Font Size: a A A

Study On Prediction Model Of Rockburst Hazard Based On Supervised Learning

Posted on:2024-06-22Degree:MasterType:Thesis
Country:ChinaCandidate:B M WuFull Text:PDF
GTID:2531307127970659Subject:Safety science and engineering
Abstract/Summary:PDF Full Text Request
The risk of coal mine rockburst is influenced by various factors such as natural geological conditions and mining technology conditions,which brings certain difficulties to the monitoring and early warning of rockburst.Currently,there are massive monitoring methods related to rockburst,and the large amount of monitoring data obtained can be further analyzed and excavated.The prediction model of rockburst risk still needs further optimization.This paper adopts supervised learning algorithms such as RF,XGBoost,Light GBM,LR,LSTM and so on,and uses the natural geological data,mining technology data,microseismic and stress monitoring data of a coal mining face in Shandong Province are used to carry out the research on the prediction model of rockburst risk based on supervised learning,Provide new methods and approaches for preventing the danger of rock burst during coal mining.The main research results are as follows:1)Combined with the relevant data of 10 rockburst risk indicators sorted out in the literature,a two-layer Stacking fusion prediction model of rockburst risk grade was established.Through the evaluation of four evaluation indicators,namely confusion matrix,F1 value,accuracy and Hamming loss,the analysis showed that the two-layer Stacking fusion model improved the accuracy of rockburst risk grade prediction,among the factors that affect the prediction of the risk level of rockburst,the sensitivity of structural conditions is relatively high.The research results have been verified in the prediction of the risk level of rock burst at a coal mining face in Shandong Province.The study area has a moderate risk of rock burst,which is consistent with the actual situation.2)Based on the time series data of microseismic energy monitored in real time in the region,the LSTM-GRU prediction model was established on the basis of GRU prediction model.The prediction performance of the model was evaluated by using two indicators,namely,average absolute error and root mean square error.At the same time,the evolution trend of microseismic energy in this region is predicted.Compared with GRU prediction model,LSTM-GRU model has better prediction performance and improved the accuracy of microseismic energy prediction.3)Based on the real-time monitoring of stress time series data in the region,a1DCNN-LSTM prediction model was established on the basis of the LSTM prediction model to predict the evolution trend of surrounding rock stress in the study area.It was found that compared to the LSTM-GRU model,the 1DCNN-LSTM model has a higher accuracy in predicting the evolution trend of stress.The research results in this paper can provide new methods and approaches for identifying the risk level of rockburst during coal mining,monitoring and predicting microseisms and stresses,and helping to adopt targeted rockburst prevention and control technologies.It has certain guiding significance for safe coal mining.Figure [30] Table [26] Reference [100]...
Keywords/Search Tags:Coal mine rockburst, Supervised learning algorithm, Stacking fusion model, LSTM-GRU model, CNN-LSTM model
PDF Full Text Request
Related items