Thorax soft tissue response for validation of human body models and injury prediction Master’s Thesis in Applied Mechanics JAN-FREDERIK RATER Vehicle Safety Division Department of Applied Mechanics CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden, 2013 Master’s Thesis 2013:07 THESIS FOR THE DEGREE OF A GRADUATED ENGINEER in MECHANICAL ENGINEERING Thorax soft tissue response for validation of human body models and injury prediction by JAN-FREDERIK RATER Vehicle Safety Division Department of Applied Mechanics CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden, 2013 Thorax soft tissue response for validation of human body models and injury prediction JAN-FREDERIK RATER c© JAN-FREDERIK RATER, 2013 MASTER’S THESIS IN APPLIED MECHANICS no 2013:07 ISSN 1652-8557 Department of Applied Mechanics Chalmers University of Technology SE-412 96 Gothenburg Sweden Telephone + 46 (0)31-772 1000 Cover: THUMS v3-M under single belt loading conditions in simulated table top tests by Kent et al. (2004) Department of Applied Mechanics Gothenburg, Sweden, 2013 Thorax soft tissue response for validation of human body models and injury prediction Master’s Thesis in Applied Mechanics JAN-FREDERIK RATER Department of Applied Mechanics Vehicle Safety Division Chalmers University of Technology Abstract Thoracic injuries like rib fractures and lung injuries are the most frequently occurring injuries in Road Traffic Collisions (RTCs). These injuries are severe and can be life- threatening. 81 % of all car occupants in fatal car accidents have thoracic injuries with an Abbreviated Injury Scale (AIS) score of 3+. Seatbelt use and air bags reduce the fatality risk by 61 % compared to unbelted car oc- cupants of vehicles without air bags. Nevertheless according to the National Highway Traffic Safety Administration (NHTSA) more than 30 000 people die each year due to RTCs in the USA. For the validation of new restraint systems and for injury prediction Anthropomorphic Test Devices (ATDs) were traditionally used. ATDs are only gross mechanical repre- sentations of the human body and thus the information to predict injuries accurately is limited. A second tool for the investigation of restraint systems and injury prediction are Finite Element Human Body Models (FE-HBMs). They offer a more detailed description of the anatomy of the human body, e.g. viscera are represented. The quality of Human Body Models (HBMs) is limited by the amount of details and the validation level of par- ticular parts. Lungs are, besides ribs, the most frequently and severely injured part of the body in RTCs. Despite this no investigations to validate human lung models under frontal car crash like conditions have been carried out and experimental data for the dynamic behaviour and injury mechanism are an exception. In this study, the state of the art of HBMs, models of the thorax and currently used mate- rial models for simulating thoracic viscera were identified. To rate and validate these material models for lungs, impact experiments on swine lungs were simulated with LS-DYNA. The time and force response of the models were com- pared to the experimental results at an impact speed of 5.4 m s . Coefficient studies with the parameter of different material models were accomplished to enhance the model response. For the best material model, low density foam, a new stress versus strain curve was also implemented, because the model tuning due to parameter optimization was limited. The deformation behaviour of the final model was close to the experimental results. Only the force response for the first part of deformation was higher than compared to the exper- iments. For rating the model quality the deformation and force response were compared V to the experimental data based on the Mean Square Error (MSE). Finally, the MSE of the optimized material model was only half of the MSE of the best model from literature. The final material model was implemented as material properties in the thoracic viscera of the Total HUman Model for Safety version 3.0 Modified (THUMS v3-M). The influence of the modified material model to the thoracic response and the biofidelity were proved against table top tests. The tuned material did not influence the thoracic response within the first 20 mm of chest deflection. Afterwards higher reaction forces occurred as tho- racic response with the tuned model, but the forces stayed clearly inside the experimental corridor. KEYWORDS: Frontal crash, thoracic injury criteria, lung injury, Human Body Model, Finite Element, model validation, THUMS, lungs modelling VI Acknowledgement This thesis was carried out at SAFER for Chalmers University of Technology, Applied Mechanics department, Traffic Safety division under the supervision of Manuel Mendoza- Vazquez and Johan Davidsson. I would like to thank several people for their encouragement, support, and patience during the work on this thesis. At Chalmers University of Technology in the Vehicle Safety Division I gratefully thank my supervisor M.Sc. Manuel Mendoza-Vazquez for the support he provided me through- out the research. Without his academic guidance, numerical knowledge and moral support this project may never have been completed. I would also especially like to thank Ph.D. Johan Davidsson that he provided me with the opportunity to carry out my work in SAFER - Vehicle and Traffic Safety Centre during an ERASMUS internship at Chalmers University of Technology. I wish to thank Ingrid Middleton for her help with her perfect English skills and endless patience whenever it was needed. I also would particularly like to thank my wonderful girlfriend Hanna for her support, love and help from the beginning until the end. You motivate me each day anew with your backing and encouraging words. Finally, I would like to thank my parents, Bärbel and Hans-Joachim, as well as my sisters, Juliane and Katharine, who endlessly encourage me in everything I do. I am eternally grateful for your support, encouragement and patience without which my accomplish- ments would not have been possible. Many thanks. VII Table of Contents List of Figures X List of Tables XIII List of Acronyms XIV 1 Introduction 1 1.1 Background and Research Justification . . . . . . . . . . . . . . . . . . . 1 1.2 Research Objectives and Scope . . . . . . . . . . . . . . . . . . . . . . . 2 2 Literature Research 4 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2 Anatomy of the Thorax . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.3 Car Crash Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.1 Abbreviated Injury Scale . . . . . . . . . . . . . . . . . . . . . . 6 2.3.2 Frontal Car Crashes . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3.3 Seatbelt Related Injuries . . . . . . . . . . . . . . . . . . . . . . 7 2.4 Human Body Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2.4.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4.2 Thoracic Modelling . . . . . . . . . . . . . . . . . . . . . . . . . 12 2.4.3 Total Human Model for Safety . . . . . . . . . . . . . . . . . . . 13 2.5 ATDs and Injury Prediction . . . . . . . . . . . . . . . . . . . . . . . . . 15 2.6 Lung Parenchyma Experiments . . . . . . . . . . . . . . . . . . . . . . . 17 2.6.1 Hoppin 1975 - Properties of Lung Parenchyma in Distortion . . . 17 2.6.2 Vawter 1978 - Elasticity of Excised Dog Lung Parenchyma . . . . 18 2.6.3 Zeng 1987 - Measurement of the Mechanical Properties of the Human Lung Tissue . . . . . . . . . . . . . . . . . . . . . . . . 20 2.7 Validation Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.7.1 Hayamizu 2003 - Measurement of Impact Response of Pig Lung . 21 2.7.2 Kent 2004 - Thoracic Response to Dynamic, Non-Impact Load- ing from a Hub, Distributed Belt, Diagonal Belt and Double Di- agonal Belts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.8 Material Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.8.1 Strain-Energy Function . . . . . . . . . . . . . . . . . . . . . . . 24 2.8.2 Viscoelastic Material Behaviour . . . . . . . . . . . . . . . . . . 26 2.8.3 Low Density Foam . . . . . . . . . . . . . . . . . . . . . . . . . 26 3 Methods 27 3.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.1.1 THUMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.1.2 Thoracic Organs . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.2 Impact Response Tests following Hayamizu . . . . . . . . . . . . . . . . 28 3.2.1 Experimental Modelling . . . . . . . . . . . . . . . . . . . . . . 29 VIII Table of Contents 3.2.2 Contact Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.2.3 Material Models . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.2.4 Coefficient Study . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.2.5 Load Curve Study . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.3 Model Validation - Table Top Tests . . . . . . . . . . . . . . . . . . . . . 34 4 Results 35 4.1 Material Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 4.1.1 Strain-Energy Function . . . . . . . . . . . . . . . . . . . . . . . 35 4.1.2 Viscoelastic Material Behaviour . . . . . . . . . . . . . . . . . . 35 4.1.3 Low Density Foam . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.1.4 Impact Speed Variation . . . . . . . . . . . . . . . . . . . . . . . 38 4.2 Coefficient Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4.2.1 Strain-Energy Function . . . . . . . . . . . . . . . . . . . . . . . 39 4.2.2 Viscoelastic Material Behaviour . . . . . . . . . . . . . . . . . . 41 4.2.3 Low Density Foam . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.3 Load Curve Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.4 Final Model Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.5 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 5 Discussion 51 5.1 Experimental Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 51 5.1.1 Lung Tissue Experiments . . . . . . . . . . . . . . . . . . . . . . 51 5.1.2 Impact Response Tests . . . . . . . . . . . . . . . . . . . . . . . 53 5.1.3 Table Top Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 5.2 Discussion of the Simulations . . . . . . . . . . . . . . . . . . . . . . . 54 5.2.1 Strain-Energy Function . . . . . . . . . . . . . . . . . . . . . . . 55 5.2.2 Viscoelastic Material Behaviour . . . . . . . . . . . . . . . . . . 56 5.2.3 Low Density Foam . . . . . . . . . . . . . . . . . . . . . . . . . 56 5.2.4 Model Validation . . . . . . . . . . . . . . . . . . . . . . . . . . 57 5.2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 5.3 Classification of the Results . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.3.1 Comparability Between Human Lungs, Swine Lungs and the Lungs Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 5.3.2 THUMS v3.0 Lung Model . . . . . . . . . . . . . . . . . . . . . 60 5.4 Injury Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 6 Conclusions 62 6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 References XV Appendix A XXI Appendix B XXII IX List of Figures 2.1 Thoracic organs without muscles, ribs and sternum (a); lungs with pul- monary segments, anterior view (b); modified from Schuenke et al. (2006) 5 2.2 Injury regions and severe for drivers in frontal car crashes, Cuerden et al. (2007) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Seatbelt signs by Hayes et al. (1991) . . . . . . . . . . . . . . . . . . . . 9 2.4 An overview of recent FE-HBMs, Yang et al. (2006) . . . . . . . . . . . 11 2.5 Stress versus strain curves from experiments and modified curves for low density foam, Mendoza-Vazquez et al. (2012); Vawter et al. (1978); Wang (1995) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.6 THUMS v4: AF05, AM50 and AM95, Toyota (2011) . . . . . . . . . . . 14 2.7 AIS injury severity depending to the velocity and compression for lungs and soft tissue, Lau and Viano (1986) . . . . . . . . . . . . . . . . . . . 16 2.8 Rat lung model with maximum strain (a) and best correlation metric (b) for εmax ∗ ε̇max, Gayzik et al. (2007) . . . . . . . . . . . . . . . . . . . . . 18 2.9 Schematic drawing of the tissue testing set-up (a); extension ratio un- der symmetrical loading for all axes against normalized tensile force (b), Hoppin Jr. et al. (1975) . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.10 Slabs specimen testing under uni- or biaxial loading conditions Fung (1993) (a); stress versus strain curve for eleven specimens under uniaxial loading condition (b) Vawter et al. (1978) . . . . . . . . . . . . . . . . . 19 2.11 Stress versus strain curve for a human lung tissue specimen subjected to a fixed load in x-direction and sinusoidally varied stretch in the y-direction, Zeng et al. (1987) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.12 High speed impact pictures of a swine lung after different times for 5.4 m s , Hayamizu et al. (2003) . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.13 Experimental results for lung impact experiments by Hayamizu et al. (2003) 22 2.14 Load cases for table top tests, Kent et al. (2004) . . . . . . . . . . . . . . 23 2.15 Force versus deflection and the corridors for hub, single and double di- agonal belts and distributed loading conditions; the coefficients shown in each plot refer to the quadratic equation y = αx2 +βx, Kent et al. (2004) . 24 2.16 Kelvin-Maxwell model (a), Wang (1995); behaviour of the low density foam model and the influence of the shape and decay factor (b), Livermore (2012) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1 THUMS v3-M and isolated thoracic viscera . . . . . . . . . . . . . . . . 28 3.2 Load curve for low density foam following Mendoza-Vazquez et al. (2012) for thoracic organs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.3 Model of the experimental set-up by Hayamizu et al. (2003) (isometric view) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.4 Time versus displacement curve adjusted to the impact time and scaled to the model size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.5 Schematic drawing of curve optimization with LS-OPT . . . . . . . . . . 33 X List of Figures 3.6 Experimental set-up of the table top tests by Kent et al. (2004) with a single belt; the arrow points to the place where the displacement was recorded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 4.1 Simulated experiments for different values of the strain-energy function at the moment of error termination . . . . . . . . . . . . . . . . . . . . . 36 4.2 Simulated experiments for different values of the viscoelastic material model at the moment of error termination (a) and (b) and after some par- ticular elapsed time for normal termination by Roberts et al. (2005) (c) and (d) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 4.3 Thoracic volume deformation behaviour with the low density foam mate- rial model after different times with an impact speed of 5.4 m s . . . . . . . 38 4.4 Experimental data by Hayamizu et al. (2003) and simulation data for an impact speed of 5.4 m s . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 4.5 Model response and experimental data for different impact speeds for low density foam material model . . . . . . . . . . . . . . . . . . . . . . . . 39 4.6 MSE sensitivity study for the strain versus energy function with a = α and b = β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.7 Model response for a coefficient study for the material model strain-energy function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 4.8 Stiffest ((a) and (b)) and softest ((c) and (d)) material models from the strain-energy function parameter study shown in Figure 4.7 . . . . . . . . 41 4.9 MSE sensitivity study for the viscoelastic material model, with g0 = GS, gi = GL and b = β . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 4.10 Model response for a coefficient study of viscoelastic material behaviour . 42 4.11 Simulated thoracic volume with viscoelastic material model with the low- est MSE at the state of highest deformation . . . . . . . . . . . . . . . . 43 4.12 MSE sensitivity study for the viscoelastic material model with: decay constant b = β , density ro = ρ , viscous coefficient d = DAMP, young’s modulus e = E, shape factor for unloading s = SHAPE and the hysteretic unloading factor hu = HU . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.13 Model response for a coefficient study of the material model low density foam; colours depending to the viscous coefficient d. . . . . . . . . . . . 44 4.14 Model response for the the material model low density foam at the mo- ment of highest compression ((a) and (b)) and model response plots ((c) and (d)) for the best results of the coefficient study . . . . . . . . . . . . 45 4.15 Model responses for different load curves . . . . . . . . . . . . . . . . . 46 4.16 Stress versus strain curves from curve optimization and from literature . . 46 4.17 Model response for the final material model . . . . . . . . . . . . . . . . 47 4.18 Model response of the viscus for the final material model . . . . . . . . . 48 4.19 THUMS v3-M with the modified lung material model under single belt loading conditions, the left part of the body had been removed . . . . . . 49 4.20 Force versus compression response for THUMS v3-M . . . . . . . . . . 49 5.1 Stress versus strain curves from experiments, Hoppin Jr. et al. (1975); Radford and Remington (1957); Vawter et al. (1978); Zeng et al. (1987) . 52 XI List of Figures 6.1 Force versus compression response for THUMSv3-R and THUMS v3-M, Mendoza-Vazquez et al. (2012), and the experimental table top corridors by Kent et al. (2004) . . . . . . . . . . . . . . . . . . . . . . . . . . . . XXI XII List of Tables 2.1 Lungs volume and linear dimension for male and female data sets with standard deviation, Kramer et al. (2012) . . . . . . . . . . . . . . . . . . 5 2.2 Abbreviated injury scale, States (1990) . . . . . . . . . . . . . . . . . . . 6 2.3 AIS for rib cage and thoracic injuries, States (1990) . . . . . . . . . . . . 7 2.4 Number of frontal impacts for different EES, Carroll (2009) . . . . . . . 7 3.1 Low density foam material variables from LS-DYNA for thoracic viscera from THUMS v3-M, Mendoza-Vazquez et al. (2012) . . . . . . . . . . . 28 3.2 Impactor densities for adjusted masses . . . . . . . . . . . . . . . . . . . 30 3.3 Values for the different material models from literature, Mendoza-Vazquez et al. (2012); Plank et al. (1998); Roberts et al. (2005); Ruan et al. (2003); Stitzel et al. (2005); Vawter (1980); Zhao and Norwani (2004) . . . . . . 32 4.1 Viscoelastic material parameter for the best MSE obtained from parame- ter studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 4.2 Parameter for low density foam from parameter and curve studies and the corresponding MSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 5.1 Kinetic energy for different impact speeds . . . . . . . . . . . . . . . . . 53 6.1 Material properties for lung tissue . . . . . . . . . . . . . . . . . . . . . XXII XIII List of Acronyms ATD Anthropomorphic Test Device AIS Abbreviated Injury Scale CCIS UK Co-operative Crash Injury Study CDC Collision Deformation Classification Cmax Maximum Chest Compression CT Computer Tomography DC Combined Deflection Criterion EES Energy Equivalent Speed FE Finite Element FEM Finite Element Method FE-HBM Finite Element Human Body Model GHBMC Global Human Body Models Consortium HBM Human Body Model NHTSA National Highway Traffic Safety Administration HUMOS HUman MOdel for Safety LDT Linear Differential Transducer MSE Mean Square Error MSEdis Mean Square Error for displacement MSEforce Mean Square Error for force PMHS Post Mortem Human Subjects THOR Test device for Human Occupant Restraint THUMS Total HUman Model for Safety THUMS v3.0 Total HUman Model for Safety version 3.0 THUMS v3-M Total HUman Model for Safety version 3.0 Modified PC Pulmonary Contusion RTC Road Traffic Collision VC Viscous Criterion WHO World Health Organization XIV 1 Introduction 1.1 Background and Research Justification According to the World Health Organization (WHO) nearly 1.2 million people die yearly in the world due to traffic accidents and up to fifty million people retain permanent dis- abilities, Peden et al. (2004). During the last forty years the amount of fatal car crashes has decreased continuously. In Germany in the year 1970 there were nearly 1.4 million Road Traffic Collisions (RTCs) with 19 193 fatalities. Despite increasing RTCs up to 2.4 million the amount of fatal ac- cidents reached an all-time low of 3 648 in the year 2010, Bundesamt (2012). The introduction of modern restraint systems like seatbelts and air bags contributed to this significant decline in fatal accidents, Bean et al. (2009). A statistical analysis of acciden- tal data of the National Highway Traffic Safety Administration (NHTSA) by Bean et al. (2009) showed that seatbelt use and air bags reduce the fatality risk by 61 % compared to unbelted car occupants of a vehicle without air bags. In frontal impacts the fatality risk is actually reduced by up to 74 %, NHTSA (2009). Despite this in the USA in the year 2010 11 628 restraint car occupants died in RTCs, NHTSA (2010). A study of UK Co-operative Crash Injury Study (CCIS) data by Cuerden et al. (2007) revealed that in the UK frontal car impacts are responsible for one third of all fatal car accidents. Klanner (2001) reported that in Europe even 40 % of all fatal car accidents are frontal impacts. Rib fractures and lung injuries are the most frequently occurring injuries in car crashes with a serious or more severely (Abbreviated Injury Scale (AIS) ≥ 3) injured torso, Car- roll (2009). At 81 % thoracic injuries with an AIS score of 3+ are the most frequent injuries of drivers who die in frontal car accidents, Cuerden et al. (2007). Thus, thoracic injuries are the most life-threatening injuries due to RTCs in the sample studied by Cuer- den et al. (2007). Despite the immense safety improvements due to restraint systems seatbelts do not com- pletely prevent severe or fatal injuries and can themselves cause injuries. To reduce the number of rib fractures and the severity of lung injuries it is necessary to improve current restraint systems. A benchmark is required to evaluate improvements of restraint systems. A common tool for the evaluation of restraint systems are Anthropomorphic Test Devices (ATDs). They are used e.g. in simulated vehicle impacts. However, ATDs are only a gross mechanical representation of the human body, e.g. organs are not represented. To predict injuries of the thoracic cage the deformation, acceleration, velocity and force can be recorded with ATDs. A reliable prediction of injuries car occupants may suffer due to RTCs is a desirable goal. Hence, a validated injury criterion is a prerequisite. Current criteria like the Maximum Chest Compression (Cmax) or the Viscous Criterion (VC) are based on the deformation 1 1 Introduction 1.2 Research Objectives and Scope measurement with ATDs. These criteria do not fulfil the injury mechanism of lung in- juries, Gayzik et al. (2007). The issue is that the detection of stresses and strains related to lung injuries in experiments is difficult, if not impossible, Ruan et al. (2003). Next to ATDs Finite Element Human Body Models (FE-HBMs) are a second tool to eval- uate restraint systems. The advantage of FE-HBMs is that they offer a more detailed description of the human anatomy and stresses and strains can be calculated. A reliable prediction of injuries requires a validated Human Body Model (HBM) next to an injury metric. The most important part for the prediction of life-threatening injuries is the thorax. The model of the thorax is usually validated against pendulum impact tests, sled tests or table top tests with Post Mortem Human Subjects (PMHS). Several investigations concerning mechanical properties, injury prediction and modelling of hard tissue like ribs are avail- able. In contrast, investigations concerning soft tissue like lungs are rare. The mechanical behaviour and injury prediction of lungs are not well researched, Gayzik et al. (2007); Ruan et al. (2003). Since the beginning of thoracic modelling several material models for the simulation of lungs were developed. The material models of presently used HBMs were mainly ad- justed to validate the thoracic response against PMHS experiments. Despite the relevance of lung injuries due to RTCs particular validations of the lung models under frontal car crash conditions were not carried out. 1.2 Research Objectives and Scope The objective of the research in this thesis is outlined with four main goals. The first goal is to identify the state of the art of lung modelling and to figure out material models that are used for the simulation of thoracic organs in general and lungs in particular. Further- more experiments which are suitable for the validation of a lung model are determined as input for the following part of the thesis. To modify the thoracic response of HBMs the material properties of the thoracic organs have often been used. No investigations have been done in particular to validate human lung models under conditions similar to frontal crashes. Therefore, the second goal is to rate and to compare currently used material models for lung modelling. For the rating of the material models an experiment suitable for lung validation has to be identified and to be simulated with Finite Element Methods (FEMs) with different material models. The results will be used as input for the next part of the thesis. The third goal is to improve the simulated model response against the experiments by modifying the material model. The biofidelity of a modern HBM with the modified mate- rial model has to be proved against PMHS experiments to investigate the influence of the modification to the thoracic response. 2 1 Introduction 1.2 Research Objectives and Scope The fourth and final goal of the thesis is to identify the state of the art of lung injury prediction and the lack of knowledge for lung validations. These information have to be discussed and suggestions for future work and experiments have to be developed. 3 2 Literature Research 2.1 Introduction The aim of this thesis is to compare, to evaluate and to enhance current material models of lung tissue for FE-HBMs. These models are used for automotive safety investigations. Therefore, in the first part of this chapter a short introduction to the anatomy of the thorax and important information of thoracic injuries due to RTCs will be given. Afterwards an overview of past and current human body and thoracic models will be given. Furthermore, experiments with the aim to figure out mechanical properties of lung parenchyma as well as experiments suiting for the validation of a human lung model and of the thoracic response will be summarized. The literature research is completed by presenting common material models of lung models which were used for the model validation in the further thesis. 2.2 Anatomy of the Thorax Since knowledge of the anatomy of the thoracic organs in general and the lungs in partic- ular are essential for further investigations about the lungs, the following will provide it. The heart and lungs are next to the brain two central organs of human beings and located in the thorax. The thorax is in the body between the abdomen and the neck (Figure 2.1 (a)). The thoracic cage consists of the thoracic vertebrae, the sternum and the ribs (re- moved in Figure 2.1) and protects the lungs and the heart against hits and contusion. The chamber inside the thoracic wall contains the principal organs of respiration (the tra- chea, bronchi and lungs (1)) and circulation (the heart (2) and great vessels (3)) (Figure 2.1 (a)). As it can be seen in Figure 2.1 the parietal pleura separates the chamber again into two closed chambers for the lungs and separates the thoracic cage from the abdominal organs (4). The organs of the circulation system are located between the two lung chambers, Schuenke et al. (2006); Schulte and Schumacher (2012); Standring et al. (2005). Lungs The lungs are the essential respiration organ of humans. The human lungs consist of the right and the left lung. The basic structural unit of each lung is the lobe. The right lung has three lobes, the upper, middle and lower lobe (cf. Figure 2.1 (b)). The left lung only has an upper and a lower lobe. Each lobe is further subdivided in segment wedges. The pulmonary segments are only in- completely separated from each other and they are not discernible as separate units on the lung surface. Each lung basically consists of ten segments (cf. Figure 2.1 (b)). The seg- ments then consist of segmental bronchus which get divided up to the pulmonary alveoli where the actual gas exchange happens, Schuenke et al. (2006); Schulte and Schumacher (2012). 4 2 Literature Research 2.2 Anatomy of the Thorax (a) (1) lungs, (2) heart, (3) great vessels and (arrows) pleu- ral space (b) Right and left lung with pulmonary seg- ments Figure 2.1: Thoracic organs without muscles, ribs and sternum (a); lungs with pulmonary segments, anterior view (b); modified from Schuenke et al. (2006) With a volume of 1.5 l the deflated right lung is slightly bigger than the left lung with a volume of 1.4 l. This is caused by the non-symmetric position of the heart, Schulte and Schumacher (2012). Kramer et al. (2012) analysed Computer Tomography (CT) images of 166 patients with the aim to measure the linear dimension and volume of human lungs. Contrary to the volume of deflated lungs mentioned before, these values relate to inflated lungs inside the body. The results of their study can be found in Table 2.1. Table 2.1: Lungs volume and linear dimension for male and female data sets with stan- dard deviation, Kramer et al. (2012) Male Female Combined Height [cm] Left 21.0 ± 2.1 19.9 ±2.5 19.8 ± 2.6 Right 21.0 ± 2.1 19.0 ± 2.5 20.6 ± 2.6 Max height [cm] Left 28.2 ± 2.2 26.0 ± 2.7 26.1 ± 2.6 Right 21.0 ± 2.1 26.0 ± 2.7 26.9 ± 2.7 Width [cm] Left 12.3 ± 1.1 11.1 ± 1.0 10.0 ± 1.0 Right 12.3 ± 1.1 11.1 ± 1.0 11.6 ± 1.2 Depth [cm] Left 18.0 ± 1.5 16.2 ± 1.7 17.1 ± 2.0 Right 18.0 ± 1.5 16.2 ± 1.7 16.9 ± 1.8 Volume [cm3] Left 2738 ± 533 1968 ± 505 2301 ± 636 Right 3121 ± 605 2300 ± 547 2663 ± 667 Total 5858 ± 1094 4268 ± 1028 5 2 Literature Research 2.3 Car Crash Information 2.3 Car Crash Information One key aspect of this paper are the simulations of restraint systems under frontal impact loading conditions. Therefore it should shortly be mentioned what a frontal car crash is, which injuries occur and why an investigation of these accidents is so important. 2.3.1 Abbreviated Injury Scale The intention behind developing the AIS was to get a universal and widely accepted injury scale which describes and classifies the injury level for automotive accident investigations. It was first developed and introduced by the Association for the Advancement of Auto- motive Medicine in 1969. As it is shown below in Table 2.2 the AIS divides injuries into seven levels from 0 (no injury) to 6 (maximum injury) which is also often termed virtually lethal, Forbes (2005); Nahum and Melvin (2002); States (1990). A higher AIS level means a greater life-threatening injury but it should be noted that the scale is not continuous. This means an AIS level of 4 is much more severe than two AIS level of 2, Cuerden et al. (2007). Table 2.2: Abbreviated injury scale, States (1990) AIS Score Description 0 No Injury 1 Minor 2 Moderate 3 Serious 4 Severe 5 Critical 6 Maximum Since its first publication the AIS has undergone several revisions and also various scales for specific regions of the body were published. These are e.g. scales for vascular injuries or for scull fractures. The AIS for injuries to the rib cage and the thoracic soft tissue which is shown below in Table 2.3, States (1990). 2.3.2 Frontal Car Crashes A car crash is called frontal car crash if the car hits a barrier or another car in a frontal collision between eleven and one o’clock impact direction. More than 66 % of all car impacts occur in this area, Cuerden et al. (2007) An study of CCIS data by Cuerden et al. (2007) revealed that in the UK frontal car impacts are responsible for one third of all fatal car accidents. The data were collected from June 1998 and only accidents with cars built later than 1996 were used. 6 2 Literature Research 2.3 Car Crash Information Table 2.3: AIS for rib cage and thoracic injuries, States (1990) AIS Level Rib Cage Injury Thoracic Soft Tissue Injury 1 1 rib fracture Contusions of the bronchus 2 2-3 rib fractures; sternum fracture Partial thickness bronchus tear 3 4 or more rib fractures on one side; 2-3 rib Lung contusion; fractures with hemothorax or pneumothorax minor heart contusion 4 Flail chest; 4 or more rib fractures on Bilateral lung laceration; each of two sides; 4 or more rib fractures minor aortic laceration; with hemothorax or pneumothorax. major heart contusion 5 Bilateral flail chest Major aortic laceration; lung lace- ration with tension pneumothorax 6 Aortic laceration with hemorrhage not confined to mediastinum Table 2.4 below shows the absolute frequency and the percentage of car crashes in relation to to the Energy Equivalent Speed (EES). The EES is the equivalent speed at which a particular vehicle would need to contact any fixed rigid object in order to dissipate the deformation energy corresponding to the observed vehicle residual damage. Table 2.4: Number of frontal impacts for different EES, Carroll (2009) EES (km/h) Number of front impact accidents Percentage [%] < 15 4 1 16 - 25 37 7 26 - 35 98 18 36 - 45 121 23 46 - 55 108 20 56 - 65 120 23 66 - 75 31 6 > 75 11 2 As it can be seen in Table 2.4 two third of all accidents happen between 36 km h and 65 km h EES. The risk of a moderate thorax injury (AIS 2+) is for an EES speed of 56 km h to 65 km h already nearly 30 %, Carroll (2009). In a sled test with 48 km h the chest compression rate measured with PMHS is 1 m s , Kent et al. (2004). 2.3.3 Seatbelt Related Injuries Thoracic injuries due to car accidents are severe and life-threatening and often caused through seatbelts. Therefore it is important to know what kind of injuries occur in RTCs and why injuries of the thorax are so severe. Bean et al. (2009) showed that the use of seatbelts reduced the mortality risk in RTCs by 61 %. Seatbelts were designed to prevent occupants from hitting the interior of the car 7 2 Literature Research 2.3 Car Crash Information or getting ejected out of the car. Seatbelts scatter the kinetic energy from the rapid decel- eration through the body skeleton, however seatbelts can also cause injuries themselves, Abbas et al. (2011). The first seatbelts that were implemented in cars were lap belts. Holding a body only at two points had the disadvantage that major forces are transferred directly through the lum- ber spine and lap belts did not prevent the head and chest from moving forward against the steering wheel and the windscreen. The basics of the presently used three point belt were developed in 1968 by Volvo. This seatbelt prevents the upper body from bending forward against the interior, Abbas et al. (2011). Incorrect seatbelt usage and wrong seatbelt application influence the injury severity in RTC, but the major influence on the injury severity is the velocity. There is a clear as- sociation between high speed accidents and fatal injuries. The formula for the energy (E) (Equation 2.1) explains the relationship between high velocity (v) and fatal injuries. The energy increases exponentially with increasing velocity. High energy leads to severe injuries, Abbas et al. (2011). E = 1 2 ∗m∗ v2 (2.1) As it is shown in Figure 2.2, 81 % of the drivers dying due to frontal car accidents had a thoracic injury AIS score of 3+ and only 37 % survived. In contrast 72 % of the car occupants in frontal car accidents had an AIS injury scale of 3+ for the lower extremities but 60 % of them survived. It should be noted that this data does not allow conclusions for fatal injuries because there is no information about multiple injuries and interactions. Nevertheless, these data show that thoracic injuries due to car accidents are quite common, very often severe and life-threatening. Figure 2.2: Injury regions and severe for drivers in frontal car crashes, Cuerden et al. (2007) Naturally modern three point seatbelts help to protect car occupants in RTC. But despite the enormous advantages seatbelts can cause seatbelt-related injuries. Hayes et al. (1991) called these injuries the seatbelt syndrome and divided them into different groups. The skin abrasion and contusion of the neck, chest and abdomen is called the seatbelt sign and 8 2 Literature Research 2.4 Human Body Models indicates internal injuries in 30 % of the cases (Figure 2.3). The other groups are skeletal injuries, soft-tissue and visceral injuries and vascular injuries. Injuries of the lungs belong to the group of soft tissue and visceral injuries. (a) skin contusion to the chest (b) and skin abrasion to the neck Figure 2.3: Seatbelt signs by Hayes et al. (1991) In order to analyse and describe injuries in a more detailed way the torso can be subdivided into sternum, shoulder, ribs, lungs, heart, spine and the abdomen. An analysis of the CCIS database showed that a lung injury of AIS ≥ 3 was the most common injury of all car occupants who were killed or seriously injured, Carroll (2009). Lung injuries caused by RTC are mainly Pulmonary Contusion (PC) and pneumothorax. A PC is a non penetrating contusion of the lung caused by a chest trauma. Blood or other fluids can percolate through the lung tissue as a result of a damage to the lung capillaries. This can effect the gas exchange and as a result lead to an inadequate oxygen level. A pneumothorax due to RTC is also called traumatic pneumothorax and describes the abnormal collection of air or blood in the pleural space between the lungs and the chest wall (cf. Figure 2.1 (a) black arrows). This can result in a lung collapse because the natural vacuum disappears. In contrast to a PC there is a cut or a tear of the lung tissue. This results either from a blunt trauma or a penetrating injury, for example caused by a fractured rib. Both injuries can interfere with the normal breathing and can be fatal. 2.4 Human Body Models The aim of this chapter is to give an overview of HBMs development and the current state of the art. HBMs can be used to study the interaction with restraint systems an to estimate the risk of injuries. For the simulations in this thesis the Total HUman Model for Safety version 3.0 (THUMS v3.0) and the thoracic organs from this model have been used. Therefore, a detailed introduction of THUMS v3.0 and modifications that have been done will be given. Caused by the thematic orientation of this thesis on lungs modelling, the focus lies on thoracic models and the material models used for thoracic visceral and lungs. 9 2 Literature Research 2.4 Human Body Models 2.4.1 History Numerical analyses have accompanied experimental investigations since the beginning of the computer age in the 1960s. Limited by computational speed it was necessary to simplify the mathematical model of the experimental system to a small set of derivations. Models were mainly based on lumped-mass models, multi body and Finite Element (FE) models. With the multi body method the kinematic response can be calculated, while with the FEM the dynamic and material response can be calculated. Due to the limited computational speed numerical calculations were focused on an isolated part of the body like the head, neck, thoracic, abdominal and upper and lower extremities. The highly enhanced performance of information technologies also increased the quality and the amount of details of computer models. The head was the first part of the body analysed by a numerical model. The first head lumped-mass model was developed by Hodgson et al. (1967) and was used for an inves- tigation of the dynamic response of a cadaveric scull with a simple spring-dashpot-mass model. The first finite element model was published by Chan (1974) and the head was represented by an ellipsoid. This model already contained a brain represented by a vis- coelastic material, Yang et al. (2006). In the following years lots of different models for different experiments and parts of the body were developed and published. Because of the complexity it took nearly a further thirty years until the first real human body model was published in 1995, Yang et al. (2006). Developed by Huang (1995) this FE-HBM contained 9 308 solid elements and 2 384 shell elements (cf. Figure 2.4 (a)). The model already included bones (e.g. ribs and sternum) and soft tissues like a skin and integrated a pelvis. The model was validated by a side impact test with cadavers. Several FE-HBMs have been developed in recent years, for example by Happee (1998), Lizee et al. (1998), Van Hoof (2003) and Ruan et al. (2003) (cf. Figures 2.4 (b)-(e)). The last finite element models were the Total HUman Model for Safety (THUMS) (cf. Section 2.4.3) developed by Toyota (Toyota Central R&D abs Inc., Nagekute, Aichi, Japan) and the HUman MOdel for Safety (HUMOS) by Vezin and Verriest (2005). An evaluation study of Holmqvist (2009) showed a better performance of the THUMS v3.0 compared to the HUMOS2 model for side impacts, Toyota (2011); Yang et al. (2006). The THUMS model is used by several car companies and research institutions, e.g. Chal- mers University of Technology. In 2011 Toyota released an improved version of THUMS (THUMS v4.0) incorporat- ing individual organs parts, Toyota (2011). Currently a new model, the Global Human Body Models Consortium (GHBMC), is being developed by a global consortium of seven car companies and one supplier with the purpose of advancing crash safety technology, Gayzik et al. (2012). Researchers all over the world are improving the quality and biofidelity of parts or of a whole FE-HBM. Biofidelity is defined by Wismans et al. (2005) as the process where the reliability of a model is assessed against a set of PMHSs tests. The task is to obtain a completely validated model that represents a human body for automotive safety research. 10 2 Literature Research 2.4 Human Body Models (a) Huang (1994) (b) Happee (1998) (c) Lizee (1998) (d) Van Hoof (2003) (e) Ruan (2003) (f) THUMS v3 (2003) Figure 2.4: An overview of recent FE-HBMs, Yang et al. (2006) Of course, current FE-models have a much higher quality than in the past but still there are a lot of uncertainties. This has different reasons. On the one hand there is still a huge lack of knowledge for the mechanical properties of different parts of the human body, especially for soft tissue (e.g. organs and skin). The mechanism of bone fracture is also not completely understood yet. On the other hand the understanding of an isolated part of the body is not sufficient. The interaction between different tissues and parts of the body like muscles, bones and organs has to be known as well and need to be implemented in the model. Usually, the focus lay on whole body response or on validation of parts of the body (e.g. the thorax) and on fractures and raptures of bones and ligaments. Soft tissue modelling was mainly a means to an end for human body model validation. Yet, the injury mech- anism of soft tissue is not well known. Because of these uncertainties referring to soft tissue it was a legitimate way to change the material properties and values of visceral to receive satisfactory results, e.g. Wang (1995). 11 2 Literature Research 2.4 Human Body Models 2.4.2 Thoracic Modelling In the last 50 years several models of the thorax and lungs have been developed. For visceral modelling in general and lungs modelling in particular different material models have been used. The most common material models are pseudo-elastic, viscoelastic and low density foam. For the material properties the authors chose either values randomly or used material properties from experiments with lung tissue (cf. Section 2.6). All values from the different publications presented in this section are summarized in Tables 3.3 and 6.1. Numerical simulation of the thorax started in the 1970’s with a 2-D spinal column model by Begemann et al. and a model by Lobdell et al. (1973). Lobdell’s model was tuned by Kroell (1976) and Viano et al. (1978 and 1987), Wang (1995); Yang et al. (2006). Four years later Sundaram and Feng (1977) developed a three dimensional model of the thorax using solid elements to represent the internal organs. Sundaram and Feng used non-linear homogeneous material behaviour proposed by Matthews and West (1972) as material properties. Matthews’ and West’s material data rely on experiments from Rad- ford and Remington (1957). Huang (1995) developed a human body model to investigate the biomechanics of side impacts. Therefore he compared his model with cadaveric side impact sled tests. The internal organs of the thorax were represented by one volume. He assumed a soft, vis- cous, isotropic and homogeneous material which was achieved by discrete dampers. The material properties were chosen without reference to literature. Huang argued that a gross representation of the visceral need not be proved. Models Using Pseudo-Elastic Material Models Vawter (1980) investigated the behaviour of a two dimensional lungs model loaded by its own weight. As material properties he used the pseudo-elastic model represented by a strain-energy function proposed by Fung et al. (1978). For the parameter calculation Vawter used his own experiments summarized in Section 2.6.2, Vawter et al. (1978). Another FE-HBM which used the strain-energy function was developed by Zhao and Nor- wani (2004). The experimental data for the coefficient calculation were used from Yen (1999). Gayzik (2008) developed an FE based injury metric for pulmonary contusion with a rat lung model developed by Stitzel et al. (2005) in a previous study. Therefore Stitzel et al. (2005) used an algorithm to optimize the coefficients for force versus displacement curves from experiments with rats. Models Using Viscoelastic Material Behaviour For an analytic investigation of driver thoracic response, Plank et al. (1998) exchanged the thoracic part of an existing FE-HBM with a new further developed model of the human thorax. For this experiments Plank et al. chose the material properties proposed by Her- rmann and Peterson (1968) which are based on viscoelastic stress analysis (cf. Section 2.8.2). As Young’s modulus they used the intermediate values from the heart and lungs proposed by Sundaram and Feng (1977). The density and bulk modulus were taken from Plank et al. (1998) without further literature references. 12 2 Literature Research 2.4 Human Body Models In the same year Lizee et al. (1998) developed and validated an FE-HBM of a seated 50th percentile adult male. He used viscoelastic material behaviour like Plank et al. (1998) but with different values. Again no reference for the material parameter was given. Ruan et al. (2003) and Roberts et al. (2005) used this viscoelastic material model for non-penetrating ballistic and pendulum impact tests as well. Models Using Low Density Foam as Material Model Wang (1995) developed an FE human thoracic model for side impacts. For the properties of the heart and lungs he used non-linear stress versus strain curves. Wang used the experimental data from Vawter et al. (1979) (cf. Section 2.6.2) as values for the load curve. To approximate the assumed response he increased the values ten times without giving further reasons. He used a highly compressible foam as material model. The same material model was used for the thoracic viscera of THUMS v3.0 as well. The plotted stress versus strain curves can be seen in Figure 2.5, Kimpara et al. (2005). Figure 2.5: Stress versus strain curves from experiments and modified curves for low density foam, Mendoza-Vazquez et al. (2012); Vawter et al. (1978); Wang (1995) 2.4.3 Total Human Model for Safety THUMS is an FE-HBM developed by Toyota Motor Corporation and Toyota Central R&D Labs., Inc. The model aims to simulate human body kinetics and injury responses in car crashes. The material properties are defined by constitutive material laws and the geometries of the human body parts are represented by finite element meshes, Toyota (2011). There are different versions and variations of THUMS. The basis of all THUMS versions is an average sized adult male (AM 50th %-ile) with a height of 175 cm and a weight of 75 kg. A small sized woman (AF 5th %-ile) and a large sized male (AM 5th %-ile) have also been developed. All models exist in a sitting and a standing posture representing a 13 2 Literature Research 2.4 Human Body Models car occupant and a pedestrian, respectively, Toyota (2011). THUMS v1, the first version of THUMS was published in the year 2000. The model al- ready contained bones and ligaments but the brain and internal organs were simplified as solid parts. The total amount of elements was around 80 000 with an average mesh size of 15 mm. The aim of the model was to simulate bone fractures and ligament raptures in car crashes. The second version (THUMS v2) was completed in 2004 and included a modification of the facial bones. THUMS v3.0, the third version has been available since 2008 and includes a new brain model for simulating brain injuries. The model consisted roughly of 150 000 elements and 110 000 nodes. Joints were modelled anatomically including the major ligaments and bone to bone contact. Currently, this version has established itself in several companies and research institu- tions. Chalmers University of Technology is using this model for research projects in cooperation with different partners for automotive safety research. THUMS v3.0 is the basis model on which the further research is based. The latest version of THUMS (THUMS v4) was published in 2010 and different internal organs are integrated. The total number of elements is around 2 000 000. The three standing pedestrian versions of a small sized woman, an average sized man and a large sized man can be seen in Figure 2.6. However, this model is not yet established due to the high number of elements that make calculation time high and due to projects still in progress using THUMS v3.0. Figure 2.6: THUMS v4: AF05, AM50 and AM95, Toyota (2011) Modifications of THUMS v3.0 Different modifications have been made to improve the biofidelity of THUMS v3.0. Mu- rakami et al. (2006) used the table top tests by Kent et al. (2004) (cf. Section 2.7.2) for an 14 2 Literature Research 2.5 ATDs and Injury Prediction evaluation study. They found out that changed properties of the rib cartilage can improve the model response compared to the experimental results. Pipkorn and Kent (2011) modified the mesh and material data and added muscles to THUMS v2.21. Their modified model reacted in a similar way to the PMHS in Kent’s table top tests. An important modification of the THUMS v3.0 was carried out by Mendoza-Vazquez et al. (2012). For a study on the human rib response using an FE-HBM he modified parts of the thorax. The original THUMS v3.0 terminated in some simulations with errors when contact with a seatbelt was involved. To increase the numerical stability and robustness Mendoza-Vazquez deactivated the element elimination and remeshed the intercostal mus- cles, bones and flesh of the ribcage according to Mroz et al. (2010) and Pipkorn and Kent (2011). Afterwards the cross sectional width of the ribs seven and eight were changed to increase the elastic stiffness too experimental values. Because the response of the thorax was too stiff compared to Kent’s table top experiments the material properties of the flesh and the thoracic organs were changed. The thoracic organs in THUMS v3.0 were modelled as a highly compressible foam with an input curve of stress versus strain. The original values for the curve stemmed from experiments by Vawter et al. (1978) As mentioned above, manipulated values were used for THUMS v3.0, Kimpara et al. (2005). Mendoza-Vazquez et al. (2012) decreased this stress versus strain curve again by multi- plying the original curve from THUMS v3.0 with 10−6. This stress versus strain curve is plotted in Figures 2.5 and 3.2. In order to counteract numerical instability by negative element volume due to high com- pression Mendoza-Vazquez increased the stress versus strain curve by 90 % of strain. This modification increased the stiffness for high deformations. This model is called Total HUman Model for Safety version 3.0 Modified (THUMS v3-M). The biofidelity of this modified model was approved by comparing the model response with the table top tests by Kent et al. (2004) summarized in Section 2.7.2. The results for the simulated table top test of the THUMS v3.0-R and THUMS v3-M can be seen in Figure 6.1. This shows that the model response of THUMS v3-R was out of the experimental corridor for three of four load cases. In contrast the model response for THUMS v3-M was almost always inside the experimental corridor for each load case. This modified version THUMS v3-M has also been used for the biofidelity verification with simulated table top tests in this thesis. The visceral model with the modified mate- rial properties of THUMS v3-M was the initial model for further lungs simulations. 2.5 ATDs and Injury Prediction For the evaluation of improved restraint systems or e.g. a new designed interior of cars, tools are required to predict injuries car occupants may sustain in a specific impact. There- fore usually ATDs are used in sled tests or car crash tests. Thus, ATDs are only a gross mechanical representation of the human body, e.g. viscera are not represented in a ATDs. 15 2 Literature Research 2.5 ATDs and Injury Prediction The most commonly used ATDs Hybrid III was validated against pendulum impacts to the mid sternum, e.g. by Foster et al. (1977). The injury criterion for thoracic injury prediction Cmax was developed by Kroell et al. (1974) and is defined as the ratio of chest deflection to the initial chest depth. This crite- rion allows prediction about rib fractures. The VC from Lau and Viano (1986) takes beside the deformation the time into account. The VC is a time function generated by the product of the velocity of deformation and the compression of the thorax. According to Lau and Viano (1986) a VC of 1.0 m s cor- responds to a 25 % risk of a severe thoracic injury (AIS ≥ 4). For soft tissue injuries in general Lau indicates that the VC can be used for deformation velocities below 3 m s . The deformation velocity for restraint car occupants is usually below 3 m s . Lau and Viano (1981) investigated the influence of impact velocity and chest compression to the injury severity of rabbit lungs. For this study they used impact velocities of 5, 10 and 18 m s . The severity of lung injuries increased with chest compression at a constant velocity (cf. Figure 2.7 (a)). Regions of similar injury severity could be separated by hyperbolas. In addition Lau found out, that the alveolar region of the lungs was more sensitive to the rate of loading than the vascular region, Lau and Viano (1986). From other experiments with soft tissue Lau and Viano (1986) developed an injury metric in dependency of velocity and compression for soft tissue. As it can be seen in Figure 2.7 (b), velocities smaller than 3 m s lead to crushing injuries of soft tissue. For velocities smaller 1 m s than Lau stated that the compression criterion is the best indicator for injuries. The compression velocity of the lungs in frontal crashes is usually ≤ 3 m s for restraint occupants. (a) Lung injury severity as func- tion of the impact speed and chest compression (b) Range of validity for velocity criterion and compression criterion Figure 2.7: AIS injury severity depending to the velocity and compression for lungs and soft tissue, Lau and Viano (1986) The VC and Cmax injury criteria were developed to be assessed with ATDs. They are validated for the midline sternum chest compression measurements from Hybrid III and 16 2 Literature Research 2.6 Lung Parenchyma Experiments are not sensitive to modern restraint systems. For example modern seatbelts lead to asym- metrical loads of the left and right side of the thoracic cage. Thus there is a need for new tools and criteria which take this specific loadings better under account. To meet these requirements, the intendant successor of Hybrid III the Test device for Human Occupant Restraint (THOR) is able to measure three dimensional displacements at four different points of the thorax, Mendoza-Vazquez (2012). Through the availability of more detailed deformation data, Song et al. (2011) recently suggested the Combined Deflection Crite- rion (DC). This criterion takes the sternal compressions as well as the different deflection in the right and left sides of the ribcage into account. The dynamic human thoracic responses and injuries associated with frontal impacts, side impacts and belt loadings were investigated by Ruan et al. (2003) using an FE-HBM . He compared the simulated results with PMHS experiments. As injury criterion Ruan used the VC and compared the VC with the occurring pressure. For the velocity and deflection two points were chosen related to the load case. Ruan found out, that the lungs had the lowest pressure of all organs, whereby the left lung had a higher pressure than the right lung. He assumed that a VC ≥ 1.58 indicates some tissue damage. This VC correlated well with organ damage seen in the experiments. A comparison of the VC with the pres- sure showed, that a pressure of 16 kPa indicates a lung laceration injury. Gayzik et al. (2007) developed an FE based injury metric for PC using CT images of injured rat lungs. Gayzik induced PC on rat lungs through direct impacts with an impact velocity of 5 m s on in vivo rat lungs. CT scans were taken 24 hours, 48 hours, one week and one month after the impact happened. A numerical simulation was performed of the experiments with the impactor, the rat lungs and surrounding structure. Several injury predictors, like e.g. maximal shear strain and maximum shear stress, were used and com- pared to the CT images (cf. Figure 2.8). As it can be seen in Figure 2.8 the CT images of the PC and the calculated injury metric correlated well. He obtained the best results for the maximal principle strain ∗ strain rate (εmax ∗ ε̇max) for the PC after 24 hours. 2.6 Lung Parenchyma Experiments The availability of mechanical properties of human lung parenchyma is very limited. De- spite reviewing literature extensively it was only possible to identify a few publications with experiments concerning lung tissue. As mentioned in Section 2.4.2 these studies were used from several authors as input for the material properties of their lung models. In this section these experiments are summarized to give an overview of the state of the art and the way these experiments were designed. 2.6.1 Hoppin 1975 - Properties of Lung Parenchyma in Distortion The purpose of the study by Hoppin Jr. et al. (1975) was to provide a basis for comparing analytical models and to provide data for evaluating and developing models for lung dis- 17 2 Literature Research 2.6 Lung Parenchyma Experiments (a) Rat lungs with impactor, shaded contour is maximum principle strain (b) Best calculated injury predic- tion (light grey) and CT-based injury (dark grey) Figure 2.8: Rat lung model with maximum strain (a) and best correlation metric (b) for εmax ∗ ε̇max, Gayzik et al. (2007) tortion. Therefore Hoppin cut cubes out of frozen lungs from healthy mongrel dogs’ (15 to 25 kg) frozen lungs. The cubes had a side length of 1 cm and masses from 1.2 to 1.4 g. In order to impose nearly uniform stresses they placed sixteen small hooks into each of the six surfaces before loading with different weights. The overall dimensional changes of the specimen were recorded with a Linear Differential Transducer (LDT) and a camera (cf. Figure 2.9 (a)). The results showed hysteresis behaviour from loading to unloading for symmetrical load- ing and only moderate differences of extensibility of the axis. Under asymmetrical load- ing the behaviour of the lung tissue was similar but with greater compliance and less hysteresis (cf. Figure 2.9 (b)). Hoppin et al. assumed that the lung parenchyma showed elastic and slightly hysteresis behaviour but the experimental data do not allow a prediction about isotropy. 2.6.2 Vawter 1978 - Elasticity of Excised Dog Lung Parenchyma Vawter et al. (1978) used an experimental procedure which had previously been developed by Lanir and Fung (1974) for soft tissue measuring. Vawter et al. measured the stress strain relationship on rectangular slabs of excised dog lungs under different conditions. Thus the group extracted lung tissue from anesthetized mongrel dogs (25-30 kg) and cut it into slabs with a dimension of 5.0 x 5.0 x 0.5 cm. To minimize the effects of boundary conditions and gravity Vawter et al. used slabs and tested them under uni- and biaxial loading conditions (cf. Figure 2.10 (a)). Compared to Hoppin Jr. et al. (1975) these had the advantage that the loads could have been applied easily with clamps. Following the theory of St. Venant the edge effect is limited to the boundary region. The strain rate 18 2 Literature Research 2.6 Lung Parenchyma Experiments (a) (b) Figure 2.9: Schematic drawing of the tissue testing set-up (a); extension ratio under sym- metrical loading for all axes against normalized tensile force (b), Hoppin Jr. et al. (1975) was varied over a factor of 250 with a speed of 0.03 cm s . Each specimen was stretched for about 100 to 200 times. Vawter et al. discovered that the dog lung tissue has highly non-linear stress versus strain and slightly hysteresis behaviour. Figure 2.10 (b) shows the stress versus strain curve for uniaxial loading conditions for eleven different specimens. Vawter et al. also detected a slight effect of biaxial loading conditions. Under uniaxial loading the tissue was stiffer for high tension values than under biaxial loadings. (a) (b) Figure 2.10: Slabs specimen testing under uni- or biaxial loading conditions Fung (1993) (a); stress versus strain curve for eleven specimens under uniaxial loading condition (b) Vawter et al. (1978) 19 2 Literature Research 2.7 Validation Experiments 2.6.3 Zeng 1987 - Measurement of the Mechanical Properties of the Human Lung Tissue Zeng et al. (1987) measured the mechanical properties of human lung tissue in a state of biaxial tension. The human lungs were obtained through an autopsy within 48 hours after death. After degassing the lungs they were frozen and specimens with a dimension of 3 x 3 x 0.4 cm were cut out. Force and deformation were measured with an optical device like Vawter et al. (1978) and the forces in x- and y-direction were also recorded with a force transducer. The specimens were loaded with a fixed load in x-direction and with a sinusoidal load with a fixed amplitude and a frequency of 0.04 - 0.002 Hz in y-direction. The stress-strain-relationship was similar to Vawter’s et al. with a strong non-linearity and hysteresis (cf. Figure 2.11). Zeng et al. also recorded a rapid creep in the first few seconds to the extent of three to six per cent. Figure 2.11: Stress versus strain curve for a human lung tissue specimen subjected to a fixed load in x-direction and sinusoidally varied stretch in the y-direction, Zeng et al. (1987) The lung parenchyma gets characterized as a highly non-linear pseudo-elastic material. Zeng et al. also derived a strain-energy function according to Fung et al. (1978) to describe the experimental curves by an equation. Comparing the constant C, which determines the overall stress level in the strain-energy function, in the data from Vawter et al. and from Zeng et al. shows that the human lung tissue is stiffer than the dog parenchyma. 2.7 Validation Experiments A lot of different studies for thoracic modelling and validation available (cf. Section 2.4.2) but the focus in them lies on the thoracic response and not on soft tissue and visceral val- 20 2 Literature Research 2.7 Validation Experiments idation. The only one study which is suitable for validation of a lung model was by Hayamizu et al. (2003). He measured the dynamic response of an impactor on lungs. These experi- ments were used for lung material model validation. Therefore these experiments are the essential part of this thesis and they are summarized in this section. The table top tests by Kent et al. (2004) are frequently used by different authors as valida- tion experiments (e.g. Mendoza-Vazquez et al. (2012); Murakami et al. (2006); Pipkorn and Kent (2011)), cf. Section 2.4.2). These experiments were also chosen for this thesis to prove the biofidelity of the modified lung material model. 2.7.1 Hayamizu 2003 - Measurement of Impact Response of Pig Lung The aim of Hayamizu’s experiments was to record swine lungs’ dynamic response for lung model validation. For this he placed the lungs on a table and dropped an impactor with a diameter of 80 mm and a weight of 1.7 kg from different heights onto the lungs. Hayamizu et al. chose the heights for dropping the impactor so that the impact speeds were 3.5, 4.4, 5.4 and 6.1 m s , Hayamizu et al. (2003). To keep the kinetic impact energy constant the weights of the impactor were changed. For the speeds of 3.5, 4.4, 5.4 and 6.1 m s the weights were 2.8, 1.7, 1.7 and 0.9 kg, respectively. The initial thicknesses of the lungs were 129± 16 mm. Hayamizu et al. recorded the response of the lungs with a high speed camera and measured displacement and force. Three pictures of the impact process recorded with the high speed camera can be seen in Figure 2.12. The pictures show the lung before the impact happens (a), at the deepest impact point (b) and the state after the impact process (c). (a) moment before impact happens (b) highest compression after 0.4 s (c) end of impact process after 0.9 s Figure 2.12: High speed impact pictures of a swine lung after different times for 5.4 m s , Hayamizu et al. (2003) Unfortunately, the pictures are of poor quality but the principle deformation behaviour of the lung can be seen. The lung is positioned with its lower limb in the front of the picture. The lower limb is moving upwards when the impactor compresses the lung. The highest point of the lower limb is reached at the moment of highest compression through the impactor in the middle of the lung (cf. Figure 2.12 (b)). It can be assumed that the upper limb in the background of the picture is also moving upwards. 21 2 Literature Research 2.7 Validation Experiments The experiments were carried out five times for the impact speed of 5.4 m s . For the other impact velocities the experiment was carried out only twice each. The graphs for the impact speed of 5.4 m s can be viewed in Figures 2.13 (a) to (c). The graphs of the other speeds are shown in Figure 2.13 (d). (a) Time - displacement plot for 5.4 m s (b) Time - force plot for 5.4 m s (c) Deformation - force plot for 5.4 m s (d) Deformation - force plot for 3.5, 4.4 and 6.1 m s Figure 2.13: Experimental results for lung impact experiments by Hayamizu et al. (2003) Figure 2.13 (a) shows the time displacement curve of the impactor. The displacement starts on a height of approximately 180 mm and decreases continuously down to the lowest point after t ≈ 0.042 s with a displacement of approximately 20 mm. Afterwards the graphs of the different experiments increase to an average displacement of 125 mm, except for one graph which only increases up to 60 mm. Figure 2.13 (b) shows the time versus force plots of the experiments. It reveals that the first force occurs after t ≈ 0.018 s. With one exception of a short stagnation after 0.02 seconds the graphs increase continuously up to a peak force between 720 N and 980 N after t ≈ 0.042 s. Afterwards all graphs decrease continuously to 0 N after 0.06 to 0.08 seconds. The deformation over the force is plotted in Figures 2.13 (c) and (d) for the experiments with 5.4 m s and for the other speeds, respectively. In Figures 2.13 (c) it can be seen that the deformation of the lung is up to 93 %. At a compression rate of approximately 85 % a peak force of nearly 1000 N occur. The lower peak force of 680 N is conspicuously smaller at the same compression rate. In Figure 2.13 (d) it can be seen that the highest force of F ≈ 1080 N occurs at an impact speed of 3.5 m s . At F ≈ 550 N the average peak force for 4.4 and 6.1 m s is clearly less, for 5.4m s the forces are between 700 and 990 N. For all speeds the peak forces occur between 80 % and 90 % of deformation. 22 2 Literature Research 2.8 Material Models 2.7.2 Kent 2004 - Thoracic Response to Dynamic, Non-Impact Loading from a Hub, Distributed Belt, Diagonal Belt and Double Diagonal Belts Kent et al. (2004) performed table top tests with 15 PMHS to quantify the force deflec- tion response of the same thorax under different loading conditions with dynamic, non- impact, restraint-like loadings. The thoracic response corridors were measured for the four loading conditions single belt, double diagonal belts, distributed and hub loading. The schematic load cases are shown in Figure 2.14. The subjects were placed on a rigid table and a high speed material testing machine was used to provide controlled chest deflection at a rate of 1m s which corresponds to restrained frontal-impact PMHS sled testing experiments at a speed of 48 km/h. (a) Diagonal belt (b) Double diagonal belt (c) Hub (d) Distributed Figure 2.14: Load cases for table top tests, Kent et al. (2004) A load cell was placed between the back of the PMHS and the table to measure the reac- tion force. The thoracic response was the midline sternum deflection recorded by a string potentiometer attached to the hub, belts or band. The subjects were tested four times at a nominally non-injurious level. After the non- injurious loading conditions one single loading condition was used for a final, injurious test by nominal 40 % chest deflection. The recorded mid-sternum deflection and the posterior force were scaled to a 50th per- centile male-based size and modulus. The plotted corridors for chest deflection versus reaction force for the four load cases are shown in Figure 2.15. The corridors are showing the ±1-standard-deviation and were developed for 20 % deflection level of the 50th percentile male. 2.8 Material Models In Section 2.6 three experiments with lung parenchyma were described. The recorded material behaviour in these experiments was highly non-linear and hysteresis. Also rapid creep was recorded in the first few seconds. To describe this complex material behaviour mathematically is a challenging task. The three material models most frequently used for lung tissue will be presented in this 23 2 Literature Research 2.8 Material Models Figure 2.15: Force versus deflection and the corridors for hub, single and double diagonal belts and distributed loading conditions; the coefficients shown in each plot refer to the quadratic equation y = αx2 +βx, Kent et al. (2004) section. They can describe the experimental stress versus strain curves including hystere- sis behaviour recorded in experiments. As mentioned before the values used by different authors for lung modelling are summarized in Table 6.1. Most of them refer to these ma- terial models. These material models were used in this thesis for the simulation of a lung model aiming to figure out which material model represents lung tissue best. 2.8.1 Strain-Energy Function In a publication and later on in his book "Biomechanics - Mechanical Properties of living tissue" Fung et al. (1978) proposed using a pseudo-elastic material model for living soft tissue. The idea behind a pseudo-elastic material is that a material behaves differently for loading and unloading. Therefore the material is treated with the pseudo-elastic material model as an elastic material for loading and another elastic material for unloading. It is a convenient description of the stress versus strain relationship in specific cyclic loadings. Fortunately the advantage of the pseudo-elastic material model is, that the very complex property of tissue is more simply described. This concept describes the hysteresis of liv- ing tissue like lung parenchyma easily. Several authors proposed different analytic expressions to describe the pseudo-elastic ma- terial model based on idealized models of the lung structure (e.g. Fung et al. (1978); Hoppin Jr. et al. (1975); Lee and Frankus (1975); Vawter (1980); Vawter et al. (1979)). 24 2 Literature Research 2.8 Material Models The most used equation was the one proposed by Vawter (1980) based on the the theory of strain and surface energy by Fung et al. (1978). For the energy W applies: W (I1, I2) = C 24 e(αI2 1+β I2) (2.2) where C, α and β are material constants determined by experiments and4 is the typical alveolar diameter when unstressed. I1 and I2 are the known strain invariants. Fung also proposed a relationship for the surface energy density E given as: E = 12 4 ∫ A 1 γ(A)dA (2.3) where the surface area A is given as: A2 = 4 3 (I1 + I2)−1 (2.4) The surface tension γ varies with the area, but since the exact variation of γ is not well known, Vawter simplifies the relationship by the following: γ =C1AC2 (2.5) From Equation 2.3 and 2.5 follows: E = 12C1 4(1+C2) (A1+C2−1) (2.6) where C1 and C2 are constants and A is given in Equation 2.4. Finally, the strain and surface energy equation for lung parenchyma follows from Equa- tions 2.2 and 2.6 as: W (I1, I2) = C 24 e(αI2 1+β I2)+ 12C1 4(1+C2) (A1+C2−1) (2.7) This equation is used in LS-DYNA (LSTC, Livermore, CA, USA) as material model 129 MAT_ LUNG_ TISSUE, Livermore (2012). Additionally Fung et al. (1978) proposed an analytic expression for the strain-energy func- tion. This one was also used by some authors (e.g. Lee and Frankus (1975); Vawter et al. (1979); Zeng et al. (1987)) and should be mentioned briefly. ρ0W = 1 2 ce(a1E2 x+a2E2 y+2a4ExEy)+ 1 2 ce(a1E2 x+a2E2 z +2a4ExEz)+ 1 2 ce(a1E2 z +a2E2 y+2a4EyEz) (2.8) Here c, a1, a2 and a4 are material constants and Ex, Ey and Ez are short forms of Exx, Eyy and Ezz. If the lung tissue is assumed to be isotropic, then: a1 = a2 (2.9) and the number of constants gets reduced to three, Zeng et al. (1987) 25 2 Literature Research 2.8 Material Models 2.8.2 Viscoelastic Material Behaviour Materials are called viscoelastic when they exhibit both viscous and elastic characteristics when undergoing deformation. Elastic materials strain immediately when stress is applied and return quickly return to their original state when the stress is removed. Contrary, viscoelastic materials strain linearly when stress is applied. Holmes demonstrated that soft-tissue can be described with viscoelastic material behaviour, Holmes (1986). The equation for viscoelastic materials is given in Equation 2.10, Herrmann and Peterson (1968), G(t) = GL +(GS−GL)ε −β t (2.10) with G = shear modulus, GS = short term shear modulus, GL = long term shear modulus and β = decay constant. Several authors used this material model for internal organs in their finite element analysis (e.g. Lizee et al. (1998); Plank et al. (1998); Roberts et al. (2005); Ruan et al. (2003)). The values for the coefficients were adjusted to experimental results from PMHS impact experiments, Plank et al. (1998), or obtained by the help of an identification process, Lizee et al. (1998). The Equation 2.10 is the background of the material model 006 MAT_VISCOELASTIC from LS-DYNA, Livermore (2012). 2.8.3 Low Density Foam This material model was first used by Wang (1995) for the visceral in his FE-HBM (cf. Section 2.4.2) and it is also used in THUMS v3.0 for the internal thoracic organs, Kimpara et al. (2005). (a) (b) Figure 2.16: Kelvin-Maxwell model (a), Wang (1995); behaviour of the low density foam model and the influence of the shape and decay factor (b), Livermore (2012) Low density foam is a material model of LS-DYNA (057 MAT_LOW_ DENSITY_FOAM) and is related to strain-energy material models. This model can be seen as a Maxwell fluid which consists of a damper and a spring in series (cf. Figure 2.16 (a)), Livermore (2012). The influence of the hysteretic unloading factor and the shape factor is illustrated in Figure 2.16 (b). With this factors it is possible to affect the unloading behaviour. 26 3 Methods The theoretical background given in Chapter 2 is the basis of the simulations which are presented in this chapter. The thoracic viscera from the FE-HBM THUMS v3.0 modified by Mendoza-Vazquez et al. (2012) was isolated and used for an investigation of the material models presented in Section 2.8. Therefore the experiments of Hayamizu et al. (2003) were modelled with the pre and post processor LS-PrePost (v3.2, LSTC, Livermore, CA, USA) and the finite element solver used was LS-DYNA (R6.0, LSTC, Livermore, CA, USA). To optimize the parameter of the material models the optimization tool LS-OPT (4.2, LSTC, Livermore, CA, USA) were used. Afterwards, the modified material model was implemented in the FE-HBM THUMS v3-M and the biofidelity was approved with a simulation of the table top test by Kent et al. (2004). 3.1 Model The following two paragraphs depict the models used for the simulation of the experi- ments to validate the lung model and to prove the thoracic response. 3.1.1 THUMS The model used for the simulations was originally THUMS version 3.0 presented in Sec- tion 2.4.3 in sitting posture from a 50th percentile male occupant with a mass of 77 kg and a stature of 1.75 m. The bones were represented by shell elements for the cortical bones and by solid elements for trabecular bones. The THUMS version AM50 occupant can be seen in Figure 3.1 (a). This model is shown as it was used for the simulation of the table top tests without lower limbs. In this model the skin, flesh and bones were removed partly, to show the inner structure. As mentioned in Section 2.4.3, this model was modified in some parts by Mendoza- Vazquez et al. (2012) to increase numerical stability and robustness and to improve the biofidelity. This modified model THUMS v3-M was used as the basis for the simulations in this thesis. 3.1.2 Thoracic Organs For the simulations of the experiments by Hayamizu et al. (2003) the viscera were isolated from THUMS v3-M. As it can be seen in Figures 3.1 (b) and (c) they were modelled by two volumes out of solid elements. These volumes were surrounded with a membrane out of shell elements to apply the contact between the viscera and the thoracic wall. 27 3 Methods 3.2 Impact Response Tests following Hayamizu (a) lower limbs, skin, flesh and bones has been removed partly (b) perspective, anterior view (c) superior view Figure 3.1: THUMS v3-M and isolated thoracic viscera A viscus consists of 1 389 nodes and 962 elements and the volume of each viscus is V ≈ 3.44 l with a density of ρ = 1∗103 kg m3 . Thus, the mass of each viscus is m≈ 3.44 kg. The material model was LS-DYNA MAT_ 057 LOW_ DENSITY_ FOAM (cf. Section 2.8.3). The values of the variables of this material can be seen in Table 3.1 and the stress versus strain curve is plotted in figure 3.2. Table 3.1: Low density foam material variables from LS-DYNA for thoracic viscera from THUMS v3-M, Mendoza-Vazquez et al. (2012) Variable Description Value RO Density [ kg m3 ] 1.0∗ e3 E Youngs’ modulus [ N mm2 ] 0.1 HU Hysteretic unloading factor 0.1 BETA Decay constant to model 0 creep in unloading DAMP Viscous coefficient to 0.1 model damping effects SHAPE Shape factor for unloading. Values 1 less than 1 reduce, greater than 1 increase energy dissipation 3.2 Impact Response Tests following Hayamizu The impact experiments from Hayamizu et al. (2003), summarized in Section 2.7.1, were the only experiments suitable for lung model validation simulations. Therefore these 28 3 Methods 3.2 Impact Response Tests following Hayamizu Figure 3.2: Load curve for low density foam following Mendoza-Vazquez et al. (2012) for thoracic organs experiments were used for the comparison and evaluation of material models for lungs. In this section the simulation of the experimental set-up will be explained and a short introduction of the used material properties and the modification process will be given. 3.2.1 Experimental Modelling One of the isolated volumes was placed slightly over a rigid table and loaded with gravity in z-direction to find the contact between the viscus and the table in a steady state. The state when the viscus was in equilibrium with the table was used as the new initial state for the impact simulations. The impactor was modelled as a cylinder like in the experiments by Hayamizu et al. (2003) with a diameter of d = 80 mm containing of solid elements. The chosen material for the impactor was elastic (LS-DYNA MAT_ ELASTIC_ 001) with: ρ = 7.85 ∗ e3 kg m3 , E = 40 kN mm2 and ν = 0.3. The height for the impactor resulted from the chosen density to fit with the experimental weight. For the main experiments with an impact speed of v = 5.4 m s the weight was 1.7 kg. The volume of a cylinder is calculated by following equation. V = A∗h = 1 4 π ∗d2 ∗h (3.1) In this equation A is the base of the cylinder and h the height. The weight is calculated by: m =V ∗ρ (3.2) From Equation 3.1 and 3.2 it follows for the height: h = 4 π ∗ m ρd2 (3.3) with the values for d,ρ and m follows: h = 4 π ∗ 1.7 kg 7.85∗ e3 kg m3 ∗ (80 mm)2 = 43.08 mm (3.4) 29 3 Methods 3.2 Impact Response Tests following Hayamizu The weights for the other impact speeds were adjusted by changing the density instead of the height. The densities are summarized in Table 3.2. Table 3.2: Impactor densities for adjusted masses Impact speed [m s ] Impactor mass [kg] Density [ g cm3 ] 3.4 2.8 13.145 4.4 1.7 7.85 5.4 1.7 7.85 6.1 0.9 4.225 The model of the experimental set-up can be seen in Figure 3.3. The viscus is already in a steady state and in contact with the table through gravity. In the experiments the impactor was guided on a rail, thus it could only move in z-direction and was fixed for rotating around each axes and translation in x-y-directions. The same boundary conditions were added to the impactor. Figure 3.3: Model of the experimental set-up by Hayamizu et al. (2003) (isometric view) According to the densities from Table 3.2 the initial speeds were referred to the impactor. The results from Hayamizu et al. (2003) (shown in Figure 2.13) were digitised with MAT- LAB (R2012a, The MathWorks Inc., Natick, Ma, USA). As mentioned before, one of the recorded time versus displacement curves from the experiments with 5.4 m s was much smaller than the others. Therefore this curve was considered as outlier and was ignored for the rating of the simulations. As it can be seen in Figure 2.13 (a) the time versus displacement curve starts at a height of approximately 180 mm, though the thicknesses of the lungs were only 129± 16 mm, which were reached after approximately 0.02 seconds. It can be seen in Figure 2.13 (b) that the first forces occurred after 0.018 - 0.02 seconds. This suggests that the plotted curves already show the displacement of the impactor before the impact on the lungs hap- pens. This hypothesis is also supported by the high speed pictures of the impact procedure in Figure 2.12. The highest compression occurs between 0.3 s and 0.4 s and in contrary to the plotted curves not between 0.4 and 0.5 s. Therefore the data sets needed to be adjusted to the impact time. Furthermore, the displacement height had to be scaled from the thickness of the experimental lungs to the maximum model thickness in the impact 30 3 Methods 3.2 Impact Response Tests following Hayamizu area of 100 mm (thickness in equilibrium after gravity load). The modified curve is plotted in Figure 3.4. The time versus force curves were also adapted to the changed time. These curves were used for for the evaluation of the model response. Figure 3.4: Time versus displacement curve adjusted to the impact time and scaled to the model size 3.2.2 Contact Conditions As contact conditions between the table and the viscus the contact automatic surface to surface were chosen. The viscus was the slave segment and the table the master segment. To figure out the best contact conditions between the viscus and the impactor several simulations were run. Finally, the contact condition automatic surface to surface was chosen according to Gayzik et al. (2011). For this the viscus were defined as master and the impactor as slave segment. Because there was no lubricant in the experiments between the impactor and the viscus, a high friction coefficient was chosen by Al-Mayah et al. (2008b) and Loring et al. (2005) at 0.3. To prevent the impactor of penetrating the lung model, it was partly necessary to add a soft constraint formulation with the sub-option pinball edge to edge contact. With this option the calculation of contact stiffness was based on stability consideration and the timestep was taken into account. This prevents penetration when soft materials are in contact with stiff materials. 3.2.3 Material Models The three most common material models presented in Section 2.8 were used as material properties of the lung model for the simulation of the impact experiments by Hayamizu et al. (2003). The initial material parameter were chosen from publications using this ma- terial properties for lungs. An overview of different material properties used in literature can be viewed in Table 6.1. The values for the material models which were used for the simulations are summarized in Tables 3.3 and 6.1. Three different material properties were chosen for viscoelastic 31 3 Methods 3.2 Impact Response Tests following Hayamizu material behaviour, five for the strain-energy function and one for low density foam. It is conspicuously that the values for the materials are widely spread between different pub- lications. E.g. the bulk modulus K varies from 0.05 N mm2 , Zhao and Norwani (2004), to 2880 N mm2 , Roberts et al. (2005). No values were given for the typical alveolar diameter by Vawter (1980) and Zhao and Norwani (2004), therefore the typical alveolar diameter of a human male were chosen at 0.2 mm after Tenney and Bartlett (1967). Table 3.3: Values for the different material models from literature, Mendoza-Vazquez et al. (2012); Plank et al. (1998); Roberts et al. (2005); Ruan et al. (2003); Stitzel et al. (2005); Vawter (1980); Zhao and Norwani (2004) Material Model Symbol Viscoelastic Strain-energy function Low d. foam [Unit] Plank et al. Roberts Ruan Stitzel et al. Vawter Zhao Mendoza Density ρ [ kg m3 ] 917 600 600 118 365 700 1000 Bulk modulus K [ N mm2 ] 2.875 2880 2.6 0.1124 0.1124 0.05 Decay constant β 100 0.1 0.25 0 Young’s modulus E [ N mm ] 0.1 Viscoelastic Short time shear modulus GS [ N mm2 ] 7.387∗ e−3 7.39 0.022 Long time shear modulus GL [ N mm2 ] 2.358∗ e−3 2.36 0.008 Strain-energy Function Material Coefficient C [ N mm ] 5.035∗ e−4 2.45∗ e−3 3.88∗ e7 Material Coefficient α 8.227∗ e−2 0.183 5.85 Material Coefficient β -2.46 -0.291 -3.21 Hyperelastic coefficient C1 [ N mm ] 6.535∗ e−6 1.93∗ e−5 1.27∗ e−8 Hyperelastic coefficient C2 2.876 2.71 2.71 Typical alveolar diameter 4 [mm] 0.0702 0.2 0.2 Low density foam Tension cut off stress TC 1∗ e14 Hysteretic unloading factor HU 0.1 Viscous coefficient DAMP 0.1 Shape factor for unloading SHAPE 1 Most of the material models from literature did not terminate normally. Therefore, the values had to be adjusted manually to find a working model, before a coefficient study for optimizing the model response could be carried out. 3.2.4 Coefficient Study After the first simulation part had been completed and a working model set-up had been found for the different material models, a parameter study was used to optimize the ma- terial models. The simulated material models with values from literature showed a gap to the time versus displacement and force behaviour of the experiments by Hayamizu et al. (2003). There- fore a parameter study was used to optimize the values from the different material models aiming to receive a simulated time versus displacement and force response similar to the experimental results. For the optimization the software LS-OPT was used. It was necessary to define the mate- rial coefficients as variables in the k-file (the LS-DYNA script file of the model), so that LS-OPT was able to identify them as variables. Afterwards the range and starting values 32 3 Methods 3.2 Impact Response Tests following Hayamizu had to be defined for each variable as well as the optimization algorithm and the number of iterations. The density was partly changed to investigate the influence of all parameter to the model response. Because the weight of the whole THUMS model had to keep constant, the den- sity of the final material model had to be ρ = 1000 kg m3 as in the original THUMS version. The optimization argument was the minimization of the Mean Square Error for dis- placement (MSEdis) of the impactor and the modified time versus displacement curve by Hayamizu et al. (2003) as well as the Mean Square Error for force (MSEforce). 3.2.5 Load Curve Study For the material model low density foam the load curve was a further tuning opportunity besides the material coefficient optimization. Therefore a pre-processor had to be im- plemented in LS-OPT which calculated the new curve depending on curve variables. A schematic example can be seen in Figure 3.5. Figure 3.5: Schematic drawing of curve optimization with LS-OPT For the input curve either fixed points (cf. Figure 3.5 point 0,0), points with a fixed x or y-value (point 0.8, y3) or points with flexible x and y-values (e.g. point x1, y1) could be defined. In addition to to the initial load curve of THUMS v3-M by Mendoza-Vazquez et al. (2012) the experimental stress versus strain curves of lung parenchyma (cf. Section 2.6) were digitised and also used as load curves (cf. Figure 5.1). Because Radford’s and Rem- ington’s (1957) original publication was not available, the stress versus strain curve was obtained from Matthew’s and West’s (1972) publication. The load curve with the most promising model response was used as the initial curve for the curve optimization. For the curve optimization only the first point (0, 0) was defined as a fixed one. Four points were defined as alterable in x and y direction, which means eight variables had to be defined. It should be noted that the number of simulations per iteration and the calculation time increased exponential with an increasing amount of variables. The proposed minimum 33 3 Methods 3.3 Model Validation - Table Top Tests simulations for each iteration were already 68 for eight variables. Therefore multiple optimization runs were executed with adjusted ranges for the variables. 3.3 Model Validation - Table Top Tests The optimized material properties were implemented as modified lung material in the THUMS v3-M. With this modified model the table top tests, described in Section 2.7.2 by Kent et al. (2004), were simulated to approve the biofidelity and to investigate the influence of the tuned material to the thoracic response. Thus the model THUMS v3-M was placed on a table, the pelvic slightly lower than the back and the head 87 mm higher. Gravity was applied until a steady state was reached (cf. Figure 3.6). Belts and the band were applied using the Seatbelt Fitting add-on in LS-PrePost. Figure 3.6: Experimental set-up of the table top tests by Kent et al. (2004) with a single belt; the arrow points to the place where the displacement was recorded On both ends of the belts and the hub a displacement was applied, so that the compres- sion rate matched the experimental values. For the force versus compression response, the vertical component of the contact force between the plate and the skin at the back of the model was used. The chest compression was obtained from the displacement of the belts and the band above the sternum in the height of the third rib (cf. black arrow in Figure 3.6). 34 4 Results In this chapter the results of the simulations described in the chapter before will be given. The focus lies on the model response of different material models to the impact speed of 5.4 m s . This is due to the limited data for the other impact speeds of the experiments by Hayamizu et al. (2003). 4.1 Material Models In this chapter the results of the simulations of the three different material models will be given. The material models strain-energy function and viscoelastic material behaviour were simulated with different material properties from literature. 4.1.1 Strain-Energy Function In the simulation of the experiments all material models of strain-energy function termi- nated with an error. In Figure 4.1 some simulations are shown in the state before termi- nation. The material model according to Vawter (1980) (Figure 4.1 (a)) terminated after 0.019 s because of negative volume in some elements. The viscus was deformed under the impactor 92 mm at the moment of termination. The surrounding of the viscus nearly remained at its initial thickness and a sharp edge was formed between the compressed part under the impactor and the surrounding. The upper and lower limb nearly stayed in their original positions and did not move upwards as observed in the experiments (cf. Figure 2.12). The material models according to Stitzel et al. (2005), Gayzik et al. (2007) and Gayzik (2008) already terminated after 0.009 s, 0.015 s and 0.011 s, respectively. The progress of deformation of the viscus was not sufficient enough to make a statement about the de- formation behaviour. The simulation with the material model according to Zhao and Norwani (2004) terminated immediately after the calculation was started. 4.1.2 Viscoelastic Material Behaviour From the three different material data for the viscoelastic material model summarized in Table 3.3, the models from Plank et al. (1998) and Ruan et al. (2003) terminated with an error because of negative volume in some elements. The deformation of the model in the state of error termination can be viewed in Figures 4.2 (a) and (b). The model according to Plank et al. terminated after 0.006 s. The model was deformed mainly in the area under the impactor since the rest of the viscus was still in its initial state. A sharp edge evolved between the impact area and the rest of the model (cf. Figure 4.2 (a)). 35 4 Results 4.1 Material Models (a) Vawter (1980) termination after 0.019 s (b) Stitzel et al. (2005) termination after 0.009 s (c) Gayzik et al. (2007) termination after 0.015 s (d) Gayzik (2008) termination after 0.011 s Figure 4.1: Simulated experiments for different values of the strain-energy function at the moment of error termination The model after Ruan et al. terminated after 0.015 s. The impactor deformed the model 62.7 mm before the simulation terminated. Compared to the deformed model of Plank et al. the whole model underwent a deformation through the impactor. The deformation ran continuously from the impact area smoothly to the rest of the model. The upper and lower limbs stayed in their original positions on the table (cf. Figure 4.2 (b)). The material model according to Roberts et al. (2005) terminated normally. In Figures 4.2 (c) and (d) the model is shown after 0.003 s and after 0.029 s, respectively. The highest compressions already occurred already after 0.003 s with a deformation of only 8.5 mm. Afterwards the impactor was rebounded by the viscus back in z-direction. The viscus also began to bounce off the table through the impact energy (cf. Figure 4.2 (d)). 4.1.3 Low Density Foam The material model low density foam with the material parameters from Mendoza-Vazquez et al. (2012) terminated normally. The model in its initial state can be seen in Fig- ure 4.3 (a). In Figure 4.3 (b) the model is shown after 0.035 s in the state with the highest compression. In the front the impactor is partly covered by the lung tissue. The upper and lower limb moved upwards through the energy of the impactor. In Figures 4.3 (c) and (d) the viscus is shown after 0.1 and 0.2 s. The experiments by Hayamizu et al. (2003) were recorded only for 0.1 s. In the simulation the viscus and the impactor reached a steady state after 0.2 s in which the viscus was still a little bit compressed. 36 4 Results 4.1 Material Models (a) Plank et al. (1998) termination after 0.006 s (b) Ruan et al. (2003) termination after 0.015 s (c) Roberts: maximum deformation after 0.003 s (d) Roberts: bouncing viscus and impactor after 0.029 s Figure 4.2: Simulated experiments for different values of the viscoelastic material model at the moment of error termination (a) and (b) and after some particular elapsed time for normal termination by Roberts et al. (2005) (c) and (d) The time versus displacement plot of the simulated model with low density foam material properties for an impact speed of 5.4 m s can be seen in Figure 4.4 (a). Here the dis- placement is assigned negative because the impactor moved against the z-direction. The impactor compressed the viscus from 100 mm to a minimum of 27.6 mm after 0.035 s. That means the impactor had a displacement of 86.3 mm. The results of Hayamizu et al. (2003) scaled to the size of the model are also plotted in 4.4 (a). It can be acknowledged that Hayamizu recorded a displacement of the impactor of 85 mm. In the experiments the impactor was reflected by the lungs up to the original height. Con- trary to the experiments, in the simulation the impactor was reflected only to a height of 54 mm after 0.1 s. In the steady state after 0.2 s a maximum reflection to a height of 37 mm was reached (cf. Figures 4.3 (c) and (d)). In Figure 4.4 (b) the deformation versus force plot of the simulation and the experiments can be seen. In the simulation a peak force of 675 N was reached at a compression rate of 85 %. The peak force was nearly at the height of the experimental peak force at a compression rate of 85 %. The resulting forces between a deformation of 25 % and 60 % were higher than the forces in the experiments. 37 4 Results 4.1 Material Models (a) 0 s (b) 0.035 s (c) 0.1 s (