[JFT131] Evaluating Degrees of Similarity for Geometry Using NGnet Weight

A License ID or User ID is required to download data.

Guide / Sample Data


Attempts are made at topology optimization that can make substantial changes to the geometry itself for optimized design within the electrical appliance design space. Because multiple design proposals are obtained as a result, it is conceivable to group these corresponding to the degrees of similarity between topology, wherein evaluations, etc., may then proceed.
One method of optimization is the on / off method using NGnet, where geometry is defined by the weight of the Gaussian functions positioned within design regions. With JMAG, the weight of Gaussian functions can be output after optimization calculations have completed.
This document describes a method where after executing topology optimization using NGnet, degrees of similarity for geometry from NGnet weight are evaluated.


Topology Optimization, Similarity, Clustering, on / off method, NGnet, Gaussian function, Weight

Search Filter

  • All Categories