Research On Prediction And Application Of Subgrade Settlement Of High Speed Railway Based On Grey Combination Model

With the rapid development of high-speed railway construction in our country,the safety of railway operation has become an important travel guarantee for the safety of people and property.In order to make the high-speed railway have a safe and stable operation environment,it is necessary to strictly control the deformation of its off-line projects,especially in the subgrade section which is most prone to settlement deformation.Although the single prediction method has good prediction effect and ability in buildings,dams,slopes and foundation pits,it is insufficient for the high-speed railway subgrade with settlement controlled at 15 mm.Therefore,this paper studies the settlement mechanism of high-speed railway subgrade and the prediction mechanism of related prediction models,optimizes the initial value,residual value and background value of a single prediction model to improve its prediction accuracy,and proposes to use the gray gm-bp neural network model optimized by wavelet,relying on the settlement monitoring projects of Lanxin high-speed railway and Zhonglan high-speed railway to predict the model number According to the in-depth analysis and research,a better result is obtained.Considering that the accuracy of the model is usually evaluated only by relative error and posterior error,which is too single to evaluate the prediction effect of the model comprehensively,this paper uses multiple evaluation methods such as closeness,square sum of error,standard deviation,average absolute error and average absolute percentage error to evaluate together,which greatly improves the accuracy of model evaluation.The main work of this paper is as follows:(1)Through the research on the subgrade structure of high-speed railway,the possible subgrade settlement diseases in the aspect of high-speed railway subgrade are determined.Various factors affecting the subgrade settlement of high-speed railway are analyzed and studied,and the corresponding observation scheme and technical basis are put forward according to various situations.(2)The gray system is used to predict the subgrade settlement data.Through the contrast test between the gray GM(1,1)and the gray Verhulst prediction model,the gray GM(1,1)model is used as the gray model of the gray combination model.The initial value,residual value and unequal interval sequence are improved,and the accuracy of the prediction results is improved to a certain extent.(3)Matlab wavelet toolbox is used to select wavelet function and threshold value,and wavelet soft threshold method is used to denoise subgrade settlement data,expecting more accurate prediction results in the future.The BP neural network algorithm is partly improved by using the genetic principle.The further improved BP neural network has the problems of low learning efficiency,slow convergence speed and so on.It has been applied to the deformation prediction and achieved preliminary results.(4)Based on the study of the combination mode,it is finally determined to use gm-bp series mode to carry out the combination prediction of subgrade settlement,to establish the gray gm-bp neural network prediction model optimized by wavelet,to predict the subgrade settlement data with the help of MATLAB software,and to obtain the gray GM(1,1),BP neural network and wavelet optimized GM(1,1)The results show that the prediction effect of the grey gm-bp neural network model optimized by wavelet is better than that of other single prediction models,and it is more reliable,more accurate and more applicable in the prediction of subgrade settlement of high-speed railway.