Numerical analysis and prediction of lateral-torsional buckling resistance of cellular steel beams using FEM and least square support vector machine optimized by metaheuristic algorithms
Mohamed El Amine Ben Seghier, Hermes Carvalho, Caroline Corrêa de Faria, José A.F.O. Correia, Ricardo Hallal Fakury
Abstract
This study presents an advanced framework for modeling the lateral-torsional buckling behavior of cellular steel beams, which combines hybrid intelligent models with numerical simulation. The proposed hybrid intelligent models employ a large dataset-based finite element method (FEM) for training and validation the framework, as well as metaheuristic algorithms for optimal auto-hyper-parameters selection. A total of 1535 numerical models are examined in order to evaluate the lateral-torsional buckling behavior. Following that, the least square support vector machine (LSSVM) optimized using four metaheuristic algorithms (ME): particle swarm optimization (PSO), ant lion optimization (ALO), grey wolf optimizer (GWO), and Harris hawks optimization (HHO) algorithms, is utilized to estimate accurately the lateral-torsional buckling resistance. According to the findings of a comprehensive performance evaluation utilizing statistical and graphical comparing criteria, the suggested LSSVM-ME predicts the lateral-torsional buckling behavior with excellent accuracy. LSSVM-HHO, in particular, outperforms the other hybrid intelligence models, with an RMSE of 41.72 kN.m and an NSE of 0.99. Overall, the results indicate that the proposed framework has a great potential for use as a practical tool for estimating the lateral-torsional buckling behavior of cellular steel beams.