基于声振信号融合的IRCMMDE离心泵损伤检测方法  

Damage detection of centrifugal pump based on IRCMMDE with acoustic vibration signal fusion

作　　者：陆春元[1] 焦洪宇 LU Chun-yuan;JIAO Hong-yu(School of Mechanical and Electrical Engineering,Suzhou Vocational University,Suzhou 215104,China;School of Automotive Engineering,Changshu Institute of Technology,Suzhou 215500,China)

机构地区：[1]苏州市职业大学机电工程学院,江苏苏州215104 [2]常熟理工学院汽车工程学院,江苏苏州215500

出　　处：《机电工程》2023年第6期952-959,共8页Journal of Mechanical & Electrical Engineering

基　　金：国家自然科学基金资助项目(51605046);江苏政府留学奖学金资助项目(JS-2017-188)。

摘　　要：离心泵早期的损伤特征比较微弱,难以有效提取其故障特征。针对这一问题,提出了一种基于声振信号融合的改进精细复合多元多尺度散布熵(IRCMMDE)和GWO-SVM的离心泵损伤检测方法。首先,利用多个传感器收集了离心泵在不同损伤状态下的声音和振动信号,并将声音和振动信号进行了融合,以充分利用不同类型信号中所蕴含的损伤特征信息;随后,针对多元多尺度散布熵(MMDE)不稳定的缺陷,对MMDE的粗粒化处理进行了优化,提出了改进精细复合多元多尺度散布熵(IRCMMDE)的复杂性测量指标;接着,利用IRCMMDE对声振融合信号进行了损伤特征提取,构建了各个损伤状态下的特征矩阵;最后,利用灰狼算法优化的支持向量机分类器,对各个损伤状态下的特征矩阵进行了识别,得到了最终的离心泵损伤检测结论。研究结果表明:采用基于声振信号融合的离心泵损伤检测方法,其最高可达到99.2%的故障识别准确率,相比于基于MMDE和RCMMDE的损伤检测方法,其能够更准确地识别出离心泵的损伤;该方法还能有效缓解单一信号检测时的不确定性,并且在多次实验验证下,其仍具有很高的检测精度。In order to solve the problem that the early damage characteristics of centrifugal pumps were weak and difficult to extract effectively,an improved refined composite multivariate multiscale dispersion entropy(IRCMMDE)and grey wolf optimizer(GWO)-support vector machine(SVM)method based on acoustic vibration signal fusion was proposed.Firstly,the acoustic and vibration signals of the centrifugal pump under different damage conditions were collected by using multiple sensors and fused to make full use of the damage feature information contained in different types of signals.Secondly,in view of the unstable defect of multivariate multiscale dispersion entropy(MMDE),the coarse granulation processing of MMDE was optimized,and the complexity measurement index of improved refined composite multivariate multiscale dispersion entropy(IRCMMDE)was proposed.Then,the IRCMMDE was used to extract the damage features of the acoustic vibration fusion signal,and the feature matrix of each damage state was constructed.Finally,the support vector machine classifier optimized by gray wolf algorithm was used to recognize the feature matrix,and the final damage detection conclusion was obtained.The research results show that the damage detection scheme based on acoustic vibration signal fusion can achieve the highest fault identification accuracy of 99.2%.Comparing with the methods based on MMDE and RCMMDE,it can identify the damage of centrifugal pumps more accurately.This method also effectively alleviates the uncertainty of single signal detection,and it has high detection accuracy in many experiments.

关 键 词：声振信号融合 离心泵损伤检测 改进精细复合多元多尺度散布熵 灰狼算法 支持向量机 

分 类 号：TH311[机械工程—机械制造及自动化]