Simulation-based reliability analysis of automotive wind noise quality

You are here: Homepage > Search

Simulation-based reliability analysis of automotive wind noise quality

Zou, T. / Mourelatos, Z.P. / Mahadevan, S.

An efficient simulation-based reliability analysis method for series systems with large number of random variables is developed in this paper. The method is applied to estimate an automotive wind noise quality problem and the result is satisfactory. The efficiency of the proposed adaptive importance sampling method is better than that of the basic MCS, when the MPPs of each limit state are used as the initial samples of the multi-modal adaptive importance sampling. The efficiency can be further improved if the MPPs of only the three most important limit states are used as the starting point. It should be noted that when applied to problems with smaller failure probability, the basic Monte Carlo method requires larger number of samples whereas the proposed multi-modal AIS method needs similar number of samples if appropriate sampling function is used. The comparison between the proposed method and other existing methods shows the superiority of the former in terms of both accuracy and efficiency. In the multi-modal AIS method, the accuracy and efficiency are dependent on the starting points. For system-level reliability analysis, it is desirable to use the MPPs for each limit state as the starting points. This will require that an efficient method be used to find the MPPs. In this paper, it is found that for the door wind noise problem the component limit state is not highly nonlinear, therefore FORM can converge to find the MPP in only a few iterations. In case FORM fails to converge, an adaptive response surface approach with trust region updating may be used as an alternative.