Research On Differential Settlement Control Standard And Prediction Method Of Road Subgrade In Seasonal Frozen Area

The seasonal frozen soil area of China is up to more than50%of the land area of thenation, so the development and use of seasonal frozen soil occupies an extremely importantstrategic position in China’s economic construction and social development. With thedeepening of China’s western development strategy and the revitalization of northeast oldindustrial base strategy, Speeding up infrastructure construction will be in these areas rapidsocio-economic development of effective protection, where road construction is the mostimportant basic premise. However, the road subgrade frost damage has been an importantissue for the problem in China’s road construction season frozen zone.Subgrade uneven thawing settlement deformation caused by subgrade seasonal freezingand thawing has become one of the important causes of the subgrade and pavement earlydamage in season frozen zone. Due to the subgrade and pavement as a whole, subgradeuneven thawing settlement deformation leads to inhomogeneous deformation of thepavement, and additional stress appears in the pavement structure layer. When thenon-uniform deformation value exceeds a certain limit, the sum of large additional stress andload stress of pavement is greater than the allowable tensile stress of pavement materials sothat the pavement structure layer is damaged. Subgrade strength and stability is the guaranteeof the basic conditions of the road normal use, which is mainly reflected in the size of thesubgrade settlement deformation. Therefore, the establishment of subgrade differentialsettlement control standard and accurate prediction of subgrade settlement deformation lawhas the vital significance to ensure safe operation in the road.In order to ensure that the roadbed has a sufficient stability, it is necessary to obtain thereal-time status of subgrade strength and carry out subgrade strength monitoring. Subgrademoisture content, degree of compaction, wet and dry type have exponential relationship withthe resilient modulus in seasonal frozen zone, so the resilient modulus can better reflect theoverall subgrade conditions, which is used to scientifically and reasonably determine thestability of subgrade. At present, the determination method of subgrade resilient modulus arebearing plate method, beckman beam method and falling weight deflection instrumentmethod, etc. Because the whole operation of the site bearing plate method and Beckmanbeam method is time-consuming and laborious, which is greatly influenced by human factors and the environment so that the precision is low, the two methods cannot satisfy the rapiddetection of large area and data acquisition of the pavement management system. For thefalling weight deflectometer method, its cost is expensive, its test load is too large, the tplastic deformation of pavement structure layer affects its test results, and its backcalculation is a very difficult problem. Therefore, it is necessary to conduct in-depthscientific research on the dynamic monitoring method of resilient modulus of subgrade soil,in order to provide scientific testing methods and accurate reference value for asphaltpavement structure design, which has important practical value and significance.Because of the climatic environment and geological conditions and other reasons in theseasonal frozen area, frost heave in the winter and thawing settlement in the spring occurevery year on highway, and subgrade is very prone to non-uniform deformation. Meanwhile,excessive differential settlement can make subgrade structure damage, so the subgradedeformation prediction is particularly important in the seasonal frozen zone. Many predictivemodels and methods are mostly confined to the modeling and prediction of individualmonitoring points at present, without considering the interaction relationship between eachmonitoring point, which is only a local deformation analysis research of monitoring object.In fact, the deformation of a single monitoring point is affected by other monitoring points,but also affects the deformation of other monitoring points in the process of subgradesettlement, which is a systematic process of changing with mutual influence restrictionbetween monitoring points. As a result, subgrade settlement deformation law should isresearched from the system perspective, and settlement observation data is properly treatedas a whole, in order to make an accurate prediction of settlement deformation, which hasgreat realistic significance and broad application to reduce road disasters, ensure drivingsafety, and improve economic efficiency.This paper relies on the National High Technology Research and Development Project(863Project) of China (Project No.2009AA11Z104) named “Research on a wide range ofroad hazard parameters monitoring and identification of early warning system in seasonalfrozen area”. Subgrade differential settlement control standard, subgrade stability judgmentanalysis method, and improved gray multivariable predictive model for subgrade settlementprediction are researched systematically, mainly in the following aspects:1. Through analyzing pavement mechanical response under different conditions ofdifferential settlement, the size of differential settlement is calculated quantitatively when theroad is destroyed, and then the differential settlement control standard is established based requirement for differential settlement, the differential settlement control standard of roadsubgrade in the seasonal frozen area is determined, which is divided into five levels: safer,safe, dangerous, more dangerous, and very dangerous.2. Since the resilient modulus can well reflect the overall subgrade conditions, itscientifically determines the subgrade stability. Based on the multi-layer elastic systemtheory, there is an inherent relationship between subgrade resilient modulus and strain ofbase top surface, and the strain of base top surface can be monitored in real-time andaccurately. By using BP neural network algorithm as an inversion method, the mathematicalinversion model of subgrade resilient modulus is established, in which the strain of base topsurface in the characteristic cross-section is taken as an input variable and subgrade resilientmodulus is considered as as an output variable.3. Subgrade settlement deformation is a complex system process in seasonal frozen area.The commonly used mathematical prediction models are limited to modeling and forecastingof time series data for a single monitoring point, without considering the mutual influencebetween each monitoring point, so they are not enough to reflect the overall deformationtrend of subgrade. Considering the deformation of monitoring points in the deformable bodyfrom a systems perspective, the multivariable MGM (1, n) model is an extension of thesingle variable GM (1,1) model in the case of n variables, so as to achieve building andforecasting of the deformation prediction of interactional multiple monitoring points.Through analyzing the calculation error of background value for traditional multivariableMGM (1, n) model, a new calculating formula of background value is put forward, and theoptimized multivariable MGM(1, n) model is established by using non-homogeneousexponential function fitting the accumulation generation sequence of each variable in themodel.4. In actual subgrade settlement monitoring process, there is usually the unequalinterval sequence problem, namely the sampling period of monitoring data is difficult tomaintain consistent. The original non-equidistant monitoring data greatly reduces theaccuracy and application range of predictive model. Therefore, through theoretical analysisfor modeling mechanism of the non-equidistant sequence, the non-equidistant multivariableMGM (1, n) model is established to fit and predict the non-equidistant sequence of subgradesettlement monitoring for the multivariable with mutual influence and restriction relationship.At the same time, the calculation method of background value is one of important factors which affect the accuracy and adaptability of gray forecast model, so the non-homogeneousexponential function is used to fit the accumulation generation sequence of multivariableMGM (1, n) model, an optimization approach of background value for the model is putforward in order to improve the prediction effect of the non-equidistant multivariable MGM(1, n) model.