基于MOMEDA与LMD的往复压缩机活塞杆沉降信号故障特征提取方法研究

Research on fault feature extraction method for reciprocating compressor piston rod settlement signal based on MOMEDA and LMD

针对传统经验模态分解(EMD)等方法在当前往复压缩机活塞杆故障诊断中故障特征提取能力不足的问题,本文通过电涡流传感器获得往复压缩机活塞杆的沉降信号,利用多点最优最小熵解卷积算法(MOMEDA)对信号周期进行自适应调整去干扰处理,再对其进行局部均值分解(LMD),得到信号所对应的多个乘积函数(PF)分量的特征参数因子,包括偏度系数gi、峭度系数qi和总能量比Ei/E。对比活塞杆正常和故障状态(支撑环磨损、紧固元件松动和早期裂纹)下的特征参数变化,结果显示:在活塞杆支撑环磨损情况下,g1和q3的值将分别达到-0.02和1.60,与正常值相差3~5倍;活塞杆紧固原件松动情况下,g1,g3,q1,q3均会出现大幅度偏差,甚至呈现出超过正常值10倍以上的差距;活塞杆早期裂纹情况下,低阶分量g4和q4会出现一些变化,分别达到-1.30和1.60;MOMEDA与LMD相结合的方法,能够准确、有效地对往复压缩机活塞杆沉降信号进行判断,相比于传统的EMD信号分析方法,该方法在活塞杆故障诊断领域展现出更高的实用性。

In response to the shortage of insufficient reliability of traditional empirical mode decomposition(EMD)methods in the current fault diagnosis of reciprocating compressor piston rods,this paper obtains the settlement signal of reciprocating compressor piston rods through eddy current sensors,uses multipoint optimal minimum entropy deconvolution algorithm(MOMEDA)to adaptively adjust the signal period to eliminate interference,and then performs local mean decomposition(LMD)on it to obtain the characteristic parameter factors of multiple product function(PF)components corresponding to the signal,including skewness coefficient gi,kurtosis coefficient qi,and total energy ratio Ei/E.Comparing the changes in characteristic parameters of the piston rod under normal and faulty conditions(support ring wear,loose fastening components,and early cracks),the results show that under the condition of piston rod support ring wear,the values of g1 and q3 will reach around-0.02 and 1.60,respectively,which is 3~5 times different from the normal values;When the fastening components of the piston rod are loose,g1,g3,q1,q3 will all show significant deviations,even exceeding the normal value by more than 10 times;In the case of early cracks in the piston rod,there will be some changes in the low order components g4 and q4,reaching around-1.30 and 1.60,respectively;The combination of MOMEDA and LMD method can accurately and effectively judge the settlement signal of reciprocating compressor piston rod.Compared with the traditional EMD signal analysis method,this method has shown higher practicality in the field of piston rod fault diagnosis.