Distributed Parallel Solver System for Genetic Optimization of structures

Comment

Comment by Jon Mirtschin on May 16, 2012 at 2:34am

Congratulations, this is an impressive development.

Does the frame topology remain unaltered and the profile assignment get changed?

How did you enable the multi-processing, I hadn't realized that Galapagos can spawn/thread each solution step?

Comment by Sangsu Lee on May 16, 2012 at 2:16am

Genetic Algorithm is a very powerful tool for optimization of various problems.
However, until now, applying genetic algorithm onto optimization of structure was very difficult because the execution of genetic algorithm requires a lot of iterative calculations, but the computing power needed for analyzing structural models is very huge, and it also takes much time. And modifying the structual model according to genetic algorithm was not so easy for real-scale complex structures.

So, we've developed StrAuto, a parametric structural modeler and optimizer based on Rhino/Grasshopper. StrAuto can execute iterative structural analysis and optimization automatically, and the most powerful feature of it is that it can execute the analysis task (the most time consuming part of optimization) in multi-processed way on distributed systems in network.

Now, in the movie you can see the galapagos G.A. optimizer in the left monitor (the main system) is running and many instances(15) of structural solver Sap2000 are running along with it. The middle monitor shows 5 worker systems that are running 3 instances of Sap2000 for each under the control of the main system.

The number of sap instances can be extended as much as your SAP license permits.
(one license can launch 3 instances of SAP)
With StrAuto, structural G.A. optimization can be performed 10 ~ 30 times faster so that even very complex real scale structures can be optimized.

If you are interested in StrAuto, you can see more at
http://tree8.chang-soft.co.kr/

Sangsu Lee, changsoft I&I

• ‹ Previous
• 1
• 2
• Next ›
• Page