摘要
为降低道路地形沉降的预测误差,提出一种基于优化卷积神经网络-双向长短期记忆网络(Convolutional Neural Network-Bidirectional Long Short-Term Memory Network,CNN-BiLSTM)的道路地形沉降预测模型。首先,结合CNN和双BiLSTM在提取空间和时间特征上的优势,构建CNN-BiLSTM模型;然后,采用海象优化算法(Walrus Optimization Algorithm,WaOA)对模型的学习率、迭代次数和隐含层神经元数进行优化,提出优化的CNN-BiLSTM模型;最后采用优化的CNN-BiLSTM模型对道路地沉降数据进行预测,并在仿真平台上进行验证。结果表明,该模型对道路地形沉降预测的均方根误差(RMSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)分别为1.56、1.31、1.40 mm;相较于其他模型,RMSE、MAE、MAPE低了约4 mm。结果证明,本文模型可降低道路地形沉降的预测误差,为评估道路长期稳定性提供了技术支持。
To reduce the prediction error of road subsidence,an optimized convolutional neural networks-bidirectional long short-term memory networks(CNN-BiLSTM)model for road subsidence prediction is proposed.Firstly,leveraging the advantages of CNN and BiLSTM in extracting spatial and temporal features,a CNN-BiLSTM model is constructed.Then,the walrus optimization algorithm(WaOA)is employed to optimize the learning rate,iteration count,and number of hidden neurons in the model,resulting in an optimized CNN-BiLSTM model.Finally,the optimized model is used to predict road subsidence data and validated on a simulation platform.The results show that the proposed model achieves an root mean square error(RMSE),mean absolute error(MAE),and mean absolute percentage error(MAPE)of 1.56,1.31,and 1.40 mm,respectively,for road subsidence prediction.Compared to other models,the RMSE,MAE,and MAPE are reduced by approximately 4 mm.The results demonstrate that the proposed model can minimize the prediction error of road subsidence and provides technical support for assessing long-term road stability.
作者
高静
GAO Jing(CCTEG Beijing Huayu Engineering Co.,Ltd.,Tianjing 300000,China)
出处
《国外电子测量技术》
2025年第12期123-129,共7页
Foreign Electronic Measurement Technology
