安全与环境工程2025，Vol.32Issue(5)：56-65,10.DOI:10.13578/j.cnki.issn.1671-1556.20250153
基于机器学习的深埋长隧道沿线地应力场反演方法
Inversion method of stress field along deep buried long tunnel based on machine learning
摘要
Abstract
The geostress field along deep-buried long tunnels is a key factor affecting the optimal design and rational construction of tunnels.However,due to the scarcity of measured data,geological complexity,and model uncertainties,obtaining the geostress distribution along the entire tunnel alignment from a limited number of measurement points remains a challenging research problem.Taking the Fozhaoshan Tunnel on the Fangxian—Wufeng Expressway as an example,this study conducts an in-depth analysis of the geological structural background,geostress field characteristics,and the geomechanical parameters of the surrounding rock under a medium-to-high stress environment.Based on the variation of geostress values with depth,data augmentation techniques were employed to effectively increase the training dataset for geostress inversion,thereby alleviating the overfitting issues of the BP neural network.A back propagation(BP)neural network inversion method for the stress field along deep-buried long tunnels based on geostress measured data augmentation is proposed,and this method was used to obtain the regional geostress field of the Fozhaoshan Tunnel.The research results indicate that the inverted geostress field is consistent with the characteristics of the regional geostress field,demonstrating that the method can reliably reproduce the geostress distribution along deep-buried long tunnel alignments.关键词
深埋长隧道/地应力场反演/反向传播(BP)神经网络/机器学习/数据增强技术Key words
deeply buried long tunnel/geostress field inversion/back propagation(BP)neural network/machine learning/data augmentation technique分类
资源环境引用本文复制引用
周正,杨旺,侯东波,商会州,曹世奇,郑虹..基于机器学习的深埋长隧道沿线地应力场反演方法[J].安全与环境工程,2025,32(5):56-65,10.基金项目
国家自然科学基金项目(52479112) （52479112）
