Experimental and ANN Analysis of Cold-Formed Steel Build-Up Columns with and without Intermediate Web Stiffeners under Axial Compression

Abstract

To evaluate the performance of cold-formed steel (CFS) build-up columns with and without intermediate web stiffeners, an experimental study and an artificial neural network (ANN) analysis were performed on six column specimens: three with web stiffeners and three without. The performance was investigated based on their failure mechanisms, maximum strengths, stiffness parameters, and load-displacement trends in the experiments. The axial shortening and buckling behavior of cold-formed steel (CFS) build-up columns are load-dependent, whether the columns are battened and laced with or without stiffeners or are single C-sections with or without stiffeners. Based on the experimental observations, CFS build-up columns with stiffeners exhibit greater stiffness than those without stiffeners under axial compressive loading. The outcomes of the experimental investigation are discussed in detail in this article. ANN models were employed to predict the buckling failure of all six specimens. The performance of the ANN model was analyzed using statistical criteria such as R, RMSE, and MAE, with the optimal ANN architecture containing hidden layers. The results indicate that the optimal ANN model is a highly effective predictor of buckling, with R values of 0.9996, 0.99853, 0.9953, 0.99968, 0.99946, and 0.99938 in the testing phase. It is concluded that the optimal ANN model is an extremely effective machine-learning algorithm for failure prediction, providing significant results.