多故障源混合信息的特征分离与精确辨识  

Feature Separation and Precise Identification of Multiple Fault Source Mixed Signals

作　　者：张建宇[1,2] 王国峰 杨洋 ZHANG Jianyu;WANG Guofeng;YANG Yang(Faculty of Materials and Manufacturing,Beijing Key Laboratory of Advanced Manufacturing Technology,Beijing University of Technology,Beijing 100124,China;Beijing Engineering Research Center of Precision Measurement Technology and Instrument,Beijing 100124,China;Beijing Goldwind Science and Creation Windpower Equipment Co.,Ltd.,Beijing 100176,China)

机构地区：[1]北京工业大学材料与制造学部北京市先进制造技术重点实验室,北京100124 [2]北京市精密测控技术与仪器工程技术研究中心,北京100124 [3]北京金风科创风电设备有限公司,北京100176

出　　处：《噪声与振动控制》2022年第1期88-94,共7页Noise and Vibration Control

基　　金：国家自然科学基金资助项目(51675009)。

摘　　要：多故障源的耦合问题,一直以来都是诊断领域面临的最大难题之一,能否实现多源信号解耦将直接影响故障诊断的准确性。在信源卷积混合的前提下,以多通道反卷积理论为基础,首先研究多故障源混合信息的特征分离方法。接着借助卷积混合后的仿真信号,研究反卷积系统的关键参数——初始滤波器长度对分离效果的影响。进而提出一种自适应的多源信息分离方法。为了确定分离系统的输入通道数量,基于小波分析与奇异值分解完成信号的源数估计;再通过设定滤波器长度的迭代区间,计算出不同参数下分离系统输出信号的时域指标,并自动选取最佳长度使得分离结果最优。最后,经滚动轴承的复合故障实验和多故障并发的工程数据验证,表明该方法在设定的滤波器长度范围内,能够根据信号的差异性自动寻优最佳参数,并成功分离出原始信号中隐含的各个故障源信息,实现机械故障的精确辨识。Coupling problem of multi-fault sources has become one of the most challengeable problems in mechanical and electrical equipment fault diagnosis.Multi-source signal decoupling will directly affect the accuracy of diagnosis.Under the premise of signal source convolution mixture and based on the multi-channel deconvolution separation technique,i.e.multi-source information separation method,the feature separation method for multi-fault source mixed information was studied.Then,in virtue of the simulation signal after the convolution mixture,the influence of the key parameter of the deconvolution separation system,namely the initial filtering length,on the separation effect,was analyzed.Furthermore,an adaptive multi-source information separation method was established.Source number estimation strategy based on wavelet analysis and singular value decomposition was introduced to determine the input channel number of the separation system.The filter length interval was set,the time domain index of each output signal was calculated and the best filter parameters were automatically selected so as to obtain the optimized separation signals.Finally,experimental and engineering data of rolling element bearings with compound faults was utilized to validate the effectiveness of above adaptive method.The analysis results show that the proposed method is robust enough to separate the multiple fault source information hidden in the raw signal.Therefore,the accuracy of fault diagnosis can be greatly improved.

关 键 词：故障诊断 卷积混合 反卷积 特征分离 多源耦合 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]