@article{Wan2026, 
author = {Chen Wan and Xingxia Wang and Hang Duan and Long Zheng and Jianwen Huang},
title = {Study on the Prediction of Underground Cavern Rock Deformation Based on GRU Neural Network},
year = {2026},
journal = {Chinese Journal of Underground Space and Engineering},
volume = {22},
number = {2},
pages = {448-458},
keywords = {underground cavern, surrounding rock, deformation prediction, GRU neural network},
url = {https://www.sciopen.com/article/10.20174/j.JUSE.2026.02.07},
doi = {10.20174/j.JUSE.2026.02.07},
abstract = {In order to enhance the prediction accuracy of surrounding rock deformation, enable real-time monitoring of deformation status, prevent deformation failure, and ensure construction safety, a novel underground cavern surrounding rock deformation temporal prediction method based on GRU neural network is proposed to tackle the low training efficiency, slow convergence, and poor generalization of traditional methods, along with the establishment of a corresponding prediction framework. Utilizing monitoring data of surrounding rock deformation from the underground powerhouse on the right bank of the Baihetan Dam, predictions are made and subsequently compared and analyzed with the forecasting results generated by the Long Short-Term Memory(LSTM) neural network algorithm. The results indicate that the GRU neural network model effectively addresses the prediction challenges associated with underground cavern surrounding rock deformation, offering advantages such as simplified structure, relatively fewer parameters, rapid training and convergence rates, and high prediction accuracy. Compared to the predictions derived from the LSTM neural network algorithm, the GRU model demonstrates a reduction in training duration by over 70%, with a corresponding decrease in prediction error of more than 50%. The relative error for cumulative maximum deformation is less than 0.3%, the probability of absolute error less than 0.9 mm is as high as 95%, and the maximum absolute error is only 2.05 mm.}
}