An Wavelet-Fractal Neural Network Used in Non-stationary Two-Dimensional Vibration Signal Monitoring

A non-stationary vibration signal monitoring system based on the combination of wavelet-fractal feature extraction and neural network classification are proposed here. Two-dimensional vibration (lateral and vertical) information is first processed by wavelet multi-scale analysis and transformed into frequency components in different resolutions. Then the fractal dimensions of components in different resolutions are computed to describe the local characteristics of the signal. At last, all the feature parameters are sent into a RBF neural network to get the vibration states of the machine. The subsystems are integrated to realize the classification automatically and adaptively. The performance of the algorithm is showed by experimental results