System identification for crash victim simulation
The work presented in this thesis concerns the identification of vehicle occupant models. Mathematical models of the vehicle occupant are used in the preliminary design and development phase of vehicle design. In the design phase, the model is used to guide the decision on restraint system feasibility. In the development phase the model is used to suggest solutions to problems associated with the dummy trajectory or restraint system performance. Current methods used -to determine such models involve independent component testing. The conditions under which the components are tested are often not typical of a crash test, hence iterations of the computer model are needed to successively improve model and test correlation. In order to address these problems which cause inaccurate specification of the mathematical models, an alternative method of data set assembly for crash victim models is suggested. This alternative method is based on the techniques of system identification which allow unknown system parameters to be determined from experimental input/output data. Initially the viability of using system identification techniques to develop a valid mathematical model of the vehicle occupant and restraint system was investigated. This initial study used input and output measurementsfr om computer simulations of the occupant in frontal impact, as source data for the identification. Effects of simulated disturbances (noise corrupted output signals) and the effects of simplified model structure on the identification are also investigated. Several methods for analysing the likely errors in the identified parameters are defined and discussed in this simulation study. Results relating to the identification of seat contact and seat belt characteristics from physical tests are also presented and these are interpreted in light of the simulation results.