基于CVA-ICA的机械故障源动态盲分离方法

发布时间：2019-01-18 19:51

【摘要】：动态盲源分离问题是多故障源盲分离的一个热点。传统的机械故障源分离方法要求满足统计特征保持稳定,且混合系统保持不变等假设,而忽略了时序信息。针对此不足,结合规范变量分析(Canonical variate analysis,CVA)和独立分量分析(Independent component analysis,ICA),提出一种基于CVA-ICA的机械多故障源动态盲分离方法。该方法的基本思想是将源信号看成状态空间的状态变量,观测信号看成状态空间的输出变量,从而将动态混合盲源分离问题转化为状态空间盲源分离问题,利用规范变量分析作为降维工具来构造状态空间,再利用传统的ICA算法对规范的观测信号进行盲源分离。仿真研究表明,在处理动态混合的盲分离中,提出的方法明显优于静态ICA方法,取得了满意的分离效果。将该方法应用到滚动轴承内圈和滚动体的故障盲分离中,试验结果进一步验证了该方法的有效性。
[Abstract]:Dynamic blind source separation (DBSS) is a hot topic in multi-fault source blind separation. The traditional mechanical fault source separation method needs to satisfy the assumption that the statistical characteristics remain stable and the hybrid systems remain invariant while ignoring the timing information. In order to solve this problem, a dynamic blind separation method for mechanical multi-fault sources based on CVA-ICA is proposed, which combines the canonical variable analysis (Canonical variate analysis,CVA) and independent component analysis (Independent component analysis,ICA). The basic idea of this method is to treat the source signal as the state variable of the state space and the observation signal as the output variable of the state space, so as to transform the dynamic mixed blind source separation problem into the state space blind source separation problem. The canonical variable analysis is used as a dimensionality reduction tool to construct the state space, and then the traditional ICA algorithm is used to separate the normal observation signals from the blind source. The simulation results show that the proposed method is better than the static ICA method in the blind separation of dynamic mixing, and the satisfactory separation effect is obtained. The method is applied to blind fault separation of rolling bearing inner ring and rolling body. The experimental results further verify the effectiveness of the method.
【作者单位】：南昌航空大学无损检测技术教育部重点实验室;
【基金】：国家自然科学基金(51261024,51075372,51265039) 江西省教育厅科技计划(GJJ12405) 湖南科技大学机械设备健康维护湖南省重点实验室开放基金(201204)资助项目
【分类号】：TH165.3;TN911.7

【参考文献】