ON THE PERFORMANCE OF A MODIFIED MULTIPLE-DEME GENETIC ALGORITHM IN LRFD DESIGN OF STEEL FRAMES

This paper investigates the performance of a multiple-deme genetic algorithm (GA)
with modified reproduction operators, in optimal design of planar steel frames according to the
AISC-LRFD specification. The design objective is to minimise the weight of frame subject to
strength, displacement and constructability constraints. A number of new crossover and mutation
operators, used alongside the standard operators are utilised in optimum design of a number of
steel frames subjected to the constraints of the AISC-LRFD specification, with and without
considering the second order effects, as set out by the code requirements. This modified GA
(MGA) is shown to have a very fast convergence and to produce relatively high-quality designs.
This paper also utilizes the concept of multiple-deme in the GA, as it has been used successfully
for other metaheuristic population-based methods. The multiple-deme GA is used alongside the
modified GA operators and the algorithm is named the modified multiple-deme GA (MMDGA).
The modified GA (MGA) and modified multiple-deme GA (MMDGA) are applied to three
benchmark problems and the results are compared to those obtained by other metaheuristic
methods. In the majority of cases, the results of comparisons suggest the superiority of the
MMDGA in terms of the quality of final design and the total number of performed finite elements
analyses.