带参考信号的频域盲解卷积算法及其在卫星微振动同频相关源定量辨识中的应用

The quantitative identification of satellite micro-vibration sources can provide the guidance and basis for the suppression of micro-vibration. However, due to the strong dependence of the same frequency in vibration source signals and the complex signal transmission path, it is difficult and challenging to separate the signals and estimate the source contribution quantitatively. To this end, frequency-domain blind deconvolution algorithm with reference signals(FBDR) is proposed. Firstly, a dependent source blind separation algorithm based on the same frequency removal is proposed. The separation matrix is calculated by mixed signals which the same frequency components are removed, and it is applied to original mixed signals to obtain the estimated signals. It provides an idea for the separation of dependent source signals with same frequency components. On this basis, a complex-valued FastICA algorithm with reference is proposed. The similarity information of reference signals is introduced into the optimization iterative objective function, so as to introduce the priori information to improve the separation performance of the algorithm. Finally, the effectiveness of the FBDR algorithm is verified by simulation analysis and the satellite cabin structure excitation test. The results show that the estimation error of vibration source contribution of the proposed algorithm is significantly lower than that of the comparison algorithm. The FBDR algorithm is applied to signals obtained from the satellite micro-vibration ground test. Test results show that the error of contribution estimation is less than 3%, which meets the engineering needs. It can provide a reference basis for satellite vibration attenuation and noise reduction and on orbit vibration source control.