【作者】 邢如飞；

【导师】 姜忻良；

【作者基本信息】 天津大学 ， 结构工程， 2007， 硕士

【摘要】 城市中深基坑工程开挖造成的土体变形会对工程和周围环境产生不利的影响。因此确定出基坑开挖过程中土体的物性参数,对于准确预测深基坑工程的开挖变形情况,指导工程顺利进行是很有意义的。本文采用了参数反分析的研究方法。应用ADINA有限元软件为正分析手段建立了模拟基坑开挖过程的模型;应用BP神经网络作为反分析工具建立了土性参数辨识系统,通过基坑支护结构实测位移数据反演修正土性参数的取值。用经过反演的土性参数作为有限元分析的输入数据,达到对深基坑的开挖过程进行较为准确模拟的目的。以有限元为代表的数值解法在深基坑开挖过程模拟中有传统方法不可替代的优势。本文选用的ADINA有限元软件带有多种常见土体材料模型,在单元生死的处理上采用单元刚度逐渐消逝的算法。因此能够很好地动态模拟深基坑工程的分步开挖施工过程。本构模型的选择和土性参数取值的确定是有限元计算面临的问题。本文较详细的介绍了土的弹塑性模型,选用了比较适合工程实际情况且参数不多的D-P本构模型。对于难以准确测定的土体弹性模量E值,选用BP神经网络对其进行了反演分析,充分利用了神经网络突出的非线性映射功能。在网络结构的选择上采用了分模块处理的方法,使得网络收敛能够顺利的完成。最后通过天津某地铁车站基坑开挖工程的计算,得到了能够反映工程实际情况的支护结构变形值预测值。证明了利用神经网络进行土性参数辨识,可以很好的应用于深基坑开挖的变形预测中。更多还原

【Abstract】 Digging the deep excavation in the city can cause earth transfiguration that would influence the works and the surroundings obviously. So it is important to determine the soil physical parameters in digging of the deep excavation to forecast the earth transfiguration levels and direct the works.This study is researched by parameter back analysis method. In order to simulate the digging process of the deep excavation accurately ADINA finite element software is used to establish the simulate model of the digging process of the excavation as forward analysis method, BP Neural Networks is used to established the recognition system of the soil physical parameters as back analysis method and the measure data of displacements of shoring of trench in the real works is used to correct the soil physical parameters with method of inversion which is being input in finite element analysis with method of inversion.Numerical solution that is represented by finite element analysis has the advantages obviously than conventional method in simulate the digging process of the deep excavation. ADINA finite element software, used in this study has kinds of usual earth material models and the element birth/death unit is calculated with unit rigidity fade away method, so it can simulate the digging process of the deep excavation step by step accurately. It is a problem for finite element analysis to choose the constitutive and determines the data of the soil physical parameters. In this study choose the D-P constitutive which is suitable for practical situations of the project and have few parameters and introduce the elastic-plastic constitutive in detail, BP Neural Networks is chose to invert and analyze the earth elastic ratio (E), which is hard to measured, can make a better use of the nonlinearity mapping function of the neural networks. Blocking method is chose in network structure to complete the network convergence smoothly. Conclusion, the model can reflect the displacement of the shoring of trench correctly and can be proved by a real excavation work of a underground metro station in Tianjin. The study indicates that Neural Networks can use in soil physical parameters recognizing, and can used to forecast the displacements in deep excavation.更多还原