基于MFDFA的往复泵泵阀声音信号故障诊断  

Fault Diagnosis of Sound Signal of Reciprocating Pump Valve Based on MFDFA

作　　者：郭攀 史洪伟 裴峻峰[2] 王红艳 周丹红 谢永 GUO Pan;SHI Hong-wei;PEI Jun-feng;WANG Hong-yan;ZHOU Dan-hong;XIE Yong(School of Chemistry and Chemical Engineering, Suzhou University, Suzhou 234000, China;School of Mechanical Engineering, Changzhou University, Changzhou 213016, China)

机构地区：[1]宿州学院化学化工学院,安徽宿州234000 [2]常州大学机械工程学院,江苏常州213016

出　　处：《榆林学院学报》2020年第2期41-45,共5页Journal of Yulin University

基　　金：宿州学院“新工科”试点专业建设项目(szxy2018xgk02);宿州学院校级专业带头人(2019XJZY28);宿州学院教授(博士)启动基金项目(2019jb03);宿州学院创新训练项目(201810379042);2017年安徽省高校自然科学研究重点项目(KJ2017 A435)

摘　　要：针对往复泵泵阀故障诊断,使用声音信号对其进行状态监测,采用多重分形趋势波动分析(MFDFA,Multifractal Detrended Fluctuation Analysis)计算时间序列声音信号的多重分形谱,并提取特征参数,将这些参数用于故障诊断。首先将时间序列的声音信号转换为时间序列的随机游走;然后进行去趋势并提取声音信号的多重分形特征参数进行比较分析;最后将多重分形参数作为特征向量输入支持向量机(SVM,Support Vector Machine)进行模式识别。经过验证分析,声音信号的波动呈现明显的多重分形特性,可以有效区分正常状态与故障状态,进而实现对往复泵泵阀的故障诊断。To the fault diagnosis of reciprocating pump valve,sound signal is used to monitor its condition,multifractal detrended fluctuation analysis(MFDFA)is used to calculate the multifractal spectrum of sound signals in time series,then the feature parameters are extracted from the multifractal spectrum,and these parameters can be used in fault diagnosis.First,convert the sound signal of the time series to a random walk of it.Then detrend and extract the multifractal characteristic parameters of the sound signal to analyze comparatively.Finally,take the multifractal parameters as the feature vector and input them into support vector machine(SVM)for pattern recognition.After verification and analysis,it is found that the fluctuation of sound signal has obvious multifractal characteristics,which can effectively distinguish the normal state and the fault state,and then realize the fault diagnosis of reciprocating pump.

关 键 词：往复泵泵阀 声音信号 MFDFA 故障诊断 SVM 

分 类 号：TH38[机械工程—机械制造及自动化]