Parametric identification of multibody models for crash victim simulation

This paper concerns the parametric identification of multibody system models used to predict vehicle occupant dynamics during crash impacts. The work considers the problem of identifying contact interactions between a standard anthropomorphic dummy and its surrounding surfaces - seat, belt, airbag, etc. - directly from measured kinematic data. The central theme is that identification is viable, even in cases where there is only limited test data available, and where serious problems of ill-conditioning arise. Though the study is restricted to simulation data, and the models are simplified, the work has potential in vehicle safety improvement, by providing validated system models for use in restraint system optimisation. The methods presented may also prove useful in the wider class of multibody system identification problems.