Deep Anomaly Detection on Tennessee Eastman Process Data

TL;DR

This paper evaluates modern deep learning-based unsupervised anomaly detection methods on the Tennessee Eastman process dataset, providing insights to guide industrial application choices.

Contribution

It offers the first comprehensive analysis of deep anomaly detection methods on this benchmark dataset, aiding practical industrial deployment.

Findings

Deep learning methods outperform traditional approaches in anomaly detection accuracy.

Certain models show robustness across different types of process anomalies.

The study identifies key factors influencing detection performance.

Abstract

This paper provides the first comprehensive evaluation and analysis of modern (deep-learning) unsupervised anomaly detection methods for chemical process data. We focus on the Tennessee Eastman process dataset, which has been a standard litmus test to benchmark anomaly detection methods for nearly three decades. Our extensive study will facilitate choosing appropriate anomaly detection methods in industrial applications.