The Establishment Of Grey Multiple Correction Model And GSA-BP Model And Application Of Deformation Prediction

In recent years, with the development of mining and transportation projects, a large amount ofside slopes have been generated, including the mining slopes and road slopes. These slopes inlarge scale and enormous quantity are unprecedentedly in history. Due to the complexity ofgeological conditions, the side slopes will suffer some extent of deformation during theprocess of construction and operation. When the deformation exceeds a given threshold, theman made geologic hazard will emerge. How to predict the deformation trend correctly hasbeen a hot and difficult issue in engineering sector for a long period. According to differentsource of data, this dissertation has divided the prediction models into two categories, one isthe statistical model and the other is the non-linear model, on the basis of researching theexisting deformation prediction model. In this paper, the grey multiple correction forecastingmodel which belongs to the statistical model has been built by adopting Grey Theory Model、Background value of integral optimization、Parameters quadratic fitting optimization、Equaldimension and new information dynamic optimization and Residual error modification. Withrespect to non-linear model, GSA-BP forecasting model is constructed by using BackPropagation neural network model, Genetic Algorithm and Simulated Annealing Algorithms.The research contents and achievements are summarized as follows:(1) At the beginning of the construction of GM(1,1) model, the accuracy of model parametersalgorithm has been improved to some extent by inputting Background value of integraloptimization and Parameters quadratic fitting optimization to optimize the innerparameters in the model.(2) In order to decrease the influence caused by the length of data series, using Equaldimension and new information dynamic optimization to optimize the model andreal-time fixing the trend term of model can improve the fitting degree of predicted value.(3) At later stage of grey multiple correction forecasting model, residual error modification bysection was adopted for gray prediction of residual error, so to further optimize the modelerror trends, and combined with the known result to compose the final forecast value ofgrey multiple correction forecasting model. (4) In non-linear prediction, the BP neural network model has the inherent drawbacks.Therefore, optimization the initial weights, threshold and training phase by utilizing theglobal optimization characteristics of genetic algorithm and the high search efficientproperty of simulated anneal arithmetic.(5) On the basis of grey multiple rectification forecasting model and GSA-BP model, thereare two landslide cases comes from different data sources which can be used to analyzeand test. The testing results prove the feasibility of the two models.