| In this study, we used a genetic algorithm (GA) procedure in conjunction with linear and geometrically nonlinear finite element analysis to obtain optimized designs. The developed genetic optimization procedure can be used in the linear and geometrically nonlinear optimization design of trusses and two- and three-dimensional (2-D and 3-D) framed steel structures in accordance with the American Institute of Steel Construction Load and Resistance Factor Design (AISC-LRFD) Specification. The procedure performs a discrete optimization where the design variables are the standard W shapes listed in the AISC-LRFD Manual. The selection of GA operators, such as crossover and selection schemes, and the appropriate setting of GA control parameters are discussed and investigated. Through several design examples, differences among optimized designs obtained using linear and geometrically nonlinear analyses are compared. From these examples and the capacity analyses of the optimized designs, we conclude that the optimized designs are not affected significantly by the P-{dollar}Delta{dollar} effect. However, in some cases a better design can be obtained by performing nonlinear analysis instead of linear analysis. The comparison between the designs obtained by the proposed procedure and those obtained by conventional optimization design techniques demonstrates that the proposed procedure can be more powerful than the conventional techniques for the optimization design of complex structures with discrete design variables. In addition, some experiments are performed that combine different heuristic techniques (genetic algorithms, tabu search, and simulated annealing) forming promising hybrid algorithms to speed up the design process and find better solutions. |
