Optimal truss-structure design using real-coded genetic algorithms

Abstract

Optimization of truss-structures for finding optimal cross-sectional size, topology, and configuration of 2-D and 3-D trusses to achieve minimum weight is carried out using real-coded genetic algorithms (GAs). All the above three optimization techniques have been made possible by using a novel representation scheme. Although the proposed GA uses a fixed-length vector of design variables representing member areas and change in nodal coordinates, a simple member exclusion principle is introduced to obtain differing topologies. Moreover, practical considerations, such as inclusion of important nodes in the optimized structure is taken care of by using a concept of basic and non-basic nodes. Stress, deflection, and kinematic stability considerations are also handled using constraints. In a number of 2-D and 3-D trusses, the proposed technique finds intuitively optimal or near-optimal trusses, which are also found to have smaller weight than those that are reported in the literature.