基于稳健独立分量分析的转子故障信息增强方法  

Method of rotor fault information enhancement based on robust independent component analysis

作　　者：段礼祥[1] 胡智[1] 张来斌[1] 

机构地区：[1]中国石油大学机械与储运工程学院,北京102249

出　　处：《中国石油大学学报（自然科学版）》2013年第2期95-101,106,共8页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：国家自然科学基金项目(51005247);北京市教委科研基地建设项目;原油泵站旋转设备的风险识别及评价技术研究项目

摘　　要：针对实际转子振动信号中信源相互叠加干扰、故障信息微弱的问题,提出一种基于稳健独立分量分析(RICA)的转子故障信息增强方法。首先引入双树复小波变换,对信号进行降噪预处理,降低分离算法对噪声的敏感程度。再用稳健独立分量分析对降噪后信号进行分离和信息增强。并对比其他2种经典的盲源分离算法,通过数值仿真比较它们的分离效果。结果表明:新方法通过优化步长因子得到全局最优值,采用代数方法得到最优步长参数,实现简单,并且避免了预白化处理,使得算法运算量降低;对小数据量信号,算法收敛速度快、信号分离质量高。此方法可以更有效地分离故障源及提取信号的本质故障特征。Considering the frequency aliasing and weakness of rotor vibration signal, a novel method of fault information en- hancement based on robust independent component analysis （RICA） was presented. Firstly, the signals were denoised using dual-tree complex wavelet transform （DTCWT） to reduce the noise so as to improve the performance of RICA algorithm. Then the denoised signals were separated and information was enhanced by means of RICA. Compared with the other two classical blind source separation algorithms, the effectiveness of the proposed method was validated with the simulative signal. The results show that the global optimal value is given by optimizing the step factor using the method. The optimal step size parameters are gotten and prewhitening is avoided using algebraic methods with lower computational cost. The method shows high convergence speed and good source separation property, especially applicable for small-data records. The method can effectively separate fault signals and extract quantitative fault characteristics.

关 键 词：稳健独立分量分析 双树复小波变换 转子系统 信息增强 故障诊断 

分 类 号：TH17[机械工程—机械制造及自动化] TN911.6[电子电信—通信与信息系统]