基于LMD和SVM的柴油机气门故障诊断  

Diesel Engine Valve Fault Diagnosis Based on LMD and SVM

作　　者：张俊红[1] 刘昱[1] 毕凤荣[1] 林杰威[1] 马文朋[1] 马梁[1] 

机构地区：[1]天津大学内燃机燃烧学国家重点实验室,天津300072

出　　处：《内燃机学报》2012年第5期469-473,共5页Transactions of Csice

基　　金：国家自然科学基金资助项目(50975192);教育部博士点基金资助项目(20090032110001)

摘　　要：针对柴油机气门故障诊断问题,在WP7柴油机上模拟了气门故障,提出了基于局部均值分解(local mean decomposition,LMD)和支持向量机(support vector machine,SVM)相结合的气门故障诊断方法.该方法首先用改进LMD方法将缸盖振动信号分解成若干个瞬时频率具有物理意义的PF(product function)分量之和,然后从缸盖振动信号和分解得到的PF分量中提取故障特征向量,以此作为SVM分类器的输入进行故障诊断.此外提出了改进粒子群算法(particle swarm optimization,PSO)用于SVM参数的优化.诊断结果显示,16组测试样本的测试结果均与实际状况相一致,诊断正确率为100%,该方法能快速准确地识别内燃机气门故障.To solve the fault diagnosis of diesel engine valve, the faults of the valve in the WP7 diesel engine was simulated, and a valve fault diagnosis method was proposed based on local mean decomposi- tion （LMD）and support vector machine （SVM）. Firstly, using the improved LMD decomposition vibra- tion signal of cylinder into a set of product functions, then extracting fault features from the product func- tions, and finally using the fault features as the input for the support vector machine diagnosis model. In addition, an improved strategy of particle swarm optimization （PSO） algorithm for optimizing the parame- ters of SVM was proposed. Test results show that the test results of 16 group samples are consistent with the real situation with diagnostic accuracy of 100%,. The method can rapidly and accurately identify the fault of diesel engine valve.

关 键 词：柴油机 局部均值分解 支持向量机 

分 类 号：TK428[动力工程及工程热物理—动力机械及工程]