Bayesian inference for the interior noise problem of railway vehicles

Abstract

Design and control of interior noise of transportation systems, such as rollingstock, is important from the standpoint of comfort. Therefore, a technical field called "NVH" has been established and various noise reduction methods have been developed. On the other hand, noise reduction measures conflict with other performances such as weight reduction and space saving. In many cases, there is a large uncertainty in the causal relationship between noise control measures and their effects. Therefore, the conventional deterministic method leads to designing on the overly safe side for achieving the noise target value, which is not desirable from the viewpoint of other performance. One solution to these problems is the introduction of probabilistic design decision making, which is used in such as the evaluation of the seismic resistance of nuclear power plants and the reliability of bearings. The authors have developed a Bayesian inference method for estimating the transfer function level , and contribution of noise source type, with the aim of introducing probabilistic decision making into interior noise design. The estimation accuracy was verified by applying the method to artificially generated data. As a result, it was confirmed that the estimation was possible at a practical level.