基于多变量GP-DE模型的隧道变形时间序列预测研究

Study on Time Series Prediction of the Tunnel Deformation Based on the Multivariable GP-DE Model

准确预测和控制隧道变形是确保隧道工程施工安全的重点。针对目前隧道围岩变形时间序列预测的不足,文章提出了一种基于多变量高斯过程(GP)-差异进化算法(DE)的隧道变形时间序列预测方法。根据隧道自动化监测结果进行多变量相空间重构,并通过主成分分析法降低输入维数。在此基础上采用GP-DE模型进行隧道变形预测研究。以吉林省高丽沟隧道围岩拱顶位移为例进行预测,将预测结果与BP神经网络、SVM模型预测结果进行比较。研究结果表明,多变量时间序列的GP-DE模型具有更高的预测精度,预测值与实测值吻合更好,是一种有效的隧道位移预测方法。

Accurate prediction and controlling of the tunnel deformation is the key point in ensuring the tunnel con⁃struction safety.Aiming at the insufficiency of current time series prediction method of tunnel surrounding rock de⁃formation,a time series prediction method of the tunnel deformation based on multivariable Gauss Process(GP)-Differential Evolution Algorithm(DE)is proposed.According to the results of tunnel automation monitoring,the mul⁃tivariable phase space is reconstructed,and the input dimensions are reduced by principal component analysis.On this basis,the GP-DE model is used to predict the tunnel deformation.Taking the Gaoligou tunnel in Jilin province as an example,the surrounding rock displacement on the vault crown is predicted,and the prediction results are compared with that of BP neural network and SVM model.The results show that the GP-DE model of multi-variable time series has higher prediction accuracy,and the predicted value is in good agreement with the measured value,proving that it is an effective method for tunnel displacement prediction.