Article
Improving the model-based predicted admittance of vehicles using Bayesian parameter calibration (en)
* Presenting author
Abstract:
Frequency response functions, like admittance, represent the vibroacoustic behavior of a vehicle and specify among others the quality of vehicles. In practice, numerical models like finite element models are widely utilized to aid the vibroacoustic design of a vehicle by predicting these vibroacoustic quantities of interest. This paper aims to improve the quality of the model-based predictions to meet the rising requirement on vibroacoustic vehicle behavior and customer satisfaction. Bayesian parameter calibration is conducted on a condensed Trimmed-Mass-Body model of the entire vehicle body. The Bayesian theorem is utilized to update the prior assumptions based on the information in reference data. A large number of simulations using parameter samples are performed to calibrate the dominant model parameters, which are determined by a global sensitivity analysis. Posterior distributions of model parameters are estimated by posterior samples, contributing to better model-based predictions of admittance. A good match of peaks and their frequencies between the posterior predictive distribution and the reference data of admittance in the frequency range from 15 Hz to 60 Hz is achieved. In addition, the underlying data uncertainty is quantified by posterior distributions of model parameters and posterior predictive distributions.