基于XGBoost-LSTM的光纤监测巷道变形预测

打开文本图片集

中图分类号：TD32 文献标志码：A 文章编号：1008-0562（2026）01-0017-08

XGBoost-LSTM based prediction of tunnel deformation from optical fiber monitoring

YANGJianfeng1²，LUO Ke1 ， CHAI Jing1，2 ，ZHANGDingding1²，JINGChao'，LIUYongliang

（1.College of Energy and Mining Engineering，Xi'an University of Science and Technology，Xi'an 710054，China;2.Key Laboratory of Western Mine Exploitation and Hazard Prevention，MinistryofEducation，

Xi'an University ofScience and Technology，Xi'an 710o54，China;3.School of MineSafety，North China Institute of Science and Technology，Langfang O652o1， China）

Abstract：To address the problem of roadway instability caused by deformation of sectional coal pillars beneath remnant pilars during close-distance coal seam mining，a distributed optical fiber sensing method was employed byembedding fibers inside thecoal pillars.Strain data from five monitoring boreholes were systematically processedand normalized.A training dataset was constructed using a sliding-window approach，and hyperparameters were optimized via grid search to develop an integrated mine pressure prediction model combining XGBoost （Extreme Gradient Boosting） and LSTM （Long Short-Term Memory） algorithms.The results show that the proposed model achieves a coefficient of determination （ （R2） of 0.922，with the root mean square error （RMSE） reduced to 4.215 and the mean absolute error （MAE）lowered to 2.135，demonstrating superior prediction accuracy，robustness，and generalization compared with single models such as XGBoost， LSTM，and random forest （RF）.The research conclusion reveals the horizontal strain distribution law of coal pilar under the condition of secondary mining，and provides reference for the deformation prediction of section coal pillar.

Keywords：distributed optical fiber;internal deformationofcoal pilars；ensemble prediction; hyperparameter optimization; error analysis

0引言

近距离煤层的多层次联合开采是煤矿高效开采的重要途径[1-2]。（剩余11054字）