<![CDATA[Tire failure is a major contributor to road fatalities, accounting for 3-15 of fatal accidents globally. Conventional Tire Pressure Monitoring Systems (TPMS) are limited by their reactive nature, as they only alert drivers to under-inflation after it occurs and cannot predict gradual wear or Remaining Useful Life (RUL). This research proposes a hybrid Predictive Maintenance framework that integrates Machine Learning with the Weibull reliability distribution to provide dynamic RUL estimations. Unlike existing simulation-centric or static models, the developed system utilizes an edge-cloud architecture centered on an ESP32-S3 gateway to capture telemetry from TPMS, OBD-II, and GPS modules. By treating the Weibull scale parameter (λ) as a dynamic covariate influenced by operational stressors such as engine load, vehicle speed, pressure, and temperature, the model continuously adapts to the tire’s degradation trajectory. Empirical results from field testing demonstrate that the covariate-driven approach significantly outperforms static baselines. The LightGBM regressor achieved the highest precision with a near-negligible MAPE of 0.0025 and a MASE of 0.2168. Specialized reliability metrics, including MQWE, further confirm the model's ability to minimize high-risk outliers, while an analysis of pessimistic error profiles ensures that maintenance is prompted before catastrophic failure. This study provides a scalable, real-time solution for in-vehicle Prognostics and Health Management (PHM), offering a robust pathway to reducing tire-related road casualties.]]>
Copyrights © 2026
