PDF全文阅读

     李正印1  ,  何巧灵2  ,  魏汝明2  ,  刘振东3

               ( 1 .  山东省建筑设计研究院有限公司 ，济南 250000 ;  2.  济南市勘察测绘研究院 ，济南 250000 ;

          3 .  济南市市政工程设计研究院（集团） 有限责任公司 ，济南 250000)

摘   要 ：为实现基坑沉降的高精度预测 ，基于基坑现场沉降监测成果 ，先通过极点对称模态分解算法开展 沉降数据的分解处理 ，得到若干模态分量和趋势分量 ，再利用灰狼算法 、门控循环单元神经网络构建组合预测 模型 ，并利用此模型开展各模态分量 、趋势分量的变形预测 ， 以得到基坑沉降的组合预测值。结果表明： 在基 坑施工过程中 ，常见沉降监测项目包括地表沉降 、坑顶沉降和建筑物沉降 ；经监测数据分析 ，坑顶沉降的剩余 变形空间相对最大 ，其次是建筑物沉降和地表沉降 ；总体来说 ，3 类沉降变形项目的剩余变形空间还是较为乐 观 ，只是局部少量监测点的剩余变形空间较少 ；在数据处理过程中 ，极点对称模态分解算法具有较强的数据处 理能力 ，且其能力明显优于小波去噪和模态分解法的数据处理效果 ，且经 GWO - GRU 预测 ，得到此模型在 3 类 沉降项目中 ，预测结果的相对误差均值是在 2% 左右 ，具有较高的预测精度 ，且其预测结果得出 3 类沉降变形在 后续的发展趋势较为一致 ，均呈小速率增加趋势 ，收敛特征均较为明显 ，侧面验证了基坑支护措施的运营效果 良好。通过研究 ，可为类似工程提供技术参考 ，具有一定的现实意义。

关键词：基坑 ；沉降变形 ；数据分解 ；灰狼算法 ；组合预测

中图分类号：U459           文献标志码：A 

          

 文章编号： 1005- 8249   (2025)  03- 0128- 06 

DOI:10. 19860/j.cnki.issn1005 - 8249.2025 .03 .023

Li Zhengyin¹  He Qiaoling²  Wei Ruming²  Liu Zhendong³

（1. Shandong Provincial Architectural Design Institute Co., Ltd ,Ji’nan, 250000 ,China; 2. Jinan Survey and Mapping Research Institute, Ji’nan, 250000 ,China; 3. Jinan Municipal Engineering Design and Research Institute (Group) Co., Ltd ,Ji’nan, 250000 ,China）

Abstract: To achieve high-precision prediction of foundation pit settlement, based on the on-site settlement monitoring results of the foundation pit, the polar symmetric mode decomposition algorithm is first used to decompose the settlement data, obtaining several modal components and trend components. Then, the grey wolf algorithm and gated recurrent unit neural network are used to construct a combined prediction model, and this model is used to predict the deformation of each modal component and trend component, in order to obtain the combined prediction value of foundation pit settlement. The analysis results indicate that common settlement monitoring items during foundation pit construction include surface settlement, pit top settlement, and building settlement; According to monitoring data analysis, the remaining deformation space of the pit top settlement is relatively the largest, followed by building settlement and surface settlement; Overall, the remaining deformation space for the three types of settlement and deformation projects is still relatively optimistic, with only a small number of monitoring points having limited remaining deformation space in certain areas; In the process of data processing, the pole symmetric mode decomposition algorithm has strong data processing ability, and its ability is significantly better than the data processing effect of wavelet denoising and mode decomposition methods. According to GWO-GRU prediction, the relative error mean of the prediction results of this model in three types of settlement projects is around 2%, which has high prediction accuracy. The prediction results show that the development trend of the three types of settlement deformation is relatively consistent in the future, showing a small rate increase trend and obvious convergence characteristics. This indirectly verifies the good operational effect of foundation pit support measures. Through research, it can provide technical references for similar projects and has certain practical significance.

Keywords: excavation pit; settlement deformation; data decomposition; grey wolf algorithm; combination prediction

作者简介：李正印  (1989—) ,  男 ，本科 ，高级工程师 ，研究方向：岩土工程勘察 、基坑边坡设计等相关技术服务。

通信作者：何巧灵  (1987—) ,  女 ，硕士 ，高级工程师 ，研究方向：岩土工程相关生产及科研工作。

收稿日期：2024 - 12 - 19