About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
Advances in Magnetism and Magnetic Materials
|
| Presentation Title |
Accelerating Traction Motor Design Using AI-Based Finite Element Surrogates |
| Author(s) |
Vandana Rallabandi, Pedro Ribeiro, Burak Ozpineci |
| On-Site Speaker (Planned) |
Vandana Rallabandi |
| Abstract Scope |
The design of electric motors is a constrained multi-objective optimization problem in which electromagnetic, thermal, mechanical and inverter requirements must be satisfied simultaneously. Interior permanent magnet (IPM) machines are widely adopted in traction because they combine high torque density, high efficiency and wide speed-range capability. However, their performance is highly sensitive to geometry requiring a systematic exploration of competing objectives and constraints. High-fidelity finite element analysis (FEA) accurately captures electromagnetic performance but is the main computational bottleneck when thousands of candidate designs must be assessed. This work proposes using a limited set of FEA simulations to generate a training dataset for an artificial neural network model that emulates FEA with substantially reduced evaluation time, thereby enabling its use in optimization. This approach is illustrated for an IPM traction motor designed to satisfy a required rated torque while considering voltage and current constraints and minimizing loss and permanent magnet volume. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Modeling and Simulation, Machine Learning, |