Multi-Objective Autoencoder for Fault Detection and Diagnosis in Higher-Order Data

Abstract

We propose a multi-objective autoencoder method for fault detection and diagnosis in multi-way data based on the reconstruction error of autoencoder deep neural network (ADNN). Multi-way data analysis has become an essential tool for capturing underlying structures in higher-order data sets. Our method fuses data from multiple sources in one ADNN at which informative features are being extracted and utilized for anomaly detection. It also uses the generated anomaly scores to asses the severity of the anomalous data and localize it via a localization layer in the autoencoder. We evaluated our method on multi-way datasets in the area of structural health monitoring for damage detection purposes. Experimental results show that the proposed method can accurately detect structural damage. It was also able to estimate the different levels of damage severity, and capture damage locations in an unsupervised aspect. Compared to the state-of-the-art approaches, our proposed method shows better performance in terms of damage detection and localization.