This paper presents a driving-cycle-aware shape and topology optimization workflow for interior permanent magnet synchronous machines used in traction drives. A k-means clustering approach reduces full driving cycles to representative operating points so that optimization remains computationally feasible while preserving realistic operating behavior. The workflow combines binary topology optimization, Normalized Gaussian Networks (NGnet), and spline-based shape optimization under electromagnetic, mechanical overspeed, and inverter voltage constraints. A Laplace-based mesh deformation strategy enables simultaneous optimization of magnet geometry and flux-barrier topology. Two optimized rotor designs are manufactured and tested experimentally. The central contribution is a validated, constraint-aware optimization pipeline that achieves permanent-magnet reduction of up to 10% while maintaining required torque capability and near-reference full-cycle efficiency.

翻译：本文提出了一种面向牵引驱动的内置式永磁同步电机驾驶循环感知形状与拓扑优化工作流。通过k均值聚类方法将完整驾驶循环简化为代表性工况点，从而在保持计算可行性的同时保留实际运行特性。该工作流结合了二元拓扑优化、归一化高斯网络（NGnet）与基于样条的形状优化，并考虑电磁、机械超速及逆变器电压约束。基于拉普拉斯网格变形策略，实现了磁极几何与磁障拓扑的同步优化。两种优化转子设计方案被制造并进行了实验验证。核心贡献在于建立了一种经实验验证的约束感知优化管道，在维持所需转矩能力与接近参考值的全循环效率的前提下，可实现永磁体用量减少高达10%。