1 / 37

Using Evolutionary Computation to explore geometry and topology without ground structures

Using Evolutionary Computation to explore geometry and topology without ground structures. IASS-IACM 2008 Ithaca, NY USA. Dr. –Ing. Peter von Buelow University of Michigan TCAUP 30 May 2008. What are ground structures?. Definition

keona
Download Presentation

Using Evolutionary Computation to explore geometry and topology without ground structures

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Using Evolutionary Computation to explore geometry and topology without ground structures IASS-IACM 2008 Ithaca, NY USA Dr. –Ing. Peter von Buelow University of Michigan TCAUP 30 May 2008 Peter von Buelow University of Michigan 1 / 37

  2. What are ground structures? Definition A set of all admissible bars connecting a set of admissible joints. (Dorn, et al. 1964) Peter von Buelow University of Michigan 2 / 37

  3. Examples of ground structures 12 nodes and 39 bars (Deb et al., 1999) 210 nodes and 21945 (max) bars (Klarbring, 1995) Peter von Buelow University of Michigan 3 / 37

  4. Effect of ground structure on results with EC • Ground structures have 2 effects on EC: • Results • Resolution • Geometry • Topology • Computation • Chromosome • Population • Analysis Peter von Buelow University of Michigan 4 / 37

  5. Effect of ground structure on results With ground structure (505 kg) without ground structure (322 kg) Other topologies found without using ground structure Peter von Buelow University of Michigan 5 / 37

  6. Effect of ground structure on GA computationGA computation depends on chromosome string length Length = 39 (actually used) Size of ground structure ½ of the incidence matrix Length = n(n-1)/2 = 45 Length = 21945 (all possible) Length = 153 Peter von Buelow University of Michigan 6 / 37

  7. Distinction of geometry and topology definitions Peter von Buelow University of Michigan 7 / 37

  8. Encoding Topology without ground structure ‘binary string’ from incidence matrix Peter von Buelow University of Michigan 8 / 37

  9. Breeding Topology one-point crossover Peter von Buelow University of Michigan 9 / 37

  10. Encoding Geometry without ground structure Array of real number nodal coordinates Peter von Buelow University of Michigan 10 / 37

  11. Breeding Geometry Half Uniform Crossover breeding Peter von Buelow University of Michigan 11 / 37

  12. Procedure M - the master sends and gathers all topologies S - each slave finds geometry for each topology Peter von Buelow University of Michigan 12 / 37

  13. Geometry Generations Random start with progenitor Peter von Buelow University of Michigan 13 / 37

  14. Geometry Generations Selection from generation 1 Peter von Buelow University of Michigan 14 / 37

  15. Geometry Generations Setup parents for generation 2 Peter von Buelow University of Michigan 15 / 37

  16. Geometry Generations Breed parents to fill generation 2 Peter von Buelow University of Michigan 16 / 37

  17. Geometry Generations Select best for parents in next generation Peter von Buelow University of Michigan 17 / 37

  18. Geometry Generations Setup parents for next generation Peter von Buelow University of Michigan 18 / 37

  19. Geometry Generations Breed parents for next generation Peter von Buelow University of Michigan 19 / 37

  20. Geometry Generations Cycle 1 Generations 4-6 Peter von Buelow University of Michigan 20 / 37

  21. Geometry Generations Cycle 1 Generations 26-28 Peter von Buelow University of Michigan 21 / 37

  22. Geometry GA Best from one cycle Peter von Buelow University of Michigan 22 / 37

  23. Topology Generations Mutate progenitor to start topology cycle Peter von Buelow University of Michigan 23 / 37

  24. Topology Generations Select best for next generation Peter von Buelow University of Michigan 24 / 37

  25. Topology Generations Set selection as parents for next generation Peter von Buelow University of Michigan 25 / 37

  26. Topology Generations Breed parents to obtain full generation Peter von Buelow University of Michigan 26 / 37

  27. Topology Generations Select best from full generation Peter von Buelow University of Michigan 27 / 37

  28. Topology Generations Set selection as parents for next generation Peter von Buelow University of Michigan 28 / 37

  29. Topology Generations Breed parents to obtain full generation Peter von Buelow University of Michigan 29 / 37

  30. Topology Best individuals from cycles 1- 4 All individualsfrom cycle 1 Peter von Buelow University of Michigan 30 / 37

  31. Topology selection list of ‘pretty good’ solutions Peter von Buelow University of Michigan 31 / 37

  32. Bridge Design automatic generation list of ‘pretty good’ solutions Peter von Buelow University of Michigan 32 / 37

  33. Interactive Group Design Idea generator Peter von Buelow University of Michigan 33 / 37

  34. User Selections Peter von Buelow University of Michigan 34 / 37

  35. Final Bridge Design Peter von Buelow University of Michigan 35 / 37

  36. 3D Trusses Peter von Buelow University of Michigan 36 / 37

  37. 3D Trusses Peter von Buelow University of Michigan 37 / 37

More Related