基于经验模式分解复杂度特征和最小二乘支持向量机的离心泵振动故障诊断

Vibration Fault Diagnosis of Centrifugal Pump Based on Empirical Mode Decomposition Complexity Feature and Least Square Support Vector Machine

针对离心泵振动信号的非平稳特征,提出一种基于经验模式分解(empirical mode decomposition,EMD)复杂度特征和最小二乘支持向量机的离心泵振动故障诊断方法。首先对振动信号进行经验模式分解,将其分解为若干个固有模态函数(intrinsic mode function,IMF),然后对每一个IMF分量提取复杂度特征作为故障特征向量,并以此作为输入参数建立最小二乘支持向量机分类器诊断故障。选用径向基函数(radial basis function,RBF)作为核函数,并采用差分进化算法进行参数选择。应用结果表明,EMD复杂度特征能准确诊断故障,参数优化后的模型具有更高的分类能力。

Aiming at the non-stationary and non-linearity characteristics of the vibration signals of centrifugal pump,a new method based on complexity feature of empirical mode decomposition(EMD) and least square support vector machine (LS-SVM) was presented.First of all,EMD method was used to decompose the vibration signals into a finite number of stationary intrinsic mode functions(IMF),and then complexity features of each IMF was extracted as the fault characteristics vectors and served as input parameters of LS-SVM classifier to diagnosis fault.Radial basis function(RBF) function was used as kernel function;the differential evolution(DE) method was proposed to select hyper-parameter of LS-SVM.Application results show that the proposed method is very effective,which can better extract the nonlinear features of the fault and more exactly diagnosis fault,optimized LS-SVM model has strong capability of classification.