基于CSMR和卷积神经网络的岩质边坡稳定性分析  

Stability analysis of rock slope based on CSMR and convolution neural network

作　　者：傅杨攀 刘勇健[1] 陈贡发 陈旭林 张友 刘海龙 FU Yangpan;LIU Yongjian;CHEN Gongfa;CHEN Xulin;ZHANG You;LIU Hailong(Institute of Geotechnical Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学岩土工程研究所,广东广州510006

出　　处：《自然灾害学报》2023年第1期114-121,共8页Journal of Natural Disasters

基　　金：国家自然科学基金项目(52078142);广东省自然科学基金项目(2021A1515011691);广州市科技计划项目(202002030194)。

摘　　要：CSMR分类体系是一种半定量的岩质边坡稳定性分析方法,其综合考虑了多因素对边坡稳定性的影响,但是计算复杂。在岩质边坡稳定性评价CSMR分类体系基础上,引入卷积神经网络原理,建立基于CSMR和卷积神经网络的边坡稳定性评价模型。首先,通过85个实测岩质边坡样本对模型进行训练,构建CSMR方法中的坡高H、高度修正系数ξ、RMR评分、结构面方位修正系数(F_(1)、F_(2)、F_(3))、开挖方法修正系数F 4和结构面条件修正系数λ共8个影响因子和边坡稳定状态的非线性映射关系。然后,用另外15个边坡样验证基于CSMR和卷积神经网络的岩质边坡稳定性分析模型的有效性。最后,将模型应用于广东清远银湖城边坡稳定性分析,其预测值和期望值基本吻合。同时与有限元分析法计算结果进行了对比,表明该方法具有较强的泛化能力,能快速预测边坡稳定性,可为山区工程建设中岩质边坡工程设计和管理提供依据和参考。The CSMR classification system is a semi-quantitative method for rock slope stability analysis.It takes into account the influence of several factors on slope stability,but the calculation process is complicated.Based on the CSMR classification system of rock slope stability evaluation,the principle of convolutional neural network(CNN)is introduced to establish the slope stability evaluation model based on CSMR and CNN.Firstly,the model was trained with 85 measured rock slope samples,and the nonlinear mapping relationship was established between slope stability and 8 influencing factors,including slope height H,height correction coefficientξ,RMR score,structural plane azimuth correction coefficient(F_(1)、F_(2)、F_(3)),excavation method correction coefficient F 4 and correction coefficient of the state of the structural planeλ.Then,other 15 slope samples were used to validate the effectiveness of the proposed stability evaluation model of rock slopes based on CSMR and CNN.Finally,the model is applied to the slope stability evaluation of Yinhu City in Qingyuan,Guangdong Province,and the predicted value is basically consistent with the expected value.At the same time,the comparisons with the simulation results of finite element analysis shows that the method has strong generalization ability and can quickly predict slope stability,which can provide basis for the design and management of rock slope engineering in mountainous engineering construction.

关 键 词：岩质边坡 卷积神经网络 CSMR分类体系 稳定性评价 

分 类 号：TU457[建筑科学—岩土工程] X43[建筑科学—土工工程]