PRISM - Predictive Reasoning Intelligent System for Maintenance

PRISM (Predictive Reasoning Intelligent System for Maintenance) addresses a critical challenge facing transit agencies: maintaining bus fleets efficiently while preventing service disruptions. Using innovative artificial intelligence approaches, PRISM combines Knowledge Graphs and Graph Neural Networks to predict maintenance needs before failures occur. By working with Dallas Area Rapid Transit, this Phase I SBIR project will demonstrate the feasibility of representing complex bus systems and their maintenance relationships in a Knowledge Graph (KG) structure that captures both component interactions and temporal degradation patterns. The project will develop and validate GNN architectures using the KG and specifically designed for transit maintenance prediction, with emphasis on learning from statistically rare failure data while providing transparent, explainable recommendations. PRISM will help transit agencies transition from reactive to proactive maintenance approaches, improving safety, reducing costs, extending vehicle lifespans, and enhancing service reliability. The resulting system will provide maintenance staff with clear, actionable insights about developing problems and recommended interventions, creating a partnership between human expertise and artificial intelligence that optimizes maintenance planning and resource allocation.