The repository contains the experimental studies for "Generative Design of Physical Objects using Modular framework" paper.

The experimental setup consists of:

• Synthetic cases:
  • design of structures with various number of polygons in optimal solution - script
  • design of structures with various number of vertices in optimal solution - script
  • design of structures with various number of search domain size - script
  • design of structures with various complexity - script
• Breakwaters design case study:
  • breakwaters design based on differential evolution - script
  • breakwaters design based on SPEA2 - script
  • breakwaters design based on combination of surrogate and physics-based models - script
• Microfluidic device case study:
  • microfluidic design based on Generative Neural Network - script
  • microfluidic design based on GEFEST Standard Sampler - script
• Heat-source system design:
  • heat-source design based on Generative Neural Network and Deep learning estimator - script
• Oil field design case study.

GEFEST framework is available in main repository.

• Breakwaters

  Below are the convergence results in terms of hypervolume and the best configurations found by each toolkit are presented.
  

  

• Microfluidics

  Convergence of the target variable and the best found objects by each toolkit are presented.
  
  

  

• Heat source

  The best average temperature found in the design process is presented.
  
  

              Generative sampler and Deep learning estimator