Numerical Simulation And Neural Network Prediction Of Sand Wave Evolution

The migration and growth of submarine sand waves due to the action of tidal current and wave may lead to pipeline suspension and hence the pipeline fatigue damage.It can also cause the tilt of offshore platforms,threatening the production safety of oil and gas industry.Therefore,it is of great engineering significance to study the mechanism of migration and growth for sand waves.Based on Xbeach model,numerical simulations of seabed sand wave evolution under tidal currents have been carried out in this paper.The effects of asymmetry coefficient,sand wavelength,relative water depth and sand particle size on sand wave migration and growth were analyzed.The relative impact degree of model parameters was also compared and discussed.It is found that the sand wave changes slowly with small sand wave slope,large relative water depth,and coarse sand particle,and vice versa.For sand wave migration,relative water depth is found to be with the greatest influence,while the effect of asymmetry coefficient can be ignored.For sand wave growth,effects of all parameters should be taken into consideration.Next,the BP(Back Propagation)neural network model and GAN(Generative Adversarial Network)model were established,and the models were trained based on the results from numerical simulations to predict the migration and growth of sand waves.The correlation coefficient r between BP neural network model prediction results and numerical simulation results is greater than 0.95,with an average relative error of less than 6.5%.The average relative error of GAN model is found to be less than 7.5%.Results indicate that both the BP neural network model and GAN model can predict the migration and growth of sand waves well,moreover,the GAN model improves the generalization ability of the model with little change in the accuracy of the prediction results.At last,MIV(Mean Impact Value)algorithm is used to evaluate the correlation between asymmetric coefficient,sand wavelength,relative water depth,sediment particle size and migration rate,as well as growth rate of sand waves.MIV values are in good agreement with numerical simulation results in terms of the absolute value and the sign,indicating that MIV algorithm can quantify parameter correlation well.