Research On The Prediction Of High-speed Railway Subgrade Settlement Based On The Gobi Land Area Of The BP Optimizing Network

Since the beginning of the21st century，with the accelerated pace of high-speed railwayconstruction,more and more roadbed needs to be built above the Gobi land-based，and how toachieve accurate prediction of subgrade settlement will be the key to a reasonable choice ofengineering control measures. There are a lot of methods of subgrade settlement prediction,and most of them need to establish the geotechnical model，with some of the parameters to beknown，so it is disjointed to the the specific engineering. Therefore，to seek a more reasonableforecasting methods becomes a top priority. For these reasons，for the example of high-speedrailway roadbed of China’s Gobi land area for a certain period，make the research on thecharacteristics of subgrade settlement， settlement observation program， BP networkoptimization and BP network prediction.Firstly the thesis introduces Gobi soil subgrade and analyze the factors that affect thesubgrade settlement of the the Gobi land area，and elaborate the requirements of subgradepaving. After the scheme comparison，make the use of conjugate gradient method to the BPoptimizing neural network based on Matlab software programming. After the parameterselection，build a network to train the samples and make the network test，then establish BP tooptimize network model. This model is applied to the prediction of the Liuyuan Gobi soilsubgrade settlement by comparison of the predicted data and the actual data，then it is foundthat the conjugate gradient method to optimize the BP network has a strong operationalforecast for settlement in the Gobi land area roadbed. Finally，several issues on the subgradesettlement are summarized by the thesis, and prospect the forecast of subgrade settlement.