Electric Machine Parameterization Using Vector Geometry AI Training

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for training artificial intelligence algorithms for electrical machine design and parameterization face challenges in representing diverse geometries and topologies due to limitations in 2D simulations, leading to issues like resolution problems, systematic errors, and the inability to consider freeform shapes, which hinder the discovery of new designs.

Innovation Solution

The method involves training an AI algorithm using vector graphics and polygons to represent electrical machine components, allowing for a more precise and flexible modeling of magnetic and mechanical behaviors without pixel-by-pixel subdivision, enabling the discovery of new geometries and arrangements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If pixel-based approaches are used to represent 2D geometry, then independence from templates is achieved, but resolution issues and systematic errors occur

Engineering Contradiction:

Improveindependence from templatesVSAvoidresolution accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent replaces the pixel-based mechanical representation system with a vector graphic system. Instead of using discrete pixels to represent geometry, the invention uses continuous vector graphics (polygons with vertices and edges) to describe component arrangements. This substitution eliminates the inherent resolution limitations and discretization errors of pixel-based approaches while maintaining template independence, as vector graphics can represent any geometry with arbitrary precision.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The invention changes the fundamental parameter representation from discrete pixel values to continuous vector parameters (coordinates, vertices, edges). By transforming the geometric representation from a raster grid to vector mathematics, the system achieves both template independence and high precision simultaneously, as vector graphics can be scaled and manipulated without loss of accuracy.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If pixel-based approaches are used to represent 2D geometry, then any 2D geometry can be represented, but processing of a large number of parameters is necessary

Engineering Contradiction:

Improvegeometry representation flexibilityVSAvoidparameter processing complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent substitutes the computationally intensive pixel-based parameter processing system with a more efficient vector graphic processing system. Instead of manipulating thousands of individual pixel values, the invention processes a much smaller set of vector parameters (vertices, edges, polygons), dramatically reducing computational complexity while maintaining the ability to represent any 2D geometry.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The invention segments the geometric representation into meaningful vector components (polygons representing different components, vertices defining boundaries, edges forming shapes) rather than processing every pixel individually. This segmentation allows the system to handle complex geometries by processing only the essential defining parameters of each component, significantly reducing the total parameter count.

Inventive Principle:

Principle #1Segmentation

3Adaptability or versatility

If pixel graphics are used for geometry representation, then template independence is achieved, but manufacturing validation requires interpolation techniques

Engineering Contradiction:

Improvedesign freedomVSAvoidmanufacturing validation

Core Design Contradiction:

Adaptability or versatilityVSEase of manufacture

Solution Approach 1:

The patent replaces pixel-based geometry representation with vector graphics, which are inherently suitable for manufacturing validation. Vector graphics provide exact mathematical definitions of component boundaries and positions, eliminating the need for interpolation techniques required by pixel-based approaches. This substitution ensures that designs can be directly validated for manufacturability without additional processing steps.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4651006A1Electric machine design method using a trained algorithm for electric machine design and/or parameterisation

Publication Date: 2025.11.19 ZF FRIEDRICHSHAFEN AG

Figure 1A~1B
Figure 2~3
Figure 4~5

AI summary

The approach presented here provides a method (400) for training an artificial intelligence algorithm (125) for the design and/or parameterization of an electric machine (100). The method (400) includes a step of reading (410) a data set (115, s), wherein the data set (115, s) comprises a plurality of data, each representing an arrangement of components (105) of an electric machine (100) and at least one parameter (127) of the electric machine (100) associated with each of these arrangements. The arrangement of the components (105) in the electric machine (100) is represented by at least one vector graphic (200) and/or a multi-pixel polygon (210) in a region representing the electric machine (100). Furthermore, the method (400) includes a step of training (420) the artificial intelligence algorithm (125) using the data set (115, s).