基于SSA-VMD的气阀信号故障特征提取方法研究  

Research on Fault Feature Extraction Method for Gas Valve Signal Based on SSA-VMD

作　　者：温丹阳 张秀珩[1] 刘佳音 孙林 WEN Dan-yang;ZHANG Xiu-heng;LIU Jia-xin;SUN Lin(School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110159,China)

机构地区：[1]沈阳理工大学机械工程学院,辽宁沈阳110159

出　　处：《压缩机技术》2024年第3期17-22,共6页Compressor Technology

摘　　要：为了解决VMD算法因人为预设分解层数K及惩罚因子α两个参数而导致的信号过分解和欠分解的问题,提出了一种基于SSA-VMD的隔膜压缩机气阀故障特征提取方法。该方法以各分量的最小包络熵作为适应度函数,对VMD进行参数优化。将SSA-VMD方法与原始VMD方法进行对比分析,分别计算两种分解方法各分量的多尺度散布熵(MDE)值,选取合适的熵值作为特征向量,利用随机森林算法进行故障的分类识别,从而实现对隔膜压缩机气阀各类故障的诊断。研究结果表明:使用原始VMD方法分解后的数据,故障识别的平均准确率为90.63%,而利用SSA-VMD方法故障识别的平均准确率高达99.32%。该结果表明使用SSA-VMD方法分解信号,故障诊断的准确率明显提高。In order to solve the problem of signal overdecomposition and underdecomposition caused by the artificial preset decomposition layer number K and penalty factorαof VMD algorithm,a method for extracting fault features of diaphragm compressor based on SSA-VMD is proposed.This method optimizes the VMD with the minimum envelope entropy of each component as a fitness function.By comparing the SSA-VMD method with the original VMD method,the multi-scale dispersion entropy(MDE)values of each component of the two decomposition methods are calculated separately.The appropriate entropy values are selected as feature vectors,and the random forest algorithm is used to classify and identify the faults,so as to diagnose the various faults of the diaphragm compressor air valve.The results show that the average accuracy of fault identification using the original VMD method is 90.63%,while the average accuracy of fault identification using the SSA-VMD method is as high as 99.32%.This result indicates that using SSA-VMD method to decompose signals significantly improves the accuracy of fault diagnosis.

关 键 词：VMD 麻雀搜索算法 故障诊断 多尺度散布熵 气阀 

分 类 号：TH457[机械工程—机械制造及自动化]