i QUANTITATIVE ANALYSIS OF MASTICATORY PERFORMANCE IN VERTEBRATES By SRIKANTH KANNAN August, 2008 A thesis submitted to the Faculty of the Graduate School of the State University of New York at Buffalo in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Mechanical and Aerospace Engineering State University of New York at Buffalo Buffalo, New York 14260
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
i
QUANTITATIVE ANALYSIS OF MASTICATORY PERFORMANCE IN VERTEBRATES
By
SRIKANTH KANNAN
August, 2008
A thesis submitted to the Faculty of the Graduate School of the State University of New York at Buffalo in partial fulfillment of the requirements
for the degree of
MASTER OF SCIENCE
Department of Mechanical and Aerospace Engineering State University of New York at Buffalo
Buffalo, New York 14260
ii
Acknowledgement First, I would like to express my sincerest gratitude to my advisor, Dr. Venkat Krovi, for
giving me an opportunity to work under him as a Research Assistant. He was not only my
mentor, but as a friend, he always provided me with valuable suggestions whenever I
needed them most. I would like to express my gratitude to the committee members, Dr.
Frank Mendel and Dr. Andres Soom for serving on my thesis committee and reading
through my thesis and providing me with valuable suggestions.
I would like to thank Dan Murray, Bill McDougall, Brian Wolf, and David Eley for
giving me an opportunity to work with Fisher-Price, Inc and for giving me access to the
3D laser scanner and SLA machine. I would like to thank my lab members Rajan, Chin
Pei Tang, Leng-Feng Lee, Anand Naik, Kun Yu, Hao Su, Qiushi Fu, Yao Wang and
Patrick Miller for lending me an helping hand whenever I needed it the most. I convey
my special thanks to Madu for being my project partner right from the first semester till
the end of my thesis. I would also like to thank my roommates Arun, Sriniwas, Vijay,
Parthiban, Govind, Amol for providing good support and entertainment at home.
And especially I would to thank my parents, Mr. K.R. Kannan and Mrs. K.
Vijayalakshmi and my other family members Priya, Kumar, Adeep, Ayush and
Rajeshwari for being affectionate and encouraging me right throughout my education.
THANKYOU ALL ONCE AGAIN
iii
Abstract To quantitatively measure mechanical performance signals such as forces and motions
and mechanical breakdown of food during mastication, it is imperative to accurately
reproduce the mastication motion. Reproduction of the mastication motion of a vertebrate
with a robotic device will allow us to estimate muscle and bite forces required for
different animals while chewing/biting different regimen and relate them to masticatory
muscle recruitment patterns and would be used to quantitatively evaluate the dynamic
breakdown of foods during chewing, which is vital information required in the
development of new pet foods. We also examine the use of a robotic solution where a
generic parallel manipulator with six degrees of freedom (Stewart platform) was modeled
and simulated using virtual prototyping tools to reproduce the 3D mandible trajectory. To
this end, a high fidelity (speed/resolution) motion capture system was used for capture the
3D mastication motion of different vertebrates. 3D laser scanning technology and image
processing techniques were used to obtain CAD model of a skull and mandible of a
bulldog which was then rapidly prototyped and casted to create a dentition. Architectural
parameters of muscle for a human jaw were obtained from Koolstra et al. and for a
bulldog jaw by conducting dissection of masticatory muscles. A musculoskeletal model
of the vertebrate jaw was created in AnyBody Modeling System to measure the forces
acting in the masticatory muscles and temporomandibular joints. We formulate and verify
the forward dynamics of the Stewart platform using three methods: 1. S-Function in
Simulink 2. DynaFlexPro model 3. Visual Nastran Plant. A combination of Newton-Euler
and Lagrangian method was used to formulate the inverse dynamics of a 6 DOF parallel
manipulator. Feedback linearization was implemented in Matlab/Simulink, using the
inverse dynamics and forward dynamics block, to control the motion of the moving
platform. Actuator forces were determined by implementing vertebrate mastication
trajectory as inverse dynamics using Matlab/Simulink and Visual Nastran. Results from
the inverse dynamic simulations and motion control of the Stewart platform show that the
Stewart platform enables the mastication motion to be reproduced.
iv
TABLE OF CONTENTS Acknowledgement ......................................................................................... ii Abstract......................................................................................................... iii LIST OF FIGURES .................................................................................... vii LIST OF TABLES ...................................................................................... xii 1. Introduction............................................................................................ 1
1.1 Motivation................................................................................................................. 1 1.2 Virtual Prototyping/ Simulation Based Design......................................................... 4 1.3 Musculoskeletal System Analysis............................................................................. 7 1.4 Research Tasks........................................................................................................ 10 1.5 Thesis Organization ................................................................................................ 11
2. Literature Survey................................................................................. 12
2.1 Masticatory Biomechanics...................................................................................... 12 2.1.1 Jaw Muscles and Movements .......................................................................... 12 2.1.2 Redundancy...................................................................................................... 13 2.1.3 Dynamics of Masticatory System .................................................................... 13 2.1.4. Influence of Muscles and Hill Muscle Model................................................. 14 2.1.5 Active and Passive Elements ........................................................................... 15
3.3.1 Jacobian Matrix Based On Vector Loop Closure Equation............................. 37 3.3.2 Screw Theory Based Jacobian Analysis .......................................................... 39
3.4 Jacobian-Based Performance Measures (JBPM).................................................... 42 3.4.1 Singular Value Decomposition (SVD) and Manipulability Ellipsoid ............. 42 3.4.2 Yoshikawa’s Measure of Manipulability......................................................... 44 3.4.3 Condition Number ........................................................................................... 44 3.4.4 Isotropy Index .................................................................................................. 45
3.7 Musculoskeletal System Analysis........................................................................... 50 4. Technological tools............................................................................... 52
4.1 Laser Scanning Technology.................................................................................... 52 4.1.1 Why 3D Scanning? .......................................................................................... 52 4.1.2 Different Types of 3D Scanning Technology [50]: ......................................... 53 4.1.3 Commercial Scanners ...................................................................................... 54 4.1.4 Generation of CAD Model of Vertebrate Skull and Mandible........................ 57
4.5 Musculoskeletal Model of Vertebrate Jaw ............................................................. 73 4.6 CAD Model of Stewart platform Type 6 DOF Parallel Manipulator ..................... 82 4.7 Forward Dynamics Model in DynaFlexPro............................................................ 84
5.1 Inverse Dynamic Analysis of Human Jaw Model in Anybody Modeling System. 93 5.1.1 Case Study I: One Temporalis Muscle and Muscle Model I and No External Force ......................................................................................................................... 94 5.1.2 Case Study II: Three Temporalis Muscles and Muscle Model I and No External Force........................................................................................................... 97 5.1.3. Case Study III: One Temporalis Muscle, Muscle Model 3E, No External Force ......................................................................................................................... 99 5.1.4 Case Study IV: Three Temporalis Muscles, Muscle Model 3E, No External Force ....................................................................................................................... 101
5.1.5 Case Study V: Three Temporalis Muscles, Muscle Model 3E, With External Force ........................................................................................................................... 104 5.2 Inverse Dynamic Analysis of Bulldog Jaw model in AnyBody........................... 107 5.3 Inverse Dynamic Analysis of Sabertooth Jaw Model in Anybody....................... 111 5.4 Manipulability Measures for Workspace Analysis............................................... 113
5.4.1 Case Study I: Varying the Radius of the Moving platform ........................... 114 5.4.2 Case Study II: Varying Radius of the Workspace ......................................... 117 5.4.3 Case Study III: Varying the Vertical Distance between Two platforms ....... 119
5.5 Dynamic Simulation of Stewart platform............................................................. 120 5.5.1 Simulation Using S-Function:........................................................................ 123 5.5.2 Simulation Using DynaFlexPro Model: ........................................................ 126
5.6 Simulation of the CAD Model of Stewart platform: ............................................ 129 5.6.1 Simulation Using DynaFlexPro Model.......................................................... 130 5.6.2 Simulation Using S-Function Block .............................................................. 131 5.6.3 Simulation Using Visual Nastran Plant ......................................................... 132
vi
5.7 Simulation of Jaw Motion Using Visual Nastran ................................................. 135 5.7.1 Dynamic Simulation of First Human Subject Jaw motion ............................ 136 5.7.2 Dynamic Simulation of Second Human Subject Jaw motion........................ 139 5.7.3 Dynamic Simulation of Bulldog Subject Jaw motion.................................... 141 5.7.4 Dynamic Simulation of Sabertooth Cat Jaw motion...................................... 143
6. Conclusion and Future Work ........................................................... 145
LIST OF FIGURES Figure 1-1: Development of Simulation Tools in Engineering .......................................... 2 Figure 1-2: Measure-Estimate-Test cycle........................................................................... 3 Figure 1-3: Vertebrate Mastication Motion Simulator Framework.................................... 3 Figure 1-4: Virtual Prototyping [4]..................................................................................... 5 Figure 1-5: Virtual Prototype of Piston (Left from Visual Nastran) and sabertooth Tiger (Right) [1] ........................................................................................................................... 5 Figure 1-6: Increase in Complexity of Musculoskeletal Modeling [1] [2] ......................... 8 Figure 1-7: Musculoskeletal system modeled as Articulated Multi-Body System with Redundancy [1]................................................................................................................... 9 Figure 1-8: Conventional Design Approach ....................................................................... 9 Figure 1-9: Virtual Prototyping Approach.......................................................................... 9 Figure 2-1: Human Masticatory Muscles [18].................................................................. 14 Figure 2-2: Dog Masticatory Muscles .............................................................................. 14 Figure 2-3: Elements of Hill Muscle Model [2] ............................................................... 15 Figure 2-4: Force Length profile of CE, SE and PE elements [2] .................................... 15 Figure 2-5: Force Length and Force-Velocity curve for two muscles with different mass [44].................................................................................................................................... 16 Figure 2-6: Force Length and Force-Velocity curve for two muscles with different fiber length [44]......................................................................................................................... 16 Figure 2-7: Framework of 3D Scanning Technology ....................................................... 17 Figure 2-8: 3D model of the Patient Skull [14] ................................................................ 18 Figure 2-9: Point cloud processed and surface reconstruction of the tray [14] ................ 18 Figure 2-10: SLA model and Titanium Prosthesis [14].................................................... 18 Figure 2-11: Cantilevered Maxillary Implant designed using a stereolithography biomodel [13].................................................................................................................... 19 Figure 2-12: Stresses in Shell Body Prosthesis [13]......................................................... 19 Figure 2-13: Motion Capture Setup with Experimental Devices [11].............................. 20 Figure 2-14: Markers positioning and figure of special pointer on Human Subject [11]. 20 Figure 2-15: The new facebow attached to a Human Subject [16]................................... 20 Figure 2-16: Snapshots of the display system [16]........................................................... 20 Figure 2-17: Ultrasonic Jaw Motion Analyzer (JMA) from Zebris GmbH [15] .............. 21 Figure 2-18: 3D jaw animation in 3D Studio Max [15].................................................... 21 Figure 2-19: CT images of the cranial part [17] ............................................................... 22 Figure 2-20: The optical 3D tracking device Polaris [17] ................................................ 22 Figure 2-21: Dry Skull for Validation Experiment [17] ................................................... 22 Figure 2-22: A display of the result of 4-dimensional analysis [17] ................................ 22 Figure 2-23: Jaw opening and closing Cycle [20] [21]..................................................... 23 Figure 2-24: Kinematic Structure of Jaw [18] [19] .......................................................... 24 Figure 2-25: SimMechanics model of Robotic chewing device [18] [19]........................ 24 Figure 2-26: Physical robot of the mastication system with Linear Actuation [21] ......... 25 Figure 2-27: Physical Kinematic model of Robotic Jaw [21]........................................... 25 Figure 2-28: Robotic Chewing Device with Crank Actuation [23].................................. 26 Figure 2-29: Co-ordinate System of RSS Linkage [23].................................................... 26 Figure 2-30: One leg of the RSS linkage [24] .................................................................. 26 Figure 2-31: WY series [31] ............................................................................................. 27
viii
Figure 2-32: WJ Series [22].............................................................................................. 28 Figure 2-33: WJ robots [22].............................................................................................. 28 Figure 2-34: JSN/2A Simulator [26]................................................................................. 29 Figure 2-35: Six bar linkage for chewing simulation [20]................................................ 29 Figure 2-36: Robotic Chewing Device [20]...................................................................... 29 Figure 3-1: Universal Joint Angles [35] ........................................................................... 31 Figure 3-2: 6DOF Stewart platform [33] .......................................................................... 31 Figure 3-3: Vector Loop of one Leg [33] ......................................................................... 31 Figure 3-4: Prismatic Actuators [33] ................................................................................ 31 Figure 3-5: Screw Co-ordinate Theory ............................................................................. 39 Figure 3-6: Manipulability Ellipsoid [34]......................................................................... 43 Figure 4-1: Sample ATOS 3D scanner generated point cloud and STL polygonal mesh images [45]........................................................................................................................ 52 Figure 4-2: Simplified serial depiction of an iterative generic Concept through Sustaining Engineering Process [45] .................................................................................................. 53 Figure 4-3: 3D Digitizer from Immersion ........................................................................ 53 Figure 4-4: Co-ordinate Measuring Machine ................................................................... 53 Figure 4-5: NextEngine Scanner [46] ............................................................................... 56 Figure 4-6: NextEngine Scanned Teeth Model [46]......................................................... 56 Figure 4-7: ATOS 3D Laser Scanner................................................................................ 57 Figure 4-8: ATOS Laser Scanner ..................................................................................... 58 Figure 4-9: Object painted in Grey color.......................................................................... 58 Figure 4-10: Scanning mandible in Capture3D ................................................................ 58 Figure 4-11: Scanned model in Capture3D ...................................................................... 58 Figure 4-12: Surface patches and holes in the geometry .................................................. 58 Figure 4-13: Cleaned Geometry in Geomagics ................................................................ 58 Figure 4-14: Scannning skull in Capture3D ..................................................................... 59 Figure 4-15: Scanned model in Capture3D ...................................................................... 59 Figure 4-16: Surface patches and holes in geometry ........................................................ 59 Figure 4-17: Cleaned geometry in Geomagics ................................................................. 59 Figure 4-18: Importing tiff images into MIMICS............................................................. 61 Figure 4-19: Setting Image and Pixel information ........................................................... 61 Figure 4-20:Specify the Orientation ................................................................................. 61 Figure 4-21: Calculate 3D to get 3D model...................................................................... 62 Figure 4-22: 3D model of bulldog .................................................................................... 62 Figure 4-23: STL import................................................................................................... 63 Figure 4-24: Importing medium resolution STL .............................................................. 63 Figure 4-25: CAD model of bulldog in Pro/E .................................................................. 64 Figure 4-26: SLA model of Bulldog................................................................................. 64 Figure 4-27: Rapid Prototype model of mandible ............................................................ 64 Figure 4-28: Rapid Prototype model of bulldog skull ...................................................... 64 Figure 4-29: Casting of the bulldog Skull and Mandible ................................................. 64 Figure 4-30: Calibration Grid from one camera view ...................................................... 66 Figure 4-31: Calibration Grid from Second Camera ........................................................ 66 Figure 4-32: Specifying the 3D Co-ordinates of the Calibration Points........................... 67 Figure 4-33: Checking the Calibration System for accuracy............................................ 67
ix
Figure 4-34: Specifying the Marker Points....................................................................... 67 Figure 4-35: Motion Capture Snapshot of a Human Subject............................................ 67 Figure 4-36 : Calibration box in three views .................................................................... 68 Figure 4-37: Calibration Grid (Left) and Motion Capture of a dog (Right) ..................... 69 Figure 4-38: Digitizing Canine for determinning Trajectory of mandible in Side Camera........................................................................................................................................... 70 Figure 4-39: Digitizing Canine for determinning Trajectory of mandible in Front Camera........................................................................................................................................... 70 Figure 4-40: Stick Figure of Mandible motion ................................................................. 70 Figure 4-41: 3D co-ordinates of RightTMJ, LeftTMJ and FrontIncisor .......................... 70 Figure 4-42:Transformation of Skull in Rhino ................................................................. 74 Figure 4-43: Transformation of mandible......................................................................... 74 Figure 4-44: Dissection of masticatory muscles (Left) and weighing muscle mass (Right)........................................................................................................................................... 76 Figure 4-45: Skull and Mandible model of bulldog.......................................................... 77 Figure 4-46: Human Jaw model in AnyBody ................................................................... 77 Figure 4-47: Human Skull and Mandible model in different views ................................. 78 Figure 4-48: sabertooth Cat Model ................................................................................... 79 Figure 4-49: Mandible model in top view ........................................................................ 79 Figure 4-50: Mandible model in front view...................................................................... 79 Figure 4-51: Script for Specifying the Joints.................................................................... 80 Figure 4-52: Muscle Attachment Points for bulldog skull and mandible......................... 80 Figure 4-53: Script for Specifying the drivers and Motion Capture data ......................... 81 Figure 4-54: Script for Specifying Muscle Models .......................................................... 81 Figure 4-55: Script for Specifying the Muscle Parameters............................................... 82 Figure 4-56: Linmot Linear Motors.................................................................................. 82 Figure 4-57: CAD model of Stewart platform in Solidworks........................................... 83 Figure 4-58: Forward Dynamic model of Stewart platform in Visual Nastran ................ 84 Figure 4-59: Forward Dynamic model of Stewart platform in DynaFlexPro................... 86 Figure 4-60: Upper and Fixed Body Frame properties..................................................... 87 Figure 4-61: Actuator Frame Properties ........................................................................... 87 Figure 4-62: Mass and Inertia of all the bodies ................................................................ 88 Figure 4-63: Universal Joint Angles ................................................................................. 88 Figure 4-64: Universal Joint Properties ............................................................................ 89 Figure 4-65: Prismatic and Spherical Joint properties...................................................... 89 Figure 4-66: Force Driver for actuators ............................................................................ 90 Figure 4-67: Free joint and Co-ordinate Selection properties .......................................... 90 Figure 4-68: Model Construction and Equation of Motion Generation ........................... 91 Figure 4-69: Simulink Block Generation.......................................................................... 92 Figure 4-70: Simulink Block Diagram to find state variables .......................................... 92 Figure 5-1: Representation of Case Studies in 3D............................................................ 94 Figure 5-2: Plot of Muscle Force and TMJ Reaction Force for Case I............................. 95 Figure 5-3: Plot of Muscle Activities for Case I............................................................... 96 Figure 5-4: Plot of Muscle Force and TMJ Reaction Force for Case II ........................... 98 Figure 5-5: Plot of Muscle Activities for Case II ............................................................. 99 Figure 5-6: Plot of Muscle Force and TMJ Reaction force for Case III......................... 100
x
Figure 5-7: Plot of Muscle Activities for Case III .......................................................... 101 Figure 5-8: Plot of Muscle Force and TMJ Reaction Force for Case IV........................ 103 Figure 5-9: Plot of Muscle Activities for Case IV.......................................................... 104 Figure 5-10: Plot of Muscle Forces and TMJ Reaction Forces for Case V.................... 105 Figure 5-11: Plot of Muscle Activities for Case V ......................................................... 106 Figure 5-12: Plot of Muscle Force and TMJ Reaction Force for Bulldog Jaw w/o Bite Force ............................................................................................................................... 108 Figure 5-13: Plot of Muscle Activities for bulldog Jaw w/o Bite Force......................... 109 Figure 5-14: Plot of Muscle Forces & Activities and TMJ Reaction Forces of Bulldog jaw with Bite Force................................................................................................................ 111 Figure 5-15: Plot of Muscle Force and TMJ Reaction Force for Sabertooth Jaw .......... 112 Figure 5-16: Plot of Muscle Activities for Sabertooth Jaw ............................................ 113 Figure 5-17: Plot of Manipulability Measures for r=0.125m ......................................... 116 Figure 5-18: Plot of Manipulability Measures for r=0.15m ........................................... 116 Figure 5-19: Plot of Manipulability Measures for r=0.2m ............................................. 117 Figure 5-20: Plot of Manipulability Measures for R=60mm.......................................... 118 Figure 5-21: Plot of Manipulability Measures for R=100mm and 200mm.................... 119 Figure 5-22: Plot of Manipulability Measures Height=200mm ..................................... 120 Figure 5-23: Plot of Manipulability Measures for Height=300mm................................ 120 Figure 5-24: Plot of Actuator Force without Mass (Left) and with Mass of Legs (Right) for Case I......................................................................................................................... 121 Figure 5-25: Plot of Actuator Force without Mass (Left) and with Mass of Legs (Right) for Case II ....................................................................................................................... 122 Figure 5-26: Simulink Diagram for Simulation using S-Function ................................. 124 Figure 5-27: Actuator Forces for Case I using S-Function............................................. 125 Figure 5-28: Stewart platform Tracking Line (Left) and Error Plot (Right) for Case I.. 125 Figure 5-29: Actuator Forces for Case II using S-Function ........................................... 126 Figure 5-30: Stewart platform Tracking Line (Left) and Error Plot (Right) for Case II 126 Figure 5-31: Simulink Diagram using DynaFlexPro Model........................................... 127 Figure 5-32: Plot of Actuator Forces for Case I (Left) and Case II (Right) ................... 128 Figure 5-33: Error Plot (Left) and Stewart platform Tracking Line (Right) for Case I.. 128 Figure 5-34: Error Plot (Left) and Stewart platform Tracking Line (Right) for Case II 128 Figure 5-35: CAD model of Stewart platform in Solidworks......................................... 130 Figure 5-36: Simulink diagram with DynaFlexPro model (Left) and Actuator Force plot (Right) ............................................................................................................................. 130 Figure 5-37: Stewart platform tracking a Circle (Left) and Error plot (Right) using DFP model............................................................................................................................... 131 Figure 5-38: Simulink diagram with S-Function block (Left) and Actuator Force plot (Right) ............................................................................................................................. 132 Figure 5-39: Stewart platform tracking a Circle (Left) and Error plot (Right) using S-Function .......................................................................................................................... 132 Figure 5-40: Simulink diagram with Visual Nastran block (Left) and Actuator Force plot (Right) ............................................................................................................................. 134 Figure 5-41: Stewart platform tracking a Circle (Left) and Error plot (Right) using VN model............................................................................................................................... 134 Figure 5-42: Transformation from Base to Moving platform......................................... 135
xi
Figure 5-43: Transformation from Origin of Calibration to moving reference frame.... 135 Figure 5-44: Vectors along Marker Points...................................................................... 135 Figure 5-45: Stewart platform Tracking Jaw motion...................................................... 137 Figure 5-46: Trajectory tracked by Stewart platform ..................................................... 137 Figure 5-47: Trajectory of the Front Incsior from Motion Analysis .............................. 137 Figure 5-48: Actuator Force Plot for External Force of 0N (T-L), 50N (T-R), 100N (B-L), 200N (B-R)................................................................................................................ 138 Figure 5-49: Stewart platform tracking Human Jaw motion in Visual Nastran ............. 138 Figure 5-50: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Human Jaw......................................................................................................................................... 139 Figure 5-51: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right) ............................................................................................................................. 139 Figure 5-52: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Human Jaw II...................................................................................................................................... 139 Figure 5-53: Actuator Force Plot for External Force of 0N (T-L), 50N (T-R), 100N (B-L), 200N (B-R) ..................................................................................................................... 140 Figure 5-54: Platform tracking Human Jaw Motion in Visual Nastran.......................... 140 Figure 5-55: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right) of Bulldog........................................................................................................... 141 Figure 5-56: Actuator Forces for External Force of 0N (T-L), 100N (T-R), 200N (B-L), 400N (B-R) ..................................................................................................................... 142 Figure 5-57: Platform tracking Bulldog Jaw motion in Visual Nastran ......................... 142 Figure 5-58: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Bulldog Jaw .................................................................................................................................. 142 Figure 5-59: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right) of Sabertooth ...................................................................................................... 143 Figure 5-60: Actuator Forces for External Force of 0N (T-L), 100N (T-R), 500N (B-L), 1000N (B-R) ................................................................................................................... 144 Figure 5-61: Platform tracking Sabertooth Jaw motion in Visual Nastran..................... 144 Figure 5-62: Actuator Stroke for Sabertooth Jaw........................................................... 144
xii
LIST OF TABLES Table 4-1: NextEngine Specifications .............................................................................. 55 Table 4-2: ATOL Laser Scanner Specifications............................................................... 56 Table 4-3: Muscle Parameters of Human Model.............................................................. 75 Table 4-4: Muscle Relative % Weights of Canine and Felis ............................................ 76 Table 4-5: Muscle Parameters of Bulldog Model............................................................. 76 Table 4-6: Muscle Parameters of Sabertooth Model ........................................................ 77 Table 4-7: Mass and Inertia of Mandible.......................................................................... 77
1
1. Introduction 1.1 Motivation
The goal of this project is to quantitatively measure mechanical performance signals such
as forces, motions and pressures during mastication with the intent of subsequently
characterizing mechanical breakdown of food in various vertebrates (including humans).
Such an understanding would be of tremendous importance from a variety of
perspectives. From a science perspective, it is useful to know how various animals
(including humans) preprocess the food for subsequent digestion. From an economic
perspective it potentially enables food manufacturing/processing companies to design and
process foods based on the “chewability index”. Such knowledge could potentially enable
the orthotists design, develop, fit and manufacture dental orthoses to support or correct
musculoskeletal deformities and or abnormalities of vertebrate jaws.
To aid us in this process of quantitatively measuring various mechanical output
parameters such as force/motion/pressure, we will examine use of current technological
tools and paradigms in measure-estimate-test cycle as shown in Figure 1-2 by a
combination of “Virtual Prototyping” and “Physical Prototyping”. In the last decade, the
science and engineering domains have been revolutionized by the ubiquitous availability
of computational power coupling with the advances in computational tools, algorithms
and methodologies as depicted in Figure 1-1 [1]. While the engineering-related fields
have seen the greatest benefits, the lack of significant and advanced computational
biomechanical tools has hindered progress in other arenas such as the biological sciences.
Such tools could potentially assist biologists to perform various parametric “what if” type
analyses to test various hypotheses.
However, there exist several problems when expanding into and exploring other scientific
domains such as biological sciences. However, these fields are now poised to take
advantage of the systematic and parametric techniques developed in the engineering
realm to enhance the ability of biological scientists to analyze study and verify theories
and hypotheses. For instance, it is difficult to model and simulate a living tissue, muscle,
2
tendon and ligament when compared to the far simpler mechanical systems. In particular,
the inhomogeneity and irregularity that is the norm in biological specimens necessitates
considerable biomechanical expertise for accurate modeling and simulations. In this
project, we explore the use of advanced virtual prototyping simulation tools for analyzing
both mechanical as well as biological multi-body systems. From a mechanical
engineering viewpoint we analyze 6DOF parallel manipulators for reproducing
mastication motion using virtual prototyping tools and determine the actuator forces and
mastication performance. From a biological viewpoint, a musculoskeletal model of the
vertebrate jaw will be created and analyzed to determine muscles forces required for
various biting tasks. We examine this in greater detail specifically using the case studies
of bite force measurement and masticatory performance measurement in vertebrates as
shown in Figure 1-3.
Figure 1-1: Development of Simulation Tools in Engineering
5.1.3. Case Study III: One Temporalis Muscle, Muscle Model 3E, No External Force In this case the muscle has been modeled as Hill Muscle model by considering
muscle’s force-length and force-velocity relationships. Hence in this case we see a
reaction force at the two TMJ for compensating the passive forces of the muscles
during the wide opening of the jaw. This muscle model is more realistic as compared
to the simple muscle model since it considers the passive stiffness of muscles.
100
(Masseter Muscle Force ) (Temporalis0 Muscle Force )
(Medial Pterygoid Muscle Force ) (Lateral Pterygoid Muscle Force )
(Digastrics Muscle Force ) (TMJ Reaction Force )
Figure 5-6: Plot of Muscle Force and TMJ Reaction force for Case III Now we can see an increase in the magnitude of muscle forces in Figure 5-6 due to the
Complex Muscle model which considers other architectural properties of the muscle as
mentioned above. The muscle activation level also increases in the third case study as
Figure 5-16: Plot of Muscle Activities for Sabertooth Jaw
5.4 Manipulability Measures for Workspace Analysis
Performance measures play a very vital role in quantitatively evaluating the workspace
quality of a manipulator in engineering design for the use in subsequent evaluation and
optimization of the performance. They play an important role in design, analysis,
evaluation and optimization for a robotic mechanical system. Workspace analysis was
performed to determine the manipulability of the Stewart platform type parallel
manipulator in Matlab. This analysis allows estimation of actuator lengths and size of
114
upper and fixed platforms required to reproduce the jaw motion trajectory of vertebrates.
By implementing workspace analysis it is possible to perform parametric analysis, by
varying the actuator lengths and platform dimensions, and determine the reachable
workspace of the parallel manipulator. As a first step, the Jacobian matrix of the 6DOF
parallel manipulator was computed as derived in Chapter 3. Singular value
decomposition of this Jacobian matrix was computed using the svd() function in Matlab.
The manipulability ellipsoid, yoshikawa measure and isotropy index were computed after
computing the singular value decomposition of Jacobian matrix, as shown in chapter 3.
The length of the actuator, radius of the upper platform and the radius of the workspace
in polar co-ordinates were parametrically varied to determine their optimum value for
best workspace quality. To suit the simulation of jaw motion, the radius of the fixed
platform was fixed at 0.2m. The upper platform’s radius was varied from 0.125m to 0.2m
and the range seems to be compatible with the size of the vertebrate jaw. The actuator are
required to produce a force of about 200-300N for withstanding the biting force. Hence
Linmot linear actuators were chosen for actuating the parallel mechanism. The lengths of
these actuators are varied from 0.175m to 0.3m to generate the required force. All these
actuators of different lengths have the capability of generating force ranging from 150-
300N which is sufficient to reproduce the mastication motion of vertebrates.
5.4.1 Case Study I: Varying the Radius of the Moving platform In the first case study the radius of the workspace was fixed at 40mm (for mastication
motion), the length of the actuator of the height between the upper and lower platform
was fixed at 0.2. The radius of the upper platform r was varied from 0.125 m to 0.2m to
determine the workspace quality of the moving platform for different dimensions.
Stewart platform configuration, surface plot of isotropy index of manipulability,
manipulability ellipsoid, and surface plot of yoshikawa measure of manipulability is
plotted in Figure 5-17 below. The manipulability ellipsoids are scaled down for clarity
and it shows that the manipulator is able to move the upper platform more along the
major axes and less along the minor axes. Since the manipulability ellipsoid does not
shrink and become straight line along minor axes, the end effector has a reasonable
workspace quality and does not get affected by singularity. Such visualization tool is
particularly useful in studying the motion transmission capability of a robotic system. We
115
also examine the behavior of all the manipulability indices described in the previous
section, i.e. yoshikawa measure of manipulability and isotropy index of the manipulator
over a grid specified in polar co-ordinates with radius of 40mm. We can from the surface
plots that the manipulator has lesser manipulability with lesser radius of workspace. It
can be seen that the isotropy index varies from 0 to 1 and has the best numerical
behavior. Isotropy index of 1 indicates the best manipulability and in this case we can see
that the isotropy index is about 0.345 to 0.39 which indicates there is not singularity.
From the workspace analysis we try to find out the point of workspace at which the
parallel manipulator becomes singular. We would evaluate the workspace quality of the
parallel manipulator to avoid singularity and damage to the actuators. As the radius of the
moving platform is increased from 0.125m to 0.2m, the manipulability of the manipulator
increases and the ellipsoids tends towards isotropy and the yoshikawa measure of
manipulability and isotropy index increases from 0.2 to 0.47 and from 0.4 and 0.8
respectively as seen from Figure 5-17- Figure 5-19.
Stewart platform Configuration Manipulability Ellipsoid
116
Isotropy Index Yoshikawa Measure
Figure 5-17: Plot of Manipulability Measures for r=0.125m
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure
Figure 5-18: Plot of Manipulability Measures for r=0.15m
117
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure Figure 5-19: Plot of Manipulability Measures for r=0.2m
5.4.2 Case Study II: Varying Radius of the Workspace From the above case study the radius of the upper platform r was fixed at 0.175m for a
better workspace quality. Also the manipulator does not hit singularity with this radius as
it moves along the workspace discretized in polar co-ordinates. However in this study,
the radius of the discretized workspace was increased from 40mm to 100mm just for
evaluation purpose. It can be concluded from these plots that the manipulator has a good
workspace quality even for increased radius of the workspace and does not hit the
singular region. For a mastication motion application, the range of motion in vertical
direction is about 50 mm and about 20mm in surge and sway direction. Hence it can be
asserted that the manipulator with a radius of upper platform as 0.175m can reach jaw
motion workspace without causing singular issues. The upper platform reaches the
workspace specified by mastication motion quite easily without any singular problems
and hence the design of the manipulator and actuators are quite safe. But the manipulator
looses its manipulability at the radius of more than 100 mm as seen from Figure 5-20 &
5-21.
118
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure Figure 5-20: Plot of Manipulability Measures for R=60mm
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure
119
Isotropy Index Yoshikawa Measure Figure 5-21: Plot of Manipulability Measures for R=100mm and 200mm
5.4.3 Case Study III: Varying the Vertical Distance between Two platforms In this study the height of the moving platform was varied from 150mm to 300mm and
the workspace quality of the manipulator was evaluated for these different heights to
accommodate the vertical motion of the mastication. Similar measures of manipulability
and manipulability ellipsoids are plotted as the height of the platform is increased from
150mm to 300mm in Figure 5-22 & 5-23. In this study the radius of the upper platform
was fixed at 0.175m and the radius of the workspace at 60mm. We choose the actuator
length as 200mm for a vertical mastication of 50 mm in humans and bulldog.
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure
120
Isotropy Index Yoshikawa Measure Figure 5-22: Plot of Manipulability Measures Height=200mm
Stewart platform Configuration Manipulability Ellipsoid
Isotropy Index Yoshikawa Measure Figure 5-23: Plot of Manipulability Measures for Height=300mm
5.5 Dynamic Simulation of Stewart platform
From the equations of motion derived in Chapter 3, inverse dynamic simulation of
Stewart platform was implemented using Matlab. Firstly the actuators were assumed to
have zero mass and inertia and equations of motion were derived considering only the
mass and inertia of the moving platform. Configuration of the platform and other system
parameters such as ball joint locations, mass and inertia properties of the platform and
legs were taken from Tsai et al.
121
Case Study I:
In first case study, orientation of the moving platform remains constant while the center
of mass moves along a lone passing through the origin and pointing in [1 1 1] direction
with a sinusoidal motion and represented as:
( )( )( )
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
+
+−
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
000sin2.01
sin2.0sin2.05.1
tt
t
zyx
z
y
x
p
p
p
ωω
ω
φφφ
Case Study II:
In the second case study, the moving platform rotates about the z axis with a sinusoidal
motion while the center of mass remains stationary. It can be represented as:
( )⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡ −
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
t
zyx
z
y
x
p
p
p
ωφφφ
sin35.00010
5.1
Figure 5-24: Plot of Actuator Force without Mass (Left) and with Mass of Legs (Right) for Case I
122
Figure 5-25: Plot of Actuator Force without Mass (Left) and with Mass of Legs (Right) for Case II The above Figure 5-24 and 5-25 show the forces acting in each of the six actuators as the
moving platform is tracking desired end effector trajectory as specified in case study I
and II. From the above two plots we can see that the actuator force has a sinusoidal form
and its magnitude is ranging between 0 N to 11N in first case and 3.4N to 6.6N in second
case. Now the actuators were assigned mass and inertia properties and the equations of
motion included the contribution of the six actuators to the mass/inertia matrix. Similarly
the above plots show the forces acting in each of the six actuators, by including the
dynamics of the legs as well, as the moving platform is tracking desired end effector
trajectory as specified in case study I and II. From the above two plots we can see that the
actuator force has a sinusoidal form and its magnitude is ranging between 0 N to 15N in
first case and 4.5N to 9.5N in second case. As the results from the second set of study
including the dynamics of leg produces more accurate results when compared to the
results from Tsai, it can be concluded that the dynamics of six actuators plays a important
role while deriving the equations of motion. For the purpose of sliding mode control,
some researchers have neglected the dynamic properties of legs for simplicity and
implemented the control algorithm. However from the above results it can be stated that
the dynamics of the legs contributes to the actuator forces and increases the magnitude of
these six actuator forces. Hence in the subsequent inverse dynamic simulation for
simulating the jaw motion, dynamics of the six actuators will be considered while
deriving the equations of motion.
123
A PD control based feedback linearization technique has been implemented using
Matlab/Simulink to make sure that the moving platform tracks the desired trajectory and
the error between the desired trajectory and the end-effector (moving platform) trajectory
converges to zero. Trajectory control of a six DOF Stewart platform type parallel
manipulator was simulated using 1. Matlab/Simulink 2. DynaFlexPro. For validation
purposes, the system parameters and the desired trajectory information were taken from
Tsai to verify if our Lagrangian formulation, DynaFlexPro model and control algorithm
worked efficiently. In these simulations inverse dynamic block computes forces required
at the six actuators and feeds to forward dynamics block for computing current position
and orientation of platform and the current position is first fed to the PD control block
which measures the difference between desired and end-effector trajectory and uses
control gains to minimize and converge the error to zero. It is then fed back to the inverse
dynamics block to compute the actuator forces and closed loop PD control goes on. In the
first method the forward dynamics block, which computes the current end-effector
position and orientation, is implemented based on Lagrangian Formulation using an S-
Function block. In the second method, a DynaFlexPro Forward Dynamics model was
built to compute the end-effector position and orientation based on actuator forces and
this replaces the S-Function block. The two methods are simulated for tracking the
desired trajectories as mentioned in two case studies above and actuator forces obtained
from these methods are then compared to the results of Tsai and [49] to validate our
inverse and forward dynamic simulation and control framework.
5.5.1 Simulation Using S-Function: Case Study I
Firstly inverse and forward dynamics of Stewart platform type 6 DOF parallel
manipulator with PD controller was simulated completely in Matlab/Simulink.
Configuration of the platform and other system parameters such as ball joint locations,
mass and inertia properties of the platform and legs were taken from Tsai. Following
simulations were performed to verify the results and the implementation from both
Matlab and DynaFlexPro. It is assumed that the gravitational force is the only external
force acting on the links. For this simulation the orientation of the moving platform
124
remains constant while the center of mass moves along a lone passing through the origin
and pointing in [1 1 1] direction with a sinusoidal motion.
Figure 5-26 shows the Simulink block diagram for simulating the forward and inverse
dynamics along with a PD controller for motion control of the moving platform.
Figure 5-26: Simulink Diagram for Simulation using S-Function
The actuator forces acting at the six prismatic joints are calculated as functions of time
and plotted in Figure 5-27. These results are matching with [35] within close limits. By
implementing the control algorithm the moving platform is controlled to perform the
desired motion and the inverse dynamics formulation computes as the moving platform
performs desired motion. The control gains are tuned to get the error between the desired
trajectory and end effector trajectory converge to zero. Also we can see that if the
forward dynamic block is written as an S-Function block, then the system settles in
almost no time. From the error plot in Figure 5-28 we can see that the error in the
solution is very small and moving platform tracks the desired trajectory with almost zero
error.
125
Figure 5-27: Actuator Forces for Case I using S-Function
Figure 5-28: Stewart platform Tracking Line (Left) and Error Plot (Right) for Case I
Case Study II
Similar to the above simulation, the second case study was simulated where the moving
platform rotates about the z-axis and the center of mass of the moving platform remains
stationery. Above plots show that the actuator force plot in Figure 5-29 is very similar to
the one in Tsai and the error in the solution is very small and the system settles down in
no time. As mentioned above, since the forward dynamics block is implemented as an S-
Function block, the error in between desired trajectory and moving platform trajectory is
almost zero. Figure 5-30 also shows the Stewart platform configuration as it is
performing desired motion.
126
Figure 5-29: Actuator Forces for Case II using S-
Function
Figure 5-30: Stewart platform Tracking Line (Left) and Error Plot (Right) for Case II
5.5.2 Simulation Using DynaFlexPro Model: Similar simulation was run in Simulink but in this study the S-function block was
replaced by a forward dynamics DynaFlexPro model as shown in Figure 5-31. Results
from both the simulation match each other within close limits. This simulation validates
the forward dynamics DynaFlexPro model with the S-Function block. Both the Matlab S-
Function block and DynaFlexPro model behave similarly and produce results that match
within close limits with Tsai’s paper.
127
Figure 5-31: Simulink Diagram using DynaFlexPro Model
Case Study I and II
Similar to the above simulation run using an S-Function block, a forward dynamics block
was created using DynaFlexPro to compute the moving platform’s current position and
orientation based on the actuator forces as the input. Firstly trajectory mentioned in Case
Study I was simulated and the actuator forces were plotted vs. time to validate the
DynaFlexPro model. The actuator forces acting at the six prismatic joints are calculated
as functions of time and plotted in the Figure 5-32 below. These results are matching
within close limits with Tsai’s paper results and above mentioned S-Function block. The
system settles after 0.1s and the force plot matches well with the benchmark solution.
However the setting time depends upon the Baumgarte’s stabilization parameters
mentioned in Chapter 4. In this case the Baumgarte’s stabilization parameters were set as
10=α and 10=β . Unfortunately there is no guideline to set these parameters and they
were tuned to ensure that the error between the desired trajectory and end effector’s
current trajectory converges to zero. Figure 5-32 below show the forces acting in the
actuators as the moving platform performs the desired motion as specified in the two case
studies. The magnitude and the sinusoidal nature of the force plot are very similar to the
benchmark solution after a settling time of 0.1s and error in the solution converges to
zero exponentially. Stewart platform configuration tracking the desired motion is plotted
in below figure. Figure 5-33 and Figure 5-34 shows the actuator force plot, Stewart
128
platform configuration plot and the error plot as the moving platform tracks the desired
trajectory mentioned in case study II.
Figure 5-32: Plot of Actuator Forces for Case I (Left) and Case II (Right)
Figure 5-33: Error Plot (Left) and Stewart platform Tracking Line (Right) for Case I
Figure 5-34: Error Plot (Left) and Stewart platform Tracking Line (Right) for Case II Figure 5-33 and 5-34 shows the Stewart platform configuration tracking a straight line
and rotating about z axis respectively. Also PD controller controls the motion of the
upper platform reasonably well and the error in the solution converges almost to 0 in 1 s.
129
Thus the forward and inverse dynamic analysis and PD control of 6DOF Stewart
platform type parallel manipulator has been performed using two above mentioned
methods and the results are matching well with the benchmark solutions provided by
Tsai. Also the motion controller has been implemented by tuning the gain factors
appropriately. Both the case studies have been evaluated to validate the S-Function block,
DynaFlexPro forward dynamics model and the Lagrangian formulation. In the
subsequent study a CAD model of the 6 DOF Stewart platform type parallel manipulator
will be built in SolidWorks and the above mentioned methods will be employed to
perform the inverse dynamic simulation and the motion control of the moving platform.
5.6 Simulation of the CAD Model of Stewart platform:
For this simulation, a CAD model of a 6Dof Stewart platform type manipulator has been
created in SolidWorks as shown in Figure 5-35. All the parts such as moving platform,
fixed platform, joints and actuators are assigned plain carbon steel material properties.
The dimensions of the upper and fixed platforms were parameterized based on radius of
the respective platforms. The length and the diameter of the actuator were modeled based
on the specifications of the Linmot linear motors. These linear motors are about 200 mm
long and have a diameter of 5mm. These actuators can produce forces close to 500N and
it is sufficient for the jaw motion simulation. To validate the S-Function block,
DynaFlexPro forward dynamics model and the Visual Nastran forward dynamics model,
a simple circular trajectory has been specified. The model has been imported into Visual
Nastran for forward dynamics simulation of the platform. In this simulation the center of
mass of the upper platform tracks a circle trajectory with its center at origin and radius of
20mm in the X-Z plane. Again all the simulations have been performed using three
methods for the same circular trajectory:
( )
( )
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
+−
+−
=
⎥⎥⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢⎢⎢
⎣
⎡
000
sin152.0000589.0
cos02.000419.0
t
t
zyx
z
y
x
p
p
p
π
π
φφφ
130
1. Forward Dynamics using S-Function block
2. Forward Dynamics using DynaFlexPro model
3. Forward Dynamics using Visual Nastran Plant
Figure 5-35: CAD model of Stewart platform in SolidWorks
5.6.1 Simulation Using DynaFlexPro Model System parameters such as ball joint locations, mass and inertia properties of upper
platform and legs have been taken from the SolidWorks CAD model. In this simulation
the upper platform tracks a circle trajectory as specified above and the forces acting in the
6 actuator are plotted against time. The Stewart platform configuration tracking the
circular trajectory is shown in the Figure 5-37 and Actuator Forces in Figure 5-36.
Figure 5-36: Simulink diagram with DynaFlexPro model (Left) and Actuator Force plot (Right)
131
The error in the solution which is the difference between the desired trajectory and the
end-effector trajectory has been showed in the Figure 5-37. It can be stated that the
system settles in 0.1s and starts to track the desired trajectory. Again the model produces
some error at the beginning due to the Baumgarte stabilization parameters. We can see
from these plots that the results match with the results produced by employing S-
Function forward dynamic block. This validates our CAD model as well as the
DynaFlexPro forward dynamics model built for the same CAD model. This also ensures
that the control algorithm based on feedback linearization has been implemented in an
efficient manner.
Figure 5-37: Stewart platform tracking a Circle (Left) and Error plot (Right) using DFP model
5.6.2 Simulation Using S-Function Block For computing the current position and orientation and their derivatives of upper platform
based on the actuator forces, an S-Function block was created by hand coding the
dynamics equations of motion using Lagrangian formulation in task space. System
parameters such as ball joint locations, mass and inertia properties of upper platform and
legs have been taken from the SolidWorks CAD model. In this simulation the upper
platform tracks a circle trajectory similar to the above study and the forces acting in the 6
actuator are plotted against time. This method was adopted to verify the results obtained
from Lagrangian formulation. From the Figure 5-39 it can be inferred that the PD control
algorithm reduces the error to almost zero and the upper platform tracks a circle
trajectory as shown below. In this case since there is no such use of a simulation tool to
build the forward dynamics block, the error in the solution is almost zero right from the
beginning of the simulation. The computational time required to run this simulation is
132
much lower than the simulation with DynaFlexPro model. From a computational point of
view, simulation using the S-Function block is much faster but the equations of motion of
a complex n DOF system cannot be derived easily without the aid of multi-body dynamic
simulation tool. We can also see that the forces acting in the six actuators are ranging
from 0 to 3N and vary sinusoidally in both the simulations as seen in Figure 5-38 and
thus verify our modeling and simulation procedure in Matlab/Simulink environment.
Figure 5-38: Simulink diagram with S-Function block (Left) and Actuator Force plot (Right)
Figure 5-39: Stewart platform tracking a Circle (Left) and Error plot (Right) using S-Function
5.6.3 Simulation Using Visual Nastran Plant As mentioned above, though the simulation using the S-Function block computationally
runs much faster than the DynaFlexPro model (mainly due to symbolic equations of
motion and uncontrollable Baumgarte stabilization parameters), deriving the equations of
motion of a multi degree of freedom parallel manipulator by hand could be very
133
laborious. At the same time, creating a model in DynaFlexPro needs a good knowledge of
the frames of reference, generalized co-ordinates for the initial modeling. DFP still does
not possess the conveniences that commercial-off-the-shelf packages offer in terms of
model creation and simulation. Features such as 3D visualization capabilities, automated
mass and inertial calculations (to name a few) are currently missing. Finally, while the
automated processing to create the EOMs has been adequately shielded from the user, the
user is expected to have a good grasp of Maple programming concepts and data-storage
constructs to effectively use the results. Hence in the next study, MSC Visual Nastran
was used to build the forward dynamics block. The best feature about this product is that
it can be integrated seamlessly with SolidWorks. Building a CAD model of a complex
system is much simpler than deriving its equation of motion or building a DFP model.
Once the CAD model is built it can be transferred seamlessly to the Visual Nastran tool
to create dynamic model. All the constraints were checked before running the simulation
and the Visual Nastran Plant of this CAD model was added to the Simulink control loop
to perform the forward dynamics. Similar simulation was performed using a forward
dynamics Visual Nastran plant for simulating the forward and inverse dynamic analysis
of Stewart platform. All these simulations were performed using Euler method or ODE1
solver at time step of 0.001s. The forces acting in the six prismatic actuators are plotted
against time in the Figure 5-40. Also the error in the solution converges to 0 pretty
quickly and the upper platform tracks the circle trajectory with minimal error in the
solution (Figure 5-41). However since the Visual Nastran solver does not have the
velocity and the acceleration information at the first time step and also the initial position
is little away from the desired trajectory, we see a larger actuator force to compensate for
the error in the solution. But the system settles at about 0.05s and the moving platform
tracks the desired circular trajectory. From above plots it can be inferred that the forward
and inverse dynamic simulation of Stewart platform type manipulator using these three
methods produces similar results and PD controller works well with suitable gain factors
to converge the error to zero for tracking desired motion. This also validates our forward
dynamics block created using commercial simulation tools such as DynaFlexPro and
Visual Nastran. Due to simplicity and ease of use we will be using only Visual Nastran
for determining the actuator forces as the moving platform reproduces the jaw motion of
134
various subjects captured using motion capture analysis. This analysis estimates the
actuator forces required for performing the mastication movement and will help size the
actuator. From chapter 4, we chose design and size of linear actuators from Linmot since
they have the required stroke and force producing capability which will be suitable for
reproducing the jaw motion. Hence such inverse dynamics open loop analysis will verify
and validate the specifications of the linear actuators and could be of use while building a
physical prototype of the 6DOF Stewart platform type manipulator.
Figure 5-40: Simulink diagram with Visual Nastran block (Left) and Actuator Force plot (Right)
Figure 5-41: Stewart platform tracking a Circle (Left) and Error plot (Right) using VN model
135
5.7 Simulation of Jaw Motion Using Visual Nastran
3D co-ordinates of the three marker points, RightTMJ, LeftTMJ and Front Incisor,
obtained from the motion capture analysis of mastication motion of different subjects are
w.r.t. the inertial frame of reference fixed at the origin of the calibration grid. Hence in
order for the moving platform of the 6DOF parallel manipulator to reproduce the
mastication motion, the jaw motion trajectory has to be mapped to the base reference
frame of the Stewart platform. We use Gram Schmidt Orthogonalization principle to map
the 3D co-ordinates of the jaw motion from inertial frame of reference of calibration grid
into base reference frame of Stewart platform. This mapping would ensure accurate input
jaw motion trajectory to the moving platform. From the Figure 5-43, the inertial frame of
reference of the calibration box is termed as {O} and the frame of reference attached to
the first point in the marker data as {I}. There is a transformation matrix 0A associated for
transforming the co-ordinates from frame {O} to frame {I}. The three marker points
tracked are RightTMJ, LeftTMJ and Front Incisor as shown in Figure 5-44. Let vector 1v
be the vector from Front Incisor to the LeftTMJ and 2v from Front Incisor to RightTMJ.
Figure 5-42: Transformation from Base to
Moving platform
Figure 5-43: Transformation from Origin of
Calibration to moving reference frame
Figure 5-44: Vectors along Marker Points
136
The unit vectors along these vectors can be computed as
11
1
vev
= (1)
1 23
1 2
ˆ v vev v×
=×
(2)
2 3 1ˆ ˆ ˆe e e= × (3)
Hence the Rotation matrix is given by
[ ]1 2 3ˆ ˆ ˆIOR e e e= (4)
The homogenous transformation matrix 0A is given as
11 12 13
21 22 230
31 32 33
0 0 0 1
x
y
z
r r r dr r r d
Ar r r d
⎡ ⎤⎢ ⎥⎢ ⎥=⎢ ⎥⎢ ⎥⎣ ⎦
(5)
Similarly the transformation matrix ( )A t can be found to transform the co-ordinates of
each point in the marker dataset as a function of time. Then the relative transformation
matrix ( )rA t is defined as the transformation of each point in the trajectory data with the
first point of the motion capture data and written as
( ) ( )10rA t A A t−= (6)
Similarly the transformation matrix 0B is defined as the transformation between base
reference frame { }B of Stewart platform and platform reference frame { }P and is
derived in the Chapter 3 and shown in Figure 5-42. Now the overall homogenous
transformation matrix to map the 3D co-ordinates of the motion capture data with respect
to base reference frame is given as
( ) ( )0 rB t B A t= (7)
5.7.1 Dynamic Simulation of First Human Subject Jaw motion
Motion capture analysis was performed to measure the 3D co-ordinates of the Front
Incisor point on the mandible during mastication cycle. This data served as the input
trajectory for performing inverse dynamic analysis of the 6DOF Stewart platform using
Matlab/Simulink and Visual Nastran. Four different subjects, two male human, one dog
137
and a saber tooth jaw simulator were considered for the study. Trajectory of the front
incisor point was specified as the desired trajectory for the moving platform and the
forces acting in the actuators are computed. For each of these case studies, we simulated
by varying the external bite force on the moving platform from 0 to 300N. Typically for
human and bulldog, the bite force measurements were done by researchers and reported
to be close to 150-300N [43]. These parametric analyses were performed to verify the
size and specification of the actuator available from Linmot Motor. Figure 5-45 & 5-49
below shows the Stewart platform tracking jaw motion of a human subject along with the
actuator force plot in Figure 5-48 while trajectory tracked by platform is shown in Figure
5-46 which is similar to the trajectory of the Front Incisor (Figure 5-47). Figure 5-50
shows the actuator stroke and velocity profile and it can be stated that we need an
actuator to slide about 50mm at 100mm/s for reproducing human jaw motion.
Figure 5-45: Stewart platform Tracking Jaw motion
Figure 5-46: Trajectory tracked by Stewart
platform
Figure 5-47: Trajectory of the Front Incisor from
Motion Analysis
138
Figure 5-48: Actuator Force Plot for External Force of 0N (T-L), 50N (T-R), 100N (B-L), 200N (B-R)
Figure 5-49: Stewart platform tracking Human Jaw motion in Visual Nastran
139
Figure 5-50: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Human Jaw
5.7.2 Dynamic Simulation of Second Human Subject Jaw motion In this case the motion of the mandible of another human subject was captured and
simulated using Matlab and Visual Nastran. Figure 5-51 & 5-54 below shows the jaw
motion trajectory captured by motion capture analysis and tracked by Stewart platform.
The forces acting in the actuators for different external forces are plotted in Figure 5-53.
Figure shows 5-54 the Stewart platform in Visual Nastran tracking the desired chewing
trajectory. From the actuator force plot it seems that as the biting force is increased from
0 to 200N, the forces acting in the actuator increases from 3N to 100N as seen in Figure
5-53. Figure 5-52 shows the actuator stroke and velocity profile and it can be stated that
we need an actuator to slide about 50mm at 100mm/s for reproducing human jaw motion.
Figure 5-51: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right)
Figure 5-52: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Human Jaw II
140
Figure 5-53: Actuator Force Plot for External Force of 0N (T-L), 50N (T-R), 100N (B-L), 200N (B-R)
Figure 5-54: Platform tracking Human Jaw Motion in Visual Nastran
141
5.7.3 Dynamic Simulation of Bulldog Subject Jaw motion
In next case the motion of the mandible of bulldog subject was captured and simulated
using Matlab and Visual Nastran. Figure 5-55 & 5-57 below shows the jaw motion
trajectory captured by motion capture analysis and tracked by Stewart platform. The
forces acting in the actuators for different external forces are plotted in Figure 5-56.
Figure 5-57 shows the Stewart platform in Visual Nastran tracking the desired chewing
trajectory. From the actuator force plot it seems that as the biting force is increased from
0 to 400N, the forces acting in the actuator increases from 3N to 200N as seen in Figure
5-56. Figure 5-58 shows the actuator stroke and velocity profile and it can be stated that
we need an actuator to slide about 60mm at 125mm/s for reproducing bulldog jaw
motion.
Figure 5-55: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right) of
Bulldog
142
Figure 5-56: Actuator Forces for External Force of 0N (T-L), 100N (T-R), 200N (B-L), 400N (B-R)
Figure 5-57: Platform tracking Bulldog Jaw motion in Visual Nastran
Figure 5-58: Plot of Actuator Stroke (Left) and Actuator Velocity (Right) for Bulldog Jaw
143
5.7.4 Dynamic Simulation of Sabertooth Cat Jaw motion In this case the motion of the mandible of sabertooth Tiger subject was captured and
simulated using Matlab and Visual Nastran. Figure 5-59 & 5-61 below shows the jaw
motion trajectory captured by motion capture analysis and tracked by Stewart platform.
The forces acting in the actuators for different external forces are plotted in Figure 5-60.
Figure 5-61 shows the Stewart platform in Visual Nastran tracking the desired chewing
trajectory. From the actuator force plot it seems that as the biting force is increased from
0 to 1000N, the forces acting in the actuator increases from 3N to 330N as seen in Figure
5-60. Figure 5-62 shows the actuator stroke and velocity profile and it can be stated that
we need an actuator to slide about 120mm for reproducing sabertooth jaw motion. In this
case since we could not capture the jaw motion of an actual sabertooth tiger, we could not
estimate the actuator velocity profile. Based on anatomical experience, sabertooth cats
kill their prey by closing the jaw slowly and hence the actuator velocities will be lesser in
this case.
Figure 5-59: Trajectory tracked by Stewart platform (Left) and Front Incisor Trajectory (Right) of
Sabertooth
144
Figure 5-60: Actuator Forces for External Force of 0N (T-L), 100N (T-R), 500N (B-L), 1000N (B-R)
Figure 5-61: Platform tracking Sabertooth Jaw motion in Visual Nastran
Figure 5-62: Actuator Stroke for Sabertooth Jaw
145
6. Conclusion and Future Work
6.1 Conclusion
The primary goal of this project to use Virtual Prototyping tools to estimate:
1. Force acting in the actuators as the Stewart platform is reproducing the
mastication motion of vertebrates.
2. Muscle Forces and TMJ reaction Forces as the musculoskeletal model of
vertebrate jaw is performing mastication motion.
Although prior work has been done in developing human jaw motion simulators, no work
was done in reproducing the mastication motion of animals. And these human jaw motion
simulators were built for specific patient. Hence in this thesis work, we modeled and
simulated a more generic 6DOF parallel manipulator i.e. Stewart platform for
reproducing the jaw motion of vertebrates. The other reason for simulating a Stewart
platform was that there is no size limitation on the actuators as the skull and mandible
will be mounted on top of the moving platform. For achieving the first task, a CAD
model of the Stewart platform was developed in SolidWorks. Before building the CAD
model, workspace analysis was performed to measure various measures of manipulability
and its dependence on parameters such as radius of the upper platform and the length of
the actuators. These two parameters were then determined for attaining the best
workspace in the context of reproducing jaw motion. Then the inverse dynamics of the
Stewart platform was formulated based on the Lagrangian and Newton Euler Method.
The Jacobian matrix of the 6DOF parallel manipulator was derived as described in
Chapter 3. Initially to verify the inverse dynamic code, the inverse-forward dynamics
simulation was implemented using an S-Function block and a DynaFlexPro forward
dynamics model and the results were compared with the benchmark solutions for specific
case studies provided by Tsai. Feedback linearization was successfully implemented to
verify that the moving platform tracks the desired trajectory with the error converging to
zero. Then the motion capture analysis of different subjects including humans was
performed to determine the 3D trajectory of the front incisor on the mandible. This
trajectory was then used as the input trajectory for the inverse dynamic simulation of the
146
CAD model of the Stewart platform and the actuator forces were determined. Dynamic
simulations of the Stewart platform tracking jaw motion were performed in MSC. Visual
Nastran and Matlab to determine the actuator forces required for different subjects under
varying external/bite force. It was observed that the forces generated by the actuators
range between 100-200N for a human subject depending upon the external force acting
on the moving platform. Similarly the actuator forces increases up to 200-300N for dogs
and up to 400-500N for the sabertooth tiger depending upon the external force. This
study helped us to determine the size of actuators required for reproducing different jaw
motion trajectories. However, motion analysis of the subjects need to be performed at
different chewing frequencies to determine the forces, workspace and speed of the
actuators for different rates of chewing. The model that has been built and simulated
already would prove as a useful tool to perform various “what if” type of parametric
analysis before finalizing the specification of the Stewart platform. 3D CAD model of
skull and mandible of a bulldog and sabertooth was generated from high resolution CT
Scans using biomedical image processing techniques in MIMICS. Also the CAD model
of a bulldog was obtained from laser scanning a physical skull and mandible object of a
bulldog. Although the laser scanning method produced high quality STL files of skull and
mandible, it was very tedious working with the scanner and cleaning up the geometry in
Geomagics. From the STL files of skull and mandible obtained from these two methods,
SLA model of the skull and mandible were built using a SLA machine at Fisher-Price,
Inc. Finally, a dentition of the bulldog skull and mandible was created which could be
mounted on top of the Stewart platform with some fixtures to reproduce the jaw motion.
For achieving the second task, the biomechanics of the masticatory system was reviewed
to understand the function of each masticatory muscles and temporomandibular joints.
Literature from Koolstra et al. was reviewed to comprehend the modeling of TMJ. The
architectural properties of muscle such as muscle origin and insertion points, muscle fiber
length, maximum muscle force, tendon length etc. were extracted from Koolstra et al. for
human subject. As a first step, a musculoskeletal model of human jaw was built in
AnyBody Modeling system consisting of seven masticatory muscles on each side and the
motion capture date of front incisor point was used to drive the mandible. The condyles
147
were constrained to be in touch with the fossa and the sideways movement of one of the
TMJs was constrained. Muscle forces required to perform such desired jaw motion under
varying bite force were plotted. The modeling procedure was verified by comparing the
muscle activities plot with the one created by Mark De Zee et al. It was decided to model
the muscles as Hill type muscle models and model the temporalis muscles as set of three
vectors for producing more realistic results. The work was then extended to create a
musculoskeletal model of bulldog jaw and sabertooth cat jaw. In this case, the
architectural properties of the muscle were obtained from dissection of the dog head
muscles. Muscle mass were obtained from Turnbull book to determine the muscle
volume. STL files of the skull and mandible of both the animals were imported into
AnyBody for visualization and verification of muscle attachment points. Motion capture
data of the front incisor was used to drive the mandible and muscle forces and TMJ
reaction forces were determined for different bite force simulation. It was observed that
the muscle forces, TMJ reaction forces and muscle activities increases with increase in
bite force. Also the lateral pterygoid and the digastrics perform most of the work upfront
in opening the jaw, whereas the other masticatory muscles contribute little in terms of
muscle activities. In this way, a musculoskeletal model of the vertebrate jaw was built
and is designed to perform various other “what if” type parametric analyses to test
various hypothesis. Reproduction of the mastication motion of a vertebrate with a robotic
device enabled us to estimate muscle and bite forces required for different animals while
chewing/biting different regimen and relate them analyze masticatory muscle recruitment
patterns and performance and will be used to quantitatively evaluate the dynamic
breakdown of foods during chewing in the future, which is vital information required in
the development of new pet foods.
6.2 Future work
1. Sophisticated Dissection:
Dissection of the Masticatory muscles must be performed in a sophisticated
manner using 3D digitizer, laser diffraction techniques to measure the
architectural muscle parameters accurately. Also the dataset should be estimated
for specific subject for the sake of consistency.
148
2. Accurate Motion Capture Analysis:
A more sophisticated fixture should be designed to restrain the animal subject
from moving its head while performing mastication. This ensures consistent
motion capture data without loss of information in between the cycle.
3. Bite Force Measurement:
Instead of simulating chewing force, force transducers should be used to measure
the biting force while chewing different regimen. This force could then be applied
as the external force to the Stewart platform and musculoskeletal jaw model.
4. EMG measurements for validation:
Electromyography measurements of the masticatory muscles of animal subjects
must be measured during the mastication cycle. The actual muscle activities
obtained from virtual prototyping could then be compared with the experimental
EMG measurements to validate the musculoskeletal model in AnyBody.
5. Frequency of Mastication:
Since the frequency of chewing affects the speed at which actuators move or the
muscle forces required to move the mandible along desired trajectory, we need to
measure the motion capture data at different known frequencies based on the type
of regimen being chewed.
6. Physical Prototyping
Finally we need to build a 6DOF parallel manipulator and perform real time
control using Matlab Simulink xPC toolbox. The platform needs to be tested for
biting different regimens to measure the food breakdowns and masticatory
performance and forces in the actuators using force sensors.
149
Bibliography
1. M.J. Delsignore, “A Screw -Theoretic Framework for Musculoskeletal System
Modeling and Analysis”, Department of Mechanical & Aerospace Engineering.
Buffalo, State University of New York at Buffalo, Master of Science, 2005
2. K.S. Konakanchi, “Musculoskeletal modeling of Smilodon Fatalis for Virtual
Functional Performance Testing”, Department of Mechanical & Aerospace
Engineering. Buffalo, State University of New York at Buffalo, Master of
Science, 2005
3. M. Narayanan, S. Kannan, L-F. Lee, V. Krovi, and F. Mendel, “Virtual
Musculoskeletal Scenario-Testing Case-Studies”. Proceedings of 2008 Virtual
Rehabilitation, Vancouver, Canada, 2008.
4. G.G.Wang, "Definition and Review of Virtual Prototyping", Journal of
Computing and Information Science in Engineering 2(3): 232-236, 2002.
5. V. Krovi, V. Kumar, G.K. Ananthasuresh, and J.M. Vezien, "Design and Virtual
Prototyping of Rehabilitation Aids", ASME Journal of Mechanical Design
121(3): 456-458, 1999.
6. B. May, S. Saha, M. Saltzman, "A three dimensional mathematical model of