基于VMD和拉普拉斯分值的柴油机故障诊断  

Diesel Engine Fault Diagnosis Based on VMD and Laplacian Score

作　　者：吉哲 傅忠谦[1] 

机构地区：[1]中国科学技术大学信息科学技术学院,合肥230026 [2]海军蚌埠士官学校二系,安徽蚌埠233012

出　　处：《组合机床与自动化加工技术》2017年第10期129-133,137,共6页Modular Machine Tool & Automatic Manufacturing Technique

摘　　要：针对柴油机声信号非平稳非线性的特性,提出了一种基于变分模态分解(VMD)和拉普拉斯分值(LS)的柴油机故障诊断方法。首先对柴油机声信号进行变分模态分解,从分解得到的各模态函数中进行统计特征提取,组成初始特征集;然后利用改进的拉普拉斯分值算法进行特征排序,以支持向量机(SVM)为故障分类器,实现柴油机的故障诊断;最后通过设计接受者操作特性(ROC)指示器,确定故障诊断的最优维。将该方法应用到6135D型柴油机四种常见故障的诊断中,实验结果表明该方法能有效提取柴油机声信号特征并具有较高的诊断精度。Aiming at the acoustic signal of diesel w ith features of nonstationary and nonlinear,a fault diagnosis method for diesel engine based on Variational M ode Decomposition( VM D) and Laplacian Score( LS)w as proposed. Firstly VM D w as employed to decompose acoustic signals of diesel,the statistical features w ere extracted from the decomposition of each modal function to form the initial feature set; then improved LS algorithm w as used for feature ranking,using Support Vector M achine( SVM) as a fault classifier,the fault diagnosis of diesel engine w as realized; finally through the design of Receiver Operating Characteristic( ROC) indicator,the optimal dimension of fault diagnosis was determined. The method is applied to the diagnosis of four common faults of type 6135 D diesel engine,the experimental results show that this method can efficiently extract the features of acoustic signals of diesel and has high diagnostic accuracy.

关 键 词：变分模态分解 拉普拉斯分值 特征提取 支持向量机 

分 类 号：TH165.3[机械工程—机械制造及自动化] TG506[金属学及工艺—金属切削加工及机床]