A long span bridge and a greenhouse roof truss structure optimized by means of a consistent genetic algorithm with a natural crossover

摘要

Purpose - The purpose of this paper is to introduce a novel methodology that has the capability of finding symmetrical and nonsymmetrical solutions in complex design domains without additional tuning when changing the design domain. These go from an academic design domain to a practical one. Design/methodology/approach - Various crossovers operators are applied over the same representation using a genetic algorithm for truss structural optimization cases where literature solutions have a tendency to forced symmetry in order to find an optimal design with fewer iterations. Continuous-discrete representations were cross-bred by a uniform-sbx simultaneous crossover, called zygote crossover. Specialized mutations operations are proposed to generate localized changes to improve the solution according with the design domain. Findings - Design solutions found were lighter and stiffer when comparing against cases reported in current literature and in engineering practice. Also these solutions were found in fewer iterations. Practical implications - The cases solved herein are complex and are a challenge for any optimization routine however practical design limitations are observed in the sense of out plane stability. Further comparisons cases are required in order to generate a less adjusted design, this is because the greenhouse solution had to be stiffened with out of plane bars to give it enough lateral stability. Originality/value - Continuous-discrete representations were cross-bred by a uniform-sbx simultaneous crossover, called natural crossover. Specialized mutations operations are proposed to generate localized changes to improve the solution according with the design domain. This scheme along with a less restrictive environment allows a wider exploration of search space.