Noise contribution analysis of road noise with transfer path analysis based on neural network in the box car system

Abstract

This paper introduces a new approach for transfer path analysis (TPA) combined with neural network technique. For the TPA experiment, the box car model consisting of active and passive systems were manufactured, and road noise simulation experiment were conducted through shaker excitation at various locations. Time domain data was measured and converted into a large amount of complex-valued frequency domain data according to the acquisition block for training the neural network model. The number of layers and nodes of the model were determined by considering the physical characteristics of the system, and normalization and momentum were adjusted to improve accuracy and calculation efficiency. The reliability of the model was verified by comparing the noise contribution results with the classical TPA method. Unlike the existing TPA methods, only the operating data is needed to perform the neural network based TPA. This proposed method reduces the cost of measurement step such as impact test to identify the propagation paths of the noise.