机械信号处理的 BSS 算法及其比较研究

The effectiveness of separation and identification of mechanical signals which usually aggregate from some independent component vibrations is crucial to successful fault diagnosis in the condition monitoring and diagnosis of complex machines. In this paper, in order to get the most effective algorithm of blind vibration signal separation, the performance of five different blind-source-separation (BSS) algorithms have been compared. A set of new evaluation criteria have been proposed. Their effectiveness in mechanical signal processing for machinery fault diagnosis has also been fully evaluated by means of simulation data similar to typical mechanical data. After the completion of simulation, two sets of vibration data generated from mechanical components in real industrial machine were used to further verify the results. The outcomes of the study show the characteristic differences between the studied BSS algorithms and the types of mechanical signals. The outcomes may help people to determine right algorithms.