Citation

Lainé S, Kayvantash K. Int. J. Crashworthiness 2009; 14(3): 287-294.

Copyright

(Copyright © 2009, Informa - Taylor and Francis Group)

DOI

10.1080/13588260902986044

PMID

unavailable

Abstract

Nowadays, virtual testing is used as a traffic light. After the CAD has been defined, after a material has been selected in the norm, the simulations are realised in order to give, in some cases, just an idea of the chance of success. If they are satisfactory, the validation plan is extended to reality. Is this a robust way to judge of the robustness of a system and of its ability to answer solicitations with a reliable response in spite of its own scatterings? To answer this question, different existing methodologies have been adapted and evaluated while keeping in mind the necessity of remaining industrialisable. Firstly, investigations based on a direct stochastic approach introduce the problem of shorter and shorter cycles of development in the automotive industry. The need for some thousands of simulations for a repeatable mono-load case virtual test is a stumbling block; the computer's FLOP power increase is used up by human modelling demands like Global Human Body Model (GHBM). In the models of several other domains of engineering - civil, maritime, aeronautic, atomic - reliability analysis has been successfully coupled with explicit simulation techniques, finite elements and multi-bodies (MB). It is nevertheless the case that the question of the management of an acceptable risk is, henceforth, asked. In fact, whatever the finite element analysis (FEA) or MB is, a simulation model is a simplified mathematical model with modelling hypothesis. In the same spirit, the model of reliability analysis is not only submitted to mechanical modelling errors but to statistical modelling errors too. So, the risk management needs to quantify suppliers' qualities and process capabilities in determining the most probable (mass-) produced system; this is in order to calibrate safety margins on performance criteria and associated supervision plans, in spite of the sampling statistical representativeness and the real test uncertainties. But, even if the finality remains unchanged, improving development efficiency and time in order to sell a cheaper and more protective vehicle to probable buyers, a new question appears: 'Which is the most exposed population?'