Monitoring And Research For Settlement Of Changchun No.1Line Subway

With the traffic modernization continues to heat up in major cities, Beijing, HongKong, Tianjin, Shanghai, Shenzhen and other cities have built subways, China hasentered a period of rapid development of the subway construction. Firstly, the subwaycan be decomposed ground traffic to the ground, and make full use of undergroundspace which can ease the traffic pressure on the ground in a large extent; Secondly, interms of speed, stability and transport capacity, the subway should be higher than theother modes of transportation on the ground in densely populated areas; Finally, avery important point, power is the power source of the subway, which can not onlysave energy, reduce environmental pollution, but also meet the national advocacy of"low-carbon life" concept which is exactly the goal of modernization. However, inrecent years, environmental pollution, surface subsidence all affect the safety of thesubway, so the subway settlement problem is particularly important.The research of subway settlement is a multi-disciplinary problem whichinvolves surveying, geotechnical, mechanical and geological, however, the eformationmonitoring data are easily affected by the factors such as weather conditions, soilcomposition, and has the problem of limited resourcesand the failure mechanism isnot clear, so the traditional modeling ideas will not be able to accurately forecast. Thispaper use BP neural network method to forecast the settlement of the subway, whichmake full use of its high fault tolerance, adaptability and the strong nonlinear mappingability in the processing of non structural and non regular objects. This papercombines with Changchun Subway Line of FanRong Road Station settlement project,and compares BP neural network model with the traditional GM (1.1) model theDGM (1.1) model and the Verhulst model，and quantitatively analyzes the precision ofall the models. Results proves that the BP neural network model can better use theoriginal observation data for complex learning and information processing capabilitiesand has high fault tolerance and robustness. It also shows that the method cancomprehensively consider the effect of various factors, which can be applied to the actual deformation monitoring, and provide reference for later maintenance.