The genetic algorithm seeks to improve design by a process of selection and recombination within a population of designs proceeding generation-by-generation. The problem of 'combinatorial explosion' caused by the combination of many discrete values of the design variables can be resolved by a process of progressive combinatorial space condensation. The application takes place on-line under adaptive controls with initial values specified by the designer.

The temporary suspension of design activity while the structure is reanalysed is time-consuming since each change in the design variables requires a fresh analysis of the structure. A more economical process would obtain if a reanalysis tool capable of automatically accommodating all changes in design data were available during the optimization process.

If the structure has a more-or-less standard form such as a rectangular multi-storey frame or a structural grillage then a neural network is ideally suited to this purpose. In these circumstances it is a simple matter to train a network to map input variables to output (design) quantities. The resulting reanalysis tool is then suitable for integration into a total design process.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £75 +P&P)