Modelling vehicles NVH performance: a probabilistic approach
Abstract
The problem of the extensive exploitation of vehicle NVH measurements in real driving conditions is addressed. It is shown that through a probabilistic Bayesian approach a model of the noise level inside the cabin can be built from measurements. Two fully Bayesian algorithms are introduced: a Gaussian regression and a Scaled Mixture of Gaussians (SMOG) regression. Both are shown to be superior to a simple linear regression and the SMOG regression seems to be the most robust algorithm even in the presence of measurement noise. The model proposed can be used to get an estimation of the NVH performance of a vehicle, knowing only the operating conditions.