Predicting the evolution of local thermal fields within frictional contacts remains a major challenge in tribology, due to the multi-scale and time-dependent nature of interface transformations. In dry pin-on-disc systems, the contact interface is closed and inaccessible, making direct observation of load-bearing zones impossible during operation. This work versus transient tribological regimes. Furthermore, by modifying the normal force input, the model highlights the strong sensitivity of thermal behavior to loading conditions, suggesting a pathway toward force-based stabilization strategies and data-driven smart braking. This study provides a robust experimental-AI methodology for monitoring, predicting, and potentially controlling frictional systems where direct access to the interface is not possible.