高校化学工程学报2025，Vol.39Issue(5)：876-889,14.DOI:10.3969/j.issn.1003-9015.2025.05.010
基于BDGL-CMI的传递熵网络因果推理的工业过程故障溯源
Fault tracing of industrial process based on causal reasoning of transfer entropy network based on BDGL-CMI
摘要
Abstract
To address the problems of long tracing times and poor fault tracing results caused by the interrelated effects of multiple factors in complex industrial processes,we proposed a fault tracing algorithm that integrates the block drop form of the graphical lasso(BDGL)with transfer entropy and conditional mutual information(CMI).First,we introduced an optimized graph lasso to model subgroups,effectively reducing computational com-plexity.Then,we applied transfer entropy for causal analysis to obtain the main variable pairs of transfer entropy.By calculating conditional mutual information,we analyzed the direct and indirect causal relationships of the prominent variables.Finally,based on the tracing process,we determined a concise industrial process fault tracing graph and analyze the root cause of the fault,enabling process recovery.Using data from a chilled water system and the Tennessee Eastman industrial process,we conducted causal analysis for fault tracing,verifying the effectiveness and practicality of this tracing method.This method not only improves the efficiency of fault tracing but also enhances the accuracy and interpretability of the results.关键词
图形套索/传递熵/条件互信息/因果推理/工业过程/故障溯源Key words
graphical lasso/transfer entropy/conditional mutual information/causal reasoning/industrial process/fault traceability分类
计算机与自动化引用本文复制引用
魏淑娟,齐咏生,刘利强,李永亭,高学金..基于BDGL-CMI的传递熵网络因果推理的工业过程故障溯源[J].高校化学工程学报,2025,39(5):876-889,14.基金项目
国家自然科学基金(62363029,62241309) （62363029,62241309）
内蒙古科技计划(2020GG0283,2021GG0164) （2020GG0283,2021GG0164）
内蒙古自然科学基金(2022MS06018,2021MS06018). （2022MS06018,2021MS06018）
