Study On The Prediction Of Deformation And Rock Pressure Of Overburden Monitored By Distributed Optical Fiber Based On Machine Learning

The overburden deformation of stope is the process of rock deformation,fracture and instability driven by mine pressure,which.is manifested in large-scale deformation of overburden,development of separation layer,fissures,etc.,resulting in mi.ne disasters such as working face strong pressure.In order to grasp the law of mine pressure appearance and to solve the problem of difficult prediction of rock internal deformation,the distributed optical fiber is used to monitor the rock internal deformation,and the average frequency shift of optical fiber is introduced as the index of judging period to verify the optical fiber characterizing the deformation mechanism of overburden,and the monitored data is used as the sample set,combined with the machine learning algorithm to build the prediction model of rock pressure.In this paper,39 groups of excavation data of shallow coal seam geological condition experiment in Daliuta coal m.ine are taken as the experimental samples.The reconstruction of the data phase space,the reconstruction of the data,take the last 11 excavation as the test set,a total of 2 times of mine pressure.Using a variety of machine learning algorithms such as:neural network,support vector machine,integrated algorithm:random forest,gbdt,xgboost algorithm.the prediction model of mine pressure time series is established.On the premise that the training samples and test samples remain unchanged,the BP neural network type regression model(BPNN)is established to successfully predict one cycle weighting,and the support vector machine type regression model(SVR)is established to predict one cycle weighting The integrated learning is very good in two times of periodic weighting,among which xgboost regression algorithm(xgbr)is the best one in predicting 'mine pressure,which not only successfully predicts two times of periodic weighting,but also is the best in calculation speed and model index,which is significantly higher than other models.Single geological data can't explain the prediction performance of xgboost.Therefore,taking 60 groups of excavation data of 3D model monitoring data of Yima coalfield huge thick conglomerate as samples,taking the last 12 excavation as the test set for 5 times of mine pressure appearance,the general applicability of the algorithm can be explained only by establishing different geological data models.A large-scale three-dimensional model is established to obtain the optical fiber sensing data,and a machine learning model is established based on the frequency shift value of the overburden deformation.Three representative machine learning methods,neural network,support vector machine and xgboost,are compared to make the prediction model.The experimental results show that the integrated algorithm xgboost is better than the other two algorithms,and it can successfully predict the law of 5 times of ore pressure appearance,which can provide a scientific method for ore pressure prediction.Through the research of this paper,a prediction model integrating optical fiber,mine pressure and algorithm is established,and the response relationship between them is revealed,which can be well combined to solve the problem of mine pressure prediction and provide quantitative scientific basis for the prediction of mine pressure caused by the deformation of overlying strata in Intelligent Mining.