Damage predictive model of submarine pipeline impacted by falling object based on BP neural network

Abstract

In order to predict the damage of submarine pipeline on saturated clay seabed by falling objects, a dynamic finite element model of interaction between saturated clay seabed and submarine pipeline under the impact of falling objects was established. The influence of six parameters, namely, energy of falling objects, diameter of the pipeline, wall thickness, grade of the steel material, internal pressure, and undrained shear strength of seabed soil on the damage of the submarine pipeline was analyzed by combining with the range of changes of the actual working conditions of the submarine pipeline. Taking the six parameters as input parameters, the damage of pipeline as output parameters, and the numerical simulation results as training samples, a BP neural network model for submarine pipeline damage prediction was constructed through learning and training. The results show that the larger the impact energy of the falling object is, the larger the pipeline damage is, and the pipeline damage growth rate tends to slow down with the increase in the impact energy of the falling object. The increase in pipeline diameter, wall thickness, internal pressure, and the pipeline yield strength leads to a decrease in the pipeline damage. The saturated soil undrained shear strength increases, while pipe damage increases. The established BP neural network prediction model for submarine pipe damage needs only six parameters, namely, falling object impact energy, pipe diameter, wall thickness, steel grade, internal pressure and undrained shear strength of seabed soil. It is simple and convenient, and can better predict the damage of submarine pipeline impacted by falling objects in saturated clay seabed. The numerical examples cover the working conditions of common saturated clay seabed submarine pipelines, so the model has good applicability, which provides new ideas for the prediction of submarine pipeline damage.

Keywords

• ocean engineering; submarine pipe damage; falling object impact; bp neural network; predictive model; saturated clay seabed; undrained shear strength