The Study On Prediction Of Overtopping Under Oblique And Multi-directional Waves Based On Artificial Neural Network

China has a long coastline,and many coasts are vulnerable to erosion and damage by waves and tides.The establishment of breakwaters and revetment structures can withstand the damage by waves,and protect the coastal buildings behind the breakwaters.It is of great importance to determine a suitable crest elevation of embankment engineering.Overtopping is a main index to determine the crest elevation of breakwaters.Accurate calculation of crest elevation is of great significance to ensure the safety,economy and ornamental value of embankment engineering.However,the marine environment is very complicated,and the overtopping of offshore embankments is affected by various complex factors,which is a nonlinear process,and the factors influence each other,so the relationship between the factors is difficult to confirm.In recent years,the development of information technology,especially the artificial intelligence,has provided new ideas for the prediction and calculation of overtopping.At present,most of the studies on overtopping prediction are aimed at unidirectional waves or normal incidence waves,while the actual waves are multi-directional irregular waves,which accompanied by oblique action sometimes,so there are still great deficiencies in the current researches in this field.Therefore,this paper adopts neural network to establish a method for predicting the overtopping under oblique and multi-directional irregular waves in different forms of structures.This study summarizes the previous research results,collects plenty of overtopping data under oblique and multi-directional irregular waves,and establishes a neural network dataset of overtopping under oblique and multi-directional irregular waves.The BP neural network model is used to establish the prediction models of mean overtopping discharge under oblique and multi-directional irregular waves,for vertical breakwaters and sloping breakwaters respectively.In the model,Adam algorithm is used to search for the optimal parameter of network weight value,Tanh function is used as the activation function of the neuron,and Dropout method was used to solve the over-fitting problem of BP neural network.The neural network model is optimized and improved by analyzing and adjusting relevant parameters.The model is trained by the established overtopping dataset,and the prediction accuracy and generalization ability of BP neural network model are compared and verified.The final results show that the neural network established in this study can well predict the mean overtopping discharge of multi-directional irregular waves or oblique waves on vertical breakwaters and sloping breakwaters,and the prediction accuracy and generalization ability of the model are higher than that of empirical formula and the Eur Otop neural network calculation tools.The overtopping model has certain convenience and accuracy,and is suitable for the prediction of overtopping induced by interaction of multi-direction waves or oblique waves,which provides certain basis and assistance for coastal engineering constructions.