Prediction Of Pier Scour Depth Based On GA-MIV-BP Neural Network

In recent years,the pier foundation is exposed due to water scouring,which threatens the safety of the bridge and even leads to the damage of the bridge.Therefore,it is of great practical significance to study how to accurately predict the scouring depth of bridge piers and provide theoretical basis for the depth of pier foundations and later protection.This paper firstly summarizes the relevant researches of domestic and foreign scholars on the influence of general scour,local scour and sand mining on bridge scour of bridge piers.The influencing factors are analyzed by combining the mechanism of general and local scour.Taking the Ganjiang South Branch Bridge located in the lower reaches of the Ganjiang River as the engineering background,the scour depth of the bridge is studied.The specific research contents and conclusions are as follows:Using the main domestic scour formulas to calculate the general and local scour parts of the bridge,it is found that the superimposed value of the two is much smaller than the actual scour value,and the influence of riverbed sand mining on the scour depth needs to be considered.If it is larger,it is easy to cause value deviation and cannot meet the forecast requirements well.For the complex nonlinear problem of general scour and local scour of bridge piers,on the basis of previous research,this paper proposes a BP neural network optimized by Genetic Algorithm(GA)for prediction.To screen,the results show that the MIVBP and GA-MIV-BP models have a greater improvement in the prediction accuracy than the traditional BP model,indicating that the genetic algorithm and the screening of scour influencing factors can optimize the BP network.The Bland-Altman consistency test was carried out between the predicted value of the model and the measured value of the test sample,and the results showed that the predicted value of the GA-MIV-BP model was in good agreement with the measured value.Finally,through the verification of another measured sample,it is found that its prediction accuracy is not lower than the existing empirical calculation formula,which proves that the model prediction has good extrapolation,and has certain application in the prediction of general scour and local scour of bridge piers value.The mechanism of traceable scouring is analyzed,and the calculation formula of the traceable scour depth of bridge piers caused by sand mining pits is obtained through its generalized model.However,due to the complexity of traceability scouring,it is difficult to obtain the balance ratio parameter in the formula.Therefore,it is proposed to use a large number of traceability scour samples to establish a BP network model for the balance ratio through training to predict it.The comparative analysis of the network determines that the input factors of the network are the single-width flow,the sediment content of the water flow and the average particle size of the sediment,and combined with the average influence value of the MIV,it is verified that the selected factors meet the modeling requirements of the BP network.Finally,combined with the genetic algorithm for optimization,the established GA-MIV-BP model is used to predict the test samples.The results show that the established network prediction effect is good,and it can be applied to the pier traceability scour depth prediction caused by sand mining pits.Through the network model established above,it is predicted that the general scour and local scour depths of the Nanzhi Bridge of the Ganjiang River under the design frequency flood are 0.645 m and 2.18 m respectively,and the maximum traceable scour depth caused by the sand pit on the bridge is 4.597 m.The predicted maximum scour depth of the bridge is 7.422 m.Compared with the measured value of the bridge under non-flood conditions,the result is reasonable and can provide a theoretical reference for bridge foundation design and later protection and reinforcement.