Simultaneous fault detection in satellite power systems using deep autoencoders and classifier chain

A satellite system’s health is heavily dependent on the proper functioning of the Satellite Power System (SPS), which is regarded as the core component of a satellite. Fault detection and diagnosis (FDD) plays a vital role in maintaining the stable and efficient operation of the SPS and ensuring the success of a satellite mission. Automated FDD can reduce the burden and false alarm rate associated with manual level checking of individual sensors, by effectively leveraging the correlation between sensor measurements. In many real-world applications, multiple faults can occur simultaneously and SPS is not an exception. Simultaneous fault diagnosis is especially challenging, involving detection of multiple faults occurring concurrently. In this paper, this problem is addressed using a multilabel classification model. A Deep Autoencoder and Random Forest based Classifier Chain is employed for this purpose. The proposed model is used not only for fault detection and classification, but also for localizing single as well as simultaneous sensor faults in SPS. NASA’s Advanced Diagnostic and Prognostic Testbed (ADAPT) dataset has been used for validating the system, yielding an accuracy as high as 94.63% and precision, recall, and F1-score of 0.95, 0.91 and 0.93 respectively.