AI-mag: Fast and Accurate Inductor Modeling and Design with FEM/ANN

Inductors for Power Electronics

T. Guillod, P. Papamanolis, J.W. Kolar

Lightning Fast

Get thousands of designs per second
with artificial neural networks!

Extremely Accurate

Based on accurate 3D finite element
thermal and magnetic simulations!

Free / Open-Source

Complete code and data
under BSD license!

Pareto Fronts

Explore the Pareto fronts
with a graphical user interface!

Scientific Paper

Scientific background and results
available in a paper (IEEE OJ-PEL)!


Programmed in MATLAB with
the help of Python and COMSOL!

Hybrid FEM /ANN

Combine the speed of neural networks (ANNs) and the accuracy of finite element (FEM) for obtaining a versatile tool for inductor computation and optimization (buck, boost, resonant, filter, etc.)

Explore Pareto Fronts

Explore the trade-offs (power density, mass, losses, load profile, cost, etc.) with a graphical user interface and select the optimal design.

Select the Optimal Designs

Get all the properties of a selected design down to the smallest detail (geometry, equivalent circuit, field, losses, temperature, etc.).

ETHZ and PES Logo


AI-mag was developed and brought to you by the Power Electronic Systems Laboratory at ETH Zurich. The source code and the data are under BSD license.

Get more information!