基于分步动态核主元分析的故障诊断方法  

Fault Diagnosis Approach Based on Step Dynamic KPCA

作　　者：袁哲[1] 石怀涛[1] 

机构地区：[1]沈阳建筑大学交通与机械工程学院,辽宁沈阳110168

出　　处：《沈阳建筑大学学报（自然科学版）》2013年第6期1092-1097,共6页Journal of Shenyang Jianzhu University：Natural Science

基　　金：国家自然科学基金项目(51105258);住房和城乡建设部研究开发项目(2012-K3-41)

摘　　要：目的研究复杂工业系统动态、非线性特点,提出分步动态核主元分析(Kernel Principal Component Analysis,KPCA)的故障诊断方法.方法该方法首先构造增广矩阵,然后将增广矩阵分成一系列子矩阵,将各子矩阵的构建一个新的数据增广矩阵,并对每个子矩阵使用KPCA提取变量数据的非线性空间相关特征,最后通过监测统计量监测出系统故障,用贡献度的方法识别发生故障变量.结果该方法改进了传统的动态方法,引入分步动态的定义,并且能充分考虑工业过程中的非线性和动态性,更精确的描述工业过程特性,更精确的监测复杂工业系统的故障,并准确的识别出故障变量.结论对热连轧过程中活套故障诊断的仿真结果表明:基于分步动态KPCA的故障诊断方法能准确有效地诊断出故障,并识别出产生故障的原因.According to the dynamic and nonlinear characteristics of complex industrial systems, fault diagno- sis approach based on step dynamic KPCA （Kernel Principal Component Analysis）is proposed in the paper. Firstly, augmented data matrix is constructed by using the＂ time lag shift＂ method, and separated into some sub - augmented matrices. Secondly, it applies KPCA to each sub - matrix for extracting the nonlinear spa- cial - correlations feature, respectively. And then, all nonlinear Principal Components are used to construct a new augment data matrix. The proposed method can effectively extract dynamic feature of industrial process by improving traditional dynamic method, and fully consider the nonlinear feature of industrial process so that industrial process characteristics can be described accurately, the fault can be detected precisely and the fault causes are identified. Simulation of the looper fault in hot continuous rolling mill demonstrates that the proposed method achieves better performance for computational efficiency, fault diagnosis and fault identifi- cation.

关 键 词：故障诊断 活套故障 核主元分析 分步动态方法 

分 类 号：TP311.51[自动化与计算机技术—计算机软件与理论]