The software tool developed by Stony Brook University uses self-supervised learning to detect long-term solar equipment damage weeks or years before manual inspections find it.