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

Area of Science:

• Space physics
• Electrical engineering
• Materials science

Background:

• Spacecraft solar arrays are vulnerable to plasma-induced arcing, which damages components and degrades performance.
• Highly negative biases on solar arrays increase the risk of frequent and sustained arcing events.
• Understanding arc dynamics is crucial for enhancing the longevity and reliability of space missions.

Purpose of the Study:

• To investigate the spectral characteristics of discharging and arcing on solar cell surfaces.
• To develop a predictive model for analyzing arc behavior and identifying defective solar cells using image data.
• To gain insights into the spatial and intensity variations of arced regions.

Main Methods:

• Utilized Deep Learning (DL) methodologies, including Convolutional Neural Networks (CNN) and Transfer Learning.
• Employed image processing tools and algorithms (Python, Maxim-DL) for arc image analysis.
• Analyzed arc images to examine intensity and spatial variations in arced regions.

Main Results:

• Developed a robust DL model for predicting arc behavior and detecting defective cells.
• Identified intensive arcs predominantly occurring on mid-cells and interconnectors.
• Provided insights into the dynamics and evolution of sustained arc events.

Conclusions:

• Image analysis of arcing events offers a pathway to better understand arc evolution.
• Findings can inform the development of improved mitigation strategies for solar array systems.
• Enhanced understanding of arcing contributes to increased spacecraft system reliability and performance.