Fault Diagnosis In Satellite Power Systems Using PCA And Deep Learning
Fault detection and diagnosis is one of the key technologies for monitoring the functions of power systems in satellites. Most of the machine learning approaches used currently in spacecraft fault diagnosis have limitations in representing complex functions within small samples and cells. In addition, the ability of these methods to generalise is limited. Therefore, in this paper, we use Principal Component Analysis to identify the features and Deep Neural Network to diagnose the faults. Usually a classifier trains faulty data to identify the causes of the anomalies and this aspect has generally limited the use of model-driven approaches to fault detection tasks. The paper suggests a technique for using methods of machine learning for purposes of fault detection and diagnosis. The proposed method is more suitable for characterizing complex features of equipment information, allowing for more accurate identification of equipment health status under complex monitoring tasks. Here, the dataset used is acquired from the Advanced Diagnostic and Prognostic Testbed (ADAPT). The accuracy obtained through this method is 80% and thus can prove to be efficient in healthcare management of systems.