Acoustic Source Localization in Metal Plates Using BP Neural Network

Abstract

This study introduces a methodology for detecting the location of signal sources within a metal plate using machine learning. In particular, the Back Propagation (BP) neural network is used. This uses the time of arrival of the first wave packets in the signal captured by the sensor to locate their source. Specifically, we divide the aluminum plate into several areas, design eight receiving points for receiving the excitation signal, and determine the location of each sound source. In order to train and test the machine learning network, the aluminum plate model was established using the COMSOL numerical simulation platform and the propagation of five peak waves was simulated. Correspondingly, experimental verification was carried out and a scanning laser Doppler vibrometer (SLDV) was used to build an experimental platform to collect the corresponding wave field information to obtain a data set for machine learning. The results show that the trained BP neural network can classify the sound source region in both environments.