DAMAGE PREDICTION OF THE STEEL ARCH BRIDGE MODEL BASED ON ARTIFICIAL NEURAL NETWORK METHOD

Abstract

Failure in the advance prediction of bridge structure collapse requires an enormous cost of rehabilitation. In most cases, the projection of decreases or damage to the structure due to difficulty in the testing condition. Therefore, this study analyses the damage and identification of the critical structural components' severity on the steel girder arch bridge. Using the Artificial Neural Networks (ANNs), this research has tested a parametric steel girder arch bridge. The numerical model of the 146 supported girder has analysed by epoch 500 values of ANNSs's parameter. The stiffness of 10th element is assumed to drop 10%, 20%, 30%, and 40% of whole the tested. The architecture model of ANNs was three neurons in the input layer, five neurons in the hidden layer and one neuron in the output layer. The simulation of the data set were 90:10, 80:20, 70:30, and 50:50. ANNs shows the damage' severity in this the stiffness reduction tested by applying the damage index methods. In this research, the ANNs' simulation has been reliable to predict 98% for identifying structural damage. Thus, the results confirm the feasibility of the technique and its application in predicting structural failure.