基于小波包分解与DAG-SVM的柱塞泵故障诊断  

Fault Diagnosis of Piston Pumps Based on Wavelet Packet Decomposition and DAG-SVM

作　　者：蔡伟[1] 黄坤阳 戴民强[1] 杨志勇[1] 

机构地区：[1]第二炮兵工程大学兵器发射理论与技术国家重点学科实验室,陕西西安710025

出　　处：《液压与气动》2015年第12期12-17,共6页Chinese Hydraulics & Pneumatics

基　　金：国家自然科学基金(61102170)

摘　　要：针对柱塞泵检测诊断中故障特征模糊、成因复杂、难以准确定位的问题,结合决策树与支持向量机提出一种基于小波包分解与DAG-SVM的柱塞泵故障诊断方法。该方法预先对所用C-SVM和RBF核函数的参数进行优化,而后采用db5小波包对泵体振动信号进行三层分解以提取特征向量,将特征向量输入支持向量机完成其训练及模式识别过程。同时设计了柱塞泵故障诊断的一体化装置,通过模拟不同故障,利用已知故障样本完成支持向量机的训练过程,进而对待测样本进行故障模式识别。诊断结果与样本已知状态相符,验证了该方法的准确性。According to the problems that fuzzy fault feature of piston pumps has complex causes and is difficult to locate accurately, a fault diagnosis method based on wavelet packet decomposition and DAG-SVM is proposed com- bining decision tree and support vector machine. Parameters of C-SVM and RBF kernel function were optimized in advance, then feature vectors of faults were computed with db5 wavelet packet by three-ply decomposing pump body vibration signals, which were imported to SVM to complete training procedure and pattern recognition. This method completed training procedure using known fault sample after simulating different faults, then fault patterns of test samples were recognized. Besides the integrated piston pump diagnosis device was designed. Finally diagnosis re- suits fit the states of known test samples, which verified the accuracy of this method.

关 键 词：柱塞泵 故障检测诊断 小波包分解 DAG-SVM 参数优化 

分 类 号：TH137[机械工程—机械制造及自动化]