【作者】 朱明；

【导师】 周建亮；

【作者基本信息】 中国矿业大学 ， 工程管理（专业学位）， 2021， 硕士

【摘要】 现在城市中超高层建筑越来越多,摩天大楼已经是每个城市地标性建筑的象征,是每个城市靓丽的建筑名片和现代化文明的标志,随着建筑物地面以上高度不断增加,地下结构的深度也越来越深和平面尺寸也越来越大,以及城市轨道交通项目、地下综合管廊项目在各个城市的兴起,导致深基坑工程项目越来越多。深基坑工程向超深、超大、超紧的趋势发展的同时,也造成了许多安全事故。设计失误是造成深基坑事故的首要因素,因此,要想保障深基坑工程的安全,深基坑工程设计是关键,在设计阶段采取科学的管理方法,使设计达到零失误,深基坑安全事故将会大量地减少。深基坑支护方案设计又是深基坑工程设计中的关键,在对深基坑支护进行细部设计前,首先要选取一个最优支护方案。在此背景下,本研究首先通过对大量基坑支护案例的分析和查阅基坑工程相关的书籍和手册,从影响深基坑支护方案优选的众多因素中选出9个主要影响因素,然后用改进的主成分分析法选出主要影响因素的主成分,最后用改进的BP人工神经网络结合Matlab软件代码程序通过试算对比确定了科学的、客观的深基坑支护方案优选模型。主要研究内容如下:⑴在影响深基坑支护方案优选的众多因素中,选出基坑开挖深度、基坑平面尺寸、主导土层类别、土层数、土的平均重度、土的平均粘聚力、土的平均摩擦角、地下含水量和基坑与周边建筑物邻近关系的安全等级这9个主要影响因素,作为深基坑支护方案的影响因素集,对定性因素采用相关标准进行量化,为影响因素集的主成分确定奠定了基础。⑵确定深基坑支护方案影响因素集的主成分,将9个有相关性的影响因素减少为5个不相关的主成分,加快BP人工神经网络模型的收敛速度。⑶从网络拓扑结构、梯度下降法、激活函数三方面提出一些对传统的BP人工神经网络改进的方法。用改进的BP人工神经网络并运用Matlab软件编写代码程序,通过试算对比建立深基坑支护方案优选模型。该论文有图50幅,表11个,参考文献76篇。更多还原

【Abstract】 Nowadays,there are more and more super high-rise buildings in cities.Skyscrapers have become the symbol of landmark buildings in every city,the beautiful building name card of every city and the symbol of modern civilization.With the increasing above ground height of buildings,the depth of underground structure is getting deeper and the plane size is getting bigger and bigger.As well as the rise of urban rail transit projects and underground comprehensive pipe gallery projects in various cities,there are more and more deep foundation pit engineering projects.With the development of deep foundation pit engineering to ultra deep,super large and super tight,many safety accidents have also been caused.Design error is the primary factor that causes deep foundation pit accidents,therefore,to ensure the safety of deep foundation pit engineering,deep foundation pit engineering design is the key,adopt scientific management method in the design stage,make the design reach zero error,deep foundation pit safety accidents will be greatly reduced.The design of deep foundation pit support scheme is the key to the design of deep foundation pit engineering.Before the detailed design of deep foundation pit support,an optimal support scheme should be selected first.In this context,this study first analyzes a large number of foundation pit supporting cases and references to books and manuals related to foundation pit engineering to select 9 main influencing factors from many factors affecting the optimization of deep foundation pit supporting scheme,and then uses the improved principal component analysis method to select the principal components of the main influencing factors.Finally,a scientific and objective optimization model of deep foundation pit supporting scheme is determined by trial calculation comparison with improved BP artificial neural network and MATLAB software code program.The main research contents are as follows:(1)In the numerous factors influencing the deep excavation,the select,the size of the foundation pit excavation depth,the number of dominant soil type,soil,the soil average heavy,soil,on average,cohesive force,the average friction Angle of soil,underground water and the safety of foundation pit and surrounding buildings neighboring relations the nine main influencing factors,as the influence factors of deep foundation pit support scheme set,The qualitative factors are quantified by the relative standard,which lays a foundation for determining the principal component of the influencing factor set.(2)Determine the principal components of the influence factor set of deep foundation pit supporting scheme,reduce the 9 related factors to 5 unrelated principal components,and accelerate the convergence speed of BP artificial neural network model.(3)Some methods to improve the traditional BP artificial neural network are proposed from three aspects of network topology,gradient descent method and activation function.By using improved BP artificial neural network and using Matlab software to write code program,the optimization model of deep foundation pit support scheme is established through trial calculation and comparison.There are 50 figures,11 tables and 76 references in this thesis.更多还原