Application Of Combined Model In Surface Settlement Monitoring Of Mining Area

Coal has ushered in a revival after a few years of depression.Although new energy technologies are continuously developing,coal will still be our country's main energy source for a long time to come.Along with the mining,a series of problems related to it have arisen.One of them is the ground settlement caused by mining.The surface settlement of the mining area will cause disasters such as the destruction of cultivated land and the damage of buildings(structures).The effective monitoring of surface subsidence over a long period of time and mastering the laws of subsidence are effective means to avoid disastersIn this paper,the surface subsidence caused by the mining of No.11303 working face of Heze mine is taken as the research object,aiming to assist in guiding the mining behavior by establishing a suitable surface settlement prediction model,so that the surface settlement caused by mining can be effectively controlled,and disasters can be avoided,reducing direct or indirect economic losses caused by disasters arising.Around this theme,this paper has done the following work:1.Optimize a single model of surface settlement in the mining area.The non-equal interval gray multivariate model and BP neural network model are introduced.The modeling mechanism is discussed respectively.Based on this,the unequally spaced gray multivariate model based on background value optimization and the BP-based optimization are introduced.The neural network model uses the actual measured data of the mine to establish the pre-and post-optimization model in the matlab software.The prediction residual value and the sum of squared error are taken as the accuracy evaluation standard,which proves the feasibility of the model optimization method in the prediction of surface settlement in the mining area.2.Establish a combination model.This paper introduces related theories of the combinatorial model,uses optimized two kinds of single predictive models,and adopts the form of optimal fixed weight combination to establish a combinatorial model.The results showed that the residuals of the combined model at points A01,A02,and A03 were-0.6 mm to 4.8 mm,-7.6 mm to 3.7 mm,and-4.7 mm to 3.1 mm,respectively.The residuals of unequally spaced multivariate grey prediction models at points A01,A02,and A03 were-10.5 mm to 6.1 mm,-10.4 mm to 9.6 mm,and-2.1 mm to 10.1 mm,respectively.The residuals of the BP neural network models were-1.5 to 6.8mm,-8.4mm to 2.7mm,-9.6mm to-0.4mm,respectively.The research results show that the combined forecasting model is more accurate than the single predictive model,and the change curve of the predicted value is closer to the measured value.From this,we can conclude that it is feasible to apply the combined forecasting model to the prediction of surface settlement in mining areas and that it has worth promoting in practical engineering applications...