Application of deep learning method in web crippling strength prediction of cold-formed stainless steel channel sections under end-two-flange loading

This paper proposes a deep-learning framework, specifically, a Deep Belief Network (DBN), for studying the web crippling performance of cold-formed stainless steel channel sections (lipped and unlipped as well as fastened and unfastened) with centered and offset web holes under the end-two-flange loading condition. G430 ferritic, S32205 duplex and 304 austenitic stainless steel grades are considered. A total of 17,281 data points for training the DBN are generated from an elasto plastic finite element model, validated from 69 experimental results reported in the literature. When a comparison was made against a further 53 experimental results reported in the literature, the DBN predictions were found to be conservative by around 10%. When compared with Backpropagation Neural Network (a typical shallow artificial neural network) and linear regression model based on PaddlePaddle, it was found that the proposed DBN outperformed these two methods, using the same big training data generated in this study. Using the DBN predictions, a parametric study is then conducted to investigate the effect of web holes, from which unified strength reduction factor equations are proposed. Finally, a reliability analysis is conducted, which shown that the proposed equations can predict the web crippling strength of coldformed stainless steel channel sections under the end-two-flange loading condition.