Application of a gradient-based algorithm to structural optimization

Abstract

Optimization methods have shown to be efficient at improving structural design, but their use is limited in the engineering practice by the difficulty of adapting state-of-the-art algorithms to particular engineering problems. This study proposes the use of a robust gradient-based algorithm, whose adaptation to a variety of design problems is more straightforward. The algorithm was first applied to truss geometry and beam shape optimization, both forming part of the increasingly popular class of structural form-finding problems. The results showed that the gradient-based method is an appropriate tool for defining shapes in structures. The robustness of the algorithm was verified, as a series of structural configurations were treated with similar efficiency. The gradient-based method was also applied to a more traditional structural design problem through the optimization of a steel girder, resulting in a hybrid scheme featuring a truss stiffener. Throughout the study, emphasis was laid on the practical computer implementation of the gradient-based algorithm in interaction with structural analysis tools.