基于神经网络的Boussinesq模型在珊瑚礁地形波浪破碎判据确定方法OA

The existing FUNWAVE-TVD model effectively simulates wave hydrodynamic changes in a flat terrain.How-ever,when simulating these changes in a coral reef terrain,its accuracy decreases because of an inaccurately set break-ing criterion.To overcome this limitation and accurately predict the optimal breaking criterion,an artificial neural net-work(NN)algorithm is introduced in this study.By exploring various acquired and preprocessed variables,an NN training dataset is constructed,and a corresponding backpropagation NN prediction model is designed.This model is subsequently used to predict the breaking criteria for four sets of regular waves acting on coral reef terrain.Applying these predictions to the FUNWAVE-TVD model,a comparative analysis of the simulated wave heights is performed against those generated using the default breaking criterion.The results reveal substantial improvement:the average goodness of fit(GoF)with measured wave heights for the four conditions is 0.8698,which is approximately 30%higher than the average GoF achieved using the default breaking criterion.Further analysis confirms the accuracy of this NN model:comparing the predicted breaking criteria for the four conditions against the empirically most suitable criteria re-veals the average relative error of only 0.0627 for the two datasets.This low error value definitely indicates that this NN model can effectively and reliably predict the optimal breaking criterion for the FUNWAVE-TVD model.