Stochastic modeling of the vibro-acoustic behavior of production cars

Abstract

Production cars -as any industrial product- are subject to various causes of variability including process uncertainties or product diversity. Many authors have shown that vibroacoustic problems sensitivity to small uncertainties increases dramatically with frequency until only statistical approaches remain relevant in the high frequency range. Moreover, modeling uncertainties due to numerous model simplifications induce similar dispersion effects on the computed responses. Both kind of uncertainties may be addressed when using a non-parametric stochastic modeling, based on the random matrices theory. Such a modeling, appears to be very practicable for industrial vibroacoustic problems while relying on a strong mathematical background. In a first part, the application of the non-parametric modeling of uncertainties to vibro-acoustics problems will be addressed. Stochastic aspects are controlled by only 7 dispersion parameters that provide most of the dynamic behaviors that can be observed experimentally. A Monte-Carlo simulation is performed to provide converged statistics of the stochastic problem solution. In a second part, the dispersion parameters are identified so that the stochastic model fits experimental data. Frequency Response Functions of 25 production vehicles were measured for this purpose and compared to the computed results in the low frequency range (<200 Hz).