GC-GA | BT-Lab

Overview

This project is a cooperative effort between the University of Michigan and TU Delft. It was presented at the 2nd Sino – Italian Workshop on Bridges and Special Structures, IUAV, University of Venice – 12 December 2008.

This site presents an initial trial use of Bentley's Generative Components with a Genetic Algorithm for the exploration of structural form. STAAD-Pro FEA software is used for analysis. All members are designed using HSS steel tubing sections for an ASHHTO 2 lane bridge load. The GA then uses least weight as the fitness function. The selection of breeding pairs of trusses was made interactively by architecture students at the University of Michigan – TCAUP. The designer's selection is made based on visual (aesthetic) criteria. This potentially allows the GA to explore solutions more in tune with the designer's intentions.

Each parent population contains 20 trusses. Once 20 children have been produced the next generation of parents is formed by selecting the lightest 20 trusses from the combined population of 40 parents + children. During 2 one hour sessions, 5 generations were produced. These are shown under the "Generations" tab. The students selected parents under the "Population" tab and breed one child. The child is downloaded to the local computer in the form of a .csv file (you can try this yourself). This file must be read into Excel and converted to .xls format so that it can then be read into GC.

Using the foster5 GC script, the input.xls file is read into GC and the child truss is generated. Using .dxf format to import the geometry to STAAD-Pro, the truss is analyzed and members sized. STAAD will calculate the total weight in the “steel take-off” table. The designer then renames the original child .csv file with the weight as well as the .jpg image of the truss produced by GC and uploads the files to the working directory. The files are automatically read and placed in the current population.
The necessary files for the process can all be obtained under the “Download” tab. If you would like to try it you can email me your child .csv and .jpg and I will add them to the population.

pvbuelow@umich.edu

This project is a cooperative effort between the University of Michigan and TU Delft. It was presented at the 2nd Sino – Italian Workshop on Bridges and Special Structures, IUAV, University of Venice – 12 December 2008. This site presents an initial trial use of Bentley's Generative Components with a Genetic Algorithm for the exploration of structural form. STAAD-Pro FEA software is used for analysis. All members are designed using HSS steel tubing sections for an ASHHTO 2 lane bridge load. The GA then uses least weight as the fitness function. The selection of breeding pairs of trusses was made interactively by architecture students at the University of Michigan – TCAUP. The designer's selection is made based on visual (aesthetic) criteria. This potentially allows the GA to explore solutions more in tune with the designer's intentions.
	Each parent population contains 20 trusses. Once 20 children have been produced the next generation of parents is formed by selecting the lightest 20 trusses from the combined population of 40 parents + children. During 2 one hour sessions, 5 generations were produced. These are shown under the "Generations" tab. The students selected parents under the "Population" tab and breed one child. The child is downloaded to the local computer in the form of a .csv file (you can try this yourself). This file must be read into Excel and converted to .xls format so that it can then be read into GC.
Using the foster5 GC script, the input.xls file is read into GC and the child truss is generated. Using .dxf format to import the geometry to STAAD-Pro, the truss is analyzed and members sized. STAAD will calculate the total weight in the “steel take-off” table. The designer then renames the original child .csv file with the weight as well as the .jpg image of the truss produced by GC and uploads the files to the working directory. The files are automatically read and placed in the current population. The necessary files for the process can all be obtained under the “Download” tab. If you would like to try it you can email me your child .csv and .jpg and I will add them to the population.
	pvbuelow@umich.edu