Comparative evaluation of optimization algorithms for truss shape design

Purpose The purpose of this study is to develop and benchmark a unified computational framework for shape optimization of truss structures, enabling a fair comparison of commonly used optimization algorithms under identical problem formulations. Design/methodology/approach The proposed framework integrates finite element analysis (FEA) with three optimization algorithms: derivative-free optimization (DFO) using the Nelder–Mead method, particle swarm optimization (PSO) and genetic algorithm (GA). A standardized problem setup is employed to ensure consistent evaluation criteria. The algorithms are implemented using Python-based libraries, and their performance is assessed in terms of convergence speed, optimization quality, consistency, flexibility and computational effort. Several truss configurations, including cross-braced bays and bridge systems, are used as benchmark cases. All three optimization algorithms achieved significant reductions in structural displacement across the examined truss configurations. Findings The results reveal distinct performance trade-offs: DFO provides rapid and consistent solutions with minimal computational overhead; PSO achieves fast convergence with high-quality solutions; and GA demonstrates strong adaptability to complex design spaces, albeit with higher computational cost. Originality/value This study presents a reproducible and standardized benchmarking framework that allows direct comparison of different optimization algorithms for structural shape optimization. By evaluating multiple algorithms within the same computational environment, the work provides practical guidance for algorithm selection and highlights the effectiveness of Python as a platform for applied engineering computations.