Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Experimental evaluation of the ulag glulam beam simulation program Timusk, Paul Christopher 1997

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

Item Metadata


831-ubc_1997-0292.pdf [ 12.36MB ]
JSON: 831-1.0075268.json
JSON-LD: 831-1.0075268-ld.json
RDF/XML (Pretty): 831-1.0075268-rdf.xml
RDF/JSON: 831-1.0075268-rdf.json
Turtle: 831-1.0075268-turtle.txt
N-Triples: 831-1.0075268-rdf-ntriples.txt
Original Record: 831-1.0075268-source.json
Full Text

Full Text

EXPERIMENTAL EVALUATION OF THE ULAG GLULAM BEAM SIMULATION PROGRAM by PAUL CHRISTOPHER TJMUSK B.ScE., The University of Toronto, 1993 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Wood Science We accept this thesis as conforming to the required standard THE TforVERSiTY OF BRITISH COLUMBIA April 1997 © Paul Christopher Timusk, 1997 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. Department of f~prg5T"r The University of British Columbia Vancouver, Canada Date Apr\ \ 28 , \<\<\^ DE-6 (2/88) ABSTRACT An experimental study on the bending capacity of glued-laminated beams was undertaken to facilitate the verification of the stochastic finite element beam simulation program ULAG (Ultimate Load Analysis of Glulam). For the material data base, 600 tension and stiffness tests were done on two different grades of machine stress rated laminating stock, consisting of 2x6 (38mm x 140mm) spruce and pine lumber. Thirty 16-foot (4.88m) glulam beams of eight 2x6 laminations each and thirty 24-foot (7.31m) glulam beams of twelve 2x6 laminations each were made in a commercial glulam plant and tested in bending to failure on a 200 kip universal testing machine. The mean modulus of elasticity values and ultimate bending loads were compared with analytical predictions. The laminating stock test specimens of 100 16-foot (4.88m) 'B\ 100 16-foot ' C \ 100 8-foot (2.44m) 'B', 100 8-foot ' C \ 100 8-foot (2.44m) finger jointed 'B' and 100 8-foot finger jointed 'C' 2x6 (38mm x 140mm) lumber were E-rated using a Cook Bolinders AGAF machine, and tested in tension parallel to grain to failure on a Metriguard tension testing machine. After verification with the physical test data, ULAG was used to investigate various glulam beam properties such as size effect and the effect of different beam lay-ups. u T A B L E O F C O N T E N T S ABSTRACT ii TABLE OF CONTENTS iii LIST OF TABLES vi LIST OF FIGURES vii ACKNOWLEDGMENTS x UNITS xi 1. INTRODUCTION 1 1.1 WHY GLULAM 1 1.2 PREVIOUS RESEARCH 2 2. OBJECTIVES 4 3. ULAG 5 3.1 PURPOSE 5 3.2 BACKGROUND 6 3.3 ULAG INPUT 7 3.4 ULAG OUTPUT 8 4. MATERIALS 9 4.1 LUMBER 9 4.2 LUMBER SORTING 11 4.3 GLULAM BEAM MANUFACTURE 12 5. TEST PROCEDURES 17 5.1 16-FOOT BEAMS . 17 iii 5.2 24-FOOT BEAMS 23 5.3 E-RATING 24 5.4 TENSION TESTING 29 6. RESULTS/DISCUSSION 29 6.1 GLULAM BEAMS 29 6.1.1 FAILURE TYPES 36 6.1.2 FAILURE PATTERNS 41 6.1.3 RATE OF LOADING 44 6.1.4 MOISTURE CONTENT 45 6.2 TENSION TESTS 46 6.2.1 LENGTH EFFECT 53 6.2.2 GRADE EFFECT 57 6.2.3 FINGER JOINT FAILURES 58 6.3 ULAG ANALYTICAL SIMULATIONRESULTS 61 6.3.1 ULAG VERIFICATION 61 6.3.2 ULAG SIMULATIONS 66 BEAM LAY-UP EFFECT 66 BEAM SIZE EFFECT 70 7. CONCLUSIONS/RECOMMENDATIONS 75 7.1 SUMMARY 75 7.2 CONCLUSIONS 76 7.3 RECOMMENDATIONS 77 iv REFERENCES 78 v LIST O F T A B L E S 6.1 Beam Test Summary 30 6.2 Tension Test Failures 51 6.3 Tension Test Failures 52 6.4 Tensile Test Summary Data 54 vi LIST OF FIGURES 4-1 Project Procedure Flowchart 10 4.2 Glulam Beam Manufacturing Process 13 4.3 Glulam Clamping During Manufacture 15 4.4 Glulam Finishing 15 4.5. 16-Foot and 24-Foot Glulam Beams 16 4.6 Glulam Beam Lay-ups 16 5.1 Tinius Olsen Testing Machine 18 5.2 Tinius Olsen Testing Machine 18 5.3 Beam End Support 19 5.4 Beam Test Set-up 19 5.5 Beam Test Load Cell 20 5.6 Beam Test L V D T and Support Bar 20 5.7 General Beam Failure 22 5.8 Weighing of Beams 22 5.9 Cook Bolinders A G A F E-rating Machine 25 5.10 Cook Bolinders Input-End View 25 5.11 Cook Bolinders Schematic 26 5.12 Metriguard Tension Testing Machine 27 5.13 Metriguard Tension Testing Machine 27 5.14 Tension Test Set-up 28 6.1 Distribution Function of Beam Capacities 31 Vll 6.2 CDF of 16-Foot Beam Max. Loads 32 6.3 CDF of 24-Foot Beam Max. Loads 33 6.4 CDF of 16-Foot Beam MOEs 34 6.5 CDF of 24-Foot Beam MOEs 35 6.6 Sloped Grain Effect 38 6.7 Beam Fingerjoint Failure 39 6.8 Beam Fingerjoint Failure 39 6.9 Beam Shear Failure 42 6.10 Beam Compression Failure 42 6.11 Tension Test Knot Failure 47 6.12 Tension Test Sloped Grain Failure 47 6.13 Tension Test Tensile Failure 48 6.14 Tension Test Shear Failure 48 6.15 Tension Tested Lumber 50 6.16 Tension Test CDF Plots 55 6.17 FJ ' C Tension Test CDF Plot 59 6.18 FJ 'B' Tension Test CDF Plot 60 6.19 Comparison of ULAG and 16-Foot Beam Tests _ 63 6.20 Comparison of ULAG and 24-Foot Beam Tests 64 6.21 Comparison of ULAG and 24-Foot Beam Tests, No FJ 65 6.22 ULAG Lay-up Effect 67 6.23 Lay-up Effect Plot 68 viii 6.24 ULAG Size Effect 71 6.25 Size Effect Plot, Max. Load , 72 6.26 Size Effect Plot, Extreme Stress 73 ix ACKNOWLEDGMENTS The author would like to gratefully thank his supervisor, Dr. Helmut G.L. Prion for his continued support and.guidance throughout his masters and this thesis, without which it would not have been possible. Special thanks to J. Weider and P. Symons for their help throughout the testing phase of the project, to Dr. R. Foschi, Dr. F. Lam and Dr. J. Biernacki for their tireless question answering and to his classmates for their insight and support. Financial support for this project was provided by Western Archrib and the Canadian Wood Council. Last but not least, special thanks are extended to the author's family for financial and moral support; especially his father for guidance and his wife Karen for putting up with a graduate student; and to Kiisu for serving as a constant source of distraction and procrastination. x UNITS Metric units are used throughout for the reporting and discussion of experimental and analytical data. For convenience to the reader and to conform with common terminology, lumber sizes and lengths are referred to by their nominal size designations in imperial units. To ease the conversion, both units are given where appropriate. xi 1. INTRODUCTION 1.1 WHY GLULAM As a result of increasing public pressure, changing laws and continued harvesting, the supply of large beam quality logs is diminishing. At the same time, there is an ever increasing demand for housing and other structures to satisfy the needs of the exponentially growing world population. As a result, more and more pressure is exerted on the forest industry to change to more efficient production methods in order to provide wood products from an ever decreasing wood supply. One solution to the problem is to "re-constitute" natural wood in the form of wood composites or engineered wood products. Glued laminated members (or glulam) are an example of such a wood product. The production of glulam was brought on by the lack of large cross-sectional members without major strength reducing defects. Thus the idea was born to glue small plank sections together to form large members of any size or shape, as required. Glulam is a more uniform and predictable product than solid sawn timber. • It optimizes the use of its constituent lumber components by placing higher grade material in high stress zones, while making more efficient use of logs and producing less waste. Increased uniformity and reliability are achieved by the randomization of the locations of knots and other strength reducing defects inherently present in timber. The act of sawing logs into boards, finger jointing the boards end to end, cutting the now continuous board into prescribed lengths and then gluing them one on top of another to form a beam, randomizes the defects. This procedure lowers the probability of several critical defects occurring at the same location. 1 The glulam manufacturing process also allows for the optimal placement of the different grades of lumber. High grade and high strength lumber can be used in high stress concentration areas, such as for the bottom laminates of a simply supported beam where tensile strength is most important. At the same time lower grade and lower strength lumber can be used in low stress areas where strength reducing defects are not as critical. Glulam also makes it possible to produce large beams from small trees which would otherwise only be suitable for making 2x4's (38mm x 89mm)or 2x6's (38mm x 140mm). The sizes and shapes of glulam beams are restricted only by the size of the manufacturing facility and its equipment. Finger jointing of the lamstock end to end allows for the removal of severe strength reducing defects and makes it possible to produce any length of beam. Face and edge gluing allow for the manufacture of any width or depth of beam desired. This in turn increases the recovery factor for producing beams from logs. 1.2 PREVIOUS RESEARCH In recent years, a small amount of other research has been conducted in an attempt to develop models for the prediction of strength properties of glued-laminated beams. Hilson et al. (1979) developed two numerical models for calculating the deflection of glulam beams which allow for non-conformities in the properties of individual laminations. One model is of the finite element type, and the other is of the finite interval type. The authors found that the predicted results from the two models compared well with actual beam deflection tests. The numerical models also indicated that British practice at the time over-2 estimated glulam beam deflections by approximately 17%. The models have also been used to study the sensitivity of beam deflection to glue line properties and lay-up. In 1980, Foschi and Barrett developed a computer simulation model called GLULAM to predict the strength and stiffness of glulam beams in either bending, tension or compression. The program uses a finite element approach to estimate the variability in the strength and stiffness of beams. The input data required are the average modulus of elasticity (MOE), tensile strength, and compression strength distributions of the laminating stock as well as the distribution of knots and the availability of lengths for each grade of lumber. Also required are the beam span, width, depth, lay-up, minimum distance between end joints in adjacent laminations, and loading and support conditions. In the modelling process, each beam length lamination is divided into a series of six inch (152.4mm) long elements, which are then assigned an MOE value, a tensile strength value and a compressive strength value. The model assumes that tensile failures will occur at knots, and that tensile tests for each grade have been conducted in which the knot size has been noted at each failure. It then links tensile strength to knot size for each grade through a mathematical model. The model also assumes that the localized modulus of elasticity can be determined as a function of knot size or tensile strength. In the analysis it is assumed that the glulam beam behaves elastically to the first failure, and that the first failure will occur when the tensile strength of any element parallel to grain is exceeded. An extended version of the program also exists called PGLULAM, which is capable of studying progressive beam failure. 3 The main weakness of the model is its requirement for detailed information on the distributions of knots and the method in which it determines localized modulus of elasticity values. For practical applications, the detailed knot information is too difficult to obtain. The availability of localized moduli of elasticity data along the length of a board would allow for a major improvement in the model. It would eliminate the need for detailed knot surveys and "increase the accuracy of the model by improving the precision of the stress distribution obtained from the finite element analysis" (Foschi, 1985). Folz (1997) subsequently undertook to expand the capabilities of the GLULAM program in that only tension strength and stiffness properties of the laminating stock would be required for the simulation process. A more detailed description of the ULAG (Ultimate Load Analysis of Glulam) program is provided in chapter 3. 2. O B J E C T I V E S The primary objective of this research project was to produce the physical data necessary for the verification of the ULAG (Ultimate Load Analysis of Glulam) glulam beam structural analysis computer program. Secondary objectives were: 1. To create a physical test data base for a population of glulam beams and corresponding population of lamstock to facilitate the study of the correlation of their strength properties. 4. 2. To study the properties of various grades and lengths of finger jointed and non-finger jointed lamstock, E-rated and then tested in tension parallel to grain to failure. 3. To study the properties and failure mechanisms of glulam beams tested in bending to failure. 4. To use ULAG to investigate the effects of beam size, lay-up and various other parameters on the strength of glulam beams. 3. T H E U L A G C O M P U T E R P R O G R A M 3.1 PURPOSE The purpose of the ULAG (Ultimate Load Analysis of Glulam) glulam beam structural analysis computer program is to serve as a glulam beam design tool for the manufacturing industry and, to some extent, design engineers and researchers. It will help reduce the amount of inherent overdesign and consequent waste of materials in the glulam industry by more accurately predicting the specific beam requirements for each loading situation. It will at the same time increase the level of confidence by reducing the chance of beams being underdesigned. ULAG can also be used to establish design criteria for inclusion in the code by quantifying the effects of: 1. Various beam lay-up configurations with different species and grade mixes 2. Different lamination lengths and thus number of finger joints 3. Various beam loading configurations 4. Size effect 5 5. Effect of proof loading finger joints and E-rating lam stock before beam manufacture. 3.2 BACKGROUND The ULAG glulam beam structural analysis computer program was developed at U.B.C. by Folz and Foschi (Folz, 1997). The program predicts the load carrying capacity and stiffness of glulam beams of various dimensions, lay-ups and load conditions. The program requires as input the strength and stiffness properties of the laminating material and the fingerjoints. The program assumes the load-carrying capacity to be governed by tensile fracture within the lamination material and / or at fingerjoints between lamination segments. The glue bond between laminations is assumed not to be critical. The program then proceeds to "build" as many glulam beams as specified by the user as the simulation size. The beams are "built" in much the same way as in industry, by assigning strengths, stiffnesses and lengths to laminates as they are used with fingerjoints where they are needed. It then uses a stochastic finite element analysis, combining a one-dimensional finite element model with a stochastic model for the strength and stiffness of the lamination material. Wood is an anisotropic engineering material, exhibiting spatially varying random strength and stiffness properties. These random variations are often of a sufficient scale that they become evident at the component level within a structural system. For this type of problem, a deterministic or conventional statistical structural analysis approach may prove 6 inadequate. The stochastic finite element method takes into account these uncertainties in a structural system, and determines the statistical characteristics of the response under deterministic or random loading. The tensile strength model component of the program considers both the variability between the different lamination segments as well as along their length. Material strength is assumed to be governed by a maximum tensile stress failure criterion under which brittle fracture occurs wherever the tensile strength is exceeded. Post-failure material behaviour results in a reduction in the tensile strength and stiffness to zero at the damaged locations. At a given level of loading, this localized damage may not result in the total failure of the beam and further load increase may be possible. The damage may be arrested through stress redistribution because strength and stiffness varies spatially along the lamination lengths, as well as through the cross-sectional lay-up of the beam. If the reduction of stiffness is too severe, however, a redistribution of stresses may not be possible and beam collapse will result. 3.3 ULAG INPUT For each specific species group and grade, the MOE and tensile strength must be known. The MOE can be obtained from a bending test over a known span, for example, from a Cook Bolinders type E-rating machine averaged over the board length, or with a hammer type impact grading machine. The tensile strength data must come from destructive tensile testing conducted in the parallel to grain direction, with the longest test span possible to avoid 7 interfering effects from the machine grips. A random test population of 100 or more boards (n >100) is typically required for each different lamstock grade or species group. Fingerjoint maximum tensile load data is also required by ULAG for each type of fingerjoint and for each different size in cross sectional area of lamstock fingerjointed. The fingerjoint tensile tests are to be carried out parallel to grain as are the other tensile tests. The fingerjoints to be tested must be a random sample from the plant where the beams are made. The proportions in which the lengths of each individual grade of lamstock are available to the glulam plant are required by ULAG for determining the number and location of fingerjoints. ULAG also needs the dimensions of the beam to be simulated, the kind of lay-up or grade, the laminate thickness, support conditions, the loading configuration for the given application, the number of simulations to be run and the element length to be used in the analysis. After the input is completed, the beam with the appropriate lamination lay-up with color-coded grades, and the loading and support conditions specified are displayed on the computer screen. This serves to check if the input information is correct, and allows for changes to be made. If changes are requested, the program takes the user back to the step to be changed, until the input is correct, after which the simulations are run. 3.4 ULAG OUTPUT After ULAG has successfully performed the requested number of beam simulations and loadings, it generates a screen of summary data consisting of: 8 1. Maximum beam deflection data 2. Mean load factor applied to nominal loads to produce failure data 3. 2-P Weibull fit to the load factor cumulative distribution function (CDF) 4. Computer run time 5. Statistics for each set of the many beam simulations (Maximum and minimum load factors, coefficient of variation (C.O.V.) and standard deviation). 6. A plot of the lower 25th percentile of the load factor (the factor applied to the nominal load to produce a failure) C.D.F. 4. M A T E R I A L S A flowchart depicting the allocation of materials to the various aspects of the project and the main steps involved is presented in Figure 4.1. All of the materials for this project were provided by the Edmonton plant of Western Archrib, who also fabricated the glulam beams for testing. Quality control and source identification were done by employees of the company. 4.1 LUMBER All the nominal two by six (38mm x 140mm) lamstock material provided for this study was spruce-pine-fir (SPF) B and C grade lumber produced by the same sawmill, which obtained its sawlogs from northern Alberta. The species mix consisted of approximately 99% spruce and 1% lodgepole pine, as determined by one of the graders at Western Archrib. The lumber was sawn, kiln dried and planed at the sawmill. 9 Fig. 4.1 Project Procedure Flowchart Lumber Sort by MOE Beams 30 16-Foot 30 24-Foot Bending Test Bending Test 150 16-foot 'B' Cook Bolinders 100 8-Foot 'B' 100 16-Foot'B" Test ing 150 16-foot "C Cook Bolinders 100 8-foot ' C 100 16-foot •C Tension Test Tension Test 100 8,footFJ 'B' 100 8-foot FJ ' C Tension Test 10 After shipping to Western Archrib, the lumber was graded and E-rated by graders at Western Archrib. An E-rating spot check was conducted by the U.B.C. researchers with a portable Metriguard E-rating apparatus. The lumber for tension testing was again E-rated later at U.B.C. with the more accurate Cook Bolinders AGAF machine. The moisture content was checked at Western Archrib by an in-line Wagner moisture meter, which was in turn calibrated with a Delmhorst two-pin moisture meter. The moisture content range required was 9% to 15%. If the lumber was too wet, the moisture meter would mark the board with paint, and it would be pulled from the line and re-dried in a kiln to the desired moisture content. The thickness of the lumber was also checked before beam manufacture. This was done manually with a metal gauge with an opening cut to the desired board thickness. The desired thickness tolerances were +/- 1 mm (Fargey, 1995). After grading and before sorting, 100 'B' and 100 ' C grade boards were randomly sorted out to be cut in half and fingerjointed back together, using the standard procedure at Western Archrib. These were later tension tested at U.B.C. 4.2 LUMBER SORTING At Western Archrib, the 'B' and ' C grades of SPF nominal 2x6 lumber were each separated into two piles, one for MOE / tensile testing designated as the 'test' pile and the other for beam fabrication designated as the 'beam' pile. For the purpose of this study, the test and beam piles of lumber had to be statistically similar. The MOE readings inscribed on the boards were used as a basis for the sorting. To achieve this, the lumber had to be 11 physically sorted through two times. On the first sort, the MOE values from the Western Archrib hammer impact type grading machine, written on each board by the graders at Western Archrib, were read out and entered into a computer. An identification number was assigned to and written on each board. On the computer, the boards were then ranked according to their MOE values and divided into two statistically similar groups. On the second sort, as the number of each board was read out, it was determined from the computer read-out whether it was to be placed in the test pile or into the beam pile, resulting in two piles of lumber with statistically similar E-distributions. Although the boards could have simply been randomly divided into two groups, there would not have been any guarantee that the piles would be statistically similar. By ranking the boards and then sorting them, it was assured that the two groups would be as statistically similar in MOE as possible. 4.3 GLULAM BEAM MANUFACTURE Thirty 16-foot and thirty 24-foot glulam beams were manufactured at Western Archrib, according to the recognized glulam manufacturing standards CSA 0.122 and CSA 0.177. Figure 4.2 is a simplified flowchart of the general glulam beam manufacturing process (Canadian Wood Council, 1994). After the grading procedures, the various quantities and grades for a given group of beams were selected and passed through a saw which trimmed off the ends of boards in the case where a knot was within three knot diameters of the end, since such a defect would significantly affect the strength of the fingerjoints. Next, the boards passed through the 12 Figure 4.2 Glulam Manufacture Lamstock receiving and storage and finishing 13 fingerjointer, which in this case cut fingerjoints of the vertical type. The fingerjoints were then coated with phenol resorcinol formaldehyde resin, and fitted end to end to form a continuous piece of lumber. The continuous piece then passed through a curing tunnel where the fingerjoints were radio frequency cured for 15-20 seconds at a temperature of 90° C. The continuous piece was then cut to the desired lamination length with an automated flying cross-cut saw. The saw continuously measured the amount of lumber passing make the cuts afthe correct locations (Fargey, 1995). The laminates passed through a curtain type glue spreader which coated the top of each board with an LT-75/FM282 phenol-resorcinol-formaldehyde resin/catalyst mixture from Borden. After coating, the laminations were laid up in a jig in the appropriate configuration and clamped together to cure at room temperature (Fig. 4.3). Typical curing time was approximately nine hours at 21°C with a clamp pressure of 100 psi (690 kPa) as measured with a torque wrench. Open and closed assembly times vary depending upon the environmental conditions of the gluing area on the day of gluing, but remain within the listed requirements of the resin manufacturer (Fargey, 1995). After curing, each beam was passed through a planer and then sanded to remove any roughness (fig. 4.4). The beams were then marked with the grade and up-side and wrapped in a waterproof material for shipping. The thirty 16-foot (4.88m) beams were made of nominal 2x6 (38mm x 140mm) ' C grade lumber only, eight laminations deep each, with no fingerjoints (fig. 4.5 and 4.6). 14 FigM: Glulam Brushing. 15 BEAM AND TEST CONFIGURATIONS: Figure 4.5 16-Foot and 24-Foot Glulam Beams 16-Foot Beams PI P2 — =zA x— X X 1577mm 1577mm 1577mm -w 127mm 305mm 24-Foot Beams PI P2 J r i r 457mm ' -X -X-2712mm 1626mm Figure 4.6 Glulam Beam Lay-ups ~X 127mm 2712mm c c c c c c c c c c. c c c c. c c R c R 16-foot 24-foot 16 The thirty 24-foot (7.31m) beams were made with two nominal 2x6 (38mm x 140mm) 'J3' grade laminations on the bottom, and ten nominal 2x6 (38mm x 140mm) ' C grade laminations on top, with fingerjoints occurring randomly (fig. 4.5 and 4.6). All beams were planed and sanded after gluing, resulting in a final average width of 5.0 inches (127mm). For both beam types, no design stress values were given because the given grade combinations had not been previously used or tested. As a quality control measure, Western Archrib periodically conducts several tests. Shear, block tests are conducted by taking shear block samples from the first, middle and last beam of every set of beams manufactured. This test is to ensure sufficient bonding between laminations. Cyclic delamination tests are conducted every six months by a testing agency according to CSA standards. Fingerjoints are tested every 3000 lineal feet (914m). They must average at least 3.0 times the allowable unit stress in bending for the given species, and no single test value is allowed to fall below 2.0 times the allowable unit stress in bending (Fargey, 1995). 5. TEST PROCEDURES 5.1 16-FOOT BEAMS All 16-foot (4.88m) beams were tested on the Tinius Olsen 200 kip universal screw type testing machine in the Civil Engineering Structures Lab at UBC (Fig. 5.1 and 5.2). The test span was 4.731m (Fig. 4.5). Loads were applied at third points through a pivoting spreader bar with pivoting feet at each end (Fig. 5.3, 5.4 and 5.5). The purpose of the pivots was to eliminate secondary forces in planes other than the vertical. Between each pivoting support and the test specimen was a steel bearing plate, 150 mm x 252 mm x 13 mm, 17 Fig.52: Tinius Olsen testing machine. 18 Fig. 5.3 Beam End Support GLULAM BEAM END ROLLERS 3 H H LINE LVDT HOLLOW STEEL BAR Fig5^: LVDT mounted on steel bar, on glulam beam. TJO oriented with the plate long axis perpendicular to the long axis of the test specimen. Under each end of the test specimen was a similar bearing plate (Fig. 5.5). Under one end was a 50 k N load cell with a roller plate between the load cell and the bearing plate (Fig. 5.5). Under the other end was a pivot with a bearing plate. The purpose of the steel bearing plates, on top of and below the beams was to eliminate localized crushing due to load concentrations. Load cell readings were compared with the load indicating dial of the testing machine, which is attached to a scale measuring the total force exerted on the test bed. The beam center deflection with respect to the support points was measured with a linear variable differential transformer (LVDT) mounted at mid span on a reinforced tubular steel bar (Fig. 5.4 and 5.6). The bar rested on bolts drilled into the test specimen above the reaction points at each end. The end of the line from the LVDT was then attached to the test specimen with a short steel bar. This arrangement provided a stationary reference point from which beam deflection could be measured while eliminating deflections caused by beam crushing or other, factors which could affect the beam deflection readings. The load cell and the LVDT were connected to a computer for data acquisition. Data points were taken at set intervals of 2 seconds. As each beam was being tested, the corresponding load - deflection curve was plotted on the monitor of the computer. Before testing, the moisture content of each beam was measured with a Delmhorst Instrument Co. RC-1C portable moisture detector. Readings were taken once in each lamination, approximately one meter in from one end, at a probe depth of approximately 25 mm. The moisture detector was calibrated before testing began. 21 Fig.58: Weighing of glulam beam. 23-The weight of each beam was measured with a 300 lb. (136 kg) Accu-weigh hanging dial scale which was suspended from the overhead crane hook with a steel cable (Fig. 5.8). The length of each beam was measured along its top with a tape measure as it was set to be tested on the Tinius Olsen testing machine. The width and depth of each beam were measured at approximately 1.5 meters from each end with a ruler. All specimens were tested to failure at a loading rate of approximately 0.3 mm/sec or 18 mm/min, resulting in a time to failure of approximately five to ten minutes in each case. 5.2 24-FOOT BEAMS The test procedure for the 24-foot (7.31m) beams was the same as for the 16-foot beams, except for the following differences. The test span was 7.050m. Loads were applied with the same spreader bar, but at a distance of 1626 mm apart (Fig. 4.5). Also, two 50 kN load cells were used, one at each end, rather than one load cell, because the scale on the Tinius Olsen could not be used as a check for such long span specimens. The same top bearing plates were also used, only they were oriented with the plate long axes parallel to the long axis of the test specimen. The 24-foot beams were weighed with the same 300 lb. (136 kg) Accu-weigh hanging dial by hanging one end of the beam on the scale, while placing the other end on a pivoting support, and then multiplying the measured weight by two. This procedure was necessary since the 24-foot beams were too heavy for the 300 lb. (136 kg) capacity scale. 23 5.3 E-RATTNG All 16-foot (4.88 m) 2x6 (38 mm x 140 mm) 'B' and ' C grade lumber was "E-rated" using a Cook Bolinders AGAF machine (Fig. 5.9 and 5.10). One operator rolled a board into the machine edgewise, painted end first. The board passed through the machine, and the other operator guided it out the other end, dampening interfering vibrations as much as possible. The board was then rotated 180° about it's longitudinal axis and passed back to the first operator to be run through the machine painted end first again. In this way, the MOE relative to each of the two faces of each board was found, and could later be averaged to obtain an average MOE for each board. Readings were taken every 2 mm along the length of each board, excluding the first and last 450 mm of each. The clamp pressure for the feed rollers was set at 2 bar, and the off-set was set at 4.5 mm. The span between the first and last roller was 900 mm, and the feed rate was approximately 300 mm per second (Fig. 5.11). As each board passed through the machine, the load applied by the middle roller required to achieve the 4.5 mm deflection was plotted against the coordinate along the length of the board on the monitor of the PC collecting the data, and then saved on disc. Before tension testing, the 150 16-foot 'B' grade and 150 16-foot ' C grade boards had to be sorted again according to MOE to determine which 50 of each grade were to be cut in half to make the 100 8-foot (2.44 m) 'B' grade and 100 8-foot ' C grade boards. The sorting was done in the same way as the first sort on the computer, only this time the MOE values used were from the Cook Bolinders AGAF machine at U.B.C. The boards selected to be cut were sawn in half and labeled 'A' and 'B' for later identification. 24 Fig5lo: Cook Bolinders input end view. IS Figure 5.11 Cook Bolinders Schematic COOK BOLINDERS AGAF MACHINE ELEV. I 450 m m — 4 5 0 m m — H 26 Metriguard tension testing machine. 2.7 Figure 5.14 Tension Test Set-up 8-Foot 16-Foot '- - '- 7 i J 610 mm 1220 mm 610 mm grip gnp \ — J I , , 1 •> 610 mm 3658 mm 610 mm 28 5.4 TENSION TESTING All 'B' grade and 'C' grade, fingerjointed and non-fingerjointed 2x6 (38 mm x 140 mm) lumber was tested to failure in tension parallel to grain in a Metriguard tension testing machine (Fig. 5.12 and 5.13). The grip length at each end was set at two feet (610 mm) to avoid any slipping of the lumber in the grips, leaving four feet (1.22 m)between grips for the eight-foot lumber, and twelve feet (3.66m) between grips for the sixteen-foot lumber (Fig. 5.14). The loading rate was set to result in a time to failure of at least five minutes, which is generally accepted as a quasi-static loading without severe load-duration effects. A few boards with strength far below the average load, failed in less than five minutes. The moisture content of each board was measured with the same Delmhorst portable moisture detector as used for the beams. The width and depth of each board were also measured, using an electronic micrometer. (The cumulative probability distributions of maximum tensile load values for each set of 100 boards are shown in Figures 6.16 to 6.18). 6. R E S U L T S A N D D I S C U S S I O N 6.1 GLULAM BEAMS Test results are summarized in Table 6.1 and plotted in Figures 6.1 to 6.5. The average modulus of elasticity (MOE) and average maximum failure load for the 16-foot (4.88m) glulam beams were respectively 9936 MPa and 87.65 kN. The average modulus of elasticity (MOE) and average maximum failure load for the 24-foot (7.31m) glulam beams 29 Table 6.1 Beam Test Summary 16-FOOT BEAMS i ! | 24-FOOT BEAMS : MOE ; MOR ! Max Load ! MOE MOR Max Load Beam No. I (MPa) (Mpa) i (kN) j Beam No. (MPa) (Mpa) (kN) 16-1 | 10970 j 35.01 | 85.96! 24-1 9358 37.05 124.52 16-2 j 10995: 33.24J 82.68} 124-2 10205 45.81 149.02 16-3 ! 10550 ! 30.67 I 76.10! !24-3 10160 37.89 124.78 16-4 ! 10812 ! 48.22 ! 119.70! 124-4 9778 32.46 107.12 16-5 j 9638| 23.55 ! 58.66| J24-5 10279 48.58 158.66 16-6 9688 47.47 i 118.08; ;24-6 10058! 42.95 140.82 16-7 9592 35.37; 87.82J 124-7 10486! 32.88 109.70 16-8 9941 40.54 1 101.64; J24-8 10014 ' 39.43 130.40 16-9 9947 37.63 93.04i 124-9 9611 40.72 133.50 16-10 * 31.37 77.84) 124-10 10419 44.83 147.30 16-11 ; 10007 38.56 95.30! ! 24-11 9866 36.48 119.80 16-12 10147 38.59 79.60! 24-12 9309 34.00 112.50 16-13 9413 24.51 60.541 24-13 10670 41.79 136.40 16-14 9456 33.18 82.22! 24-14 9607 35.36 116.70 16-15 10039 39.35 98.28| 24-15 9828 39.56 128.40 16-16 9974 41.55 102.40; 24-16 9981 35.38 115.50 16-17 9900 42.02 104.26! 24-17 10231 33.12 107.50 16-18 10235 39.74 98.28! 24-18 10216 34.51 112.00 16-19 9369 32.60 80.34 j 24-19 10626 43.13 141.40 16-20 9710 26.59 65.76! 24-20 9752 36.25 119.50 16-21 9505 24.76 61.661 24-24 9964 42.27 139.20 16-22 10099 41.00 102.40 24-22 9827 34.74 114.40 16-23 9486 31.65 79.22! 24-23 9974 37.65 124.80 16-24 10567 36.27 89.68i 24-24 10221 36.51 120.50 16-25 9581 38.74 95.66! 24-25 10098 38.92 129.30 16-26 9808 35.58 88.56; 24-26 , 8510 31.97 87.76 16-27 9480 27.62 68.761 24-27 10059 36.71 121.40 16-28 9255 36.06 90.06! 24-28 | 9984 38.49| 127.30 16-29 10082 36.51 91.18| 24-29 | 9965 42.58! 139.90 16-30 9901 37.56 93.801 24-30 j 9937 36.86| 119.11 Mean: 9936 35.52 87.651 Mean: 9966 38.30| 125.31 Std Dev: 474 6.18 15.38! Std Dev: 421 4.22 14.84 * Missing value 30 16-FOOT BEAM MAXIMUM LOAD DISTRIBUTION >-o UJ z> a UJ cc 1 r o CD O m o i o CD O 00 O O O) I o 00 o o o o CM O o MAXIMUM LOAD (kN) 24-FOOT BEAM MAXIMUM LOAD DISTRIBUTION >-o LU Z> o UJ al 9 8 -7 -6 5 4 3 2 1 0 o o o o o o o o o CN •sr m CD o ^ 00 o o o o o o o o CM CO sr i n MAXIMUM L O A D S (kN) Fig. 6.1 Distribution Funct ion of B e a m Capaci t ies . 31 CO LU O < LU m r -o o CN] Q O o O O LO o o LO r--ns 0. LU o o LO CM V) LU o CO CD CO o o S J -CM Q o LO CO o LO oo o LO CNJ co o o o o o A;;i;qeqoJd aAueiniuno 35 were respectively 9966 MPa and 125.31 kN. The modulus of rupture and failure loads are well represented by two parameter Weibull distributions. All glulam beams were tested in bending to failure. Many different failure types or modes were observed. However, most of the failures occurred so fast that it was difficult in most cases to conclude exactly which mode of failure was the primary cause of failure. A post-mortem analysis did not always provide conclusive evidence of the cause of failure either. For example, if a beam failed in tension in the bottom two laminates, at a zone of sloped grain in the bottom lamination and at a knot directly above in the second lamination, it was virtually impossible to determine which of the defects was the primary cause. Most failures seemed to be governed by tension fracture in the bottom laminations, initiated in the region of maximum bending stress between the two loading points. Failures were typically initiated at or propagated through knots, regions of sloped grain, or fingerjoints in the 24-foot beams. In some cases, however, failures were initiated in areas of apparently clear wood. Data on each individual glulam beam including load-deflection curves, exact dimensions, weights, moisture content levels and photographs of each beam are presented in "Glulam Beam Testing Program, Progress Report #1" by Timusk et al. (1994). A video recording of each beam test is also available. 6.1.1 FAILURE TYPES The failures at defects such as knots or sloped grain can be explained by considering the combination of wood fibre geometry and the stress field in the beam. The grain in these 36 cases deviates from being parallel to the length of the beam (x), resulting in a vertical (y) and/or horizontal (z) stress component(s) (Fig. 6.6). The principal bending stresses in the beam are parallel to the longitudinal axis and are highest in the fibres closest to the bottom of the beam, parallel to the beam's long axis. Consequently, for sloped grain, stresses result in tension perpendicular to grain and in shear parallel to grain, both of which are weaknesses in wood. The highest degree of anisotropy in wood is in tension, at 48:1 for Douglas fir (Dinwoodie, 1979). This means that in tension parallel to grain, it is 48 times stronger than in tension perpendicular to grain. Even a slight deviation in angle of grain can therefore result in significant strength reductions. The approximate strength value at any angle can be calculated with the Hankinson formula, provided the values at both parallel and perpendicular to grain are known: ae = (ap aq) / (ap sin" 9 + a q cos" 9) where: ae is the strength property at angle 9 from the longitudinal or parallel to grain a p is the strength parallel to grain a q is the strength perpendicular to grain n is an empirically determined constant (n - 1.5-2 in tension) Failures at fingerjoint locations are most often attributed to the fact that the tips of the fingers do not penetrate the entire depth of the grooves between the fingers on the opposite board (Fig. 6.7 and 6.8). This results in a row of "perforation" type voids and a weak plane, along which a failure can propagate. Fingerjoint failures could also be attributed to inadequate bonding between fingers, which could be caused by a number of reasons such as too thin a resin line, insufficient curing time, a too low curing temperature, insufficient 37 Fig. 6.6 Sloped Grain Effect LOADING ON B E A M ^ a) MOMENT DIAG b) BENDING STRESSES IN BEAM c) NEUTI#\L_ AXIS "!" COMPRESSION TENSION STRESS O N SLOPED G'RAIN d) 38 Fig. 18: Glulam beam fingerjoint failure underside view. 31 pressure during curing, glazed finger surfaces from dull cutting knives, or problems with the resin itself. If no-other reasons can be found to explain insufficient fingerjoint strength, the fingers could simply be too short due to an attempt to save wood, resulting in an inadequate bonding surface area. The areas of apparently clear wood where failures have initiated could contain other defects such as reaction wood, juvenile wood or pitch pockets, which are difficult to detect with the naked eye, even after testing. Reaction wood (called compression wood in softwoods and tension wood in hardwoods) is denser and more brittle than normal wood. As a result of being denser, it can cause abnormally high and uneven shrinkage along the grain during drying, which in turn can cause stress concentration and cross breaks, resulting in cracking and weak areas in the wood (Hoadley, 1990). Compression wood is also weaker than normal wood in MOR, MOE bending, compression parallel to grain and in tension parallel to grain. As an example, the following numbers compare compression wood to normal wood for Ponderosa pine at two different moisture content levels: -—SHRINKAGE MOR MOE Comp. Tens. Bending Parallel Parallel cw M C % 87 Sp.Gr. 0.467 Long. 0.80 Rad. 2.2 Tang. 5.1 (psi) 6120 (psi* 1000) 842 (psi) 3300 (psi) 9690 cw 12.6 0.499 11710 1019 5970 nw 133 0.354 0.21 3.9 6.4 4640 1074 2340 11780 nw 12.0 0.372 9840 1345 5210 cw: Compression wood Sp.Gr.: Specific gravity nw: Normal wood (Panshin and De Zeeuw, 1980) 40 Juvenile wood is formed close to the pith of the tree. It is usually formed during the first few years of a tree's growth, but can continue to form for up to fifteen years. Juvenile wood cells are much larger than the cells of mature wood, and have a microfibrilar angle in the S2 layer (the middle and thickest layer in the wood cell wall) greater than that of normal wood. The result of juvenile wood cells being larger is a reduction in wood density, and a consequent reduction in wood strength. The greater microfibrilar angle results in a reduction in strength in the direction parallel to grain, due to a greater tension perpendicular to the microfibril component. Pitch pockets are usually a result of an injury to a tree, causing resin to collect in an area somewhere other than in the usual resin canals. The pitch or resin in these voids can be either a liquid or a crystalline solid. These pockets in turn can result in weak areas within a piece of lumber (Hoadley, 1990). In many of the beams with higher failure loads, compression wrinkles appeared in the upper laminations, in the region of highest stress, in the middle third of the beam (Fig. 6.10). The compression wrinkles did not seem to lead to any further beam damage or cause any significant decrease in load carrying capacity, but indicate the occurrence of permanent damage. They represent the transition from the elastic to the plastic region of the load-deflection or stress-strain curve in compression. 6.1.2 FAILURE PATTERNS Each glulam beam tested can be classified as having failed in one of two ways; either in a stepwise failure pattern or as a single sudden and complete failure. Load-deflection 41 Fig.6.10: Glulam beam compression failures on top. 42. curves for each beam are presented in the test report: "Glulam Beam Testing Program, Progress Report #1" (Timusk et al, 1994). In beams with stepwise failures, fractures often propagated from the point of initiation along a glue line or along the grain in the form of a shake, to another weak knot or area of sloped grain in the lamination above. This sometimes continued to a third or fourth failure before total beam collapse. These stepwise failures can be seen on the load displacement curves as incremental steps. At each failure, the load would drop suddenly and then, if the beam still had load carrying capacity, the load would begin to increase again until the next failure. Each failure was associated with a loud cracking noise, often preceded by several smaller cracking noises. Beams which did not exhibit a stepwise failure pattern, failed completely in a sudden and brittle manner accompanied by a loud crack. These beams generally disintegrated into many pieces and splinters. Of the two failure types, the stepwise failure is much more desirable in most construction applications. The cracking and stepped failure provide warning that the structure in question is being loaded beyond it's elastic range to the point where unrecoverable damage is occurring. This may provide valuable time for taking action such as evacuating buildings or removing loads. Another benefit of the stepwise type failure is that of being able to carry some residual load beyond the initial failure. This is especially valuable in load sharing applications, where overall safety would be increased if the failed member continues to carry part of its original load. 43 6.1.3 RATE OF LOADING Wood is described as being a viscoelastic material, and so it's mechanical properties are time related (Dinwoodie, 1979). This is commonly explained by the theory that timber fails when a certain critical deflection or strain has been reached. When timber is loaded slowly, viscous flow or creep can occur, allowing the timber to strain and thus fail at a lower load. For this reason, the rate of loading is an important factor. This time-load relationship is said to be particularly strong when timber is loaded in bending, and since tensile failures often govern in bending, the time-load relationship should also be important in tensile tests. The relationship between time to failure and maximum bending stress (MOR) is thought to be logarithmic, the MOR decreasing with the logarithm of time. Pearson (1972) developed the following relationship: cr = 91.5 - 7 log i 0 t where: a is the stress level in percent t is the duration of maximum bending load in hours. There is supposedly very little scatter from the relationship due to variation in species, specimen size, moisture content or whether or not the specimen is solid sawn or laminated. As a result, an attempt was made to maintain beam bending test durations at a minimum of five minutes, with a maximum of approximately ten minutes. The five to ten minute range results in stress levels of 99% to 97% respectively from the standard test value. In CAN / CSA086.1-M89, The Canadian National Standard on Engineering Design in Wood (Limit States Design), the load duration effect is treated as an adjustment to the design 44 stress by the load duration factor KD. Experiments were conducted at the University of British Columbia (Foschi, 1989) to develop a model to predict the accumulation of damage due to the intensity of the applied stress and the amount of damage already accumulated. The model was then used to develop the load duration factors in the code. The adjustment factor, KD, takes into account dead loads, which are considered permanent such as the weight of the building components themselves, and live loads which are non-permanent such as snow loads, occupancy loads, wind loads and earthquake loads. KD varied from 1.15 for short term loads to 0.65 for permanent duration loads. 6.1.4 MOISTURE CONTENT Most strength properties of wood are inversely related to the moisture content below the fiber saturation point. This is thought to occur because of the swelling and shrinking of wood as it picks up or loses moisture, effectively changing its density. Because density is strongly related to wood strength properties, moisture content also becomes related to strength. For example, a fifteen percent increase in moisture content from the oven dry state can cause the static bending MOE to drop approximately twenty-five percent and the modulus of rupture to drop forty-five percent, in any kind of wood regardless of species (Panshin, 1980). It was for these reasons that the moisture content of each beam was determined. As it turned out, there were no significant variations in moisture content from beam to beam. The moisture content of each beam would have also been necessary had one wanted to determine 45 the density of the beams. It was also for this reason that each beam was weighed before testing. No significant weight differences between beams were observed. 6.2 TENSION TESTS All tension tests were conducted to failure with the load directed parallel to grain. Failures were classified according to the major type(s) of defect(s) or failure mechanism involved. The failure categories were fingerjoint (FJ), knot, shear, sloped grain (SG), tension and shake, or any combination of two categories. The categories were defined as follows: Fingerjoint: Any failure occurring entirely in the region between the tips of opposite fingers, or breaking 30% or more of the cross-sectional area in the region between the fingers. Knot: Any failure starting at or propagating through a knot or the sloped grain surrounding a knot (Fig. 6.11). Shear: A failure in clear wood which fails along some angle up to 45° from parallel to grain (Fig. 6.14). Sloped Grain (SG): A failure in a region where the grain deviates significantly from the surrounding grain running parallel to the length of the specimen (Fig. 6.12). Tension: A failure in clear wood across the grain in which fibers have been pulled apart (Fig. 6.13). 46 Fig.til: Knot failure. Fig. 6.12; Sloped grain failure. Fig.fc.13: Tensile failure. Shake: A failure in clear wood occurring along the grain along an annual ring, separating early wood from late wood. This type of failure leaves a smooth failure surface. Failures were classified visually by examining each failed specimen. Any combination of two failure categories were also recorded, with the dominant failure type listed first. If three or more failure categories were observed, the dominant two were recorded (Tables 6.2 and 6.3). Tension test times were maintained at or above five minutes each for the same reasons as described in the preceding section on rate of loading of glulam beams (6.1.3). The few specimens that failed in less than the five minute target time were exceptionally weak, and these premature failures were difficult to avoid. All specimens failed in the region between the grips with a test span of four feet (1.22m) for the 8-foot specimens and a test span of 12 feet ( 3.66m) for the 16-foot specimens (Fig. 5.14). Failure loads, stresses and specimen moisture content levels at the time of test are listed in Table 6.4. A full table of tension test data for each specimen is presented in the appendix. Failures were mostly instantaneous, with the two specimen halves completely separating from one another in most cases (Fig.6.15). Each failure was accompanied with a loud bang, often preceded by smaller cracking noises produced by small microfailures. No slipping of specimens within the grips was observed, as the grip pressure and grip length were maintained at maximum. The grip surfaces were also periodically cleaned between tests to avoid slippage. 49 Fig. US: Tension tested lumber. So Tbl. 6.2 TENSION TEST FAILURES FINGER JOINTED 8-FOOT "C FINGER JOINTED 8-FOOT 'B' FAILURES Number FAILURES Number FJ 19 FJ 39 knot 39 knot 12 shear 1 shear 3 S G 0 S G 1 tension 0 tension 1 shake 0 shake 2 FJ + knot 14 FJ + knot 10 FJ + shear 6 FJ + shear 10 FJ + S G 3 FJ + S G 0 FJ + tens 0 FJ + tens 1 FJ + shake 0 FJ + shake 4 knot + shear 14 knot + shear 8 knot + S G 3 knot + S G 0 knot + tens 0 knot + tens 1 knot + shake 0 knot + shake 0 shear + S G 0 shear + S G 6 shear + tens 0 shear + tens 0 shear + shake 0 shear + shake 0 S G + tens 0 S G + tens 2 S G + shake 0 S G + shake 0 tens + shake 0 tens + shake 0 Total 99 Total 100 51 TbI. 6.3 Tension Test Failures 16-FOOT ' C Number 16-FOOT 'B' Number knot 32 knot 13 shear 2 shear 24 S G 4 S G 1 tension 1 tension 4 shake 1 shake 7 knot + shear 24 knot + shear 15 knot + S G 15 knot + S G 2 knot + tens 8 knot + tens 8 knot + shake 6 - knot + shake 4 shear + S G 2 shear + S G 9 shear + tens 1 shear + tens 3 shear + shake 2 shear + shake 8 S G + tens 2 S G + tens 1 S G + shake 1 S G + shake 2 tens + shake 1 tens + shake 1 Total 102 Total 102 8-FOOT ' C Number 8-FOOT 'B' Number knot 16 knot 14 shear 5 shear 43 S G 7 S G 2 tension 8 tension 4 shake 3 shake 5 knot + shear 15 knot + shear 10 knot + S G 2 knot + S G 1 knot + tens 18 knot + tens 4 knot + shake 6 knot + shake 2 shear + S G 8 shear + S G 0 shear + tens 5 shear + tens 4 shear + shake 3 shear + shake 9 S G + tens 3 S G + tens 1 S G + shake 0 S G + shake 0 tens + shake 3 tens + shake 3 Total 102 Total 102 52 6.2.1 LENGTH EFFECT As expected, the 16-foot (4.88 m) specimens in each grade failed at a lower average load than the 8-foot (2.44 m) specimens (Table 6.4 and Figure 6.16). The 16-foot and 8-foot ' C mean failure loads were 127.56 kN and 164.26 kN respectively. The 16-foot and 8-foot 'B' mean failure loads were 175.46 kN and 204.33 kN respectively. This can be explained by the length effect, which states that because the tensile strength of lumber is directly related to the random occurrence of strength reducing defects, the longer the gauge length or test specimen, the higher the probability of the occurrence of a defect. This in turn increases the probability of failure at a given load, reducing the mean failure load for a group of test specimens of a given length. One theory often used to describe the effect of length on the strength of lumber is Weibull's weakest link theory, as described in the paper "Effect of length on the tensile strength of lumber" by Lam and Varoglu (1990). The theory assumes that the stresses throughout the volume in tension are uniform and parallel to grain, and that the tensile failure in lumber may be considered as brittle. Then from the two parameter Weibull model, assuming statistical independence between the tensile strengths of the components making up the length of the member, the effect of length on the tensile strength of the member as a whole can be modeled by (Lam, 1990): 53 Tbl. 6.4 Tensi le Test Summary Data Fingerjointed 8-Foot ' C Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 27970.71 124.46 26487.27 10.56 Stdev.: 6509.78 28.97 6164.53 0.93 Min: 15000 66.75 14204.469 8.00 Max: 47500 211.36 44980.82 12.50 Fingerjointed 8-Foot 'B' Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 35548.00 158.18 33662.70 10.98 Stdev: 6982.37 31.07 6612.05 0.90 Min: 16400 72.98 15530.22 8.00 Max: 52600 234.06 49810.34 13.50 8-Foot 'C Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 36913.73 164.26 30875.60 12.10 Stdev: 7631.57 33.96 6383.24 1.09 Min: 19200 85.44 16059.38 9.00 Max: 55900 248.74 46756.22 14.00 8-Foot 'B' Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 45918.63 204.33 38407.54 11.26 Stdev: 10475.90 46.62 8762.32 1.22 Min: 17400 77.43 14553.81 4.50 Max: 73700 327.95 61644.60 13.00 16-Foot ' C Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 28666.67 127.56 23977.55 11.33 Stdev: 7174.09 31.92 6000.60 1.33 Min: 9700 43.16 8113.33 6.50 Max: 50700 225.60 42406.80 14.00 16-Foot'B' Load (lbs) Load (kN) Strs(kPa) MC (%) Average: 39430.39 175.46 32980.61 11.35 Stdev: 7267.91 32.34 6079.07 0.88 Min: 24300 108.13 20325.15 9.00 Max: 59000 262.54 49349.14 13.50 54 Where: k = scale parameter m = shape parameter vj = volume of specimen 1 v2 = volume of specimen 2 Li = length of specimen 1 L2 = length of specimen 2 ai = stress or strength of specimen 1 (5i = stress or strength of specimen 2 v0 = reference volume For the scale parameter the exponent -1.085 was used as it is described as being a good estimation by Leicester. The value of k can also be estimated from regression analysis of the logarithm of tensile strength vs. the logarithm of test specimen length. 56 o Predicted Stress Ratios and Actual Stress Ratios: Length Grade k Predicted Actual Difference Stress Ratio Stress Ratio (%) 8-Foot ' C 5.5305 1.2197 1.2877 5.3 16-Foot ' C 4.4952 1.2768 1.2877 0.8 8-Foot 'B' 4.9699 1.2474 1.1645 7.1 16-Foot 'B' 6.2639 1.1917 1.1645 2.3 6.2.2 GRADE EFFECT A distinct effect of grade on tensile test mean failure loads was observed (Fig. 6.16). The 16-foot 'B' grade mean failure load was 175.46 kN, 47.9 kN higher than the mean failure load for the 16-foot ' C grade at 127.56 kN. Also the 8-foot 'B' grade mean failure load of 204.33 kN was 40.07 kN higher than the 8-foot ' C grade mean failure load of 164.26 kN. This can be explained by the fact that the lumber is graded according to MOE and visual criteria to separate it into the various laminating grades such as 'B' and ' C . Fifth percentile values are of more significance in design applications as they take into account the inherent variability in material properties, and can be estimated from the CDF plots (Fig. 6.16). There seems to be a relationship between MOE and tensile strength, as is the case with MOE and MOR. The higher the MOE or stiffness, the higher the MOR or ultimate bending strength and the tensile strength. This could be explained by a stiffer or higher MOE board being denser than a less stiff board, and there exists a strong relationship between density and strength. It could also be explained by defects reducing the bending stiffness 57 because of reduced tensile stiffness in the tension region of a bending region. This then reduces the overall bending strength. Visual grading criteria separate lumber into grades on the basis of visual defects, higher grades being allowed fewer and smaller defects than lower grades. Because the higher grades have fewer and smaller defects, they are stronger than the lower grades. 6.2.3 FINGERJOINT FAILURES The fingerjointed groups of 8-foot 'B' and 'C' lumber were tested in tension to failure similar to the other groups. In addition to the earlier listed failure categories, the fingerjoint failure category was added to the list of failure classification. For the purpose of determining the strength of the fingerjoints for ULAG, only the specimens of this group failing at a fingerjoint were of importance. The specimens failing at other defects merely indicated that the fingerjoints were at least as strong as that defect. The fact that the mean tensile strength of the fingerjointed groups were lower than those of their corresponding non-fingerjointed groups can be explained by the presence of the fingerjoint simply being another defect, increasing the probability of failure over a similar group without it. It is interesting to note that within a fingerjointed group of the same grade and length, the mean failure load of the specimens failing at fingerjoints was higher than the mean failure load of the specimens failing at other defects (Fig. 6.17 and 6.18). This suggests that if one were interested in determining the strength of fingerjoints, say for the proof loading option in ULAG, one could simply load the specimens to a pre-determined load which will cause fingerjoint failures in ten to fifteen percent of the population after the failure curve for a larger population is established. This would give a good indication of the lower tail end of the population or the 58 LL Q O CO CD I-c o CO c CD H O LL o o o o « oo o o ««<1. •es ilures Failui FJ Fai Other o < Oo Oo « Oo o o oo o « o o o L O o o o o s f o o o L O co o o o (0 o ro O O o Ln o co d d CD d m d s f d co d CNJ d o d o o o o CL CO CO CD o v-o •*-> S CO CM O O O o CM Ajjijqeqojd eAjjeiniuno o Q_ Ll_ Q O -*—< co CD h-c o CO c cu H O T J CD •4—* C o CD O) c CD OO Co Q O +•» (/> o H c o "55 c <D °<>o W ZD 0) — • S= (0 LL-CC (D uT O o < <1< 00<x> <<. <a< o<> < <o^  < o* 03 d CO d O d d ci d CNJ d o d Annqeqojd 3Aji.e|niiino weakest members, which are the members of concern. In the glulam or other structural industry, one is mostly interested in the minimum load carrying capacity, because this is where the serious problems occur. The fact that the 'B' grade fingerjoint failure loads were higher than the ' C grade fingerjoint failures loads, which all occurred in clear wood, can again be explained by the MOE difference in grades. The higher MOE values of the 'B' grade may be related to density, which in turn is related to tensile strength. 6.3 ULAG ANALYTICAL SIMULATION RESULTS To verify the procedure used by the ULAG glulam beam simulation computer program, a series of simulations of 200 runs each were performed, using the material properties from tests on laminating stock and fingerjoint specimens. These results were then compared with actual glulam beam test data. Subsequently the program has been used to investigate size effects and the influence of different lay-ups on the strength of beams. 6.3.1 PROGRAM VERIFICATION The ULAG program was run to simulate a population of 200 beams for each of the two types of glulam beams. A complete plot of all the beam tests results vs. ULAG simulation results are presented in Figures 6.19, 6.20 and 6.21. 61 Total Load Carrying Capacity Test ULAG Mean Value COV (kN) (%) Mean Value COV (kN) (%) 16-Foot Beams 24-Foot Beams 87.6 17.9 (35.08) 125.3 10.9 (38.44) 82.7 14.7 (33.12) 119.4 9.9 (36.63) 122.0 10.1 (37.42) Results in italics for no fingerjoints. 200 runs (n=200) in each case. (MOR values in brackets) There is very good agreement between the ULAG simulation results and the actual beam tests for the 16-foot beams, which did not have fingerjoints. The agreement for the 24-foot beams is also good, but the ULAG results seem to be on average conservative, under-predicting the beam strengths by approximately 5%. This could be explained by the fact that during the fingerjointing processs, the ends of boards were often cut short to eliminate knots within three knot diameters of the fingerjoint. This in effect improved the grade of the laminating stock used in the beams, as compared with the raw data from untrimmed stock. When the effect of fingerjoint strength was eliminated from the model by specifying an artificialy high tension strength, it was found that the difference between the ULAG simulation and the actual results dropped to less than 3%. Regardless of the relatively crude remedy, it can be concluded that the results are within 3% to 5% of the test data on the onservative or "safe" side, which can be considered to be in "satisfactory agreement". 62 Figure 6.19 Comparison of ULAG and 16-Foot Beam Tests 40 50 60 70 80 90 100 110 120 130 140 Ultimate Load Pu ( k N ) UBC Glulam Test Program: 16* Test Beams 63 Figure 6.20 Comparison of ULAG and 24-Foot Beam Tests U l t i m a t e L o a d P 0 ( k N ) UBC Glulam Test P rogram: 24 ' Test Beams 64 Figure 6.21 Comparison of ULAG and 24-Foot Beam Tests No FJ Failures 1.0 0.9 : >s0.8 : _l 0.7 : o | 0.6 ^ 0 . 5 •*-> 10.3 10.2 ° 0 . 1 0.0 I I I I I 1 I 70 80 Test Dato. ULAG Simulation. i i i i i i i i i i 90 100 110 i i i i i i i i i i i i i i i i i i i i i 120 130 140 150 160 U l t i m a t e L o a d P u ( k N ) UBC Glulam Test Program: 24' Test Beams (No End Joint Failures) 65 6.3.2 ULAG SIMULATIONS The ULAG glulam beam simulation program ideally lends itself to study the effect of a variety of parameters on the bending strength of glulam beams, such as (a) the effects of the use of a variety of laminating stocks with differing material properties, (b) different lay-up schemes, (c) the length of laminating stock and subsequent frequency of fingerjoints, (d) the depth of beams, (e) the length of beams, (f) different loading conditions, (g) fingerjoint strength. A summary of parameters studied is given in the following subsections. BEAM LAY-UP EFFECT One of the advantages of glulam beams is that by sawing a log into lumber and then gluing it back together, one can optimize the use of the lumber by placing the higher grade material in areas of high stress, while lower grade and less expensive material is used in the remaining low-stress areas. Glulam beams are typically custom manufactured and can thus be designed for their end use. For these reasons a variety of different lay-ups corresponding to different glulam grades are commonly manufactured in industry, each with a specific application. For example, the f-E grades have high grade laminations only on the bottom, and are designed for uses where positive bending is dominant. The f-EX grades have high grade laminations both on the top and the bottom, and can be used for cantilevered beams or applications where load reversal can be expected, resulting in large positive or negative bending moments . To investigate the effect of placing a varying number of high grade laminates as the bottom laminations in single span, simply loaded beams, several ULAG simulations were performed (Fig. 6.22 and 6.23). For the investigation, the beam dimensions, number of 66 Fig. 6.22 ULAG LAY-UP EFFECT 1 1 T 38.1mm 2.4384 m 2.4384 m-Nominal Load: 20 kN Span: 4.8768 m (16') Depth: 304.8 mm (8 lams) Width: 0.1397 m (5.5") Lam Thickness : 38.1 mm (1.5") Element Length: 0.5 m No. of Simulations: 40 Lamstock Grade: 'B' and 'C Mean Lay-up Load Fctr Stdev 0-B 8-C 1.256 0.2435 1-B 7-C 1.504 0.2776 2-B 6-C 1.597 0.2349 3-B 5-C 1.578 0.2484 4-B 4-C 1.580 0.2836 5-B 3-C 1.564 0.2366 6-B 2-C 1.604 0.2283 7-B 1-C 1.550 0.3331 8-B 0-C 1.580 0.2781 67 o C L -*—• o CD CO I ch (N>l) peo-| XB | / \ | aBejaAv 68 laminations, loading configuration, nominal load, and element length were all kept constant. The number of 'B' grade laminations was varied from none (all the laminations were ' C grade) to eight (all the laminations were 'B'), substituting 'B's for 'Cs from the bottom up. If linear elastic behaviour is assumed, then the bending stress in the beam can be calculated as: cr = M * y / I where M is the bending moment, y is the distance from the neutral axis to the point where stress is to be calculated, and I is the section modulus defined as 1/12 b d3, where b and d are the width and the depth, respectively. The maximum stress occurs at the extreme fibers, compression in the top laminates and tension in the bottom laminates. Because tensile fracture generally governs in glulam beams, and because the 'B' grade laminates are stronger in tension than the ' C grade laminates, the largest increase in load carrying capacity occurs when the bottom most ' C grade laminate is replaced with a 'B' grade laminate. Replacement of the next ' C grade laminate with a 'B' grade laminate only moderately increases the load carrying capacity of the beam, because the tensile stresses are lower one laminate up, reducing the probability of failure. The substitution of the third and subsequent ' C grade laminates with 'B' grade laminates did not increase the load carrying capacity of the beam. Because the two grades have similar properties in compression, the substitution with 'B' grades in the high compression region in the top of the beam made no difference to the beam strength. 69 From this type of simulation it can be concluded that it makes no economic sense to place more than one or two high grade laminates in the bottom of the lay-up for an eight laminate glulam beam subject to positive bending. BEAM SIZE EFFECT Next ULAG was used to investigate the effect of beam size in terms of depth on load carrying capacity. Thirteen sets of ULAG simulations were run, varying the number of laminations in a beam from two to fourteen in increments of one 38mm lamination (Fig. 6.24). In each case the applied load, beam span, beam width, lamination thickness, support conditions, element size and number of simulations were kept constant. For analysis, plots were made of ultimate failure load vs. number of laminations.(Fig. 6.25) and of extreme fiber stress vs. number of laminations (Fig. 6.26). The largest drop in extreme fiber stress at failure occurred when the beam depth was changed from two laminations to three. The addition of subsequent laminations resulted in progressively smaller reductions in extreme fiber stress, and a general leveling off of the curve as illustrated in Figures 6.6, 6.7 and 6.8. In tension testing it was clear why a longer board should be weaker than a shorter board, due to the higher stressed volume and therefore a higher probability of the occurrence of a critical defect such as a knot or sloped grain (section 6.1). In the case of beams, the overall reduction in extreme fiber stress, and the non-linear relationship between beam size and load carrying capacity is not as easily explained by the size effect theory. In the case of beams where the depth is increased but the length remains constant, the relationship is not as 70 Fig. 6.24 ULAG SIZE EFFECT 38.1mm 2.4384 m 2.4384 m 1* Nominal Load: Span. Depth: Width: Lam Thickness : Element Length: No. of Simulations: Lamstock Grade: 20 kN 4.8768 m (16') 76.2 mm - 533.4 mm (2 lam - 14 lam) 0.1397 m(5.5M) 38.1mm(1.5M) 0.2 m 40 Al l 'C No. Lams Depth (m) 2 0.0762 3 0.1143 4 0.1524 5 0.1905 6 0.2286 7 0.2267 8 0.3048 9 0.3429 10 0.3810 11 0.4191 12 0.4572 13 0.4953 14 0.5334 Mean Load Fctr 0.435 0.703 1.119 1.716 2.280 2.815 3.769 4.566 5.181 6.352 7.118 8.437 9.567 71 Stdev 0.108 0.169 0.252 0.395 0.463 0.778 0.664 1.031 1.150 1.412 1.544 2.106 1.792 clear. Although the deeper beam does contain more wood resulting in a higher probability for the occurrence of a critical flaw (a higher stressed volume), the inner wood is at a lower stress level, so the inner flaws are not as critical as flaws towards the extreme fibers. The relationship between extreme fiber stress and beam depth can also be explained by examining the difference in stress distributions between shallow beams and deep beams: Deep Beam Shallow Beam From the stress distribution diagrams it is apparent that the extreme fiber stress is much greater for the shallow beam than for the deep beam. At the same time, the mean stress in the outer most laminations of each beam are the same, allowing them to be made of the same grade of laminating material. For this reason, two beams made from identical laminating material having the same MOE and tensile strength properties, may have quite different extreme fiber stress levels. As beams get deeper, the stress distribution gradient becomes less steep and the extreme fiber stress drops off. Another factor in the beam size or depth to load carrying capacity relationship may be load sharing as explained by Keenan (Keenan, 1974). In the case of a shallow beam, as the extreme fiber on the tension side reaches its maximum stress level, the next fiber towards the 74 center of the beam will be far from having reached its maximum stress level due to the steep stress gradient. This inner fiber will then be able to assume some of the load from the extreme fiber, resulting in load sharing and ductility in the stress block. On the compression side, as the extreme fiber reaches its maximum stress and wants to buckle, the fiber next to it being far from its maximum stress level can lend lateral support to the extreme fiber. This load sharing may continue for several fiber layers towards the center of the beam, increasing the overall load carrying capacity. In the case of a deep beam with a shallow stress gradient, the fibers next to the extreme fiber will already be very near their maximum stress limits, and will not be able to assume much of the extreme fiber's stress or be able to lend lateral support. 7. C O N C L U S I O N S A N D R E C O M M E N D A T I O N S 7.1 SUMMARY A population of 'B' and ' C grade 16-foot (4.88 m) 2x6 (38 mm x 140 mm) lamstock was E-rated and separated into two statistically similar groups, one for making into glulam beams and testing as beams, and the other for testing as lamstock. Thirty 16-foot 8-lamination deep all ' C grade glulam beams with no fingerjoints, and thirty 24-foot (7.31 m) 12-lamination deep 'B' and ' C grade glulam beams with fingerjoints were commercially manufactured at Western Archrib in Edmonton. Two-hundred 8-foot (2.44 m) ' C fingerjointed specimens and two-hundred 8-foot 'B' fingerjointed specimens were also produced at Western Archrib for later testing at UBC. 75 The glulam beams were tested to failure in four point bending, recording the load-displacement curves of each as well as the maximum failure load. The 'B' and ' C grade lamstock boards were fed through the Cook-Bolinders E-rating machine generating modulus of elasticity profiles for each. Each grade was then separated into two statistically similar populations, one of which was cut in half giving 100 8-foot 'B's and 100 8-foot 'C's, 100 16 foot 'B's and 100 16-foot 'C's, as well as 100 8-foot finger jointed 'B's and 100 8-foot finger jointed 'C's. All were then tested in tension parallel to grain to failure, recording the type of failure and the maximum load. The modulus of elasticity profiles and the tension test data served as input information for the ULAG utility program "ULAGPRO 1", giving ULAG the necessary information to simulate glulam beams made from such lamstock. ULAG was then run, simulating glulam beams identical to the beams which we had made and tested. The real glulam beam test data served to verify the ULAG glulam beam simulations. After verification, ULAG was used to investigate the effects of glulam beam size and various lay-ups. Various other studies were carried out with the lamstock test data, investigating grade effect, length effect and failure types. 7.2 CONCLUSIONS The four point bending test conducted on thirty 16-foot (4.88 m) and thirty 24-foot (7.31 m) glulam beams support the predictions of the ULAG glulam beam simulation computer program, using input data from a statistically similar population of lamstock. The 76 16-foot beam simulations fit the test data more closely than the 24-foot simulations. This is thought to be due to the underestimation of fingerjoint strengths. ULAG also works well as a tool for investigating properties of glulam beams such as size effect or the use of different grades in the lay-up. 7.3 RECOMMENDATIONS Although ULAG works quite well in the format it is presently in, a more user friendly operating environment would greatly enhance its effectiveness. The ability to go into an existing simulation scenario all presented on the screen at once, and then to change one or two selected parameters without having to go through numerous lines one after another would also be a great improvement. The beam lay-up, loading and support condition schematic for user verification before execution is definitely an asset. The investigation of the properties of glulam beams made of other laminating species, grades and combinations of grades would be of interest. Tensile, modulus of elasticity and fingerjoint properties for each type of lamstock would have to be determined for the investigation. 77 R E F E R E N C E S American Institute of Timber Construction. (1985). AITC 117-84 Design Standard Specifications for Structural Glued Laminated Timber of Softwood Species. Wiley-Interscience Publication, Toronto, Ontario. American National Standards Institute. (1973). Structural Glued Laminated Timber. American National Standard A 190.1-1973. Voluntary Product Standard PS 56-73. U.S. Department of Commerce. National Bureau of Standards. American Society for Testing and Materials. (1987). Standard Methods for Establishing Stresses for Structural Glued Laminated Timber (Glulam). ASTM D 3737 - 87. Philadelphia, Pennsylvania. American Society for Testing and Materials. (1992). Standard Terminology Relating to Wood. ASTM D 9-87 (Reapproved 1992). Philadelphia, Pennsylvania. Bender, D.A. Woeste, F.E. Schaffer, E.L. Marx, C M . (1985). Reliability Formulation for the Strength and Fire Endurance of Glued-Laminated Beams. Res. Pap. FPL 460. Madison, WI: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory; 43 p. Bury, K.V. (1986). Statistical Models in Applied Science. Robert E. Krieger Publishing Company. Malabar, Florida. Canadian Standards Association. (1980). Structural Glued-Laminated Timber. CSA Standard 0122-M1980. Canadian Standards Association. Rexdale, Ontario. Canadian Standards Association. (1981). Qualification Code for Manufacturers of Structural Glued-Laminated Timber. CSA Standard 0177-M1981. Canadian Standards Association. Rexdale, Ontario. Canadian Wood Council (1994). Wood Design Course Notes. Canadian Wood Council Ottawa, Ontario. Canadian Wood Council (1990). Wood Design Manual 1990. Candian Wood Council, Ottawa, Ontario. Devore, J.L. (1987). Probability and Statistics for Engineering and the Sciences (Second Edition). Brooks/Cole Publishing Company, Monterey, California. Fargey,jj. (1995). Telephone conversation with and doccument fromjFargey, Western Archrib, Edmonton, Alberta. 78 Folz, B.R. (1997). Stochastic Finite Element Modelling of Glued-Laminated Beam-Columns. Ph.D Dessertation, Department of Civil Engineering, Universtiy of British Columbia, Vancouver, British Columbia. Folz, B. and Foschi, R.O. (1995). ULAG: Ultimate Load Analysis of Glulam User's Manual Version 1.0. Department of Civil Engineering, University of British Columbia, Vancouver, British Columbia. Forest Products Laboratory (1990). Wood Engineering Handbook (Second Edition). Prentice Hall, Englewood, California. Foschi, R.O. (1985). GLULAM: A Simulation Model For Strength And Stiffness of Glued-Laminated Beams. A.I.T.C. Foschi Barrett Engineering Consultants, Inc., Vancouver, British Columbia. Foschi, R.O. Barrett, J.D. (1980). Glued laminated beam strength: a model. ASCE Journal of the Structural Division, Vol.106, No. ST8, pp. 1735 - 1754. Foschi, R.O. Folz, B. Yao, F. and Prion, H.G.L. Timusk, P.C. (1995). Evaluation and Verification of ULAG, a Computer Simulation Program for the Strength and Stiffness of Glued-Laminated Beams. University of British Columbia, Vancouver, British Columbia. Hilson, B.O. Pellicane, P.J. Smith, I. Whale, L.R.J. (1979). Towards Optimal Design of Glued Laminated Timber Beams. Hoadley, R.B. (1990). Identifying Wood: Accurate Results With Simple Tools. The Taunton Press, Inc. Newtown, Connecticut. Illston, J.M. Dinwoodie, J.M. Smith, A.A. (1979). Concrete, Timber and Metals. The Nature and Behavior of Structural Materials. Van Nostrand Reinhold Company, New York, New York. Keenan, F.J. (1974). Shear Strength of Wood Beams. Forest Products Journal, Vol. 24, No 9. Madison, Wisconsin. Lam, F. Varoglu, E. (1990). Effect of Length on the Tensile Strength of Lumber. Forest Products Journal, Vol. 40, No. 5. Madison, Wisconsin. Marx, C M . Moody, R.C. (1981). Strength and Stiffness of Small Glued-Laminated Beams with Different Qualities of Tension Laminations. Res. Pap. FPL 381. Madison, Wisconsin: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. 79 Moody, R.C. Carter, D. Plantinga, P.L. (1988). Analysis of Size Effect for Glulam Beams. Proceedings of the 1988 International Conference on Timber Engineering. Panshin, A.J. Zeeuw, CD. (1980). Textbook of Wood Technology, Fourth Edition. McGraw-Hill Publishing Company, New York, New York. Timusk, P.C Prion, H.G.L. Wyder, J. Foschi, R.O. (1994). Glulam Beam Testing Program, Progress Report #1. Department of Civil Engineering, University of British Columbia, Vancouver, British Columbia. Walpole, R.E. Myers, R.H. (1989). Probability and Statistics for Engineers and Scientists (Fourth Edition). MacMillan Publishing Company, New York, New York. Wolfe, R.W. Moody, R.C. (1979). Bending Strength of Vertically Glued Laminated Beams with One to Five Plies. Res. Pap. FPL 333. Madison, Wisconsin: U.S. Department of Agriculture, Forest Service, Forest Products Laboratory. 80 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items