⚡ electrical engineering

Multi-Material Topology Optimization with Continuous Magnetization Direction for Permanent Magnet Synchronous Reluctance Motors

This paper proposes a novel density-based multi-material topology optimization framework with continuous magnetization direction and Nitsche-type mortaring to design high-torque permanent magnet-assisted synchronous reluctance motors, utilizing a topological derivative-inspired interpolation scheme and K-means clustering to ensure technical feasibility while minimizing computational cost.

Original authors: Thomas Gauthey, Peter Gangl, Maya Hage Hassan

Published 2026-04-01

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Thomas Gauthey, Peter Gangl, Maya Hage Hassan

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are an architect tasked with designing the most efficient engine for a new electric car. But there's a catch: you aren't just drawing blueprints; you are allowed to invent the shape of the engine from scratch, deciding exactly where to put metal, where to leave empty space, and where to place powerful magnets.

This paper is about a team of researchers who developed a new "digital sculpting" tool to solve this exact problem for a specific type of electric motor called a Permanent Magnet Synchronous Reluctance Motor (PMSynRM).

Here is the story of their discovery, broken down into simple concepts:

1. The Problem: The "Goldilocks" Motor

Electric motors usually come in two flavors:

The Iron Motor: Cheap and simple, but not very strong.
The Magnet Motor: Very strong, but expensive because magnets are costly.

The researchers wanted to build a "hybrid" motor that uses just enough iron and just enough magnets to get the maximum strength without wasting money. It's like trying to bake the perfect cake where you need just the right amount of flour and sugar to make it rise, but you don't want to waste any ingredients.

2. The Old Way vs. The New Way

The Old Way (The Lego Set):
Previously, engineers designed these motors like they were building with Lego bricks. They had to decide: "Is this spot a magnet? Is it iron? Is it air?" And if they used a magnet, they had to pick a fixed direction, like "North-South." It was rigid, like trying to fit a square peg in a round hole.

The New Way (The Clay Sculptor):
This paper introduces a method that treats the motor's interior like wet clay.

Instead of snapping bricks together, the computer "molds" the material.
It can decide that a spot is 30% iron and 70% air, or that a magnet's direction can curve smoothly from one angle to another, just like a river flowing.
The Magic Trick: They created a new mathematical "interpolation" (a blending recipe) that allows the magnet's direction to change continuously. Imagine a compass needle that doesn't just point North or East, but can point anywhere in between, smoothly, depending on what the motor needs at that exact moment.

3. The Speed Bump: The "Four-Point" Shortcut

Calculating how much torque (rotational power) a motor produces is like trying to predict the weather; it requires simulating thousands of tiny moments as the motor spins. Doing this for every single design idea would take years of computer time.

The Solution: The researchers realized they didn't need to check every single second of the spin. They found a "shortcut" where checking just four specific positions of the rotor (like checking the weather at 6 AM, 12 PM, 6 PM, and Midnight) was enough to accurately predict the average performance for the whole day. This cut the computing time down from an eternity to a manageable afternoon.

4. The "Ghost" Problem and the "Clustering" Fix

When the computer molds the clay, it sometimes gets a bit messy. It might create a magnet that points in a slightly different direction for every tiny speck of material. In the real world, you can't manufacture a magnet that changes direction every millimeter; it has to be a solid block with one clear direction.

The Fix (K-Means Clustering):
After the computer finishes its perfect digital design, the researchers use a "clean-up crew" (called K-Means clustering).

Imagine you have a bucket of marbles of all different colors.
The algorithm groups them into 5 distinct piles (clusters) based on their color and location.
It then tells the engineers: "Okay, this whole section of the motor should be one solid magnet pointing in direction A, and this next section should be a magnet pointing in direction B."
This ensures the final design is something a factory can actually build.

5. The Result: A Stronger, Smarter Motor

By using this new "clay sculpting" method, the researchers found designs that were significantly stronger than traditional ones.

They managed to distribute the magnets and iron in shapes that no human engineer would have thought of (often looking like organic, flowing patterns rather than geometric blocks).
They maximized the "push" (torque) the motor gives, making the electric vehicle more efficient.

The Big Picture

Think of this paper as the difference between drawing a map and using a GPS.

Old method: Engineers drew a map based on what they knew worked before.
New method: The computer explores the entire landscape, finding hidden paths (optimal shapes) that humans missed, and then gives you a clear, drivable route (the manufacturable design) to get there.

This technology promises to make electric motors cheaper, lighter, and more powerful, helping to drive the future of electric transportation.

1. Problem Statement

The paper addresses the design optimization of Permanent Magnet-assisted Synchronous Reluctance Motors (PMSynRM). These motors offer higher average torque than standard synchronous reluctance motors (SynRM) but present a complex design challenge due to the need to simultaneously optimize:

Material Distribution: The spatial arrangement of three distinct materials: ferromagnetic iron, air, and permanent magnets within the rotor.
Magnetization Direction: Unlike traditional approaches that fix magnetization to a discrete set of angles, this study seeks to determine a continuous magnetization direction field to maximize performance.
Constraints: The optimization must adhere to manufacturing constraints (e.g., volume limits for iron and magnets) and technical feasibility (discrete magnetization directions for production).

The primary objective is to maximize the average electromagnetic torque while respecting volume constraints and the non-linear magnetic properties of the materials.

2. Methodology

The authors propose a density-based topology optimization framework integrated with advanced numerical techniques to handle the multi-material and continuous magnetization challenges.

A. Mathematical Formulation

Governing Equations: The problem is modeled using 2D magnetostatics. The state equation is a non-linear boundary value problem involving the magnetic vector potential $u$ , where the reluctivity $\nu$ depends on the flux density (non-linear iron) and the magnetization vector $\vec{M}$ is a design variable.
Torque Calculation: To reduce computational cost, the average torque is approximated using a four-point static method (evaluating torque at rotor positions $0, \pi/12, \pi/6, \pi/4$ ). This was validated against 500-point calculations, showing negligible error (<0.1%) for optimized designs.
Optimization Problem: The goal is to maximize the average torque subject to volume constraints for iron and magnets. The problem is solved using an Augmented Lagrangian method to handle inequality constraints.

B. Key Innovations in Interpolation and Mapping

Continuous Magnetization Direction:
- Instead of fixing angles, the authors use two density variables ( $\rho_{Mx}, \rho_{My}$ ) to represent the magnetization vector.
- To ensure the magnetization magnitude remains constant regardless of direction (avoiding artificial anisotropy), a Square-to-Disk transformation is applied. This maps the unit square of density variables onto a unit disk, preserving the angle while maintaining a constant maximum magnetization norm.
Novel Interpolation Scheme:
- The paper introduces a new material interpolation function inspired by the topological derivative.
- Unlike standard SIMP (Solid Isotropic Material with Penalization) which favors low-density materials, this new scheme ensures the derivative of the interpolation function matches the topological derivative at $\rho=0$ and $\rho=1$ . This is specifically tailored for the non-linear Marrocco BH-curve of the ferromagnetic material.
Handling Rotor Rotation (Nitsche Mortaring):
- To avoid remeshing the stator for every rotor position, the authors apply Nitsche-type mortaring. This allows the rotor mesh to rotate freely relative to the stator mesh while maintaining continuity of the solution across the air gap interface.

C. Post-Processing and Filtering

Filtering: A Helmholtz filter (PDE-based) is used to remove checkerboard patterns and ensure mesh independence.
Projection: A projection function (tanh-based) is applied to sharpen boundaries and reduce "grey" areas (intermediate densities).
K-Means Clustering: Since continuous magnetization is not manufacturable, a K-means heuristic is applied in post-processing. It clusters the continuous magnetization angles into a finite set of discrete directions (e.g., $k=5$ clusters) while respecting geometric proximity, ensuring the final design is technically feasible.

3. Key Contributions

Continuous Magnetization Optimization: The study successfully demonstrates a density-based approach where magnetization direction is a continuous variable during optimization, rather than a pre-defined discrete set.
Novel Interpolation Scheme: The introduction of a topological derivative-inspired interpolation scheme for non-linear magnetostatics, which improves convergence and design quality compared to standard polynomial schemes.
Efficient Computation Strategy: The combination of the four-point torque approximation and Nitsche mortaring significantly reduces the computational burden, making the optimization of complex multi-material PMSynRMs feasible.
Feasibility-Driven Post-Processing: The integration of K-means clustering to translate continuous optimization results into manufacturable designs with discrete magnetization angles.

4. Results

The method was validated on a specific PMSynRM geometry (24 slots, 1 pole pair, distributed winding).

Iron-Air Optimization: Results for SynRM (without magnets) matched literature values, validating the baseline optimization.
Multi-Material Optimization:
- Designs with varying magnet volume fractions (7.5% to 30%) were generated.
- Torque Improvement: The inclusion of optimized magnets significantly increased torque. For example, with 30% magnet volume, the average torque reached 1.9000 N.m (optimized) and 1.8671 N.m (after K-means post-processing).
- Comparison: The optimized designs outperformed a baseline SynRM (1.0768 N.m) and showed torque values comparable to or exceeding other state-of-the-art designs in the literature, even with higher magnet volumes.
Robustness: The method was tested with different phase angles ( $\phi = 3\pi/32$ ) and starting points (unbiased grey fields), consistently converging to high-performance topologies.

5. Significance and Conclusion

This research represents a significant step forward in the automated design of electric machines. By moving away from fixed magnetization angles and complex geometric parameterization, the proposed method allows the optimization algorithm to "discover" novel rotor topologies that maximize torque.

The ability to handle continuous magnetization and non-linear material properties within a unified density-based framework, while ensuring the final design is manufacturable through clustering, bridges the gap between theoretical optimization and practical engineering application. The use of Nitsche mortaring and the four-point torque approximation makes the approach computationally efficient enough for industrial-grade design cycles. Future work aims to extend this to multi-objective optimization and exhaustive comparisons with other interpolation schemes.