Abstract:Multivariate fault detection has been a requirement in various industrial systems to ensure safe operations and to improve product quality. The T-2 statistic is a widely used fault detection index for monitoring unusual changes in the mean of multivariate data. In spite of its advantages, the T-2 statistic also suffers from amplification and masking effects, especially when the number of variables to be monitored is large. The reason for the amplification effect is that the T-2 statistic often uses an enlarged control limit that is determined in the space of all variables to detect the faults caused by only a small part of variables. The masking effect is because the faulty components of the T-2 statistic are diluted by the normal (small) components to yield a T-2 value within the control limit. Both amplification and masking effects reduce the fault detection capability of the T(2 )statistic. To address this problem, improvements are made to the T-2 statistic in this paper. Based on an orthogonal decomposition of the standard T-2 (ST) statistic, an adaptive dimension reduction (ADR) method and the weighted T-2 (WT) and intersecting T-2 (IT) statistics are developed. For each observation of p variables, the ADR method searches over p subspaces (with dimensions from 1 to p) to find an optimal detection subspace, in which the difference between the computed T-2 value and the corresponding control limit is the largest. Then, the ADR method performs the ordinary T-2-based fault detection in the optimal subspace. The WT statistic is a weighted version of the ST statistic, where large components have large weights, while small components are given small weights. The IT statistic divides all components of the ST statistic into several groups, and then it performs the ordinary T-2-based fault detection in each group. The result of the IT statistic is the intersection of fault detection results of all groups. The ADR method and the WT and IT statistics all have better fault detection capabilities than the ST statistic because they are able to reduce the amplification and masking effects. Their advantages are illustrated with two case studies.

Assessment of T2- and Q-statistics for Detecting Additive and Multiplicative Faults in Multivariate Statistical Process Monitoring

New Fault Recognition Method Based on Multivariate Statistical Process Control

Statistical monitoring of dynamic multivariate processes - Part 2. Identifying fault magnitude and signature

Statistical Process Monitoring Based on the Kernel Mean Discrepancy Test

Fault Detection and Isolation of Non-Gaussian and Nonlinear Processes Based on Statistics Pattern Analysis and the K-Nearest Neighbor Method

Process monitoring and fault detection based on multivariate statistical projection analysis

SVM-based Multivariate Statistical Process Control Method and Its Application

Multivariate Statistical Process Monitoring Based on Blind Source Analysis

Improvements to the T-2 Statistic for Multivariate Fault Detection

A multivariable statistical process monitoring method based on multiscale analysis

A Comparison of Different Statistics for Detecting Multiplicative Faults in Multivariate Statistics-Based Fault Detection Approaches.

A combined passive-active method for diagnosing multiplicative fault

Fault Detection with Improved Principal Component Pursuit Method

Incipient Fault Detection with Smoothing Techniques in Statistical Process Monitoring

Root Cause Analysis in Multivariate Statistical Process Monitoring: Integrating Reconstruction-Based Multivariate Contribution Analysis with Fuzzy-Signed Directed Graphs

A Novel Dynamic Baysian Canonical Correlation Analysis Method for Fault Detection

A Distribution-Free Method for Process Monitoring

Two-Level Multiblock Statistical Monitoring for Plant-Wide Processes

Process monitoring and fault diagnosis method combining bagging DPCA‐ICA with moving window Kolmogorov–Smirnov test

Fault Detection Behavior And Performance Analysis Of Principal Component Analysis Based Process Monitoring Methods

Detection of Intermittent Faults Based on an Optimally Weighted Moving Average T^2 Control Chart with Stationary Observations