基于拉普拉斯分值和超球支持向量机的轴承故障诊断方法设计  

Fault Diagnosis Method of Bearing Based on Laplasian Score and Hypersphere Support Vector Machine

作　　者：冯慧玲[1] 常国权[1] 孔娟[1] 

机构地区：[1]安阳工学院计算机科学与信息工程学院

出　　处：《计算机测量与控制》2015年第4期1102-1105,共4页Computer Measurement &Control

基　　金：河南省科技厅基础研究项目(2012939)

摘　　要：针对轴承振动信号中的故障信息往往很微弱,同时振动样本数据分布不平衡即故障样本占总样本数的比例低,从而导致故障诊断模型训练不精确而影响诊断精度的问题,提出了一种基于拉普拉斯分值和超球大间隔支持向量机的故障诊断方法;首先,采用有标签的训练样本数据和拉普拉斯分值法提取原始振动信号中的微弱故障信息,并降低其数据维数,从而得到用于故障诊断的特征向量,然后设计了一种改进的超球大间隔支持向量机的故障诊断模型,通过最小化超球体积和最大化超球边界和故障样本之间的间隔来实现故障诊断,以解决样本的不均衡问题,最终通过将测试样本数据代入决策方程并通过投票机制确定其故障类别;在Matlab环境下对轴承故障诊断进行实验,实验结果证明了文中方法能有效解决样本的不均衡情况下的故障诊断,且相对其它方法,具有诊断精度高和收敛速度快的优点。Aiming at the bearing vibration signal usually has weak fault information, and also the vibration sample data is distributed unevenly namely the fault sample data counts small in all the sample data, leading to the inaccuracy of fault diagnosis model and fault diagnosis, a fault diagnosis method based on Laplasian score and hypersphere support vector machine is proposed. Firstly, the label sample data and Laplansian score method is used to extract the weak fault information in original vibration signal, the data dimension is reduced and the fault diagnosis feature vector is obtained, then an improved hypersphere big distance support vector machine is introduced, the fault diagnosis is implemented by minimizing the hypersphere volume and maximizing the distance between hypersphere boundary and fault sample to solve the sample uneven distribution problem, finally, the test sample data is input to decision equation using the voter mechanism to assure the diagnosis classifier. The experiment of bearing is operated in the Matlab environment and the result shows the method in this paper can realize the fault diagnosis, and compared with the other methods, it has the higher diagnosis accuracy and quick convergence.

关 键 词：故障诊断 拉普拉斯分量 轴承 超球支持向量机 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]