MUMBAI, India, Oct. 12 -- Intellectual Property India has published a patent application (202641117979 A) filed by Roopa Nayak on October 01, 2026, for System And Method For Degradation Aware Adaptive Torque Sharing In Dual Motor Electric Drives Using Reinforcement Learning And Electrothermal Feedback.
Inventors include Roopa Nayak; Dr. Rashmy Deepak; and Dr. Sumathi S.
The application for the patent was published on October 09, 2026, under issue no. 41/2026.
Abstract: Parallel and mechanically coupled electric drives commonly use equal torque sharing, although manufacturing variation, resistance growth, cooling degradation, and unequal initial temperatures can cause the two motors to operate with different electrical and thermal efficiencies. This work presents a degradation-aware adaptive torque-sharing controller for a dual induction-motor drive. The controller combines a symmetry-invariant observation structure, a deep deterministic policy gradient agent, and an electrothermal supervisor. The learned policy determines the magnitude of torque redistribution, while the supervisor assigns its direction so that the healthier motor carries the appropriate additional load. The continuous action is mapped to a bounded sharing magnitude between 0 and 0.06, preserving equal sharing in healthy operation and limiting redistribution under degraded conditions. The reward accounts for total copper loss, torque and speed tracking, peak temperature, inter-motor temperature imbalance, torque imbalance, and action smoothness. Training uses randomized ambient temperature, load torque, load-application time, degradation severity, affected motor, electrical resistance, thermal resistance, and initial temperature, thereby exposing the policy to healthy, mild, moderate, and strong degradation. The proposed method was evaluated against equal sharing, a fixed-magnitude controller, and an oracle torque-sharing map. In the standard 15-case evaluation, the selected 200-episode agent achieved a mean degraded copper-loss saving of 6.6835 W (0.9434%) relative to equal sharing, closed 96.986% of the available oracle loss-reduction gap on average, and maintained a minimum gap closure of 86.614%. All cases satisfied the imposed speed- and torque-tracking feasibility criteria, with no action-limit violations. Across 30 unseen and extrapolation cases, the same agent reduced mean copper loss by 2.9630 W and 2.7669 W, respectively, relative to equal sharing, without producing a worse or infeasible case. Aggregate comparisons further showed that the adaptive controller produced the best reward and lowest copper-energy consumption, whereas the fixed controller achieved a slightly lower peak temperature. These results indicate that the learned controller does not merely reproduce a fixed sharing ratio: it adjusts the sharing magnitude with load and degradation severity and provides a favorable energy-thermal trade-off. The approach offers a practical framework for efficiency-oriented and thermally aware load allocation in fault-tolerant multi-motor propulsion and industrial drive systems.
Disclaimer: Curated by HT Syndication.
