【作者】 王鹏；

【导师】 梁正召；

【作者基本信息】 大连理工大学 ， 建筑与土木工程（专业学位）， 2022， 硕士

【摘要】 目前,工程中基坑事故仍处于高发状态,尤其是深基坑工程。通过对事故进行分析,发现对基坑变形监测及预警的重视程度不够是造成事故发生的重要原因。基坑工程情况较为复杂,影响基坑变形的因素众多:岩土体材料的多相性、基坑深度和长宽比、现场的降雨和施工速率等因素都会影响基坑的变形和稳定。传统的监测及预警手段效果差的原因在于很难将众多影响因素全部考虑,也很难做到实时监控,而随着计算机技术的发展,实现对基坑变形的高精度实时监测及预警已成为可能,这将对基坑工程的安全管理有着重要的现实意义。本文以深圳市南山区某一深基坑项目为研究实例,建立了一套基于智能算法的基坑变形监测预警系统。整个系统由以基坑变形监测、变形预测和安全预警三大模块为主体的算法包构成。智能算法采用Python语言编程实现,通过将监测、预测和预警结合起来,层层递进,在时间维度上基于每个点的监测数据实现对未来各点变形的预测和趋势分析,同时,从两个层面考虑了基坑的安全稳定,既保证各监测点的安全,又实现了对基坑整体变形的管控。(1)在基坑监测模块,首先对现场收集到的监测数据缺失值进行线性插值处理,再根据同一监测项目位置相近监测点变形情况具有一致性这一原理,利用灰色关联度方法对监测数据进行关联度分析,监测点关联度变化情况采用动态滚动方法实现,以10-12个监测数据为一个序列,3-5个数据迭代为1个步长,进而实现对基坑变形趋势按照区域把控,方便对异常的监测传感器进行检查和更换,降低工程风险。(2)在变形预测模块,利用两种预测方法——不等时距灰色模型和卷积神经网络实现对基坑的高精度预测。在工程前期或监测点数据较少时使用单变量灰色模型预测方法GM(1,1)。当工程后期数据量较多时,采用一维卷积神经网络(CNN)进行预测。两种方法都采用动态滚动预测方法实现,以7个数据为一个序列预测变形,设置动态滚动预测步长为1。除了完成各点变形预测以外,还须对预测结果进行误差分析,并通过持续收集到的误差情况选择精度更高的方法进行预测。(3)在安全预警模块,将收集到的监测数据采用数据融合的方法对基坑整体变形情况做出判断,再根据基坑监测点安全分级表对各监测点稳定与否做出评判,接着利用层次分析法(AHP)搭建基坑安全预警框架,由监测点数据逐级分析,得到基坑整体稳定性,但仅仅得到基坑稳定性的定性评价是不够的,为了解决这一问题,将层次分析法与模糊数学中的隶属函数概念相结合,提出了定量评价基坑稳定的新举措。本文以基坑工程为研究对象,以人工智能与应用数学为方法,借助计算机编程语言Python,对基坑的变形和安全稳定做了详细分析探讨,为以后基坑工程现场管理提供了切实可行的新方法,对规避工程风险,保证工程质量有着十分重要的现实意义。更多还原

【Abstract】 At present,foundation pit accidents are still high in engineering,especially for deep foundation pit projects.By analyzing the accident,it is found that the lack of attention to the monitoring and early warning of foundation pit deformation is an important cause of the accident.The foundation pit engineering situation is complex,and there are many factors that affect the foundation pit deformation: the heterogeneity of rock and soil materials,the depth and aspect ratio of the foundation pit,the on-site rainfall and construction rate and other factors will affect the deformation and stability of the foundation pit.The reason for the poor effect of traditional monitoring and early warning methods is that it is difficult to consider all the influencing factors,and it is difficult to achieve real-time monitoring.With the development of computer technology,the realization of high-precision real-time monitoring and early warning of foundation pit deformation has become a Possibly,this will have important practical significance for the safety management of foundation pit engineering.This paper takes a deep foundation pit project in Nanshan District,Shenzhen as a research example,and establishes a set of foundation pit deformation monitoring and early warning system based on intelligent algorithms.The whole system is composed of an algorithm package with three modules as foundation pit deformation monitoring,deformation prediction and safety early warning.The intelligent algorithm is implemented by Python language programming.By combining monitoring,forecasting and early warning,it progresses layer by layer.Based on the monitoring data of each point in the time dimension,it realizes the prediction and trend analysis of the deformation of each point in the future.The safety and stability of the foundation pit are considered at each level,which not only ensures the safety of each monitoring point,but also realizes the control of the overall deformation of the foundation pit.(1)In the foundation pit monitoring module,the missing values of the monitoring data collected on the spot are firstly processed by linear interpolation,and then according to the principle that the deformation of the monitoring points in the same monitoring project is consistent,the gray correlation method is used to analyze the monitoring data.The correlation analysis is carried out,and the change of the correlation degree of the monitoring points is realized by the dynamic rolling method.With 10-12 monitoring data as a sequence,and 3-5data iterations as a step,the deformation trend of the foundation pit can be analyzed according to the region.It is convenient to check and replace abnormal monitoring sensors and reduce engineering risks.(2)In the deformation prediction module,two prediction methods,unequal time interval gray model and convolutional neural network,are used to achieve high-precision prediction of foundation pits.The univariate grey model prediction method GM(1,1)is used in the early stage of the project or when the monitoring point data is less.When the amount of data is large in the later stage of the project,a one-dimensional convolutional neural network(CNN)is used for prediction.Both methods are implemented by the dynamic rolling prediction method,which uses 7 data as a sequence to predict the deformation,and sets the dynamic rolling prediction step size to 1.In addition to completing the deformation prediction of each point,it is also necessary to carry out error analysis on the prediction results,and select a method with higher accuracy for prediction through the continuously collected error conditions.(3)In the safety early warning module,the collected monitoring data is used to judge the overall deformation of the foundation pit by the method of data fusion,and then the stability of each monitoring point is judged according to the safety grading table of the monitoring points of the foundation pit,and then Use the Analytic Hierarchy Process(AHP)to build a foundation pit safety early warning framework,and analyze the monitoring point data step by step to obtain the overall stability of the foundation pit,but it is not enough to obtain the qualitative evaluation of the stability of the foundation pit.Combining AHP with the concept of membership function in fuzzy mathematics,a new measure for quantitative evaluation of foundation pit stability is proposed.This paper takes the foundation pit engineering as the research object,uses artificial intelligence and applied mathematics as the method,and uses the computer programming language Python to make a detailed analysis and discussion on the deformation,safety and stability of the foundation pit,which provides practical and feasible on-site management of the foundation pit engineering in the future.The new method is of great practical significance for avoiding engineering risks and ensuring engineering quality.更多还原