因果关系(物理学)
格兰杰因果关系
多元统计
计量经济学
根本原因
过程(计算)
计算机科学
统计假设检验
非线性系统
故障检测与隔离
根本原因分析
统计
数据挖掘
人工智能
数学
机器学习
工程类
可靠性工程
物理
操作系统
量子力学
执行机构
作者
Han-Sheng Chen,Zhengbing Yan,Yuan Yao,Tsai-Bang Huang,Yi-Sern Wong
标识
DOI:10.1021/acs.iecr.8b00697
摘要
Multivariate statistical process monitoring (MSPM) has received a considerable amount of attention in terms of both academic research and industrial applications. Most of these efforts have been focused on fault detection and isolation, while root cause diagnosis has not yet been fully addressed. In recent years, data-driven causality analysis methods have been adopted in order to understand the complex relationship between process variables and to identify the causes of the faults triggering the alarms. Among them, the Granger causality (G-causality) test is a popular method of inferring causal associations between signals based on temporal precedence. Nevertheless, the conventional G-causality test applies only to stationary and linear time series. Additionally, it determines the relationships between the variable pairs and is not suited to multivariate cases. In this study, the use of statistical tests is proposed in order to assess whether the time series are nonstationary or nonlinear. For significant nonstationary or nonlinear signals, the Gaussian process regression (GPR) approach is integrated into the framework of the multivariate G-causality test in order to better indicate the causal relationships between the candidate process variables. The feasibility of the proposed scheme for root cause diagnosis is illustrated through case studies.
科研通智能强力驱动
Strongly Powered by AbleSci AI