Human body modelling IRCOBI Asia, Lonavala, India, April 27, 2018. Prof. Karin Brolin Assoc. Prof. Johan Davidsson Chalmers University of Technology, Gothenburg, Sweden
What is a model? All models are wrong, but some are still useful. George Box, University of Wisconsin-Madison A simulation is reliable if the differences between numerical and test results are small enough to lead to the same design conclusions Paul du Bois, independent consultant in the automotive industry It is important to verify and validate models: Are the mathematical equations correctly implemented? (Verify) Are results close to experimental observations? (Validate)
Two kinds of numerical models of occupants: 1. Crash test dummy models 2. Human body models
Other crash FE models Barrier models FTSS LSTC Moose model Dyamore
Don t make the model more complicated than what is required for its application(s)!
Human Body Models (HBM) Multi Body Dynamics Finite Element Method
Dynamic simulation techniques Multi Body Dynamics Finite Element Method Experiments (MBD) (FEM) High degree of approximation Very low cost Low degree of approximation Low cost Very low degree of approximation High cost When does it come in the product cycle? Early Early - late Late
Multi body dynamics Pedestrian truck impact Courtesy of Prof. S. Mukherjee, Indian Institute of Technology in Delhi, India.
Multi body dynamics Pedestrian 3-wheeler impact Courtesy of Prof. S. Mukherjee, Indian Institute of Technology in Delhi, India.
Finite element method Vehicle - infrastructure
Finite element method Occupant restraint interaction Erik Eliasson and Jacob Wass (M.Sc. Thesis at Chalmers in collaboration with Autoliv Research).
Finite element method Spinal response in rear end impact Östh J et al. The VIVA OpenHBM Finite Element 50th Percentile Female Occupant Model: Whole Body Model Development and Kinematic Validation. IRCOBI 2017, IRC-17-60.
Full Body FE Models THUMS HUMOS GHBMC Model 50 th percentile male 175 cm, 78 kg
HOW ABOUT THE REST OF THE POPULATION?
Recent progress (some examples) THUMS child models, 2016 https://newsroom.toyota.co.jp/en/detail/12497148 PIPER child models, 2017 & morphing tool http://piper-project.org/downloads ViVA 50 th female model, 2016 https://www.chalmers.se/en/projects/pages/openhbm.aspx Obese HBM Hwang E. et al. Stapp Car Crash Journal, Vol. 60 (November 2016), pp. 473-508.
Injury prediction with HBMs Example from the SAFER-project: Improved injury prediction using HBM.
Model and methods HBM: Total Human Model for Safety (THUMS) v. 3.0 Finite Element Code: LS-DYNA v. 970/971 Pre- and post-processing: Hyperworks LS-PREPOST MatLab Primer Toyota Motor Corporation, Toyota Central Labs Inc. 2008. Users' Guide of Computational Human Model THUMS AM50 Occupant Model: Version 3.0 080225, LSTC Inc., Livermore, CA, USA, Altair Engineering Inc., Mich. USA, The Mathworks Inc., Natick, MA, USA, Oasys Ltd., Sollihull, UK
TUNING The material parameters are changed / optimized so that simulation results fit a set of experimental data. VALIDATION Material parameters are chosen (or tuned) from one set of experimental data and the response of the model is compared to another set of experimental data.
Validate the model at the loading conditions it is intended for Local load Linear loads Combination of loads Distributed load Validation data envelope Out-ofposition loads Future occupant postures Low and high rate loads
Thorax validation: tissue level Stress - strain Cortical bone specimen Experimental data from Kemper et al. 2005.
TABLE 1 RIB STIFFNESS [N/mm] Rib level Modified SIMULATION THUMS EXPERIMENT Tests [13]* 2 7.9 7.7 ± 3.6 3 2.0 2.1 ± 0.9 4 1.2 2.2 ± 0.9 5 1.2 2.0 ± 0.6 6 1.3 2.8 ± 1.2 7 2.2 3.1 ± 1.1 8 2.0 2.7 ± 1.0 9 1.5 2.5 ± 1.0 10 1.5 2.8 ± 1.0 *Mean ± one standard deviation Thorax validation: component level Fig. 1. Single rib test configuration Experimental data from: Li Z et al., Rib fractures under anterior-posterior dynamic loads: experimental and finite-element study, J Biomech, 43, 2, 228-234, 2009. Kindig M, Tolerance to failure and geometric influenced on the stiffness of human ribs under anterior-posterior loading, University of Virginia, School of Engineering and Applied Science, 2009.
Thorax validation: body part Force - compression Kroell pendulum impact
Thorax validation: body part Experimental data from Kent et al (2005)
Thorax validation: body part Stiffness and deformation
Thorax validation: whole body Kinematics 40 km/h Experimental data from Shaw et al (2009)
Construction of AIS 2+ injury risk curves 23 PMHS tests Shear stress First principal strain Mendoza-Vazquez M, Davidsson J, Brolin K. Construction and evaluation of thoracic injury risk curves for a finite element human body model in frontal car crashes. Accident Analysis and Prevention. 2015;85:73-82.
Injury risk curve construction Test 1 Test 2 Risk of injury 1 Injury criterion Simulation 1: Simulation 2:
Mendoza-Vazquez M et al. Evaluation of Thoracic Injury Criteria for THUMS Finite Element Human Body Model Using Real-World Crash Data. I: IRCOBI Conference Proceedings - International Research Council on the Biomechanics of Injury, 10-12 September, Berlin, Germany. 2014. 528-541. Real-world data comparison Volvo Cars' Traffic Statistical Accident Database 95% conf. int.
Real-world data comparison Shear stress Mendoza-Vazquez M et al. Evaluation of Thoracic Injury Criteria for THUMS Finite Element Human Body Model Using Real-World Crash Data. IRCOBI Conference Proceedings, 10-12 September, Berlin, Germany. 2014. 528-541.
Real-world data comparison ANOTHER STUDY w. THUMS Accident reconstructions (NASS/CDS) Stochastic simulations to account for population and case variability. Iraeus J, Lindquist M, Development and validation of a generic finite element vehicle buck model for the analysis of driver rib fractures in real life nearside oblique frontal crashes, Accident Analysis and Prevention. 2016; 95 Part A:42-56. NASS/CDS Real world data FE Human Body Model simulations
We over predicted injuries compared to accident data. Validation for kinematics and stiffness. Injury prediction using strain. How can we improve the biofidelity of injury predictions? Rib strain validation Additional model detail
Improve injury prediction Hypothesis Rib fracture is strain controlled Method Develop a generic rib cage Validate rib strain as indicator of rib fracture Evaluate how PMHS results relates to real world accident data Maximum strain absolute location [%] R2=0.91 Rib fracture absolute location [%] Trosseille, X., Baudrit, P., Leport, T., Vallancien, G., (2008). Rib cage strain pattern as a function of chest loading configuration. Stapp car crash journal 52, 205-231.
Development of a generic ribcage Based on anatomical data on rib cross sections i, cortical bone thickness ii, curvature and twist ii. i. Choi, H.-Y., Kwak, D.-S., (2011). Morphologic characteristics of korean elderly rib. J Automotive Safety and Energy, 2011, Vol. 2 No. 2 ii. Shi, X., Cao, L., Reed, M.P., Rupp, J.D., Hoff, C.N., Hu, J., (2014). A statistical human rib cage geometry model accounting for variations by age, sex, stature and body mass index. Journal of biomechanics 47 (10), 2277-2285.
Experimental data for strain validation Single Rib Tests Kindig (2009) Tested 94 ribs (2-10) from 9 subjects Subjects aged 31-63 years Loading speed ZZ(t) was 1 m/s Rib ends constrained in all DOFs but Z translation Unpublished OSU (2016) Tested 222 ribs (4-7) from 175 subjects Subjects aged 0-100 years Loading speed ZZ(t) was either 1 or 2 m/s Rib ends constrained in all DOFs but Z translation Strain gauge
Experimental data for strain validation - Three physical test series Table Top and Impactor Tests Shaw (2008) Trosseille (2008) Kent (2008) Tested 5 male subjects 16 strain gages Indentor loading speed was 1 m/s 3 indentor locations (low, mid, upper) Tested 11 subjects (10M +1 F) 96 strain gages Impactor speed 4.2-6.5 m/s 3 impactor directions (frontal 0, oblique 60, and lateral 90 ) Tested 3 subjects (M +2 F) Impactor speed 1 m/s Three tissue states (intact, denuded and eviscerated) Four load conditions
Experimental data for strain validation - Three physical sled test series Sled Tests Forman (2006) Sled based on a Ford Taurus MY 1999 Tested 3 male subjects (39-49 y) ~50%ile male Delta Velocity 30km/h Standard 3p belt system, no LL or pretension Shaw (2009) Crandall (2011) Test setup referred to as Gold standard 1 Tested 8 male subjects (avg 54 y) ~50%ile male Delta Velocity 40km/h Standard 3p belt system, no LL or pretension Test setup referred to as Gold standard 2 Tested 2 male subjects (59 and 66 y) ~50%ile male Delta Velocity 30km/h Load limited belt system, F=2.9kN (at shoulder) No rib fractures All got rib fractures No rib fractures
Rib strain validation 1. Single rib tests 2a. Table top tests 2b. Impactor tests
Rib strain validation & Injury criteria evaluation 3. Sled tests 4. Accident reconstructions...back to 1 0.00 0.02 0.0 0.00 0.02 0.0 0 20 40 60 80 100 V [km/h] 0.000 0.015 0.030 0 20 40 60 80 100 Airbag trig time [ms] -50 0 50 Angle [deg] P(AIS3+ rib frac 1.0 0.8 0.6 0.4 0.2 0.0 AGE=30 AIS3+ 0 20 40 60 80 100 120 Iraeus, J., (2015). Stochastic finite element simulations of real life frontal crashes. Doctoral thesis, comprehensive summary. Umeå University. NASS/CDS Winsmash V [km/h] Simulation
Model morphing using SAFER THUMS v9.0 by Johan Iraeus. Female, 150cm, BMI=20 Female, 175cm, BMI=25 Female, 175cm, BMI=40 Male, 175cm, BMI=25 Male, 175cm, BMI=40 Johan = Male, 195cm, BMI=26
Modelling muscle response Example from the SAFER-project: Active Human Body Model.
Muscle elements Line elements Hill-type material model
Control strategy tt dddd tt uu tt = kk pp ee tt + kk ii ee ττ dddd + kk dd 0 dddd Musculoskeletal model Reference position Neuromuscular feedback control model PIDcontroller u(t) Activation dynamics Activation level Muscle material model Muscle force HBM Joint angle Shortening velocity Muscle length Neural delay Joint angle
Tuning to sled test 1 Sled test with lap belt only (Ejima et al. 2007) 10 m/s -2 acceleration over 0.2 s Instruction to be relaxed Passive Active 1
Resulting muscle activity Östh J, Brolin K, Carlsson S, Wismans J, Davidsson J (2012) The Occupant Response to Autonomous Braking: A Modeling Approach that Accounts for Active Musculature. Traffic Injury Prevention 13(3):265 277. 44
Tuning to Passenger Autonomous Braking Active 1 too soft => Active 2 Östh J, Brolin K, Carlsson S, Wismans J, Davidsson J (2012) The Occupant Response to Autonomous Braking: A Modeling Approach that Accounts for Active Musculature. Traffic Injury Prevention 13(3):265 277.
Tuning to sled test 2 Sled test with 3-point belt (Ejima et al. 2008): 8 m/s 2 acceleration over 0.6 s Active 2 => Active 3 Tune controller gains: 144 simulations single stage iteration meta model tt dddd tt uu tt = kk pp ee tt + kk ii ee ττ dddd + kk dd 0 dddd Ejima et al. (2008) Östh J, Brolin K, Bråse D (2014) A Human Body Model with Active Muscles for Simulation of Pre-Tensioned Restraints in Autonomous Braking Interventions. Traffic Injury Prevention, in press. 46
Results Ejima et al. (2008) Active 2 Active 3 Other sim. 47
Validation to Autonomous Braking 11m/s 2 autonomous braking interventions (Östh et al. 2013) Östh J, Brolin K, Bråse D (2014) A Human Body Model with Active Muscles for Simulation of Pre-Tensioned Restraints in Autonomous Braking Interventions. Traffic Injury Prevention, 16 (3) s. 304-313, 2015.
Schematic of HBM validation belt pay-out EMG EMG seat deformation
Validation experiments New experimental volunteer tests series: Drivers and passengers in test vehicle Muscle activity, vehicle data and 3-D kinematics Autonomous events and driving: Lane change w/o braking and with braking Braking U-turns All data collected and data analyses ongoing
Driver Braking Anticipatory postural response present in voluntary driver braking. Modelled by changing the reference positions in proportion to the acceleration load. 0.2 Reference Position (rad) 0.1 0-0.1-0.2 0 0.5 1 1.5 2 2.5 Time (s) Östh J, Eliasson E, Happee R, Brolin K (2014) A Method to Model Anticipatory Postural Control in Driver Braking Events, Gait & Posture. 40 (4) s. 664-669, 2014.
* Östh J, Brolin K, Bråse D. A Human Body Model with Active Muscles for Simulation of Pre-Tensioned Restraints in Autonomous Braking Interventions. Traffic Injury Prevention, 16 (3) s. 304-313, 2015. ** Östh J, Eliasson E, Happee R, Brolin K. A Method to Model Anticipatory Postural Control in Driver Braking Events, Gait & Posture. 40 (4) s. 664-669, 2014.
Ongoing / unpublished work Implemented muscle control for omnidirectional events using feedback control of Body angles, and Muscle length Experimental study on autonomous and driver Braking, and Steering. Validation
Omni-directional SAFER A-HBM
Omni-directional SAFER A-HBM Lane Change with Braking
Future outlook In 2025, will human body models be used and for what?
Acknowledgements The modelling work presented has been carried out in association with the SAFER - Vehicle and Traffic Safety Centre at Chalmers, Sweden. C3SE (Chalmers Centre for Computational Science and Engineering) is acknowledged for supplying computational resources.