Design optimization of N-shaped roof trusses using reactive taboo search

Abstract

Design optimization of a class of plane trusses called the N-shaped truss (NST) is addressed. The parametric model of NST presented is intended for real-world application, avoiding simplifications of the design details that compromise the applicability. The model, which includes 27 discrete variables concerning topology, configuration and sizing of the truss, presents a challenging optimization problem. Aspects of such challenge include large search space dimensionality, absence of a closed-form objective function (OF) and constraints, multimodal objective function and costly CPU time per objective function evaluation. Three implementations of general-purpose genetic algorithms (GAs) are tested for this problem, along with a version of taboo search called reactive taboo search (RTS). In this study, the raw version of RTS exhibited better performance than the tested versions of GA but lacks some of the GA capabilities to span the search space. A modification of RTS that uses a population-based exploitation of the search history is proposed. The optimization results show that the introduced modification can further improve the performance of RTS.