Abstract:To solve the problem of low precision of vibration source signals extraction caused by the strong convolution effect of vibration signals in the process of rocket shell propagation, an adaptive extraction method of rocket vibration signals based on convolution blind source separation is studied. Based on the independence among the vibration sources, the method establishes the objective function of the fourth order statistics, and uses the random gradient method to realize the adaptive updating of parameters. The inverse filter structure of the rocket shell system is optimized to obtain the original vibration source signal. By analyzing the high frequency vibration data of a carrier rocket engine, the results show that, compared with the linear adaptive blind source separation method and the nonlinear adaptive blind source separation method, the convolution adaptive blind source separation method can obviously find the frequency corresponding to turbine rotation and combustion, as well as some broadband excitation, due to considering the convolution effect and underdetermined conditions in practice, which verifies the effectiveness of the proposed method.