DEGREE PROJECT IN MEDICAL ENGINEERING, SECOND CYCLE, 30 CREDITS
STOCKHOLM, SWEDEN 2019
Improvements and Validation of THUMS Upper Extremity
Refinements of the Elbow Joint for Improved Biofidelity
KRISTÍN SVERRISDÓTTIR
KTH ROYAL INSTITUTE OF TECHNOLOGY
SCHOOL OF ENGINEERING SCIENCES IN CHEMISTRY,
BIOTECHNOLOGY AND HEALTH
| i
Abstract
Introduction One out of five reported motor vehicle collision injuries occur to the upper extremities. Certain parts of The Total HUman Model for Safety(THUMS) lack validation against experimental data, including the elbow. The aim of this project is to refine and validate the elbow joint of THUMS, with focus on anatomical response of the elbow during axial impact applied to the wrist.
Methods Internal contacts in the elbow were modified and new contacts assigned between bones and ligaments of the elbow. The posterior part of the radial- and ulnar collateral ligaments, and joint capsule was implemented to the model. Elastic modulus of the cortical bones of the elbow was increased as well as the shell thickness of the humeral cortical bone. The updated model was validated against an experiment where an axial load was applied to the wrist of a female cadaver.
The experimental resultant force in the wrist was then compared with the wrist force obtained from the simulations.
Results The correlation between the experimental and simulation resultant wrist force for the updated model resulted in a CORA score of 0.882. This gave a 6.7% higher CORA score compared with the original model. Hourglass energy was reduced from 63.52% of internal energy to 0.78%. Energy ratio and contact energies indicated that the simulation was stable.
Discussion Movement of elbow bones was assessed to be more anatomically correct, by accounting for the posterior ligament and elbow capsule support. The contact peak force in the humerus was lower and occurred earlier in the simulation in the updated model compared to the original. This is believed to be due to the reduced gap between the elbow bones after increasing the shell thickness of the humeral cortical bone. The model setup resembled the experiment in a good manner.
Conclusion The upper extremity of THUMS was refined for improved biofidelity, with focus on the anatomical response of the elbow joint under an axial impact. However, further model improvements are suggested as well as extended validated against other experimental impact results.
Keywords
Upper extremity, elbow, biofidelity, validation, Total HUman Model for Safety
(THUMS), Finite Element Modeling (FEM)
ii |
Sammanfattning
Introduktion En av fem rapporterade krockskador med motorfordon förekommer i de övre extremiteterna. Vissa strukturer hos Total HUman Model for Safety (THUMS) saknar validering gentemot experimentell data, där armbågen är ett av dem. Syftet med detta projekt är att förfina och validera armbågsleden hos THUMS, med fokus på dess anatomiska respons under axiellt islag applicerad på handleden.
Metod Interna kontakter i armbågen modifierades och nya kontakter tilldelades mellan ben och ligament. De posteriora delarna av kollateralligament hos radius och ulna implementerades i modellen, så även armbågens ledkapseln.
Elasticitetsmodulen hos de kortikala benen i armbågen höjdes och skalets tjocklek i det humerala kortikala benet utökades. Den uppdaterade modellen validerades mot ett experiment där en axiell belastning hade applicerats mot en kvinnlig kadavers handled. Den resulterande kraften i handleden från experimentet jämfördes sedan med erhållen kraft i handleden från simuleringarna.
Resultat Korrelationen mellan den experimentella kraften och simulerade kraften hos den uppdaterade modellen resulterade i ett CORA-poäng på 0,882. Detta är en ökning med 6,7% jämfört med den ursprungliga modellen. Hourglassenergin reducerades från 63,52% av inre energi till 0,78%. Energiförhållandet och kontaktenergier indikerade stabil simulering.
Diskussion Rörelsen av armbågens ben bedömdes vara mer anatomiskt korrekt, med hänsyn till stödet från de posteriora ligamentet och armbågens ledkapsel. Den maximala islagskraften i humerus minskade och uppträdde tidigare i simuleringen hos den uppdaterade modellen jämfört med originalet. Detta tros bero på reducerat avstånd mellan armbågens ben genom ökandet av skaltjockleken hos det humerala kortikala benet. Modelluppställningen motsvarade experimentets uppställning.
Konklusion De övre extremiteterna av THUMS förfinades i syfte att förbättra biofideliteten. Fokus låg på armbågens anatomiska respons under ett axiellt islag. Både ytterligare förbättringar av modellen och utökad validering mot andra experimentella islag rekommenderas.
Nyckelord
Övre extremitet, armbåge, biofidelitet, validering, Total HUman Model for Safety
(THUMS), Finite Element Modellering (FEM)
| iii
Acknowledgements
First and foremost, I would like express my gratitude to my supervisor, Madelen Fahlstedt for giving me the opportunity to work on this interesting and challenging project. Thank you for exceptional instruction and support through the course of my thesis work, for reviewing my texts, and for always keeping your door open for my questions.
I want to thank Pooya Sahandifar and Victor Alvarez who also supervised me on this project. Thank you Pooya for your time spent teaching me and providing valuable suggestions towards the progress of this project; and thank you Victor for your inputs and support on the project.
I also want to recognize my fellow master students within the Neuronics Department, Steinunn, Jia Cheng, Ekant, Sina, Nicole, and Beatrice, who provided a positive work environment. Particularly I would like to thank my group mates, Jiota, Aðalheiður, Rebekka, Cristina, and Simon, for valuable feedback and discussions. Special thanks to Jiota for helping me with translating my abstract to Swedish.
As this thesis marks my last work within my six years of engineering studies, I want to acknowledge my family. Mamma, pabbi, Vigdís, and Kristjón, thank you for being my closest support and motivation.
Stockholm, June 2019
Kristín Sverrisdóttir
iv |
Author
Kristín Sverrisdóttir, sverris@kth.se Medical Engineering
KTH Royal Institute of Technology
Supervisors
Madelen Fahlstedt, Neuronic Engineering, KTH Pooya Sahandifar, Neuronic Engineering, KTH Victor Alvarez, Autoliv
Reviewer
Svein Kleiven
Hälsovägen 11, 141 52 Huddinge Stockholm, Sweden
KTH Royal Institute of Technology
Examiner
Mats Nilsson
Hälsovägen 11, 141 52 Huddinge Stockholm, Sweden
KTH Royal Institute of Technology
CONTENTS | v
Contents
1 Introduction 1
2 Methods 3
2.1 Model Improvements . . . . 4
2.1.1 Internal Contacts . . . . 4
2.1.2 Modelling of Elbow Joint . . . . 5
2.1.3 Material Improvements . . . . 7
2.1.4 Element Quality . . . . 7
2.2 Validation of UE of THUMS . . . . 8
2.2.1 Experiment . . . . 8
2.2.2 THUMS Setup . . . . 9
2.3 Data Analysis . . . . 12
2.3.1 CORA . . . . 12
3 Results 13 3.1 Biofidelity of Model . . . . 13
3.2 Stability of Model . . . . 15
4 Discussion 16 4.1 Model Improvements . . . . 16
4.2 Experiment . . . . 17
4.3 Future Work . . . . 18
5 Conclusion 20
References 21
Appendix A - Literature Study 26
Appendix B 27
1. INTRODUCTION | 1
1 Introduction
Upper Extremity (UE) injuries resulting from motor vehicle collisions (MVCs) have received limited attention within traffic research [1]. More efforts have been put into investigating injuries with higher fatality, such as serious spinal cord injuries and brain injuries [2, 3]. Although UE injuries are usually not fatal, studies have shown that they can impact daily life with long-term impairment, which is associated with significant societal cost [4–8]. With improved occupant restraint systems in cars, especially airbags, number of fatal injuries has decreased.
However, number of injuries to the upper- and lower extremities has increased, and is predicted to continue raising in the future [4, 5, 9–11].
Several attributes of human surrogates are used within injury biomechanics research for accident reconstruction [12]. Examples include: human cadavers, human volunteers, animals, anthropomorphic test devices, and computational models. The surrogates have different advantages when approximating for response of a living human, and the injuries that various impacts can lead to. Simulating real-life crashes using computational models allows for analysis of factors which are considered too costly to investigate by conventional experimental approaches [12].
Human body model (HBM) constructed using finite element (FE) method is a computational model. HBMs provide information about both global- and local injury criteria, including injury prediction based on obtained stress and strain values [12]. Depending on model quality, HBMs can be computationally expensive, but can also provide the most accurate results of all computer models used in injury biomechanics. The accuracy of simulation results depends significantly on the quality of the model, meaning its biofidelity in terms of anatomy and material properties.
Total Human Body Model for Safety (THUMS) is a numerical HBM developed by Toyota Corporation and Toyota R&D Lab. THUMS is widely used in industry for car crash simulations, and is available for both pedestrian and vehicle occupant.
The original model represents a 50th percentile American adult male of 175 cm and 77 kg, height and weight respectively [13].
Validating FE HBMs is important to ensure that the model response shows
a reliable representation of a human body response. HBMs are generally
validated against data from cadaveric experiments, or from human volunteers
for pre-crash scenarios [14]. Certain body regions of the THUMS model have
been validated against experimental data, such as the thorax, abdomen, head,
2 | 1. INTRODUCTION
neck, and lower extremities [15]. Some validation work of the UE has been carried out, including a single bone validation of the radius, ulna and humerus and a validation of the forearm as a biomechanical system [16]. However, the impact response of the elbow joint of the UE has not been validated, to the best of the author’s knowledge.
Other limitations of the UE of the THUMS include extensive anatomical simplifications. Certain tissues are excluded which affects biofidelity of the model. Examples of those tissues include: soft tissues in hands, elbow joint capsule, and articular cartilage in joints. Additionally, mechanical properties assigned to biological tissues of THUMS are simplifications of real human properties. This leaves many aspects remaining for investigation, and can therefore be improved with more reliable values from published literature.
The aim of this project is to refine and validate the elbow joint of THUMS, with focus on anatomical response of the elbow during an axial impact loading applied to the wrist.
This will be achieved by investigating biofidelity of the model’s elbow joint, and
adapt the model for improvements. The UE model will then be validated against
data from a frontal impact applied on the wrist in a cadaveric experiment obtained
by Duma et al. [17].
2. METHODS | 3
2 Methods
Literature study was conducted to gain background knowledge of the thesis topic, covering the following topics: anatomy of the UE focusing on the elbow, mechanical properties of biological tissues, UE injuries from MVCs, experiments of UE impacts, FE method, and validation of simulations to experiments.
Summary of the literature study can be found in Appendix A.
Evaluation and updates were performed to the UE of THUMS AM50 Occupant SAFER v9.0.1, hereafter referred to as THUMS SAFER. The model consists of bones, ligaments, soft tissues, skin and muscles. For this study, only the passive contribution of the muscles of the forearm was included. Muscles of the upper arm were excluded from the model. Further information on the model can be found in section E.2 under Appendix A.
LS-PrePost v4.6 (Livermore Software Technology Corporation, Livermore, USA) was used for pre- and post processing, and an explicit dynamic solver in LS- Dyna v.9.7.1 (Livermore Software Technology Corporation, Livermore, USA) was used for running simulations [18].
(a) Complete UE (b) Flesh and skin of UE removed
Figure 2.1: UE of THUMS AM50 Occupant SAFER v9.0.1 (THUMS SAFER)
4 | 2. METHODS
2.1 Model Improvements
This section aims to describe modifications that were made to the UE of THUMS SAFER to gain improved biofidelity and model robustness. Focus was on the elbow joint, but several modifications were made to other parts of the UE.
2.1.1 Internal Contacts
Prior to model updates, there was extensive penetration between ligaments and bones in the elbow during an impact situation, see Figure 2.2. The model had two surface to surface contacts assigned for the elbow: one contact between the distal humerus and the proximal radius and ulna, and one contact between the annular ligament (AL), radial collateral ligament (RCL) and the distal humerus.
The contact between the bones was kept but the contact between the ligaments and the humerus was modified.
Figure 2.2: Lateral view of the radius (red) penetrating through AL (green) and RCL (blue).
Automatic surface to surface contact was generated between the cortical bones and ligaments in the elbow, in order to eliminate penetration. Distal end of the humerus and proximal ends of the ulna and radius acted as a slave with the four ligaments of the elbow, AL, RCL, quadrate ligament (QL), and ulnar collateral ligament (UCL) as master (Figure 2.3 (a)).
Tied contact was added between nodes on the AL and a segment set of the
radial neck (Figure 2.3 (b)). This was required as there was connection defined
2. METHODS | 5
between movement of the two in the model, although they are connected in the human elbow. The anatomical purpose of the AL is to restrain rotations of the radial head and to keep it in place adjacent to the ulna.
(a) Bones (yellow) and ligaments (red)
(b) AL (green nodes) and segment of radius (black boxes below)
Figure 2.3: Contacts added to the elbow joint of THUMS SAFER, (a) surface to surface contact, and (b) tied nodes to surface offset contact.
Existing contacts of the UE, as well as the surface to surface contact generated, were assigned soft constraint, segment-based contact option and search depth in automatic contact within the automatic surface to surface contact card. This was done to comply with the model requirements of THUMS SAFER [19]. See Table 2.1 for assigned contact setting.
Table 2.1: Updated contact card (card 9) for contacts of the UE in THUMS.
SOFT (constraint option) SBOPT (segment-based contact option) DEPTH (search depth in automatic contact)
2 (pinball segment based) 3 (warped segment check) 5 (checked for surface and edge-to-edge penetration)
2.1.2 Modelling of Elbow Joint
To reduce the gap between the elbow bones which is present in the model, shell
thickness was added to the cortical bones in the elbow. A parametric study was
performed to understand the influence on the model response. Resultant force
in the wrist, and the contact force in the cortical bone of the humerus varying shell
thicknesses were analyzed. Thickness values investigated were 1mm (original
thickness), 2mm, and 3mm. Shell thickness of 3mm was chosen. Dynamic
6 | 2. METHODS
and static friction coefficients were set to 0.1 as friction of articular cartilage in human synovial joints is reported to be very low under dynamic loading [20]. In order to account for shell reference surface offset in the contact treatment of the humerus, as the thickness was altered, offsets were set to be treated using shell thickness (CNTCO = 1). Reference surface of cortical bones of the elbow and the elbow ligaments was set to the bottom surface (NLOC = -1).
Elbow ligaments were remodelled to improve element quality. Attachment points of ligaments with surrounding bones were adjusted for improved stability. See Figures F.3 - F.6 under Appendix B for changes. The RCL and UCL are simplified in the model, resulting in lack of support for stability on the posterior side of the elbow [21]. Additionally, the elbow joint capsule is not modelled, but the capsule plays a role in stabilizing and strengthening the joint [22, 23]. The ligaments on the posterior side of the elbow were modeled with beam elements (Figure 2.4), and material properties found in Table 2.2. Same material properties were chosen as are assigned to the existing elbow ligaments, but limited studies have been published reporting mechanical properties of the elbow ligaments.
Although Smith et al. [24] reported elastic moduli for the anterior and posterior bundles of the UCL to be 13.77 MPa and 1.96 MPa respectively, the values currently used in the model for the elbow ligaments were kept to avoid inducing model instabilities.
Table 2.2: Material model and properties of beam elements with a combined function of elbow ligaments on the posterior side and the elbow capsule.
Part Material Model Young’s Modulus, E [GPa] Density, 𝜌 [kg/m
3] Poisson’s Ratio
Elbow Capsule ELASTIC 82.65 1000 0.22
Figure 2.4: Beam elements (pink) in the posterior side of the elbow joint representing
the function of the posterior aspects of RCL and UCL which is missing and the elbow
capsule. Beam elements thickened for better illustration.
2. METHODS | 7
2.1.3 Material Improvements
Mechanical properties of trabecular bone has been shown to vary greatly between anatomical regions [25], with most previous studies focusing on the femur and the tibia [26, 27]. However, Dunham et al. [28] conducted an experiment of mechanical properties of trabecular bone of the distal humerus and showed the elastic modulus to vary from 2.9 MPa to 1041.7 MPa with mean of 309.8 MPa. Elastic modulus of trabecular bone of humerus was changed from 40 MPa to 310 MPa. Due to lack of published experimental data, same values were chosen for the trabecular bone of the radius and ulna.
2.1.4 Element Quality
The following modifications were performed due to hourglass energy being outside of acceptable range. Element formulation of the flesh of the lower arm, and the spongy parts of the proximal and center ulna and radius, and distal humerus were changed from constant stress solid element (ELFORM 1) to fully integrated solid (ELFORM 2).
Element formulation of ligaments in hand and wrist was changed from Belytschko-Tsay shells (ELFORM 2) to fully interated Belytschko-Tsay membranes (ELFORM 9). Elbow ligaments and the interosseous membrane were changed from Belytschko-Tsay membranes (ELFROM 5) to the same as the hand and wrist ligaments (ELFORM 9).
Hourglass control type of all solids and shells was changed from Flanagan- Belytschko viscous form (IHQ 2) to the same control type with with exact volume integration for solid elements (IHQ 3), and to a configuration which activates the full projection warping stiffness for shell elements (IHQ 8).
The model checker tool within the student edition of HyperWorks v.2017.2 (Altair
Engineering Inc, Troy, USA) was used to check for element quality both before
and after model improvements.
8 | 2. METHODS
2.2 Validation of UE of THUMS
2.2.1 Experiment
To validate the UE of THUMS, experiment by Duma et al. [17] was replicated.
Axial load was applied across the wrist joint using a pneumatic impactor, and the resultant force in the wrist was measured. The test was configured in a way to resemble a loading situation of a side airbag deployment, see Figure 2.5.
Figure 2.5: Lateral view of the test configuration. Wrist and elbow joints are forced into compression by axial load applied when a pneumatic impactor (not shown) strikes the transfer piston. Published from Duma et al. [1] with permission from authors.
The UE was dissected at the mid-shaft of the humerus and supported by two cables which held the elbow in 90° flexion. The forearm was kept intact to preserve the load distribution provided by the interosseous ligament which connects the ulna and the radius longitudinally. The handgrip assembly was constructed of aluminum in a manner to provide a narrow and rigid contact surface with the wrist only. This position was chosen to minimize applied moment, and to resemble what was considered the most vulnerable position of the wrist in a side airbag deployment situation. No pre-load was applied to the wrist, and the wrist was lightly fixed to the handgrip assembly, allowing it to translate away from it after impact. Data was filtered using channel frequency class (CFC) 180 filter.
The experiment was performed on 19 female cadavers. The experimental
results (Figure 2.6) used in this study were from a UE of a female cadaver of
weight 55 kg [17]. Section 2.2.2 explains how the load curve was used in the
2. METHODS | 9
simulation and correlation analysis. See section D under Appendix A for further information on the experiment.
Figure 2.6: Axial force measured in wrist of female cadaver, used in this study for comparison of simulated results [17].
2.2.2 THUMS Setup
First, the UE of the THUMS was isolated from the full-body model. Positioning was performed to attain a 90° flexion of the elbow (see Figure 2.7). This was achieved by simulating a pull onto the hand outwards which aligned the radius in 90° with the humerus.
(a) Angle: 98.5° (b) Angle: 90°
Figure 2.7: Angle between humerus and radius before and after positioning of model.
All other parts of the model are hidden for better visualization of the angle between the
bones.
10 | 2. METHODS
Hand was positioned by applying a load to the palm and fingers to allow for a direct impact to the wrist. All parts besides the hand were kept rigid to avoid any distortion of the arm. Nodal position of the hand and wrist which corresponded to the experimental setup were then used for the validation simulations.
Two tension-only cables of length two meters each and properties of steel, were attached to the forearm and the upper arm [29]. The cable that supported the lower arm was constrained to a node on the surface of the lower-arm band which was made from a set of nodes on the lower arm replicating the cuff from the experiment, see Figure 2.5. The cable supporting the upper arm was constrained to a node in the center of the humerus, but the whole upper arm was made rigid, as the upper arm had been dissected in the experiment at the mid-shaft of the humerus. Both cables were constrained using a rigid body constraint.
The impactor was modeled as a rigid rectangular plate with material properties of aluminum. The impactor was constrained to only allow for translational movement in the negative x-direction. Material properties of impactor and cables can be seen in Table 2.3 and the replicated experimental setup can be seen in Figure 2.8.
Table 2.3: Material model and properties of impactor and cables used in simulation.
Part Material Model Young’s Modulus, E [GPa] Density, 𝜌 [kg/m
3] Poisson’s Ratio
Impactor RIGID 69 2700 0.22
Cables CABLE DISCRETE BEAM 200 7800 N/A
2. METHODS | 11
Figure 2.8: Medial view of the replicated experimental setup. Green arrow indicates applied axial rigid body load applied in negative x-direction. Cables are shortened and thickened for illustration.
The load curve (Figure 2.6) was applied as a rigid body load to the wrist and
the impactor was put in contact with the cortical part of the wrist bones (Figure
2.9) only using automatic surface to surface contact. The resultant force in
the wrist was measured in a cross-section in the distal ulna and radius, see
Figure 2.9. Contact force in the distal humerus with the proximal ulna and radius
was also analysed in conjunction with the wrist force, to investigate difference
between the response of the original and the updated model. Data obtained
from simulations was filtered with channel frequency class (CFC) 180 filter using
Matlab (MathWorks, Natick, USA).
12 | 2. METHODS
Figure 2.9: Cross section of distal ulna and radius (black arrows) where resultant wrist force is measured from simulations. Wrist bones are shown as well (red arrows), where the load is applied.
2.3 Data Analysis
2.3.1 CORA
CORrelation and Analysis (CORA) is a method which provides an objective overall evaluation of correlation between whole response curves obtained from experiments and simulations [30]. CORA consists of set of algorithms which combine two independent sub-methods: a corridor rating and a cross-correlation rating. The corridor rating evaluates the fitting of a curve into user-defined or automatically calculated corridors, and the cross-correlation rating evaluates phase-shift, shape and area below curves. Rating of results ranges from 0 to 1, where 0 shows no correlation and 1 indicates a perfect match. See Section F.2 for further information on CORA.
For this thesis, CORA v4.0.4 (PBD - Partnership for Dummy Technology and
Biomechanics, Gaimersheim, Germany) was used to compare the simulation
results for resultant force in the wrist to the experimental results. The cross-
correlation method was used, meaning that the phase-shift, shape, and area
below curves was evaluated. The parameteres were evaluated with set weight
factors of 0.5 for the shape of the curve, and 0.25 for both phase-shift and area
below the curve.
3. RESULTS | 13
3 Results
3.1 Biofidelity of Model
Difference in peak of resultant force measured in the wrist in the simulation of the updated model compared with the original was 280 N. Difference between the updated model and the results from the cadaveric experiment was 290 N. The load curves obtained from simulations and experiment can be seen in Figure 3.1.
Figure 3.1: Experimental results of resultant force at the wrist compared to simulated results using original model and updated model.
CORA rating of the curves of wrist axial force of the experiment and the simulation before and after model updates can be found in Table 3.1.
Table 3.1: CORA rating of the resultant wrist force obtained from experiment and simulations, using the cross-correlation method.
Original Model Updated Model Improvement
CORA Rating 0.827 0.882 6.7%
14 | 3. RESULTS
Biofidelic assessment of the response of the elbow bones under impact was performed by analysing the movement of the bones in relation to one another.
The medial view of the elbow bones of the original and updated model 20 ms after initial impact can be seen in Figure 3.2. The posterior ligament support which has been modeled as beam elements can be seen in Figure 3.2 (b).
(a) Original model. (b) Updated model.
Figure 3.2: Position of elbow bones, ulna (purple), radius (red), and humerus (grey) 20 ms after initial impact. Elbow ligaments, flesh and skin are hidden in the images.
Resultant force of the humerus generated in the contact with the proximal ulna and radius from simulations is presented in Figure 3.3.
Figure 3.3: Resultant force in cortical bone of humerus before and after model
modifications.
3. RESULTS | 15
3.2 Stability of Model
Positioning of the model showed more stable results after performing the model updates. That can be seen by comparing the penetrations of ligaments and bones in the elbow in the last state of the positioning simulation, see Figure 3.4.
(a) Original model. (b) Updated model.
Figure 3.4: Position of the elbow after positioning of the arm. Yellow circle marks penetration and the red circles mark loose nodes.
Simulation the experiment for both original and updated model model had no negative hourglass energy. Maximum hourglass- versus internal energy ratio, as well as minimum and maximum values for energy ratio, sliding energy (total contact energy) and total mass can be seen in Table 3.2. Requirements are stated according to the THUMS SAFER requirement document [19]. The total contact energy in defined contacts for ligament versus bone was positive.
Table 3.2: Stability and computational parameters before and after model updates.
Requirements are according to reported values for THUMS SAFER [19].
Parameter Requirement Original Model Updated Model
Energy ratio (min) 0.95 0.987 0.987
Energy ratio (max) 1.05 1 1
HG energy/Int energy <10 % 63.52 % 0.78 %
Total sliding energy N/A 1.21 0.90
Added mass <5 % 1.56 ⋅ 10 −4 % 1.67 ⋅ 10 −4 %
Smallest timestep N/A 1.06 ⋅ 10 −3 ms 5.79 ⋅ 10 −4 ms
Computational time N/A 68 sec (1 min, 8 sec) 116 sec (1 min, 58 sec)
16 | 4. DISCUSSION
4 Discussion
4.1 Model Improvements
Penetrations in elbow were eliminated by applying contact between bones and ligaments and modifying existing contacts, explained in Chapter 2.1.1. This was evident in both positioning of the UE and in the simulated experiment.
Although number of internal contact definitions should be kept to minimum, it was necessary to increase the number of contacts by one. This was due to the ignored relationship between the annular ligament and the radial head. The function of this ligament is to limit dislocations of the radial head but still allow for rotation of the radius during pronation and supination [31], and therefore an offset contact was defined.
The absence of the posterior part of UCL and RCL in THUMS, in combination with exclusion of the elbow joint capsule caused an unrealistic gap between the elbow bones under impact (see Figure 3.2. Beam elements representing the posterior parts of the RCL and UCL as well as the elbow capsule were added for posterior elbow stability. The elbow capsule has been reported to have similar mechanical structure to the glenohumeral joint of the shoulder, apart of the elbow joint having less capability to stretch [32]. The elastic modulus of the glenohumeral joint has been shown to vary from 28.4 MPa to 56.8 MPa depending on region [33]. Therefore the currently assigned properties for the elbow ligaments which have an elastic modulus of 82.65 MPa was chosen also for the beam elements.
Remodeling of ligaments was required due to their unstable response. The ligaments had nodes which were not connected to nodes of the surrounding tissue, leading to behavior of the ligaments which is anatomically incorrect.
Increased shell thickness of the cortical bone of the distal humerus from 1mm to 3mm served as a compensation to some extent to the absence of cartilage in the elbow joint of the THUMS.
Large difference in stiffness of cortical and trabecular bone of the of the proximal
ulna and radius, and distal humerus, as well as their poor element quality (see
warpage and aspect ratio values under Table F.2 under Appendix B caused
instability in the model. Elastic modulus of cortical parts of the same bones
is assigned with 17.6 GPa, and that difference resulted in negative volumes of
the trabecular bone under large deformations. THUMS SAFER has advantage
of being low in computational cost and is therefore a desired option to use in
4. DISCUSSION | 17
industry, although there are FE HBMs including THUMS v4 and GHBMC, used in research with finer mesh [14].
The element formulation of the flesh of the lower arm was changed from constant stress solid element to a fully integrated solid. This along with exact volume integration hourglass control type eliminated hourglass energy issues with the model. However, the computational time of the simulation was increased from 68 seconds to 116 seconds, or 70.6%.
4.2 Experiment
The experiment by Duma et al. [17] was replicated to validate the model response and to compare the response of the THUMS SAFER before and after modifications were implemented.
According to the information provided about the experiment, the simulation set- up resembled the experiment in a good manner. During the simulation, the impactor did not lose contact with the wrist until at the very end of the simulation, around 25 ms after impact. This corresponds well with the experiment where the handgrip held until the end of the impact. This indicates that the force and momentum were transferred from the wrist trough the radius, interosseous membrane, and ulna in the forearm, as intended. This was confirmed by measuring the force at the impact location and at the distal end of the forearm, but the same tendency of the resultant force was observed.
The resultant force measured in the wrist in the updated THUMS SAFER compared with the original model showed improvement in correlation with the simulated results, or a CORA rating of 0.882 compared to 0.827. This showed an improvement of 6.7%, although both correlations were within the limits of the biofidelity loadcase requirements for the THUMS SAFER of normal termination and CORA rating of 0.7 [19].
When comparing the tendencies of the resultant wrist force curves of the original
and updated models, a few interpretations can be made. The wrist force (Figure
3.1) was analyzed state by state in conjunction with the contact force of the
distal humerus with the proximal ulna and radius (Figure 3.3). A correlation was
observed between the contact force in the elbow and the resultant force in the
wrist. The force in the humerus of the original model showed one distinct peak of
around 1250 N at 12 ms, while the updated model showed two peaks, first one
at 630 N and the second one at 800 N. Another major difference between the
original and updated model is the trend from 20-25 ms. This difference can be
18 | 4. DISCUSSION
explained by looking at Figure 3.2. The implemented posterior constraint forces the humerus to stay in contact with the ulna as the joint is moving as a unit, as if it would be encapsulated. However, in the original model the humerus separated from the ulna under the impact loading. By understanding the behavior of the elbow joint under the impact, the resultant force in the wrist can be explained having the direct load translation from the wrist through the longitudinal section of radius and ulna towards the elbow joint. The original wrist force curve showed two peaks close to each other in magnitude, but they occurred directly following to the peak force in the humerus. Same accounts for the two peaks of the wrist force for the updated model. The contact in the elbow is also showed to occur around 3 ms earlier compared to original model which is anticipated to be due to the increased shell thickness of the bone. The response of the updated model is claimed to be more biofidelic as the elbow joint is enclosed within a capsule.
The experiment used has its limitations. The load curve used was chosen due to the peak force of the resultant force in the wrist being within the range of likely resulting in a fracture of the wrist. Duma et al. [28] reported 50% chance of a wrist fracture at an axial load of 1700 N. This is a limitation to this validation, but only three resultant wrist curves were presented from the experiment and the peak force values of the other two were considered too low severity or too high severity for an impact of interest to analyze. In addition, the velocity of the impact was not identified. Therefore, the load curve measured in the wrist, reported in the experiment was implemented as a rigid body load in the simulation.
The experiment was performed on 5th percentile female, using results from a 55kg female while THUMS SAFER represents a 50th percentile male of 76 kg.
By scaling the load curve applied by a factor of 1.2, equivalent peak force to the experiment was obtained. By using anthropometric data, the mass of the UE was estimated and turned out to be close to the THUMS SAFER UE, or both around 3.2 kg.
4.3 Future Work
The shell thickness implemented closes the open spaces in the elbow and
compensates for the absence of cartilage to a certain degree. However, it does
not account for mechanical properties of cartilage. By accounting for damping,
contact mechanics of the elbow joint is anticipated to be improved, resulting in
a more biofidelic impact response.
4. DISCUSSION | 19
Nevertheless, by adding detailed modelling to THUMS SAFER which is modelled with a fairly course mesh, the computational time would be expected to increase. Therefore, the trade-off between biofidelity versus computational time and stability has to be evaluated for each model adjustment.
The wrist of the THUMS SAFER requires modelling improvements and contact definitions, as it currently has penetrations and low stability of ligaments.
Element quality of trabecular bones of the UE was observed to have low quality index values which did not meet the criteria for the model, see Table F.2 under Appendix B.
More comprehensive study is required to confirm the biofidelity of the elbow
joint of THUMS SAFER. But as mentioned prior in this report, only axial load
applied directly to the wrist and reported in the study used. The study aimed to
apply a load across the wrist joint similar to what occurred in the side airbag
test deployment performed. However, a car crash situation could result in
different kinds of impacts than the one studied. Duma et al. [34] performed
a drop test impacting the elbow, forcing it into hyperextension. This study is
suggested to further investigate the response of the elbow and the UE of THUMS
SAFER. The study reports the peak elbow moment for the test population of
24 cadaveric arms, and plots for the moment measured over 60 ms for two
cadaveric specimen. Dynamic hyperextension injury criteria for the female
elbow joint was developed.
20 | 5. CONCLUSION
5 Conclusion
The aim of this study was to assess the UE of the THUMS SAFER and update
the model according to its weaknesses in terms of biofidelity and stability. This
was achieved with modelling adjustments to the ligaments of the elbow joint, by
adjusting contact parameters and other model parameters such as hourglass
control, element formulation and shell thickness. The ratio of hourglass energy
compared with internal energy was significantly improved, however the energy
ratio remained the same. Implementing model changes which resulted in both
biofidelic improvements without sacrificing the intented function and stability
measures of the model was a challenge. The wrist resultant force measured
in the simulations compared to the experiment used showed better correlation
after implementing model changes and gave a CORA rating of 0.884 compared
to 0.838 using the original model. To conclude, the biofidelity of the elbow
joint of THUMS SAFER has been improved, although there are yet further
improvements that can be made to the model, as suggested in Section 4.3.
REFERENCES | 21
References
[1] Mark Chong et al. “Upper extremity injuries in restrained front-seat occupants after motor vehicle crashes”. In: Journal of Trauma and Acute Care Surgery 70.4 (2011), pp. 838–844.
[2] MEL Van den Berg et al. “Incidence of spinal cord injury worldwide: a systematic review”. In: Neuroepidemiology 34.3 (2010), pp. 184–192.
[3] Mark J Swierzewski et al. “Deaths from motor vehicle crashes: patterns of injury in restrained and unrestrained victims.” In: The Journal of trauma 37.3 (1994), pp. 404–407.
[4] Kai-Uwe Schmitt et al. Trauma Biomechanics: Accidental injury in traffic and sports. Springer Science & Business Media, 2010.
[5] Carol Conroy et al. “Upper extremity fracture patterns following motor vehicle crashes differ for drivers and passengers”. In: Injury 38.3 (2007), pp. 350–357.
[6] Anand Hammad et al. “Development of Upper Extremity Finite Element Model for Elderly Female: Validated Against Dynamic Loading Conditions”. In: ASME 2017 International Mechanical Engineering Congress and Exposition. American Society of Mechanical Engineers.
2017, V012T16A006–V012T16A006.
[7] Robert Kaufman et al. “Severe soft tissue injuries of the upper extremity in motor vehicle crashes involving partial ejection: the protective role of side curtain airbags”. In: Accident Analysis & Prevention 102 (2017), pp. 144–
152.
22 | REFERENCES
[8] King-Hay Yang. Basic Finite Element Method as Applied to Injury Biomechanics. Academic Press, 2017.
[9] Matthew W Goldman et al. “The association between restraint system and upper extremity injury after motor vehicle collisions”. In: Journal of orthopaedic trauma 19.8 (2005), pp. 529–534.
[10] Gerald McGwin Jr et al. “Association between upper extremity injuries and side airbag availability”. In: Journal of Trauma and Acute Care Surgery 64.5 (2008), pp. 1297–1301.
[11] Peter G Martin, Jeff R Crandall, and Walter D Pilkey. “Injury trends of passenger car drivers in frontal crashes in the USA”. In: Accident Analysis
& Prevention 32.4 (2000), pp. 541–557.
[12] Jeffrey Richard Crandall et al. “Human surrogates for injury biomechanics research”. In: Clinical anatomy 24.3 (2011), pp. 362–371.
[13] Dirk Fressmann et al. “FE Human Modelling in Crash–Aspects of the numerical Modelling and current Applications in the Automotive Industry”.
In: LSDYNA Anwenderforum, DYNAmore GmbH, Frankenthal, Germany, pp. FI-23–FI-34 (2007).
[14] Erik Eliasson and Jacob Wass. “Industrialisation of a finite element active human body model for vehicle crash simulations”. PhD thesis. Master’s thesis). Chalmers University of Technology, 2015.
[15] Masami Iwamoto, Yuko Nakahira, and Hideyuki Kimpara. “Development
and validation of the total human model for safety (THUMS) toward further
understanding of occupant injury mechanisms in precrash and during
crash”. In: Traffic injury prevention 16.sup1 (2015), S36–S48.
REFERENCES | 23
[16] Bengt Pipkorn and Lukas Zehnpfennig. “Upper Extremity Injury THUMS Validation”. In: Unpublished Material (2017).
[17] Stefan M Duma et al. “Injury risk function for the small female wrist in axial loading”. In: Accident Analysis & Prevention 35.6 (2003), pp. 869–875.
[18] Dr Brian Gladman. LS-DYNA KEYWORD USER’S MANUAL, VOLUME I. LIVERMORE SOFTWARE TECHNOLOGY CORPORATION (LSTC), 2019.
[19] Bengt Pipkorn. SAFER THUMS Model Numerical, Robustness, Biofidelity
& User Requirements. Tech. rep. Jan. 2019.
[20] R Bruce Martin et al. Skeletal tissue mechanics. Vol. 190. Springer, 1998.
[21] Anne M Gilroy, Brian R MacPherson, and Lawrence M Ross. Atlas of Anatomy. Thieme Stuttgart, 20012.
[22] Jeroen de Haan et al. “Stability of the elbow joint: relevant anatomy and clinical implications of in vitro biomechanical studies”. In: The Open Orthopaedics Journal 5 (2011), p. 168.
[23] Bernard Morrey and Kai-Nan An. “Articular and ligamentous contributions to the stability of the elbow joint”. In: The American journal of sports medicine 11.5 (1983), pp. 315–319.
[24] Matthew V Smith et al. “Mechanical Properties and Microstructural
Collagen Alignment of the Ulnar Collateral Ligament During Dynamic
Loading”. In: The American journal of sports medicine 47.1 (2019),
pp. 151–157.
24 | REFERENCES
[25] Steven A Goldstein. “The mechanical properties of trabecular bone:
dependence on anatomic location and function”. In: Journal of biomechanics 20.11-12 (1987), pp. 1055–1061.
[26] Dennis R Carter and Wilson C Hayes. “The compressive behavior of bone as a two-phase porous structure.” In: The Journal of bone and joint surgery. American volume 59.7 (1977), pp. 954–962.
[27] CM Schoenfeld, EP Lautenschlager, and PR Meyer. “Mechanical properties of human cancellous bone in the femoral head”. In: Medical and biological engineering 12.3 (1974), pp. 313–317.
[28] Cheryl E Dunham et al. “Mechanical properties of cancellous bone of the distal humerus”. In: Clinical Biomechanics 20.8 (2005), pp. 834–838.
[29] Dr Brian Gladman. LS-DYNA KEYWORD USER’S MANUAL, VOLUME II, Material Models. LIVERMORE SOFTWARE TECHNOLOGY CORPORATION (LSTC), 2019.
[30] Christian Gehre, Heinrich Gades, and Philipp Wernicke. “Objective rating of signals using test and simulation responses”. In: Proceedings:
International Technical Conference on the Enhanced Safety of Vehicles.
Vol. 2009. National Highway Traffic Safety Administration. 2009.
[31] Stefan Fornalski, Ranjan Gupta, and Thay Q Lee. “Anatomy and biomechanics of the elbow joint”. In: Sports medicine and arthroscopy review 11.1 (2003), pp. 1–9.
[32] DS Kaltsas. “Comparative study of the properties of the shoulder joint
capsule with those of other joint capsules.” In: Clinical orthopaedics and
related research 173 (1983), pp. 20–26.
REFERENCES | 25
[33] Michael J Bey et al. “Structural and mechanical properties of the glenohumeral joint posterior capsule”. In: Journal of shoulder and elbow surgery 14.2 (2005), pp. 201–206.
[34] Stefan M Duma et al. Upper extremity interaction with a helicopter side airbag: injury criteria for dynamic hyperextension of the female elbow joint.
Tech. rep. SAE Technical Paper, 2004.
26 | REFERENCES
Appendix A - Literature Study
REFERENCES | 27
Contents - Literature Study
A Biomechanically Relevant Anatomy 28
A.1 Upper Extremity (UE) . . . . 28 A.2 Elbow Joint . . . . 29 A.2.1 Bones and Articulations of the Elbow . . . . 29 A.2.2 Ligaments and Stabilizers of the Elbow . . . . 30 B Mechanical Properties of Biological Tissues 32 B.1 Cartilage . . . . 32 B.2 Ligament . . . . 33 B.3 Bone . . . . 35
C UE Injuries from MVCs 37
C.1 Injury Trend . . . . 37 C.2 Prevalence and Types of Injuries . . . . 37
D Experiments of UE Impacts in MVCs 39
E Finite Element Method (FEM) 41
E.1 FEM in Traffic Safety Research . . . . 41 E.2 Total Human Model of Safety (THUMS) . . . . 41
F Validation of Simulations to Experiments 46
F.1 Previous Validation of THUMS SAFER UE . . . . 46 F.1.1 Forearm . . . . 46 F.1.2 Humerus . . . . 47 F.1.3 Clavicle . . . . 48 F.2 Correlation and Analysis (CORA) . . . . 48
References 49
Appendix B 57
CONTENTS - LITERATURE STUDY