一种基于GMM的多工况过程故障诊断方法  

Multi-mode process fault diagnosis method based on GMM

作　　者：孙贤昌 田学民[1] 张妮[1] 

机构地区：[1]中国石油大学(华东)信息与控制工程学院,山东青岛266580

出　　处：《计算机与应用化学》2014年第1期33-39,共7页Computers and Applied Chemistry

基　　金：国家自然科学基金项目(61273160);山东省自然科学基金项目(ZR2011FM014)

摘　　要：故障发生后,为了查出故障原因,需要辨识出故障变量实现故障诊断。传统多元统计故障诊断方法通常假设过程只运行在单一工况下,从而不能适用于多工况过程。针对这个问题,提出了1种基于高斯混合模型(Gaussian Mixture Models,GMM)的多工况化工过程故障诊断方法。首先建立GMM对过程数据进行聚类,自动估计出工况数以及各工况的分布参数,同时对数据进行标准化处理,使得各个变量在后续处理中占有相同比重,然后对各个工况建立PCA(PrincipalComponentAnalysis)模型。当检测到故障发生后,本文通过结合相对贡献量和后验概率,构造全局相对贡献量得到全局相对贡献图,用于多工况过程故障变量的识别。CSTH过程的仿真结果表明,本文提出的方法能够有效地辨识出故障变量。When a fault occurs in the chemical multi-mode process, the fault diagnosis is needed to find out the cause of the fault by identifying the fault variables. Traditional multivariate statistical fault diagnosis methods are designed for a single operating condition and may produce erroneous conclusions if used for multi-mode process. A novel multi-mode chemical process fault diagnosis approach based on GMM is proposed in this paper. First, the GMM model is established to automatically cluster the process data to get the number of modes and corresponding parameters. In order to make the variables have the same weight, the standardization of the data is needed. Then a principal component model is established for individual operating mode to describe the statistical features of whole operating process. When the fault is detected, this paper constructs the global relative contribution rate to get the global contribution plot by combining the relative contribution and posterior probability. Simulation results of CSTH process demonstrate that the proposed approach can achieve good identification in multi-mode processes.

关 键 词：多工况 高斯混合模型 故障诊断 贡献图 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]