Research On The Application Of Data Mining In Unconsolidated Layer Settlement

In the eastern part of China,the crust is generally distributed with deep loose layers.Due to the dehydration and consolidation deformation of the loose layer of the crust,more than one hundred coal mine shafts in the Huaibei,Xuzhou,and Quzhou mining areas have successively undergone varying degrees of Deformation or even damage,affecting safety in production.The research of using fiber grating monitoring system is an effective method to prevent deformation settlement of loose layer of the earth's crust.For the data monitored by the FBG monitoring system,the abnormal information should be identified first,and then it can be effectively used to predict and analyze the future settlement of the loose layer of the earth's crust in order to improve the reliability of the prediction of the deformation of the loose layer of the earth's crust.And mine safety assessment capabilities.This paper presents a method based on optical wavelength anomaly data detection algorithm for anomaly detection of data collected by fiber grating monitoring system.Starting from the principle of the monitoring system,the threshold value is first used to determine the abnormality of the collected wavelength data.Secondly,the detected abnormal data is classified from the model characteristics into three categories:"single anomaly","consecutive anomaly" and "abnormal sequence".Then correct the different types of abnormal data in a reasonable way.Eliminate the "corrected data exceeds the threshold phenomenon." The experimental results show that the anomaly detection algorithm proposed in this paper can identify and correct abnormal data,and the results are good.A gray Verhulst-BP model based on a sliding window is proposed in this thesis for the prediction of formation subsidence.During the experiment,the traditional Gray Verhulst model and time series ARIMA model are used to predict the ARIMA model.The ARIMA model has good prediction effect,the root mean square error is 1.3124,and the root mean square error of the gray Verhulst model is 7.2719.Secondly,the gray Verhulst model can To predict the trend,but the numerical prediction error is larger,the model is improved by setting the sliding window and the step size,and a gray Verhulst model based on the sliding window is proposed.The experimental root mean square error is 1.8664,but The method has a phenomenon that the local residual is too large;Taking into account that the BP neural network function approximation effect is good,so the use of BP neural network to predict the residual,based on the sliding window gray Verhulst model predicted value,plus prediction residuals as a true predictor.The RMSE of the gray Verhulst-BP model based on the sliding window obtained by experiments is 0.2506.The effects of the four prediction models are compared.Finally,a gray Verhulst-BP model based on the sliding window is selected for the prediction of loose crust settlement.