中文摘要：

为了有效提取含噪机械故障信号中的故障特征信息,研究了一种基于小波半软阈值消噪的盲源分离方法。利用小波半软阈值对故障信号进行消噪处理;采用联合近似对角化算法对信号进行盲源分离;考虑在噪声干扰下预消噪常常不足以消除全部噪声,因此在盲源分离后再进行适当的消噪处理,以提高其分离性能。实验验证了所提出方法的有效性和可行性。

英文摘要：

In order to extract fault feature informations from the mechanical malfunction signals with noise, a method of blind source separation was proposed based on wavelet semi-soft threshold de noising. First, wavelet semi-soft threshold was used to filter the failure signals. Then, joint approximate diagonalization was used as blind source separation method to separate signals. Pretreatment was often not enough to eliminate all noises, therefore, it was necessary to denoise again to improve the separation performance. Finally, the feasibility and validity of this method was verified by experiments.