基于 GF-GPR 的地铁车站基坑变形预测与应用研究

为解决受噪声影响地铁车站基坑变形预测精度受到限制的问题,文章首先使用高斯滤波(Gaussian filter, GF)算法对监测数据进行降噪处理,再采用高斯过程回归(Gaussian process regression, GPR)算法预测基坑变形,构建一种GF-GPR基坑变形预测模型,并将GF-GPR模型应用于成都某车站地铁基坑的变形预测。结果表明:原始监测数据存在大量噪声,变形不连续,经过GF算法降噪后基坑变形序列变得平稳,同时有用的突变信息仍然被保留。降噪后数据的信噪比(signal-to-noise ratio, SNR)为12.884~17.139,均方误差(mean square error, MSE)为0.430~0.875 mm;所提出的GF-GPR模型的变形预测结果与基坑实际变形趋势一致,GF-GPR模型的预测精度相较于单一GPR算法提高了31%~81%,最大均方根误差降低了0.4367~1.2881 mm。该研究成果可为基坑变形智能预测、施工事故防范提供参考。

In order to solve the problem that the prediction accuracy of foundation pit deformation in subway stations is limited due to noise, Gaussian filter(GF) algorithm is used to reduce the noise of monitoring data, and then Gaussian process regression(GPR) algorithm is used to predict foundation pit deformation, and a GF-GPR foundation pit deformation prediction model is constructed. The GF-GPR model is applied to the deformation prediction of a subway station foundation pit in Chengdu City. The results show that there is a lot of noise in the original monitoring data, and the deformation is discontinuous. After the noise reduction by GF algorithm, the deformation sequence of foundation pit becomes stable, and the useful mutation information is still retained. The signal-to-noise ratio(SNR) and mean square error(MSE) range from 12.884 to 17.139 and 0.430 mm to 0.875 mm, respectively. The deformation prediction results of the proposed GF-GPR model are consistent with the actual deformation trend of the foundation pit. Compared with the single GPR model, the prediction accuracy of the GF-GPR model is increased by 31%-81%, and the maximum root mean square error(RMSE) is reduced by 0.436 7-1.288 1 mm. The research results can provide references for the intelligent prediction of foundation pit deformation and the prevention of construction accidents.