Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Dynamic modeling and design evaluation of interactions of fish and a head-removal machine Bussani, Franco Steven 1993

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

Item Metadata

Download

Media
831-ubc_1994-0075.pdf [ 3.48MB ]
Metadata
JSON: 831-1.0080844.json
JSON-LD: 831-1.0080844-ld.json
RDF/XML (Pretty): 831-1.0080844-rdf.xml
RDF/JSON: 831-1.0080844-rdf.json
Turtle: 831-1.0080844-turtle.txt
N-Triples: 831-1.0080844-rdf-ntriples.txt
Original Record: 831-1.0080844-source.json
Full Text
831-1.0080844-fulltext.txt
Citation
831-1.0080844.ris

Full Text

DYNAMIC MODELING AND DESIGN EVALUATION OFINTERACTIONS OF FISH AND A HEAD-REMOVAL MACHINEByFranco Steven BussaniB. A. Sc. University of British ColumbiaA THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTERS OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE STUDIESMECHANICAL ENGINEERINGWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIADecember 1993© Franco Steven Bussani, 1993In presenting this thesis in partial fulfilment of the requirements for an advanced degree atthe University of British Columbia, I agree that the Library shall make it freely availablefor reference and study. I further agree that permission for extensive copying of thisthesis for scholarly purposes may be granted by the head of my department or by hisor her representatives. It is understood that copying or publication of this thesis forfinancial gain shall not be allowed without my written permission.Mechanical EngineeringThe University of British Columbia2075 Wesbrook PlaceVancouver, CanadaV6T 1W5Date:c7IAbstractCommercial salmon processing is a five-hundred million dollar industry in British Columbia. This research is part of the work undertaken at the Industrial Automation Laboratory of the University of British Columbia to address the need to increase the rawproduct recovery rate during the processing of salmon, in view of dwindling stocks andincreasing production costs. The scope of this research is limited to the analysis of positioning the salmon for head removal on a conventional machine (Iron Butcher). A betterunderstanding of the dynamic interactions in this stage of processing will lead to establishing design improvements that will increase the raw product recovery rate of the IronButcher.Finite element models are developed to simulate the static and dynamic interactionof a salmon with the Iron Butcher. The finite element models are verified by comparingexperimental results to simulation results obtained from the models. Verification ensuresthe simulation results are accurate and also provides an estimate of the error in the finiteelement models.The finite element model is first used to analyze the fish—machine interaction duringthe conveying process. Simulations provide information on how to better configure themachine so that reliable conveying of fish in different situations is achieved. The finiteelement model is also used to analyze the operation of the Iron Butcher in its presentconfiguration to determine the causes of positioning error, termed overfeed and underfeed.Once the mechanisms which cause these errors are identified, the model of the processingmachine is modified so that the effectiveness of these modifications can be determinedprior to the building of a prototype machine.iiThese simulation results lead to the establishment of possible design improvements forconveying, indexing and holding the salmon. Recommendations are made for the designand configuration of machine components which will result in an improved performanceof the the Iron Butcher.HiTable of ContentsAbstractList of Tables viiiList of Figures xAcknowledgement xvi1 Introduction 11.1 History of Fish Processing 11.2 The Iron Butcher 31.3 Motivation of the Research 51.4 Research Objectives 91.5 Organization of the Thesis 122 Literature Review 142.1 Manipulation of Sliding Objects 142.2 Impact, Contact and Grasp 162.3 Material Properties of Fish Tissue 183 Model of the Fish 223.1 The Finite Element Method 223.1.1 Solution Procedure in the Finite Element Method 233.1.2 Application of the Finite Element Method to Biological Systems 26iv3.2 Material Properties of Salmon Tissue3.2.1 Material Properties of Muscle3.2.2 Material Properties of Organs3.2.3 Material Properties of Bone3.3 Model Assumptions3.3.1 Assumptions Pertaining to Material Properties of Biological Tissues3.3.2 Assumptions Pertaining to the Fish Geometry3.4 Details of the Model3.4.1 Development of the Fish Geometry3.4.2 The Collar—bone and Gill Cover3.4.3 Assigning Values for Material Properties3.5 Verification of the Finite Element Model of a Fish3.5.1 Model Validation3.5.2 Mesh Refinement Study4.3.2 Simulation Results 5726272829292931313236373838444 Model of the Processing Machine 484.1 Model Assumptions 484.2 Model of the Iron Butcher 494.2.1 The Table Surface 494.2.2 The Conveying System 504.2.3 The Holding Mechanism 514.2.4 The Indexing System 524.2.5 The Fish—Machine Interface 534.3 Verification of the Conveying Model 544.3.1 Order of Magnitude Calculation 55V4.3.3 Discussion of Results.4.4 Verification of the Indexing and Holding Model4.4.1 Operation of the Iron Butcher4.4.2 Simulation Results4.4.3 Discussion of Results4.5 Limitations of the Model6 Investigation of Indexing Errors6.1 Investigation of Underfeed Error6.2 Investigation of Overfeed Error6.2.1 The Role of Friction in Overfeed Error6.2.2 The Role of Structural Deformities6.3 Summary of Results and Design Recommendations7 Implementation of Design Improvements 1087.1 Laboratory Prototype 1087.2 Industrial Prototype 1118 Conclusions and Recommendations 1138.1 An Overview of the Research 1135859596061625 Finite Element Simulation of Conveying5.1 Conventional Conveying5.2 Modified Iron Butcher Conveying5.2.1 Pulsed Conveyor Motion of the Current Iron Butcher5.2.2 Conveying Using Staggered Lugs .5.2.3 Conveying Using Holding Pans5.3 Summary of Results and Design Recommendations6666677275828493939798104104vi8.2 Recommendations for Design Improvement 1148.3 Recommendations for Further Work 115Bibliography 117viiList of Tables2.1 The modulus of deformability and the upper limit on the linear region ofthe true stress—strain curves for various species of fish, as determined byJohnson et al 202.2 Recorded values of Young’s modulus and the associated coefficient of variation for various species of fish, as determined by Borderias et al 203.1 Material properties for soft and firm fish used in the finite element simulations 393.2 Experimental results for the bending of Fish No.1 and Fish No.3 403.3 Simulation results for the bending of Fish No.1 and Fish No.3 423.4 The percentage error between experimental and simulation results for thebending of Fish No.1 and Fish No.3 443.5 Convergence curve data for the model of Fish No.3. CPU times are forsimulations on a Sun SPARC station IPX using release 5.0 of ANSYS. . 464.1 Overshoot and steady—state resting position for the simulations of conveying firm and soft fish 584.2 A comparison of the actual size of the machine components and the corresponding component size in the finite element models 626.1 Variation in the minimum gill cover—fish gap size for successful indexingusing the current design of the Iron Butcher 95viii6.2 Variation in the minimum gill cover—fish gap size for successful indexingwhen the nose push—down device is added to the Iron Butcher 97ixList of Figures1.1 The Iron Butcher: (a) positioning (conveying and indexing) the fish forthe head cut (b) feeding fish onto the conveyer. Conveyer motion is fromright to left 41.2 An illustration of anatomical features of salmon which are of interest whenprocessing salmon on the Iron Butcher 51.3 Positioning a salmon on the feeding table 61.4 Lugs, used for conveying the salmon, are shown on either side of an indexer 71.5 An indexer is locked into the collar—bone region of a fish and is indexingthe fish for the head cut while a weight rests on the tail—end of the fish. 81.6 The indexer retracts from the fish just prior to the removal of the fish headby the cutter 91.7 A schematic representation of the Iron Butcher, showing the crucial components 101.8 A comparison of the amount of meat attached to the fish head (waste)resulting from a good cut (on the right) with that from a cut where overfeeding error occurred (on the left) 112.1 The orthogonal axes for fish muscle, as defined by Petrell et al 193.1 The metameric structure of fish muscle. The pattern of lines on the cross(1) and longitudinal (2) sections represents the arrangement of sheets ofconnective tissue (myosepta) in the muscle 273.2 Simplification of the fish geometry 32x3.3 The equipment setup for acquiring the image of a salmon 333.4 Calculation of the major principal axis and 8, the angle of orientation ofthe principal axis 343.5 Measurement of the perpendicular distance from points on the contour ofthe fish to the major principal axis 353.6 The cross—section of the finite element model of the fish (right) is basedon the drawing (left) 373.7 The captured image of the bending experiment of Fish No.1 (on the left)and Fish No.3 (on the right) 413.8 Dimensionalized experimental results for the bending experiments 413.9 Simulation results for the bending of fish using the material property valuesfrom Petrell et al.: (a) Material property values from the low end of therange (b) Material property values from the high end of the range 423.10 Simulation results for the bending of fish using material property valueswhich are an order of magnitude smaller than those cited by Petrell et al.:(a) Material property values from the low end of the range (b) Materialproperty values from the high end of the range 423.11 Simulation results for the bending of (a) Fish No.1 (b) Fish No.3 433.12 Models used in the mesh refinement study 453.13 Convergence curves for the finite element model of the fish 474.1 Two views showing the finite element model of the Iron Butcher 504.2 Orientation of the lugs for (a) straight conveying (b) staggered conveying 514.3 Displacement profiles (motion sources) that specify the motion of the lugsfor (a) continuous conveying (b) pulsed conveying 52xi4.4 The force—deflection relationships for a CONTACT49 element specifiedwith (a) rigid friction (b) elastic friction 554.5 Position of the fish mass center for the simulation of the conveying of (a)firm fish (b) soft fish 584.6 Schematic representation of the Iron Butcher showing the location andmotion of the lugs, indexer and holder during the indexing and holdingprocess. (1) Initial position (2) Position when the indexer and holder droponto the fish (3) Final position when indexing is completed 604.7 Simulation results for the validation of indexing and holding processes ofsmall fish 644.8 Simulation results for the validation of indexing and holding processes oflarge fish 655.1 Constant speed conveying of small, soft fish using the existing configuration of the Iron Butcher. The transient “rocking” motion is sufficientlysettled at Time = 0.61 s 685.2 Constant speed conveying of small, firm fish using the existing configuration of the Iron Butcher. The transient “rocking” motion is sufficientlysettled at Time = 1.01 s 695.3 Constant speed conveying of large, soft fish using the existing configurationof the Iron Butcher. The transient “rocking” motion is sufficiently settledat Time = 0.81 s 705.4 Constant speed conveying of large, firm fish using the existing configuration of the Iron Butcher. The transient “rocking” motion is sufficientlysettled at Time = 1.61 s 71xii5.5 Pulsed conveying of a small, soft fish using the existing configuration ofthe Iron Butcher shows that an error in the lateral position of the fish isintroduced after only two cycles of the conveyor motion 745.6 Simulation results for the pulsed conveying of a large, soft salmon with asingle parallel holder show that a successful hold is not achieved 765.7 Simulation results for the pulsed conveying of a large, soft salmon withtwo parallel holders show that a successful hold is possible 775.8 Simulation results for the pulsed conveying of a large, soft salmon whentwo parallel holders are used: (a) the holding forces, (b) the undeformedstate of the fish prior to starting the holding process and (c) the deformedstate of the fish during the holding process 785.9 Simulation results for the pulsed conveying of a large, soft salmon whenone slanted holder is used, show that a successful hold is possible 795.10 Simulation results for the pulsed conveying of a large, soft salmon whentwo slanted holders are used, show that a successful hold is possible. .. 805.11 Simulation results for the pulsed conveying of a large, soft salmon whenone slanted holder is used: (a) the holding force, (b) the undeformed stateof the fish prior to starting the holding process and the deformed state ofthe fish (c) viewed from the dorsal (back) side of the fish (d) viewed fromthe ventral (belly) side of the fish 815.12 Simulation results for the pulsed conveying of a large, soft salmon whentwo slanted holders are used to hold the fish: (a) the holding force, (b)the undeformed state of the fish prior to starting the holding process andthe deformed state of the fish (c) viewed from the dorsal (back) side of thefish (d) viewed from the ventral (belly) side of the fish 82xlii5.13 (a) The approximate model for the staggered—lug conveying process and(b) the resulting free—body diagram for the pusher 835.14 Constant speed conveying of a large, firm fish, using staggered lugs. . 845.15 Pulsed conveying of a large, soft fish using staggered chain lugs, with theholders oriented perpendicular to the line created by the lugs 855.16 Pulsed conveying of a large, soft fish using staggered chain lugs with theedges of the holders parallel to the corresponding sides of the table. . 875.17 A sketch of the holding pan used in the simulations 885.18 Pulsed conveying of a large, soft fish using a pan and parallel holders. . 895.19 Pulsed conveying of a large, soft fish using a pan and slanted holders. . 905.20 Holding forces for the pulsed conveying of a large, soft fish using a holdingpan with (a) parallel holders (b) slanted holders 915.21 Pulsed conveying of a large, soft fish using a rotated pan and slanted holders. 926.1 The fish—gill cover gap 946.2 Components of the push—down mechanism: (a) the finite element modeland (b) simulation result for indexing 966.3 Overfeed motion (in the x—direction) of the fish when the fish conforms tothe table surface under gravity loading (ftsh_taUe 0.1) 996.4 Overfeed motion (i—direction) of the fish as a function of the coefficient offriction between the fish and indexer (if:sh_tab1e = 0.1). Simulation resultsfor indexing large fish with (a) firm muscle (b) soft muscle 1006.5 Overfeed motion (z—direction) of the fish as a function of the coefficientof friction between the fish and indexer (fish_tab1e = 0.05). Simulationresults for indexing large fish with (a) firm muscle (b) soft muscle 101xiv6.6 Overfeed motion (x—direction) of a large, soft fish when the holder is notused (ILfiah_taUe = 0.1) 1026.7 Overfeed motion (x—direction) of a large, soft fish as a function of thecoefficient of friction between the fish and indexer, when the position ofthe holder is corrected (f&taUe = 0.1) 1036.8 A movable carousel system to permit the manual adjustment of the holderposition 1077.1 The laboratory prototype head—butchering machine 1097.2 The industrial prototype head—butchering machine 111xvAcknowledgementI would like to thank Dr. C. W. de Silva for his constant supervision and guidancethroughout this research. Funding for this research was provided by the Natural Sciencesand Engineering Research Council’s Industrial Automation Chair held by Dr. C. W. de Silva,and through a G. R. E. A. T. scholarship awarded by The B. C. Science Council. I thankMr. A. Beatty for his help with the vision component of the research. I would like tothank my family for their constant support and encouragement. Last but not least, Ithank my wife Angelina for her support, understanding and patience.xviChapter 1Introduction1.1 History of Fish ProcessingThe principles of canning had been established by 1809 and have not changed significantlyto this day, although canning methods and canning machinery have changed considerably.The world’s first salmon cannery was established on the Sacramento River in 1864 bythe Hume Brothers, William, G.W., and R.D., and by Percy Woodson [1]. At that timeall the operations on the canning line were performed manually. By the early 1900’s,several labor—saving machines were developed. The Jensen can filler automatically filledcans to the required weight. A topper was employed to automatically seal the lids on thecans. In 1903, the first Iron Butcher (also known as the Iron Chink), was introduced in acannery in Bellingham, Washington, [1]. This machine greatly increased the productionrates for preparing round (whole, undressed) salmon for canning and enabled the industryto expand rapidly. Over the years the Iron Butcher has been refined in an attempt toincrease the throughput rates and to improve the raw product recovery, but the basicmechanical design remains virtually unchanged.As advances in food science and mechanical engineering were made, new machinerywere developed for the fish processing industry. Filleting machines employ two parallelknives which are separated by a distance slightly greater than the diameter of the fishbackbone. Feeding a fish lengthwise into the machine results in two cuts being madealong the length of the fish and on either side of the backbone. The backbone and the1Chapter 1. Introduction 2attached strip of fish meat is discarded from the two resulting fillets. Skinning machinesare comprised of a steel drum which uses a refrigerant to freeze the outer surface of thedrum, and a conveyer to transport fish past the drum. The drum rotates so that thesurface speed of the drum is identical to the speed of the conveyer. As a fish movespast the drum the skin touches and freezes to the drum, and is peeled from the fish.Head cutting machines and large vacuum pumps for unloading fish from boats have alsobeen developed. Technology employed in the fish processing industry is for the mostpart mechanical in nature. These first—generation machines are typically hydraulicallydriven and employ little or no sensing to perform their desired task. Sensors used havebeen almost exclusively limited to primitive mechanical sensors. For example, the weighstation, which automatically weighs cans after they are filled, uses a lever—arm balanceto determine if the cans are under—weight, over—weight, or filled to the correct weight.Correctly filled cans are sent to the topper for the lids to be fit and are vacuum sealed,while those cans which are under—weight are diverted to filler stations for correction.The use of electronics and image processing techniques in fish processing has receivedattention only recently. Some recent developments can be pointed out. Automatedweighing stations grade fish on the basis of weight, sending fish in a specific weightrange to a specific location. The HD1S Herring Sex Discriminator 1 uses an infra—redlight source and sensor to separate herring according to the sex of the fish. In order tostandardize the procedures and to ensure that high government standards are met whencooking the cans of salmon, the retorts have been retrofitted from manual operation tocomputer control with manual supervision.The Fisheries Technology Group Ltd, Canada, has developed a Fish Monitoring System (FMS 1000) 2• The FMS 1000 captures a computer image of backlit fish and uses a‘HDlS Herring Sex Discriminator is manufactured by Neptune Dynamics Ltd. 180—6751 Graybar,Richmond, British Columbia, Canada2FMS 1000 is a registered trademark of Fisheries Technology Group Ltd, P.0. Box 9190, St. Johns,Chapter 1. Introduction 3fish species recognition formula to determine the particular species of fish. Results fromthe image analysis and an automated weighing station are used to determine the weightof the fish and generate drive signals for operating the gates for sorting the fish by speciesand weight.A machine that was first tested in the fish processing industry but that has potential applications throughout all food processing industries is Sentinel Vision Inc.’s SeamSensor . The Seam Sensor uses image processing to inspect, in real—time, the seal oneach can after it is sealed. Improperly sealed cans are removed from the processing line,reducing the likelihood of botulism in the canned product.1.2 The Iron ButcherThe Iron Butcher, which was developed to replace the skilled manual labour of butchers,employs fixed average settings. As long as the size of the fish being processed is thesame as what the Iron Butcher is set—up to process, assuming that the machine operatesproperly, the raw product recovery rate of the Iron Butcher approaches that of manualoperation, while the processing efficiency of the machine exceeds that of manual labor.However, since the Iron Butcher has no means of adjusting to the size of each fish runthrough it, the processing efficiency quickly falls below that of manual labor as the rawproduct recovery rate falls.Salmon placed on the Iron Butcher is headed (the head is removed), split, eviscerated,and further cleaned before moving on for subsequent processing. Figure 1.1 is a pictureof the Iron Butcher from the feeding to cutting operations. The operation of the IronButcher is as follows: A worker places a salmon on the feeding table so that the salmonis oriented with its head pointed towards the cutter and its belly facing away from theNewfoundland, Canada AlA 2X93Sentinel Vision Inc. 4—7449 Hume, Ladner, British Columbia, CanadaChapter 1. Introduction 4Figure 1.1: The Iron Butcher: (a) positioning (conveying and indexing) the fish for thehead cut (b) feeding fish onto the conveyer. Conveyer motion is from right to left.direction of the conveyer motion. The worker must also align the salmon so that the endof the gill cover, which also corresponds to the collar—bone location, lies between twomarkers (see Figures 1.2 and 1.3). The feeding table has a horizontal door that pivotsabout one end, allowing the salmon to slide onto the conveyer bed. Pivoting motion ofthe feeding door is synchronized with the conveying mechanism.Conveying of the fish is accomplished by three chains which move in recessed slotsin the machine bed. Lugs, extending 5.5 cm beyond the level of the machine bed, areattached to the chains with an inter—lug spacing of 23 cm. These lugs push the fishin the direction of the conveyer motion (see Figure 1.4). Lateral indexing (incrementalpositioning) of the salmon starts when a metal foot (indexer) drops onto the fish; . weightalso drops onto the tail end of the fish simultaneously, to hold down the fish. The indexerand weight move along the machine bed at the same speed as the chains, but in additionthe indexer simultaneously moves laterally towards the cutter, as a result sliding acrossthe lengthwise direction of the fish toward its head. When the indexer foot reaches the(a) (b)Chapter 1. Introduction 5Figure 1.2: An illustration of anatomical features of salmon which are of interest whenprocessing salmon on the Iron Butcher.collar—bone, near the gill cover, it is expected to lock in with this region of the fish andthe entire fish is pulled laterally by the indexer toward the cutting line (see Figure 1.5).The indexer positions the fish so that the collar—bone of the fish is just beyond the cuttingline. The indexer then retracts from the fish just prior to the removal of the head bythe cutter (see Figure 1.6). The cutting blade is specially shaped to remove both thefish head and pectoral fin while minimizing the amount of meat (i.e., waste) which isleft attached to the head. A schematic diagram of the Iron Butcher, showing its maincomponents, is shown in Figure 1.7.1.3 Motivation of the ResearchWhen the Iron Butcher is adjusted correctly for the size and the species of fish beingprocessed, the head is severed from the body no more than 6 mm posterior of the collarbone and pectoral fin; this is considered a good cut. Two major errors can occur whenindexing the salmon for a head—removal operation, and both involve improper positioningof the fish with respect to the cutter.Chapter 1. Introduction 6— Hinged side ofthe feedingtable— Gill-positionmarkers— Feeding TableFigure 1.3: Positioning a salmon on the feeding table.The first type of error, underfeed, occurs when the fish is not moved enough, sorather than the cutter removing the head and collar—bone from the body, the cutter cutsthrough the head itself, leaving part of the head and all of the collar—bone attached tothe body. This happens when the indexer foot does not engage with the collar—bone, buteither depresses the soft gill cover towards the nose, or completely slips over the gill cover.Underfeed requires the fish to be re—processed, either by the Iron Butcher or manually.In either case, the throughput of the machine is reduced and added expense is requiredto remove the head and collar—bone correctly. The second type of error, overfeed, occurswhen the fish is moved too much, so that the cutter removes the head from the bodymore than 6 mm posterior of the collar—bone (see Figure 1.8). This situation occurs whenthe indexer foot becomes engaged with a different structural region prior to reaching thecollar—bone, possibly due to softness or physical defect of the fish body. In many casesthe wastage can be so significant that the positioning error is as high as 75 mm posteriorof the collar—bone.Chapter 1. Introduction 7The ability of the fish processing industry in British Columbia to maintain its globalcompetitiveness will depend upon several factors. Increasing the productivity and reduction of the wastage it currently incurs while processing the raw product are two suchcrucial factors. In the past, when the price of the raw material was low and the fishstocks were apparently high, the processors could afford to accept the wastage that theselabour—saving machines inherently produced. But with limitation of the catch, mostlydue to dwindling stocks, and the sharp increase in price of the raw product over the lastfew years, processors are finding it more difficult to remain competitive while acceptingthe wastage incurred when these machines are used for processing. The head cut on theIron Butcher accounts for the largest percentage of raw product loss during the processingof salmon for canning. Every one percent increase in the raw product recovery rate willsave the fish processing industry in British Columbia approximately five—million dollarsannually . It follows that wastage reduction has direct economic benefits, if not being a199O figures— IndexerFigure 1.4: Lugs, used for conveying the salmon, are shown on either side of an indexer.Chapter 1. Introduction 8WeightIndexerLugFigure 1.5: An indexer is locked into the collar—bone region of a fish and is indexing thefish for the head cut while a weight rests on the tail—end of the fish.requirement for survival of the industry.A pioneering effort in developing advanced technology for fish processing automationhas been undertaken with the establishment of the Industrial Automation Laboratoryin the Department of Mechanical Engineering at the University of British Columbia [2].Specific projects include knowledge—based hierarchical control of a fish processing cell toaddress the concerns of raw product recovery, throughput rate and the quality of cuttingwhile subject to various specifications. Model—based vision is being used to infer complexfeatures of an object from simple geometric attributes of the object. This methodologyhas been successfully applied to the generation of cutting contours for Pacific salmon.Coordinated manipulation and control is being applied to the holding and manipulationof a fish while it is being cut. The grasping and handling of a non—homogeneous objecthas been addressed by the development of an innovative gripper.Chapter 1. Introduction 9Indexer— SalmonLeading edgeof the cutterFigure 1.6: The indexer retracts from the fish just prior to the removal of the fish headby the cutter.1.4 Research ObjectivesPositioning a salmon for the head cut on an Iron Butcher is neither an accurate nor highlyrepeatable operation. The ultimate goal of the research described in this dissertation isto recommend design improvements which may be used in a redesign of the Iron Butcherthat will result in better accuracy and repeatability. These improvements will lead toan increase in the raw product recovery rate. In order to accomplish this, the followingobjectives are defined:1. Develop a model of the fish—machine interactions for the processing offish on the Iron Butcher. The model is an analytical computer—simulation toolused to investigate the problems encountered during the conveying and indexingprocesses on the Iron Butcher. Simulation has the benefit of easy variation ofChapter 1. Introduction 10Feedingtablesystem parameters and evaluation of the dynamic response, thereby providing insight into the underlying problems with the Iron Butcher without the cost of timeand resources required to construct or modify prototypes and conduct controlledexperiments.2. Identify the magnitudes and locations of the forces to be applied toa salmon for proper holding and conveying. While the salmon is beingconveyed, it is desirable to constrain the fish to maintain a fixed orientation withrespect to both the cutter and the indexer while achieving a stable motion priorto cutting. Also, different stages of relative motion may be needed. For example,initially the indexer foot will slip over the top surface of a fish, but the fish itselfwill not slip laterally on the conveyer. Next the fish is expected to slip laterally aspushed by the indexer, but there should not be any slip between the indexer andthe fish. The magnitude of holding forces should be minimized to reduce bruisingCutterFigure 1.7: A schematic representation of the Iron Butcher, showing the crucial components.Chapter 1. Introduction 11Figure 1.8: A comparison of the amount of meat attached to the fish head (waste)resulting from a good cut (on the right) with that from a cut where overfeeding erroroccurred (on the left).of the fish body, while maintaining them at sufficient levels to achieve the requiredmotion stages. Two types of conveying are investigated, conveying at a constantspeed and pulsed or intermittent conveying.3. Identify and quantify the parameters which will lead to the elimination ofboth underfeed and overfeed errors. Many parameters and variables are activeduring indexing, but the role each plays in determining if the salmon is improperlyindexed (primarily underfeed or overfeed) has to be investigated. After identifyingpossible “problem” parameters and variables, the significance of each variable indetermining the success/failure of the indexing process can be systematically determined. Finally, an acceptable operating range for each problem parameter isChapter 1. Introduction 12required in order to determine design guidelines for the new machine.4. Establish design guidelines and make recommendations for the crucialcomponents of a typical butchering machine. The results of objectives 1—3 will lead directly to the establishment of design guidelines for the conveying,indexing and holding mechanisms. These recommendations may take many forms(e.g., parametric, geometric, active or conceptual).1.5 Organization of the ThesisThis chapter has provided an introduction of the development of fish processing in BritishColumbia. The operation of the Iron Butcher has been explained. Finally, the motivationfor and the formal objectives of this research were presented. Chapter 2 contains a reviewof literature relevant to this work.Chapter 3 begins with an introduction to the finite element method, followed by adiscussion of the material properties of salmon tissue. Assumptions which are requiredfor the formulation of the finite element model of the salmon, along with details of themodel development, are then presented. The chapter ends with the verification of thefish model.Chapter 4 deals with the model of the processing machine. The details of the modelare presented after the assumptions required for modeling the processing machine arediscussed. Next, the verification of the model of the processing machine is presented,and the chapter concludes with a discussion of the limitations of the model.Chapter 5 provides an analysis of the conveying process on the Iron Butcher. Simulations examine the operation of the machine in both its existing mode and in possiblemodified modes. Some modifications to the design of the machine are made and aresimulated to determine if the operation of the machine, in its new configuration and inChapter 1. Introduction 13the new mode, is satisfactory. The analysis of the indexing process of the Iron Butcher ispresented in Chapter 6. Simulation results identify mechanisms or dynamic interactionswhich may be responsible for underfeed and overfeed errors. Modifications to the designof the Iron Butcher are simulated to show that the design can be improved. Both Chapter5 and Chapter 6 conclude with discussions of the design recommendations for improvedperformance of the Iron Butcher.Chapter 7 discusses the implementation of design suggestions on two prototype head—cutting machines. Finally, Chapter 8 summarizes the thesis, outlines the main contributions, and makes recommendations for future work.Chapter 2Literature ReviewThe task of conveying and lateral indexing a salmon for the head cut on the Iron Butchermay be considered primarily as a problem of how to manipulate a flexible object thatis able to slide, by pushing it on a flat surface. Because of the similarities that existin the manipulation of sliding objects and the analysis and design of part—feeders, withthe conveying and positioning of fish on the Iron Butcher, literature dealing with thesetopics is reviewed in section 2.1. In the analysis of the dynamic interactions between afish and the Iron Butcher, the high processing speed and the interaction between objectsthat undergo relative motion on the Iron Butcher necessitate the inclusion of impactand contact dynamics of deformable bodies. A review of the impact, contact and graspin robotic applications is presented in section 2.2, as these applications were the onlyrelevant sources found in the literature review conducted. Biomechanic and food—texturestudies which experimentally studied the material properties of fish tissue are reviewedin section 2.3, in view of the importance of such experimental data in this investigation.2.1 Manipulation of Sliding ObjectsAn important issue in positioning a fish for the head cut is how to deal with the uncertainties in the initial location of the fish and shape of the fish. Mason [3] introduces amethod that is different from structuring the environment or using sensors to attack thisproblem. In that work, it is proposed that the task mechanics be used to eliminate theuncertainty without the need for sensing. A funnel has been defined as any operation14Chapter 2. Literature Review 15that eliminates uncertainty mechanically, and converges towards the positioning objective. The funnel will achieve the goal of positioning an object despite variation in theinitial location and shape of the object. Task mechanics are invaluable when manipulating flexible materials, as pointed out in that work. Rather than directly controlling allmotion degrees of freedoms of the flexible material, one must only exploit the naturaltendencies of that material.Mason [4], having recognized the importance of pushing as an essential component inmany robotic manipulator tasks, presents a theoretical basis for analyzing and planningpushing operations. Coulomb friction is assumed and the frictional forces are consideredto dominate over inertial forces. The motion of the object is determined, which dependson whether or not the object rotates, and on the direction of rotation.Brost [5] presents a method for the automatic planning of robot grasp motions fora parallel—jaw gripper that is insensitive to bounded uncertainties in the location ofthe object. His method uses the physics of friction to generate all grasp plans thatutilize either a squeeze—grasp, offset—grasp, or push—grasp motion. These three motionsare essentially examples of Mason’s mechanical funnel [3] as applied to a parallel—jawgripper. Brost only examines planar motion and uses Mason’s work in [4] to determinethe direction of rotation of the object.Peshkin and Sanderson [6] have developed a method to plan the manipulation of anobject that is free to slide without knowing the exact distribution of frictional forcesbetween the object and the surface it is sliding over. They have simplified the problemby assuming the following:• Zero friction at the pusher—object contact.• The support force distribution is confined to a disk.• The object is a 2—D rigid body which is pushed across a level surface.Chapter 2. Literature Review 16• A point contact exists between the pusher and the object.• Coulomb friction exists between the object—surface interface.• All motions are slow (quasi—static).Peshkin and Sanderson have extended their work [7, 8] to compute configuration mapsthat provide a basis for planning the operation sequences which occur in parts—feederdesigns or in more general sensorless manipulation strategies for robots.Gilmore [9] provides a rule—based algorithm to predict the dynamic motion of parts—feeders or pushing operations of manipulators. The algorithm automatically determines,predicts or detects the kinematic—constraint changes and reformulates the dynamic equations of motion, automatically simulating the dynamic behavior of systems with unpredictable or unforeseen kinematic—constraint changes. The rules have been developed using kinematics, system dynamics and geometric modeling, but ignoring frictional effectsand body deformations.The methods outlined thus far have in common Mason’s idea of a funnel. Except forGilmore [9], all assume slow motions. If the motion of the pusher and object is not slow,and if instantaneous relative motions occur, then impact dynamics must also be considered. Furthermore, the methods presented do not address the pushing of deformablebodies.2.2 Impact, Contact and GraspImpact dynamics must be considered for robotic and pusher operations when impulsive—type forces exist and inertial effects are not negligible compared to frictional effects.Wang and Mason [10] have explored the planar impact of two objects and developed agraphical method for predicting the mode of contact, the total impulse and the resultantChapter 2. Literature Review 17motions of the objects. The analysis includes the effects of elasticity and friction and isbased on the seminal work by Routh [11].Once the pusher and object are in contact, the relative motion between them is alsoof interest. Cai and Roth [12] have derived equations for the study of roll—slide motionsbetween rigid curves with point contact under planar motion, by considering only theinstantaneous time based kinematics. It has been assumed that a tactile sensor [13] isavailable to measure the relative motion at the point of contact and that the elasticdeformation of the sensor is negligible. Equations have been derived for the case of twopoint contacts as well, and this work has been further expanded to examine the 3—D caseof planar motion with point contact [14]. The derived equations for both the 2—D and3—D case were applied to sensor—based robotic path planning.In a different approach which does not require sensing of the motion of the contactpoint, Montana [15] uses differential geometry to derive a set of contact equations whichdescribes the motion of a point contact over the surfaces of two touching objects inresponse to a relative motion of these objects. Two assumptions, that the objects are rigidbodies and that there is only point contact between the objects, are made. Montana useshis contact equations to investigate the rolling of a sphere between two arbitrarily shapedfingers and for fine grip adjustment. He then extends this work to non—rigid bodies andderives compliant contact equations to model the kinematics of contact with compliance[16]. Results of his experiments show that compliance can cause large deviations fromthe rigid—body model, which supports the requirement of a compliant, or deformable, fishmodel in studying the dynamic interaction during the conveying, indexing and holdingoperations of the Iron Butcher.Saliba and de Silva [17] have studied the dynamics of planar objects on a flat surface,during grasping with a robotic hand. A hand has been designed and built for the experimental investigations. The innovative hand uses fewer actuators than it has degrees ofChapter 2. Literature Review 18freedom, and furthermore, a novel switch based on Coulomb friction has been employedin order to eliminate the need for electronic sensors in limit sensing and overload protection. The effects of Coulomb friction and various geometric and kinematic parameterson the dynamics of the grasped object have been analyzed and studied through bothlaboratory experimentation and computer simulation. The use of a gripper of this typein handling fish for an Iron Butcher—type machine has been investigated [18].Although certain aspects of the above work are applicable to the problem at hand, certain limitations exist. The various studies outlined above investigate specific tasks, whilethe problem at hand requires an integrated study that encompasses pushing, impact,contact and grasp of a flexible body with nonlinear and anisotropic material properties.2.3 Material Properties of Fish TissueMeasurement of mechanical properties of biological tissues is used in the study of biomechanics and in texture studies of food. Biomechanics seeks to understand the mechanicsof living systems, and data is available primarily for the human body. Texture studyof food attempts to provide a relationship between palatability (a subjective measure)and mechanical properties of the food (an objective measure). Data collected for texturestudies of fish deal with mechanical properties of both raw and processed fish.Petrell et al. [19] performed a preliminary study to determine the elastic properties ofthe transversely anisotropic structure of salmon muscle subjected to compression tests.The test samples consisted of blocks of muscle tissue, and consisted of both myotomes(muscle fibre) and myosepta (connective tissue between muscle fibres). They found alarge variation in the values of the moduli of elasticity E11 and £22 along the orthogonalaxes 1 and 2 (see Figure 2.1), but established the approximate range of values as:• 300 kPa 625 kPaChapter 2. Literature Review 19• 700 kPa E22 1100 kFa• 300 kPa < C12 550 kPa 1The moduli E11 and E22 were calculated using engineering stress and strain so thatthey are only valid for small strains. Many factors have been proposed as contributingto the wide variability in these elastic constants. “These include high intramuscularfat levels, insufficient exercise, pre—slaughter stress, pre—slaughter starvation, improperpost—mortem handling and time since slaughter” [19].EFigure 2.1: The orthogonal axes for fish muscle, as defined by Petrell et al.Johnson et al. [20] developed a procedure to measure the response of both fresh andcooked fish muscle subjected to compression by an Instron Universal Testing Machine.The modulus of deformability, M, defined as M UT/CT, where UT and CT are thetrue stress and true strain respectively, was used to characterize the overall resistanceof the material to deformation, instead of Young’s modulus E, because, strictly speaking, Young’s modulus is only applicable to linear elastic strains. Experimental results1Fjm a discussion with R.J. Petrell, August 6, 1992.MyotornesMyosepitiChapter 2. Literature Review 20indicated the linear region of the true stress versus true strain curves ranges from 20%to 40% and depends strongly on the species of fish. The deformability modulus variedbetween species of fish as well as between samples within a species of fish and the valuesare summarized in Table 2.1.Table 2.1: The modulus of deformability and the upper limit on the linear region of thetrue stress—strain curves for various species of fish, as determined by Johnson et al.Fish Species [ M (kPa) Upper limit of linear regionrTst #1 Test #2 (% strain)Hake 17—31 23—38 20—25Pollock 20 — 54 15— 39 20 — 25Flounder 15 — 88 25— 35(Pseudopleuronecetes americanus)Bluefish 16 25(Potatomus soltatrix)Cod 73 25—30( Godus morhua)Table 2.2: Recorded values of Young’s modulus and the associated coefficient of variationfor various species of fish, as determined by Borderias et al.Test #1 Test #2Fish Species Eaverage coefficient Eaverage coefficient(kPa) of variation (kPa) of variationTrout 86 18 137 19Sardine 53 26 98 8Conger 120 36 129 16Horse Mackerel 76 28 88 21Blue Whiting 111 32 80 48Borderias et al. [21] attempted to establish a link between sensory analysis of fishtexture with instrumental testing of the mechanical properties of fish flesh. The experimental investigation of texture was performed using an Instron model 1140 TexturometerChapter 2. Literature Review 21and each test consists of five replicates. A summary of the Young’s moduli and the corresponding coefficients of variation from compression tests on various species of fish areshown in Table 2.2.Petrell et al. [19] did not report any values of Poisson’s ratio for salmon muscle,but this parameter is estimated to be in the range of 0.3 ii 0.5, and is dependenton the water content of the muscle tissue, as disclosed in private conversations 2• Whenthe fish is alive the water content of the tissue is at its maximum (ii 0.5). As the timesince slaughter increases, the water content of the tissue decreases (v —* 0.3). Chow andOdell [22] performed a finite element analysis to investigate the deformations and stressesin the buttocks of a sitting person. They measured the muscle material properties in vivoand found Vhuman muscle = 0.49.Due to anatomical and structural mutations occurring post—mortem, the mechanicalproperties of soft tissue change, normally within a few hours of death. Yamada [23]analyzed data obtained from stress—strain tests performed on a variety of human skeletalmuscles. He found that, “in general, the post—mortem decrease in the ultimate strength ofmuscle tissue occurs quite rapidly during the first 24 hrs, slows down somewhat between24 and 36 hrs, and becomes substantially constant at 48 hrs.” Although Yamada’sobservation is for human muscle, it illustrates the rapid post—mortem deterioration inthe strength of muscle tissue which is common to all animal species.2From a discussion with R.J. Petrell, August 6, 1992.Chapter 3Model of the FishIt is required to analyze the dynamic interactions of a fish with the Iron Butcher duringthe holding, conveying and indexing processes. The purpose of the analysis is to predictthe dynamic behaviour of a fish when acted upon by various components of the IronButcher during processing. Parametric studies based on this may be employed to predictwhether specified conditions and structural modifications will result in a better holdon the fish while conveying and indexing, and what conditions are necessary to betterconvey and index the fish. The main objective is accurate positioning and the reductionof wastage at the head cut. The development of a model of the fish is the first steptoward these goals. The present chapter begins with a brief introduction of the FiniteElement Method in section 3.1. A discussion of the material properties of salmon tissue insection 3.2 provides the framework for establishing the model assumptions in section 3.3.Details of the fish model are elaborated upon in section 3.4, and the chapter concludeswith the verification of the fish model in section 3.5.3.1 The Finite Element MethodHuebner and Thornton [24] write:Continuum problems are concerned with fields of temperature, stress, massconcentration, displacement, and electromagnetic and acoustic potentials, toname just a few examples. These problems arise from phenomena in naturethat are approximately characterized by partial differential equations and their22Chapter 3. Model of the Fish 23boundary conditions. . . . However, the number of problems with exact solutions is severely limited, and most of these have already been solved. Theyare the classical problems.In such cases where the problem is so complex that an exact closed—form solution cannot be found, numerical methods of analysis are a viable alternative. Of these, two of themost popular are the finite difference method and the variational methods (e.g., the Ritzmethod). Because they do not support non—rectangular or non—uniform meshes, thesemethods encompass difficulties in defining boundary conditions along curved boundariesand difficulties in accurately modeling geometrically complex domains. With the finiteelement method (FEM), the assumed trial functions are defined over the element, not theentire solution domain, and must only satisfy certain continuity conditions, not boundaryconditions. As a result, the FEM can be applied to geometrically complex domains.In the processes under investigation, the loading on a fish varies with time. The fishis a complex structure comprised of hard tissue (bone) and soft tissue (muscle, internalorgans, connective tissue and skin). As a result, the numerical method of analysis chosenmust be capable of both static and dynamic analyses, and must account for the nonlinear,deformable and anisotropic material of the fish. Because the FEM is able to handle suchrequirements, this particular method is used in the present research to model a fish.3.1.1 Solution Procedure in the Finite Element MethodThe solution of a problem by the finite element method follows an orderly process. Thefollowing steps summarize in general terms how the solution process of the finite elementmethod proceeds.1. Discretize the solution domain. The solution domain is divided into subregions called elements. A variety of element shapes is available to be selectedChapter 3. Model of the Fish 24from, and if care is taken, more than one element shape may be used in the samesolution domain.2. Select interpolation functions. For each element, assign nodes and then choosethe type of interpolation function (shape function) to represent the variation of thefield variable over that element.3. Determine the element properties. Select the parameter values for the elementproperties and determine the matrix equations expressing the properties of theindividual elements.4. Obtain the system equations by assembling the element properties. Combine the matrix equations describing the behavior of the elements to form the matrixequations defining the behaviour of the entire solution region or domain. The system equations are modified to account for the boundary conditions of the problem.A vector for the loading applied to the system has to be defined at this stage.5. Solve the system equations. The system equations (a set of second—order simultaneous differential equations in the general dynamic case) are solved to obtainthe unknown nodal values of the field variable. Solutions can be obtained whetherthe system equations are linear or nonlinear. For a dynamic analysis, the equationto be solved is[MJ{k} + [C]{k} + [K]{X} = {F} (3.1)where [M], [C] and [K] are the global mass, damping and stiffness matrices respectively, {?}, {.k} and {X} are the global acceleration, velocity and displacementvectors respectively, and {F} is a vector of applied forces.6. Perform additional computations if required. The solution of the systemequations can be used to calculate other important parameters. For example, if theChapter 3. Model of the Fish 25solution of the equations for a structural analysis gives the displacement at eachnode in the solution domain, these results can be used to approximate the stressfield in the solution domain.The above procedure is followed to arrive at a solution using the finite element method.However, once a solution is obtained, it must be validated. Validation usually involvescomparing the finite element solution results to one or more of the following:• Experimental results.• An exact solution, if it exists. (Sometimes an exact solution may exist for a specialcase of the problem being investigated. The finite element model is developed fora general situation, which then is reduced to this special case for which an exactsolution exits, so that a comparison can be made. If the results are in agreementwithin an allowable tolerance, then, assuming that the special case is sufficientlyrepresentative, the finite element model of the general case is considered valid, byassociation).• Other simulation results that are available and known to be valid.• An order of magnitude calculation.In addition to validating the finite element solution, an estimate of the error in thissolution, arising from the assumptions made to simplify the model (e.g. geometric approximations) and from round—off error, is required [25]. A convergence study, also knownas a mesh refinement study, has to be performed as well, to generate a convergence curve.To generate the convergence curve, the finite element simulation is solved several times,each time using a finer element mesh for successive simulation runs. One of the field variables being solved for is selected and is graphed against the simulation run—time. TheChapter 3. Model of the Fish 26resulting curve should show a decrease in simulation error with increasing solution times.Simulation error is estimated by the change in the value of the selected field variable fromtwo successive simulation runs. This approach will provide information for selecting themesh size for a particular problem.3.1.2 Application of the Finite Element Method to Biological SystemsSince its introduction, use of the finite element method has progressed from analyzingsimple structural problems through thermo—fluid processes to complex biological systems.Biological models typically consider either the bone structure (hard tissue), the soft tissue,or some combination of the two. Modeling soft tissue presents the most difficulty sincethe normal behaviour of soft tissues is time dependent. In addition, the experimentaldata regarding the mechanical properties of soft tissues are either scarce or unavailable.Even when data for the soft tissues are available, the conditions under which they arecollected often generate more questions than answers. As a result, assumptions regardingthe mechanical properties of soft tissues are extremely difficult to stipulate.3.2 Material Properties of Salmon TissueA salmon is a complex biological system comprised of hard tissue (bone) and soft tissue(muscle, connective tissue, organs, skin and fat). Because the mechanical behaviourof soft tissues is time dependent while the mechanical behaviour of hard tissue is timeindependent, in general, the material properties of hard tissue are more accurately knownand are more widely available. Nonetheless, the material properties of both soft and hardtissue must be defined for the salmon to be modeled using the finite element method.Chapter 3. Model of the Fish 273.2.1 Material Properties of MuscleIn a fish such as salmon, the flesh muscle is divided into layers of contractile fibres calledmyotomes which are held together by a thin connective tissue called myosepta. Themyotomes and myosepta are structured in a complex geometry, as shown in Figure 3.1.In addition to the complexity of the organization of the myotomes and the myosepta, theFigure 3.1: The metameric structure of fish muscle. The pattern of lines on the cross (1)and longitudinal (2) sections represents the arrangement of sheets of connective tissue(myosepta) in the muscle.[26]myotomes themselves may be curved, and have a non—uniform cross—section that oftenresults in coning (the tapering of the cross—section to a point). The complexity of thepattern of myotomes and myosepta, combined with the shape of the muscle fibre results inthe fish muscle behaving like an anisotropic, composite material. However, certain regionsof the fish muscle (the mid—anterior in either the epaxialis or hypaxialis) are straight andare of uniform cross—section. In these areas, the fish muscle can be considered to betransversely isotropic and can be characterized by the independent elastic moduli, E11and E22 in the two orthogonal directions (i.e., along the muscle fibres and perpendicularto the muscle fibres), the Poisson’s ratio ii, and an in—plane shear modulus C12.connective tissue (myosepta)(1) (2)Chapter 3. Model of the Fish 28As outlined in section 2.3, Petrell et al. [19] measured the elastic moduli B11 andB22 of salmon tissue prepared from these special regions of the fish. However, theirresults are one order of magnitude larger than results obtained by Johnson et al. [20]and Borderias et al. [21], who measured the modulus of deformability (M oT/ET)and Young’s modulus (B = cr/e), respectively, of samples prepared from fillets of variousspecies of fish. It is unlikely that this order of magnitude difference is the result ofthe method used to calculate B or M, since both Petrell et al. and Borderias et al.calculated B using engineering stress and strain, while Johnson et al. used true stress andstrain to calculate M. Further comparison of B for other animals shows that for beef,6.3 kPa < < 126 kPa [271, for sheep, E.e 157 kPa [28] , and for humans,15 kPmuscle a.As will be shown in section 3.5.1 of this work, reducing the values of B11, B22 andG12 which Petrell et al. report, by one order of magnitude, results in simulation resultsobtained using the finite element model of the fish which are characteristic of a fish withfirm muscles, as would be expected from fish in rigor mortis. The values of the materialproperties for characterizing soft fish are determined by comparing simulation results toexperimental results.3.2.2 Material Properties of OrgansBy observation, the belly region (comprised of muscle tissue and the organs in the bellycavity) of a salmon is softer than the dorsal region (comprised of solid muscle and somebone) of the salmon. Post—mortem changes also cause a build—up of gas in the bellycavity. No data are available to define the material properties of the contents of the bellycavity, but it can be characterized as a spongy solid with a Young’s modulus lower thanthe Young’s modulus for fish muscle. The Poisson’s ratio for the contents of the bellycavity is considered to be about 0.49.Chapter 3. Model of the Fish 293.2.3 Material Properties of BoneBone is a nonhomogeneous, anisotropic and composite material primarily composed ofcollagen and hydroxyapatite. Bone is hard and has a stress—strain relationship similarto many engineering materials [29]. A comparison of the Young’s moduli for bone andmuscle shows that in humans, the stiffness of bone tissue is approximately 670,000 timeslarger than that of muscle tissue (Emt l5kPa, Eb0 10, 000, 000kPa [23]). Thiscomparison reveals that under identical loading, deformation of bone is negligible whencompared to the deformation of muscle (soft tissue). Although approximate values forthe Young’s modulus for fish bone is not available, experience shows that, as in the caseof human tissue, fish bone is much harder than fish muscle.3.3 Model AssumptionsA fish is a complex biological system, and simplifying assumptions have to be madebefore it can be modeled mathematically. This simplification is achieved by makingassumptions regarding the material properties of the tissues and the geometry of thefish. These assumptions then serve as a basis upon which a finite element model can bedeveloped.3.3.1 Assumptions Pertaining to Material Properties of Biological TissuesThe following assumptions are used to simplify the representation of the material properties of biological tissues in the finite element model of a salmon.1. The fish muscle is transversely isotropic. The associated elastic constants are:Young’s moduli EUSCIe, E1e corresponding to orthogonal axes 1 and 2 (seeFigure 2.1), Poisson’s ratio v8die, and the in—plane shear modulus Gr.Chapter 3. Model of the Fish 302. The elastic constants J2mUSCIC and GU3de, and the Poisson’s ratio 1,muc1efor fish muscle are assigned the following range of values:• 15.3 kPa < <62.5 kPa• 31 4 kPa < <110 kPa• 15.7 kPa QmU8cle 57.9 kPa•musde 0.53. The fish muscle has a structural damping factor of 0.10 /3muscle 0.12, which isused to calculate the viscous damping matrix [C] in the dynamic equations.4. The effects of the fish skin are insignificant compared to the effects of the fishmuscle.5. The internal organs in the belly cavity behave like a spongy solid and are characterized by a Young’s modulus of E11 = (0.1) x a Poisson’s ratio of= 0.49, a shear modulus of G11 = (0.1) x and a structural dampingfactor of 13beIiy =j3muscie6. The fish bone is linear elastic and isotropic. The Young’s modulus for the fish boneis one order of magnitude greater than the Young’s modulus for fish muscle, i.e.,= 10 x Similarly, Gb01 = 10 x GU8CIe. The Poisson’s ratio for fishbone is estimated as = 0.3 [30].7. The mass density of all tissue is approximately the same, and is in the range1141 kg/rn3 p 1373 kg/rn.Chapter 3. Model of the Fish 313.3.2 Assumptions Pertaining to the Fish GeometryThe following assumptions are used to generate the geometry of the finite element modelof the salmon.1. The geometry of the model is based on a 1.48 kilogram, farmed, Chinook salmonthat was selected at random. It is assumed that the geometry of this fish is representative of the geometry of all fish (salmon) processed on the Iron Butcher.2. The fish is divided into three distinct regions, the head, the back, and the belly(see Figure 3.2).• The head of the fish is solid bone. The boundary between the head and thebody of the fish is defined by the collar—bone.• The back region is solid muscle.• The belly region is a muscle layer surrounding the spongy belly cavity.3. The collar—bone is straight but need not be perpendicular to the long axis of thefish, which runs from the nose to the tail.4. The nose of the fish is cut—off to facilitate meshing of the finite element model.5. All fins are removed.3.4 Details of the ModelIn this section, the procedure for developing the geometry for the finite element modelof the fish from an actual salmon is discussed, and furthermore, the limitations whichare imposed by the procedure followed are outlined. The manner in which the materialproperty values for the salmon tissue are assigned to the fish model is also presented.Chapter 3. Model of the Fish 32Figure 3.2: Simplification of the fish geometry.3.4.1 Development of the Fish GeometryThe profile and thickness of the finite element model of fish is based on the geometry of a1.48 kg, farmed, Chinook salmon made available by B. C. Packers Ltd. Standard imageprocessing techniques are used to extract the required data. The geometry of the cross—section is based on the cross—section shown in Figure 3.1, and not from the cross—sectionof the actual fish, because the required steak of the fish would lose much of its structuralrigidity, so that the observed cross—section would not be accurate.The entire model is set up in a non—dimensional form that is related to the fish length.By changing only one model variable, the fish length, finite element models for fish of anyarbitrary size can be easily created. However, it is noted that although a fish of any sizemay be easily modeled, the proportional size of various features of the fish is assumed toremain the same from model to model, (i.e., models are “geometrically similar”) since allBack RegionHead Region Belly RegionChapter 3. Model of the Fish 33features are related to the fish length. This method of defining all coordinates of a nodein terms of the fish length implies that the problems associated with conveying, indexingand holding are either independent of, or are weakly associated with, the slight variationin fish geometry that occurs from fish to fish. This assumption may be easily relaxed byincreasing the number of representative geometric parameters, but is not done so herefor the sake of brevity.The fish model is established such that the profile is defined in the z—y plane (asshown in Figure 3.2), with the length of the fish measured as the x—coordinate, the fishwidth (the belly to back distance) measured as the y—coordinate and the fish thicknessmeasured as the z—coordinate.Determining the Fish ProfileThe profile of the fish is determined by placing the salmon on a white, horizontal surface.The camera is positioned vertically above the fish and structural lighting is used to reducethe shadows cast by the fish onto the background (see Figure 3.3). The lens apertureFigure 3.3: The equipment setup for acquiring the image of a salmon.is set so that the contrast between the background and the fish is enhanced. An imageChapter 3. Model of the Fish 34is captured and sent to a SHARP GPB—1 image processing board residing in an IBM—compatible personal computer based on the 486 processor. Standard image processingtechniques are used to analyze the image. First, the image is thresholded to distinguishthe fish from the background. Next, the silhouette of the fish is established by identifyingthe boundary. The major principal axis of this silhouette is calculated along with 6, theangle the major principal axis makes with the horizontal (see Figure 3.4). The fishMajorPrincipalAxisFigure 3.4: Calculation of the major principal axis and 6, the angle of orientation of theprincipal axis.silhouette is rotated about its center of area so that the principal axis is horizontal andis divided into n (n = 100) equal sections. Finally, the perpendicular distance from theprincipal axis to the edge (boundary) of the silhouette is determined at each of the msections along the principal axis (see Figure 3.5). These distances are saved in a file as afraction of the overall fish length.The number of points selected on the back profile of the fish equals the numberof points selected on the belly profile in order to simplify the eventual meshing of theChapter 3. Model of the Fish 35•MajorPrincipalAxisFigure 3.5: Measurement of the perpendicular distance from points on the contour of thefish to the major principal axis.finite element model. The straight line joining the corresponding points on the back andbelly represents a cross—section of the fish (e.g., the line joining points T1 and B1, inFigure 3.5, forms a cross—section). The number of cross—sections, and hence the numberof points defining the profile, is determined by the results of the mesh refinement study,as described in section 3.5.2.Determining the Fish ThicknessUsing a similar set—up to what is used for determining the profile of the fish, an image ofthe thickness of the fish is obtained. An image of the fish thickness is acquired and imageprocessing measures the distance, perpendicular to the major principal axis, of points onthe edge of the image, and saves this data as a fraction of the overall fish length. Thisnon—dimensional data is a measure of the maximum fish thickness, as a fraction of thefish length, at discrete locations along the fish length. A linear regression is performedon this data to find a fourth order polynomial expression to describe the maximum fishthickness along the fish length. This equation is used directly in the finite element codefor the fish model.B,Chapter 3. Model of the Fish 36Geometry of the Fish Cross—sectionThe fish cross—section is based on the cross—section shown in Figure 3.1. Nodes areselected so as to:• adequately define the contour of the outer surface and the belly cavity• allow the cross—section to be easily meshed with the minimum number of elements• allow the coordinate axes of the elements to approximate the orientation of themuscle fibres• ensure the elements which are formed will satisfy the geometric constraints imposedon them by the mathematics used to formulate the element type (e.g., for 3—D brickelements, the recommended range for the angle between adjacent edges is 900±450).The cross—section of the finite element model uses 30 nodes and 23 elements to defineits geometry while observing the above constraints (see Figure 3.6). The coordinatesof the nodes on the cross—section are defined as a fraction of the maximum width andmaximum thickness at the cross—section. These fractions are kept constant along thelength of the fish.3.4.2 The Collar—bone and Gill CoverThe combination of the physical characteristics of the collar—bone and gill cover enablethe indexer to lock into this region of the fish and position the fish for the head cut.These structures are simplified and combined into one structure in the finite elementmodel of the fish. Three—dimensional spar elements (ANSYS LINK8 elements) have oneend defined at the junction of the fish head and body, along the side of the fish. Theother end of the LINK8 element extends above the side of the fish body. An indexerChapter 3. Model of the Fish 37Figure 3.6: The cross—section of the finite element model of the fish (right) is based onthe drawing (left).sliding on the top surface of the fish body will contact the protruding end of the LINK8elements (i.e., lock into this region of the fish), enabling the indexer to push the fish untilit is positioned for the head cut. The distance which the LINK8 elements extend abovethe side of the fish can be altered to simulate the degree of opening of the gill cover (e.g.,a gill cover that is depressed into the head or that is raised above the collar—bone).3.4.3 Assigning Values for Material PropertiesThe results given by Petrell et al. [19] provide an estimate of the Young’s moduli of agroup of muscle fibres in two orthogonal directions (E11 and E22). Although a range ofvalues is given for both and E22, that work does not provide any data on the pairingof the Young’s moduli for each sample. Therefore, in the present work, the maximumvalue of E11 is paired with the maximum value of E22, and the minimum value of E11is paired with the minimum value of E22. A linear distribution of parameter values isassumed within the range of these two extremes.Having created a criterion to generate pairings of E11 and E22 of fish muscle in thisThickness aChapter 3. Model of the Fish 38manner, and by assuming the value given by Chow and Odell [22] for Poisson’s ratio(ii = 0.49), it is possible to estimate a value for the shear modulus of the muscle fibre.Specifically, the shear modulus for fish muscle is defined according toE11 + E22G12= 2(1 + zi) (3.2)This results in values for C12 which fall within the range reported by Petrell.The freshness of the fish, which may be categorized by the firmness of the fish muscle,is one parameter which must be varied to determine the role it plays in the effectivenessof a fish processing machine in accurately positioning various salmon for processing. Inorder to reduce the required number of simulations to a practical level, however, onlythose material property values corresponding to what can be loosely classified as “soft”and “firm” fish are used in the simulations. A fish categorized as firm is indicative of afish which is extremely fresh, has been properly stored and handled, and may be in rigor,while a fish categorized as soft has either been improperly handled, improperly stored,or has been previously frozen and thawed. Table 3.1 lists the material properties to beused in modeling soft and firm salmon.3.5 Verification of the Finite Element Model of a FishVerification of the finite element model of a fish is a two—step process. First, the simulation results are validated by comparing them to data which are known to be accurate.Second, the error in the finite element model of the fish is estimated and a suitable meshdensity is determined from the results of a mesh refinement study.3.5.1 Model ValidationSince there are no known results which relate the response of a salmon to external forcingfunctions applied to it, it is necessary to conduct an experiment to determine such results.Chapter 3. Model of the Fish 39Table 3.1: Material properties for soft and firm fish used in the finite element simulations.[ Parameter Soft Fish Firm Fish(kPa) 15.3 62.5(kPa) 31.4 110GU8 (kPa) 15.7 57.9Vmu8 0.49 0.49f3muscle 0.10 0.12E11 (kPa) 1.53 6.25G11 (kPa) 1.57 5.79be11y 0.49 0.49,3beUy 0.10 0.12Eb0I (kPa) 314 1100Gb0 (kPa) 157 579Vbone 0.30 0.30In the experiment, a salmon is simply supported on two horizontal rails and is allowedto bend due to gravity. The rails are constructed from a square channel section ofaluminum that is rigidly mounted on a wall such that the rails are horizontal. Results ofthe experiment are recorded using the computer vision system and procedure discussedin section 3.4.1.Although experimental and simulation data exist for many points along the length ofthe fish, accurately matching the data points of the experiment with corresponding datapoints in the simulation is difficult because of the slight geometric differences which existbetween the actual fish and the idealized finite element model of the salmon, and becauseof the large deformations which occur when the fish bends over the rails. However, twopoints on the fish are easily identifiable and correspond to the same location on the fishin both the experiment and the simulation, and are used in the comparative studies. Thefirst point is at the intersection of the side of the fish which rests on the supports and thetip of the fish nose, and is referred to as the nose. The second point is at the location onChapter 3. Model of the Fish 40the side of the fish which rests on the supports where the caudal (tail) fin joins the body,and is referred to as the tail. These points are marked in Figure 3.8. Measurementsof the nose and tail are made relative to the rail nearest the nose and nearest the tailrespectively, and are given in polar coordinates.Experimental ResultsExperiments are conducted for two fish, designated Fish No.1 and Fish No.3. Both fishwere stored in a freezer prior to the experiment, but were completely thawed beforebeing placed on the supports. Slowly freezing biological tissue (i.e., taking it under roomtemperature conditions and placing it into a freezer) will allow ice crystals to form insidethe cells, and as the ice crystals grow they may rupture the cell wall. When the tissueis subsequently thawed it is softer and weaker than before it was frozen, since the cellswith ruptured walls do not contribute to either the structural rigidity or strength of thetissue. Because the fish used in the experiments have gone through a process of slowfreezing, they can be considered as representative of very soft fish.For each experiment, an image is captured and processed to obtain non—dimensionalmeasurements (see Figure 3.7). The data is dimensionalized using the distance whichseparates the supports, and is plotted in Figure 3.8. Table 3.2 lists the location of thenose and the tail relative to the supports.Table 3.2: Experimental results for the bending of Fish No.1 and Fish No.3.FISH No.1 FISH No.3Nose Tail Nose Tailr (m) I 8 (deg) r (m) I 8 (deg) r (m) I 8 (deg) r (m) 1 6 (deg)0.1288 206.4 0.1462 318.0 0.1525 222.4 0.1429 328.2Chapter 3. Model of the Fisho.oo-0.1c>-0.200.0\I‘‘0.00C-0.10>-0.200.0Figure 3.8: Dimensionalized experimental results for the bending experiments.Simulation Results41Simulations were run to verify that the material properties chosen to represent soft andfirm fish lead to general agreement between experiments and simulations which incorporate the finite element model of the fish developed in this chapter. The simulations inthis section are static analyses, which investigate the bending of a fish which is simplysupported in the same manner as in the experiments.Figure 3.9 shows the simulation results for the case when the fish material propertiesFigure 3.7: The captured image of the bending experiment of Fish No.1 (on the left) andFish No.3 (on the right).Experimental Results forthe Bending of Fish No.1I-7’/L-I ——--7\ \Experimental Results forthe Bending of Fish No.3I /s‘.L-l DT\-Nrts %\se I1—---:__-Ese rai—0.10 0.20 0.30 0.40Horizontal Position (m)0.10 0.20 0.30 0.40Horizontal Position (m)Chapter 3. Model of the Fish 42cited by Petrell et al. are used. The results for the case where the values of the fish material properties cited by Petrell et al. are reduced by one order of magnitude, are shownin Figure 3.10. Finally, Figure 3.11 shows the simulation results when the fish materialproperties are modified to obtain general agreement with the experimental results for thebending of soft fish (Fish No.1 and Fish No.3). The location of the nose and tail relativeto the supports for the simulations of the bending of soft fish is given in Table 3.3.Ii(a) (b)UFigure 3.9: Simulation results for the bending of fish using the material property valuesfrom Petrell et al.: (a) Material property values from the low end of the range (b) Materialproperty values from the high end of the range.(b)Figure 3.10: Simulation results for the bending of fish using material property valueswhich are an order of magnitude smaller than those cited by Petrell et al.: (a) Materialproperty values from the low end of the range (b) Material property values from the highend of the range.Table 3.3: Simulation results for the bending of Fish No.1 and Fish No.3.FISH No.1 FISH No.3Nose Tail Nose Tailr (m) S (deg) r (m) S (deg) r (m) I 6 (deg) r (m) S (deg)0.1265 189.8 0.1355 313.9 0.1586 215.0 0.1365 314.6(a)Chapter 3. Model of the Fish 43Figure 3.11: Simulation results for the bending of (a) Fish No.1 (b) Fish No.3.Discussion of ResultsWhen the material properties cited by Petrell et al. are used in the model of the simplysupported fish, virtually no bending occurs. However, when these values are reduced byan order of magnitude, the simulation results are representative of a firm fish, such asfish in rigor. These values provide an upper limit on the values of the material propertieswhich can be input to the finite element model of the fish.The percentage error between the experimental and simulation results for the bendingof Fish No.1 and Fish No.3, listed in Table 3.4, is calculated according to the equations—errorrrrmtht1x 100% (3.3)9exper:ment—8s:mulattonerrors = x 100% (3.4)where r and 8 are the polar coordinates and i denotes a measurement for the nose or thetail. The larger error between the experimental and simulation results for Fish No.1 isattributed to the process by which the material properties for soft fish are determined;the material properties are varied until the percentage error between the experimentaland simulation results for Fish No.3 are below 5%. As an additional check of the validityof the model of the fish, the same material properties are used in the simulation of Fish(a) (b)Chapter 3. Model of the Fish 44Table 3.4: The percentage error between experimental and simulation results for thebending of Fish No.1 and Fish No.3.Fish No.1 Fish No.3Nose Tail Nose Tailerror,. error8 error,. errors error, error8 error, error9(%) (%) (%) (%) (%) (%) (%) (%)-1.80 -8.04 -7.30 -1.29 3.98 -3.33 -4.14No.1. A general agreement between the experimental and simulation results for FishNo.1 indicates that these material properties are adequate for modeling soft fish usingthe model developed here.3.5.2 Mesh Refinement StudyThe mesh refinement study provides an estimate of the error in the analysis results. Thiserror is inversely related to the number of degrees of freedom in the model; as the numberof degrees of freedom increases, more CPU time is required to run the simulation, so areduction in the analysis error comes only at the expense of added CPU time. Basedon the information provided by the mesh refinement study, a suitable mesh density isselected for further simulation runs.A simulation of a simply supported fish is carried out using a coarse mesh for the fish.Three additional simulations, each having a successively finer mesh than the previoussimulation (see Figure 3.12) are carried out to generate the convergence curve data shownin Table 3.5. The convergence curves of the data are plotted in Figure 3.13, which revealthe characteristic knee. The steeper part of the curve, before the knee, indicatesthat a small increase in the number of degrees of freedom results in a sizable change inthe solution. The shallower section of the curve, after the knee, indicates that a largeincrease in the number of degrees of freedom results in a small change in the solution;Chapter 3. Model of the Fish 45Figure 3.12: Models used in the mesh refinement study.therefore it buys little added accuracy. If it is accepted that the model with the greatestnumber of degrees of freedom (3959) yields an approximation that is closer to the exactsolution, and if it is accepted that this approximation is a converged solution, then it isevident that the percentage difference between the 3959 degree of freedom approximationand the 1709 degree of freedom approximation is really a measure of the error in the1709 degree of freedom approximation. The 3959 degree of freedom approximation canbe considered converged since the changes in LiX and LZ from the 1709 degree offreedom approximation are less than 5 mm. Thus the error in the 1709 degree of freedomapproximation of the fish is approximately 8.0 %. Although the 3959 degree of freedommodel has less error associated with it than the 1709 degree of freedom model, the addedCPU time prohibits its use in the study of the fish—machine interactions in the subsequentwork. For the work which is reported in the following chapters, the model from the 1709degree of freedom approximation is modified so that the cross—sections are distributedmore evenly over the length of the fish.1709 degrees of freedom 3959 degrees of freedomChapter 3. Model of the Fish 46Table 3.5: Convergence curve data for the model of Fish No.3. CPU times are forsimulations on a Sun SPARC station IPX using release 5.0 of ANSYS.D. 0. F. CPU Time LiX (m) Change in iiZ (m) Change in(s) LiX (%) IXZ (%)539 582.510 0.017571 -0.083743Nose 989 1925.620 0.027771 57.7 -0.099534 18.91709 2966.860 0.035029 26.4 -0.11165 12.23959 78460.488 0.038004 8.49 -0.11679 4.60539 582.510 -0.025580 -0.078353Tail 989 1925.620 -0.048834 90.9 -0.10379 32.51709 2966.860 -0.055368 13.4 -0.11162 7.53959 78460.488 -0.058717 6.05 -0.11497 3.00Chapter 3. Model of the Fish 47I II 11111 P 11111:ose!o.000 1’iose-0.060-0.120—.ii.c0 500 1000 1500 2000 2500 3000 3500 4000 4500Simulation Degrees of FreedomFigure 3.13: Convergence curves for the finite element model of the fish.Chapter 4Model of the Processing MachineIn the preceding chapter, a finite element model for a salmon has been established. Next,it is necessary to identify the various components of the Iron Butcher and the interactionsbetween the fish and the machine. Operation of the Iron Butcher is recorded on videoand studied in play—back, and particularly in slow—motion, to gain an understanding ofthe entire processing sequence. Geometric measurements of the Iron Butcher provideadditional spatial information. Section 4.1 lists the assumptions which are required tomodel the machine. The model details are presented for each of the machine componentsin section 4.2. This section includes a discussion of modeling the contact behaviourbetween the fish and machine. The verification of the model of both the processingmachine and the fish—machine interactions is presented in sections 4.3 — 4.4. The modellimitations are discussed in section 4.5.4.1 Model AssumptionsThe following assumptions are made to simplify the process of modeling the Iron Butcherusing finite element analysis.1. All machine components on the Iron Butcher are rigid compared to the fish.2. The process of moving the fish from the feeding table on to the conveyor is notspecifically modeled. Instead, the simulations are initiated with the fish properlyoriented (i.e., the back of the fish facing the direction of the conveyor motion and48Chapter 4. Model of the Processing Machine 49the fish head on the side of the conveyor edge nearest the cutter) and lying flat onthe table of the Iron Butcher. The fish is initially at rest.3. The table surface is flat, horizontal and continuous (for example, the table does nothave recessed slots for housing the chains which convey the fish).4.2 Model of the Iron ButcherA complete description of the components and operations of the Iron Butcher is given insection 1.2. From this description and the assumptions of section 4.1, the finite elementmodel of the Iron Butcher is developed. The following sections discuss in detail themethod by which the various machine components and the fish—machine interactions aremodeled.4.2.1 The Table SurfaceThe table on the Iron Butcher measures approximately 1.80 m by 0.70 m and has three,2.0 cm deep slots to house the chains which are part of the conveying system. Thetable is modeled by a single, 8—node brick element (refer to the ANSYS manual for acomplete discription of the finite elements used) measuring 1.80 m by 0.70 m by 0.02 m,thereby providing a flat, horizontal and continuous surface on which the fish move (seeFigure 4.1). The eight nodes which define this element are constrained to have zerodisplacement in the three coordinate directions. Because this element type has linearshape functions, constraining all the nodes of the element in this manner results in aperfectly rigid surface, regardless of any force or pressure loading applied to it.Chapter 4. Model of the Processing Machine 504.2.2 The Conveying SystemThe conveying system on the Iron Butcher consists of the lugs and the chains; the lugs,which are attached to the chains, extend above the table surface and contact the fish,while the chains reside in the recessed slots of the table and transfer the locomotivepower from the hydraulic drive to the lugs. The finite element model is simplified bycombining the functions of both the chains and lugs into the lugs alone, and modelingthem as motion sources. Each lug in the model is defined by one 8—node brick element(ANSYS SOLID45 element). The lugs may be positioned for straight conveying (i.e.,the lugs form a straight line which is perpendicular to the conveyor motion) or staggeredconveying (i.e., the lugs form a straight line at an angle to the conveyor motion) as shownin Figure 4.2.\___________cuttingConveyor Lugs/ Line Indexer HolderFigure 4.1: Two views showing the finite element model of the Iron Butcher.Chapter 4. Model of the Processing Machine 51AIDirection ofConveyor LugsMotion(b)Figure 4.2: Orientation of the lugs for (a) straight conveying (b) staggered conveying.Motion of the lugs is defined by specifying the displacement profiles of the type shownin Figure 4.3 to the lug nodes which define the surface which touches the fish. As thesenodes on the contacting surface have displacement constraints imposed upon them, thesurface will be rigid compared to the fish. Only a single set of three lugs, which touchesthe belly of the fish, is required for the model. The size and position of the lugs areinitially taken to approximate those of the Iron Butcher. However, these parametersmay be varied in subsequent simulations in order to improve the existing design.4.2.3 The Holding MechanismThe function of the holding mechanism is threefold: (1) to prevent fish from being pushedby the indexer before the latter is engaged with the collar—bone of the fish, (2) to preventthe fish from rotating away from the lugs during the indexing process, and (3) to hold thefish during the cutting process. To accomplish this, a pad measuring 10 cm by 7.5 cm by2.0 cm and located between the two lugs furthest from the cutting line, applies a 8.4 Nforce to the fish. Although the pad and indexer are dropped onto the fish at the sameLugs(a)Chapter 4. Model of the Processing Machine 52Displacement Displacement(a) (b)Figure 4.3: Displacement profiles (motion sources) that specify the motion of the lugsfor (a) continuous conveying (b) pulsed conveying.time, the exact instant of contact is determined by the initial distance between the fishand the holder. The pad is modeled by a single 8—node brick element (ANSYS SOLID45element). The pad moves in the same direction and at the same rate as the conveyor,and in addition there is the vertical motion which causes contact between the fish andthe pad. The displacement profiles defining the motion of the pad are applied to thenodes which form the contacting surface. The Young’s modulus of the pad is two ordersof magnitude larger than that of the fish muscle, so the pad appears to be rigid withrespect to the fish. The location, size and magnitude of the holding force, as well as thenumber of holders, may be varied to establish the optimum method of holding the fish.4.2.4 The Indexing SystemThe indexing system on the Iron Butcher is expected to properly align the collar—bone ofthe fish with the cutter. The system consists of the foot, which comes into contact withthe fish, and associated mechanisms to move the foot, as necessary for the conveyingprocess. For simplicity, the functions of these associated mechanisms are combined withTime TimeChapter 4. Model of the Processing Machine 53the finite element model of the foot. The foot is modeled by a single 8—node brickelement (ANSYS SOLID45 element) with approximate dimensions of 7.5 cm by 3.6 cmby 2.5 cm and a mass of 0.85 kg. The size of the model of the indexer is slightly largerthan the size of the actual indexer so that contact between the fish and indexer canbe modeled while adding as few degrees of freedom as possible. The two perpendicularcomponents of motion (i.e., the conveying motion and the indexing motion) of the footare achieved by defining the displacement profiles of these motions at the nodes whichform the two surfaces which come into contact with the fish. The vertical motion ofthe foot is unconstrained after the foot drops onto the fish, so that the weight of thefoot alone will determine the vertical force between the foot and the fish. The Young’smodulus of the indexer foot is chosen to be two orders of magnitude larger than that ofthe fish muscle, thereby approximating the relative rigidity of the foot.4.2.5 The Fish—Machine InterfaceIn finite element analysis, a family of elements, known as contact elements, are usedto represent contact and sliding between two surfaces or between nodes. In the caseof contact between the fish and the Iron Butcher, the exact location of contact is notknown beforehand, so it is not practical to define in the contact model all possibilitiesof node—to—node contact which may occur. Instead, the surfaces which may come intocontact are specified, and a suitable contact element is defined to “link” these surfacesand model contact between these surfaces. The CONTAC49 element is selected as itrepresents contact and sliding between two surfaces in three dimensions.To model contact between the fish and the table surface, the nodes on the side ofthe fish which may come in contact with the table are selected to define the first contactsurface. Next, the nodes which represent the upper surface of the table are selected todefine the second contact surface. CONTAC49 elements are then defined to “link” theseChapter 4. Model of the Processing Machine 54two surfaces, permitting contact and sliding to be modeled. The procedure is repeated tomodel fish—lug, fish—holder and fish—indexer contacts. Unless otherwise specified, 1u = 0.1is used as the coefficient of friction between all surfaces.If a finite element model can be separated into a linear part and a nonlinear part, atechnique known as “substructuring” can sometimes be used to greatly reduce the CPUtime required to run a simulation. Substructuring can be used only if the nonlinearsolution options, such as plasticity, creep, swelling, stress stiffening and large deflection,are not used. In this instance, the elements which represent the fish and the componentsof the Iron Butcher (the table, lugs, holder, indexer and cutter) are linear, while thecontact elements are nonlinear. Because the linear and nonlinear portions of the modelcan be separated, the linear elements are submodeled using the MATRIX5O elementwhile the CONTAC49 elements are defined between the appropriate surfaces of the fishand machine components. In this work, the static analyses of section 3.5 do not usesubstructuring, as the large deflection option is used. However, in the simulations inChapters 4— 6, substructuring is used since the large deflection option is not used.4.3 Verification of the Conveying ModelCONTAC49 elements are used to model contact and sliding between the fish and themachine components. When friction has to be introduced, CONTAC49 elements arecapable of modeling either rigid friction or elastic friction, which are defined by thecharacteristic curves in Figure 4.4. Although rigid friction closely approximates thefriction in the real world, elastic friction has the advantage of allowing the solutionto converge more easily than if the rigid friction option is used. However, if K3, theparameter for the elastic friction option, is not properly evaluated, large errors can beintroduced into the simulation results. For the case of an object with an initial velocity,Chapter 4. Model of the Processing Machine 55FAI.iIFNoI—SlidingDistanceIFNoIFor FNo, < 0, and noreversed loading(a) (b)Figure 4.4: The force—deflection relationships for a CONTACT49 element specified with(a) rigid friction (b) elastic friction.sliding on a flat surface with a coefficient of friction , the object will exhibit oscillationsabout its steady—state resting position if the elastic friction option is used. The amplitudeof the oscillation is inversely related to the value of the elastic friction parameter, so carehas to be exercised in choosing this value.Verification is required to ensure that the assumptions for the Iron Butcher are valid,that the finite element model is properly defined, and that the use of elastic friction doesnot adversely affect the simulation results.4.3.1 Order of Magnitude CalculationConsider an object which is characterized by a mass m, damping constant c and stiffnessk, which is free to slide over a flat, horizontal surface. A coefficient of friction, , existsFi.’ IFNoISlidingDistancei IFNoIChapter 4. Model of the Processing Machine 56between the surface and the object. Motion of the object over the surface can be describedby equation (4.1), where , , and x are the acceleration, speed and position of the object,respectively, and f is the sum of the applied forces.mà+cc+kx=—f (4.1)If the only forces acting on the object are gravity and friction, thencx=kx=0 (4.2)and equation (4.1) reduces tox=—pg (4.3)By successively integrating equation (4.3) we getth= —pgt + ci (4.4)and— pg+cit+c2 (4.5)Equations (4.4) and (4.5) require initial conditions to solve for c1 and c2. For example,if at time t = 0 seconds the object is moving at 0.5 rn/s (the conveying speed on theIron Butcher), and the corresponding instantaneous location of the object is taken as thereference point, thenth (0) = 0.5 rn/s x(0) = 0.0 rn (4.6)The initial conditions given by equation (4.6) are used to solve for the constants inequation (4.4) and (4.5). This results inx= —pgt + 0.5 (4.7)x=+0.5t (4.8)Chapter 4. Model of the Processing Machine 57By solving equation (4.7) for the time at which the object has zero speed, we gett1 = seconds (4.9)pgBy substituting ti into equation (4.8) we can determine the distance the object travelsbefore coming to a stop, as(t) = 0.125 (4.10)Equation (4.10) approximates the distance the center of mass of a fish having an initialvelocity of 0.5 rn/s will travel while sliding on the table of the Iron Butcher. With acoefficient of friction p = 0.3, this distance is computed to be x(ti) = 0.0425 meters.4.3.2 Simulation ResultsNonlinear, transient, dynamic simulations are run to verify the conveying model for firmand soft fish. In these simulations, a fish starts from rest on the table of the Iron Butcherwith the lugs located just 4 mm away from the fish belly. The lugs move 0.10 m at the rateof 0.5 rn/s, pushing and accelerating the fish. The lugs then instantaneously stop moving,freeing the fish to slide on the table surface. Figure 4.5 shows the location of the fishmass center as a function of time. The mass center overshoots the steady—state restingposition and oscillates about this point before coming to rest. The actual overshoot andthe steady—state resting position of the fish mass center for the simulations of conveyingfirm and soft fish are listed in Table 4.1. Overshoot is defined as the difference between themaximum position and the final resting position of the mass center, while the steady—stateposition is the difference in the mass center location between the final resting positionand the position when the lugs stopped moving (time = 1.2 seconds).Chapter 4. Model of the Processing Machine 58Figure 4.5: Position of the fish mass center for the simulation of the conveying of (a) firmfish (b) soft fish.Table 4.1: Overshoot and steady—state resting position for the simulations of conveyingfirm and soft fish.Simulation Overshoot (m) Steady—state position (m)Firm fish 0.000860 0.041990Soft fish 0.001033 0.0421554.3.3 Discussion of ResultsThe simulation results are in close agreement with the order of magnitude calculations,with a 3.1% error for the conveying of firm fish and a 0.81% error for the conveying ofsoft fish. Because the order of magnitude calculation is, at best, only an estimate ofwhere the fish should come to rest, these error values cannot be interpreted as the actualerror in the conveying model. These errors only indicate that the simulation results areintuitively correct.The simulation results indicate that the use of elastic friction with the proper valuesI0.260.220.180.140.100.06/—-/0.260.220.180.140.100.06/:——12ZZZZ1.0 1.6 1.81.2 1.4Time (s)(a)I1.0 1.2 1.4 1.6 1.8Time (s)(b)Chapter 4. Model of the Processing Machine 59for the elastic friction parameter, does not produce large errors. The maximum overshootobserved in the simulation is 0.001 m, which is acceptable considering the scale of thefish and Iron Butcher, and the motions involved.The close agreement between the order of magnitude calculations and the simulationresults indicates that the assumptions pertaining to the conveying system of the IronButcher are valid, the conveying model is properly defined, and the values of the elasticfriction parameters are correctly chosen.4.4 Verification of the Indexing and Holding ModelContact of the fish, with both the indexer and the holder is modeled using CONTAC49elements. As was explained in section 4.3, the value of the elastic friction parameter K8has to be carefully selected to ensure that the error in the simulation resulting from theuse of the elastic friction model is acceptable. Verification of the indexing and holdingmodel is also required in order to determine if the assumptions pertaining to the holdingand indexing system of the Iron Butcher are valid, and if the finite element model isproperly defined.4.4.1 Operation of the Iron ButcherFrom the spatial measurements of the Iron Butcher and by observing recorded video—tapeimages of the indexing and holding process, the motion and timing of the indexer andholder are determined. This information forms the basis upon which the finite elementmodel of indexing and holding is evaluated.The initial location of the holder and indexer are described in sections 4.2.3 and 4.2.4respectively. After the lugs, along with the indexer and the holder move down the tablethrough 0.35 m, both the indexer and the holder are released and dropped onto the fishChapter 4. Model of the Processing Machine 60body. At this instant, the indexer also starts to move laterally to position the fish forthe head cut. The lugs, the indexer and the holder move through 0.30 m down the tableas the indexer moves through 0.09 m laterally. When the indexer completes its lateralmotion, it is lifted off the fish so that the head cut can be made. Now the fish should bepositioned with its collar—bone on the cutting line and the belly resting firmly againstthe lugs. This entire process takes approximately 1.3 seconds. Figure 4.6 is a schematicrepresentation showing the motion of the lugs, indexer and holder.CuttingLinemFigure 4.6: Schematic representation of the Iron Butcher showing the location and motionof the lugs, indexer and holder during the indexing and holding process. (1) Initialposition (2) Position when the indexer and holder drop onto the fish (3) Final positionwhen indexing is completed.4.4.2 Simulation ResultsA nonlinear, transient, dynamic analysis of the indexing and holding process is carriedout for both a small and a large salmon. The small salmon has a length and mass ofChapter 4. Model of the Processing Machine 610.442 m and 1.48 kg respectively, while the large salmon has a length of 0.645 m anda mass of 4.59 kg. The coefficient of friction, 1, between all surfaces is taken as 0.10.The fish are initially located on the Iron Butcher such that the collar—bone is in—linewith the lug nearest the cutter. To reduce the solution CPU time, the lugs, indexerand holder are made to move just 0.05 m before the indexer and holder drop onto thefish, and furthermore, the indexer is not lifted off the salmon after indexing is complete.These are the only differences between the operation of the actual Iron Butcher and thesimulation.Figures 4.7 and 4.8 show the simulation results, for the small and large salmon respectively, at discrete points in time. Time = 0.01 s, Time = 0.11 s and Time = 0.71 scorrespond to stages (1), (2) and (3), respectively, in Figure 4.6. In both simulations thefish deforms from the weight of the indexer and holder when contact occurs. In addition,the location of contact between the fish and these components changes; the indexer slidesover the fish toward the gill cover before engaging with the gill cover, and the locationof contact between the fish and holder changes after the indexer engages with the gillcover and positions the fish. In both cases the fish are ultimately positioned with thecollar—bone at the cutting line.4.4.3 Discussion of ResultsComparison of the simulation results with the actual operation of the Iron Butcher yieldsfavorable results. Motion of the holder and indexer mirror those of the actual IronButcher. Contact and sliding motion between the fish, and both the holder and theindexer, is realistic. The indexer also correctly engages with the gill cover while indexingand the fish is correctly positioned for the head cut. These results indicate that the elasticfriction coefficient, K9, is properly chosen, the assumptions pertaining to the holding andindexing system are valid, and the finite element model is properly defined.Chapter 4. Model of the Processing Machine 624.5 Limitations of the ModelLimitations on the finite element model of fish processing can be attributed to the conflictbetween the opposing goals of accuracy and computing time. Although high solutionaccuracy is desired, it comes only at the cost of excessive computing time.The conflicting goals of simulation accuracy versus computational load must somehowbe balanced. This balance is achieved in section 3.5.2 when the mesh density for the finiteelement model of the fish is established. However, the trade—off between accuracy andcomputing time imposes limitations on the development of the model of the processingmachine.The chosen mesh density for the fish is characterized by an z—direction spacing of thecross—sections which is 6% of the fish length. As it is the case that, all other things beingequal, the larger the number of finite elements in a model the greater the computing timerequired to obtain a solution, every effort is made to reduce the number of elements used.Thus, given the mesh density of the fish model, the least computationally expensive wayto model contact between the fish and the machine components requires the machinecomponents to have an x—direction length greater than 6% of the fish length. Thisrequirement specifies chain lugs and an indexer to have x—direction dimensions which arelarger than those of the actual machine (see Table 4.2).Table 4.2: A comparison of the actual size of the machine components and the corresponding component size in the finite element models.Component Actual size (m) Finite element size (m)Small Fish Model Large Fish Modellug 0.025 x 0.0032 x 0.045 0.036 x 0.0032 x 0.045 0.053 x 0.0032 x 0.070indexer 0.025 x 0.075 x 0.025 0.036 x 0.075 x 0.025 0.053 x 0.075 x 0.025holder 0.10 x 0.075 x 0.02 0.10 x 0.075 x 0.02 0.10 x 0.075 x 0.02Chapter 4. Model of the Processing Machine 63Furthermore, for contact to occur, a node on the fish (which only exists at a cross—section) must lie within the bounds of a machine component. If nodes on adjacentcross—sections come in contact with the same component simultaneously, the modeledcontact will closely approximate the actual contact on the machine. But if nodes on onlyone cross—section touches a component, then the contact occurs at only one point, so theresults do not closely approximate the actual contact.Although this technique of modeling contact may seem to have a large detrimental effect on the simulation results, it in fact does not. Actual contact can be closelyapproximated by summing the contact forces and applying them at a single point. Thesimulation error which results due to contacts can be attributed to the difference betweenthe location of the resulting contact force and the location of where the resulting contactforce should be applied.Cl)I-’.oqI—’P-j-4 Cl)‘-3 0 b a 0) Ct CD C Ca‘-3 Ct 9 —1 a Cl, Ct C) C Ca Cl,I 0)-a.—CDcl•00 F I-i IF° C b ci, CD 0 C ci,CC,’Chapter 5Finite Element Simulation of ConveyingThe development and validation of the finite element models of the fish and the processing machine were presented in chapters 3 and 4. With these models it is possible toinvestigate the operation of the Iron Butcher in its present configuration, or in some modified configuration, without having to physically modify the machine. Since the currentIron Butcher cannot automatically compensate for errors in the position and orientationof the fish introduced during conveying, the fish must be reliably transported from onestage of processing to the next in such a manner that the error in its final position andorientation is minimal.The conveying process of the Iron Butcher in it present configuration is investigatedin section 5.1, to determine if the machine can accurately and reliably convey fish ofdifferent sizes and firmness. Section 5.2 then presents the investigation of conveyingwhen the configuration of the Iron Butcher is altered. Recommendations for the designof the conveying system are listed in section 5.3.5.1 Conventional ConveyingIn this section, the performance of the conventional Iron Butcher, during the conveyingprocess, is investigated. The simulations will address the question of whether or not theIron Butcher, in its present configuration, can reliably convey fish of different sizes andmuscle firmness.Nonlinear, transient, dynamic analyses are performed for four types of fish: small fish66Chapter 5. Finite Element Simulation of Conveying 67with soft muscle, small fish with firm muscle, large fish with soft muscle and large fishwith firm muscle. In each case, both the fish and the chain lugs are initially at rest.The lugs move through 1.0 m in 2.0 s at a constant speed of 0.5 rn/s. The coefficient offriction at both the fish—table interface and the fish—lug interface is set at 0.10.The simulation results are shown in Figures 5.1 — 5.4. In each case, a similar patternof motion is observed, which is characterized by an initial transient phase followed bya steady—state phase. Initially, the fish comes in contact with all three lugs. Then thecontact between the fish and the left—most lug is lost as the fish rotates clockwise. Atsome point, the fish reverses its direction of rotation so that the gap between the fishand the left—most lug is closed. Fish with soft muscle stop this “rocking” motion whenthe fish touches the left—most lug, so that from this point on the relative position of thefish with respect to the lugs remains fixed. The “rocking”continues for the firm fish untilthe contact between the fish and the right—most lug is broken and then re—established.The duration of the transient (rocking) period varies from approximately 0.6 s for theconveying of small, soft fish to 1.60 s for the conveying of large, firm fish.These results indicate that the Iron Butcher, in its present configuration, can reliablyconvey fish of different sizes and muscle firmness. Furthermore, the fish should be conveyed through 0.8 m before the next stage of processing is started so that the transient“rocking” motion can be sufficiently settled.5.2 Modified Iron Butcher ConveyingPrior to cutting the fish, the fish and cutter must be properly aligned. To accomplishthis, either the fish can be manipulated, the cutter can be manipulated, or both the fishand the cutter can be repositioned so that they are properly aligned for the head cut.In the current Iron Butcher design, the fish is moved to the cutter. However, becauseCDOq‘-1‘-1OCD 0CDijCl) H t1 o .0 :1 Cl) I-;- Cl) Cl)Cl) O. aq CD (DCD ‘Cl) CDa-. 0 Ho CDO II a0I-.CD Oq 00(tiQqI—4‘-1‘-iOCb 3CDI-srjC,CDqII-IcoOI.-.-I CD-.-. II b I-. ) -i)(tI II I- ào I- C’,I. COCDOqI-I•-1OCDAZc,cCDi,3‘lCDP3CD—•(DO•0Ii pqOCD•cl 0C,CDCDCDQq C,3.0CD 000.•.FC II I) b .. (A,11TH( -‘u1I vsI,—yI -3CDQ.iI-iOCD(D [I . 0 I-i OSD cO cI o CD (DCD —-I. - . 0 -“ CD II èo c0I-.CD I -4 I.Chapter 5. Finite Element Simulation of Conveying 72the fish is a flexible body whose exact shape, material properties and dynamic responseare not exactly known, it may be easier to position the cutter and not move the fish inthe final stages of “fine manipulation”. The appeal of this method arises from the factthat the control of robots and similarly controlled mechanical linkages, which the cuttermay be considered as a special case, is more widely researched and implemented than isthe control of flexible bodies. If the decision is made to move the cutter, then sensingis required to locate the position on the fish body to which the cutter should be movedfor performing the head cut. De Silva and Riahi [31] have shown that computer visioncan be used to successfully determine the location on a fish where the head cut shouldbe made. Recording the image of the fish is simplified and the accuracy is improved ifthe fish is stationary during imaging. Furthermore, the cutter design can be simplifiedand the accuracy of cut will be higher if the fish is stationary during cutting. For thesereasons, the results of pulsed (intermittent) conveying using the current configuration ofthe Iron Butcher is investigated in section 5.2.1.Gamage [32] has shown that the maximum yield for a straight cutter used for theremoval of salmon heads will occur if the fish is rotated clockwise through 300. Forthis reason, staggered conveying in which the fish is rotated clockwise through 30° isinvestigated in section 5.2.2. The use of pans to hold and transport the fish may be aviable alternative to the present system of chain lugs, and is investigated in section 5.2.3.5.2.1 Pulsed Conveyor Motion of the Current Iron ButcherIf conveying on the current Iron Butcher is changed from a constant speed to a pulsedmotion without making additional changes to the Iron Butcher, it is obvious that problems will occur during the conveying process. First, if a holding mechanism is not addedfor the conveying process, then the fish will remain unconstrained when the conveyorstops, so the momentum of the fish will cause it to continue moving until it collides withChapter 5. Finite Element Simulation of Conveying 73the leading set of chain lugs. Second, the lack of a proper holder may allow the fish tomove laterally during the cycling of the conveyor motion, which would result in a poorlylocated head cut.The results from a nonlinear, transient, dynamic analysis, in which the conveyormotion of the conventional Iron Butcher is modified from a continuous motion to apulsed (intermittent) motion, shows that an error of 0.01 m in the lateral position ofthe fish occurs after the fish is conveyed through just two cycles (see Figure 5.5). In thesimulation, the coefficient of friction at both the fish—table and at the fish—lug interfaceis set at 0.10. The conveying cycle is one second in length, consisting of a constantconveying speed of 0.5 rn/s for the first half—second, followed by a stationary period ofa half—second. These results illustrate the need for a holder that will securely hold thefish if a pulsed motion is used for its conveying.Two general types of holders are investigated to determine their effectiveness in holding the salmon during the pulsed conveying on the conventional Iron Butcher. Thebottom surface (i.e., the surface that is in contact with the salmon) of the first holderis parallel to the table surface, while the bottom surface of the second holder makes a7.6° angle with the table surface, with the edge of the holder that is away from the lugsbeing inclined towards the table top. The first type of holder is referred to as a parallelholder, since the surface touching the fish is parallel to the table top, while the secondtype of holder is referred to as a slanted holder because the surface which touches thefish is inclined towards the table top. The effectiveness of both a single holder and apair of holders of the same type is determined. All simulations are nonlinear, transient,dynamic analyses of conveying. Again, the coefficient of friction at the fish—table andfish—lug interfaces are set at 0.10, while the coefficient of friction at the fish—holder interface is set at 0.40. To maintain a processing rate of two fish per second, the periodof conveying is made to equal 0.5 s, which consists of a conveyor motion of 0.8 rn/s forCD C C’,I)CD C 0 00 C’, CD p 00 C’,f ,o CD020) 00CD)-+•02I— 0÷CD.,.C.q00 I-I—h0020÷ CDCi) 0÷ CD p,0200202.• ‘-4÷OCD CD- CD 02OqC0CD0 I—hCD I-1CDCD b C’,HFI-.CD I. -4CD p Ci)tffIiLllCD‘IC C’,Chapter 5. Finite Element Simulation of Conveying 75the first 0.25 s, followed by a pause in the conveyor motion for the final 0.25 s. All simulations are carried out through two conveying cycles to determine if a successful holdis achieved. A successful hold for these simulations is defined as a hold which maintainscontact between the fish and at least one chain lug throughout the conveying process.The first simulation tests the effectiveness of a single parallel holder. The simulationresults indicate that a single parallel holder cannot achieve a successful hold using theparameters defined above (see Figure 5.6). However, if a pair of parallel holders are used,a successful hold can be achieved (see Figure 5.7). Although this hold satisfies thedefinition of a successful hold, the holding forces and the resulting deformation of thefish are so large that a great deal of bruising and mutilation of the fish muscle wouldresult (see Figure 5.8). For these reasons, parallel holders cannot be used here.Successful holds are achieved for a single, slanted holder and for a pair of slantedholders (see Figures 5.9 — 5.10). The magnitude of the holding forces and the resultingdeformation of the fish that are required to obtain a successful hold with slanted holdersare much smaller than those experienced when parallel holders are used (see Figures 5.11— 5.12). Slanted holders will produce minimal bruising of the fish muscle and will notmutilate the fish at all. The use of a pair of slanted holders is preferred to a single,slanted holder for this holding task because the holding force is smaller in the formercase, where the motion of the fish head and tail is also smaller during the pause in theconveyor motion.5.2.2 Conveying Using Staggered LugsAs was previously noted, aligning the lugs to form a line at an angle of 30° to thecutter blade will maximize the yield of useful meat when a straight cutter is used atthe head cut. In this section, the conditions which must exist for a fish to slide alongthe staggered lugs during constant speed conveying are first estimated. Next, a suitableChapter 5. Finite Element Simulation of Conveying 76Time = 0.01 S Time = 0.11 s Time = 0.21s Time = 0.31sTime = 0.41 s Time = 0.51s Time = 0.61 S Time = 0.71 sTime = 0.81 s Time = 0.91sFigure 5.6: Simulation results for the pulsed conveying of a large, soft salmon with asingle parallel holder show that a successful hold is not achieved.holder is determined for the case of pulsed conveying, with staggered chain lugs.The conditions which must exist for a fish to slide along the staggered lugs during constant speed conveying are estimated from a kinematic analysis of the idealized process.First, the three lugs are replaced by a rigid, rectangular pusher, and the fish is similarlysubstituted by a rigid rectangular block. The pusher is rotated clockwise through 300.The approximate model and the free—body diagram for the pusher are shown in Figure 5.13. In this figure, F1 is the frictional force opposing the motion u of the fish overTime = 1.01 sChapter 5. Finite Element Simulation of Conveying 77Time = 0.11 STime = 0.51sTime = 0.91sTime = 0.21sTime=0.61sTime = 1.01 SFigure 5.7: Simulation results for the pulsed conveying of a large,parallel holders show that a successful hold is possible.soft salmon with twothe table surface, while F2 and F3 are the components of F1 parallel and perpendicularto the pusher surface. Assuming Coulomb friction between the fish and table,F1 = JL1 mf13h g (5.1)where t1 is the coefficient of friction between the table and the fish. From the geometry,Time = 0.01 s Time = 0.31sTime = 0.41 sTime =0.81 sF2 = t1 mj13h g SZfl8 (5.2)Chapter 5. Finite Element Simulation of Conveying 78z____________2________ __ __ _ __ __________0Figure 5.8: Simulation results for the pulsed conveying of a large, soft salmon when twoparallel holders are used: (a) the holding forces, (b) the undeformed state of the fishprior to starting the holding process and (c) the deformed state of the fish during theholding process.For the fish to slide along the pusher,F2 P2 mf3h g (5.3)where P2 1S the coefficient of friction between the fish and pusher. Substituting equation(5.2) into (5.3) and simplifying the result gives,P2simO— (5.4)With 6 = 300, sliding will occur if and only if1l 2p (5.5)In the present design of the Iron Butcher, P1 P2, so the fish will not slide along theTail Holder Head HolderII(b)0.25 0.50Time (s)1.25(a)(c)staggered chain lugs during the conveying process.Chapter 5. Finite Element Simulation of Conveying 79Time = 0.41sTime = 0.11 sTime = 0.51sTime = 0.21 s Time = 0.31sTime= 0.71 sTime = 0.81s Time = 0.91s Time= 1.OlsFigure 5.9: Simulation results for the pulsed conveying of a large, soft salmon when oneslanted holder is used, show that a successful hold is possible.Results from a nonlinear, transient, dynamic simulation for the staggered lug, constant speed conveying of a large, firm fish are in agreement with this kinematic analysis,as the fish does not slide along the lugs (see Figure 5.14). In this simulation the conveying speed is set at 0.5 m/s, while the coefficient of friction between the fish—table andfish—lug surfaces is assigned the value 0.1.Having shown that the fish will not slide along the lugs during constant speed conveying using staggered chain lugs, we shall proceed to develop suitable holders for pulsedTime =0.OlsTime = 0.61 sChapter 5. Finite Element Simulation of Conveying 80Time = 0.11 s Time = 0.21s Time = 0.31sTime = 0.51sTime = O.81s Time = 0.91 s Time= 1.OlsFigure 5.10: Simulation results for the pulsed conveying of a large, soft salmon when twoslanted holders are used, show that a successful hold is possible.conveying, again using staggered chain lugs. Nonlinear, transient, dynamic simulationsare run to determine a suitable design for the holders. Drawing on the simulation resultsof section 5.2.1, only the case of two slanted holders is investigated. The conveying speedis 0.8 rn/s and the period of the conveying cycle is 0.5 s. The coefficient of friction atthe fish—table and the fish—lug surfaces is 0.10, while the coefficient of friction at thefish—holder surface is 0.40. The orientation of the holders, with respect to the lugs, isvaried to determine its effect on the quality of hold. When the holders are rotated so thatTime = 0.OlsTime = 0.41s Time = 0.61 s Time = 0.71 sChapter 5. Finite Element Simulation of Conveying 81U-.---ziH:460o 0.25 0.50 0.75 1.00 1.25Time (s)(a)Figure 5.11: Simulation results for the pulsed conveying of a large, soft salmon when oneslanted holder is used: (a) the holding force, (b) the undeformed state of the fish priorto starting the holding process and the deformed state of the fish (c) viewed from thedorsal (back) side of the fish (d) viewed from the ventral (belly) side of the fish.they are perpendicular to the line created by the lugs, the fish is held so that it maintainscontact with the lugs. Unfortunately, the fish is squeezed forward as the holding forceis increased, thereby introducing an error in the position of the gill cover of the fish (seeFigure 5.15). However, if the holders are oriented such that the sides of the holdersare parallel to the corresponding sides of the table, the fish maintains contact with thelugs without introducing any error in the lateral position of the fish (see Figure 5.16).Therefore, two slanted holders, oriented as shown in Figure 5.16, are suitable for holdingthe fish during pulsed conveying, using staggered chain lugs.40_-‘I.’)-800-120- —---------- -(b)(c)Chapter 5. Finite Element Simulation of Conveying5.2.3 Conveying Using Holding Pans82The investigation of pulsed conveying with straight or staggered lugs in sections 5.2.1 and5.2.2 shows that although a successful hold, as defined in those sections, can be achieved,the motion of the head and tail of the fish is still quite large. The motion of the tail isnot a problem since this region of the fish is not processed here. However, the large headmotion during the pulsed conveying cycle will introduce positional errors in the locationof the collar—bone, which will adversely affect the yield of useful meat at the head cut.In this section, the use of a simple holding pan with holders is investigated to determineif a more reliable hold, with less undesirable motion of the fish head and tail, can beachieved.Tail Holder Head Holder-‘.IvonC11 hj4I— —— — — ——0— — — —— ——AnLTl]Ok\4I---—— — ————— — —-—----H ad I [o1 r(b)- 160o(c)0.25 0.50 0.75 1.00 1.25Time (s)Head Holder Tail Holder(a) (d)Figure 5.12: Simulation results for the pulsed conveying of a large, soft salmon when twoslanted holders are used to hold the fish: (a) the holding force, (b) the undeformed stateof the fish prior to starting the holding process and the deformed state of the fish (c)viewed from the dorsal (back) side of the fish (d) viewed from the ventral (belly) side ofthe fish.Chapter 5. Finite Element Simulation of Conveying 83Figure 5.13: (a) The approximate model for the staggered—lug conveying process and (b)the resulting free—body diagram for the pusher.The pan which is used in the simulations is shown in Figure 5.17. The pan is dividedinto two sections which are separated by 0.04 m to give adequate clearance for the cutter.The finite element model of the pan is developed in the same manner as that of the tableand the fish—table contact. Nonlinear, transient, dynamic analyses of pulsed conveyingare carried out with both the parallel holder and the slanted holder. For all cases,the coefficients of friction are maintained at the same values, with 1fish—pn = 0.1 andfish—holder = 0.4. The simulations proceed through two conveying cycles, with the periodof each conveying cycle being 0.5 s in length and with a conveyor speed of 0.8 ni/s.The simulation results show that a more reliable hold is achieved when the lug—holderconfiguration is replaced by the pan—holder configuration (see Figures 5.18 — 5.19). Themotion of the head and tail is greatly reduced and the magnitude of the holding forces areless (see Figure 5.20) when the pan—holder configuration is used. There is no differencein the motion of the fish whether a pair of parallel holders or a pair of slanted holders areused, however, the magnitude of the holding forces is slightly less for the pair of slantedUF2F1(a) (b)Chapter 5. Finite Element Simulation of Conveying 84Time = O.llsFigure 5.14: Constant speed conveying of a large, firm fish, using staggered lugs.holders. When the holding pan is rotated clockwise through 300, as in Figure 5.21, thenew configuration of the pan and holder convey the fish without the large, undesirablemotion of the head and tail which is present when the lug—holder configuration is usedduring pulsed conveying.5.3 Summary of Results and Design RecommendationsThe Results of this chapter are summarized below.5.1 When using lugs to convey at a constant speed, the “rocking” motion sufficientlysettles down after the fish are conveyed through 0.8 m.Time = 0.01 sChapter 5. Finite Element Simulation of Conveying 85Figure 5.15: Pulsed conveying of a large, soft fish using staggered chain lugs, with theholders oriented perpendicular to the line created by the lugs.5.2.1 When straight lugs are used with a pulsed conveying motion, a pair of holders arerequired to hold the fish. In addition, slanted holders provide a better hold with alower holding force than parallel holders.5.2.2 When staggered lugs convey fish at a constant speed and holders are not used, thefish will not slide along the lugs provided ufjahtabje<2[Lfiah_lugs. When staggeredlugs follow a pulsed motion, two holders are required to hold the fish. The holdersshould be slanted to provide a better hold with a lower holding force, and the sidesof the holders should be parallel to the sides of the table.5.2.3 Holding pans convey fish so that the undesirable motion of the head and tail isgreatly reduced from that seen when chain lugs are used for conveying. Slantedholders and parallel holders are equally satisfactory at restraining the fish, but theTime = 0.31 s Time = 0.41 s Time = 0.51 sChapter 5. Finite Element Simulation of Conveying 86holding force of the slanted holders is approximately 10—20% lower than that of theparallel holders.The following recommendations are made to improve the conveying system of theIron Butcher:1. Replace the chain lugs with holding pans. By using holding pans instead ofchain lugs to convey the fish, the unwanted motion of the head and tail is greatlyreduced, whether conveying is at continuous speed or intermittent (pulsed). Thisimproved positional accuracy will directly translate into less waste at the head cut.If constant speed conveying is desired, then the fish need be conveyed only through0.1 m (as compared to 0.8 m for chain lugs) before progressing to the next stageof processing, as the holding pans help settle the transient motion more quicklythan is possible with the lugs. This benefit means that a more compact machinecan be built, thus saving vaiuable work space and will have a lower cost since fewerresources go into producing the machine. Finally, even a simple design for the holding pan, as was used in the simulations, produces significantly improved conveyingresults. Stainless steel or aluminum would be a suitable material for the holdingpans for reasons of hygiene (easy to clean and disinfect) and manufacturability.2. Use contoured holders to hold the fish. Slanted holders are shown to besuperior to parallel holders, yet slanted holders provide improved restraint onlywhen the conveyor pauses in pulsed conveying. If the holders are contoured toapproximate the contour of the fish, then the fish would be better restrained whenthe conveyor starts to move as well as when the conveyor pauses. The motion of thehead would be smaller, so the waste at the head cut, as a result of this extraneousmotion of the head, would be reduced. The contour should conform with a rangeof fish, at least in an average sense, however.CDJoqoqCDui-0CD N CD oq c+ I-Ii CDq cn CD Oc, CDc, c-i- C. CD C÷Qq P’CD Q_i-i i—.CD CD00I-.CD I.p 00 00 JIIr p 00Chapter 5. Finite Element Simulation of Conveying 88zz40.0740.110.164085All dimensions in meters_________0.07 _1/0.010.60L -“H 0.0250.075Figure 5.17: A sketch of the holding pan used in the simulations.CD 9’ I-. 00 CD C-) 0 CD 0 CD 0 1 CD 0 CDCD C C C’)CD C 00 C’, CD C C’)CD p :• C’)CD I. 00 CoI-’.oq I-i CD 9’0p.toC 00 C’,02 CD C, 0 CD cc::—.oC2I-..CD a. Cs I. to CChapter 5. Finite Element Simulation of Conveying 91-— -0.25 0.5 0.75Time (s)0.25 0.5 0.75Time (s)oli er1.0 1.25Figure 5.20: Holding forces for the pulsed conveying of a large, soft fish using a holdingpan with (a) parallel holders (b) slanted holders.-vzz 0HadIoler-F’,I:zz::z1e d IIT ii I Ok er-2U—— —4:______-T ii I olc r100o 1.0 1.25 -lOOo(a) (b)Chapter 5. Finite Element Simulation of Conveying 92Time = 0.51 sFigure 5.21: Pulsed conveying of a large, soft fish using a rotated pan and slanted holders.Chapter 6Investigation of Indexing ErrorsOn the present Iron Butcher, the wastage at the head cut results primarily from theimproper indexing of the fish. The extreme cases of indexing error, underfeed and overfeed, are examined in this chapter. The possible mechanisms which cause these errorsare studied through simulation. Design improvements are then proposed and tested todetermine their effectiveness in preventing these errors from occuring. Recommendationsfor the redesign of the machine to reduce the frequency of underfeed and overfeed errorsare then made.6.1 Investigation of Underfeed ErrorUnderfeed error occurs when the indexer does not engage with the collar—bone and gillcover, but instead slips past both these structures and continues to move toward thecutting line without indexing the fish. Two explanations are available to describe whythe indexer does not properly engage with the gill cover and collar—bone.1. The gill cover is pushed into the gill cavity. During handling, the gill coveris sometimes pushed into the gill cavity. When this happens, the gill cover is belowthe surface of the fish, so it is impossible for the indexer to engage the gill cover.With frozen and thawed fish, the gill cover could be stuck in this manner.2. The gap between the gill cover and the side of the fish is not largeenough for the indexer to engage with the gill cover. The fish—gill cover93Chapter 6. Investigation of Indexing Errors 94gap is illustrated in Figure 6.1.If the gill cover is pushed into the gill cavity, the gill cover should be freed from thegill cavity before the fish is placed on the Iron Butcher. If this is not done, the indexingprocess will be more likely to fail, since the indexer cannot engage with the gill cover. Itis expected that a minimum gap size is required for the indexer to engage with the gillcover for indexing, but it is not known if this minimum gap size varies with the location ofthe gill cover relative to the collar—bone. Estimating the minimum gap size for successfulindexing, and how this gap size varies with the position of the gill cover, is determinedthrough computer simulation.Nonlinear, transient, dynamic analysis of the indexing process is performed to determine the minimum gill cover—fish gap size for successful indexing. The model for a large,firm fish is used, with the coefficients fish. xer /.Lftgh_holder and Ifish—lugs all set to0.1. An initial static load step is used to specify the simulation initial conditions, whichare zero velocity and acceleration for the fish, and the fish lying fiat on the table undergravity loading. The time at the end of this static load step is 0.01 s.Fish-Gill Cover GapCollar-BoneFishFigure 6.1: The fish—gill cover gap.Chapter 6. Investigation of Indexing Errors 95Table 6.1: Variation in the minimum gill cover—fish gap size for successful indexing usingthe current design of the Iron Butcher.Distance the gill cover is posterior Minimum initial gillof the collar—bone (m) cover—fish gap size (m)0.000 0.0020.0025 0.00090.005 0.0007The location of the end of the gill cover, which is engaged by the indexer, is not fixed.As can be expected of a natural resource, there is some variation in the location of thisfeature, but this feature will not normally be superior of the collar—bone and will be up to0.005 m posterior of the collar—bone. For positions within this range of possible locationsof the end of the gill cover, the gill cover—fish gap size is varied until the minimum gapsize for successful indexing is obtained. The results of the simulations are summarized inTable 6.1. The results show that the minimum gap size for successful indexing increasesas the location of the gill cover approaches the collar—bone. This trend can be explainedby the fact that while the fish muscle deforms under the weight of the indexer, the collarbone does not deform under this weight. Since the fish muscle immediately adjacent tothe collar—bone is attached to the collar—bone, this muscle would deform less than muscletissue further from the collar—bone when the same loading is applied to each. If the tissuebelow the indexer deforms more due to the weight of the indexer, then the initial gillcover—fish gap size (i.e., the gill cover—fish gap size before the fish muscle deforms underthe weight of the indexer) can be smaller for successful indexing to occur.From the above results, it is apparent that increasing the gill cover—fish gap size sothat it exceeds the minimum gap size for successful indexing, will significantly reduceor altogether eliminate underfeed error. In order to increase the gill cover—fish gap sizefor each fish that is processed on the Iron Butcher, a mechanism to push the fish headChapter 6. Investigation of Indexing Errors 96down, and as a result increase the gill cover—fish gap size, is proposed. The push—downmechanism in Figure 6.2 consists of a flat plate with a mass of 0.85 kg and measures0.06 m x 0.15 m x 0.01 m. This plate can drop onto the nose of the fish and move withII_Push-Down PlateFigure 6.2: Components of the push—down mechanism: (a) the finite element model and(b) simulation result for indexing.the conveyor, or it can remain stationary and be hinged to allow the fish head to passbelow it as the fish is conveyed toward the cutter. In addition to this plate, a 0.01 mhigh bar is added to the surface of the table and runs the length of the table. The fishhead is lifted by the bar to increase the clearance between the fish head and the tableso that the plate can push the nose of the fish down and increase the gill cover—fish gapsize.Nonlinear, transient, dynamic analyses are performed to determine the effectiveness ofthe push—down mechanism in preventing underfeed error. The simulations use the modelof a large, firm fish, and the coefficients of friction 1tfishtabte, ILfiah1ugs, ILfish—indexer,LifterI I(a)Push-Down PlateChapter 6. Investigation of Indexing Errors 97Table 6.2: Variation in the minimum gill cover—fish gap size for successful indexing whenthe nose push—down device is added to the Iron Butcher.Distance the gill cover is posterior Minimum initial gillof the collar—bone (m) cover—fish gap size (m)0.000 0.0010.0025 0.00010.005 0.0003Pfish—holder, Pfzah—pushdownplote and /.Lf:shlifter are all assigned a value of 0.1. For thesimulations, the push—down plate drops onto the fish head and moves with the fish as itis conveyed. The simulation results for the minimum initial gill cover—fish gap size forsuccessful indexing are summarized in Table 6.2. These results indicate that the pushdown mechanism successfully increases the size of the gap from its initial value to a largervalue that is required for successful indexing to occur, without adversely affecting theconveying or indexing processes.6.2 Investigation of Overfeed ErrorFour possible causes for overfeed error have been identified. These are:1. Friction between the fish and indexer, and/or between the fish and the table surface.2. Structural deformities on the fish. Examples of these are cuts or gouges in thefish flesh, permanent crease lines caused by nets during fishing or caused duringtransport or storage, or the absence of scales in a region of the fish.3. The design and/or configuration of the machine components.4. Softness of the fish muscle.Chapter 6. Investigation of Indexing Errors 98There is some inter—relationship between these suspected causes of overfeed error. Forexample, a cut may be modeled as a structural deformity, but the absence of fish scalesfrom a region of the fish, although classified as a structural deformity, may be bettermodeled as a change in the coefficient of friction between the fish and indexer in thatregion. The following simulations of indexing terminate prior to the indexer engagingwith the gill cover. As a result, the motion of the fish can be attributed to mechanismswhich cause overfeed error.6.2.1 The Role of Friction in Overfeed ErrorIn order to determine if friction plays an active role in the occurrence of overfeed error,a series of simulations are performed. The purpose of the first simulation is to determinethe motion of the fish when the fish comes into contact and conforms to the table surfaceunder gravity loading. The second set of simulations determine the motion of the fishas the fish—indexer and fish—table coefficients of friction are varied. The third set ofsimulations determine if the holder has any affect on overfeed error. The last set ofsimulations determine if repositioning the holder has a positive effect in reducing thelikelihood of overfeed error occuring. All simulations have an initial static load step topermit the fish to come into contact and conform to the table surface, after which pointdynamic load steps simulate various combinations of holding and motion of the indexersliding over the fish.Motion of the Fish Under Gravity LoadingIn all the finite element simulations, the fish is initially located slightly above the tableof the Iron Butcher. A static load step is first performed to permit the fish to contactand conform to the table surface under gravity loading. After this static load step, thefish is at rest on the table surface and dynamic load steps are then performed to modelChapter 6. Investigation of Indexing Errors 99various stages of processing. The coefficient of friction at the fish—table interface is 0.1,and the model of a large, soft fish is used. Figure 6.3 shows the a,—direction motion of thefish as the fish comes into contact with and conforms to the table surface. It should benoted that in the simulations, indexing occurs in the negative z—direction of the globalcoordinate system. These results indicate that the fish moves through approximately0.005 m in the negative a,—direction when it conforms to the table surface. Any motionin the the global x—direction in excess of -0.005 m in the following simulations can beconsidered to be indexing error.-0.02EU,‘ -0.060z-0.08I I I I I0 0.16 0.32 0.48 0.64 0.80Time (s)Figure 6.3: Overfeed motion (in the i—direction) of the fish when the fish conforms tothe table surface under gravity loading (I’fish_table = 0.1).Overfeed Error as a Function of the Coefficients of FrictionA series of simulations are performed to determine how the z—direction motion of the fishchanges as the coefficient of friction between the fish and indexer changes. Simulationsare performed for a large, firm fish and a large, soft fish. The coefficient of friction forChapter 6. Investigation of Indexing Errors 100Figure 6.4: Overfeed motion (—direction) of the fish as a function of the coefficient offriction between the fish and indexer (I’fish—table = 0.1). Simulation results for indexinglarge fish with (a) firm muscle (b) soft muscle.In addition to the effect the fish—indexer coefficient of friction has on overfeed, intuitively, reducing the fish—table coefficient of friction should increase the possibility ofoverfeed. A series of simulations are carried out to determine the effect of the fish—tablecoefficient of friction on overfeed error. In these simulations, the coefficient of frictionifish—tab1e is set to 0.05, instead of 0.10 as in the above simulations. Figure 6.5 shows theresults of these simulations.A comparison of the curves in Figure 6.4 with those in Figure 6.5 immediately providesfour general results. First, friction between the fish and indexer is a possible mechanismcausing overfeed error. Second, overfeed error increases with increasing fish_jadexer.0t = 0.1= 0.3= 0.5= 0.7the other surfaces are fish_table = ILfsh_ho1der = 0.1. The results of the simulationsare summarized in Figure 6.4. These curves show the a—direction motion of the fish fordifferent values of ifsah—Jexer. The motion of the indexer is also shown so that theresults can be interpreted relative to that motion.0!-O.02-0.04-0.06-0.08= 0.9-0.02• -0.04-0.06-0.08= 0.1i.t = 0.5= 0.7Fish—— —— Indexer0Fish— ——— Indexerii = 0.9Time (s)(a)I I0 0.16 0.32 0.48Time (s)(b)0.64 0.80Chapter 6. Investigation of Indexing Errors 1010 1.6 3.2 4.8 6.4 8.0 0 1.6 3.2 4.8 6.4 8.0Time (s) Time (s)Figure 6.5: Overfeed motion (z—direction) of the fish as a function of the coefficient offriction between the fish and indexer (1Lfh_tab1e = 0.05). Simulation results for indexinglarge fish with (a) firm muscle (b) soft muscle.Third, overfeed is more likely to occur as IUf:8htab1e decreases. Fourth, overfeed error ismore likely to occur, and is larger, for soft fish than it is for firm fish.Closer examination of the curves in Figure 6.4(b) provides further information asto the behaviour of the interactions as the indexer slides across the fish. The curves for0.1 ILfishtndexer 0.7 all have an instantaneous slope that is smaller in magnitude thanthe slope of the line denoting the motion of the indexer. For these cases, even though thefish is being pushed laterally, there is slip between the indexer and fish, with the indexermoving at a higher rate than the fish. For the case of Pfish—zexer = 0.9, the same holdstrue during the period of 0.0 s to 0.46 s. However, after 0.46 s, the instantaneous slopeof the curve is greater in magnitude than the slope of the line denoting the motion of theindexer. After 0.46 s, the fish is moving faster than the indexer, so another mechanismmust be working to introduce the overfeed error.0= 0.1= 0.5= 0.7ii = 0.9-0.06z-0.08=0.10I-0.02Ea)tO.04-0.06z-0.08Fish \\\IndexerFishii = 0.5t = 0.9(a) (b)Chapter 6. Investigation of Indexing Errors 102Effect of the HolderSince it has been determined that another mechanism, in addition to friction, is responsible for overfeed error, a simulation is run that does not include a holder. These simulationresults will indicate if the holder contributes to overfeed error. A large, soft fish is modeled with /I!jsh..tabje = 0.1 and /Afjsh..jndextj. 0.3. Figure 6.6 shows that the fish movesapproximately 0.002 m in the negative x—direction when the holder is not used duringthe indexing process. From Figure 6.4(b), the movement of the fish is approximately0.015 m when the holder is used during indexing. The holder contributes to the overfeederror because as it presses down on the fish there is also a component of the force thatcauses the fish to slide in the negative x—direction. From this result, it can be concludedthat the interaction of the fish with the holder is a mechanism in overfeed error.04- Ary) \ \-0.04\\\-0.06oZ Fish-0.08 —— — —- IndexerI I I0 0.16 0.32 0.48 0.64 0.80Time (s)Figure 6.6: Overfeed motion (z—direction) of a large, soft fish when the holder is not used(t1ish—table = 0.1).Chapter 6. Investigation of Indexing Errors 103Redesign of the Indexing SystemTo minimize the component of the holding force which causes overfeed, the holder ismoved so that it is located over the thickest part of the fish. A series of simulations arethen carried out for a range of values of the coefficient of friction Pfish indexer to see if theoverfeed error is reduced. The results are given in Figure 6.7. By comparing these curveswith those of Figure 6.4(b), it is seen that for a large, soft fish, the contribution thatthe holder makes to overfeed error is significantly reduced when the holder is positionedover the thickest part of the fish. Furthermore, when the holder is positioned over thethickest part of the fish, overfeed will likely occur only if pjah-..indexer > 0.5.= 0.1—O.04• -0.06oZ Fish————-Indexer-0.08I I I I I0 0.16 0.32 0.48 0.64 0.80Time (s)Figure 6.7: Overfeed motion (x—direction) of a large, soft fish as a function of the coefficient of friction between the fish and indexer, when the position of the holder is corrected(PI:sh—taue — 0.1).Chapter 6. Investigation of Indexing Errors 1046.2.2 The Role of Structural DeformitiesIf the structural deformity is best modeled by a change in fish—ixer in a region of thefish skin, then results of section 6.2.1 are applicable. In such a case the action proposedin section 6.2.1 should suffice to eliminate most instances of overfeed error of this type.If on the other hand, the structural deformity is best modeled by a physical structurewhich protrudes above the surface of the fish, the results of section 6.1 would apply. Toreduce the incidence of overfeed of this type, the indexing process should be abandonedin favor of a system which senses the location of the gill cover and positions the cutterfor the head cut, or the fish should not be placed on the Iron Butcher for processing, butshould be processed manually.6.3 Summary of Results and Design RecommendationsThe results of this chapter are summarized below.• The minimum initial gap size between the gill cover and the fish for successfulindexing to occur is established.• A modification to the Iron Butcher (i.e. the push—down mechanism) is proposed toincrease the size of the gap between the gill cover and the fish so the indexer willmore easily engage with the gill cover and collar—bone.• Operation of the Iron Butcher with the push—down modification is simulated todetermine if the push—down mechanism reduces the likelihood of underfeed erroroccuring.• Mechanisms responsible for overfeed error are identified.Chapter 6. Investigation of Indexing Errors 105• Simulations are used to determine the role each mechanism plays in overfeed error.Simulation results are also used to develop guidelines for the design of the indexingsystem to reduce the occurrence of overfeed error.The following recommendations are made to improve the indexing system of the IronButcher.Recommendations to Correct Underfeed Error1. Modify the Iron Butcher to include a push—down mechanism for the indexing process. This mechanism will increase the size of the gap between the gill cover andthe fish, and enhance the protrusion of the collar—bone beyond the fish side, sothat the indexer can engage the gill cover and collar—bone and index the fish forthe head cut.Recommendations to Correct Overfeed ErrorFrom the above analysis, two modifications to the machine will greatly reduce the incidence of overfeed error from friction between the fish and indexer and from the effects ofholding.1. Spraying water on the fish will act as a lubricant as the indexer slides over the fish.Since the coefficient of friction is reduced, the chance of overfeed occuring fromfriction alone is reduced. At the same time, grooves should be cut into the tablesurface to carry the excess water away, as overfeed is sensitive to Pltah-..tabie.2. Modify the Iron Butcher to permit adjustment of the position of the holder. Foroptimum results the machine should adjust the position of the holder for each fish.Since this would require sensing, actuation and control, it would be expensive toChapter 6. Investigation of Indexing Errors 106implement. However, a manual adjustment which could be made daily or hourlybased on the size of the fish running through the machine would reduce overfeedfrom the holder holding the fish near the tail where the fish is thin. The carouselin which the holders move could be mounted in a track so that the location of theholders could be easily adjusted and locked into place. A conceptual sketch of themovable carousel is shown in Figure 6.8.Chapter 6. Investigation of Indexing Errors 107____Positioning TracksConveyorCutterCarousel forthe HoldersConveyor MotionPositioning TracksLiFigure 6.8: A movable carousel system to permit the manual adjustment of the holderposition.Chapter 7Implementation of Design ImprovementsTo complement the analytical work in this dissertation and other research carried outin the Industrial Automation Laboratory, Department of Mechanical Engineering atU. B. C., several design improvements are implemented on two seperate butchering machines. The first machine is a laboratory prototype used for research, while the secondmachine is a prototype built and tested for industrial use. Both of these prototypes haveundergone significant change from the current Iron Butcher, and there are some simularities in the design of these prototypes. This chapter describes the configuration andoperating mode of these machines. The effects the design changes have on the machineprocessing of fish are discussed.7.1 Laboratory PrototypeThe components of the laboratory prototype are shown in Figure 7.1. A description ofthe prototype follows.Feeding System A fish is placed in a holding pan so that the salmon is oriented withits head pointed towards the cutter and its belly facing away from the conveyormotion. In addition, the collar—bone must lie between two markers.Conveying System The profile of the holding pan is an arc. The holding pan consistsof two components, a small pan to hold and support the head of the fish duringconveying, and a large pan to hold and convey the body of the fish. The two108Chapter 7. Implementation of Design Improvements 109Structured HolderLightingHolding PanCutterFigure 7.1: The laboratory prototype head—butchering machine.components of the holding pan are aligned and rotated through 30°. The conveyormoves at a constant speed of 0.5 rn/s.Holding System Holding is accomplished by a group of stationary, active holders. Theholders are cylindrical in shape. The cylinders rotate so that the point on thecylinder which is in contact with the fish has the same linear speed as the conveyor.The holders are spring loaded to accomodate different sizes of fish automatically.Holding starts after the feeding stage and continues to the onset of cutting.Indexing System The mechanical indexing system of the existing Iron Butcher is notused on this protype. Instead of indexing the fish, the cutter is moved to therequired location to perform an accurate head cut. Careful placement of the fishin the holding pans during feeding ensures that only a small adjustment in thecutter position is required to obtain an accurate cut. A CCD camera with anNose HolderChapter 7. Implementation of Design Improvements 110electronic shutter captures an image of the head region of the fish as the fish isconveyed toward the cutter. The location on the fish where the cut is to be made isdetermined by image processing, and a control signal is sent to a servo—controllerto position the cutter. An ultrasound displacement sensor records the thicknessof the fish so that the cutter can be adjusted to compensate for the variation inthickness between fish.Cutting System A double—bladed rotary cutter replaces the old chopping blade of theexisting Iron Butcher. The new cutter creates a clean cut that is able to extractmeat that was previously lost with the old cutter.Although not perfect, the laboratory prototype is an improvement over the existingIron Butcher, due primarily to the improvements to the conveying system and the newapproach regarding indexing. The combination of holding pans and active holders providea firm grip of the fish during the conveying process. This is an important factor indetermining the accuracy of the cut, since any lateral movement of the fish with respectto the pan after the CCD camera images the head region of the fish will adversely affectthe cut accuracy. The detection of the collar—bone and sensing of the fish thickness usingultrasound are both accurate and reliable. The accuracy of the cut on entry is improvedfrom approximately 6 mm posterior of the collar—bone on the existing Iron Butcherto approximately 1 mm posterior of the collar—bone on this prototype. Although theaccuracy of the cut on entry is higher, the accuracy of the rest of the cut diminishesto that of the existing Iron Butcher because of the lack of holding during the cuttingprocess. As the fish is cut, the cutting forces tend to either pull the fish toward and intothe cutter or push the fish away and out of the cutter, depending on the direction thecutter blades are rotating, causing the cut to deviate from the calculated optimal cut.Chapter 7. Implementation of Design Improvements 1117.2 Industrial PrototypeFigure 7.2 shows the components of the industrial prototype. A description of the prototype follows.Pneumatic RamCCD CameraHolderLugs ___ACutting BladeFigure 7.2: The industrial prototype head—butchering machine.Feeding System The fish is fed onto the table surface with the same orientation usedin the laboratory prototype. The fish must also be aligned so that the collar—bonelies within positioning markers.Conveying System Conveying is accomplished by three lugs, as in the existing IronButcher. However, unlike the constant speed conveying of the existing Iron Butcher,the conveying motion of this prototype is intermittent.Holding System The holders on the prototype maintain a fixed position over the table.Three holders are located at each position corresponding to the location of the fishChapter 7. Implementation of Design Improvements 112when the conveyor pauses. The holders approximate the contour of the fish andare spring loaded to accomodate various sizes of fish. Pneumatics are used to liftthe holders while the conveyor is moving.Indexing System The indexing system employs the same approach as the laboratoryprototype. That is, the head region of the fish is imaged using a CCD camera andimage processing determines the location on the fish where the cut should be made,and the cutter is moved to this position.Cutting System A guillotine—style cutter blade is mounted on a single—axis table toposition the cutter at the desired cutting location. A pneumatic piston controlsthe chopping action.Although the conveying process works well to move the fish along the machine, thehead and tail of the fish are not properly constrained, allowing them to oscillate back andforth as the conveyor cycles through its stop—start motion. Because of this extraneousmotion of the fish head, the cutter cannot be positioned exactly at the location determinedby the imaging system, but must instead be positioned slightly posterior of this position.This results in positional accuracy of cut of between 1 mm and 6 mm, which is animprovement on the existing Iron Butcher, but is not as good as the accuracy of theentry cut on the laboratory prototype. The guillotine-style cutter produces a clean cut,unlike the chopping cutter of the existing Iron Butcher. After the head is removed, onlythe two tail lugs touch the fish, causing the fish to rotate and jam in the conveyor system.Employing pans would eliminate the extraneous motion of the head and tail of the fishas it is conveyed so that the accuracy of cut could be improved to 1 mm. Furthermore,the pans would prevent the fish from jamming after the head is removed.Chapter 8Conclusions and RecommendationsIn this final chapter, the main accomplishments of this research are summarized. Recommendations for improving the design of the Iron Butcher are listed. Finally, recommendations for future work on this topic are presented.8.1 An Overview of the ResearchThe main accomplishments of this research are summarized below:1. Finite element models of the fish, the Iron Butcher and the fish—machine interactions have been developed. These models permit the static and dynamic analysis ofmachine processing of salmon on the Iron Butcher, and serve as a tool for analyzingthe operation of the machine in various configurations.2. An investigation of the conveying process on the Iron Butcher concentrates ondesign requirements when new configurations and operating requirements are specified.• The operation of the Iron Butcher in its present configuration and under thepresent operating specifications is examined.• Pulsed conveying with straight chain lugs, conveying with staggered chain lugsat a constant and pulsed conveyor motion, and the use of holding pans as analternative to lugs are all investigated.113Chapter 8. Conclusions and Recommendations 1143. The mechanisms which cause overfeed and underfeed indexing error are investigated. A systematic approach is employed to determine the effect of each mechanism when more than one mechanism is active in producing the indexing error.4. Design guidelines have been developed from the simulation results and serve to aidthe designer in improving the performance of the Iron Butcher.5. This research demonstrates the usefulness of finite element analysis as a design toolfor materials handling problems, particularly when the objects being handled areflexible and the dynamics of the process cannot be ignored.6. The Iron Butcher has been modified with design changes to the conveying, indexingand cutting systems and has been demonstrated to provide better results.8.2 Recommendations for Design ImprovementThe following recommendations are made to improve the conveying and indexing processes of the Iron Butcher.1. Use contoured holders to hold the fish. The shape of the holder should roughlyconform with a range of fish. At least two holders should be used per fish.2. Replace the chain jugs with holding pans. The holding force is greatly reducedwhen the chain lugs are replaced by holding pans. Furthermore, holding pansgreatly reduce the unwanted motion of the head region when the fish is conveyed.3. Add a nose push—down mechanism. The nose push—down mechanism increasesthe fish—gill cover gap size, thereby decreasing the likelihood of underfeed error.4. Spray water on the fish. Water acts as a lubricant, reducing the fish—indexercoefficient of friction, thereby decreasing the chance of overfeed error.Chapter 8. Conclusions and Recommendations 1155. Permit adjustment of the holder position. Proper position of the holder isrequired to reduce overfeed error.8.3 Recommendations for Further WorkThis research has been limited to the investigation of the conveying and indexing processes on the Iron Butcher. Although these two processes are responsible for the extremecases of waste at the head cut, even if the fish is properly conveyed and indexed, thetearing action of the cutter creates waste with each cut. Therefore, a logical progressionof this work would examine the cutting process. Although finite element analysis currently has the capability of modeling processes which have as a feature the removal ofmaterial, the criteria used to determine when and how the material should be removed isthe critical part of the finite element model. If the generated criteria do not accuratelyreflect the actual physical process of material removal, then the simulation results willprovide information about a quite different process, and the usefulness of these resultswill be severely limited.The task of modeling the cutting process becomes a problem of material sciences. Thefirst step is to determine the material properties of the muscle tissue. This descriptionwould account for the anisotropic and nonlinear properties of the muscle tissue. Next,an understanding of how muscle fibre fails is required. Controlled experiments mayreveal the failure modes of muscle fibre for different loading conditions, but ultimatelyan understanding of the failure modes during different cutting processes (e.g., chopping orslicing) is required. If an understanding of these processes are attained, would they stillbe valid for a grouping of muscle fibre? The last step before modeling the process is totranslate the understanding of the cutting process into a set of constraint equations whichwill control the removal of material and the distribution of cutting forces in the finiteChapter 8. Conclusions and Recommendations 116element model. The model of the fish and machine must then be modified to incorporatethese changes. The process of acquiring an understanding of the cutting process shouldenable the proper selection of a cutter for a specific cutting task, while simulation resultsshould provide information about how to hold the fish during the cutting process.Bibliography[1] Browning, R. J., FISHERIES OF THE NORTH PACIFIC, History, Species, Gear& Processes (Rev. ed.), Alaska Northwest Publishing Co., Anchorage, Alaska, 1974.[2] de Silva, C. W., “Research Laboratory for Fish Processing Automation”, Proceedingsof the 3rd International Symposium on Robotics and Manufacturing, Vol. 3, A.S.M.E.Press, N.Y., pp. 935—940, 1990.[3] Mason, M. T., “The Mechanics of Manipulation”, 1985 IEEE International Conference on Robotics and Automation, pp. 544—548.[4] Mason, M. T., “Mechanics of Pushing”, nd International Symposium on RoboticsResearch, pp. 421—428, August 1984.[5] Brost, R. C., “Automatic Grasp Planning in the Presence of Uncertainty”, 1986IEEE International Conference on Robotics and Automation, pp. 1575—1581.[6] Peshkin, M. A. and Sanderson, A. C., “Manipulation of a Sliding Object”,Proceedings- 1986 IEEE International Conference on Robotics and Automation,pp. 233—239.[7] Peshkin, M. A. and Sanderson, A. C., “Planning Robotic Manipulation Strategiesfor Sliding Objects”, Proceedings- 1987 IEEE International Conference am Roboticsand Automation, pp. 696—701.[8] Peshkin, M. A. and Sanderson, A. C., “Planning Robotic Manipulation Strategiesfor Workpieces that Slide”, IEEE Journal of Robotics and Automation, Vol. 4, No. 5,pp. 524—531, October 1988.[9] Gilmore, B. J. and Streit, D. A., “A Rule-Based Algorithm to Predict the DynamicBehaviour of Mechanical Part Orienting Operations”, 1988 IEEE International Conference on Robotics and Automation, pp. 1254—1259.[10] Wang, Y. and Mason, M. T., “Modeling Impact Dynamics for Robotic Operations”,1987 IEEE International Conference on Robotics and Automation, pp. 678—685.[11] Routh, E. J., Dynamics of a System of Rigid Bodies (Seventh Edition), Dover Publications, New York, 1960.117Bibliography 118[12] Cai, C. S. and Roth, B., “On the Planar Motion of Rigid Bodies with Point Contact”,Mechanüm and Machine Theory, Vol. 21 No. 6, pp. 453—466, 1986.[13] de Silva, C. W., Control Sensors and Actuators, Prentice-Hall, Inc., EnglewoodCliffs, N. J., 1989.[14] Cai, C. S. and Roth, B., “On the Spatial Motion of a Rigid Body with Point Contact” ,IEEE 1987 International Conference on Robotics and Automation, pp. 686—695.[15] Montana, D. J., “The Kinematics of Contact and Grasp”, The International Journalof Robotics Research, Vol. 7 No. 3, pp. 17—32, June 1988.[16] Montana, D. J., “The Kinematics of Contact with Compliance”, 1989 IEEE International Conference on Robotics and Automation, Vol. 2, pp. 770—774.[17] Saliba, M. and de Silva, C. W., “An Innovative Robotic Gripper for Grasping andHandling”, Proc. IEEE Conference on Industrial Electronics, Control, and Instrumentation, Kobe, Japan, pp. 1590—1595, October 1991.[18] de Silva, C. W. and Saliba, M., “Instrumentation Issues in the Handling of Fish forAutomated Processing”, Proc. IEEE Conf. on Industrial Electronics, Control, andInstrnmentation, San Diego, CA, November 1992 (In Press).[19] Petrell, R. J., Rodriguez, A. C. and Nordin, D. M., “Anisotropic Elastic Propertiesof Farmed-Salmon Muscle”, Proceedings of the 1990 International Winter Meeting ofthe American Society of Agricultural Engineers, Paper No. 906572, December 1990.[20] Johnson, E. A., Segars, J. G., Kapsalis, M. D. and Peleg, M., “Evaluation of theCompressive Deformability Modulus of Fresh and Cooked Fish Flesh”, Journal ofFood Science, Vol. 45, pp. 1318—1326, 1980.[21] Borderias, A. J., Lamua, M. and Tejada, M., “Texture analysis of fish fillets andminced fish by both sensory and instrumental methods”, Journal of Food Technology,Vol. 18, pp. 85—95, 1983.[22] Chow, W. W. and Odell, E. I., “Deformations and Stresses in Soft Tissues of aSitting Person”, Journal of Biomechanical Engineering, Vol. 100, pp. 79—87, 1978.[23] Yamada, H., Strength of Biological Materials, The Williams & Wilkins Company,Baltimore, 1970.[24] Huebner, K. H. and Thornton, E. A., THE FINITE ELEMENT METHOD FORENGINEERS (Second Edition), John Wiley & Sons, Inc., New York, 1982.Bibliography 119[25] Rizzo, A. R., “Estimating Errors in FE Analyses”, MECHANICAL ENGINEERING/CIME, May 1991, PP. 61—63.[26] Dunajski, E., “Texture of Fish Muscle”, Journal of Texture Studies, Vol. 10,pp. 301—318, 1979.[27] Segars, R. A., Nordstrom, H. A. and Kapsalis, J. G., “Textural Characterstics ofBeef Muscles” Journal of Textural Studies, Vol. 5, No. 3, pp. 283—297, Oct. 1974.[28] Bouton, P. E. and Harris, P. V., “The effects of some post-slaughter treatments onthe mechanical properties of bovine and ovine muscle”, Journal of Food Science,Vol. 37, pp. 539—543, 1972.[29] Fung, Y. C., Biomechanics, Mechanical Properties of Living Tissues, Springer-Verlag, New York, N. Y., 1981.[30] Evans, F. 0., Mechanical Properties of Bone, Charles C. Thomas, Springfield, Illinois, 1973.[31] de Silva, C. W. and Riahi, N., “System Integration and Image Processing Techniquesin a Fish Processing Workcell”, Advances in Instrumentation, DSC - Vol. 30, A. S.M. E., N. Y., pp. 73—78, December 1991.[32] Gamage, L. D. K. B. and de Silva, C. W., “Use of Image Processing for the Measurement of Orientation with Application to Automated Fish Processing”, IECQN’90,The 16th Annual Conference of IEEE Industrial Electronics Society, pp. 482—487,Nov. 1990.

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0080844/manifest

Comment

Related Items