摘要
利用声信号对往复泵进行状态监测,针对往复泵的声信号是具有非平稳性、非线性等复杂特征的信号,采用多重分形去趋势波动分析(MFDFA)计算时间序列声信号的多重分形谱,并提取作为故障特征量。分别用支持向量机(SVM)、遗传算法(GA)改进的SVM、混合蛙跳算法(SFLA)改进的SVM进行故障识别。通过实验测取往复泵的原始信息信号并分析,验证了声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,对比研究三种识别方法表明了基于混合蛙跳算法优化(SFLA)改进的支持向量机识别效果最好,基于MFDFA和SFLA-SVM相结合的故障诊断方法能准确地提高往复泵泵阀的故障诊断准确率,是往复泵故障诊断方法的一种新的有效方法。
Acoustic signal is being used to monitor the condition of the reciprocating pump.The fault signal of the Reciprocating Pump is the complex signal of the non-stationary and nonlinear characteristics.MFDFA method is applied to calculate the multifractal spectra of acoustical signals in time series,and extracted as fault feature quantities.The SVM,the SVM improved by GA and the SVM improved by shuffled frog leaping algorithm are used to identify fault.The original information signal of reciprocating pump is measured and analyzed by experiment.it proves that the fluctuation of acoustic signal indicates obvious multifractal characteristics,the normal state and fault state can be distinguished effectively.Comparing the three identification methods indicates that the SVM improves by SFLA has the best recognition effect.The method based on Support Vector Machine based on shuffled frog leaping algorithm and MFDFA can improve the accuracy of the fault diagnosis of the reciprocating pump valve accurately.It is an innovative and effective method to diagnose the fault of reciprocating pumps.
作者
裴峻峰
严安
彭剑
赵钧羡
PEI Jun-feng;YAN An;PENG Jian;ZHAO Jun-xian(College of Mechanical Engineering,Changzhou University,Jiangsu Changzhou213000,China;Jiangsu Hexin Oil Machinery Group Co.,Ltd.,Jiangsu Yancheng224000,China)
出处
《机械设计与制造》
北大核心
2020年第4期199-203,207,共6页
Machinery Design & Manufacture
基金
国家自然科学基金(51505041)。
关键词
往复泵
声信号
MFDFA
故障诊断
SVM
优化
Reciprocating Pump
Sound Signal
MFDFA
Fault Diagnosis
SVM
Optimization
