Research On Prediction Of Local Scour Depth At Bridge Piers Under Ice Cover

The mechanism of local scour of bridge piers under ice cover is very complicated,and it is very important for the safety design of bridge engineering to correctly calculate the local maximum scour pit depth near the bridge piers.When the traditional regression equation is used to calculate the maximum scour pit depth,it is difficult to comprehensively consider the related control variables and boundary conditions.Based on the experimental data of local scour of bridge piers under clear water in laboratory,the relationships among flow velocity,with or without ice cover,pier diameter,bed sediment size,water depth and local scour pit depth of bridge pier are analyzed respectivelyBased on the principle of dimensional analysis,BP neural network was used to analyze the related factors that affect the local scour of bridge piers.3/4 of the experimental data were taken as the training data set of the prediction model and 1/4as the test data set of the prediction model.When predicted the local scour pit depth of bridge pier under the condition of open flow,the input factors of the network were as follows:flow Frude number Fr,ratio of water depth to pier diameter h/D,ratio of median particle size to pier diameter d₅₀/D,ratio of water depth to median particle size of river bed sediment h/d₅₀,output factor:ds/D.When predicted the local scour pit depth of bridge piers under ice sheet conditions,the input factors of the network were as follows:flow Frude number Fr,ratio of water depth to pier diameter h/D,ratio of median particle size to pier diameter d₅₀/D,ratio of water depth to median particle size of bed sediment d₅₀/D,ratio of surface roughness under ice cover to surface roughness of bed sediment n_i/n_b,output factor:ds/D.Correlation coefficient（CC）,root mean square error（RMSE）,mean absolute percentage error（MAPE）and determination coefficient（R²）were used as the evaluation indexes of the calculated results,and the predicted results were compared with the experimental results.When BP neural network is trained with dimensionless data,it can predict the equilibrium scour pit depth better than the original scour data.A Sensitivity analysis showed that the influence of velocity of the flow and Froude number（Fr）on the equilibrium scour depth is greater than that of other independent parameters.When BP model was used to predict the local scour pit depth of bridge piers under the condition of open flow,the correlation coefficient of predicted results is 0.9541,MAPE is 29.67%,R²is 0.9105,RMSE is 0.1335.When predicting the depth of scour pit under the condition of ice cover,the correlation coefficient of predicted results is0.9337,MAPE is 28.58%,R²is 0.8550,RMSE is 0.1911.The results showed that the BP neural network has a high accuracy in predicting the depth of local scour pit of bridge piers under the conditions of open current and ice cover,which can provide a certain reference for practical engineering.