Data-driven deep neural network for structural damage detection in composite solar arrays on flexible spacecraft

A data-driven approach based on Deep Neural Network (DNN) techniques is here proposed for Structural Health Monitoring of large in-orbit flexible systems. Damage scenarios are generated via a Finite Element commercial code to train and test the machine learning model, by considering equivalent properties of the composite material of the solar panels. The fully coupled 3D equations for the flexible spacecraft are integrated to test typical profiles of attitude manoeuvres in case of different damages. The DNN model is trained using sensor-measured time series responses, with each response associated with the label of the corresponding damage scenario, and tested via k-folding approach. This methodology offers a promising approach to detect structural damage in solar arrays on spacecraft using machine learning techniques.