苏州地铁车站基坑变形的空间效应分析及预测研究

Spatial effect analysis and prediction of foundation pit deformation of Suzhou metro station

为有效掌握苏州地铁基坑的空间变形特征,实现对其变形的高精度预测,利用数理统计及极限学习机开展基坑变形的空间效应分析及预测研究。首先,基于基坑变形监测成果,开展基坑变形的空间效应分析;其次,利用小波去噪、极限学习机及混沌理论等构建了基坑变形预测模型,以达到基坑变形的时间效应分析。结果表明,在空间效应分析方面,基坑竖向沉降变形及侧向水平位移均以西侧相对最大,其与近接既有住宅楼相关,近接既有建筑物对基坑变形的影响较大,且在不同施工阶段条件下,基坑侧位移随深度变化存在一定差异,第二、三阶段的侧位移增量相对最大;在时间效应分析方面,小波去噪能很好分解基坑变形的趋势项和误差项,但去噪参数对去噪效果影响较大,去噪过程应进行参数的优化筛选;粒子群算法能有效提高极限学习机的预测精度,混沌理论也能有效弱化误差序列,所得变形预测结果的相对误差均值均小于2%,验证了预测模型的有效性,且通过外推预测,得出基坑变形在后4个周期仍将进一步增加,但增加速率相对较小,后期在时间效应条件下趋于稳定。研究结果对开展不同区域地质条件下的基坑变形特性研究、更好地指导现场施工具有参考价值。

In order to effectively grasp the spatial deformation characteristics of Suzhou metro foundation pit and realize its high-precision deformation prediction,mathematical statistics and limit learning machine were used to study the spatial effect analysis and prediction research of foundation pit deformation.Firstly,based on the monitoring results of foundation pit deformation,the spatial effect of foundation pit deformation was analyzed;secondly,the prediction model of foundation pit deformation was constructed by using wavelet denoising,limit learning machine and chaos theory to achieve the time effect analysis of foundation pit deformation.Case study shows that:in the aspect of spatial effect analysis,the vertical settlement deformation and lateral horizontal displacement of foundation pit can reach a relative maximum value in the west side,which is related to the adjacent existing residential buildings.In addition,the adjacent existing buildings have a greater impact on the deformation of foundation pit,and under different construction stages,the lateral displacement of foundation pit rate varies with the depth,and the increment of lateral displacement in the second and third stages is used to be relatively largest.In the aspect of time effect analysis,wavelet denoising can decompose the trend term and error term of foundation pit deformation well,but the denoising parameters have a great influence on the denoising effect,so the parameters should be optimized in the denoising process;meanwhile,particle swarm optimization algorithm can effectively improve the prediction accuracy of the limit learning machine,chaos theory can also effectively weaken the error sequence,and the relative error mean value of the deformation prediction results can be obtained.All of them are less than 2%,which proves the validity of the prediction model.Through extrapolation,it shows that the foundation pit deformation will increase further in the last four periods,but the increase rate will be relatively small and tend to be stable under effect condition the later time.The research result provides some reference for the study of foundation pit deformation under regional geology conditions and guidance of on-site construction.