基于朴素贝叶斯方法和权值分析方法的电机轴承故障诊断

Fault diagnosis of motor bearing based on naive Bayes and weight analysis methods

为了分析电动机轴承故障类型,首先采用小波包分析对轴承振动信号进行了高低频分解及重构,然后以各频带的能量值构成了轴承振动信号的特征向量,最后通过朴素贝叶斯网和提出的权重分析两种方法进行了电机轴承故障分类;采用朴素贝叶斯网对已知电机轴承故障类型的样本数据进行训练,获得参数后识别未知样本的故障类型,利用权重分析法计算未知与已知类型的电机轴承振动样本的相关系数,然后构建权重,并按照权值和的大小获取未知样本的故障类型。仿真结果表明,朴素贝叶斯网能较好地实现电机故障诊断,所提出的权重分析方法也能较好地对电机故障进行诊断。

In order to analyze the types of motor bearing faults,wavelet package analysis was firstly used to decompose and reconstruct the signals of the ball bearings into different frequency bands,then thevalues of energies on each bands were used to compose feature vectors of ball bearings＇ signals.Those vectors,which were considered as samples,were used in naive Bayes and weight analysis models respectively to complete the classification of motor bearing faults.Naive Bayes network trained the traing samples（whose type are known）,and then classified the testing samples（whose type are unknown）.In weight analysis model,the Euclidean distances of each testing samples and training samples were computed,and the types of all testing samples were obtained through the constructed weights.The simulation results show that,through wavelet package analysis,Naive Bayes can deal with motor bearing fault well,and the weight analysis method can also analyze the fault signals effectivly.