基于支持向量机的机械故障特征选择方法研究

在机械故障诊断中,对机器状态信号进行处理可得到故障特征集。但是此特征集中通常含有冗余特征而影响诊断效果。特征选择可以去除原始特征中的冗余特征,提高诊断精度和诊断效率。本文提出采用支持向量机(SVM)作为决策分类器,研究了使用SVM的错误上界如半径-间距上界代替学习错误率作为特征性能评价,并且使用遗传算法对特征集进行寻优的特征选择方法。此方法由于只需要训练一次SVM,相比常用的分组轮换方法有较高的计算效率。数值仿真和减速器的轴承故障特征选择试验中,采用此方法对生成特征集进行选择,并与常用的分组轮换法进行了对比。结果显示此方法有较好的选择性能和选择效率。

Research on SVM-Based Machine Fault Feature Selection

Fault features set can be obtained by processing machine condition signal in machine fault diagnosis.Unfortunately,redundant features usually exist in the fault feature setand decrease diagnosis correctness.Feature selection can remove redundant features to enhance the effect of diagnosis.This paper adopts support vector machine(SVM) as a fault classifier,takes into account the existing bounds on the generalization error such as radius-marginbounds for evaluation of SVM,and applies genetic algorithm(GA) to solving feature selection problem.This is computationally much faster than k-folds cross-validation,since we do not have to retrain SVM many times on each feature subset but just once.Results of simulated data and gearbox bearing fault diagnosis example show that this method possesses excellent optimization feature subset property,and is faster than the existing methods.