Understanding human injury thresholds and injury mechanisms are two of the most critical steps taken in setting up criteria for occupant protection. They rely on the testing of human surrogates in order to obtain these data. Past and present research has provided crucial data needed in order to make vehicles safer and in order to design better crash test dummies. However, more data are needed for the design of safer automobiles. Because testing of volunteers cannot be conducted at an injurious level, testing of human cadavers is the next best alternative. Unfortunately, testing of human cadavers is not only hindered by a wide scattering of data due to biological variations, but also by the fact that these tests are very difficult and expensive to conduct. Most recently, threats of contracting infectious diseases such as AIDS have made it even more difficult for researchers to conduct such studies.

     Another approach is to develop and validate numerical human models for the evaluation of vehicle crashworthiness. While a computer model may never be used to replace a physical experiment, a validated model can be used to provide general trends and to guide future experiments. With current advancements in computer technology, many design configurations can be tested on the computer before a physical test is conducted so that the predictions of the model can be confirmed.

     Finite element models have been considered to be the best tool for the modeling of objects with complex geometry, multiple material compositions and complicated loading conditions. Obviously, such a model requires that several assumptions, such as material constitutive laws, material properties, and geometrical and boundary conditions, are made before the model can be used. Without a complete knowledge of those assumptions, the model may yield errors. Thus, a model must be rigorously tested before it can be applied to a crucial task such as designing a restraint system. The purpose of this study is to develop a finite element model of a 50^th percentile male human thorax. The model will be validated against data obtained from lateral pendulum impact tests.

     The selection of developing the human thorax model for side impact has been made based on the fact that there are three side impact dummies available, namely the Department of Transportation side impact dummy (SID), biofidelic side impact dummy (BIOSID) and European side impact dummy (EUROSID). The issue regarding which side impact dummy is best suited for the assessment of automotive side impact injury has been controversial ([8]). Previously few thoracic models have been developed. Lobdell et al. ([11]) developed a lumped parameter model for the thorax and validated it against cadaver tests. Viano ([18], [19]) used this model to study the characteristics of padding and its effects on injury to the thorax. Although lumped parameter methods generally provide the easy access, control and study of each parameter, these models are much less descriptive than those using finite element methods.

     Sundaram and Feng ([15]) created a finite element model of the thorax, but they assumed that the thorax was symmetrical and validated it statically. Chen ([2]) used only beam elements to model the rib cage and lumped the internal organs in with the thoracic wall. He showed that the dynamic responses of the model favorably correlated with cadaver tests. Yang and King ([23]) presented a two-dimensional model for the study head-neck motions during lateral pendulum impact to the thorax. The boundaries between internal organs were not identified in the model.

     A recent study of a three dimensional human upper torso was created by Plank and Eppinger ([12]). This model includes the spine and the thorax. The nonlinear interactions of internal organs during impact were not included. The focus of this study is to develop a human thorax model with detailed internal organ descriptions. The major difference between this model and all previous models is its inclusion of proper nonlinear interfacing among organs. It is hoped that through continued improvements in the determination of material constitutive laws and through the use of new data for additional validations, the model may be used in the future to predict crash injuries.