Feature Learning for Fault Detection in High-Dimensional Condition-Monitoring Signals

TL;DR

This paper introduces an unsupervised feature learning and one-class classification method using stacked Extreme Learning Machines for fault detection in high-dimensional industrial sensor data, trained only on healthy conditions.

Contribution

It presents a novel integrated approach combining HELM autoencoders with one-class classifiers for effective fault detection using only healthy data for training.

Findings

Outperforms traditional methods like PCA and SVM in fault detection accuracy.

Effective in high-dimensional, non-informative sensor data scenarios.

Validated on synthetic and real power plant data.

Abstract

Complex industrial systems are continuously monitored by a large number of heterogeneous sensors. The diversity of their operating conditions and the possible fault types make it impossible to collect enough data for learning all the possible fault patterns. The paper proposes an integrated automatic unsupervised feature learning and one-class classification for fault detection that uses data on healthy conditions only for its training. The approach is based on stacked Extreme Learning Machines (namely Hierarchical, or HELM) and comprises an autoencoder, performing unsupervised feature learning, stacked with a one-class classifier monitoring the distance of the test data to the training healthy class, thereby assessing the health of the system. This study provides a comprehensive evaluation of HELM fault detection capability compared to other machine learning approaches, such as stand-alone one-class classifiers (ELM and SVM), these same one-class classifiers combined with traditional dimensionality reduction methods (PCA) and a Deep Belief Network. The performance is first evaluated on a synthetic dataset that encompasses typical characteristics of condition monitoring data. Subsequently, the approach is evaluated on a real case study of a power plant fault. The proposed algorithm for fault detection, combining feature learning with the one-class classifier, demonstrates a better performance, particularly in cases where condition monitoring data contain several non-informative signals.