Estimates of Internal Forces in Torsionally Braced Steel I-Girder Bridges Using Deep Neural Networks

The bracing components in steel I-girder bridge systems are essential structural for the bridges to restrain their rotation due lateral torsional buckling (LTB). Current design specifications require be installed prevent sections from unexpectedly twisting instability. To estimate internal forces acting on elements, we can use approximate equations that provide considerably conservative values. Otherwise, it is necessary conduct a thorough finite element analysis considering initial imperfections obtain accurate systems. This study aims estimation models based deep neural network (DNN) algorithms more accurately compared with current methodology when LTB occurs. conducted by constructing response data geometrically nonlinear forces, namely moments (Mbr) and (Fbr). propose prediction models, 16 input three output variables were selected training data. Furthermore, parametric hyperparameters used DNN was analyzed number of hidden layers, neurons, epochs. Based statistical performance indices (i.e., RMSE, MSE, MAE, R2), estimated values using evaluated determine best models. Finally, (Mbr, Fbr) results depending (numbers epochs), proposed.