Optimal discrete truss design using improved sequential and genetic algorithm

Abstract

The potential of two distinct approaches applied to the truss discrete optimization problem is presented in the paper. The sequential discrete optimization method SDO (which is a deterministic procedure, using heuristics based on the idea of fully stressed truss design) and the genetic algorithm GA (a stochastic search method, inspired by the natural evolution model) are compared. The minimum weight design of truss structures subjected to stress and displacement constraints is investigated, including the case of multiple load conditions. The discrete design variables are areas of members, selected from a finite catalogue of available sections. Benchmark 2D and 3D problems are presented in numerical examples. The effectiveness of two approaches is discussed. The improvements of both algorithms and GA integrating the results of SDO method are proposed. They enable us to accelerate the convergence, diminish the number of structural analyses and guide to refined “near optimal” solutions.