Medalist learning algorithm for configuration optimization of trusses

Abstract

Truss configuration optimization, involving discrete sizing variables and continuous layout variables, is crucial for various practical applications. This paper introduces a highly efficient metaheuristic algorithm known as the Medalist Learning Algorithm (MLA) to address this optimization problem. Inspired by the learning behavior observed in group dynamics, the MLA offers a concise implementation procedure with two key operations: identifying medalists and facilitating individual learning. By leveraging the learning efficiency derived from a Logistic function, the MLA strikingly balances exploration and exploitation capacities throughout the learning period. To demonstrate its effectiveness, four classical truss structures with up to 44 design variables under multiple loading conditions are utilized to evaluate the MLA's performance in solving sizing and shape optimization problems of varying scales. Comparative analysis between the MLA's results and previously reported findings establishes its superiority in achieving the final feasible best weight. Statistical evidence further illustrates the MLA's consistent ability to deliver competitive solutions for a wide range of truss configuration optimization problems. The MLA's simplicity in implementation ensures its widespread applicability and potential for practical usage in the field of structural optimization.