) (Do (SPProduction) (SPProduction)) ) Figure 5.25 Technical Feasibility Diagnostic Rule The f i r s t condition checks to see i f the state of technical f e a s i b i l i t y i s False in order to perform a diagnosis. The second condition ensures that the driving condition i s "soft" which commonly suggests the use of the "In_Pairs" driving strategy. The third condition e x p l i c i t l y checks the variable "Pile_Driving_Conditions" value. The "Show" operator displays the text f i l e of "Tech_Fea.txt" to explain the process to the user. The hypothesis "Check_Feasibility" i s the same as the previous one with a different rule (in NExpert, a hypothesis embraces one or more rules) . In order to f i r e the f i r s t diagnostic rule, i t s inference category (QINFCAT shown in figure 5.25) i s set to three which is NExpert's priority Chapter 5. CMSA Implementation 214 mechanism that fires rules within the same hypothesis in ascending order of their inference category. To continue, assume this rule's conditions were satisfied and that the system control strategy altered the pi l e driving strategy from pile driving "In_Pairs" to pile driving "In_Singles". Then, the "Reset" operator resets the SP_Production hypothesis to Unknown such that control strategy backtracks and re-runs the "Drive.c" numerical routine with changed input variables. The "Do" operator then forces the re-evaluation of the Unknown hypothesis. 5.3.5 CMSA Chaining and Reasoning (Control Strategy) In this section, we explain the CMSA chaining and reasoning features for each major operator: Design, Predict and Analyze. Currently, CMSA i s implemented using backward chaining in the NExpert Object sense of chaining definitions. Its solution propagates in a forward reasoning fashion. NExpert, originally a rule based system, i s a hybrid system which makes i t non-trivial to rationalize i t s chaining and reasoning approach with conventional terminology. For instance, technical f e a s i b i l i t y , covered in section 5 . 3 . 4 , employs a t r i a l and error procedure for the Design/Predict/Analysis cycle in order to modify a design attribute value, i.e. a design alternative, construction Chapter 5. CMSA Implementation 215 resource, or construction strategy. This f a c i l i t y is labelled as non-monotonic reasoning in NExpert Object, i.e., making assumptions and retractions. What follows i s a description of the chaining and reasoning process in CMSA. Figure 5.26 shows a CMSA network of hypotheses — rules grouped in categories and inter-connected for chaining and inference propagation purposes. Shown in this figure are control strategy clauses used in the methods synthesis process. CMSA accomplishes the control strategy task using the RHS "Do" operator after an hypothesis has proved to be true. The Do operator performs two operations. It triggers a specific knowledge evaluation and then passes control to the next operation. For example, the following clause from the Design box of figure 5.26, is extracted from figure 5.20, and i s interpreted as f i r e the (Select_Pile_Driver) hypothesis. It is triggered i f the previous control clause, extracted from figure 5.12, Do (Select_Pile Driver) (Select Pile Driver) Do (Select SSP) (Select SSP) Chapter 5. CMSA Implementation 216 is successfully fired — i.e. select steel sheet p i l e . Once the p i l e driver selection i s done successfully, then control passes to applying the hammer dynamic formula embedded in the hypothesis (SP_Production), i.e. predict a productivity rate and check i f the pile reaches i t s refusal depth. The clause that triggers this hypothesis i s found in the Predict box in figure 5.26, i.e. Do (SP_Production) (SP_Production) CMSA starts with the Suggest operator which currently is a surrogate for the preliminary f e a s i b i l i t y knowledge base. The user i s prompted with a choice of GWSS alternatives. The Design task follows risk evaluation i f the level of risk i s acceptable. The design element subtasks exhibit forward reasoning and chaining modes. The hammer selection task i s more complicated. The hammer type i s based on s o i l p r o f i l e . Hammer size i s based on compatibility conditions with the design elements of steel sheet and soldier piles and goal requirements, (e.g. production rate), although i t is assumed with CMSA that the maximum energy hammer w i l l be the default choice. Thus hammer selection involves mixed modes of chaining with a forward mode of reasoning. By contrast, "Rl" an expert system that configures VAX computer systems, exhibits forward chaining with backward reasoning (McDermott 1984). Chapter 5. CMSA Implementation Figure 5.26 CMSA Model of Chaining and Reasoning Chapter 5. CMSA Implementation 218 Once operator method attributes are generated, the Predict operator selects a suitable method (procedural routines) for measuring the method performance for the given project context. Next, the Analyze operator applies analysis and/or interpretation routines to the results from the Predict operator. This, in general involves backward modes of chaining and reasoning. During the Predict/Analyze process, alternative ranking is based on minimum costs. Further, the Analyze operator may exhibit non-monotonic reasoning by making and retracting assumptions. For instance, i f the pi l e driving construction strategy state of In_Pairs is "True", but the method did not satisfy a goal criterion, (say production rate), or method elements compatibility resulted in a failed solution, the original plan could be altered by the CMSA control structure by retracting the pile driving strategy state of In_Pairs to be "False" and assessing the value "True" to the In_Singles state value. Thus, a mix forward mode of chaining and reasoning are used for the Analyze operator. 6. The Prototype Example 6.1 Introduction Features of the prototype CMSA implemented are described in this chapter. Input/output data, solution strategy processing, sensitivity of decisions to input changes, and the explanation f a c i l i t y are described f i r s t . An example problem i s demonstrated using a step by step approach. The example consists of two Cut-and-Cover shoring methods in parallel — steel sheet piles and soldier piles and lagging. The second part of this chapter i s devoted to a detailed example of the risk assessment process, implemented in NExpert Object as an independent module. 6.2 Example Problem Description The example problem i s a proposed tunnel, 1000 f t long, 60 f t deep and 20 f t wide. The s o i l profile consists of two layers: a 40 f t top layer of loose sand, and s t i f f clay below that. The contract duration for this project i s estimated to be a maximum of 240 days (or 5 ft/day), with a unit cost of $2,800 per foot ± $500. It i s assumed that upper and lower bounds for unit costs and production rates are given for the Cut-and-Cover tunnel alternatives. A record of a session i s provided, with user input and CMSA system responses. Example screens are shown as Chapter 6. Prototype Example 220 appropriate. Due to the lack of NExpert's explanation f a c i l i t i e s , explanation f i l e s with an extension "txt" were used extensively to explain some of the CMSA operations and query processes to the user. 6.2.1 Session Start A session commences by suggesting either a "datum" for input variables or suggesting a NExpert "hypothesis" as shown in screen 6.1. This window i s invoked by the command "Suggest" followed by the command "Knowcess" from the Expert Command Menu. From this window, any hypothesis could be highlighted and put in the Suggest/Keep corner before NExpert starts the session, e.g., the hypothesis "Select_A_GWSS" i s used to trigger the CMSA session. Also this window i s used to l i s t any datum (a premise in a rule) and triggers the beginning of the session by f i r s t evaluating the premise (as opposed to the hypothesis). Control i s then passed to evaluate the rule. Once a hypothesis or datum i s placed in the Suggest/Keep corner, "OK Knowcess" command is selected, the session starts. Chapter 6. Prototype Example 221 Screen 6.2 CMSA Overview Rule Network window Chapter 6. Prototype Example 222 7 Name 10 (1) a 9 Show "Textl .txt" @KEE JB(1) GWSS Is "Soldiet_P 9 =>Do (1) Select_GWSS Name 10 (1) a Show "Textl txt" @KEE (1) GWSS Is "Steel_Sh« =>Do (1) Select_GWSS ^ IRULE NETWOR Select_GWSS_of_SPL (114) [1] Select_GWSS_of_SSP (113) Screen 6.3 CMSA Rule Network Window After f i r i n g "Select_A_GWSS", a text f i l e (CMSA.txt) appears on the screen, introducing the user to the Cut-and-Cover tunnelling problem. Thus, CMSA is set in a "backward-chaining" mode. For the loaded CMSA knowledge base, screen 6.2 shows the overall rule network using the Overview Rule Network (ORN) window. The highlighted branches represent rules that are fired successfully during the session. NExpert Command Menus are shown (File, Edit, Expert, etc.) in the NExpert Environment Screen. By zooming in on the dotted box of ORN, the "Rule Network Window" focuses visually on a individual rule or a group of rules in the CMSA knowledge base, as shown in screen 6 . 3 . The Rule Network Window (RNW) i s enlarged to the size of the Chapter 6. Prototype Example 223 screen. In the upper right hand side of the slide, the Rule Network Overview (RNO) i s shown. In the middle of the RNW, the rule "Select_GWSS_of_SSP" i s shown fired (a condition or hypothesis i s indicated by the following icons: "True" by a check mark, "False" by a highlighted check mark, "Unknown" by a question mark, "Not Known" by an empty box, "Being Currently Investigated:" by a target, and "Being Evoked:" by an asterisk. Continuing with the session, CMSA provides the user with the following potential feasible alternatives as shown in screen 6 . 4 . » Select A GWSS ? 1. Steel Sheet Pile (SSP). 2. Soldier Pile and Lagging (SPL). The system asks the user to choose one GWSS alternative to start the detailed KB part. For example purposes, the two alternatives of Steel Sheet Piles and Soldier Piles and Lagging are assumed to have survived the preliminary screening process. Assume the user chooses the steel sheet p i l e alternative. The next part of the session involves specifying the s o i l context. Chapter 6. Prototype Example 224 6.2.2 Problem Context Specification What i s the Number of SoilLayers ? 1. One 2. Two » TWO The user i s prompted for either a single s o i l layer or a two s o i l layer scenario. The latter has been selected as shown in screen 6 . 5 . . CMSA asks for the s o i l type for the top layer as shown in screen 6.6. Alternatively, a second input format i s based on a Standard Penetration Test (SPT) Profile. If the f i r s t input format i s adopted , assume you select Loose_Sand as the top layer. Screen 6.7 shows the input for the 40 foot loose sand layer depth. Screen 6.8 shows that s t i f f clay i s chosen as the second s o i l layer. Screen 6.9 shows the water level input. Other queries include the tunnel depth in feet (60), and the tunnel length in feet (1000). This input format i s selected for simplicity. If the input for the SPT profile i s selected, the user w i l l be required to input SPT readings at 5 foot intervals for the depth of the tunnel or excavation. Chapter 6. Prototype Example 225 SESSION CONTROL What i s t h ^ Value o f GWSS ? Screen 6.4 GWSS Feasible Alternatives SESSION CONTROL! What i s the Ualue oF Number_of_Soil_Layers. ? T u o S o i l L a y e r s O n e S o i l L a y e r Screen 6.5 Soil Profile Specification (1) SESSION CONTROL What is - the Ualue» of- SoiI_Type_1 ? Hoose-Saridl OK Dense_Sand G m S o f t Clay S t i f f Clau ~ NO TK NO UN' Screen 6.6 Soil Profile Specification ( 2 ) Chapter 6. Prototype Example Screen 6.7 Soil Profile Specification (3) gsESSION CONTROL] What i s the Ualue of S o i l T y p * > 2 ? • fst i ff ClauB OK DenseSand k Soft Clay NOTKNOWN Screen 6.8 Soil Profile Specification (4) •SESSION C O N T R O L H H B H B I What i s the depth of Water Table ? OK 20; Screen 6.9 Water Table Level Input Chapter 6. Prototype Example 227 GWSS Technical Feasibility Assessment: The CMSA System then processes the following operations: i ? 1. Calculate pressure and moments-1': A rule i s used to do the computations for this scenario; see Appendix A for the computation procedures. The hypothesis "Calculate Pressures and Moments for "Two_Soil_Layers Scenario" is fired. 2. Select a suitable sheet p i l e i f i t exists with the given data base of "ssp.nxp" (see screen 6.10). The following operations are carried out by CMSA. Retrieve Steel Sheet Pile database "ssp.nxp"; Fire Section Modulus Rule which incorporates moments based on Retaining System Spacing; Select ASTM Steel Sheet Pile Section of PZ_27; Attach Selected PZ_27 to Matched_SSP; Inherit PZ_27 properties to Matched_SSP; If this fires successfully, then pass control to Select Pile Driver. Pressure and moments calculations are based on the "default" spacing of the retaining system (15 f t vertically and 12 f t horizontally). Chapter 6. Prototype Example 228 I -I STOP J 9Bf1 ] Select_SSP_of_PSA_32 (13S II] Select_SSP_of_PZ_27 (138) 9 (1) Section_Modulus > = Ki') SectionModulus < 9 <ISelected_SSP|>.desic 9 =>Do <|Selected_SSP|> 9 =>CieateOb|ect <|Selec ? =>Do (1) Select_Pile_D (1) Section_Modulus > = (1) SectionModulus < <ISelected_SSP|>.desic Show "hammei.txt* @KI =>Do <|Selected_SSP|> =>CreateObject <|Selec 9 =>Do (1) Select_Pile_D Screen 6.10 Hypothesis "Select_Suitable_Sheet_Pile" i s Fired Successfully After successfully selecting a suitable SSP data base (SSP.nxp), control i s passed to the "Select a Suitable Hammer" hypothesis. After a pile section has been successfully selected, the system selects a suitable hammer as follows. Based upon the s o i l s t r a t i f i c a t i o n and sequencing, a generic type of hammer i s selected. For this example, the impact hammer type i s more suitable than a vibratory one. A Double Acting Air Hammer (DAAH) i s found to be an appropriate impact hammer. Once the type of hammer i s specified (the hypothesis "Select Pile Driver" i s fired) , i t i s then sized using Chapter 6. Prototype Example 229 heuristic rules (the hypothesis "Selec^Vibratory^rJDAAH^amme^PD" i s fired, see screen 6.11). The CMSA proceeds as follows: Retrieve DAAH database "DAAH.nxp"; Based on experiential rules, the DAAH i s sized in terms of "Rated Delivered Energy". If the search i s not successful and no such size exists within the database, a message appears offering the alternatives of abandoning the search and quitting; selecting another hammer; or selecting another shoring alternative. The default strategy i s to pick the hammer with the highest delivered energy; If a DAAH hammer of the required size is found, i t w i l l be attached to the "Matched^ammer" object and i t s properties (model, delivered energy, unit cost, etc) are inherited from the "Selected_Hammer" l i s t . Cost is not considered as part of the criterion at this stage; Control i s then passed to the driving strategy knowledge base. The selected hammers class and dynamic objects created after f i r i n g the above rule are shown in screen 6.12. Chapter 6. Prototype Example (1) Soil_Type_1 Is "Loo (1) Soil_Type_2 Is "Sofi Retrieve 'dan nxp' @T" Name MAX(<|Selected_ <|Selected_Hammei|>T =>Let (1) Pile_D nver "D =>Let (1J Harnmei_Type =>CieateObjecl <|Selec =>DeleteOh|t:i:t <|Selec =>Do <|Suitable_Hamrri' 9 =>Do (1J SP_Pioductioi ? [1] Single_Pile 13T0P |1) Select_Viboiatoiy_PD_oi_Dj« Screen 6.11 Hypothesis "SelectSuitableHammer" i s Fired Successfully H Model - Unknown Selected Hammer LenQth_of_ Stroke « Unknown flaai.Weight * Unknown Strokes_per_Min » Unknown Theor_Enefgy - Unknown {+)Hammer_14 )Hammer_15 Screen 6.12 SelectedHammers Class and i t s Dynamic Objects Chapter 6. Prototype Example 231 To predict the production rate for the hammer/pile/soil scenario using the dynamic formula (the hypothesis "SP_Production" i s fired), the system next asks the following question. What i s the Efficiency of the Double Acting Air Hammer (DAAH)? The user i s requested to provide a hammer efficiency estimate (this could be automated i f desired). Under different project and,soil conditions, and indeed within the same site, the pile driver efficiency may vary. Assume in place of the CMSA default value, the user has estimated an efficiency of 71%. Screen 6.13 shows the hammer efficiency input by the user. The p i l e driving strategy is an important factor which must be determined prior to running the technical f e a s i b i l i t y routine (Drive.c routine), where the s o i l , pile, hammer, and driving strategy are combined. Soil conditions and stratifications determine the state of driving conditions. A "driving strategy text f i l e " i s displayed to explain to the user the different state variables of this parameter. Do you want to choose a driving strategy or leave i t to the system? Yes, No. If yes, then the user has to choose: Chapter 6. Prototype Example 232 1. Drive Piles In_Singles, 2. Drive Piles In_Pairs. A text f i l e "Drive.txt" i s displayed to explain and recommend the use of either strategy. Screen 6.14 shows one page of the Drive.txt f i l e which outlines the assumptions for using each p i l e driving strategy. For this example, we select "Drive Pile In Pairs". Screen 6.13 Hammer Efficiency Input Screen 6.14 "Drive.txt" Explanatory F i l e Chapter 6. Prototype Example 233 » Let CMSA Choose ? "Drive I n P a i r s " The input for the "Drive.c" routine is written into the f i l e s "hammer.nxp" and "soil.nxp". The f i r s t f i l e contains information about the hammer properties (hammer type, frequency, rated energy, etc) . The second f i l e contains information regarding the Standard Penetration Test profile, s o i l types, their sequences, and thickness. - The compiled "Drive.c" routine, "Drive.exe", i s executed (when the rule "SP_Production" is fired, this premise i s evaluated) by CMSA (see Appendix B for the derivation, coding, and input and output f i l e s for the "Drive.c" routine). The output parameters are written into the f i l e "out.nxp" for the NExpert interface, and "out.out", which allows for viewing the output results as a stand alone application. NExpert thus checks i f the state of Technical Feasibility i s "True" or "False", with the text f i l e "Tech_fea.txt" displayed to explain the underlying logic. "True" implies that the combination of the strategy of pile driving, matched sheet piles and matched DAAH hammer satisfy the pile driving constraint (blows/ft fx 120 and the performance measures f a l l within a required range). The quantitative output variables from the "Drive.c" routine are: incremental and cumulative number of hammer blows; skin f r i c t i o n ; end bearing f r i c t i o n ; average set; and speed of pile driving. The qualitative parameter i s the technical f e a s i b i l i t y state, which would be set to "False" i f the threshold of the average set/average speed of pile driving has been violated. The technical f e a s i b i l i t y state i s interpreted by CMSA (hypothesis "Check Feasibility" i s fired), as "True" for a Chapter 6. Prototype Example 234 feasible combination of the methods attributes. If the state i s found to be "False", then the diagnostic part of the technical f e a s i b i l i t y i s invoked to offer a plausible method attribute. For instance, i f the pi l e driving strategy was "In_Pairs", then i t i s possible to change this strategy to driving "In_Singles". The "Drive.c" routine finds the variable time component of the production rate. The total production time per pile, or set of piles, i s the sum of the fixed and variable components. The fixed time component i s obtained from the user in order to compute the total duration of the pile driving activity, and thus the production rate in terms of piles per day, foot run per day, and total cost. The following questions are used to perform a quantity take-off and estimate total project duration. What i s the fixed time to position, splice, and seat a pile? (Default 5 minutes) » 6 minutes What i s the Standard Length of the Steel Sheet Piles? SSPs come in different sizes, including 25 f t , 40 f t , 50 f t , 70 f t ; with the CMSA default being 25 f t . » 50 f t Production rate and quantity take-off computations for the steel sheet pile GWSS alternative are then performed. Chapter 6. Prototype Example 235 The detailed costs of this alternative include material, labor, and equipment. Next, control i s passed to the risk assessment KB. The second part of this chapter presents how risk i s quantified in monetary terms. For ease of risk component integration with CMSA and for computation speed, the risk computation has been encoded in a "Risk.c" routine. This routine is fired by CMSA and returns the risk cost fraction, as percentage of total cost, for SSP alternative based on risk categories and probability data input f i l e "SSP_Risk.nxp". The risk fraction is then added to the total cost of the SSP alternative. Selected variables relevant to this alternative are then written to the "Results_l.nxp" f i l e , as shown in screen 6.15. RESUTLS1.TXT \ T e c h n i c a l _ F e a s i b i l i t y . S t a t e \ = " T r u e " \Nunuer_of P i l e s \ = " 1 ,333" \ T o t a l _ P r o d u c t i o n T i n e _ i n _ D a y s \ = , , U 8 " \ P i o d u c t i u i t y _ i n _ N u n b e r _ o f _ P i l e s _ P e r _ D a y \ = " 2 8 ' \ S e l e c t e d _ S t e e l _ P i l e \ = " P Z _ 2 7 " \ P i l e D r i u e r \ = " D o u u l e _ f l c t i n g _ A i r Hanner" \Hanner_Hodel \="11 B3" * # * * * * * * * * ID Screen 6.15 "Results!.nxp" F i l e Next, the system considers the second feasible solution of soldier piles and lagging, queries the user in a way Chapter 6. Prototype Example 236 similar to the f i r s t alternative and writes the results to "Results_2.nxp". Once a preliminary feasible alternative has proved to be feasible at the low level, i t s attribute values are stored in i t s results frame, and the criterion of minimum total cost i s applied to rank the alternatives. For our example, CMSA selects the steel sheet p i l e alternative. Based upon other factors, such as production rate, preferred resources, and other attribute values generated in the CMSA session, the user may wish to over-ride this recommendation. Figure 6.1 displays the context variables (the high level primitives for the represented major concepts) that are instantiated during the CMSA session. The direction of inference shown is from bottom to top. At the bottom, three major input categories are identified: tunnel dimensions, s o i l profile and conditions, and risk input. The structural members f e a s i b i l i t y instantiation is dependent upon site conditions and tunnel dimensions. Risk input data i s necessary for assessing the risk component. Once technical f e a s i b i l i t y i s proven, the session is terminated, and the session results are printed in output f i l e s . H Specify Retaining System (Type and Spacing) SSP Risk Is Feasible (funnel Dimensions) (JSoll Profile and Conditions) (^ Rlsk Conditions^) W CO Chapter 6. Prototype Example 238 6.2.3 Modified Example Results from a second example are presented to demonstrate the a b i l i t y of the CMSA to respond to a change in input conditions. A variation of the f i r s t example i s used to ill u s t r a t e the in f e a s i b i l i t y of a GWSS alternative in a different project context. For the previous example s o i l profile, suppose the following variables are changed: the top layer i s changed from loose sand to dense sand and the second layer changes from soft clay to s t i f f clay; and the hammer efficiency i s changed from 70% to 50%. Assume a l l other project conditions are the same. The system executes the control strategy and selects a PZ_32 steel sheet pile section and a Vulcan single acting a i r hammer of model "0" (hammer energy i s 24,375 l b - f t per blow). A 50% efficiency for the Single acting a i r hammer i s then specified. Given the s o i l s t r a t i f i c a t i o n nature (dense sand on s t i f f clay i s considered to be a compacted soil) , the system recommends using the driving strategy "Drive In Singles" After executing the compiled "Drive.c" routine, "Drive.exe" is executed (the hypothesis "SP_Production" is fired). A text f i l e "Tech_feal.txt" is displayed. This text f i l e explains that the technical f e a s i b i l i t y state i s "False" and thus the search for a feasible solution has failed. By retrieving the data encyclopedia notebook Chapter 6. Prototype Example 239 "resultsl.nxp", i t i s found that pil e driving was halted at a depth of 55 f t . A possible remedy includes using a larger hammer than the one contained in the existing SAAH knowledge base. Therefore, for this project context, the CMSA informs the user that i t failed because i t i s not possible to drive the selected p i l e to i t s refusal depth. Thus, i t declares that the SSP alternative i s infeasible at the detailed level. 63 Risk Component Assessment Implemented 6.3.1 Introduction The risk component was implemented by using NExpert Object and by writing an evaluation routine. The NExpert module takes the user through a step by step input data process and displays text f i l e s to explain the decision tree. The evaluation routine is used to embed the risk component in CMSA to speed up the session and to minimize user input. Both implementations follow the risk assessment framework as described in section 4.4. 6.3.2 NExpert Risk Implementation An example session i s used for demonstration (see figure 4.12, Risk Assessment Tree). Two shoring alternatives are considered in parallel: Steel Sheet Piles (SSP), and Soldier Piles and Lagging (SPL). Chapter 6. Prototype Example 240 A session commences as follows: What i s the I n i t i a l Cost for Steel Sheet Piles? This cost i s assumed to be obtained from the detailed f e a s i b i l i t y part. » S 3.00 M What i s the I n i t i a l Cost for the Soldier Piles and Lagging \? » $ 2.80 M This cost i s assumed to be obtained from the detailed f e a s i b i l i t y part. What i s the Likelihood of Worse_Than_Expected Subsurface Conditions ? » 33 % The questions that follow treat the assessment of the risk component for the Worse_Than_Expected scenarios. The user i s f i r s t asked i f there i s a chance of a catastrophic damage risk for this alternative, or i f the likelihood of this catastrophic risk exceeds a pre-specified value. If yes, then the alternative w i l l be eliminated without asking further risk related questions. If the answer i s no, the computation then takes place for the total risk cost. Chapter 6. Prototype Example 241 The system starts with the computation of the Worse_Than_Expected risk branch. If the risk costs associated with this branch are bounded, then the alternative w i l l most l i k e l y be feasible risk-wise. To show how NExpert Object i s used to implement this part, the queries for risk categories and probabilities proceed as follows for the Steel Sheet Piles GWSS alternative. What i s the Likelihood of Catastrophic Damage of SSP ? If catastrophic damage likelihood i s less than a specified threshold, the solution carries on. If not, the plan f a i l s , and SSP w i l l be infeasible. Assume i t is 0. » P_ What i s the Likelihood of Element_Damage_Given_Worse_Than Expected_Conditions ? » 0.22 Figure 6.2 maps the probability input into the risk assessment decision tree. Shoring Alternatives Event Chances Consquences Costs (Outcomes) Encountered Geological More Favorable Steel Sheel Pile System Damage (Catastrophic Damage) Infinity oo Risk Category Risk Cost Equipment Loss 10% Labor Loss 10% Material Loss 10% Life Loss 10% Subsurface Sub-sidence Loss 10% Season Loss 10% Other Losses 10% Chapter 6. Prototype Example 243 Next, a breakdown of the costs, by resource type and site conditions, i s specified by the user, in terms of a fraction of the i n i t i a l cost for each alternative previously input by the user. Everything i s priced against the single input number of expected cost as estimated by the detailed f e a s i b i l i t y component. Starting with the Worse_Than_Expected conditions scenario, the following risk categories costs are relevant to Element_Damage_Given_Worse_Conditions scenario. The risk categories values are shown below: - What i s the Equipment Loss = 0.10 - What i s the Labor Loss =0.10 - What i s the Material Loss = 0.10 - What i s the Life Loss = 0.10 - What i s the Subsurface Subsidence Loss = 0.10 - What i s the Season Loss = 0.10 - What i s the Other Losses = 0.10 A calculation takes place to find the likelihood of no element damage given worse than expected condition scenario, Likelihood of No_Element_Damage_Given_Worse_Conditions =1-(Likelihood of Element_Damage_Given_Worse_Conditions) -(Likelihood of Catastrophic_Damage_Given_Worse_Conditions) The likelihood for the catastrophic system failure i s assumed to be 0, therefore, the no damage likelihood i s : = 1 - 0.22 - 0 = 0.78 Chapter 6. Prototype Example 244 Risk categories relevant to this component are l i s t e d below: - What i s the Equipment Loss - what i s the Labor Loss - What i s the Material Loss - what i s the Life Loss - what i s the Subsurface Subsidence Loss - what i s the Season Loss - What i s the Other Losses What i s the Likelihood of "As_Expected" conditions ? » 40 % Similarly, other risk categories, relevant to the AsExpected scenario of encountered s o i l conditions, w i l l be considered: - What i s the Equipment Loss - What i s the Labor Loss - What i s the Material Loss - What i s the Life Loss - What i s the Subsurface Subsidence Loss - What i s the Season Loss - what i s the Other Losses The likelihood of the More_Favorable_Than_Expected subsurface conditions i s : = ( 1 - Likelihood of Worse_Than_Expected Conditions -Likelihood of As_Expected Conditions) = ( 1 - 0.33 - 0.40) = 0.27 Similarly, other relevant risk categories to the More_Favorable_Than_Expected scenario of encountered s o i l conditions w i l l be treated; noting that a negative risk fraction indicates savings. = 0.00 = 0.00 = 0.00 = 0.00 = 0.00 = 0.00 = 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Chapter 6. Prototype Example 245 What i s the Equipment Loss What i s the Labor Loss What i s the Material Loss What i s the Life Loss What i s the Subsurface Subsidence Loss What i s the Season Loss What i s the Other Losses 0.01 0.01 0.01 0.01 0.01 0.01 0.01 Costs may now be calculated as follows: Total Cost for the SSP Alternative = I n i t i a l Cost for SSP + SSP Risk Since risk i s expressed as percentage of total cost, then Total Cost = I n i t i a l Cost ( 1 + SSP_Risk fraction)) Total Cost for the SSP Alternative = $ 3.00 M * (1 + 0.03) = $ 3.09 M SSP Risk = $ 90,000 Similarly, the process w i l l be repeated for the GWSS of the Soldier Piles and Lagging alternative. Table 6.1 summarizes a l l input data for both GWSS alternatives. SPL costs are calculated as follows: Total Cost for the SPL Alternative = I n i t i a l Cost for SPL + SPL_Risk Total Cost for the SPL Alternative = 2.80* (1+0.14) $ 3.19 M SPL Risk = $ 390,000 Based upon total costs including risks, the steel sheet pi l e alternative is preferred to the soldier p i l e and lagging alternative. At the end of the session, a text f i l e w i l l be displayed, indicating both alternatives are feasible and ranking the alternatives in a descending order of total Chapter 6. Prototype Example 246 costs. Thus, the steel sheet pile alternative w i l l be ranked f i r s t , followed by soldier piles and lagging. Figure 6.3 shows a flow chart of the risk assessment algorithm as implemented. 6.3.3 Risk Routine The risk component is implemented in "Risk.c" routine as an external program. The executable f i l e "risk.exe" i s fired by using a rule in the CMSA prototype. The routine basically performs the risk assessment computation as described in section 4.4. The "Risk.c" input data f i l e includes the risk cost categories and probabilities for each alternative. The routine returns the results to an output f i l e where CMSA then retrieves the output variable for further computation. For example, table 6.2 shows the input f i l e "SSP_Risk.nxp" for the SSP alternative. The output f i l e "Riskl.nxp" contains the variable "SSP_Risk.Fraction = 0.0319" which is the sum of the risk assessment u t i l i t y as a fraction of the i n i t i a l cost. Thus, the total cost for the SSP alternative including risk i s equal to (1 + SSP_Risk.Fraction) times the i n i t i a l cost as estimated by CMSA. Chapter 6. Prototype Example 247 Likelihood of Scenarios Conditions - Worse_Than_Expected - Catastrophic_Damage - Element_Damage - No_Damage - As_Expected - Favorable_Than_Expected Risk Categories (Consequences) Risk Categories for Worse Than Expected Catastrophic Damage SSP Alternative 0.33 0.00 0.22 0.78 0.40 0.27 0.00 SPL Alternative 0.33 0.00 0.33 0.67 0.40 0.27 0.00 ElementDamage Equipment Loss 0.10 0.20 Labor Loss 0.10 0.20 Material Loss 0.10 0.20 Life Loss 0.10 0.20 Subsurface Subsidence Loss 0.10 0.20 Season Loss 0.10 0.20 OtherLosses 0.10 0.20 No_Element_Damage Equipment Loss 0.00 0.00 Labor Loss 0.00 0.00 Material Loss 0.00 0.00 Life Loss 0.00 0.00 Subsurface Subsidence Loss 0.00 0.00 Season Loss 0.00 0.00 OtherLosses 0.00 0.00 Risk Categories for AsExpected Equipment Loss 0.00 -0.01 Labor Loss 0.00 -0.01 Material Loss 0.00 -0.01 Life Loss 0.00 -0.01 Subsurface Subsidence Loss 0.00 -0.01 Season Loss 0.00 -0.01 Other Losses 0.00 -0.01 Risk Categories for MoreFavorableThanExpected Equipment Loss -0.01 -0.02 Labor Loss -0.01 -0.02 Material Loss -0.01 -0.02 Life Loss -0.01 -0.02 Subsurface Subsidence Loss -0.01 -0.02 Season Loss -0.01 -0.02 Other Losses -0.01 -0.02 Table 6.1 Risk Assessment Data Input Summary c ^ ^ GWSS Sheet a Steel Piles "(Alternative Is Not Feasible) Get Element Damage Likelihood Given VTT E Conditions /Get Risk Category Cost ~N V Given Element Damage^ / -Labor -Material - Equipment -Season Loss - Ground Inflow - Retaining System -OtherLosses Oat Risk Category ttem_CosfNi Given No Element T Kg -Material -Equipment - Season Loss - Ground Inflow - Retaining System -OtherLosses As Expected Conditions^  ~ Likelihood (given1 Get Risk Category Cost As Expected Condi -Labor -Material - Equipment -Season Loss • Ground Inflow - Retaining System - Other Losses Get More FavorableThan Conditions Likelihood (IT F T E) (fea Risk CalegoryltecrTco^ ) - Labor -Material - Equipment - Season Loss -Ground Inflow - Retaining System -OtherLosses O •5 »t o n h o ct o rr ft H M H ft Rank Feasible Alternatives Try Another Alternative fit II rO 00 Chapter 6. Prototype Example 249 \Favorable_Than_Expected.Probability\="0.27" \Equipment_Loss_for_FTE\="-0.010" \Labor_Loss_for_FTE\ = "-0.010" \Material_Loss_for_FTE\="-0.010" \Subsurface_Subs_Loss_for_FTE\="-0.010" \Season_Loss_for_FTE\="-0.010" \Life_Loss_for_FTE\"-0.010" \Other_Losses_for_FTE\ = "-0.010" \Favorable_Than_Expected.Finish\="" \As_Expected.Probability\="0.40" \Equipment Loss_for_AE\="0.00" \Labor_Losr for_AE\="0.00" \Material_Loss_for_AE\="0.00" \Subsurface_Subs_Loss_for_AE\="0.00" \Season_Loss_for_AE\ = "0.00" \Life_Loss_for_AE\"0.00" \Other_Losses_for_AE\ = "0.00" \As_Expected.Finish\ = "" \Worse_Than_Expected.Probability\ = "0.33" \No_Damage_Given_W_T_E.Probability\ = "0.78" \Equipment_Loss_for_ND\="0.00" \Labor_Loss_for_ND\ = "0.00" \Material_Loss_for_ND\ = "0.00" \Subsurface_Subs_Loss_for_ND\ = "0.00" \Season_Loss_for_ND\ = "0.00" \Life_Loss_for_ND\"0.00" \Other_Losses_for_ND\ = "0.00" \No_Damage_Given_W_T_E.Finish\="" \Element_Damage_Given_W_T_E.Probability\ = "0.22" \Equipment_Loss_for_ED\ = "0.10" \Labor_Losr for_ED\ = "0.10" \Material_Loss_for_ED\ = "0.10" \Subsurface_Subs_Loss_for_ED\="0.10" \Season_Loss_for_ED\ = "0.10" \Life_Lxjss_for_ED\"0.10" \Other_Losses_for_ED\ = "0.10" \Element_Damage_Given_W_T_E.Finish\ ="" \System_Damage_Given_W_T_E.Probability\ = "0.00" \Equipment Loss_for_SD\ = "0.00" \Labor_Losr for_SD\ = "0.00" \Material_Loss_for_SD\="0.00" \Subsurface_Subs_Loss_for_SD\="0.00" \Season_Loss_for_SD\ = "0.00" \Life_Loss_for_SD\"0.00" \Other_Losses_for_SD\ = "0.00" \System_Damage_Given_W_T_E.Finish\ ="" \Worse_Than_Expected.Finish\ = "" ****************************** Table 6.2 "SSP_Risk.nxp" Input F i l e for SSP Alternative 7. Conclusions and Recommendations for Further Research 7.1 Summary This thesis reveals that the problem of construction methods selection i s complex and ill-structured. A tool i s needed to categorize, rank, prune, and synthesize methods alternatives. A step in this direction was provided through a KBES framework approach, which further led to the concept of a construction methods selection shell. A prototype system called CMSA was designed and implemented. This system interacts with the user to suggest, design, predict, and analyze the assembled method. Presently, CMSA may be considered as a developmental stage prototype, i.e., an expert system which has passed the conceptual stage, and where a f i r s t prototype has been built and run but not completely validated. Several method attributes, goals, and constraint constructs (or primitives) have been implemented. The structure adopted w i l l allow other researchers to add other methods, alter some rules, and adapt CMSA to different construction problem domains as desired. Chapter 7. Conclusions 251 7.2 Contribution of The Thesis The main contributions of this thesis are: 1. Structuring of the problem of method selections by setting out a generalized definition of the construction method selection problem (Chapter 4) which can accommodate a diverse range of methods selection problems. 2. The devising of an expert system conceptual model involving declarative and procedural knowledge which reflects a practical approach to generating and then reducing a feasible set of alternatives. A two tiered process of preliminary and detailed f e a s i b i l i t y is proposed to screen, prune, and synthesize a construction method. 3. The implementing of a small prototype, called CMSA, to demonstrate the f e a s i b i l i t y of the detailed f e a s i b i l i t y part architecture proposed in the conceptual model. Most of the features identified in the detailed f e a s i b i l i t y part of the conceptual model were implemented in CMSA. Chapter 7. Conclusions 252 73 Further Research Listed below are several avenues to further pursue in order to f u l l y explore the potential for knowledge based systems to make a substantive contribution to productivity improvement through development of a construction methods selection tool. 1. Attention should be focused on exploring a range of technology rich example problems such as Cut-and-Cover tunnelling, structural cycle design for high rise construction, bridge construction and so forth in order to further generalize the conceptual model proposed. 2. The preliminary f e a s i b i l i t y phase component of the model should be implemented and further elaborated on; undertaking further knowledge acquisition, implementing the proposed data base for methods, enlarge the regulation knowledge component, and so forth. 3. At the detailed f e a s i b i l i t y level, the use of optimization techniques to affect tradeoffs between fixed and variable costs (e.g. spacing of retaining system members) should be explored. Chapter 7. Conclusions 253 4. Experiment implementing CMSA using a purely Object Oriented Programming (OOP) approach. For instance, this can be applied for technical f e a s i b i l i t y of a method as a generalized and parental object. Such objects can begat other objects such as method designs, strategies, model processes, and construction processes. This involves object design, different inference mechanisms (message exchanges), and should be compared to the previous implementation in terms of possible code size reduction, and ease of understanding. 5. Addressing validation issues for CMSA through more input from practitioners. In some instances, where knowledge was represented as high primitives, i t i s preferable that such knowledge be available in the system in a more objective format. For instance, in the risk assessment of a shoring alternative, i t i s assumed that the user estimates the value of GWSS wall failure probabilities based on project context and water table conditions. 6. More documentation of methods and the procedural and experiential knowledge that accommodates them, i s needed. For instance, there are incomplete knowledge Chapter 7. Conclusions 254 bases such as the vibratory hammer KB which need further procedures to estimate productivity. Most design KBs (SSP, SPL, Struts, and wales) employed one or two design methods. Other design methods, which may be more suitable under a specific project context, could be included. 255 Bibliography [I] Adeli, H., and Paek, Y J . (1986)., "Computer-Aided Design of Structures Using Lisp.", Computers & Structures, Vol.22, No.6, pp.939-956. [2] AISC (Nov 1978)., Specification for the Design, Fabrication and Erection of Structural Steel Buildings., Chicago: American Institute of Steel Construction Inc. [3] Ashley, D.B. (1980).,"Simulation of Repetitive Unit Construction.", Journal of Construction Engineering and Management, Vol.106, No.C02, pp.185-195. [4] Ashley, D.B. et al. (1983)., "Critical Decision Making During Construction.", Journal of Construction Engineering and Management, Vol.109, No.2, pp. 146-162. [5] Ashley, D.B. et. al. (1979)., "Advantage and Limitations of Adaptable Tunnel Design and Construction Methods.", Proceedings of Rapid Excavation and Tunnelling Conference, Atlanta, Georgia, U.S., pp.989-1011. [6] Ayyub, B., and Haider, A. (Dec 1985). "Decisions in Construction Operations.", Journal of Construction Engineering and Management, Vol.111, No.4, pp.343-357. [7] Barber, G. H. (1987)., "Fitting the Hammer to the Pile and the Soil", Pile Buck, Pile Buck Inc., First June Issue, P.O.Box 1056, Jupiter, FL 33468-1056, pp. 26-34. [8] Bernold, L.E. (1987)., "Goal Oriented Expert-Simulation For Building Construction.", Transaction of the Building Cost Research Conference on Building Cost Models and Computers, pp.409-418. [9] Bickel, J.O., and Kuesel, T.R. (1982)., Tunnel Engineering Handbook, New York: Van Nostrand Reinhold Co. [10] Blanning, R.W. (1984)., "Management Applications of Expert Systems.", Information and Management, No.7, pp.311-316. [II] Bobrow, D.G., and Stefik, M. (1983)., The Loops Manual, Xerox Corporation. [12] Bonissone, P.P., and Tong, R.M. (1985). "Editorial: Reasoning with Uncertainty in Expert Systems.", International Journal of Man-Machine Studies, No.22, pp.241-250. Bibliography 256 [13] Bowles, J.E. (1977)., Foundation Analysis and Design, New York: McGraw-Hill Inc. [14] Bruce, D. A., and Jewell, R. A. (1987)., "Soil Nailing: Applications and Practice, Part (2).", Ground Engineering, Vol.20, No.l, pp. 10-15. [15] Buchanan, B.G, and Shortliffe, S.H.(1984)., Rule Based Expert Systems, Reading, Mass.: Addison-Wesley Pub. Co. [16] Canadian Foundation Engineering Manual (1978)., Canadian Geotechnical Society, Montreal, Quebec, Canada. [17] Caterpillar Performance Handbook (1982)., Caterpillar Tractor Co., Peoria, Illinois, U.S.A. [18] Charniak, E., and McDermott, D. (1985)., Introduction to Artificial Intelligence., Reading, Mass.: Addison-Wisely Pub. Co. [19] Chin, M . W. (1987)., "Some Potential Applications of Expert Systems in Construction Engineering and Management in the West Indies", Managing Construction Worldwide, Vol.1, London, U.K., pp.591-597. [20] Clancey, W. J . (Dec 1985)., "Heuristic Classification", Artificial Intelligence, Vol. 27, pp.289-350. [21] Clocksin, W.F., and Mellish, C. S. (1981)., Programming in Prolog, Berlin: Springer-Verlag. [22] CSA Standard 085-1976 (1976)., "Code for the Engineering Design of Wood.", Canadian Standard Association, Rexdale, Ontario, Canada. [23] Duda, R.O., Gasching, J.G., and Hart, P.E. (1979)., Model Design in the PROSPECTOR Consultant System for Mineral Exploration. Expert Systems in the Micro Electronic Age, Michie, D., ed. (1979)., Edinburgh: Edinburgh University Press. [24] Duda, R.O., Hart, P.E., Nilsson, N J . , and Sutherland, G. (1978)., "Semantic Network Representation in Rule Based Inference systems.", Pattern Directed Inference Systems, Waterman,D., and Hays-Roth, F., (Eds.) New York: Academic Press. [25] Gaarslev, A. (1977)., "Optimizing Interrelated Decisions in Construction Plannings.", Relations in Project Management, Ninth Project Management Institute Conference, Drexel-Hill, Pennsylvania, pp.344-352. [26] Gashing, J . et al. (Aug 1981)., "Developments of a Knowledge Based System for Water Resources Problems.", SRI Project 1619, SRI International, Stanford, California, U.S.A. Bibliography 257 [27] Gates, M., and Scarpa, A., (1980)., "Criteria for Selection of Construction Equipment.", Journal of Construction Engineering and Management, Vol.106, No.2, pp.207-219. [28] Gates, M., and Scarapa, A. (1984)., "Optimum Penetration of Friction Piles.", Journal of Construction Engineering and Management, Vol. 110, No.4, pp.491-510. [29] Goldberg, A. (1981)., "Introducing the Smalltalk-80 System, BYTE, Vol.6, No.8, pp.36-48. [30] Gray, CA. , and Little J . (1985)., "A Systematic Approach to The Selection of An Appropriate Crane for a Construction Site.", Construction Management and Economics, No.3, pp. 121-144. [31] Gray, C A . (1986)., "Intelligent Construction Time and Cost Analysis.", Construction Management and Economics, No.4, pp. 135-150. [32] Gruber, T., and Cohen, P. (1987)., "Principles of Design for Knowledge Acquisition.", IEEE Conference on Artificial Intelligence Applications, pp.9-15. [33] Halpin, D.W., and Bernold, L.E. (1986)., "Advance Microcomputer Simulation for Construction Managers.", Proceedings of the 4th Conference on Computing in Civil Engineering, New York, N.Y., pp. 762-772. [34] Halpin, D.W., and Woodhead, R.W. (1976)., Design of Construction and Process Operations, New York: John Wiley and Sons. [35] Harris, F., and Alkass, S. (1988)., "Expert System for Earthmoving Equipment Selection in Road Construction.", Journal of Construction Engineering and Management, Vol.114, No.3, pp.344-363. [36] Harris, F.C., and Wijesundera, A (1985)., "The Integration of an ES in the Construction Planning Process.", Proceedings of 2nd International Conference of Civil and Structural Engineering Computing, Vol.2, pp.399-405. [37] Hendrickson, C.T. et al. (1988)., "Expert Systems for Construction Planning.", Journal of Computing in Civil Engineering, Vol.1, No.4, pp.253-269. [38] Hendrickson, C.T. et al. (1987)., "Hierarchical Rule-Based Activity Duration Estimation", Journal of Construction Engineering and Management, Vol.113, No.2, pp.288-301. [39] Howard, R A . (1983)., "The Evolution of Decision Analysis.", Readings on The Principle and Applications of Decision Analysis, Vol.1: General Collections, Howard, R. A. and Matheson, J.E. (Eds.), Strategic Decisions Group, pp.5-16. [40] Hunt, H.W. (1979)., Design and Installation of Driven Pile Foundation, Associated Pile and Fitting Corp., Clifton, New Jersey, U.S.A. Bibliography 258 [41] Hurd, M.K. (1961)., "Formwork For Concrete", ACI Special Publication No. 4, Detroit, Illinois. [42] Jackson, P. (1986)., Introduction to Expert Systems., U.K., Wokingham: Addison-Wesley Pub. Co. [43] Jones, J.P., and Newcomb, F.M. (1963). "Introduction to Pile Driving", 77? No.23, Construction Institute of Stanford University, Stanford, CA., U.S.A. [44] Kim, Y.W. (1984)., Decision Support Framework in Adaptable Tunnelling, Ph.D Thesis, Department of Civil Engineering, Cambridge, MIT, U.S.A [45] Kunz, J.C. et al. (1986)., "Contingent Analysis for Project Management Using Multiple Worlds.", Proceedings of First International Conference on Application of Artificial Intelligence in Engineering Problems, London, U.K., pp.707-718. [46] Law, G.K. (1987)., Decision Support System for Construction Cycle Design, Master Thesis, Department of Civil Engineering, University of British Columbia, Vancouver, Canada. [47] Levitt, R.E., and Ashley, D.B. (1987)., "Expert Systems in Construction: Work in Progress.", Journal of Computing in Civil Engineering, Vol.1, No.4, pp.303-311. [48] Levitt, R.E., Kartam, NA. , and Kunz, J.C. (1988)., "Artificial Intelligence Techniques for Generating Construction Projects Plans.", Journal of Construction Engineering and Management, Vol.114, No.3, pp.329-343. [49] Logcher, R.D., and Leston, N.B. (1985)., "An Expert System Framework for Analyzing Construction Project Risks.", Center for Construction Research and Education, Department of Civil Engineering, MIT, U.S.A. [50] Lutz, J.D., Chang, L., and Napier, T. (1990)., "Evaluation of New Building Technology", Journal of Construction Engineering and Management, Vol.116, No.2, pp.281-299. [51] McDermott, J . (Fall 1984)., "Rl revisited: Four years in the trenches.", Al Magazine, pp.21-32. [52] Means Heavy Construction Cost Data (1987)., R.S. Means Company Inc., 1st Annual Edition, Kingston, Massachusetts, U.S.A. [53] Megaw, T.M, and Bartlett, J.V. eds. (1981)., Tunnels: Planning, Design, Construction., Vol.1 and Vol.2, U.K., Chicester: Ellis Horwood Limited. [54] Mellichamp J.M., and Wahab, A.F. (1987)., "An Expert System for FMS Design.", Simulation, Vol.48, No.5, pp.201-208. [55] Meritt, R.S. ed. (1976)., Standard Handbook for Civil Engineering, New York: McGraw-Hill Book Company. Bibliography 259 [56] Mikroudis, G., and Wilson, J.L. (1986)., "Selecting a Tool for Building a Knowledge-Based System.", ATLSS Report No. 86-01, Lehigh University, Lehigh, U.K. [57] Miller, H.D., and Grant, D. (1987). "Tunnel Simulation Computer Model", Science Council of British Columbia, Vancouver, B.C., Canada. [58] Minsky, M., (1968)., Semantic Information Processing. Cambridge, Mass.: MIT Press. [59] Minsky, M., (1975)., "A framework for Representing Knowledge.", Artificial Intelligence, Winston, P.H., ed., 1986, Mass.: Addison-Wesley Pub. Co. [60] Mohan, S. (1990)., "Expert Systems Applications in Construction Management and Engineering", Journal of Construction Engineering and Management, Vol.16, No.l, pp.87-99. [61] Murray, K J . , and Sheppard, S.V. (1988)., "Knowledge Based Simulation Model Specification.", Simulation, Vol.50, No.3, pp. 112-119. [62] Neuron Data Inc. (1988)., NExpert Object Fundamentals: PC Version 1.1, Palo Alto, California, U.S.A. [63] O'Keefe, R.M. (1986)., "Simulation and Expert Systems - A Taxonomy and some Examples.", Simulation, Vol. 46, No.l, pp. 10-16. [64] Palmer, D J , and Tomlinsan, M. (1981)., "Are piles necessary and what determine choice?", U.S. Department of Transportation Reports FHWA-RD-73-40 dWi, pp.111-120. [65] Paulson, B.C. et al. (1981)., "Simulation and Analysis of Construction Operations.", Journal of Construction Engineering and Management, Vol.109, No.2, pp. 89-104. [66] Paulson, B.C. et al. (1987)., "Simulation and Analysis of Construction Operations by Microcomputers." Journal of Construction Engineering and Management, Vol.113, No.2, pp.302-314. [67] Peck, R.B. (1984)., "State of The Art : Soft Ground Tunneling.", Proceedings of two sessions at Geotek'84, Tunnelling in Soil and Rock, Atlanta, Georgia, pp.1-11. [68] Peurifoy, R.L. (1970)., Construction Planning, Equipment, and Methods, 2nd Edition, New York: McGraw-Hill Book Co. [69] Ratay, R.T. (1984)., Temporary Structures in Construction, New York: McGraw-Hill Book Co. Bibliography 260 [70] Reddy, Y.V., and Fox M. (1982)., "Knowledge Representation in Organization Modeling and Simulation: A Detailed Example.", Proceedings of 13th Annual Pittsburgh Conference, April 1982, pp.685-692. [71] Riggs, L.S. (1980)., "Simulation of Construction Operations"., Journal of Technical Issues, Vol.106, No.l, pp.145-163. [72] Ringwald, R.C. (1987)., "Bunching Theory Applied to Minimize Cost", Journal of Construction Engineering and Management, Vol.113, No.2, pp.321-326. [73] Schroeder, W.L. (1980)., Soils in Construction, 2nd ed., New York: John Wiley & Sons. [74] Selinger, S. (1980)., "Construction Planning for Linear Projects." Journal of Construction Engineering and Management, Vol. 106, No.2, June 1980, pp. 195-205. [75] Shafer, G. W. (1976)., A Mathematical Theory of Evidence. Princeton, New Jersey: Princeton University Press. [76] Shannon, R.E. (1986). "Intelligent Simulation Environments.", SCS Simulation Series,Vol. 17, No.l, pp. 150-156. [77] Shannon, R.E. (1987)., "Models and Artificial Intelligence.", Proceedings of the 1987 Winter Simulation Conference, pp. 16-23. [78] Shen, C.K. et al. (1981a)., "Field Measurement of an Earth Support System.", ASCE GT12, pp.1625-1642. [79] Shen, C.K. et al. (1981b)., "Ground Movement Analysis of Earth Support Systems.", ASCE GT12, pp.1609-1624. [80] Tatum, C.B. (1987)., "Improving Constructibility During Conceptual Planning.", Journal of Construction Engineering and Management, Vol.113, No.2, pp.191-207. [81] Tatum, C.B. (1988)., "Classification System for Construction Technology.", Journal of Construction Engineering and Management, Vol.114, No.3, pp.344-363. [82] Tomlinson, M J . (1975)., Foundation Design and Construction, 3rd ed., U.K., London: Pitman Pub. Co. [83] Toolan, F. E., and Fox, D. A. (1977). "Geotechnical Planning of Piled foundations for Offshore Platforms.", Proceedings of Civil Engineering Institute Part-1, May 1977, p.221-244. [84] Touran, A., and Toshyuki, A. (1987)., "Simulation of Tunnelling Operations.", Journal of Construction Engineering and Management, Vol.113, No. 4, pp.554-568. Bibliography 261 [85] Trimble, G. (1987)., "Expert Systems in Construction: Their Potentials and the Methodology of Knowledge Acquisition.", Managing Construction Worldwide, Vol.1, London, U.K., Sept 1987, pp.610-619. [86] Turban, E., and Watkins, P. (1985)., "Integrations of Expert System and Decision Support System.", DSS-85, pp.52-63. [87] Touran, A. (1987)., "Integration of Expert Systems with Simulation.", Journal of Construction Engineering and Management, Vol.11, No.3, pp.480-493. [88] Wager, D.M. (1987)., "Can Expert Systems Help the Construction Industry.", Proceedings of the second International Conference in Civil and Structural Engineering Computing, Vol.2, London, U.K.,pp.387-389. [89] Warszwaski, A. (1985). "Decision Models and Expert Systems in Construction Management.", Building and Environment, Vol.20, No.4, pp.201-210. [90] Winterkorn, H.F., and Fang, H., Eds. (1975)., Foundation Engineering Handbook, New York: Van Nostrand Reinhold Co. Inc. [91] Winston, P.H. (ed.), (1986)., Artificial Intelligence, 2nd edition, Reading, Mass.: Addison-Wesley Pub. Co. [92] WEAP Program (1983)., Goble and Associates Inc., U.S. Department of Transportation., Washington D.C., U.S.A [93] Zadeh, L. A. (1975)., "Fuzzy Logic and Approximate Reasoning", Syntheses, Vol.30, pp.407-428. 26% Appendix A Earth Pressure and Moments Computations A.1 Introduction Soil i s an extremely heterogeneous material that has been clas s i f i e d under different classification systems into many types. For purposes of this thesis, s o i l types are restricted to a subset of Sand (loose sand and dense sand), and Clay (soft clay and s t i f f clay). Furthermore, s o i l p r o f i l e layering i s restricted to one or two layers for these s o i l types. Thus, 14 s o i l layering scenarios are possible, e.g., loose sand for one layer, loose sand on s t i f f clay for two layers, etc. Two major s o i l interactions relevant to the design of ground wall support structure are lateral pressure and moments. A.2 Lateral Pressure Calculations Lateral pressure is a function of s o i l type, retaining system specification (vertical and horizontal spacing, i t s type, e.g., anchorage vs. struts), and rate of construction. Conventional earth lateral pressure and moment computation methods (Peck 1969, Terzaghi and Peck 1967, and Winterkorn and Fang 1975) are available as well as sophisticated Finite Element Methods. The former are adopted for the prototype CMSA implementation as we wish to explore alternative solutions, rather then refine the design of specific GWSSs. Formulas, set out in several references (Ratay 1984, and Winterkorn and Fang 1975) were adopted for estimating earth pressure and moments. The following nomenclature i s adopted. Nomenclature (phi) = shearing resistance angle of sand in degrees K(a) = active Rankine earth pressure coefficient, dimensionless P(a) = active earth pressure, l b / f t A 2 q(u) = unconfined compressive strength of clay in ksf (kips per square foot) C = Cohesion Coefficient for Clay ra(w) = water unit weight in l b / f t A 3 ra(wet) = wet unit weight in l b / f t A 3 ra(Subm) = submerged unit weight in l b / f t A 3 Appendix A. Earth Pressure and Moments Computations 263 ra(c) = unit wet weight of clay in l b / f t A 3 ra(s) = unit wet weight of sand in l b / f t A 3 S = Section Modulus for structural members, i n A 3 . H = depth of excavation in f t H(sp) = vertical spacing between the supporting struts in f t H(w) = water level in f t q = average surcharge pressure, l b / f t A 2 SPT = Standard Penetration Test, dimensionless. Soil properties adopted for the CMSA prototype are shown in table A . l . SoilType ra ra phi q(u) SPT (wet) (subm) 1. Loose Sand 100 58 0.70 4-10 2. Dense Sand 129 70 2.6 30-50 3. Soft Clay 110 60 30 2-4 4. StiffClay 130 70 40 8-15 Table A.l Soil Types Properties Employed in CMSA Next, calculations for pressures and moments envelopes w i l l be derived for one layer and two layer scenarios. Appendix A. Earth Pressure and Moments Computations 264 Surcharge Time* mi FORM Pre* (a) (C) (b) PraMtir* Diagram* Surcharga (d) Taal Praauit (e) Figure A.l Pressure and Moments Envelopes (from Winterkorn and Fang 1975) Figure A.l (a) shows a sketch of a retaining system along with the system reactions (struts/anchors) and pressure envelopes. The pressure is decomposed into s o i l pressure (b), water (c), and surcharge (d), and their sum i s shown in figure (e) . Figures A.l (f) and (g) represent moment diagrams that correspond to total s o i l profile pressure for simply supported beam (hinged) and continuous (fixed at each end) . For the CMSA prototype, the surcharge pressure was not considered. Next, lateral pressure and moment relation for one and two layer s o i l models are presented. A.2.1 One Soil Layer Case (a): When Soil i s Loose Sand/Dense Sand Using the Terzaghi-Peck approach (1967), the unit pressure for sand i s given by (Winterkorn and Fang 1975): P(Sand) 0.*65 * K(a) * ra(sand) * H (A.l) Appendix A. Earth Pressure and Moments Computations 265 where ra(Sand) = dry unit weight for the sand under water table, and wet unit weight above water table K(a) = (tan (45 - (phi))) A2 (A.2) For temporary structures design, only active Rankine pressure forces have been considered, i.e., passive Rankine pressure i s ignored. Rankine Pressure coefficients are computed using Coulomb pressure envelopes which do not account for wall s o i l angle f r i c t i o n . Maximum unit pressure for water, P(Water) = ra(water) * H(w) (A.3) Therefore, maximum total pressure i s P(Total) = P(Sand) + P(Water) (A.4) Static water conditions are assumed (no upflow into the bottom of excavation). Maximum moments (assume uniformly distributed load) a) for continuous spans, M(max) = P * H(sp) A2 / 10 (A.5) where, H(max) = Maximum vertical spacing between the struts (15 f t is the vertical spacing adopted for CMSA). b) for simple spans: M(max) = P * H(sp) A2 / 8 (A.6) For CMSA, equations A.3 and A.5 w i l l be adopted with default retaining system spacing of 12 f t horizontally and 15 f t verti c a l l y in Ratay (1984), horizontal spacing varies from 8 f t to 12 ft) . Case (b): When Soil i s Soft Clay/Stiff Clay (Terzaghi-Peck (1967)) For cohesive soils, the water table should be assumed at the s o i l surface, and total stresses are computed with (phi) = 0. This applies to short term excavation open for an extended period without standing water. Appendix A. Earth Pressure and Moments Computations 266 For Soft Clay P = K(a) * (ra) * H (A.7) For S t i f f Clay P = F * (ra) * H (A.8) where F =0.30 ra(Clay) = saturated unit weight for the clay under the water table and wet unit weight above the water table. A.2.2 Two Soil Layers For two or more s o i l layers (see figure A.2), Peck (1943) proposed the following unit pressure for excavations in layered soils (sand and clay), with sand overlaying clay: P(m) = K'(a) * ra' * H + ra(w) * (H - H(w)) (A.9) where K'(a) = 1 - 2 * q' (u) / (ra' * H) (A.10) and q'(u) i s the average compressive strength for sand and clay, and ra' i s the average unit weight for sand and clay. q'(u) = (1/H) * [ra(s) * H(s) A2 * K(s) * tan(phi) + (H-H(s) * n * q(u)] (A.11) The s subscript refers to sand stratum and n = ratio of f i e l d unconfined compressive strength to that determined in the laboratory. Also, ra' = (1/H) *[ra(s) * H(s) + ra(c) * H(c)] (A.12) in which c subscript refers to the clay stratum. Pressure relationships (and pressure diagrams) for design of retaining elements in mixed soils which have values of c and phi have not been developed (Winterkorn and Fang 197 5). Conventional techniques may be employed for design of bracing elements in shallow excavations. Lambe (Winterkorn and Fang 1975) has reviewed various methods of determining pressures in excavations and s o i l movements and has compared the results with data from actual excavations. It was shown that movements of the s o i l outside of the excavation and strut loads cannot be adequately predicted under most f i e l d conditions. Appendix A. Earth Pressure and Moments Computations 267 Ha He Sand: K-1 (phf)« **(•) Clay: q(u) n(c) Earth Pressure H Figure A.2 Soil Profile for Two Soils Scenario [from Winterkorn and Fang 1975] A3 Design Principles for Structural Members A.3.1 Structural Member Design (Wall Members) This section describes design principles for sizing steel sheet piles and soldier piles. 1. Steel Sheet Piles a) For continuous spans the maximum moment for 3 or more spans i s equal: M(max) = P * (H(sp) A 2)/l0 b) For simple spans: M(max) = P * (H(sp)A2)/8 (S) = M(max)/ (1.5 * F(b)) (A.5) (A.6) S i s the required section modulus, 1.5 i s the overstress factor for short term loading, and F(b) i s the allowable steel working stress which i s assumed to be 25 ksf for A328 Steel. Appendix A. Earth Pressure and Moments Computations 268 Example: Size a steel sheet p i l e for an excavation for 45 f t deep excavation of loose sand. Retaining system spacing i s 15 ver t i c a l l y . The water table elevation i s 10 f t below the surface. From equation A . l , maximum Pressure for Sand 0.65 * K(a) * Subm. Unit Weight for loose sand * Excavation depth (H+5) 0.65 * 0.33 * 0.058 * 50 = 0.62 ksf The depth of sheeting below excavation level i s assumed to be 5 f t as a default in CMSA to prevent or minimize s o i l heave or boiling. For more accurate results, a t r i a l and error procedure to find i t s value can be found in foundation texts (Winterkorn and Fang 1975, Ratay 1984) where a 20 to 40 % i s typically added to the calculated length as a safety factor in some cases. From equation A.2, maximum Pressure for Water ra(water) * (H - 10) 0.062 * 40 = 2.48 ksf Total Pressure (P) = 0.62 + 2.48 = 3.10 ksf Assuming hinges between the lower two struts (simple supported span assumption), the maximum moment, using equation A.6 i s Moment = P * H(max)A2/ 8 = 3.10 * 15A2/ 8 = 87.19 kips.ft Then from equation A.9, the required section modulus S = Moment*12/(working stress * overstress factor(1.5)) 27.9 i n A 3 / f t of wall From the ASTM A328 data base lookup, a section modulus of 27.9 i n A 3 / f t f a l l s between PZ_27 sheet (S = 45.30 in A3) and PDA_27 sheet (S = 14.30 in A3). The f i r s t one w i l l be selected and tried. Note that i f tie-back anchors were used, the p i l i n g would have to be sized for the downward force component of the anchors a combined stress situation. 2. Soldier Piles Similarly, soldier piles could be designed for 12 f t wide strip and 15 f t (strut spacing) high for earth pressure. Appendix A. Earth Pressure and Moments Computations 269 A3.2 Retaining System Members 1. Wales The procedure for sizing wales i s : 1. Compute bending moment; 2. Pick t r i a l section and compute bending stress; 3. Determine axial stress (wales may be axially loaded due to deflection of end retaining walls) and slenderness ratio (Kl/r) for both axes; 4. Select F(a) from AISC Column Tables; 5. Check effect of combined stress [Sect. 1.6.1-AISC]; and 6. Repeat steps 2 through 5 i f necessary. 2. Struts To compute the reaction on a strut, multiply the lateral pressure by the load carrying area (12 f t by 15 f t spacing of struts): 1. Find the maximum reaction force on strut; 2. Try a WF section and determine slenderness ratio (1/r) where r = radius of gyration, and 1 = the unbraced strut length; 3. Determine the allowable stress (Fa) from current AISC Column Tables; and 4. Repeat steps 2 and 3 until strut capacity > = strut load For maximum moment calculations, struts are assumed to be simply supported and axially loaded. Struts or tie-back reactions are calculated assuming each support carries the load developed over half the distance to the next retaining reaction force position. A33 Lagging Lumber may be surfaced on one side (SIS) , two sides (S2S) , two edges (S2E), a combination of sides and edges (SISIE, SIS2E, S2SE) , or on a l l four sides (S4S, Hurd 1961). the length of lumber ordered, when this can be specified, should be such that i t can be used to the best advantage with l i t t l e waste. Most common lengths commercially available range up to 20 -24 f t in even-numbered increments of 2 f t . Lagging Design: First, arching is assumed to take place in granular soils and s t i f f to hard cohesive s o i l s . Arching may not take place in soft cohesive s o i l , where the overburden pressure approaches four times the cohesive strength of the s o i l . The arching effect occurs when s o i l pressure causes lagging to deflect (timber overstress) and Appendix A. Earth Pressure and Moments Computations 270 thus decrease overall lateral pressure. Therefore, arching may only be assumed under special conditions and for these f u l l active lateral pressure must be accounted for. For CMSA, the maximum moment for lagging i s computed as (p*lA2/12) when arching i s present (Ratay 1984). Three c r i t e r i a are used to check timber lagging design: maximum moments allowable, maximum deflection, and maximum shear. For the CMSA prototype, the f i r s t one was chosen as the criterion to select a section for lagging. Moments (M) = f * S (A. 13) where M i s maximum moments (Resisting moments), f i s allowable stress in extreme fiber, S i s section modulus (b * H A2). There are several commercial designations and species that are cl a s s i f i e d into groups i s reported in CSA (CSA Standard 085-1976) . For the sake of the prototype application, we adopt the Douglas-fir-larch Species Identification, and Select Structural grade (allowable bending stress of 1850 psi (see pp. 31 Table-7). Therefore, applying following formula: M = f * S where f M M S 1850 psi, P * LA2/8 for no arching, or P * LA2/12 for Arching b * hA2/6 For 1 = Unbraced Lagging Length = 12 f t (horizontal spacing of struts) So, for b = 4 in Nominal (3 1/2 effective), S b * HA2/6 = b * hA2 b * hA2 For b = M/f P * LA2/8/1850 P * 12A2 * 6/ (8*1850) P * 0.058 3.5 in, h = sqrt(P * 0.058/1.5). A data base for the lagging members used in the prototype i s shown in table A.2. Appendix A. Earth Pressure and Moments Computations 271 Member Designation Section Modulus (b*hA2) (inA3) 2 4 3.06 3 4 5.10 4 4 7.15 4 6 17.65 6 6 27.7 Table A.2 Lagging Members Appendix B Pile Driving Production Derivation B.l Introduction This appendix i s divided into two parts. The f i r s t one reviews basic methods and formulas for calculating s o i l resistance to pile driving. A formula for total driving time as implemented in CMSA i s derived. The second part u t i l i z e s dynamic formulas for equating theoretical hammer energy with s o i l resistance in order to estimate the speed of p i l e driving. It should be mentioned that the literature does not contain consistent relationships among s o i l , p i l e and hammer properties. The essence here i s to adopt some of those formulas to develop our model for predicting p i l e driving progress rate. The adequacy, validity, and verification for these relationships are beyond the scope of this thesis. Several methods exist for s o i l resistance calculations in relation to pile driving. They include Cone Landa, API (Toolan and Fox 1977), Cone Penetration Test (CPT), Standard Penetration Test (SPT) (Tomlinson 1975), and/or use of ultimate (upper bounds) unit skin and end bearing f r i c t i o n (Peurifoy 1970). Soil, p i l e , and hammer interactions significantly govern the pace of pi l e driving. We start with a review of skin and end bearing formulas. SPT in conjunction with the ultimate bounds for s o i l resistance for the CMSA prototype. B.2 Soil/Pile Friction Calculations The following nomenclature is used. Q(b) Q(s) Q(u) A(p) A(s) P(d) N(q) K(s) SF p i l e end bearing resistance, l b / f t A 2 p i l e skin f r i c t i o n resistance, l b / f t A 2 total p i l e f r i c t i o n resistance, l b / f t A 2 p i l e cross section area, f t A 2 surface area for the pi l e shaft, f t A 2 earth pressure at the ti p of the p i l e , l b / f t A 2 coefficient for computing the skin f r i c t i o n , earth pressure coefficient. Skin Friction, lb Appendix B. Pile Driving Production Derivation 273 The basis of the "static" or s o i l mechanics method of calculating the ultimate carrying capacity of a p i l e i s that the ultimate carrying capacity i s equal to the sum of the ultimate resistance of the base of the p i l e and the ultimate skin f r i c t i o n over the embedded shaft length of the pi l e , i.e.: Q(U) = (Q(b) + Q(s))/2.5 (B.l) where Q(b) = base resistance, Q(s) = shaft resistance, and 2.5 i s a safety factor. Knowing the angle of shearing resistance of the s o i l at the base level, Q(b) can be calculated for different s o i l conditions as described below. B.2.1 Sands Total Base Resistance = Q(b) = A(p) * P(d) * (N(q) - 1) (B.2) where A(p) = Cross section area of the p i l e P(d) = Effective Overburden Pressure at p i l e base level N(q) = Bearing Capacity Factor that i s relevant to angle of shearing resistance of sand and SPT in N blows/300 mm. Total Skin Friction = SF . = K(s) * P(d) * tan (delta) (B.3) where K(s) = Earth Pressure Coefficient Delta = Angle of Wall Friction Therefore, SF force i s = Unit Skin Friction * Average effective pressure * Surface Area = K(s) * Avg. P(d) * tan (Delta) * A(s) where Appendix B. Pile Driving Production Derivation 274 Avg. P(d) i s the average effective overburden pressure over the embedded depth of pile, and A(s) i s the embedded surface area. Broms (Tomlinson 1975) has related values of K(s) and delta to the effective angle of shearing resistance (phi) for various p i l e materials and relative densities as shown in table B.l below: Pile Delta K(s) Material Low Density High Density Low Density High Density (Loose Sand) (Dense Sand) (Loose Sand) (Dense Sand) Steel 20 0.5 1.0 1.0 Table B.l Values for Angle of Internal Friction [from Bowels 1977] Unit skin f r i c t i o n , in a uniform cohesionless s o i l , increases linearly with increasing depth. A peak value of 2.3 0 ksf for unit skin f r i c t i o n i s used for straight-sided piles. Where piles are driven deeper than 2 0 times their diameters, rule of thumb values can be used for the average skin f r i c t i o n over the whole shaft. Table B.2 relates the sand s o i l density to an upper bound of skin f r i c t i o n based on experimentation. Relative Density Average Skin Friction < 0.35 (loose) 0.209 ksf 0.35 - 0.65 (medium dense) 0.21 - 0.52 ksf 0.65 - 0.85 (dense) 0.52 -1.46 ksf > 0.85 (Very dense) 1.46 ksf-2.30 ksf Table B.2 Ultimate Skin Friction for Sands [from Bowels 1977] B.2.2 Clays and S i l t s End Resistance: Qb = Nc * Cb * Ab where (B.4) Appendix B. Pile Driving Production Derivation 275 Nc = 9 i f p i l e driven more than 5 diameters Cb = undisturbed shear strength Ab = Cross section area Skin Resistance (1): Qs = alpha * Cd * As (B.5) where alpha = adhesion factor. It varies for piles of the same type on the same site between 0.4 to 1.0 for clays of 1.50 ksf shear strength, and between 0.25 and 0.45 for clays of 2.58 to 3.1 ksf shear strength (Ratay 1984). Cd = undisturbed shear strength adjacent to the shaft, As = surface area of the shaft Meyerhof (Tomlinson 1975) states that a maximum skin f r i c t i o n of 1.13 ksf (calculated on a l l surfaces of flanges and web) should be used for H piles. Factors such as pore pressure (water may work as lubricant), ground heave (and re-consolidation), lateral vibration, smearing, and "strain-softening" are of significance but may vary in i t s effect even on the same site for adhesion computation for clays. Ultimate loads on piles driven into cohesive soils are as follow: For end resistance, Q(b) = N(C) * C(b) * A(b) (B.6) where N(c) = bearing capacity factor C(b) = Shear strength at the base of the pi l e A(b) = Cross Section Area of the pi l e t i p . For skin f r i c t i o n , Q(s) = alpha * Avg. C(b) * A(s) (B.7) Appendix B. Pile Driving Production Derivation 2 7 6 adhesion factor. average undisturbed shear strength of the clay adjacent to the shaft Surface area of the shaft. For uniform clays, the shear strength increases with depth. The average value of shear strength over the whole shaft length i s taken for C(b). When clays exist in layers, e.g. soft clay over s t i f f clay, the skin f r i c t i o n i s computed for each layer using the corresponding adhesion factor appropriate to the shear strength and overburden conditions. B.2.3 Time Required for Driving a Sheet Pile A relationship i s required to estimate production rate of pil e driving as a function of soil/pile/hammer interaction. Dynamic formulas and more exact methods such as Weave Equation Analysis for piles that f i t a class of impact hammers have been developed (Peurifoy 1 9 7 0 , WEAP 1 9 8 3 ) . No relationships are available so far for vibratory hammers. Suppose that a soil/hammer/pile scenario i s given and we want to know production rate for a single steel sheet p i l e (or segments of piles) to be driven to refusal depth. Driving time i s divided into two components: variable time and fixed time. The f i r s t counts for the time when the pile driver i s in driving mode. The latter counts for the time when the crane i s used to seat the pile , the crew i s f i t t i n g the hammer on the pile , the crane moves from one place to another, and so forth. Thus, Production Rate per Single Steel Sheet Pile or Soldier Pile = Fixed Time + Variable Time (B.8) This i s based on the assumption that the crane i s dedicated to driving one sheet pile at a time. That implies that both hammer and crane are tied up together. Often the driving pattern follows a wave. Adjacent piles are driven for short distances and at equal intervals and alignments. Fixed Time = Grabbing + Raise + Positioning + Seating + Welding (function of i : 3Adhesion factors are calculated for three scenarios (Winterkorn and Fang 1 9 7 5 ) : 1 . For sands or sand gravels over s t i f f clays; 2 . for soft clay over s t i f f clays; and 3. for s t i f f clay. alpha 1 3 where, lp Avg.C(b) A(s) Appendix B. Pile Driving Production Derivation 277 Number of vertical pieces of installed sheet pile) Average Fixed Time per One Pile = n * (Grab + Raise + Position) + (n - 1) * (Welding + Seating) (B.9) in which n i s the number of segmented or spliced p i l e lengths. Segments of steel sheet piles may come in different lengths of 40 f t , 50 f t , 60 f t up to 80 f t depending on mode of transportation, depth of excavation, project site restriction, crane and hammer available, and existing p i l e stock for the contractor. Therefore, n = Excavation Depth / Length of Sheet Piles) - 1 ( i f f i r s t term in not integer) Total Number of Pile Segments = l c * Number of Pile Driving Cycles. ... (B.10) where l c i s the pile segment length. Variable Time per One Cycle of Pile Driving = Production Time (Driving Mode) = Function of (pile segment length, s o i l condition, Tunnel Depth, sheet pi l e type, Pile Driver type) ... (B.ll) The derivation for this component i s treated in the next section where dynamic formulas are employed. Number of Pile Driving Cycles = Tunnel Length * 2 * 12 / One Pile Installation Width (Pile Width) ... (B.12) where One Pile Installation Width denotes one unit width of pil e installation, i.e., there might be (n) pi l e segments (say 2 to 3 1/2 segments) installed for one unit wall width (say one foot) for one face of the excavation (2 for two faces, and 12 to convert feet into inches when the pi l e width i s specified in inches). Total Pile Driving Activity Time = ((Fixed Time per One Cycle) + (Variable Time per One Cycle)) * Number of Pile Driving Cycles. Assuming 8 working hours per days (could be variable i f desired), then Appendix B. Pile Driving Production Derivation 278 Number of Working Days = (Total Time / 8) days (B.13) B3 Pile Driving Production Rate Estimation B.3.1 Pile Driving Production Formula Derivation The author developed the production rate formula from consideration of f i r s t principles. With minor variations, a similar analysis was described in Gates and Scarpa (1984). The following nomenclature i s used: R = Soil Resistance force for Piles, lbs S — Average penetration per blow (Set) for last 5 or 10 blows, in/blow f — Number of hammer blows/foot driving increment, E — manufacturer's maximum rated energy per blow, ft-lbs w(r) — weight of ram, lbs W(p) — weight of p i l e , including driving - appurtenances, lbs e — coefficient of restitution H height of hammer drop, f t R = Skin Friction + End Bearing (B.14) For non-displacement piles of steel sheet piles and soldiers piles, end bearing can be ignored. Therefore, R = Skin Friction (B.15) In theory, theoretical energy delivered by the impact hammer is equated to s o i l resistance during p i l e driving. The relationships expressed by dynamic formulas, empirically derived, relate displacement of sheet p i l e (S) to hammer delivered energy and s o i l resistance. For each class and type of impact hammer, there i s a dynamic formula. For a drop hammer, the formula i s R = 2 * W * H + 0.10 (B.16) For single acting air/steam hammer, the dynamic formula i s R = 2 * W * H / S , S(i) (2 * W * H / R(i)) - 1 (B.17) Appendix B. Pile Driving Production Derivation 279 in which R(i) and S(i) are incremental s o i l resistance and corresponding p i l e penetration rate per unit depth (in ft) . Equations B.16 and B.17 are based on Engineering News Formula (Peurifoy 1970). An alternate dynamic formula for drop hammer (as well as other hammers) based on Highway Department Modified Engineering News Formula (Peurifoy 1970) i s : R = (2.5 * E(same as W*H)) * (W(r) + e**2 * W(p)) / (W(r) + W(p)) (B.18) in which E = W * H (potential energy with no losses) . Similarly, the Modified Highway Department Modified Engineering News Formula for single acting air/steam hammer i s : R(s) = ((2*(W+A(p))* H)/ (S +.10 * P/W) (B.19) where, A(p) = Area of the piston of the ram P = Air or steam pressure on the piston. The hammer theoretical energy, E = W * H, i s used in the equations. The net hammer energy could be less (80%-2 0%) than the theoretical energy due to hammer energy losses (impact, f r i c t i o n , etc.) and hammer efficiency. Many dynamic formulas have been developed. Recently, Wave Equation Analysis for Piles (WEAP) has emerged as more accurate than dynamic formulas (WEAP 1984). However, WEAP programs are quite big, complex, restricted to a class of diesel hammer, and requires an intelligent front end and back end for automation. Therefore, for the CMSA prototype, the Engineering News dynamic formulas were adopted for ease of understanding, coding, and interfacing. An assumption adopted i s that there are no energy losses, and the hammer efficiency i s specified by the user. Thus, the basic form of the equation adopted for predicting R, the s o i l resistance for piles i s of the form: R = Unit Skin f r i c t i o n (at depth z) * Pile Surface Area (B.20) R can be approximated using interpolation at different depths, R at depth z, using a static formula i s : Appendix B. Pile Driving Production Derivation 280 R = Skin f r i c t i o n (at z depth) * SSP-surface area/12, R(l) = Skin f r i c t i o n (at depth z(l)) * SSP-Surf. area/12, R(i) = SSP-surface area (constant) * SF (at z ( i ) ) , R(i) = f (depth z(i)) (B.21) SSP-surface area corresponds to the surface area of any non-displacement piles (e.g. steel sheet piles, H or soldier p i l e s ) . Skin f r i c t i o n increases with depth and thus rate of change can be emulated as a step function or linear relationship versus depth. For CMSA, i t i s assumed to be linear and the analysis i s performed in 1 foot increments. R at a depth interval of every x feet (say one foot) , and using equation B.17 i s : S(i) = (2*W*H/R(i)) - 1.0 (B.22) For each foot depth interval ( z ( i ) - l ) , duration (T(i)) for driving p i l e one foot form the surface can be approximated as follows: T (i) = (x foot * 12 in/foot) / ((S(i) in inches/blow) * f (blows/minutes)) in minutes (B.23) Therefore, the total variable time per p i l e driving cycle i s : Variable Time = Sum (T(i)) (B.24) = f(R(i), S ( i ) , d(z)) (B.25) Once the variable time component value i s computed in the Drive.C routine, i t i s passed to the CMSA prototype which then adds i t to the fixed time component to obtain the total p i l e driving time (see figure 4.10 of Drive.c routine interface with CMSA). To show the relationship between the number of blows i t takes the a hammer to drive a pi l e 1 f t and the s o i l resistance at time of driving (SRD), figure B.l (a) presents such a profile for a displacement p i l e . Figure B.l (b) shows the expected blow count/depth relationship for a particular combination of driving plant for predicting i n i t i a l penetration of jackets below the sea bed in normally consolidated clays (Toolan and Fox 1977). These curves are predicted based on semi-empirical equations which are then considered during wave equation analysis computation. Although figure B.l i s for an offshore displacement pil e , there i s a parallel for non-displacement piles installation Appendix B. Pile Driving Production Derivation 281 behavior. For instance, hammer efficiency influences hammer performance where hammer blow count varies with driving depth and s o i l resistance. 100 110 120 130 "2 i« n 0) 160 a 0) 180 O 170 3 180 X {ll90 0 • 200 210 220 230 240 20 Blow count: blows/ft *o , ao ao ioo 120 Juriiigeo out IrM 18 I Jf« ! 0.7 \ *< v Ondrtv fig a tort IIIM tup Loc •Don PB Me ick7 100 | t ^ . n a B irlni een Ine J mil avli uudf 1 0.78' X A n I rlrt ring i V > V c 1 C jrlni In OOfl BW» Umi ' m i rMn •r ) y 1,7a c i d * Inq i Iter full* / l itnjo ! Ho KO. d0. h» imei lendi a ! (a) Blow Count/Resistance Curves 9000 a 8 0 0 0 2 £ 7000 I 6000 •a | 5000 1 3000 2000 0 0.75 0.65 < / Pile Driving Information Pte Penetration Onvtty Coitneetom PI l« Segment Length Pile tip In Hammer Cushion Spring constant anvil Coer. of mtttutlon Total iMa nalatanoa total point resistsnoe Computarnm : 54 In. dta. open pipe* : 208 It (83m) :2by13ft :3m : Sand : Menck 7000.1 segment : 12* hard wood :17S00n/cm :0.7B : aS8.0.78 : 1875 Hps (85011) : varying : 750308/3 40 80 120 160 200 Blow count: blows/ft 240 280 (b) Predicted Blow count/Depth, 54 in dia. Piles Figure B.l Hammer Blow Count versus Soil Resistance [from Toolan and Fox 1977] Appendix B. Pile Driving Production Derivation 282 B . 3 . 2 Example A proposed tunnel of 12 00 f t long, 66 f t deep, 20 f t wide i s to be constructed in a remote area. The s o i l profile consists of two layers: 40 feet of loose sand layer on top of s t i f f clay. After running CMSA, a steel sheet pile alternative was selected for consideration. Piles properties relevant to the Drive.c routine are: - Sheet Pile Designation i s PZ_27 (ASTM A36) - PZ_27 Cross Section Area i s 11.91 in A2, and - PZ_27 Surface Area i s 4.94 f t A 2 / f t (SA). Other properties such as cross section area are used for setting an upper bound for hammer energy without damaging the p i l e while driving. As well, i t i s used in another variation of the Drive routine when SPT is used to interpret the s o i l properties. CMSA, based on project context, further specifies a hammer u t i l i z i n g i t s construction and geotechnical knowledge. Properties that apply to the Drive.c routine are: - Hammer_Type is Double Acting Air Hammer (DAAH), - Hammer Frequency is 95 cycles/min, and - Rated Hammer Energy is 11,490 l b - f t . Other properties, such as hammer ram weight are used to indirectly select a suitable crane and recommend a pile segment length, i f required. In order to pass the parameters to Drive.c routine, CMSA creates two NExpert f l a t (ASCII files) data bases. The f i r s t one contains the s o i l profile and is called "Soil.nxp" and has this format: \ Soil_type_l\ "Loose_Sand" \ S t a r t _ l \ "1" \ F i n i s h _ l \ "40" \ Soil_type_2\ "Stiff_Clay" \ Start_2\ "41" \ Finish_2\ "71" /* start p i l e driving at depth 1 f t */ /* S t i f f Clay starts at depth 41 f t */ /* Refusal depth i s 71 f t . * / Appendix B. Pile Driving Production Derivation 283 The second f i l e , "Hammer.nxp", includes the hammer and sheet pi l e relevant properties. \ Refusal_Depth \ "71" \ Frequency \ "95" \ Energy \ "11490" \ Surface area \ "4.94" /* f t A 2 / f t */ \ Hammer_Type\ "DAAH" \ Hammer_Model\ "Vulcan of 11_3B" ********************** Note that the sheet p i l e should be driven 5 f t below excavation level (5 + 66 = 71 f t deep) to prevent or minimize s o i l heave for clays and s o i l boiling for sands. There are two output f i l e s of "Out.nxp" and "Out.out". The f i r s t one i s retrieved by CMSA to interpret and analyze selected variable values. The second f i l e , "Out.out" contain more details about the same run session. A sample of "Out.nxp" i s shown below: \Incremental_Depth.amount\="1.00" \Depth.amount\="71.00" \S o il_Res istance.amount\="33188.82" \Penetration.amount\="0.8039" \Incremental_Time.amount\="0.3 0" \Variable_Time.amount\="8.49" /* Cumulative variable Time */ \Technical_Feasibility.State\="True" ***************** A sample of "Out.out" results i s shown in table B.3. Figure B.2. presents the selected hammer (DAAH, efficiency 71%) blow count versus driving depth. Appendix B. Pile Driving Production Derivation 284 Mouth (It) S o i l Type 1 2 3 4 5 6 7 8 9 L o o s e S a n d L o o s e ' S a n d L o o s e ~ S a n d L o o s e ' S a n d L o o s e ~ S a n d L o o s e ~ S a n d L o o s e ' S a n d L o o 8 e ~ 8 a n d L o o s e ' S a n d 1 0 L o o s e ' S a n d 1 1 L o o s e ~ 8 a n d L o o s e ~ 8 a n d L o o s e ~ S a n d L o o s e ~ S a n d L o o s e ~ 8 a n d L o o s e ' S a n d L o o s e ' S a n d L o o s e ~ S a n d L o o s e ' s a n d L o o s e - S a n d L o o s e ~ S a n d L o o s e ' S a n d L o o s e ' S a n d L o o s e ' S a n d L o o s e " ~ 8 a n d L o o s e ' S a n d L o o s e ' S a n d L o o s e - S a n d L o o s e - S a n d L o o s a ' S a n d L o o s e - S a n d L o o s e ~ S a n d L o o s e ~ S a n d , L o o s e ' S a n d 3 5 L o o s e ' S a n d 3 6 L o o s e - S a n d L o o s e ~ S a n d L o o s e ' S a n d L o o s o - S a n d L o o s e - S a n d 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 2 4 2 5 2 6 2 7 2 8 2 9 3 0 3 1 3 2 3 3 3 4 3 7 3 8 3 9 4 0 . 4 0 S t 4 1 S t 4 2 S t 4 3 S t 4 4 S t S o i l lies, (lb) 1 4 . 5 2 4 3 . 5 8 8 7 . 1 5 1 4 5 . 2 5 2 1 7 . 8 8 3 0 5 . 0 2 4 0 6 . 7 0 5 2 2 . 9 0 6 5 3 . 6 3 7 9 8 . 8 8 9 5 8 . 6 5 1 1 1 2 . 9 5 1 3 2 1 . 7 8 1 5 2 5 . 1 3 1 7 4 3 . 0 0 1 9 7 5 . 4 0 2 2 2 2 . 3 2 2 4 8 3 . 7 7 2 7 5 9 . 7 5 3 0 5 0 . 2 5 3 3 4 2 . 3 9 3 6 3 6 . 1 0 3 9 3 1 . 3 6 4 2 2 8 . 1 9 4 5 2 6 . 5 7 4 8 2 6 . 5 2 5 1 2 8 . 0 2 5 4 3 1 . 0 9 5 7 3 5 . 7 1 6 0 4 1 . 9 0 6 3 4 9 . 6 5 6 6 5 8 . 9 6 6 9 6 9 . 8 3 7 2 8 2 . 2 5 7 5 9 6 . 2 4 7 9 1 1 . 7 9 8 2 2 8 . 9 0 8 5 4 7 . 5 7 8 8 6 7 . 8 0 9 1 8 9 . 5 9 9 5 4 3 . 0 1 9 8 9 7 . 4 6 1 0 2 5 2 . 9 5 1 0 6 0 9 . 4 7 1 0 9 6 7 . 0 4 1 1 3 2 5 . 6 4 1 1 6 8 5 . 2 8 1 2 0 4 5 . 9 5 1 2 4 0 7 . 6 7 1 2 7 7 0 . 4 2 1 3 1 3 4 . 2 1 1 3 4 9 9 . 0 4 1 3 8 6 4 . 9 0 1 4 2 3 1 . 8 0 1 4 5 9 9 . 7 4 1 4 9 6 8 . 7 2 1 5 3 3 8 . 7 3 1 5 7 0 9 . 7 8 1 6 0 8 1 . 8 7 1 6 4 5 5 . 0 0 1 6 8 2 9 . 1 7 1 7 2 0 4 . 3 7 1 7 5 8 0 . 6 1 1 7 9 5 7 . 8 8 1 8 3 3 6 . 2 0 1 8 7 1 5 . 5 5 1 9 0 9 5 . 9 4 1 9 4 7 7 . 3 7 1 9 8 5 9 . 8 3 2 0 2 4 3 . 3 4 2 0 6 2 7 . 8 8 2 1 0 1 3 . 4 5 Penetration (in/blow) 1 8 7 2 , 6 2 3 , 3 1 1 , 1 8 7 , 1 2 4 , 8 9 . 6 6 , 5 1 , 4 1 , 3 3 , 2 8 . 2 3 . 2 0 . 1 7 . 1 5 . 1 3 . 1 2 . 1 0 . 9 . 8 . 8 . 7 . 6 . 6 . 5 . 5 . 5 . 4. 4 . 4. 4 . 3 . 0 5 1 5 9 5 0 5 9 2 5 2 1 1 5 1 7 1 0 1 0 5 0 1 7 6 2 6 9 0 4 2 5 0 3 4 9 3 9 1 2 6 5 9 9 0 1 9 4 7 3 1 7 3 0 0 5 0 1 3 6 6 5 8 1 3 6 3 8 4 8 3 7 5 3 4 8 1 5 0 0 3 5 8 3 7 8 6 8 1 6 9 3 3 1 4 9 0 7 4 5 3 4 1 2 0 2 8 9 0 6 9 6 4 1 0 4 0 0 7 1 8 2 6 9 8 3 7 8 0 1 5 6 3 4 1 4 7 9 8 3 3 7 0 2 0 4 6 0 8 1 4 9 6 6 5 8 5 9 1 7 4 9 5 6 4 7 5 5 5 2 2 4 6 3 1 3 7 9 5 3 0 1 0 2 2 7 1 1 5 7 4 0 9 1 6 0 2 9 4 9 7 0 4 9 1 4 4 8 6 1 3 8 1 0 7 7 6 2 6 7 1 6 7 1 . 6 7 2 8 1 . 6 3 1 0 1 . 5 9 0 9 1 . 5 5 2 6 1 . 5 1 5 8 1 . 4 8 0 6 1 . 4 4 6 8 1 . 4 1 4 3 1 . 3 8 3 0 1 . 3 5 3 0 1 . 3 2 4 0 1 . 2 9 6 1 1 . 2 6 9 2 1 . 2 4 3 3 1 . 2 1 8 3 1 . 1 9 4 1 Inc. Time (min) 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 2 0 . 0 2 0 . 0 2 0 . 0 2 0 . 0 2 0 . 0 2 0 . 0 2 0 . 0 3 0 . 0 3 0 . 0 3 0 . 0 3 0 . 0 3 0 . 0 3 0 . 0 3 0 . 0 4 0 . 0 4 0 . 0 4 0 . 0 4 0 . 0 4 0 . 0 4 0 . 0 5 0 . 0 5 0 . 0 5 0 . 0 5 0 . 0 5 0 . 0 5 0 . 0 6 0 . 0 6 0 . 0 6 0 . 0 6 0 . 0 6 0 . 0 7 0 . 0 7 0 . 0 7 0 . 0 7 0 . 0 7 0 . 0 8 0 . 0 8 0 . 0 8 . 0 8 . 0 8 . 0 9 To til 1 1 Time Aco Skin l-'iictii Uuw Uiow (min) (lb/ft*£t) 0 9 0 9 0 9 0 . 0 9 0 . 1 0 0 . 1 0 0 . 1 0 0 . 1 0 0 . 1 0 0 . 1 1 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 0 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 1 0 . 0 2 0 . 0 2 0 . 0 3 0 . 0 4 0 . 0 5 0 . 0 6 0 . 0 7 0 . 0 8 0 . 0 9 0 . 1 0 0 . 1 2 0 . 1 4 0 . 1 6 0 . 1 8 0 . 2 0 0 . 2 2 0 . 2 4 0 . 2 7 0 . 3 0 0 . 3 3 0 . 3 6 0 . 3 9 0 . 4 2 0 . 4 6 0 . 4 9 0 . 5 3 0 . 5 7 0 . 6 1 0 . 6 5 7 0 7 4 7 9 8 4 8 9 0 . 9 4 0 0 0 6 1 2 1 8 2 4 3 0 3 7 4 4 5 1 5 8 1 . 6 5 1 . 7 3 1 . 8 0 1 . 8 8 1 . 9 6 2 . 0 5 2 . 1 3 2 . 2 2 2 . 3 1 2 . 4 0 2 . 4 9 2 . 5 9 2 . 6 9 2 . 7 9 2 . 8 9 2 . 9 9 3 . 1 0 T e c h n i c a l F e a s i b i l i t y » T r u e 1 7 . 5 0 5 2 . 5 0 1 0 5 . 0 0 1 7 5 . 0 0 2 6 2 . 5 0 3 6 7 . 5 0 4 9 0 . 0 0 6 3 0 . 0 0 7 8 7 . 5 0 9 6 2 . 5 0 1 1 5 5 . 0 0 1 3 6 5 . 0 0 1 5 9 2 . 5 0 1 8 3 7 . 5 0 2 1 0 0 . 0 0 2 3 8 0 . 0 0 2 6 7 7 . 5 0 2 9 9 2 . 5 0 3 3 2 5 . 0 0 3 6 7 5 . 0 0 4 0 2 6 . 9 8 4 3 8 0 . 8 4 4 7 3 6 . 5 8 5 0 9 4 . 2 0 5 4 5 3 . 7 0 5 8 1 5 . 0 8 6 1 7 8 . 3 4 6 5 4 3 . 4 8 6 9 1 0 . 5 0 7 2 7 9 . 4 0 7 6 5 0 . 1 8 8 0 2 2 . 8 4 8 3 9 7 . 3 8 8 7 7 3 . 8 0 9 1 5 2 . 1 0 9 5 3 2 . 2 8 9 9 1 4 . 3 4 1 0 2 9 8 . 2 8 1 0 6 8 4 . 1 0 1 1 0 7 1 . 8 0 1 1 4 9 7 . 6 0 1 1 9 2 4 . 6 5 1 2 3 5 2 . 9 5 1 2 7 8 2 . 5 0 1 3 2 1 3 . 3 0 1 3 6 4 5 . 3 5 1 4 0 7 8 . 6 5 1 4 5 1 3 . 2 0 1 4 9 4 9 . 0 0 1 5 3 8 6 . 0 5 1 5 8 2 4 . 3 5 1 6 2 6 3 . 9 0 1 6 7 0 4 . 7 0 1 7 1 4 6 . 7 5 1 7 5 9 0 . 0 5 1 8 0 3 4 . 6 0 J 8 4 8 O . 4 0 1 8 9 2 7 . 4 5 1 9 3 7 5 . 7 5 1 9 8 2 5 . 3 0 2 0 2 7 6 . 1 0 2 0 7 2 8 . 1 5 2 1 1 8 1 . 4 6 2 1 6 3 6 . 0 1 2 2 0 9 1 . 8 1 2 2 5 4 8 . 8 6 2 3 0 0 7 . 1 6 2 3 4 6 6 . 7 1 2 3 9 2 7 . 5 1 2 4 3 8 9 . 5 6 2 4 8 5 2 . 8 6 2 5 3 1 7 . 4 1 0 . 0 1 0 . 0 3 0 6 1 3 2 2 3 6 5 4 0 . 7 7 1 . 1 . 1 . 2 . 2 . 3 . 4 . . 0 6 . 4 1 :M . 9 3 . 6 0 . 3 8 5 . 2 5 6 . 2 4 7 . 3 5 8 . 5 8 9 . 9 4 1 1 . 4 3 1 3 . 0 6 1 4 . 8 2 1 6 . 7 2 1 8 . 7 5 2 0 . 9 2 2 3 . 2 2 2 5 . 6 7 2 8 . 2 5 3 0 . 9 8 3 3 . 8 5 3 6 . 8 6 4 0 . 0 2 4 3 . 3 2 4 6 . 7 7 5 0 . 3 7 5 8 . 0 0 6 2 . 0 5 6 6 . 2 5 7 0 . 6 1 7 5 . 1 4 7 9 . 8 5 8 4 . 7 2 8 9 . 7 6 9 4 . 9 8 1 0 0 . 3 6 1 0 5 . 9 3 1 1 1 . 6 6 1 1 7 . 5 8 1 2 3 . 6 7 1 2 9 . 9 3 1 3 6 . 3 8 1 4 3 . 0 1 1 4 9 . 8 2 1 5 6 . 8 1 1 6 3 . 9 8 1 7 1 . 3 4 1 7 8 . 8 8 1 8 6 . 6 1 1 9 4 . 5 3 2 0 2 . 6 3 2 1 0 . 9 3 2 1 9 . 4 1 2 2 8 . 0 9 2 3 6 . 9 6 2 4 6 . 0 2 2 5 5 . 2 8 2 6 4 . 7 3 2 7 4 . 3 8 2 8 4 . 2 3 2 9 4 . 2 8 Table B.3 "Out.out" Sample Output for Example Problem Appendix B. Pile Driving Production Derivation 285 0 20 40 60 80 100 120 140 160 Blow Count(blows/ft) Figure B.2 Hammer Blow Count versus Driving Depth Sample Run Output Variable Interpretation: The output f i l e "Out.out" contains most of the variables that are produced by the program "Drive.c" and interpreted by CMSA. The variable time, based on equation B.24, for one pi l e driving cycle i s 33.21 minutes as printed at the last driven s o i l step. Note the high speed of pi l e driving, approximately 22 ft/blow, on the top part of the s o i l layer (table B.l). The reason for this high value i s due to near zero resistance of the sand for the f i r s t driving foot resulting in very small incremental time (rounded to zero for the f i r s t four feet). Practically, the steel sheet pile by i t s own weight sinks into the sand for few feet before using the pi l e driver. So far, no adjustment has been made to account for this duration. The technical f e a s i b i l i t y state returns a Boolean value of "True" thus implies that the pi l e can be driven to refusal depth successfully without a stoppage due to high resistance or obstacle (by observing blow count during driving), and that the number of hammer blows per minute did not exceed i t s upper bound rule of thumb for p i l e damageability). From figure B.2, as the depth increases, the cumulative s o i l resistance to pi l e driving increases, and thus a higher blow count i s required to drive the pile further. The blow count i s plotted for every 10 f t of s o i l depth. In this "Drive.c" routine, the blow count i s found as the inverse of Appendix B. Pile Driving Production Derivation 286 penetration rate (in/blow) for the last inch in each driven foot which i s then multiplied by 12 i n / f t to convert blow count to blows/ft. For example, at s o i l depth 20 f t , the 20th foot blow count (in/blow) which i s divided into 12 in / f t yields 10 blow/ft. This value i s the incremental blow count at this depth. B.3.3 Drive.c Program Figure B.3 presents a flow chart for the "Drive.c" routine written in C programming language. This routine reads input data from two input f i l e s of "Hammer.nxp" and "Soil.nxp". These f i l e s are created by the CMSA prototype which contains relevant information regarding hammer/soil/pile session scenario. The routine returns output to two data f i l e s of "Out.out" and "Out.nxp". The f i r s t contains detailed output variables and the second f i l e contains selected output variables. The latter then i s retrieved by the prototype CMSA for further information processing. The routine "Drive.c" function i s to compute the cumulative variable time component based on dynamic formulas as derived in B.14 to B.25. To do so, at every foot from surface to the refusal depth during pil e driving (see table B.l), computations for the output variables are done. These variables include s o i l resistance, p i l e penetration rate, incremental and cumulative variable time. Appendix B. Pile Driving Production Derivation 287 Input RS" PropfHw Input ma •SdUnp* SoM ProAto Identify SoflTyp* ktanWyHmMrTyp* For OrMng Eich Foot InoranMnt (I), CornputK -SollRMlttano* • PanaMton Rate (Avenge Sal) - Incremental and CumuJatfva Vartatta Tlma •Skki Friction Figure B.3 Drive.c Routine Flow Chart Figure B.4 sums the progression of developing the model as finalized in equation B.15. Part B.4 (a) shows the forces applied at the p i l e in equilibrium, part (b) presents s o i l resistance, and part (c) presents penetration rate and cumulative variable driving time. If the unit depth, in part (b), i s assumed to be 1 f t , then total s o i l resistance i s computed by summing the area under the incremental resistance area. For part (c) , penetration rate i s computed at the last inch for each foot and held constant along this foot. The C source code for Drive.c i s given in l i s t i n g B.l. Defeated Hammer Enargy t Drivijig Depth | I | hcraneat Depth i Sun0Fg(As) Qp«Ap R(SoiRsstsiara>) Penotr; tion Rate rCumulative Variable Driving Time Incremental Variable * Driving Time (a) Forces on Pile (b) Soil Resistance (c) Cumulative Driving Time Appendix B. Pile Driving Production Derivation 289 #include <math.h> #include <stdio.h> #include <fcntl.h> #include <io.h> #include <float.h> #include <sys\types.h> #define BUFFSIZE 1000 #define STRSIZE 20 /* This program calculates the total time required to drive pile length without splicing. The data accepted from input file are the following: 1. L = Total depth of pile driving, ft 2. Depth = Depth as a variable to count for a unit step, ft 3. SA = Surface Area, in^ 4. E = Energy, or (W*H), lb-ft 5. F = Hammer Frequency, cycles (blows)/ minute 6. T(i) j = Incremental time per depth step (10 foot here), min 7. Variable_Time , = Sum of incremental times (total time for driving a total length of L, minutes 8. R(i) = Soil resistance for piles during driving (static or dynamic), lb 9. S(i) = Average penetration per blow (set) for last 5 or 10 blows, in/blow 10. deltad = Increment or Unit Step of Driving Depth for Calculating most of the required variables above, in this routine it is set to 1 foot */ /* Program Assumptions It is important to emphasize that piles are used as earth retaining system rather than foundation support (for sheet piles and H-piles). 1. Hammer/soil/pile is mimicked in this program. 2. Hammers denote impact type only. There are several hammers of Single Acting Air/Steam Hammer, Double Acting Air/Steam Hammer. Others will be added as required. 3. Soil stratum can be one or two layers. This description can be extended to more than that. 4. By equating total resistance to total allowable driving load, safety factors are not considered. 5. Skin friction (SF), is estimated bu using the linear relationship with depth set out in Peurifoy (1970). However, for speciality contractors, soil/pile/hammer scenarios are documented or recorded based on dynamic testing and/or from previous jobs. 6. The equation used does not consider end bearing resistance. However, in dense sand or stiff clay, this is a major consideration. This assumption is relaxed in a later stage. 7. A soil plug could may arise for H-piles driven in dense sand or soft/hard clay. It has not been accounted for. Listing B.l Drive.c Routine for Pile Driving Appendix B. Pile Driving Production Derivation 290 8. Energy losses such as hammer loss, impact loss, compression loss are not considered. However, for simplification, this may be accounted for in the hammer efficiency enquiry for the user in the CMSA before running this routine. 7 /* Following soil profiles are assumed under the conditions that each skin friction is given by Peurifoy (1970). Thus no elicitation is required for unit weights, water table, and other soil parameters */ char *soil[] = { "Loose_Sand", "Dense_Sand", "Soft_Clay", "Stiff_Clay" }; /* Hammers provided are those of Single Acting Air/Steam hammers, Drop Hammers, and Double Acting Air /Steam Hammers */ char *hammer[] = { "SAAH", "DAAH" /* Double Acting Air Hammer */ "DASH", "DH" /* Drop hammer */ }; main () { float SA, E, F, L, start, finish, AcSf, n_blows; float depth, deltad, Ri, Si, Ti, VariableTime; int fd, i, done; FILE *fl, *f2; char buffer[BUFFSIZE], valuestr[STRSIZE], c, scrout; char *bufferptr, *p; char Soil_Type[STRSIZE + l], Hammer_Type[STRSIZE + l]; int NSOILS = sizeof(soil)/sizeof(char *); int NHAMMS = sizeof (hammer)/sizeof(char *); float SF (), HF 0; /* Read data from "Hammer.nxp" file which includes refusal depth (L), Hammer Frequency, Hammer Energy, Pile Surface Area, and Hammer Type; in total five parameters */ if ((fd = open ("HAMMER.NXP", ORDONLY)) = = -1) return (0); read (fd, buffer, BUFFSIZE); close (fd); /* Extract the values out of the buffer */ bufferptr = &buffer[0]; i = 0; Listing B.l Drive.c Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 291 while (*bufferptr = = ' \ \ ' && *bufferptr != '*') { p = &valuestr[0]; while (* + + bufferptr ! = ""); while ((c = *+ + bufferptr) != "") *(P++) = c; while ((c = *+ +bufferptr) = = " 11 c = = '\r' 11 c = = V ) ; *p = '\0'; switch (i) { case 1: L = atof (valuestr); break; case 2: F = atof (valuestr); break; case 3: E = atof (valuestr); break; case 4: SA = atof (valuestr); break; case 5: p = &valuestr[0]; while (*p = = ") /* removing of leading blanks */ p++; strcpy (Hammer_Type, p); break; } } /• Files Called and Created for Input/Output 1. Input data are entered in two flat files with an .nxp extension to facilitate NExpert Object interface. These files are soil.nxp that contains the soil profile (soil type, start and end of each layer). Hammer.nxp that contains selected hammer and pile parameters. 2. Output results are shown in two files of (nxp) format. These are: out.nxp, mainly used to retrieve selected output variables, especially technical feasibility state and cumulative variable pile driving time, for further knowledge processing and manipulation. outout, mainly used to check the validity of the output variables during development under the Dos. It contains incremental and cumulative values for soil depth, skin friction, penetration rate, time, and number of blows. 7 if ((fd = open ("SOIL.NXP", ORDONLY)) = = -1) return (0); read (fd, buffer, BUFFSIZE); close (fd); /* Open the "OUT.NXP" and "OUT.OUT" files to output results 7 Listing B.1 Drive.c Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 292 if ((fl = fopen ("OUT.NXP", "w")) = = NULL) return (0); if ((f2 = fopen ("OUT.OUT", "w")) = = NULL) return (0); /* Ask the user if he wants to see the output (out.out) on the screen or Not 7 printf ("Do you want output on screen? (Y/N) "); scr_out = (char) getchar (); fprintf (f2," Depth Soil Type Soil Res. Penetration Inc. Time Total Time Acc Skin Frictn Cum Blow\n"); fprintf (f2," (ft) (lb) (in/blow) (min) (min) (lb/ft*ft)\n"); if (scr_out = = V 11 scrout = = 'Y') { printf (" Depth Soil Type Soil Res. Penetration Inc. Time Total Time\n"); printf (" (ft) (lb) (in/blow) (min) (min)\n"); } Variable_Time = 0.0; delta_d = 1.0; done = 0; AcSf = 0.0; n_blows = 0.0; /* Extract the soil profile information out of the buffer */ bufferptr = &buffer[0]; while ("bufferptr = = ' \ \ ' && *bufferptr != '*') { /* Get the soil type of this layer */ p = &valuestr[0]; while (*++bufferptr != ""); while ((c = *+ + bufferptr) != "") *(p++) = c; while ((c = * + +bufferptr) = = " 11 c = = '\r' 11 c = = '\n'); *p = '\0'; p = &valuestr[0]; while (*p = = ") /* removing of leading blanks */ p++; strcpy (Soil_Type, p); /* Get the starting depth of this layer */ p = &valuestr[0]; while (* + + bufferptr ! = '"'); while ((c = *+ + bufferptr) != "") *(p+ +) = c; while ((c = *+ +bufferptr) = = " 11 c = = '\r' 11 c = = '\n'); *p = '\0'; start = at of (valuestr); /* Get the ending depth of this layer */ p = &valuestr[0]; while (*++ bufferptr != ""); while ((c = *+ + bufferptr) != "") *(p++) = c; while((c= *++bufferptr) == " | | c== V II c== V ) ; Listing B.l Drive.c Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 293 *p = '\0'; finish = atof (valuestr); if (finish > L) finish = L; /* we do not want to go beyond the pile's length */ /* Following Pile Driving Formula is based on Engineering News formula. The calculations for soil resistance and other parameters are applicable to single-acting steam hammer. 7 /* Definitions R (i) = safe load on a pile, lb S(i) = average penetration in in/blow /* skin friction Resistance is computed iterate and cumulatively whereas end bearing is computed discretely */ for (depth = start; depth < = finish; depth + = deltad) { AcSf = AcSf + SF (depth, SoilType, NSOILS, soil); Ri = (SA * 12) * AcSf; Si = HF (E, Ri, HammerType, NHAMMS, hammer); nblows = n_blows + (12 / Si); Ti = (deltad * 12) / (Si * F); VariableTime + = Ti; fprintf (f2, "%6.0f %10s %10.2f %12.4f %8.2f %8.2f %8.2f %6.2f\n", depth, SoilType, Ri, Si, Ti, VariableTime, AcSf, nblows); if (scr_out = = y 11 scr_out = = 'Y') printf ("%6.0f %10s %10.2f %12.4f %8.2f %8.2f\n", depth, Soil_Type, Ri, Si, Ti, Variable_Time); /* Set a flag to test if the maximum allowable number of hammer blows per inch, 8 to 12 blows/inch as stated in Peurifoy (1970) and Canadian Foundation Manual (1978), has been exceeded or not. If hammer blow rate (inversely proportional to penetration rate) is more than that, then there is a chance the pile could be overstressed or damaged due to either 1) existence of an obstacle, 2) higher soil resistance, 3) Lower Hammer efficiency. So, lower bound for penetration rate (Si) can be translated as 1/12 (0.083) inch per blow (1/Si), or as Allowable Maximum Ti in minutes per ft. The first one was chosen, and therefore, if that limit is reached at a given driving depth, the program further assigns a Boolean variable of the technical feasibility state to "False". The program returns the value of technical feasibility and the depth where stopped if it did not reach the refusal depth. CMSA will interpret these results and suggest a remedy action. This program makes no allowance for obstruction */ if (Si < = 0.083) { printf ("Driving stopped at depth = %4.0f since penetration rate\n", depth); printf ("has reached its limit of 12 blows per inch.\n"); Listing B.l Drive.c Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 294 done = 1; break; } } /* Don't even look at the next layer if the pile's Refusal Depth has been reached */ if (depth > = L 11 done) break; } /* Write the results to the output file */ sprintf (valuestr, "%6.2f, delta_d); p = &valuestr[0]; while (*p == ")p++; fprintf (fl, "\\Incremental_Depth.amount\\ = \"%s\"\n", p); sprintf (valuestr, "%8.2f, depth-delta_d); p = &valuestr[0]; while (*p = = ") p+ +; fprintf (fl, "\\Depth.amount\\ = \"%s\"\n", p); sprintf (valuestr, "%10.2f, Ri); p = &valuestr[0]; while (*p = = ") p+ +; fprintf (fl, "\\Soil_Resistance.amount\\ = \"%s\"\n", p); sprintf (valuestr, "%12.4f, Si); p = &valuestr[0]; while (*p = = ") p+ +; fprintf (fl, "\\Penetration.amount\\ = \"%s\"\n", p); sprintf (valuestr, "%8.2f, Ti); p = &valuestr[0]; while (*p == ")p++; fprintf (fl, "\\Incremental_Time.amount\\ = \"%s\"\n", p); sprintf (valuestr, "%8.2f, Variable_Time); p = &valuestr[0]; while (*p = = ") p+ +; fprintf (fl, "\\Variable_Time.amount\\ = \"%s\"\n", p); if (done) { fprintf (fl, "\\Technical_Feasibility.State\\ = \"False\"\n"); fprintf (f2, "\nTechnical Feasibility = False\n"); printf ("\n Technical Feasibility = False\n"); } else { fprintf (fl, "\\Technical_Feasibility.State\\ = \"True\"\n"); fprintf (f2, "\nTechnical Feasibility = True\n"); printf ("\n Technical Feasibility = True\n"); } fprintf (fl fclose(fl); fclose(f2); Listing B.l Drivcc Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 295 float SF (depth, SoilType, NSOILS, soil) float depth; char *Soil_Type, *soil[]; int NSOILS; { int i; for (i = 0; i < NSOILS; i+ +) { if (strcmp (Soil_Type, soil[i]) = = 0) break; } switch (i) { /* Following step function equations were used for estimating skin friction. Those were derived, by interpolation, from Peurifoy (1970), table 17-4, pp. 483. They will be used as upper bound for the theoretical calculated skin friction 7 case 0: /* Soil Type 1 = Loose_Sand */ if (depth > 0.0 && depth < = 20.0) return (17.5 * depth); else if (depth > 20.0 && depth < = 60.0) return (1.88 * depth + 312.50); else if (depth > 60.0 && depth < = 100.0) return (1.88 * depth + 312.50); else if (depth > 100.0) return (500); break; case 1: /* Soil Type 2 = Dense_Sand */ if (depth > 0.0 && depth < = 20.0) return (30.00 * depth); else if (depth > 20.0 && depth < = 60.0) return (2.50* depth + 550.00); else if (depth > 60.0) return (700); break; case 2: /* Soil Type 3 = Soft_Clay */ if (depth > 0.0 && depth < = 20.0) return (7.5 * depth); else if (depth > 20.0 && depth < = 60.0) return (1.25 * depth + 255.00); else if (depth > 60.0 && depth < = 100.0) return (1.25 * depth + 255); else if (depth > 100.0) return (350); break; Listing B.l Drivex Routine for Pile Driving (continued) Appendix B. Pile Driving Production Derivation 296 case 3: /* Soil Type 4 = Stiff_Clay */ if (depth > 0.0 && depth < = 20.0) return (20.0 * depth); else if (depth > 20.0 && depth < = 60.0) return (1.25 * depth + 375.80); else if (depth > 60.0 && depth < = 100.0) return (1.25 * depth + 375.80); else if (depth > 100.0) return (500); break; case 4: /* Other Soil types = Clay_Sand, SandGravel, etc. */ break; default: printf ("unknown"); break; } } float HF (E, R, HammerType, NHAMMS, hammer) float E, R; char *Hammer_Type, *hammer[]; int NHAMMS; { int i; for (i = 0; i < NHAMMS; i+ +) { if (strcmp (HammerType, hammer[i]) = = 0) break; } switch (i) { case 0: /* Single and Double Acting Air/Steam Hammer 7 return (2*E/R -0.1); /* This is Energy Equation for SAAH/SASH, DAAH/DASH Likewise, other hammer dynamic formulas could be entered 7 case 1: /* Drop Hammer */ return (2*E/R -1.0); /* This is Energy Equation for Drop Hammer */ case 2: /* If no other hammer specified return 0 */ return (0.0); default: return (0.0); } } Listing B.1 Drive.c Routine for Pile Driving (continued) Appendix C Interviews C.l Introduction This appendix covers interviews conducted with some shoring speciality contractors. A survey was also designed to s o l i c i t input from ya large number of contractors. It was not formally distributed and i s therefore not included. Both formal and informal meetings were held with specialty contractors on the topic of shoring methods including v i s i t i n g a site where steel sheet p i l e driving activity was in progress. C.2 Minutes of Meeting with Dillingham Contractors Present: Alan Russell (ADR), Ibrahim Al-Hammad (IA), Stuart Brown (Chief Engineer, Dillingham) Date: January 30th, 1990 Background: The meeting took place at Dillingham headquarters office on Wednesday morning, 9:45-11:45, on January 30, 1990 with the above participants. Stuart Brown was the expert while ADR and Ibrahim acted as knowledge engineers. Mr. Brown is a senior engineer at Dillingham Contractors Company. He has been involved in estimating, bidding and executing shoring projects. Discussion started after ADR briefed Brown about the methods selection problem and i t s relevance to Ibrahim's research work. Brown was given a outline of the proposed interview that contained an introduction to the subject matter, objectives of the interview for knowledge acquisition, and a brief tutorial about expert systems. The discussion that ensued is summarized below. Question 1: Given there are two shoring alternatives steel sheet piles (SSP) versus soldier piles and lagging (SPL), what factors decide the selection of one at a high level without going into detailed analysis? Appendix C. Interviews 298 Answer: Fir s t , Brown indicated that SSP and SPL are similar in being positive systems: i.e. they contain water flow, and are safer to work with, and have a lower risk premium. Furthermore, ground conditions and local experience are the major considerations for shoring selection. Subsurface conditions in Vancouver are either good ground or "garbage". Brown emphasized that expertise and a v a i l a b i l i t y of the equipment in the region or local area are major considerations in selecting a shoring type. For example, in the Vancouver area, shotcrete with anchors i s the prevailing method (with some risk) whereas in Toronto, i t i s soldier piles and lagging. Shotcrete in Vancouver has been used for the last 15 to 20 years because of the good s o i l , i t i s 3 times faster than soldier piles and lagging, and i t i s more economical. It was also emphasized that standard local practice i s very important when selecting a shoring method. In Vancouver, common shoring methods are shotcrete, soldier piles and lagging, and other s o i l nailing; steel sheet piles are typically used for cellular and marine cofferdams. Two major c r i t e r i a were identified for evaluating alternatives: price, and local risk assessment. Those are quantified in terms of monetary value with expected probabilities. The bottom line i s to use the most economical system. Question 2: Does the excavation activity have a major influence on shoring selection? Answer: Shoring activity i s more expensive than excavation. Excavation i s insensitive to shoring method and i s only a secondary consideration. As for ease of excavation, i t i s preferable to use anchors, not struts. Using struts causes an extra cost of approximately 30% by requiring material handling equipment (mainly cranes). Question 3: How does the firm makes decisions about selecting a shoring method? Answer: A brainstorming session takes place to discuss alternatives globally, with the decision being based mainly on cost and confidence that the method w i l l work. A contingency allowance i s part of the decision making. Of paramount Appendix C. Interviews 299 importance are historical records of productivity for different methods in making the choice. For instance, for one of the projects they bid, the s o i l condition was of sand and gravel with high water-table for a 30 foot deep cut. Two alternatives were considered: 1. Shotcrete which requires an open face in order to be sprayed to a pre-designed thickness. Local risk i s of wall collapse and how big the open face i s . If there i s a washout failure, then, the resulting loss includes the costs of crews not working, crews fixing the problem, and the loss of equipment. The risks and other potential losses are assessed using educated guesses by going through what i f questions. 2. Soldier piles and Lagging which i s a positive system. At the end, having considering a l l the factors, shotcrete was selected. Question 4: How much detailed design i s performed before an alternative is pruned? For instance, consider SSP and SPL. Answer: Fir s t , the level of design depends on the l i k e l y alternatives, the size of the project, profit, number and identity of competitors, etc. For SSP and SPL, a minimum preliminary design that takes approximately two hours includes sizing the materials, spacing and sizing of the supports and pricing out the materials. Other considerations include purchase of materials and current material inventory. An attempt i s always made to use materials on hand as opposed to buying new materials. Sizing i s done for H-piles and Steel sheet piles, struts, wales, and anchors. Question 5: What are the standard lengths of SSP used in the f i e l d and how important i s i t to optimize their length for more efficie n t driving ? Answer: That depends upon site location, site layout (which influences the maximum length that can be welded on sit e ) , means of transportation (determines i f large sizes are Appendix C. Interviews 300 possible), manufacturers (typically they come in 30' and 60' lengths, and sometimes up to 120'), depth of the cut, size of crane and hammer for pile driving. When steel sheet p i l i n g i s available from other projects, short lengths w i l l be used, even though additional welding i s required, because of the economies involved. The biggest advantage of welding piles before driving them, i s to have minimum hammer setting time during which the crane i s idle. Typically, this i s 10-15 minutes, whereas the variable time takes 1/4 to 1/3 of total driving time (e.g. 5 minutes for 100 foot using Single Acting Air Hammer (SAAH) in loose sand). Follow-up question: If time i s a major concern and a steel sheet p i l i n g system was specified by the design engineer, does shortening the project duration alter the SSP design and are other shoring methods evaluated as well? Answer: Well, you might use heavier sheets with a retaining system of anchors, i f possible. This i s more expensive than using lighter sheets. However, there i s less risk of sheets being overstressed or damaged, resulting in delays. Question 6: When time i s an issue, what shoring method i s preferable? ANSWER: Sometimes, time can be a very important factor. For instance, for shotcrete, i t i s fast in wide open excavation whereas narrow cut may require a staged excavation (or vice versa). For a street cut project, Dillingham was asked to close the street for a week and use decks over the cut. For such a case, soldier piles and lagging are preferable since this system provides a more r i g i d support for passing cars and trucks, although i t i s less economical. The social impact of not disturbing t r a f f i c had more priority and overruled shoring economics. For shotcrete, the site i s l e f t open in stages, and you have a short time to work due to short s o i l stand up time for spraying and anchor installation. Other follow-up crews have more time to work. In order to speed up construction duration, one may consider: 1. 2 . 3. Using fast setting material; double shifting; and overtime. Appendix C. Interviews 301 Remember for SPL, a staged installation approach i s required, similar to shotcrete (excavation then lagging). Shotcrete needs sufficient length — a minimum of a 300 foot long cut. Tunnel dimensions have an influence on selecting methods. For instance, a narrow tunnel follows a linear process whereas a wide tunnel allows more than one crew in one space compartment unit. Thus for a narrow tunnel, SSP is more preferred than SPL due to d i f f i c u l t y of instal l i n g lagging, which makes i t more d i f f i c u l t to excavate. Question 7: How accurate are the estimates when alternatives are being pruned ? Answer: Normally, for preliminary design, cost estimates are within 10% of the detailed design estimate, Question 8: How do ground conditions influence the risk for shoring? Answer: If encountered ground conditions are worse than anticipates then use large SSP and big struts. This may increase the cost, but i t shortens time and i s safer (less risk) . For instance, SSP i s much faster in uniform loose s o i l where there i s no obstruction or tough driving. Question 9: How do you prune alternatives based on technical f e a s i b i l i t y without considering risk or under the same risk ? Answer: For soft s o i l s , i.e., loose sands and s i l t s with high water level, s o i l containment i s important. You can not afford to lower the water table, so the choice becomes SSP. If the s o i l i s t i l l then SSP is not feasible due to hard driving. For loose sand and gravel, use H-piles. A problem with soldier piles and lagging i s that to i n s t a l l the lagging, you need an open face; leaking water may cause a problem. H-Piles are specified and spaced to support s o i l arching, whereas, lagging i s used to contain water and s o i l . You may use pre-drilled H-Piles to reduce the chance of damage. For medium dense sand and gravel, both SSP and SPL can be used. For t i l l and sand, i t i s not economical to use SSP. For Cofferdam, heavy SSP sections are used to minimize the number of supports; anchors are preferred because they are Appendix C. Interviews 302 cheaper than struts and make construction faster. If the ground i s loose, or swampy, then anchors can't be used. Anchors should be used whenever s o i l easement, and property access permission conditions are realized. A disadvantage of using struts i s that i t ties up a crane. Question 10: What hammers are preferred for SSP driving? Answer: Fir s t , SSP i s rarely used in Vancouver since s o i l i s either good (strong ground) or "garbage". If SSP i s used, vibratory hammers are the most efficient. The speed of driving i s a function of the length of SSP, weight of the SSP unit, and vibratory hammer size. When choosing a vibratory hammer, the rule of thumb used i s to select the one with the highest energy. Keep in mind that set up time i s 3 to 4 times the driving duration. Use a greedy approach to select the vibratory hammer size. Using the biggest one without overstressing the p i l e results in putting contingency up front in case of d i f f i c u l t driving. There w i l l be no need to change the hammer, thus avoiding risks i f d i f f i c u l t conditions are encountered. Single Acting Air Hammers (SAAH), Double Acting Air Hammers (DAAH) , and Drop hammers are used for soft driving. Air hammers (every blow 8 inches to 2 foot) requires adjusting the crane. If diesel hammer is used, there has to be enough s o i l resistance for p i l e driving, i.e., i t i s more suitable for hard material which exhibits higher s o i l resistance than those of the a i r hammers. Normally, two piles are driven together, however, for tough driving drive one pile at time. Note that hammer vertical mobility and attachment are important. In essence, you do not want the pile to penetrate too much per blow such that i t slips out of the leads and thus requires set up time for each blow. Question 11: If, for example, a diesel hammer has been recommended, what model should be selected? Answer: It depends on the length and weight of the pi l e and other factors. Rules of thumb to use are: 1. The weight of the hammer ram should be 1/3 of pile weight or more; and 2. Piles could take up to 12 blows per inch. Appendix C. Interviews 303 H-piles come in 40 to 60 foot lengths. Thus, many sub-optimal situations could be contemplated by trade-offs among pil e length and weight, cost of material and equipment and construction efficiency. SSP comes in different sizes lengths (depends on mode of transportation) from 20' to 70'. The goal i s to minimize welding. Question 12: If you have two layers of s o i l , for example, the f i r s t one is loose s o i l and the second i s dense s o i l , how do you select a hammer? Answer: Start with a Drop/Air hammer for the f i r s t layer, then use diesel hammer for the second layer. Use a vibratory hammer for loose to medium dense s i l t and sand. Question 13: What i s the average setup time and driving time for pile driving? Our algorithm predicts 3 1/2 minutes for driving time only for a Double-Acting Air Hammer for loose sand and 100 feet driving depth? Answer: If i t i s really cooking, i t takes 3 to 3 1/2 minutes for pil e setup time, and i t takes 15 to 20 minutes for driving the p i l e to the refusal depth i f uninterrupted (i.e. total variable time). For the latter, in one instance, the variable time duration was 5 minutes for 100 deep excavation when single acting a i r hammer was used to drive a steel sheet p i l e in loose sand. At the end of the interview, Brown drew a number of useful articles for Pile Buck magazine to our attention. Appendix C. Interviews 304 Rules Deduced from this Interview: Two major factors are considered in evaluating shoring alternatives: price and risk assessment. Both are quantified in terms of monetary value. The following rules are a f i r s t attempt at encoding some of Mr. Brown's experience. These rules w i l l be compared to what i s available in the literature in an attempt to adapt them to CMSA prototype when appropriate. Those rules are written in IF THEN format as crude rules of thumb, shown in l i s t i n g C.l below. RULE: Correlation between Risk and Design Elements IF Steel Sheet Pile (SSP) is Recommended as a Shoring Method AND Heavy SSP are Used THEN Failure Risk of SSP is Low RULE: Speed of Operations during Construction IF Encountering Ground Conditions are worse than anticipated THEN Use Heavier SSP and Big Struts. RULE: Use SSP IF Soil is Poor and/or Loose Sands and/or Silts AND Water Table Level is High THEN Use SSP AND SSP Risk is Low RULE: Do not Use SSP IF Soil is Till THEN SSP Is Not Feasible (Even with heavy SSP, it will be difficult to drive them) AND SP&L is Feasible RULE: WaterTable and Risk IF Water_Table is High AND SP&L is Feasible THEN SP&LRisk is Moderate RULE: Using Arch Principle for Design IF Soil is Non-cohesive AND SP&L is Selected THEN Design H-Piles and Laggings Using Arch Principle RULE: Feasibility for SSP and SP&L IF Soil is Medium Sand, Gravel THEN SSP and SP&L are Feasible. Listing C.l Extracted Rules of Thumb Appendix C. Interviews 305 RULE: Anchor Use IF Ground Conditions are Loose, Swampy OR Easement Is not Possible (Property Access) THEN Anchorage Is Not Feasible RULE: Risk associated with Method (If you select highest energy vibrator, then you avoid the risk of not driving the pile to its final position). IF Vibratory Energy is Maximum AND GWSS of SSP Is Selected OR GWSSofH-Piles Is Selected THEN Method Vibratory Risk Is Very Low RULE: Number of Piles to be Driven IF Driving Is Tough THEN Drive a Single Pile at a Time ELSE Drive Two Piles at a Time RULE: Diesel Hammer Use IF Soil Resistance is High (SPT > upper bound) THEN Diesel Hammer Use is Feasible RULE: Link Tunnel Dimnsions to GWSS Selection IF Tunnel Width is Narrow AND SSP Is Feasible AND SP&L Is Feasible THEN Use SSP Listing C.l Extracted Rules of Thumb (continued) Appendix C. Interviews 306 C3 Minutes of Meeting with Quadra Construction Present: John Simonett, President, Quadra Construction, and Ibrahim Al-Hammad Date: March 24th, 1990 Background: Quadra is a small size company that has l i t t l e documented experience in selecting hammers. It i s experienced in the driving of a wide range of p i l e types. Fi r s t , I introduced the subject matter of this v i s i t by briefing John on expert systems technology and the problem of method selection. Furthermore, the importance of vibratory hammer selection was stressed (that was one of the main purposes of this meeting since a logic procedure had already been developed for Impact Hammers). The purpose of the meeting was several-fold: 1. To examine and compare the logic of the prototype regarding steel sheet p i l i n g sizing and selection with the driving equipment required (hammers and cranes). It was found that such the logic developed to date for the prototype CMSA was similar to what is practiced by this firm, which helped confirm the valid i t y of our approach. 2. An algorithm has already been developed to predict the productivity of impact hammers based on energy equations. However, sheet p i l i n g i s most effectively driven, when conditions allow, by vibratory pil e drivers (in favorable conditions, driving by vibratory i s 10 times faster than driving by impact hammer). So far, there i s nothing in the literature that relates vibratory p i l e driver productivity to i t s type or energy nor to the s o i l type. Thus, I wanted to know how a vibratory pil e driver i s selected. The answer was in the form of a rule of thumb — select a vibratory driver with the highest horse-power. Typically, a vibratory pil e driver drives steel sheet piles 2 0 ft/min in good conditions whereas i t drives them 2 ft/min in bad conditions. I presented John with an a r t i c l e from Pile Buck in which a vibratory driver i s sized based on pil e weight vs. s o i l blow count (STP) charts to get vibratory dynamic force. John was intrigued and tried to verify the charts with a pre-selected ICE 216 Vibratory hammer employed at a job site. However, there was no agreement in the results. John has commented Appendix C. Interviews 307 that this case study does not f a l l within the range of the charts and perhaps may resemble the worst condition. Subsequent discussions yielded the following information: 1. Cost breakdown for sheet p i l i n g driving revealed that the cost of material i s approximately 3/4 of total cost whereas 1/4 i s labor cost and equipment (the labor cost i s almost equal to the equipment cost). Some representative costs are shown below in Table C.l. 1. Vibrator Model Dynamic Force Rent (Tons) Model ICE 812 145.5 $11000/month Model ICE 216 36.4 $7000/month 2. Crane Rent 35 Ton $6500/month 70 Ton $8500/month 150 Ton $12000/month 3. Crew Cost 4 men / day $1100/day Table C.l Sample Pile Driving Resources Unit Cost [Quadra 1990]; In comparison with other shoring methods, material costs and savings are of major concern for SSP rather than labor or equipment. 2. Pile driving time i s 3 to 4 times the set up time. 3. Length of pile i s constrained by transportation means and the crane boom avail a b i l i t y . 4. Pile driving i s more efficient i f driven by pairs — in fact sheets have been ordered in doubles from the factory, or welded at the site. 5. A vibratory hammer typically yields a maximum rate of production of 20 ft/minute under good conditions to a minimum of 2 ft/min. 6. If the s o i l i s cohesive or very compacted, then a vibratory pil e driver i s not suitable. 7. Hammer rental cost does not vary linearly with horsepower. Appendix C. Interviews 308 8. Crane capacity and configuration must suit p i l e dimensions and the hammer weight. Normally, crane cost i s equal to the hammer cost. 9. Vibratory hammers are more productive for non-displacement piles, however, they are more expensive to rent and mobilize. 10. Hydraulic hammers are more productive due to their higher frequency for the same energy. There i s no local experience with them, and apparently, they are more expensive. 11. Risk assessment was mentioned as important. However, the discussion was not elaborated enough to understand the decision process nor how could i t be linked to methods selection. At the end, I asked Mr. Simonett i f I could v i s i t some of their nearby pile driving projects. A month later, Mr Simonett informed me about a project they were involved in. A description of the site v i s i t follows. C.4 Project Site Visit Date: April 28, 1990 A project involving Quadra included driving heavy sheet piles for a caisson. From a top view, i t consist of a rectangular shape 22 feet wide by 43 foot long with 50 foot long sheet piles and a 216 ICE Vibrator (Low Frequency). At f i r s t , an interior frame i s set up to hold the piles together at the ground level. It i s then l i f t e d by a crane to a height of 25 f t and seated on 4 H-Piles. SSP piles are f i r s t seated by their own weight to 1 to 2 foot, then they are inter-connected with their adjacent piles. The foreman explained that he designates some piles as key piles which are welded to the frame- at each cycle welding took place at mid-level frame (at 20 f t elevation) and at ground floor. Piles were driven in vertical rounds and in pairs. First round, vertical driving i s 15 f t ; second round, i t drops to 11 f t due to more resistance, and to 8 f t and so forth. Furthermore, i f piles are hard to drive then they are driven in singles. The foreman mentioned that oiled sheet piles provide less resistance between adjacent piles and that new ones are easier to drive and extract than old ones due to rust. Description of A Sheet Pile Driving Cycle: Sequential p i l e driving proceeds in waves around the cofferdam. When driving reaches the key sheets, a worker has to un-weld them from the top and bottom frames. This causes some delay Appendix C. Interviews 309 during which the crane operator, laborer, and vibratory operator are idle. It was observed that a f a i r amount of the laborers time was consumed in aligning the vibratory p i l e driver in position over the piles and in moving the pil e driver cables around. 3io Appendix D Knowledge Base Sample, Interviews, Unit Costs, and Sample Data Base D.l Introduction This appendix contains listings for: 1. A sample CMSA knowledge base using NExpert Object syntax, printed as a text f i l e . This i s a partial CMSA that excludes the Risk component. 2. Vibratory hammers selection knowledge. 3. Unit costs for materials and crews used for CMSA quantity take-off and cost estimate calculations. 4. A sample of structural members for ground wall support system (GWSS), and data bases used in CMSA prototype. Appendix D. CMSA Partial Listing and Miscellany D.2 Partial Listing of CMSA Knowledge Base 3 1 1 (©VERSION = Oil) (©PROPERTY = amount ©TYPE=Float;) (©PROPERTY = Cost ©TYPE=Float;) (©PROPERTY = CrossSectionArea ©TYPE=Float;) (©PROPERTY = depth ©TYPE = Float;) (©PROPERTY = designation ©TYPE=String;) (©PROPERTY = drivingwidth ©TYPE = Float;) (©PROPERTY = Duration ©TYPE=Float;) (©PROPERTY = .... (©CLASS = Feasibility ) (©CLASS = Matched_SSP (©PROPERTIES = Cross_Section_Area designation drivingwidth Section_Modulus Surface_Area Weight_per_Foot ) ) (©CLASS = Selected_Hammer (©PROPERTIES = HammerModel LengthofStroke RamWeight Strokes_per_Min TheorEnergy ) ) (©CLASS = ...) Listing D.l Partial Listing of CMSA Appendix D. CMSA Partial Listing and Miscellany 312 (©OBJECT = AlternativeofSPL (©PROPERTIES = Cost ) ) (©OBJECT = AlternativeofSSP (©PROPERTIES = Cost Value ©TYPE=String; ) ) (©OBJECT = PZ27 (©PROPERTIES = Value ©TYPE = String; ) ) (©OBJECT = Tunnel (©PROPERTIES = depth Length ) )... (©RULE = Find_Section_Modulus_for_SSP_of_Single_Layer ©COMMENTS = "This rule finds Section Modulus and then passes control to steel sheet pile selection (section modulus is measured in in~3). 15 foot represents vertical spacing between the struts";@WHY = "This rule finds the section modulus for a steel sheet pile in order to select a suitable section."; (@LHS = (Is (SoiLtype) ("LooseSand")) (Name (Tunnel.depth) (Tunnel.depth)) ) (©HYPO = CalPresandMomforSSLofSSP) (@RHS = (Do ((0.65*((1-SIN(0.52))/ (l + SIN(0.52))) *100*Tunnel.depth/1000)) (Lateral_ Pressure)) (Do (Lateral_Pressure*POW(15,2)/8) (MaxMoment)) (Do (Max_Moment*12/(25*1.5)) (SectionModulus)) (Retrieve ("ssp.nxp") (@TYPE = NXP;@FILL=ADD;@CREATE= | Selected_SSP |; \@PROPS = designation, Weight_per_Foot,Cross_Section_Area,\ Surface_Area,driving_width,Section_Modulus;\ @FlELDS = "Designation","Weight_per_foot","Cross_section_area",\ Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 313 "Surface_area"/'Driving_vvidth","Section_modulus";\ )) (Do (Select_SSP) (Select_SSP)) ) ) (@RULE= FindSectionModulusforSSPofTwoSoilLayers ©COMMENTS = "This rule finds Section Modulus and then passes control to selecting steel sheet pile selection (section modulus is measured in in^). 15 foot represents vertical spacing between the struts";@WHY="This rule finds the section modulus for a steel sheet pile in order to select a suitable section."; (@LHS = (Is (SoilTypel) ("Loose_Sand")) (Name (0.1) (Start_l)) (Name (Layer_l_of_Soil_Type_l.depth) (Finish_l)) (Is (Soil_Type_2) ("Soft_Clay")) (Name (Finish_l) (Start_2)) (Name (Tunnel.depth + 5) (Finish_2)) ) (@HYPO = CalPresandMomforTSLofSSP) (@RHS = (Do ((0.65*((1-SIN(0.52))/(l + SIN(0.52)))*100*Tunnel.depth/1000)) (Lateral_ Pressure)) (Do (Lateral_Pressure*POW(15,2)/8) (Max_Moment)) (Do (Max_Moment*12/(25*1.5)) (Section_Modulus)) (Retrieve ("ssp.nxp") (@TYPE = NXP;@FILL=ADD;@CREATE = | Selected_SSP | ;\ ©PROPS = designation, Weight_per_Foot, Cross_Section_Area,\ Surface_Area, driving_width, Section_Modulus;\ @FIELDS = "Designation","Weight_per_foot","Cross_section_area",\ "Surface_area","Driving_width","Section_modulus";\ )) (Do (SelectSSP) (Select_SSP)) ) (@RULE= TechnicallyFeasibleAlternative ©COMMENTS = "If pile driving conditions are soft, pile driving pattern is In_Singles. If the synthesized method under the soil/hammer/pile is not technically feasible then control resets the pile driving pattern status into In_Singles. The technical feasibility status will be checked again. If it is "True", then the latter driving strategy is adopted. If the technical feasibility state is "False", then altering hammer/pile selection is advised. Such advise is not yet incorporated in CMSA). Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 314 (@LHS = (No (TechnicalFeasibility.State)) (Is (DrivingConditions) ("Hard")) (Is (PUeDrivingPattern) ("InSingles")) (Show ("Tech_Fea.txt") (©KEEP=FALSE;@WAIT=TRUE;)) ) (@HYPO= CheckFeasibility) (@RHS = (Do (Re_check_SSP_Hammer_Comb) (RecheckSSPHammerComb)) ) ) (@RULE = Technically_Feasible_Alternative ©COMMENTS = "If the driving conditions are hard, pile driving pattern is "InPairs", and the method is technically feasible, then compute the sheet pile production rate."; (@LHS = (Yes (TechnicalJ'easibility.State)) (Is (DrivingConditions) ("Hard")) (Is (Pile_Driving_Pattern) ("InPairs")) (Show ("Tech_Fea.txt") (©KEEP=FALSE;@WAIT=TRUE;)) ) (@HYPO= CheckFeasibUity) (@RHS = (Do (ComputeProductionofSSP) (Compute_Production_of_SSP)) ) ) (@RULE = Technically_Feasible_Alternative @INFCAT = 3; ©COMMENTS = "If the driving conditions are hard, pile driving pattern is "In_Singles", and the method is technically feasible, then compute the sheet pile production rate."; (@LHS = (Yes (TechnicalJFeasibility.State)) (Is (DrivingConditions) ("Hard")) (Is (PileDrivingPattern) ("InSingles")) (Show ("Tech_Fea.txt") (©KEEP = FALSE;@WAIT = TRUE;)) ) (@HYPO= CheckFeasibUity) (@RHS = (Do (ComputeProductionofSSP) (ComputeProductionofSSP)) ) ) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 315 (@RULE= Technically_Feasible_Alternative @INFCAT=3; ©COMMENTS = "If the driving conditions are soft, pile driving pattern is "In_Singles", and the method is not technical feasible, then compute the sheet pile production rate."; (@LHS= (Yes (Technical_Feasibility.State)) (Is (DrivingConditions) ("Soft")) (Is (Pile_Driving_Pattern) ("In_Singles")) (Show ("Tech_Fea.txt") (©KEEP=FALSE;@WAIT=TRUE;)) ) (@HYPO= CheckFeasibiUty) (@RHS = (Do (ComputeProductionofSSP) (ComputeProductionofSSP)) ) ) (@RULE = TechnicallyFeasibleAlternative ©COMMENTS = "If the driving conditions are soft, pile driving pattern was "InPairs", and the method is technically feasible, then compute the sheet pile production rate."; (@LHS = (Yes (TechnicalJ'easibility.State)) (Is (DrivingConditions) ("Soft")) (Is (PileDrivingPattern) ("InPairs")) (Show ("Tech_Fea.txt") (©KEEP = FALSE;@WAIT = TRUE;)) ) (@HYPO= CheckFeasibility) (@RHS = (Do (ComputeProductionofSSP) (ComputeProductionofSSP)) ) ) (©RULE = Technically_Feasible_Alternative @INFCAT=3; ©COMMENTS = "If the driving conditions are soft, pile driving pattern is "InPairs", and the method is not technically feasible, then reset the pile driving pattern state to "In_Singles" and re-fire the technical feasibility rule (This rule is analogous to a loop construct in a conventional programming)."; (@LHS = (No (Technical_Feasibility.State)) (Is (Driving_Conditions) ("Soft")) (Is (PileDrivingPattern) ("InPairs")) (Name ("InSingles") (PileDrivingPattern)) (Show ("Tech_Fea.txt") Listing D.l Partial Listing of CMSA (continued) Appendix D . CMSA Partial Listing and Miscellany 316 (@KEEP = FALSE;@WAIT=TRUE;)) ) (@HYPO= Check_Feasibility) (@RHS = (Reset (SPProduction)) (Do (SPProduction) (SP_Production)) ) (@RULE= Technically_Feasible_Alternative ©COMMENTS = "If the driving conditions are soft, pile driving pattern is "In_Singles", and the technical feasibility state is "False", then terminate the session. Furthermore, a text file may be displayed to announce that there is no feasible solution under the session scenario. The hypothesis (Re_check_SSP_Hammer_Comb) has not been implemented yet. Its future object is to recommend changes in hammer/pile designs to arrive at feasible method synthesis"; (@LHS = (No (Technical_Feasibility.State)) (Is (DrivingConditions) ("Soft")) (Is (Pile_Driving_Pattern) ("InSingles")) (Show ("Tech_Fea.txt") (@KEEP=FALSE;@WAIT=TRUE;)) ) (@HYPO= Check_Feasibility) (@RHS = (Do (RecheckSSPHammerComb) (Re_check_SSP_Hammer_Comb)) ) ) (@RULE = Production_Measures_of_SSP_for_Driving_Single_PUe ©COMMENTS = "Total driving time of a set of piles to the refusal depth consists of two components; variable component which depends on soil/pile/hammer scenario, and fixed which is dependant on pile welding and positioning and movement of pile driver. A pile is driven in several segments, i.e., in a pile series to the refusal depth. For the variable time computation, it is assumed that there is no pile driving interruption; the fixed time component is equal to multiple of set up times ;@WHY="This is to measure the productivity performance of sheet pile and a given pile driver."; (@LHS = (Is (PileDrivingConditions) ("InSingles")) (Name (Fixed_Time_per_given_Hammer) (Fixed _Time_per_pile.amount)) (Show ("PLength") (@KEEP=FALSE;@WAIT=TRUE;)) (Name (CEIL(Tunnel.depth/SSP.Standard_Length)*Fixed_Time_per_pile.amount) ( Fixed_Time_per_SSP_Unit_Width.amount)) (Name (CEIL(Tunnel.depth/SSP.Standard_Length)) (Number_of_Piles_per_SSP_ UnkWidth)) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 317 (Name (Fixed_Timejper_SSP_Unit_Width.amount+Variable_Time.amount) (Total_Drivmg_Time_per_SSP_Unit_Width)) ) (@HYPO = ComputeProductionofSSP) (@RHS = (Do (2*12*(Tunnel.Length)/( < | Selected_SSP | > .driving_width)) (NumberofSSPUnitWidth)) (Do (Number_of_SSP_Umt_Width*Nujnber_of_PUesj)er_SSP_Umt_Width) (Number_of_Piles)) (Do(Number_of_SSP_Unit_Width*Total_Dr^^^ (Total_Production_Time_in_Days)) (Do (NumberofPiles/TotalProductionJTime_in_Days) (Productivity_in_Number_of_Piles_Per_Day)) (Do (CEIL(NumberofPiles)) (Number_of_Piles)) (Do (hypo) (hypo)) ) ) (@RULE = Production_Measures_of_SSP_for_Driving_Single_Pile ©COMMENTS = "Total time consists of two components: variable component which depends on soil/pile/hammer scenario, the fixed which is dependant on pile welding and positioning, and movement of pile driver. Theoretically the number of sheets used is the same, however, the delivery could be "In-Pairs" requiring less fixed time for the crane. The computation of the total driving time is adjusted by identifying PileDrivingConditions state (InPairs, In_Singles). For the former, driving sheets in pairs, the total driving variable time is perhaps less than the latter, however, the total driving fixed time (as number of the setups) is less than the former. The "PLength" text file explains how pile segment length effects the total driving time; ©WHY="This is to measure the productivity performance of sheet pile and a given pile driver."; (@LHS = (Is (Pile_Driving_Conditions) ("InPairs")) (Name (Fixed_Time_per_given_Hammer) (Fixed_Time_per_Pair_of_piles.amount) ) (Show ("PJLength") (©KEEP = FALSE;@WAIT = TRUE;)) (Name (CEIL(Tunnel.depth/SSP.StandardJL£ngth)*Fked_Timejer_Pairs_of_piles.amount) (Fixed_Time_per_Pair_of_SSP_Unit_Width.amount)) (Name (CEIL(Tunnel.depth/SSP.Standard_Length)*2) (Number_of_Piles_per_Pair_of_SSP_Unit_Width)) (Name (Fixed_Time_per_Pair_of_SSP_Unit_Width.amount+VariableTime.amount) ( Total_Driving_Time_per_Pair_of_SSP_Unit_Width)) ) (©HYPO = ComputeProductionofSSP) (@RHS = (Do (2*12*(Tunnel.Length)/(< | Selected_SSP | > .drivingwidth)) (Number_of_Pair_of_SSP_Unit_Width)) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 318 (Do(Number_of_Pair_of_SSP_Urut_Width*Numte ofSSPJJrntWidth) (Number_of_Pairs_of_PUes)) (Do (Number_of_Pair_of_SSP_Umt_W Pair_of_SSP_Unit_Width/(\8*60)) (TotalProductionTimeinDays)) (Do (2*Niunber_of_Pair_of_PUes/Total_Production_Tmie_in_Days) (ProductivitymNumberofPuesPerDay)) (Do (CEIL(Number_of_Pair_of_Piles)) (Number_of_Piles)) (Do (hypo) (hypo)) ) (@RULE = WriteDesiredOutputResults (@LHS = (Name (1) (a)) ) (@HYPO= hypo) (@RHS = (Write ("Results.nxp") (@TYPE = NXP;@FILL=NEW;@UNKNOWN = TRUE;@ATOMS = Techmcal_FeasibUity,\Number_of_PUes,Total_Production_Time_in_Days,\ Productivity_in_Number_of_Piles_Per_Day,Selected_Steel_Pile,\ Pile_Driver,Hammer_Model;)) (Show ("Results.nxp") (@KEEP = FALSE;@WAIT=TRUE;)) \(Do(Done) (Done)) ) ) (@RULE= Select_GWSS_of_SSP (@LHS = (Name (10) (a)) (Show ("Textl-txt") (@KEEP=FALSE;@WAIT=TRUE;)) (Is (GWSS) ("Steel_Sheet_Pile")) ) (@HYPO= SelectAGWSS) (@RHS = (Do (Select_GWSS_of_SSP) (Select_GWSS_of_SSP)) ) ) (@RULE= SelectGWSSofSPL (@LHS = (Name (10) (a)) (Show ("Textl.txt") (@KEEP=FALSE;@WAIT=TRUE;)) (Is (GWSS) ("Soldier_Piles_and_Lagging")) ) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 319 (@HYPO= SelectAGWSS) (@RHS = (Do (SelectGWSSofSPL) (SelectGWSSofSPL)) ) ) (@RULE= Soldier_Piles_is_Selected (@LHS= (Name (10) (a)) (Is (Number_of_Soil_Layers) ("Two_Soil_Layers")) ) (@HYPO= SelectGWSSofSPL) (@RHS = (Do (Calculate_Pressure_and_Moments_for_Two_Soil_Layers_of_SPL) (CalculatePressureandMomentsfor_Two_Soil_Layers_of_SPL)) ) (@RULE= Soldier_Piles_is_Selected (@LHS = (Name (10) (a)) (Is (NumberofSoilLayers) ("One_Soil_Layer")) ) (@HYPO= SelectGWSSofSPL) (@RHS = (Do (Calcmate_Pressure_and_Moments_for_Single_Soil_Layer_of_SPL) (Calculate_Pressure_and_Moments_for_Single_Soil_Layer_of_SPL)) ) ) (@RULE = Steel_Sheet_Pae_is_Selected (@LHS = (Name (10) (a)) (Is (Number_of_Soil_Layers) ("One_Soil_Layer")) ) (@HYPO= Select_GWSS_of_SSP) (@RHS = (Do(Cal_Pres_and_Mom_for_SSL_of_SSP) (Cal_Pre_and_Mom_for_SSL_of_SSP)) ) ) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 320 (@RULE = SteelSheetPileisSelected (@LHS = (Name (10) (a)) (Is (NumberofSoilLayers) ("TwoSoilLayers")) ) (@HYPO= SelectGWSSofSSP) (@RHS = (Do(Cal_Pres_and_Mom_for_TSL_of_SSP) (Cal_Pres_and_Mom_for_TSL_of_SSP)) ) ) (@RULE = SelectVibratoryPDorDAHammerPDforVeryLooseSand @INFCAT=0; ©COMMENTS="Select type of pile driver (mainly hammers) based on soil conditions (heuristics found in Hunt 1979)"; (@LHS = (Is (SoiLtype) ("Very_Loose_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Very_Loose_Sand ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics found on Hunt 1979)"; (@LHS = (Is (Soil.type) ("Very_Loose_Sand")) ) (@HYPO= Select_Pile_Driver) (@RHS = (Let (PileDriver) ("DoubleActingHammer")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Very_Loose_Sand @INFCAT=0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics found in Hunt 1979)";@WHY="Inference category (INFCAT = 0) is set to 0 in order to set a low priority for selecting a vibratory pile driver as opposed to impact hammers."; (@LHS = Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 321 (Is (Soil.type) ("Very_Loose_Sand")) ) (@HYPO= Select_Pile_Driver) (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Very_Dense_Sand ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics found in Hunt 1979)"; (@LHS = (Is (SoiLtype) ("Very_Dense_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("Double_Acting_Hammer")) (Do (RetHammerEnergy) (RetHammerEnergy)) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Very_Dense_Sand @INFCAT = 0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics found in Hunt 1979)"; (@LHS = (Is (SoiLtype) ("Very_Dense_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("Vibratory")) (Do (RetHammerEnergy) (RetHammerEnergy)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Medium_Sand ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics found in Hunt 1979)"; (@LHS = (Is (SoiLtype) ("Medium_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (Pile_Driver) ("Double_Acting_Hammer")) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 322 (Do (RetHammerEnergy) (RetHammerEnergy)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Medium_Sand @INFCAT=0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (SoiLtype) ("MediumSand")) ) (@HYPO= SelectJPile_Driver) (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Loose_Sand ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)";@WHY = "Inference category is set to 1 for DA-hammer which indicates that it has higher priority than vibratory pile driver"; (@LHS = (Is (SoiLtype) ("Loose_Sand")) (Retrieve ("daah.nxp") (@TYPE=NXP;@FILL=ADD;@CREATE = | SelectedHammer | ;\ )) (Name (MAX( < | Selected_Hammer | > .Theor_Energy)) (Max_Energy)) (= (< | Selected_Hammer | > .Theor_Energy-Max_Energy) (0)) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("DoubleActingHammer")) (Let (HammerType) ("SASH")) (CreateObject ( < | SelectedHammer | > ) (| Suitable_Hammer |)) (DeleteObject ( < | Selected_Hammer | > ) (| Selected_Hammer |)) (Do (< | SuitableHammer | > .HammerModel) (Hammer_Model)) (Do (SPProduction) (SPProduction)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Dense_Sand @INFCAT = 0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 323 (@LHS = (Is (SoiLtype) ("Dense_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (Ret_Hammer_Energy) (Ret_Hammer_Energy)) ) ) (@RULE = Select_Vibratory_PD_or_DA_Hammer_PD_for_Dense_Sand ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (SoiLtype) ("Dense_Sand")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("Double_Acting_Hammer")) (Do (Ret_Hammer_Energy) (RetHammerEnergy)) ) (@RULE = Select_Vibratory_PD_for_Very_Stiff_Clay ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (SoiLtype) ("Very_Stiff_Clay")) ) (@HYPO= Select_Pile_Driver) (@RHS = (Let (PileDriver) ("Single_Acting_Hammer")) (Do (Ret_Hammer_Energy) (Ret_Hammer_Energy)) ) ) (©RULE = Select_Vibratory_PD_for_Very_Soft_Clay @INFCAT=0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (SoiLtype) ("Very_Soft_Clay")) ) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 324 (@HYPO= SelectPileDriver) (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (RetHammerEnergy) (RetHammerEnergy)) ) ) (@RULE= Select_Vibratory_PD_for_Stiff_Clay ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (Soil.type) ("StiffClay")) ) (@HYPO= SelectJ>ile_Driver) (@RHS = (Let (PileDriver) ("DoubleActingHammer")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) (@RULE = Select_Vibratory_PD_for_Medium_Clay ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (Soil.type) ("Medium_Clay")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("DoubleActingHammer")) (Do (Ret_Hammer_Energy) (RetHammerEnergy)) ) ) (@RULE = Select_Vibratory_PD_for_Medium_Clay @INFCAT = 0; ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (SoiLtype) ("Medium_Clay")) ) (@HYPO= SelectPileDriver) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 325 (@RHS = (Let (Pile_Driver) ("Vibratory")) (Do (RetHammerEnergy) (Ret_Hammer_Energy)) ) ) (@RULE = Select_Vibratory_PD_for_Hard_Clay ©COMMENTS = "Select type of pile driver (mainly hammers) based on soil conditions (heuristics based on Hunt, Hall)"; (@LHS = (Is (Soil.type) ("HardClay")) ) (@HYPO= SelectPileDriver) (@RHS = (Let (PileDriver) ("Single_Acting_Hammer")) (Do (RetHammerEnergy) (RetHammerEnergy)) ) ) (@RULE= Select_SSP_of_PZ_38 ©COMMENTS = "Select SSP designation of PZ_38, Section_Modulus is in inA3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (SectionModulus) (38.3)) (< (SectionModulus) (46.8)) (Is ( < | Selected_SSP | > .designation) ("PZ_38")) ) (@HYPO= SelectSSP) (@RHS = (Do ( < | SelectedSSP | > .designation) (Selected_Steel_Pile)) (Do (SelectPileDriver) (SelectPileDriver)) ) ) (@RULE= Select_SSP_of_PZ_32 ©COMMENTS = "Select SSP designation of PZ_38, Section_Modulus is in in~3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (SectionModulus) (30.2)) (< (SectionModulus) (38.3)) (Is (< |Selected_SSP| > .designation) ("PZ_32")) ) (@HYPO= SelectSSP) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 326 (@RHS = (Do (< |Selected_SSP| > .designation) (SelectedSteelPile)) (Do (SelectPileDriver) (Select_Pile_Driver)) ) ) (@RULE= Select_SSP_of_PZ_27 ©COMMENTS = "Select SSP designation of PZ_38, SectionModulus is in inA3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (Section_Modulus) (10.7)) (< (SectionModulus) (30.2)) (Is ( < | Selected_SSP | > .designation) ("PZ_27")) ) (@HYPO= SelectSSP) (@RHS = (Do (< |Selected_SSP| > .designation) (Selected_Steel_Pile)) (CreateObject ( < | Selected_SSP | > ) (| Matched_SSP |)) (Do (SelectPileDriver) (Select_Pile_Driver)) ) (@RULE= Select_SSP_of_PSA_32 ©COMMENTS = "Select SSP designation of PZ_38, SectionModulus is in in/v3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (SectionModulus) (1.9)) (< (SectionModulus) (2.4)) (Is (< |Selected_SSP| > .designation) ("PSA32")) ) (@HYPO= SelectSSP) (@RHS = (Do ( < | Selected_SSP | > .designation) (Selected_Steel_Pile)) (CreateObject (< |Selected_SSP| >) (|Matched_SSP|)) (Do (SelectPileDriver) (SelectPileDriver)) ) ) (@RULE= Select_SSP_of_PSA_28 ©COMMENTS = "Select SSP designation of PZ_38, Section_Modulus is in in"3";@WHY = "This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 327 (@LHS = (>= (Section_Modulus) (2.4)) (< (Section_Modulus) (2.5)) (Is (< | SelectedSSP | >.designation) ("PSA28")) ) (@HYPO= SelectSSP) (@RHS = (Do (< |Selected_SSP| > .designation) (SelectedSteelPile)) (CreateObject ( < | SelectedSSP | > ) (| MatchedSSP |)) (Do (SelectPileDriver) (SelectPileDriver)) ) ) (@RULE= SelectJSSP_of_PMA_22 ©COMMENTS = "Select SSP designation of PZ38, Section_Modulus is in in/v3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (Section_Modulus) (2.5)) (< (Section_Modulus) (5.4)) (Is (< |Selected_SSP| > .designation) ("PMA22")) ) (@HYPO= SelectSSP) (@RHS = (Do (< |Selected_SSP| > .designation) (Selected_Steel_Pile)) (CreateObject ( < | Selected_SSP | > ) (| Matched_SSP |)) (Do (SelectPileDriver) (SelectPileDriver)) ) ) (@RULE= Select_SSP_of_PMA22 ©COMMENTS = "Select SSP designation of PZ38, Section_Modulus is in in^3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (Section_Modulus) (2.5)) (< (Section_Modulus) (5.4)) ) (@HYPO= Select_SSP) (@RHS = (Let (SSP) ("PMA22")) (Do (SelectPileDriver) (SelectPileDriver)) ) ) Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 328 (@RULE= Select_SSP_of_PDA27 ©COMMENTS = "Select SSP designation of PZ38, SectionModulus is in in/v3";@WHY="This is used to determine the type of sheet pile from which other properties can be deduced from the SSP data base (SSP.NXP) for further treatment."; (@LHS = (>= (SectionModulus) (5.4)) (< (SectionModulus) (10.7)) ) (@HYPO= SelectSSP) (@RHS = (Let (SSP) ("PDA27")) (Do (SelectJPileJDriver) (Select_Pile_Driver)) ) (©RULE = Single_Pile_Variable_Production_Time ©COMMENTS="This rule computes the variable driving time for a pile series under soft driving conditions and "InSingles" pile driving patters. The rest of the rule premisses assign variables to selected hammer and pile properties. These variables are then written to the input files "Hammer.nxp" and "Soil.nxp" for the Drive.c routine (see chapter 5)."; @WHY="This rule assumes that soil conditions state is at best and thus pile driving time will be shorter than other soil scenario."; (@LHS = (Name (Tunnel.depth+5) (L)) (Show ("Drive.txt") (©KEEP = FALSE;@WAIT=TRUE;)) (Is (DrivingConditions) ("Soft")) (Is (PilesDrivingPattern) ("InSingles")) (Name (< | SuitableHammer | > .Strokes_per_Min) (F)) (Name (< |Suitable_Hammer| >.Theor_Energy*Hammer.Total_Efficiency) (E)) (Name ( < | Selected_SSP | > .Surface_Area) (SA)) (Name ( < j Selected_SSP j > .Cross_Section_Area) (Ap)) ) (@HYPO= SPProduction) (@RHS = (Write ("Hammer.nxp") (©TYPE=NXP;@FILL=NEW;@ATOMS=L,F,E,SA,Hammer_Type;\ )) (Write ("soil.nxp") (©TYPE = NXP;@FILL=NEW;@UNKNOWN=TRUE;@ATOMS = SoilTypel A Start_l,Finish_l,Soil_Type_2,Start_2,Finish_2;\ )) (Execute ("Drive.exe") (©TYPE = EXE;@WAIT = TRUE;)) (Retrieve ("out.nxp") (©TYPE = NXP;@FILL=ADD;@FWRD = TRUE;@CREATE = | VarTime | ;\ Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 329 ©ATOMS = VariableTime.amount;)) (Retrieve ("out.nxp") (©TYPE=NXP;@FILL=ADD;@FWRD=TRUE;@CREATE = | Feasibility | ;\ ©ATOMS=Technical_Feasibility.State;)) (Do (CheckFeasibility) (Check_Feasibility)) ) ) (©RULE = Single_Pile_Variable_Production_Time ©COMMENTS = "This rule computes the variable driving time component for a pile series."; (@LHS = (Name (Tunnel.depth+5) (L)) (Show ("Drive.txt") (©KEEP = FALSE;@WAIT=TRUE;)) (Is (DrivingConditions) ("Hard")) (Name (< | SuitableHammer | > .Strokes_per_Min) (F)) (Name (< | SuitableHammer | > .Theor_Energy*Hammer.Total_Efficiency) (E)) (Name ( < | Selected_SSP | > .Surface_Area) (SA)) (Name (< |Selected_SSP| >.Cross_Section_Area) (Ap)) ) (@HYPO= SPProduction) (@RHS = (Write ("Hammer.nxp") (©TYPE = NXP;@FILL=NEW;@UNKNOWN = TRUE;@ATOMS = L , \ F,E,SA,Hammer_Type;)) (Write ("soil.nxp") (@TYPE = NXP;@FILL=NEW;@ATOMS = Soil_Type_l,\ Start_l,Finish_l,SoU_Type_2,Start_2,Finish_2;\ )) (Execute ("Drive.exe") (©TYPE = EXE;@WAIT = TRUE;)) (Retrieve ("out.nxp") (©TYPE = NXP;@F1LL=ADD;@FWRD = TRUE;@CREATE = | VarTime | ;\ ©ATOMS = Variable_Time.amount;)) (Retrieve ("out.nxp") (©TYPE=NXP;@FILL=ADD;@FWRD=TRUE;@CREATE = | Feasibility | ;\ ©ATOMS=Technical_Feasibility.State;)) (Do (Check_Feasibility) (Check_Feasibility)) ) ) (©RULE = Single_Pile_Variable_Production_Time @INFCAT = 5; ©COMMENTS = "This rule computes the variable driving time component for a pile series."; Listing D.l Partial Listing of CMSA (continued) Appendix D. CMSA Partial Listing and Miscellany 330 (@LHS = (Name (Tunnel.depth+5) (L)) (Show ("Drive.txt") (@KEEP = FALSE;@WAIT=TRUE;)) (Is (Driving_Conditions) ("Soft")) (Name ("InPairs") (Piles_Driving_Pattern)) (Name (< |Suitable_Hammer| >.Strokes_per_Min) (F)) (Name (< | SuitableHammer | >.Theor_Energy*Hammer.Total_Efficiency) (E)) (Name (< |Selected_SSP| >.Surface_Area*2) (SA)) (Name ( < j Selected_SSP | > .Cross_Section_Area*2) (Ap)) ) (@HYPO= SPProduction) (@RHS = (Write ("Hammer.nxp") (@TYPE=NXP;@FILL=NEW;@ATOMS=L,F,E,SA,Hammer_Type;\ )) (Write ("soiLnxp") (@TYPE = NXP;@FILL=NEW;@ATOMS = SoU_Type_l,\ Start_l,Finish_l,Soil_Type_2,Start_2,Finish_2;\ )) (Execute ("Drive.exe") (@TYPE = EXE;@WAIT=TRUE;)) (Retrieve ("out.nxp") (@TYPE = NXP;@FILL=ADD;@FWRD = TRUE;@CREATE = | VarTime | ;\ ©ATOMS = Variable_Time.amount;)) (Retrieve ("out.nxp") (@TYPE = NXP;@FILL=ADD;@FWRD = TRUE;@CREATE = | Feasibility | ;\ ©ATOMS=TechnicalFeasibility.State;)) (Do (CheckFeasibility) (CheckFeasibility)) ) ) (@GLOBALS = ©INHVALUP = FALSE; ©INH VALDOWN=TRUE; Listing D.l Partial Listing of CMSA (continued) Appendix D . CMSA Partial Listing and Miscellany 331 D.3 Vibratory Hammer Selection Knowledge Base Development Vibratory p i l e driver selection knowledge was extracted from the literature. This component serves to select and size a vibratory hammer for a given s o i l and pi l e scenario. What is missing, i s a model which can predict p i l e driving penetration rate as opposed to the dynamic formulas applied to the impact hammers. Furthermore, there i s no indication what kind of vibratory p i l e driver suits a s o i l profile (as opposed to impact hammers types such as SAAH, DAAH, etc.). The interview with Quadra Construction Co. (see Appendix C) revealed that vibratory p i l e driver type selection i s based on experience. In summary, there i s no mechanism to check whether a pi l e driven by a vibratory p i l e driver could reach i t s refusal depth nor how long would i t take to drive a p i l e to i t s refusal depth. Therefore, this part i s limited to vibratory p i l e driver selection without a technical f e a s i b i l i t y test. The vibratory p i l e driver selection i s based on empirical formula and charts found in Barber (1987) — where a vibratory hammer can be specified at this level by i t s dynamic force and amplitude. From graph 1 of the previous reference, for a given s o i l profile SPT and different steel p i l e unit weights, one can find a vibratory dynamic force. Table D.l presents vibratory dynamic force (in tons) computed in terms of SPT (substituted as N) against different p i l e unit weights as derived from Graph 1 (Barber 1979). For instance, the f i r s t relationship implies the following governing equation. For pi l e unit weight of 10 l b / f t , Vibratory Dynamic Force = 0.21 * N (D.l) where N is the standard penetration test (SPT). If the pi l e unit weight i s specified (as CMSA prototype specifies structural member of SSP/SPL), then the dynamic force for a suitable vibratory hammer can be determined, and vice versa. Appendix D. CMSA Partial Listing and Miscellany 332 Line Number Pile Unit Weight (lb/ft) Vibratory Dynamic Forces (Tons) 1 10 0.21* N 2 20 0.50 * N 3 30 0.75 * N 4 40 1.00 *N 5 50 1.20 *N 6 60 1.49* N 7 70 1.72 * N 8 80 2.00* N 9 90 2.33 * N 10 100 2.50 * N 11 110 2.70 * N 12 120 3.03 * N 13 130 3.23 * N 14 140 3.45 * N 15 150 3.85 * N Table D.l Vibratory Pile Drivers Sizing Graph 2 of Barber (1979) presents a linear relationship between the pi l e length and vibratory amplitude. This i s transformed into the following governing equation: Amplitude (in) = Pile Length (ft) * .016 + 0.12 (D.2) Using equation D .2, either the amplitude can be determined given a pi l e length, or vice versa. XMAS treats p i l e length (pile segment) as a variable volunteered by the user. For the CMSA prototype, equation D.l results in dynamic force designation of a vibratory size (where p i l e unit weight i s predetermined by the hypotheses Select_SSP and/or Select_SPL). Equation 2 specifies the vibratory amplitude based on user pile segment size. Crane Selection: Cranes are considered as secondary resources for p i l e driving. Although crane selection depends on several factors, for CMSA prototype development, crane selection i s determined by the weight of the hammer ram weight. Table D.2 presents rules of thumb for selecting cranes designated by their carrying capacity versus the theoretical energy of the impact hammer. For instance, a 35 ton crane may be used to carry a hammer with an upper bound Appendix D. CMSA Partial Listing and Miscellany 333 of 8750 f t - l b , 40 ton crane for hammers within the range (8750 - 15000) f t - l b , and so forth. Cranes Impact Hammer Energy (Ton) (lb-ft) 25 8750 40 15000 60 25000 100 > 25000 Table D.2 Crane Selection Format (From Means Heavy Construction Cost Data 1987) "Crane.nxp" i s the f i l e that contains the crane data base that are used to represent cranes in CMSA prototype. D.4 Unit Cost Quotations Cost quotation sources for the CMSA prototype include cost data manuals, previous projects, interviews, and local vendors. What follow are comments about cost estimates and their break down. Contractor experience from previous jobs was used to set upper and lower bounds for unit cost rates (x $/ft run), or unit cost per surface area (x $ per square foot of steel sheet pili n g ) , and production rate in (day/ft). - The interview with Quadra Construction Co. (see Appendix C) provided experienced based estimates for the crews and equipment involved in the pi l e driving activity. For instance, from the project site v i s i t , the following unit costs were obtained: For sheet p i l i n g driving, the cost of material i s approximately 3/4 of the total cost of pi l e driving operation, whereas 1/4 i s labor cost and equipment (labor cost almost i s almost equal to the equipment cost for a vibratory hammer). Representative costs employed by CMSA are shown below in table D .3. Appendix D . CMSA Partial Listing and Miscellany 334 1. Steel Sheet Piles Unit Cost and H - Piles Grades (40 and 50 ksi) $ 930/Ton 2. Lumber Unit Cost Available Length Size 2 4 $ 0.27/ft 6 to 20 ft 3 4 $ 0.50/ft 6 to 20 ft 4 4 $ 0.75/ft 6 to 20 ft 4 6 $ 1.35/ft 6 to 20 ft 6 6 $ 2.30/ft 8 to 20 ft 3. Impact Hammer Theoretical Rent Energy (lb-ft) SAAH, DAAH 8750 $ 4000/month SAAH, DAAH 15,000 $ 5000/month SAAH, DAAH 25,000 $ 6000/month 4. Crane Rent 25 Ton $5000/month 40 Ton $6500/month 60 Ton $9000/month 100 Ton $11000/month •5. Crew (Pile Driving) Cost 4 men / day $22000/month 6. Vibrator Model Dynamic Force Rent (Tons) Model ICE 812 145.5 $11000/month Model ICE 216 36.4 $7000/month Table D.3 Representative Resources Unit Costs in Vancouver, B.C, [Quadra Construction Co Ltd] In comparison with other shoring methods, for steel sheet pili n g , material cost savings are of major concern. Unit prices are as quoted from the local market (1990) . These prices vary with the amount of material quantity, length and size of the piles/lumber segments, material quantity, and so forth. Appendix D. CMSA Partial Listing and Miscellany 335 D.5 Sample Data Base Files As indicated earlier, these include the design elements for the construction method (structural members (steel sheet piles, soldier piles, struts, wales, and laggings) + construction resources (hammers + cranes). The data bases contain a sample of what could be used. CMSA uses a subset of these data bases. 1. Steel Sheet Piles The following properties and dimensions were taken from Winterkorn and Fang (1975) and are found in database "SSP.nxp". The SSP are divided into three groups according to their section modulus around the X-axis (the assumption used i s that the section modulus rather than interlock strength constitutes the basis for SSP selection including shape. A partial l i s t i n g of those used in CMSA i s show in table D.3. Group_l : This i s mainly Z-section with Section Modulus (S): 46.8 in A3 « S « 30.2 in A3 There are 3 SSP sections in this group. Group_2: This i s mainly invert U-section with Section Modulus (S): 10.7 in A3 « S « 2.4 in A3 There are 2 SSP sections in this group. Group_3: This i s mainly PSA and PSA (straight) sections with Section Modulus (S): 1.9 in A3 « S « 2.4 in A3 There are 4 SSP sections in this group which depend on using them in applications involving interlock strength rather than section modulus. Properties are shown for information purposes. Group_4: Miscellaneous sheet p i l i n g are not included. Properties treated in the data base are as follows ( l i s t i n g D.2) : 1. Designation; 2. Weight_per_foot in (lb); Appendix D. CMSA Partial Listing and Miscellany 336 3. Cross_Section_Area (a) in (in A2); 4. Surface_area in (ft A2/ft) {excludes interlock area}; 5. Driving_width in (in); and 6. Section_Modulus in (in A3). Group_l: Z-Section \SSP_l.Designation\ = "PZ38" \SSP-l.Weight_per_foot\ = "57.00" \SSP_l.Cross_section_area\ = "16.77" \SSP_l.Driving_width\ = "18" \SSP_l.Surface_area\="5.52" \SSP_l.Section_modulus\ = "46.8" \SSP_2.Designation\ = "PZ32" \SSP-2.Weight_per_foot\ = "56.00" \SSP_2.Cross_section_area\ = "16.47" \SSP_2.Driving_width\ = "21" \SSP_2.Surface_area\ = "5.52" \SSP_2.Section_modulus\ = "38.3" \SSP_3.Designation\ = "PZ727" \SSP_3.Weight_per_foot\ = "40.5" \SSP_3.Cross_section_area\="11.91" \SSP_3.Driving_width\="18" \SSP_3.Surface_area\="4.98" \SSP_3.Section_modulus\ = "30.2" Listing D.2 Sample of Steel Sheet Pile Data Base Used in CMSA "SSP.nxp" 2 . Soldiers Piles Structural properties are given for use when Standard HP-Piles are ut i l i z e d as rakes, wales or as other structural members (AISC 1978). The database "HP_Pile.nxp" contains selected soldier pil e members. The normal Material Specification i s : ASTM A36, ASTM A572 grades 42 through 60 (HP 14 * 117 i s not available in grade 60). Those H-piles are available in welded form from Kaiser Steel Corporation. Appendix D . CMSA Partial Listing and Miscellany 337 Note that other steel structural members such as WF-sections could be used. WF-sections have more flange width and thus are more resistant to lateral pressure. For the prototype application, attention was limited to H-piles. Properties treated in the data base are as follows: 1. Designation; 2. Weight_per_foot in (Lb); 3. Cross_Section_Area (a) in (in A2); 4. Surface_area in (ft A2/ft) {excluding interlock area}; 5. Driving_width in (in); 6. Driving_depth in (in); and 7. Section_Modulus in (in A3). \HP-l.Designation\ = "14_117" \HP-l.Weight_per_foot\="117" \HP_l.Cross_section_area\="34.4" \HP_l.Driving_width\ = "14.89" \HP_l.Driving_Depth\ = "14.23" \HP_l.Surface_area\ = "7.11" \HP_l.Section_modulus\ = "173" \HP-2.Designation\ = "14_102" \HP-2.Weight_per_foot\ = "102" \HP_2.Cross_section_area\ = "30.0" \HP_2.Driving_width\ = "14.78" Listing D.3 Soldier Piles Sample Data Base "HPPile.nxp" 3. Struts Properties and dimensions of American-Produced Standard (W) Shapes for Columns (Struts) and Beams (Wales) for Internal Bracing Retaining System (Table 1-22 Manual of Steel Construction, AISC) were used. The W14 series was adopted for the prototype because of the variety available in this size range. Further, Fy = 36 ksi, and Fa = 19 ksi (Winterkorn and Fang 1975). The properties in the data base are as follows. 1. Designation; 2. Weight_per_foot in (Lb); 3. Cross_Section_Area (a) in (in A2); and 4. Radius_of_Gyration r(y) in (in). Appendix D. CMSA Partial Listing and Miscellany 338 \Strut_l.Designation\="W_14_132" \Strut_l.Weight_per_foot\="132.00" \Strut_l.Cross_section_area\="38.8" \Strut_l.Radius_of_Gyration\ = "3.76" \Strut_2.Designation\ = "W14120" \Strut_2.Weightjper_foot\="120.00" \Strut_2.Cross_section_area\="35.3" \Strut_2.Radius_of_Gyration\="3.74" \Strut_3.Designation\ = "W14109" \Strut_3.Weight_per_foot\="109.00" \Strut_3.Cross_section_area\ = "32.0" \Strut_3.Radius_of_Gyration\="3.73" Listing D.4 Struts Sample Data Base "Strut.nxp 1 1 4. Lagging: Structural properties are given for use when timber i s uti l i z e d as lagging, rakers, wales or as other structural members (CSA Standard 1976). Several nominal timber sizes are available such as 2_4, 3_4, 4_4, 4_2, 4_3, 4_6, and 6_6 are stored in "lagging.nxp" database. Properties pertinent to lagging design, within CMSA, are limited to: 1. Designation: Lag_?_?; 2. Lagging_Width, in; 3. Lagging_Thickness, in ; 4. Lagging_Section_Modulus, in A3; and 5. Lagging_Unit_Cost_per_foot, in Canadian Dollars/foot. Appendix D. CMSA Partial Listing and Miscellany 339 \Lagging_l.Designation\="Lagging_2_4" \Lagging_l.Width\ = "4" \Lagging_l.Thickness\="2" \Lagging_l.Section_modulus\="3.06" \Lagging_l.Unit_Cost_per_Foot\ = "0.27" \Lagging_2.Designation\ = "Lagging_3_4" \Lagging_2.Width\ = "4" \Lagging_2.Thickness\="3" \Laggmg_2.Section_modulus\="5.10" \Lagging_2.Unit_Cost_per_Foot\ = "0.50" Listing D.5 Lagging Sample Data Base "Lag.nxp11 5. Hammers Sample Hammer databases are named according to the class of the hammer. For instance, "DAAH" stands for Double Acting Air Hammer; "SAAH" stands for Single Acting Air Hammer, etc. Listing D.6 shows a sample of impact hammers which i s implemented in CMSA. Table 17-16 of Peurifoy (1970) data on pi l e driving hammers has been adopted for the impact hammer data base. Data fields used are:: 1. Ram_Weight in (lb); 2. Stroke_per_minute in (no units); 3. Length of stroke in (in); and 4. Theoretical_Energy (ft-lb) per blow. This database contains single acting a i r hammers of type Vulcan. Appendix D. CMSA Partial Listing and Miscellany 340 \Hammer_01.Hammer_Model\ = "2" \Hammer_01.Ram_Weight\="3000" \Hammer_01.Strokes_per_Min\="70" \Hammer_01.Length_of_Stroke\ = "29" \Hammer_01.Thero_Energy\="7260" \Hammer_02.Hammer_Model\ = "1" \Hammer_02.Ram_Weight\="5000" \Hammer_02.Strokes_per_Min\ = "60" \Hammer_02.Length_of_Stroke\ = "36" \Hammer_02.Thero_Energy\ = "15000" \Hammer_03.Hammer_Model\="0" \Hammer_03.Ram_Weight\="7500" \Hammer_03.Strokes_per_Min\ = "50" \Hammer_03.Length_of_Stroke\ = "39" \Hammer_03.Thero_Energy\="24375" Listing D.6 Impact Hammer Sample Data Base "Hammer.nxp" 6. Vibratory Hammers Sample "Vibro.nxp" i s the database for variety of vibratory pil e drivers. The following different models of vibratory hammers are adopted from Peurifoy (1970) for Foster Vibro driver/extractor ( l i s t i n g D.7). Properties of interest are: 1. Maximum Energy delivered, f t - l b per sec; 2. Vibration frequency, rpm, min; 3. Vibration frequency, rpm, max; 4. Total horsepower; 5. Voltage; 6. Maximum Amplitude; 7. Cycles per sec, and; and 8. Approximate weight, lb. Appendix D. CMSA Partial Listing and Miscellany 3 4 1 \Vibratory_l.Model\="2-17" \Vibratory_l.Max_Energy\="18440" \Vibratory_l.Min_Frequency\ = "1090" \Vibratory_l.Max_Frequency\ = "1290" \Vibratory_l.Tot_Horsepower\="34" \Vibratory_l. Voltage\="440" \Vibratory_l.Max_Amplitude\ = "60" \Vibratory_l.Cycles\ = "60" \Vibratory_l Approx_Weight\="6200" \Vibratory_2.Model\ = "2-35" \ Vibratory_2.Max_Energy\ = "37970" \Vibratory_2.Min_Frequency\ = "890" \Vibratory_2.Max_Frequency\="1120" \Vibratory_2.Tot_Horsepower\="70" \Vibratory_2.Voltage\ = "440" \Vibratory_2.Max_Amplitude\ = "120" \Vibratory_2.Cycles\ = "60" \Vibratory_2 Approx_Weight\ = "9100" Listing D.7 Vibratory Hammers Sample Data Base [Vibro.nxp] 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0050498/manifest

Comment

Related Items