Minimal weight geometry design of steel structures by observer-teacher-learner based optimization

Structural response of a simulated skeletal model under static loads is majorly governed by stiffness properties. It depends not only on the member sections but also on the nodal geometries as a more critical property in spatial arrangement of steel material. The present work formulates weight minimization of truss and frame structures within an optimization problem so that geometry and sizing are simultaneously altered as design variables. Furthermore, an enhanced variant of teacher-learner-based optimization as a parameter-less meta-heuristic is employed to solve the problem. Performance of the proposed algorithm is evaluated in bridge and diagrid examples with considerable reduction in the consumption of steel material under the design code constraints. The first example is of continuous type while the second is a mixed integer-continuous problem. The results shows superiority of observer-teacher-learner based optimization over standard teacher-learner-based optimization and particle swarm optimization regarding the convergence rate and quality. It is also observed that the improved method is more robust than the others due to its less standard deviation about the mean result. As such a method can find proper solution with few tuning and computational effort it is recommendable for engineering applications

Keywords:

Weight minimization , meta-heuristic algorithm , teaching-learning-based optimization , skeletal structure