Free-Form Grid Structure Multi-Objective Optimization Associated with Integrated Design and Machine Learning

Abstract

Abstract Grid structures are among the lightest spatial structures for supporting large-span shells while constructing an efficient free-form casefree-form is exceedingly challenging. This research aims to investigate a special type of pattern for free-form grid structures using a process that combines selected design variables with their corresponding structural properties. Therefore, functional grid structures based on traditional Iranian geometries are parametrically interconnected in a workflow, where the results are compared to a base case. Iranian Chahar-Lengeh Girih as the main generative designer is generated by the Hankin Method that defines a controlling parameter (P) for the shape typology where the elements of the structure support a simple quadrilateral grid or a reciprocal one. Comparing these two validates the structural efficiency of architectural and structural performance. The workflow carries out a multi-objective evolutionary algorithm (NSGAII) to optimize the structural performance of the free-form grid structures with satisfying the architectural design requirements concomitantly. It is composed of the design-related part which includes the parameters of generating free-form grid structures, while the element diameter, element thickness, and the number of columns represents the structure-related counterparts. Consequently, reducing Mass, Displacement, and Elastic Energy are optimization objectives, besides K-Means Clustering with diverse approaches is used to explore patterns of possible solutions among the functions. The results portray that in the Overall-Best solution extracted by the workflow, with a P parameter equal to 0.8, Displacement, and Elastic Energy are improved by %88.44 and %89.82, respectively compared to its Reference Case where P = 0.5 (simple grid shell).