Heuristically Seeded Genetic Algorithms Applied to Truss Optimisation

Abstract.

A heuristic method of seeding the initial population of a Genetic Algorithm (GA) is described, which enables better solutions to discrete truss optimisation problems to be found within a shorter time period, and with a negligible increase in computational effort (compared with the simple GA). The seeding method is entirely automatic, and makes use of the problem-specific routines used to calculate fitness, already present within the GA. The GA models natural, biological evolution as a means of producing a ‘good’ solution to a problem. The GA described here is implemented in various versions. The differences between each version are in the selection procedure and/or the generation of the initial population. To compare the effectiveness of each strategy the GA variants are applied to four example problems.