Design optimization of a large industrial steel structure using an enhanced K-means clustering optimizer and finite-element model updating

Abstract

For the first time, a new method has been introduced to address the optimal design problem of large and complex steel structures with a focus on minimizing weight. The structures considered in this study represent typical steel factory structures with nonprismatic sections of columns and rafters. The process of simulating this structure, from its geometric representation to a finite-element (FE) model, poses significant challenges using conventional methods. To overcome these challenges, a program was developed using the Open Application Programming Interface (OAPI) in the SAP2000 software and MATLAB to establish a new FE model updating technique. The weight optimization process is performed using a newly devised optimization algorithm named KODE. This algorithm combines the advantages of two existing algorithms, namely the K-means clustering optimizer (KO) and the Differential Evolution algorithm (DE). The primary innovation of KODE lies in its ability to generate additional movement directions, ensuring a better balance between the ability of exploitation and exploration compared to the original KO algorithm. To demonstrate the effectiveness of KODE compared to the other algorithms, 23 classical benchmark functions and CEC2005 benchmark functions are employed as initial numerical examples. Subsequently, KODE is applied to optimize objective functions, which is established based on the AISC360-05 design standard (American Institute of Steel Construction 360-05), for optimal weight in a steel factory structure. The results in this study show the efficiency of KODE in solving optimization problems. In particular, KODE has demonstrated high effectiveness and reliability when combined with FE model updating to design optimization for large-scale steel structures.