Impulse response feature extraction based on genetic algorithm with correlation filtering

A method based on genetic algorithm (GA) together with Laplace wavelet correlation filtering (CF) is proposed in this paper to extract impulse response features from vibration signal. In this method, real-format Laplace wavelet instead of complex one is used to reduce data redundancy of CF. Exponent-scaled relative inner product of the wavelet and the same length of signal is designed as the fitness function, GA with adaptive crossover and mutation probability is used to select the wavelet atom with the maximal correlation value. The corresponding parameters of this wavelet atom can be used to identify the features of the impulse response. This algorithm is tested by simulated signal and bearing vibration signal, it can extract features of impulse response exactly, and its efficiency is superior to that of traditional Laplace wavelet CF algorithm.