AI-mag: Inductor Design Tool

Inductor Design with Artificial Neural Networks

The Power Electronic Systems Laboratory at ETH Zürich is pleased to publish "AI-mag", a new inductor optimization tool! This tool combines the accuracy of the Finite Element Method (FEM) with the evaluation speed of Artificial Neural Network (ANN).

Moreover, the software is open-source (MATLAB, Python, and COMSOL) and the working principles are described in an open-access scientific paper (T. Guillod, P. Papamanolis, and J.W. Kolar, Artificial Neural Network (ANN) Based Fast and Accurate Inductor Modeling and Design, IEEE Open Journal of Power Electronics, 2020).

Capabilities and Performances

• Complete model: thermal loss coupling, core loss map, HF losses, etc.
• Versatile model: geometry, core material, winding stranding, excitation, etc.
• Accurate model: less than 3% deviation with 3D FEM
• Fast model: compute 50'000 designs per second
• Multi-objective optimization: losses, volume, mass, cost, etc.

Some Figures

FEM/ANN Modeling Workflow

Software Screenshots

Optimized Inductor (2kW DC-DC, 2.5W of losses @ 500kHz)

Going Further

• Website - https://ai-mag.github.io
• IEEE Paper - https://doi.org/10.1109/OJPEL.2020.3012777
• GitHub - https://github.com/ethz-pes/AI-mag
• PES Laboratory: https://pes.ee.ethz.ch