Intelligent local buckling design of stainless steel I-sections in fire via Artificial Neural Network
Zhe Xing, Kaidong Wu, Andi Su, Yuanqing Wang, Guangen Zhou
Abstract
Traditional local buckling design methods of stainless steel I-sections in fire generally adopt the effective width method. However, in order to precisely consider the influence of fire conditions, traditional methods are tedious and their accuracy is not ideal. To fill this gap, an intelligent local buckling design method of stainless steel I-sections in fire via Artificial Neural Network (ANN) is proposed. A comprehensive data set including test and FE results is built and the correlation among parameters is evaluated. Based on this data set, ANN models are developed and optimized in accordance with the benchmark of Kruppa’s criteria, and then k -Fold cross-validation is conducted to avoid overfitting. Finally, the optimized ANN models are assessed and compared with traditional design methods in terms of accuracy and reliability, which indicates that ANN is suitable for the local buckling design of stainless steel I-sections in fire.