Localization of Acoustic Emission Source Based on Chaotic Neural Networks

Because of containing several model waveforms and transmission speed of each model are various, the source signal of rub-impact acoustic emission (AE) will lead to waveform distortion in propagation process, and it is difficult to achieve exact source location by traditional time difference of arrival algorithm. A chaotic neural network technique was introduced to calculate the location of AE source. Numerous researches show that rotor rub-impact fault has sufficient non-linear features, so obtain the characteristics of the non-linear dynamics which reveal the AE source form the rub-impact data by using the chaos theory and use it as the input of the neural network to get the localization. We propose a modified Gaussian Mixed Model (GMM) with an embedded Time Delay Neural Network (TDNN). It integrates the merits of GMM and TDNN. Simulation results prove, theoretically and practically, that it can locate AE source efficiently and provide the basis for the rotor rub-impact fault diagnosis, so it has good application prospect and is worth to research further more.