基于时间序列与DEGWO-SVR模型的隧道变形预测方法

为了对准确预测隧道变形的非等距性及复杂非线性特征，结合时序分析理论、差分进化算法(DE)、灰狼优化算法(GWO)和支持向量回归机(SVR)模型，提出新的隧道变形预测模型. 利用3次样条函数插值法将非等距监测数据等距化；基于时间序列原理将变形分解为趋势项及平稳随机项，并采用所提模型分别对2个分解项进行预测；将各位移分量叠加，实现隧道累积变形的预测. 以重庆市兴隆隧道实测拱顶沉降为例，预测前方ZK37+900和ZK37+910断面拱顶沉降，并与已有模型进行对比. 结果表明：所提模型预测的均方根误差分别为0.193 7 、0.086 9 mm，平均绝对百分比误差分别为1.21%、0.55%，相关系数分别为0.997 1、0.992 8. 相比于已有模型，所提模型的预测精度更高、误差更小，具有更好的适用性及应用前景.

Prediction method of tunnel deformation based on time series and DEGWO-SVR model

Combining time series analysis, differential evolution (DE), grey wolf optimizer (GWO), and support vector regression (SVR), a new prediction model of tunnel deformation was proposed, in order to accurately predict tunnel deformation has the characteristics of non-equidistant and complex nonlinear. Firstly, non-equidistant data was processed equidistantly by cubic-spline function interpolation. Then based on the time series principle, the deformation were decomposed into the trend terms and stationary random terms. The proposed model was used to predict the two terms. Finally, the displacement components were superimposed to realize the prediction of the tunnel cumulative deformation. Taking the measured vault settlement of Xinglong Tunnel in Chongqing as an example, the vault settlement of the front ZK37+900 and ZK37+910 section was predicted and compared with existing models. Results showed that the root-mean-square error were 0.193 7 mm and 0.086 9 mm, the mean-absolute-percent error were 1.21 % and 0.55 %, and the correlation coefficient were 0.997 1 and 0.992 8. Compared with the existing models, the proposed model has higher prediction accuracy and smaller error, it has better applicability and application prospects.