Deep Learning-Based Optimization of Bionic Cooling Ducts for Heavy-Duty Vehicle Electric Drive Systems

Liquid cooling has become a predominant method for electric motor thermal management due to its superior heat transfer coefficient. However, conventional cooling channel designs often struggle to achieve an optimal balance between thermal dissipation performance and pressure loss. Inspired by the streamlined morphological features of diving beetles, this paper proposes a novel bionic cooling channel design. By extracting the hydrodynamic profile of the diving beetle’s body, a series of bionic diving beetle-inspired spoiler tubes were designed and integrated into the cooling ducts to optimize flow patterns. A deep learning-based framework was established to correlate these bionic geometric parameters with the thermal–hydraulic performance of the motor. The results demonstrate that the developed ANN model achieves an average R2 of 0.95 across the three output parameters, indicating high predictive accuracy. Compared to traditional structures, the optimized biomimetic channel yielded a 16.09% reduction in pressure drop relative to the conventional reference channel at the same flow rate. Furthermore, under real-vehicle operating conditions, the maximum winding temperature of OBDBS was comparable to that of IBDBS, while the cooling-channel pressure drop was reduced by 7.34%. This study demonstrates that the integration of bionic morphological principles with the GA-ANN-NSGA-II multi-objective optimization framework can reduce the hydraulic power requirement while maintaining a comparable winding temperature, providing an efficient optimization pathway for the thermal management of high-power-density motors.