Shape and sizing optimisation of space truss structures using a new cooperative coevolutionary-based algorithm

Abstract

Optimising the shape and size of large-scale truss frames is challenging because there is a nonlinear interaction between cross-sectional and nodal coordinate forces of structures. Meanwhile, combining the shape and bar size variables creates a multi-modal search space with dynamic constraints, making an expensive optimisation engineering problem. Besides, most of the real truss problems are large-scale, and optimisation algorithms are faced with the issue of scalability by increasing the size of the problem. This paper proposed a novel Cooperative Coevolutionary marine predators algorithm combined with a greedy search (CCMPA-GS) for truss optimisation on shape and sizing. The proposed algorithm used the divide-and-conquer technique to optimise the shape and size separately. Therefore, in each iteration, the CCMPA-GS focuses on shape optimisation initially and then switches to the size of bars and tries to find the best cooperative combination of the solutions in the current population using a context vector (CV). A greedy search is embedded in the following to fix the remaining violations from the structure's stress and displacement. This novel alternative optimisation strategy (CCMPA-GS) compared with 13 established genetic, evolutionary, swarm, and memetic meta-heuristic optimisation algorithms. The comparison is based on optimising two large-scale truss structures consisting of 260-bar and 314-bar configurations. Experimental results demonstrate that the proposed CCMPA-GS method consistently outperforms the other meta-heuristic methods, delivering optimal designs for the 314-bar and 260-bar truss structures that are superior by 52 % and 63.4 %, respectively. This signifies a substantial enhancement in optimisation performance, highlighting the potential of CCMPA-GS as a powerful alternative in the field of structural optimisation.