Sarcouncil Journal of Engineering and Computer Sciences

Abstract: This article investigates the convergence of artificial intelligence, telematics, and predictive maintenance methodologies in connected vehicle ecosystems. It examines how machine learning algorithms, particularly Graph Neural Networks, can process complex telematics data streams to detect anomalies, predict component failures, and optimize maintenance scheduling. The article demonstrates that AI-integrated predictive maintenance systems can reduce unplanned downtime while decreasing overall maintenance costs compared to traditional schedule-based approaches. The architecture proposed in this study leverages edge-cloud collaborative processing, multimodal sensor fusion, and real-time streaming capabilities to enable proactive rather than reactive maintenance paradigms. By modeling vehicles as complex systems with interdependent components, it captures subtle patterns in the relationships between subsystems that precede failures. The implementation of transfer learning methodologies further enables knowledge sharing across heterogeneous fleets, while event-driven processing pipelines deliver actionable insights with minimal latency. This data-centric framework fundamentally transforms fleet management operations by converting maintenance from a reactive cost center to a proactive strategic advantage.