Explainability: Relevance based Dynamic Deep Learning Algorithm for Fault Detection and Diagnosis in Chemical Processes

TL;DR

This paper introduces a relevance-based explainability method for deep learning models to improve fault detection and diagnosis in chemical processes, demonstrating enhanced accuracy and robustness on a benchmark dataset.

Contribution

It proposes a novel relevance measure derived from Layerwise Relevance Propagation to enhance deep neural network FDD performance with small datasets.

Findings

Relevance-based feature selection reduces overfitting.

Improved fault detection accuracy on Tennessee Eastman Process.

Enhanced distinguishability between fault classes.

Abstract

The focus of this work is on Statistical Process Control (SPC) of a manufacturing process based on available measurements. Two important applications of SPC in industrial settings are fault detection and diagnosis (FDD). In this work a deep learning (DL) based methodology is proposed for FDD. We investigate the application of an explainability concept to enhance the FDD accuracy of a deep neural network model trained with a data set of relatively small number of samples. The explainability is quantified by a novel relevance measure of input variables that is calculated from a Layerwise Relevance Propagation (LRP) algorithm. It is shown that the relevances can be used to discard redundant input feature vectors/ variables iteratively thus resulting in reduced over-fitting of noisy data, increasing distinguishability between output classes and superior FDD test accuracy. The efficacy of the proposed method is demonstrated on the benchmark Tennessee Eastman Process.