Predicting of moment-rotation behavior of bolted connections using neural networks

This study focused on moment-rotation behavior of bolted top and seat angle with double web angle connections, using Artificial Neural Network. Several 3D parametric finite element models are developed where the geometrical and mechanical properties of connections are as parameter. In the models, all connection components such as beam, column, angles and bolts are modeled using eight node brick elements. The effect of all component interactions, such as slippage of bolts and frictional forces, modeled using surface contact algorithm and to evaluate the connection behavior more precisely, bolts pretension applied on bolts shanks as first load case. Results of numerical modeling are compared with test results of experimental works that has been done by researchers, and showed good agreement with test results. More models created by parametric model and obtained moment-rotation curves used to train back propagation neural network. Testing the network reveals high performance generality and effectiveness of the trained network for predicting of moment-rotation behavior of bolted connections.

کلیدواژه ها:

Bolted angle connection ، Finite element modeling ، Neural Network ، Back propagation