Generative AI for Geometric Design Optimization

STOCHOS© GEN-AI

PI Probaligence gmbh

Geometric Design Optimization and gEN-BO

Generative AI with STOCHOS©

In the past, STOCHOS has provided one of the most sample efficient probabilistic methods for the optimization of arbitrary parametric optimization problems with Bayesian Optimization.

With the newly developed and extremely efficient STOCHOS GEN-AI, there are new ways to explore design spaces non-parametrically and to generate thousands of new designs within seconds.

After a training phase the geometric DIM-GP allows any mesh-based results to be predicted and can thus replace expensive FEM / CFD simulations.

By combining GEN-AI, geometric DIM-GP and Bayesian Optimization to a new method called GEN-BO, STOCHOS now offers one of the most efficient methods for non-parametric optimization problems.

Example

Automatic design clusters

STOCHOS comes with a build-in function that automatically clusters the generated designs and visualizes them.

Bayesian Optimization with Generative AI

Typical goals of design optimization:

  • Improve structural efficiency
  • Enhance manufacturability
  • Increase durability
  • Optimize thermal performance
  • Streamline aerodynamics
  • Maximize sustainability
  • Enhance design aesthetics
  • Reduce energy consumption
  • and many more…

 

GEN-BO to further boost your optimization

In combination with the other tools, STOCHOS search for the best new candidate to run a simulation based on expected improvement (including uncertainty). 
This is possible for single-objectives (for example minimize stress) as well as multi-objectives like (minimize stress while minimize weight at the same time). The latter results in a pareto optimization.

 

Contact us to arrange a pilot with your application!

A New Era in Machine Learning

PI Probaligence

Address

Social Networks

Links List

PI Probaligence GmbH

(c) 2023 All Rights Reserved