基于LFOA优化多核支持向量机的液压泵故障诊断  

Fault Diagnosis of Hydraulic Pump Based on LFOA Optimized Multi-kernel SVM

作　　者：杜伟 房立清 齐子元 张前图 

机构地区：[1]军械工程学院火炮工程系 [2]驻356厂军代室

出　　处：《机械设计与研究》2018年第1期137-140,144,共5页Machine Design And Research

基　　金：河北省自然科学基金资助项目(E2016506003)

摘　　要：针对采用传统支持向量机（SupportVectorMachine，SVM）进行故障诊断时核函数的构造和参数的选取存在盲目性的问题，提出一种基于改进果蝇优化算法优化多核SVM的液压泵故障诊断方法。首先，对液压泵振动信号进行局部特征尺度分解（LocalCharacteristic—scaleDecomposition，LCD），从多个角度提取混合特征组成特征集。然后，基于全局核函数和局部核函数构建多核支持向量机，并利用具有Levy飞行特征的果蝇优化算法（LFOA）对核函数权值和参数的选取进行优化。最后，将特征集输入多核SVM进行识别。液压泵故障诊断结果表明，与采用FOA、GA和PSO优化算法及单核SVM相比，所提方法具备全局寻优能力强和诊断准确率高的优点，可有效应用于液压泵故障诊断。Aiming at the blindness of the construction of kernel function and the selection of parameters in the fault diagnosis using traditional support vector machine （SVM） , a fault diagnosis method of hydraulic pump based on improved fruit fly optimization algorithm optimized multi-kernel SVM was proposed. Firstly, the local characteristic- scale decomposition （LCD） of the hydraulic pump vibration signal was performed to extract the mixed feature set from multi domains. Then, the multi-kernel support vector machine was constructed based on global kernel function and local kernel function, and the fruit fly optimization algorithm with Levy flight characteristic （ LFOA ） was used to optimize the selection of kernel weights and parameters. Finally, the feature set was input to the multi-kernel SVM for identification. The results of hydraulic pump fault diagnosis show that compared with FOA, GA and PSO optimization algorithms and single-kernel SVM, the proposed method has the advantages of strong global optimization ability and high diagnostic accuracy, and can be effectively applied to hydraulic pump fault diagnosis.

关 键 词：多核支持向量机 果蝇优化算法 局部特征尺度分解 液压泵 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化]