地铁隧道地表变形小波去噪及灰色-时序组合预测模型研究

以青岛地铁3号线地表变形横向观测线实测数据为例，开展小波去噪及时序组合预测模型的研究。首先，采用小波理论对观测值进行粗差剔除与去噪处理，根据均方误差最低、信噪比最高的原则，证实dmey小波1层分解、rigrsure软阈值小波去噪方法是最优的。其次，给出地铁隧道地表变形灰色-时序组合预测模型表达式，选用等维新息GM(1,1)模型和残差时间序列模型进行地表变形叠合预测。最后，通过小波去噪后时间序列预测模型、小波去噪前灰色-时序组合预测模型、小波去噪后灰色-时序组合预测模型进行计算分析，结果表明小波去噪后灰色-时序组合模型预测精度最高，并分析了各模型预测精度差别的成因。

Study on Surface Deformation Wavelet De-noising of Subway Tunnel and Combined Prediction Model with Gray and Time Series

Taking measured data of surface deformation transverse observation line of Qingdao Metro Line 3 as an example, this paper studies a wavelet de-noising combined model. First, we use wavelet theory to eliminate observation value errors. According to the principle of lowest mean square error and highest signal to noise ratio, the calculated results show that dmey wavelet decomposition and rigrsure soft threshold wavelet de-noising are optimal. Second, we present the surface deformation predicting model expression combined with gray and time series of the subway tunnel. The settlement value GM(1,1) model and residual time series model are selected to predict surface deformation. Last, we analyze and compare the wavelet de-noising time series model and the combined wavelet de-noising gray and time series prediction model, both pre and post. The results show the post wavelet de-noising gray and time series combined model has the highest prediction accuracy. We analyze the different causes of each model.