Overview
Running topology optimization calculations increases the number of FEA executions that are required to obtain optimum solutions. This therefore increases the calculation time. In JMAG, topology optimization calculation times can be reduced by using CNN (Convolutional Neural Network) surrogate models.
This document describes as the workflow for topology optimization calculations using CNN surrogate models running topology optimization for creating a CNN surrogate model, creating the CNN surrogate model, verifying the accuracy of the CNN surrogate model, and using the CNN surrogate model to run topology optimization calculation.
Keywords
CNN, Surrogate model, Preinstalled script, Python, Topology optimization, NGnet
Notes
CNN surrogate models cannot be used with optimization calculations that correspond to any of the following.
- Combinations of topology optimization + parametric optimization
- Topology optimization (sensitivity analysis) calculations
- When measurement variables are included in constraint conditions and objective functions
