Predicting the Behavior of welded semi-rigid Connections exposed to fireUsing Artificial Neural Network

In this paper, an artificial neural networking (ANN) model is described to predict the moment-rotation response of semi-rigid beam-to-column joints at elevated temperature. Two types of beam-to-column welded joints, i.e. top and seat welded bolted angle connection with and without web angle are modeled. The applied moment and joint’s temperatures are used as input parameters to model the rotational capacity of the joint with increasing temperatures. Data from 22 experimental fire tests and verified finite element model are used for training and testing and validating the neural network models.The model’s predicted values are compared with actual test results. The results indicate that the models can predict the moment–rotation–temperature behavior of semi-rigid beam to-column joints with very high accuracy. The developed model can be modified easily to investigate other parameters that influence theperformance of joints in fire.

Keywords:

welded angle connections , artificial neural network , Purelin function , logsig function , activation function , hidden layer , fire , moment-rotation , temperature-rotation , elevated temperature