基于BP网络的深部软岩巷道围岩力学参数反演研究  

Study on Parametric Inversion of Rock Mechanics for Surrounding Rock in Deep Soft Rock Roadway by BP Neural Network Method

作　　者：夏海燕[1] 付建军[2] 时凯[3] 

机构地区：[1]湖北工程学院城市建设学院,湖北武汉410073 [2]中国水电顾问集团中南勘测设计研究院,湖南长沙4100142 [3]中国科学院武汉岩土力学研究所,湖北武汉430071

出　　处：《矿冶工程》2013年第5期25-29,共5页Mining and Metallurgical Engineering

摘　　要：为探索高地应力下深部软岩巷道围岩应力场及位移场的分布情况,建立了简化的应变软化模型,分析了此模型的关键参数;采用均匀设计法设计了数值计算中待分析的参数方案,且将各方案参数代入FLAC3D软件中进行了数值仿真分析;再将各工况下监测点计算的相对位移与岩体力学参数代入BP网络进行训练、学习,并将监测点位移相对监值输入训练的BP网络,获得参数反演值。以淮南矿业集团顾桥煤矿轨道巷为工程背景,进行了巷道深部位移监测,根据监测结果反演了简化应变软化模型的关键参数,经正分析对比验证了力学参数反演结果的准确度、可靠性,反演结果为后续工程数值计算奠定了基础。In order to explore the field of stress and displacement field of the surrounding rock in deep soft rock roadway under the condition of high stress and low strength, a simplified strain-softening model was established and key parameters of this model were analyzed. The parameter program to be analyzed in numerical computation was designed with a uniform design method, then of which all parameters were substituted into FLAC3D software for numerical simulation analysis. After that, the relative displacement and rock mass mechanical parameters of each monitoring point under different working conditions were substituted into BP neural network, for training and study, then displacement relative to monitoring was input into the trained BP neural network to obtain the real parameters. With the soft rock roadway in Guqiao Coal Mine of Huainan Mining Group as an engineering practice, key parameters in the simplified strain-softening model were processed by parametric inversion based on in-situ monitoring of the deep displacement of surrounding rock. The result has been verified to be accurate and reliable after analysis and comparison, laying the groundwork for the following engineering numerical analysis.

关 键 词：软岩巷道 围岩 高应力 应变软化模型 均匀设计 BP网络 

分 类 号：TU443[建筑科学—岩土工程]