贝叶斯优化LSTM神经网络边坡不确定性变形预测及应用  

Bayesian Optimization LSTM Neural Network Slope Uncertainty Deformation Prediction and Application

作　　者：徐霄宇 左珅[1] 左祥龙 葛玉宁 XU Xiaoyu;ZUO Shen;ZUO Xianglong;GE Yuning(Shandong Jiaotong University,Jinan,Shandong 250357,China;Shandong Luqiao Group Co.,Ltd.,Jinan,Shandong 250014,China)

机构地区：[1]山东交通学院,山东济南250357 [2]山东路桥集团有限公司,山东济南250014

出　　处：《黑龙江交通科技》2025年第2期40-43,49,共5页Communications Science and Technology Heilongjiang

摘　　要：针对目前降噪方法无法彻底消除GNSS监测到的噪声干扰,研究基于贝叶斯优化LSTM神经网络的边坡变形进行不确定性分析预测。基于LSTM神经网络算法模型基本理论,通过优化神经网络对边坡沉降变形进行预测,引用贝叶斯思想使LSTM神经网络预测实现不确定性预测功能,通过引入评价指标,以临淄至临沂高速公路工程为例验证了该方法的可行性。研究发现:贝叶斯优化后的LSTM神经网络具有更好的预测能力,随着监测时间和数据规模的增加,不确定性预测模型的精度也越来越高。In view of the fact that the current noise reduction method can not completely eliminate the noise interference monitored by GNSS,this paper intends to study the uncertainty analysis and prediction of slope deformation based on Bayesian optimized LSTM neural network.Based on the basic theory of LSTM neural network algorithm model,the slope settlement deformation is predicted by optimizing the neural network.The Bayesian idea is used to make the LSTM neural network prediction realize the uncertainty prediction function.The feasibility of the method is verified by taking the Linzi-Linyi Expressway project as an example.The study found that the Bayesian optimized LSTM neural network has better prediction ability.With the increase of monitoring time and data scale,the accuracy of the uncertainty prediction model is higher and higher.

关 键 词：边坡工程 神经网络 变形预测 贝叶斯优化 

分 类 号：U416.14[交通运输工程—道路与铁道工程]