基于最小二乘支持向量机的风机故障诊断方法研究  

Research on Fault Diagnosis System of Ventilator Based on LS-SVM

作　　者：刘延霞[1] 李建刚[1] 任子晖[1] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221116

出　　处：《淮阴师范学院学报（自然科学版）》2011年第5期414-417,422,共5页Journal of Huaiyin Teachers College;Natural Science Edition

基　　金：国家自然科学基金资助项目(60775044)

摘　　要：介绍了最小二乘支持向量机(LS-SVM)算法的诊断原理,学习算法以及技术路线.在对现场振动信号特征数据进行采集以及归一化处理的基础上,建立了风机故障数学模型以及故障样本数据库.分析了风机故障模式识别的原理,提出应用LS-SVM进行故障特征学习和分类的方法.最后对故障模型进行训练和仿真,并通过与传统的三层BP神经网络输出进行对比,验证了该方法能够及时检测到故障的发生并进行识别,是风机故障诊断的有效方法.仿真结果也证明了基于结构风险最小化原理,在兼顾泛化能力和训练误差的同时,LS-SVM在解决风机振动样本数据集非线性和容量较小的问题上优势明显,很容易建立风机故障诊断模型,在缩短自主学习时间上也有了较大的改进.This paper introduced the theory,learning algorithm and technical route of LSSVM.Though acquainting fault signals on-site and normalizing characteristic data,this method realized to establish the mathematical model and samples of database.By analyzing the principle of ventilator fault,this paper proposed the mean of LS-SVM learning and classification.Compared with the traditional BP neural network,LS-SVM had a better comprehensive performance in diagnosis of ventilator.The simulation results also prove that based on structural risk minimization principle,both training error and generalization capabilities,LS-SVM has unique advantages to solve the small sample data and nonlinear problems,particularly suitable for the establishment of fault diagnosis model,and has been greatly improved in the shorten time of independent study.

关 键 词：最小二乘支持向量机 风机 故障诊断 时域分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]