root cause diagnosis文献综述

[1]Autonomous root-cause fault diagnosis using symbolic dynamic based
causality analysis, Neurocomputing, 2020.

方法思路`

（1）symbolic dynamic-based normalized transfer entropy is proposed to generate an initial causal graph of root-cause candidates.
(2)symbolic dynamic-based normalized direct transfer entropy is proposed and utilized for pruning the initial graph.
(3)a topological approach is proposed to locate the root-cause variables according to the pruned causal graph

案例

TEP
[2]Data-driven root cause diagnosis of process disturbances by exploring causality change among variables

方法思路

（1）同时考虑正常样本信息和故障信息（ to filter the
disturbance-unrelated causality and obtain compact and useful disturbance-related information）
（2）采用Kolmogorov-Smirnov test去分析因果关系

案例

TEP
[3]Root cause analysis of industrial faults based on binary extreme gradient boosting and temporal causal discovery network,Chemometrics and Intelligent Laboratory Systems, 2022.

方法思路

(1)binary-extreme gradient boosting (Bi-Xgboost) is proposed to analyze the fault contribution of variables
(2) a fault variables screening method based on the number of variables called mean contribution threshold (MCT) is proposed for screening the appropriate number of fault variables
(3) a temporal causal discovery network (TCDN) is introduced for root cause analysis with causal time lag information

案例

TEP
[4] Root cause diagnosis for complex industrial process faults via spatiotemporal coalescent based time series prediction and optimized Granger causality

方法思路

提出基于 spatiotemporal coalescent-based prediction model (SCPM)和causality verification algorithm (CVA)的改进格兰杰因果分析方法进行故障溯源

案例

TEP and BF process
[5]Root cause diagnosis in multivariate time series based on modified temporal convolution and multi-head self-attention, 2022, JPC

方法思路

(1)The temporal convolutional network is modified through feature reconstruction and skip connection to heighten the ability of feature extraction, which are further used to mine causalities
(2)The multi-head self-attention is modified via
threshold normalization for quantifiable causal inference, elevating the accuracy and generalization
performance
(3)the calculation rule of root node score based on contrastive causal graph is proposed for root cause diagnosis

案例

(1) A numerical case
(2)TEP
[6]Causal Generative Model for Root-Cause Diagnosis and Fault Propagation Analysis in Industrial Processes, 2023,TIM

方法思路

(1) a novel metric, recovery ratio (RR), is developed to represent the ratio of the restored samples whose fault detection statistics fall below the control limit (CL). Fault tracing is effectively realized by comparing the RRs of different restored samples;
(2) the dynamic feature extraction (DFE) network is utilized to extend the method to dynamic processes with nonlinear variable relationship
(3) a hierarchical CTVGM-based fault tracing strategy
is designed, which first traces the fault at the device level, and then stretches into the variable level to obtain fine-grained diagnosis results.

案例

(1) A numerical case
(2)TEP
[6] Hierarchical Causal Graph-Based Fault Root Cause Diagnosis and Propagation Path Identification for Complex Industrial Process Monitoring, 2023, TIM

Motivation

There are numerous control loops in the processing system, and the closed-loop feedback makes units strongly couple. This makes the faults propagate
and evolve with the material and energy flow networks.

方法思路

(1)the process is divided into several subblocks according to process knowledge. The monitoring model is built for fault detection.
(2)the fault-related subblocks and variables are selected based on the contribution plot, and the causal analysis based on transfer entropy (TE) is performed on them to construct HCG.
(3)a propagation path identification method based on
causal feature long short-term memory (CF-LSTM) is designed

案例

hot strip mill process (HSMP)
[7]Multiscale Partial Symbolic Transfer Entropy for Time-Delay Root Cause Diagnosis in Nonstationary Industrial Processes, TIE, 2023

Motivation

—Nonstationary characteristics and lack of time
delays analysis remain two large obstacles to diagnosing fault root cause in industrial processes. However, ignoring these two issues, many traditional methods tend to introduce some spurious or indirect causal relationships that interfere with root cause judgments.

方法思路

(1) Two complementary causal criteria of MPSTE are designed for determining causal direction between nonstationary variables, avoiding confusing bidirectional causality
(2) A novel strategy to construct a causal diagram with time delay is developed for root cause diagnosis in nonstationary processes, tracing the fault propagation path and delay with a concise directed diagram

案例

a coal mill rig
[8]Nonlinear Dynamic Granger Causality Analysis
Framework for Root-Cause Diagnosis of Quality-Related Faults in Manufacturing Processes, IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, 2022

方法思路

(1) An optimal variable division based on minimal redundancy maximal relevance algorithm is presented to get the quality-related variables
（2）a new attention-based random disturbance gated recurrent unit is designed for nonlinear dynamic Granger causality analysis, aiming at locating the root-cause and identifying the propagation path of quality-related faults

案例

hot strip mill process
[9] Partial Cross Mapping Based on Sparse Variable
Selection for Direct Fault Root Cause Diagnosis for Industrial Processes, IEEE TNNLS, 2023

方法思路

(1)generalized Lasso-based variable selection is performed. The Hotelling T2 statistic is formulated and the Lasso-based fault reconstruction is applied to select candidate root cause variables;
(2)the root cause is diagnosed through the PCM and the propagation path is drawn out according to the diagnosis result

案例

a numerical example
the Tennessee Eastman benchmark process
the wastewater treatment process(WWTP)
the decarburization process of high-speed wire rod
spring steel.
[9]A new transfer entropy approach based on information granulation and clustering for root cause analysis,2023,CEP

传递熵方法的局限性

（1）It requires the studied process be stationary whereas the presence of faults may lead to nonstationary changes;
(2) The computational complexity of TE is high.

思路

（1）the Information Granulation based Transfer Entropy (IGTE) and Information Granulation based Direct Transfer Entropy (IGDTE) are proposed to infer causal and direct causal relations
（2）) an Ordering Points To Identify the Clustering Structure (OPTICS) clustering-based Probability Density Function (PDF) estimator is designed to estimate joint/conditional probabilities based on the information granule.

案例

（1）a two-tank system
（2）Tennessee Eastman Process
[10]Graph attention network with Granger causality map for fault detection and root cause diagnosis,2023, computers and chemical engineering.

方法思路

（1）Extract causal relationships from normal data to construct a causal map, serving as the adjacency
matrix for the Graph Attention Network (GAT). This ensures predictions depend solely on causally related variables
（2）Encode past measurements into hidden vectors as node features using 1D-CNN. The Graph Attention Network (GAT) leverages the causal map to make predictions, capturing variable interactions effectively
（3）Detect faults by analyzing prediction errors from the trained CNN-GAT model. Abnormal prediction errors in new data deviating from the training data could indicate the occurrence of faults.
（4）After fault detection, identify a set of fault-affected variables based on large prediction errors. Use the
learned causal map to pinpoint the root cause of the fault from the set of fault-affected variables.

案例

TEP