Research Of Prediction And Application Of Rock-fill Dam Material Watering Volume Based On Chaotic Time Series And Random Forest Regression

It is a common way to improve the compaction quality by watering the dam material in the construction process of rockfill dam.However,the current research on the watering control of dam material fails to consider the influence of meteorological factors on moisture content of dam material during dam construction process.In order to maintain the targeted moisture content of dam material during the rolling construction process and guaranteed the compaction quality,it is necessary to comprehensively consider the change of moisture content of dam material in the process of watering;Therefore,based on the chaotic time series and the random forest regression method,this paper establish the prediction model of watering volume of the rockfill dam material considering the effects of meteorological factors to control precisely the watering volume.The main achievements of the study are as follows:(1)In view of the complexity and variability of the existing meteorological information,it is difficult to describe the change process by traditional methods,and a short-term meteorological information prediction method based on chaotic time series is proposed.Based on the characteristics of time series of meteorological information,chaotic time series of meteorological information is analyzed using chaotic time series method.The coordinate delay method is used to reconstruct the phase space,and the time series of the meteorological information is extended to the high dimension space to confirm time delay and embedding dimension of the series,which gives it the ability to predict chaos.The weather prediction model based on the chaotic time series is constructed to predict the short-term meteorological information in the process of dam spreading.(2)In view of the nonlinear variation of moisture content in dam material caused by meteorological factors,a prediction method for moisture content change of dam material based on Stochastic Forest regression is proposed.First,the original data is constructed based on the change of moisture content of the dam material and meteorological information,and the original data are resamaged using the bagging method to ensure the difference of the training samples.Secondly,the random subspace method is used to generate regression trees to reduce the correlation between trees and trees.Finally,the prediction model of moisture content change of dam material is established based on random forest regression method,and ten-fold cross is used to test the prediction accuracy of the model.At the same time,the nonlinear change of moisture content of dam material caused by meteorological factors is predicted.(3)in view of accurate calculation of the watering volume of the dam material,watering volume prediction model is constructed based on chaotic time series and random forest regression,which is applied to a practical project to realize the accurate calculation of the watering volume of the dam material.Combining the forecast results of the moisture content change of the dam material,the initial moisture content of the dam material,the design moisture content and the dam material quality,a prediction model for dam material watering volume based on the chaotic time series and random forest regression model was constructed to accurately calculate the watering volume Based on the paving and rolling monitoring system and the transportation monitoring system,this method was applied to a practical project and compared with the existing watering method,which verified the validity and accuracy of the method,provide a scientific guidance for the fine control of watering volume of rock-fill dam material.