Application of Artificial Neural Network to Modeling of Pipeline Damage Prediction

Because the damage of pipeline is controlled by many factors, such as fault movement, pipe-soil interaction, buried depth, etc., the relationship between pipeline damage and influencing factors is complicated. In order to predict the pipeline damage, predictive model is constructed on the basis of artificial neural network (ANN), in which the damage of pipeline becomes a nonlinear function of influence factors. According to eight groups sample data, MATLAB is applied to analyze the design of predictive model; influences of model structure, concealed layer number, neuron number of concealed layer, and training function, on the predictive results are analyzed. Model parameters and preferences are optimized, and predictive model of pipeline damage is determined based on results of numerical simulation. Finally, optimum model structure is worked out and some advice for modeling and protection of pipeline is proposed.