A Semi-Supervised Meta-Heuristic Shell Form-Finding Approach

Abstract

This paper proposes a general semi-supervised form-finding approach for optimization of shells, where an initial geometry referred to as the “Directive Form (DF)” influences and directs the shape optimization process. The degree of the influence of the DF over the optimization process is controlled by the optimizer algorithm. A normal option involves stronger influence of DF within the early stages of the optimization process and a vanishing effect towards the later stages of the process. The method is illustrated through an example using the artificial intelligence for the optimization process. The selected problem is optimized with particle swarm optimization algorithm in MATLAB. The required structural analysis of the optimization steps is carried out by the finite element software SAP2000 which is linked up with the optimization algorithm via the application programming interface software. The coordinates of the nodes over the surface of the shell are selected as the optimization variables, and the semi-supervised algorithm finds the desired optimal form by minimizing the eccentricity of the forces as the target function. Finally, the paper presents a simple and workable stress calculation method for the optimized shells under gravitational loads.