Analysis of space solar array arc images based on deep learning techniques

Abstract

One of the critical challenges for solar arrays operating in space is the occurrence of discharging and arcing on the array surfaces exposed to plasma environments. Arcs can lead to severe damage to cell interconnectors due to the generated high peak currents and therefore significantly impact the performance and reliability of spacecraft systems. Solar arrays with highly negative biases are particularly prone to frequent arcs, which can escalate into sustained arcs, causing harm to spacecraft and satellite components. The presented study aims to investigate discharging and arc spectra on solar cell surfaces through the analysis of arc images. Leveraging advanced Deep Learning (DL) methodologies, including Convolutional Neural Networks (CNN) and Transfer Learning, a robust predictive model has been developed to analyze arc behavior and identify defective cells based on image data. Furthermore, algorithms and image processing tools, such as Python and Maxim-DL, are employed to examine variations in intensities and spatial variation in the arced region. The findings highlight the prevalence of intensive arcs, particularly on mid-cells and interconnectors, and provide insights into the dynamics of sustained arc events. The image analysis can advance the understanding of arc evolution, enabling improved mitigation strategies for solar array systems in space applications.