Remote Diagnosis of Fault Element in Active Phased Arrays using Deep Neural Network

It can be extremely mission-critical to identify the source of an error in satellite communication systems and in this initial work an approach based on neural networks is proposed for diagnosing the faulty antenna elements remotely from the receiver side. A simple active antenna array which consists of four power amplifiers and a four-by-one linear antenna array has been considered in this work for proof of concept. Signals captured from various fault scenarios are used to train a 4layered feed-forward neural network using amplitude and phase data. The validation results show that the trained neural network can correctly identify the fault power amplifier with an accuracy higher than 96 %, which indicates the promising potential of the proposed approach.