摘要
运用传统方法开展边坡失稳风险评估时面临着计算量大、处理过程复杂等问题。基于深度学习理论,提出一种快速评估土质边坡失稳风险的方法。考虑影响土质边坡稳定性的9个主要影响因素,利用栈式自编码器进行训练和测试,并采用误差反向传播算法计算误差的传递,在多次训练中优化网络参数,获得最优网络模型。经过测试表明:建立的模型经过训练后能够快速、准确地确定边坡失稳的风险等级。该方法摆脱了传统风险评估方法中复杂、冗长的数据预处理和确定权重的过程,可为边坡及其他工程风险等级的快速评估提供新的思路。
To overcome the existing problem in assessing instability risk grade of slope,such as huge calculations,complex data processing and so on,this paper proposes a new method to rapidly assess risk of soil slope instability based on deep learning.The stacked autoencoder in deep learning model is applied to construct predictive model after training and test considering the nine main factors of soil slope stability.The calculating error is propagated by back propagation algorithm.After a considerable training,an optimal model is obtained,and the test results show that the instability risk grade of soil slope can be assessed quickly and accurately.The proposed method avoids the complicated processes of data processing and weights determination.The research results provide new ideas for rapid assessment of slope and other engineering risk grade.
作者
于锦
谭飞
仝德富
郭永楠
马邦闯
YU Jin;TAN Fei;TONG Defu;GUO Yongnan;MA Bangchuang(Faculty of Engineering,China University of Geoscience(Wuhan),Wuhan 430074,China)
出处
《安全与环境工程》
CAS
北大核心
2020年第5期153-158,共6页
Safety and Environmental Engineering
基金
中央高校基本科研业务费专项资金项目(CUGCJ1821、CUG170645)
国家自然科学基金项目(41920104007、51879245、11672360)。
