Predictive temperature control of electric two wheeler hub motor using gradient aware neural regulation with degradation tracking and fault tolerant multi condition torque adaptation

Thermal stress in electric two-wheeler hub motors is a major bottleneck affecting drivetrain reliability, safety, and efficiency-especially under varied terrain, aggressive riding profiles, and sensor reliability degradation. This paper presents a novel Hybrid Gradient-Aware Neural Regulation (GANR) framework designed to enable predictive motor temperature control, real-time thermal degradation monitoring, and fault-tolerant torque shaping. The system integrates three distinct yet interlinked control layers: (i) gradient-sensitive neural temperature estimation, (ii) cumulative thermal damage indexing through a Motor Health Index (MHI), and (iii) multi-condition torque derating logic that adapts to ambient temperature, health state, and usage intensity. The controller combines a physics-based thermal model with a lightweight neural network estimator capable of learning under sensor drift, latency, or complete failure. A fallback and safety-mode flowchart ensure robust operation by switching to boundary-scaled torque caps and enabling a real-time decision layer that balances energy consumption, performance demand, and thermal safety. The system's intelligence also extends to an ambient-aware derating boundary and a thermal margin estimator to prevent runaway heating events. A detailed simulation framework is developed using real vehicle specifications, encompassing city, aggressive, and worst-case thermal cycles. Results reveal that the proposed control system can delay torque derating by up to 14%, improve energy efficiency by ~ 7.4%, and achieve prediction RMSE below 2.1 °C even under multiple sensor fault scenarios. In addition, a full real-time deployment analysis is presented using a mid-tier embedded platform with CPU-cycle, memory, and scheduling footprint assessments. This study offers an original, fault-resilient, and hardware-feasible solution to a largely unaddressed domain-intelligent thermal-aware torque control in electric two-wheelers, creating a new pathway toward smarter, safer, and more robust electric mobility systems.