Prediction And Analysis Of Binary Embankment Piping Danger In The Yangtze River Based On Grey Correlation And GA-DBN

The safety of levees is related to the people’s livelihood,and most of the levees in our country are made by heightening and thickening,and their geological structure is complex,which leads to frequent occurrence of piping dangers.In severe cases,major accidents such as dyke bursts and dam collapses may occur.Therefore,timely and effective response to the occurrence of piping dangers Making predictions has important research significance and practical value.The Yangtze River embankment has an important political and economic status in Wuhan and even the whole country,but it is always endangered by the danger of piping.The factors affecting the piping of the Yangtze River embankment are complex and have obvious nonlinear relationships.It is difficult to accurately predict with conventional methods.Deep Belief Network(DBN)model It is a forecasting method that has developed rapidly in recent years and has great advantages in data processing and forecasting analysis.Therefore,this paper takes the Yangtze River Binary Embankment Project as the main research object,combined with the actual project,adopts grey relational analysis to screen the main factors affecting the piping of the Yangtze River embankment,and uses the GA-DBN model to predict and analyze the piping.The main contents are as follows:(1)Combining the existing research results and the actual situation of the Yangtze River embankment project,summarize the influencing factors of the Yangtze River embankment piping,use the grey correlation analysis method to analyze the sensitivity of the influencing factors,and determine the water level,the thickness of the weakly permeable overburden,the permeability coefficient,and the compression coefficient 5.Five factors of void ratio are important influencing factors.(2)In addition,the formation reason and method of the research object of embankment piping danger are described,the basic theory of deep belief network is summarized,and various parameters are compared and analyzed through empirical formula,etc.,to study the influence of deep belief network parameters on the piping prediction model The impact of accuracy will ultimately determine the network model parameters.On the basis of gray correlation analysis,a dyke piping prediction model based on deep belief network is constructed to analyze and predict the piping danger of dykes,and obtain the prediction results,and combine the prediction results of support vector machines and the DBN model without gray correlation processing Comparing the prediction results,it is finally determined that DBN has higher prediction accuracy,and after the grey relational analysis and processing,DBN has a significant improvement in the prediction accuracy of embankment piping.(3)Aiming at the problem that the parameters of the deep belief network model do not have self-adaptation,the genetic algorithm is used to optimize the parameters of the deep belief network,and a GA-DBN-based embankment piping prediction model is established.By comparing with the accuracy of the unoptimized DBN,Finally,it is determined that the genetic algorithm can effectively improve the accuracy of dike piping prediction.(4)Select the measured data of a certain actual embankment section of the Yangtze River and replace the model’s verification set to verify the model’s prediction accuracy.The results obtained have high prediction accuracy,which proves the effectiveness and practicability of the prediction model.