【作者】 刘杰；

【导师】 阮怀宁； 王媛；

【作者基本信息】 河海大学 ， 岩土工程， 2002， 博士

【摘要】 本文系统总结评述了本课题相关领域业已出现的诸多成果及最新进展，指出了目前该领域存在的一些极需解决的问题及可能遇到的困难。在此基础上，本文采用包容多孔介质、多裂隙介质及离散裂隙介质的统一域混合模型，利用目前参数反演领域应用最为广泛的间接法，首先针对裂隙岩体渗流场单场参数反问题进行了深入研究，主要目的是改善参数反问题的适定性以提高参数反演的可靠程度。针对裂隙岩体渗流有限元分析提出了推求水头对渗透系数的偏导数的新算法，即有限元直接偏微分方法；并基于此算法进行参数敏感性分析，据此提出了通过参数敏感性分析构造优化目标泛函权函数的方法，从而最大限度的利用了敏感性好的有利量测信息改善反问题的不适定性，同时也适当利用了敏感性差的实测点水头值，增加了实测信息的数量，防止因量测资料过少而使反演结果严重失真；将有限元直接偏微分的方法结合Gauss-Newton-Marquit方法求解裂隙岩体渗流参数反问题，避免了反演计算中利用差分法求目标函数梯度向量带来的诸多不利，也使该类算法的反演效率大大提高；通过构造个体自适应概率对基本遗传算法做了改进，将改进后的算法与传统的单纯形方法结合提出了一种新型高效的混合遗传算法，克服了基本遗传算法局部搜索能力差、计算量大以及对大搜索空间适应能力差等缺陷；将以上两种算法分别结合裂隙岩体渗流有限元分析编制了大型反演分析程序，并将其成功的应用于典型重力坝坝基裂隙岩体渗流参数反演实践中。在以上研究的基础上，考虑到裂隙岩体渗流场与应力场存在强烈的耦合作用，本文基于求解稳定渗流场与弹性位移场耦合的全耦合分析方法，首先对耦合分析中的力学参数进行敏感性分析，而后利用水头、位移等多类型量测资料，应用本文提出的混合遗传算法进行裂隙岩体渗流与应力耦合的参数反问题研究；实际工程中流体在变形裂隙岩体介质中的流动是一个渗流场与应力场耦合作用的动态过程，数值分析应尽可能的反映这种动态特征，为此本文基于流体在变形介质中运动的一般微分方程进行了两场动态耦合问题的有限元分析，并提出了一种新的数值积分算法计算非恒定渗流有限元分析中的自由面边界积分项；同时考虑到现场观测资料往往是一个动态的时间序列，在参数反演过程中应该能及时利用现场量测的新增信息进行参数反演，本文针对确定两场动态耦合问题渗流参数与力学参数的动态反演方法进行了尝试性研究。基于以上的研究理论，本文编制了渗流场与应力场静态、动态耦合三维有限元分析程序，并结合本文提出的混合遗传算法编制了相应的参数反演分析程序；典型裂隙岸坡工程算例验证了以上程序的正确性。更多还原

【Abstract】 Research advances in the fields of seepage parameter inverse problems are reviewed in the paper. Thorough comments are developed on different parameter identification methods as well as suggestions are given about future studies. On the basis of a united mixed media model including porous medium, equivalent continuum fracture medium and discrete fracture medium, the seepage parameter inverse problem about fracture rock masses is studied systematically and some effective measures are presented to decrease the ill-posed property and improve the reliability of parameter identification. The indirect method widely used in parameter identification fields is applied to solve the seepage parameter inverse problem. In order to get the derivatives for water heads against seepage parameters, a new algorithm named direct differentiation algorithm is presented for FEM analysis of seepage problem in fracture rock masses; Based on the new algorithm, a new weight function for the objective function is developed. With the application of the weight function, the effects of monitored water heads with high sensitivity against objective parameters can be strengthen to decrease the ill-posed property of the inverse problem. At the same time, the monitored water heads with low sensitivity are used properly to insure enough monitored information against distortion of the back analysis result. By applying the direct differentiation algorithm based on FEM , Gauss-Newton-Marquit method is also improved in the paper. As a result, several defects due to the difference method in calculating the gradient vector of the objective function are eliminated and the efficiency of Gauss-Newton-Marquit method is raised greatly. Primary genetic algorithm is improved by the construction of individual self-adapting cross and mutation probabilities, and a new hybrid genetic algorithm with high efficiency is presented to decrease several limitations of primary genetic algorithm in practice, such as poor local search ability, premature convergent, excessive computational cost and bad adaptability to large search space. Universal computer code programs are developed separately by combining the above two optimization methods with FEM of seepage problem in fractured rock masses and applied successfully to the seepage parameter back analysis in the foundation of a typical gravitation dam. Because of strong inter action between the fluid flow and stress in fracture rock masses, the parameter inverse problem for the coupled analysis is further studied in which both water heads and displacements are regarded as monitored in formations. The four-freedom complete couple method is used to solve the flow-stress coupled forward problem. In the back analysis of the couple problem, the sensitivity analysis of displacements against mechanical parameters is made firstly, and on the basis of the sensitivity analysis, the objective mechanical parameters are determined. Then the hybrid genetic algorithm is applied to solve the parameter inverse problem. Furthermore, since seepage in deformable fracture rock masses is a dynamic process with strong flow-stress coupling effects,the corresponding numerical analysis should reflect the dynamic characteristics. Therefore, a FEM analysis for the dynamic flow-stress coupling problem is also developed in the paper based on the general partial differential equations of water flow in the deformable medium. And a new numerical integral algorithm with high precision is put forward to calculate the integral term on the boundaries of free surface in the FEM analysis of unstable seepage problems. At the same time, monitored data in situ can be considered as dynamic serials related to time, and a dynamic back analysis method is tried to determine the seepage and mechanical parameters in FEM analysis of the dynamic couple problem. On the basis of the above theories, universal computer codes are developed separately to deal with the static and dynamic parameter inverse problem. And a typical fractured rock slope case study is carr更多还原