An Application of Cluster Analysis to Dummy Injury Readings in a Frontal Crash 2012-01-0556
Public concern about the crashworthiness of vehicles has been continuously rising in recent years. Crashworthiness is evaluated under various crash configurations, including frontal collisions, in regulatory testing and in New Car Assessment Programs. Accordingly, vehicle manufacturers must deploy sophisticated product development strategies and redouble their engineering efforts in order to develop vehicles that satisfy the specified requirements for crashworthiness.
Computer simulation is one effective approach to resolving this issue in that it provides a valuable tool for conducting multiple parameter studies and iterations in a short period of time. However, it is no easy task for CAE engineers to analyze the large volumes of calculation results obtained in frontal crash simulations and to understand the phenomena involved. One reason is that a great deal of time is needed to understand the many calculation results comprehensively, despite the fact that frontal crash phenomena are interrelated in complex ways.
This paper presents an example of a parameter study in which cluster analysis was used effectively to examine front-seat passenger restraint systems in a frontal crash. In a cluster analysis, calculation results are grouped into clusters having the same response characteristics. Because the design variables are also similarly clustered, engineers can gain a deeper understanding of the phenomena of interest.
Two types of simulation models were used in this study. First, a multibody system (MBS) model was used as a simple mass-spring model to conduct a parameter study. This simple model made it possible to perform many calculations in a broad design space in a short period of time. Cluster analysis was then applied to analyze the calculation results of the parameter study. Using this method to understand simulation results enables engineers to formulate hypotheses about design guidelines for satisfying safety performance requirements. As the next step, a detailed finite element model that facilitated highly accurate simulations was used to verify the validity of the hypotheses.
This study made clear the influence of pelvis behavior on dummy chest injury readings, and the results also demonstrated the utility of cluster analysis in trying to understand the complex phenomena involved in a frontal crash.