TRAIDMA (Transit Reliability and AI Driven Maintenance Architecture)

FTL's “Transit Reliability and AI-Driven Maintenance Architecture (TRAIDMA)” leverages FTL’s experience in fleet maintenance, graph databases, and machine learning to provide a decision support system enabling transit agencies to engage in predictive maintenance to ensure fleet-wide state-of-good-repair. By leveraging AI-driven models and condition-based monitoring, agencies can predict equipment failures, optimize maintenance schedules, and estimate future capital needs for fleet replacement and rehabilitation.TRAIDMA uses AI model insights to generate a real-time fleet manager dashboard that aggregates the predictive outputs into an at-a-glance view of fleet health providing context-driven recommendations that are easy for maintenance staff to interpret, and cost-aware predictions that provide insight into the financial impact of maintenance actions. FTL will be working with the UMass Transportation Center and the Pioneer Valley Transit Authority (PVTA) to acquire and pre-process a comprehensive maintenance and vehicle telemetry dataset. Two complementary machine learning models leveraging XGBoost and LSTM architectures will be developed and evaluated to generate introspectable predictive maintenance insights. The effort will culminate in a refined system design, validating predictive performance across the entire fleet, and preparing for Phase II expansion, which will integrate broader data sources and additional failure modes for enhanced reliability and scalability.