基于ICA混合模型的多工况过程故障诊断方法

发布时间：2018-03-11 12:17

本文选题：ICA混合模型　切入点：多工况过程　出处：《化工学报》2016年09期 　论文类型：期刊论文

【摘要】：针对工业过程数据的多模态和非高斯特性,提出一种基于独立元混合模型(independent component analysis mixture model,ICAMM)的多工况过程故障诊断方法。该方法将独立元分析与贝叶斯估计结合,同时完成各个工况的数据聚类和模型参数求取,并建立基于贝叶斯框架下的集成监控统计量实时监控过程变化。在检测到故障后,针对传统的变量贡献图方法无法表征变量之间信息传递关系的缺点,提出基于信息传递贡献图的故障识别方法。该方法首先计算各变量对独立元混合模型统计量的贡献度,进一步通过最近邻传递熵描述故障变量之间的传递性,挖掘故障变量之间的因果关系,从而确定故障源变量和故障传播过程。最后对一个数值系统和连续搅拌反应釜(CSTR)过程进行仿真研究,结果验证了本文所提出方法的有效性。
[Abstract]:In view of the multi-modal and non-#china_person0# characteristics of industrial process data, a fault diagnosis method based on independent component analysis mixture model ICAMMM is proposed, which combines independent element analysis with Bayesian estimation. At the same time, the data clustering and model parameter calculation of each working condition are completed, and the real-time monitoring process of integrated monitoring statistics based on Bayesian framework is established. In view of the disadvantage that the traditional method of variable contribution graph can not represent the information transfer relation between variables, a fault identification method based on information transfer contribution graph is proposed. Furthermore, the transitivity between fault variables is described by nearest neighbor transfer entropy, and the causality between fault variables is mined. Finally, a numerical system and a continuous stirred tank reactor (CSTR) process are simulated and the results show that the proposed method is effective.
【作者单位】：中国石油大学(华东)信息与控制工程学院;
【基金】：国家自然科学基金项目(61273160,61403418) 山东省自然科学基金项目(ZR2014FL016)~~
【分类号】：TP277

【相似文献】