BIM-based intelligent optimization of complex steel joints using SVM and NSGA-II

Abstract

Steel joints are vital load-bearing components in structures. In some structures like large-scale bridges, these joints become increasingly complex. However, the current optimization of complex steel joints primarily relies on empirical knowledge and manual trial-and-error. This paper proposes an approach for the optimization of complex steel joints using SVM (support vector machines) and NSGA-II (non-dominated sorting genetic algorithms II) in BIM environment. Initially, this research utilizes Rhino, Grasshopper, and Abaqus to create a BIM framework for intelligent optimization of complex steel joints. Partial components of joints are parameterized, followed by the implementation of finite element (FE) parametric modeling. Subsequently, the outcomes from FE analysis of the parameterized model are employed to train a surrogate FE model of complex steel joints using SVM. Finally, the optimized design is achieved using NSGA-II, based on the surrogate model. Through the comparison experiment of a practical engineering case, it is proved that the proposed approach can effectively assist designers in the complex steel joint design.