Fault Detection in Industrial Systems Using Maximized Divergence Analysis Approach

doi:10.1109/ACCESS.2022.3181360

Fault Detection in Industrial Systems Using Maximized Divergence Analysis Approach

2022-01-01

3

OA

AI

DOI：10.1109/ACCESS.2022.3181360

Original

Original request for help

Save

Abstract

En 中文

Dimensionality reduction techniques including partial least-squares (PLS) and principal component analysis (PCA) have been widely applied for data-driven process monitoring. However, the objectives of PCA- and PLS-based techniques are not specific for fault detection where a superior detection performance results from a large divergence (i.e., difference) between normal operating data and faulty data. In this article, a maximized divergence analysis (MDA) method is proposed to detect faults in industrial systems. The objective of MDA is to directly maximizes the Kullback-Leibler (KL) divergence corresponding to the distributions of normal operating data and faulty data during the procedure of dimensionality reduction. An algorithm using eigenvalue-decomposition technique is put forward to efficiently solve the optimization problem of maximizing KL-divergence. Two-dimensional synthetic data and Tennessee Eastman process are used to demonstrate the effectiveness of the proposed MDA-based detection approach.

Keywords:

Fault detection

Dimensionality reduction

Principal component analysis

Loading

Process monitoring

Probability density function

Optimization

Dimensionality reduction technique

fault detection

fault diagnosis

process monitoring

Kullback-Leibler divergence

Tennessee Eastman process

Fault Detection in Industrial Systems Using Maximized Divergence Analysis Approach

Abstract

Journal

Organization