渗流场与应力场的完全耦合模型及其在深基坑工程中的应用

根据抽水试验观测资料，提出了土体渗透系数与有效应力的非线性耦合响应模型。建立了基于比奥固结理论、邓肯-张E-μ非线性本构关系和渗透系数非线性耦合响应关系的完全耦合分析模型，并对深基坑降排水及开挖过程中的土体渗流场与应力场进行完全耦合分析，预测了相应的土体变形。     

基于模糊理论的混凝土热力学参数识别方法

根据混凝土内部温度观测数据,建立了基于模糊理论的混凝土热力学参数识别方法。根据观测点的温度观测数据的误差特性,建立了模糊优化的目标函数。将热传导反问题作为非线性优化问题处理,然后采用BFGS方法求解。由于温度观测误差的随机特性,分析了混凝土热力学参数识别结果的统计特性。数值计算结果表明,在考虑到温度观测误差的条件下,所建立的参数反演方法能够有效地识别出混凝土的热力学参数。     

改进的模拟退火-变尺度优化方法在测井解释中的应用

实际中的测井参数反演是一个多参数、非线性优化问题,所采用的目标函数,即度量由参数化的理论模型得出的预测值与观测值的吻合程度,往往存在多解的现象。针对这种状况我们提出了模拟退火与变尺度综合反演方法用于参数计算,经过实际试算证明该方法效果很好。     

基于堆石坝竣工期沉降观测数据的材料非线性本构模型参数反演

为了高效和准确地确定堆石料的非线性本构模型参数,提出了基于响应面方法的参数反演方法。采用有限元方法数值模拟了堆石坝的分层填筑过程。建立了堆石坝变形观测点沉降的响应面函数,确定了多项式响应面函数的系数。根据堆石坝竣工期变形观测数据和确定的响应面函数,采用优化方法反演确定了堆石料本构模型参数。工程实际应用结果表明,该方法具有较高的计算效率,预测的堆石坝沉降变形与现场观测值吻合较好。     

一维冻结土体冻胀曲线的分形优化

分形几何理论作为研究自然界具有自相似性质的非线性问题的主要手段，被广泛应用于岩土工程中。土体的冻结涉及到土体本身的性质、温度场、水分场和应力场的共同作用，采用数值解法又必须对其进行大量的简化，使其所求得的数值解或拟解析解与实际值存在较大差距。采用神经网络方法研究表明，在冻胀初期系统难以稳定，同时对于低含水量的土体误差也较大。近年来，通过国内外学者的大量研究，发现土体冻结过程中同样存在分形现象。通过分析土体冻胀特征，提出采用分形插值方法研究一维冻结土体的冻胀问题。采用Douglas-Peuker算法和随机中点移位法相结合的方法对一维冻结土体的冻胀曲线进行了优化。研究结果表明，采用该方法对冻胀曲线进行优化，既可保留其原有的宏观结构特征也可以反映微观结构动态；经过插值处理后，曲线的特征点和特征形状并无变化；但细节部分随着分形维数的增加，逐步得到细化。由此真实地反映了冻胀过程，为研究土体的冻胀过程提供了一种新的方法。     

基于BP神经网络的土体细观力学参数反演分析

利用离散元方法对颗粒材料的细观力学特性研究, 目前确定数值计算模型的细观力学参数大多数通过反复调试获取, 效率低、可重复性差。本文采用开源的颗粒离散元程序LMGC开展了土体双轴压缩数值试验, 通过25组土体细观力学参数计算得到相应的宏观力学参数, 建立了BP人工神经网络反演系统。利用土体物理试验得到的土体宏观力学参数, 输入BP神经网络, 反演得到土体的细观力学参数。将所得细观力学特性参数输入所建立的土体数值计算模型, 得到土体破坏过程中的应力-应变关系曲线, 以及土体颗粒的力链图和旋转变形云图。所建立的土体数值试验模型能够较好地模拟土体变形破坏过程, 利用BP神经网络反演细观力学参数以及数值模型计算得到的土体宏观力学参数与物理试验吻合较好, 误差在10%左右, 土颗粒间力链云图以及旋转变形云图较好地揭示了土体变形破坏的机理。     

一种土的非线性弹性本构模型参数的反演方法

旨在提出一种土的非线性弹性本构模型参数反演的方法。以现今普遍实行的地基载荷试验为基础,依据遗传算法的组合优化理论,采用正演计算和遗传算法优化相结合的方式,建立了土层非线性弹性本构模型参数反演的方法;并依据某黄土场地地基载荷试验数据,实施了黄土土层非线性弹性本构模型参数反演的全过程。计算结果表明,所建立的方法可以实现土层非线性弹性本构模型中相互关联的多个参数的组合优化,并在对初始值要求较低的情况下,可以获得良好的参数反演精度。从而为土的变形特性分析和土与其中及相邻结构的共同作用分析,提供了较好的土体本构模型参数的确定方法。     

连续排水边界下考虑变荷载的软土一维非线性固结分析

土体往往具有非线性压缩特性,不同压缩特性下土体的固结规律存在差异。综合考虑土体非线性特性、变荷载作用以及连续排水边界条件建立了一维固结方程。采用无条件稳定的有限差分法和半解析法对固结方程进行了求解,两种解答方法的可靠性通过连续排水边界条件退化以及两种解答结果对比得到验证。在有限差分法解答的基础上,详细分析了界面参数、荷载参数以及非线性参数对土体固结的影响。结果表明：连续排水边界的界面参数取值越大,则超静孔隙水消散速率及土体沉降速率越大,但界面参数取值对最终沉降量没有影响;超静孔隙水压力在加载阶段逐渐增大,在恒载阶段逐渐消散;加载速率越大,则超静孔隙水压力峰值越大,并且土体固结速率越快,说明延长施工周期有利于降低超静孔隙水压力的影响;工程中要准确预测土体固结速率绝非易事,采用固结理论预测时需保证土体模型、边界条件以及土体计算参数等因素的准确性。     

不同破坏形式加筋红土强度分析

从加筋土体的等效围压概念出发,根据加筋土体发生拉裂破坏和剪切破坏时筋材所发挥的作用不同,提出了两种破坏形式下考虑等效围压折减系数的等效围压表达式,并根据加筋土体的极限平衡条件,确定了加筋土体的强度参数,分析了不同排水条件对加筋土体强度参数的影响。结合临界围压,建立起加筋土体强度破坏标准。按照一般加筋土体的广义等效围压,给出了两种破坏形式下等效围压折减系数的确定方法,并用实例进行验证。     

软土固结过程中的微结构变化特征

从粘性土体结构系统观点出发 ,建立了土体结构形态概念模型 ,即土体微结构形态由颗粒形态、颗粒排列方式、孔隙性和颗粒接触关系 4种结构要素定性描述 ,这 4种结构要素分别由颗粒大小、颗粒形状、颗粒定向性等 9种结构参数定量刻划。根据土体具有的非线性特征 ,运用分形几何理论提出了粒度分维、颗粒定向分维等 7项定量表征土体微结构状态的分维指标。在此基础上 ,以厦门软土为例开展了软土固结过程中的微结构变化规律研究 ,发现固结过程中软土微结构要素调整变化具有明显的阶段性 ,前期为剧烈调整阶段 ,后期为缓慢变化阶段 ,整体表现为分维指标下降的趋势 ,从而为软土结构力学模型的建立奠定了基础。     

与局部微调方法相结合的遗传算法

传统的遗传算法存在早熟收敛的缺点,不善于高精度的局部微调。为此,提出了一种与局部微调方法相结合的遗传算法。算例表明,它是有效的、稳健的,在水科学优化问题中具有广泛的应用价值。     

A hybrid Bayesian back‐propagation neural network approach to multivariate modelling

There is growing interest in the use of back‐propagation neural networks to model non‐linear multivariate problems in geotehnical engineering. To overcome the shortcomings of the conventional back‐propagation neural network, such as overfitting, where the neural network learns the spurious details and noise in the training examples, a hybrid back‐propagation algorithm has been developed. The method utilizes the genetic algorithms search technique and the Bayesian neural network methodology. The genetic algorithms enhance the stochastic search to locate the global minima for the neural network model. The Bayesian inference procedures essentially provide better generalization and a statistical approach to deal with data uncertainty in comparison with the conventional back‐propagation. The uncertainty of data can be indicated using error bars. Two examples are presented to demonstrate the convergence and generalization capabilities of this hybrid algorithm. Copyright © 2003 John Wiley & Sons, Ltd.     

基于SAGA-FCM的煤与瓦斯突出预测方法

为提高模糊C-均值聚类（Fuzzy C-Means Clustering Algorithm,FCM）算法在煤与瓦斯突出预测中的准确度,提出一种将模拟退火算法（Simulated Annealing Algorithm,SA）与遗传算法（Genetic Algorithm,GA）相结合用于模糊C-均值聚类分析的煤与瓦斯突出预测方法。该方法综合了模拟退火算法全局搜索、高精度的优点和遗传算法强大的空间搜索能力,将经遗传模拟退火算法优化后的初始值赋给FCM,避免了由于聚类中心初始值选择不当造成FCM算法收敛到局部极小点上。结合典型突出矿井数据进行分析,结果表明:遗传模拟退火算法优化后的FCM算法较单一,预测准确度高。      

改进的遗传算法及其在渗流参数反演中的应用

利用水头实测资料,以裂隙组的渗透系数比例因子为待反演的参数向量,在采用基本遗传算法进行参数反演研究的基础上,针对裂隙岩体无压渗流参数反问题计算量过大,目标参数众多以及参数可能变化范围大等特点,提出了一种混合遗传算法求解此类问题,力求克服简单遗传算法在解决此类问题时存在的局部搜索能力弱、易出现早熟收敛及计算量大等缺陷,并通过典型岩坡渗流算例进行验证,同时给出了基本遗传算法、传统单纯形算法的反演成果。计算结果表明,该方法保持了基本遗传算法优点,并有效地提高了算法的运行效率,从而为求解裂隙岩体无压渗流参数反问题等计算量大的系列问题提供了新的途径。     

Implicit numerical integration for a kinematic hardening soil plasticity model

Soil models based on kinematic hardening together with elements of bounding surface plasticity, provide a means of introducing some memory of recent history and stiffness variation in the predicted response of soils. Such models provide an improvement on simple elasto‐plastic models in describing soil behaviour under non‐monotonic loading. Routine use of such models requires robust numerical integration schemes. Explicit integration of highly non‐linear models requires extremely small steps in order to guarantee convergence. Here, a fully implicit scheme is presented for a simple kinematic hardening extension of the Cam clay soil model. The algorithm is based on the operator split methodology and the implicit Euler backward integration scheme is proposed to integrate the rate form of the constitutive relations. This algorithm maintains a quadratic rate of asymptotic convergence when used with a Newton–Raphson iterative procedure. Various strain‐driven axisymmetric triaxial paths are simulated in order to demonstrate the efficiency and good performance of the proposed algorithm. Copyright © 2001 John Wiley & Sons, Ltd.     

基于蚁群算法的含水层参数识别方法

根据渗流场的水头和流量观测数据,建立了基于蚁群算法的地下含水层参数识别方法,含水层参数识别反问题的不适定性由解的不唯一性和不稳定性所表征。与传统的基于梯度的优化方法相比较,对于参数识别反问题蚁群算法能够收敛到全局最优解。为了将蚁群算法引入到参数识别反问题,介绍了一些数值算例,并且将参数识别结果与数值模拟结果进行了比较。研究表明,所提出的参数识别方法具有鲁棒性、全局收敛性和抗观测噪音的能力。     

Implicit integration of a mixed isotropic–kinematic hardening plasticity model for structured clays

In recent years, a number of constitutive models have been proposed to describe mathematically the mechanical response of natural clays. Some of these models are characterized by complex formulations, often leading to non‐trivial problems in their numerical integration in finite elements codes. The paper describes a fully implicit stress‐point algorithm for the numerical integration of a single‐surface mixed isotropic–kinematic hardening plasticity model for structured clays. The formulation of the model stems from a compromise between its capability of reproducing the larger number of features characterizing the behaviour of structured clays and the possibility of developing a robust integration algorithm for its implementation in a finite elements code. The model is characterized by an ellipsoid‐shaped yield function, inside which a stress‐dependent reversible stiffness is accounted for by a non‐linear hyperelastic formulation. The isotropic part of the hardening law extends the standard Cam‐Clay one to include plastic strain‐driven softening due to bond degradation, while the kinematic hardening part controls the evolution of the position of the yield surface in the stress space. The proposed algorithm allows the consistent linearization of the constitutive equations guaranteeing the quadratic rate of asymptotic convergence in the global‐level Newton–Raphson iterative procedure. The accuracy and the convergence properties of the proposed algorithm are evaluated with reference to the numerical simulations of single element tests and the analysis of a typical geotechnical boundary value problem. Copyright © 2007 John Wiley & Sons, Ltd.     

Series Expansion-Based Genetic Inversion of Wireline Logging Data

An evolutionary approach is applied to solve the nonlinear well logging inverse problem. In the framework of the proposed interval inversion method, nuclear, sonic, and laterolog resistivity data measured at an arbitrary depth interval are jointly inverted, where the depth variation of porosity, water saturation, and shale volume is expanded into series using Legendre polynomials as basis functions. In the interval inversion procedure, the series expansion coefficients are estimated by using an adaptive float-encoded genetic algorithm. Since the solution of the inverse problem using traditional linear optimization tools highly depends on the selection of the initial model, a heuristic search is necessary to reduce the initial model dependence of the interval inversion procedure. The genetic inversion strategy used in interval inversion seeks the global extreme of the objective function and provides an estimate of the vertical distribution of petrophysical parameters, even starting the inversion procedure from extremely high distances from the optimum. For a faster computational process, after a couple of thousand generations, the genetic algorithm is replaced by some linear optimization steps. The added advantage of using the Marquardt algorithm is the possibility to characterize the accuracy of the series expansion coefficients and derived petrophysical properties. A Hungarian oil field example demonstrates the feasibility and stability of the improved interval inversion method. As a significance, the genetic inversion method does not require prior knowledge or strong restrictions on the values of petrophysical properties and gives highly reliable estimation results practically independent of the initial model and core information.

     

A new displacement back analysis to identify mechanical geo‐material parameters based on hybrid intelligent methodology

Displacement back analysis is a common method to identify mechanical geo‐material parameters using the monitored displacement. How to obtain a global optimum solution in large space search of highly non‐linear multimodal is a key point of optimum back analysis. The paper presents a new back analysis that is an integration of evolutionary support vector machines (SVMs), numerical analysis and genetic algorithm. The non‐linear relationship between the mechanical geo‐material parameters to be identified and the corresponding displacement values of key points is learned and represented by evolutionary SVMs in global optimum. Numerical analysis is used to create training and testing samples for recognition of SVMs. Then, performing a global optimum search on the obtained SVMs using genetic algorithm can identify the mechanical geo‐material parameters. The proposed algorithm is tested by back analysis of an elastic plate and an elastic–plastic plate and used to recognize mechanical parameters of subclay, strongly weathered tuff and weakly weathered tuff of Bachimen slope, Funing expressway, Fujian, China. The results indicate that applicability of the proposed algorithm with enough accuracy. Copyright © 2004 John Wiley & Sons, Ltd.     

异构并行算法快速构建全球扰动重力梯度全张量图

扰动重力梯度是扰动重力位的二阶导数,相对于其他重力场元素能更多地反映变化的不规则地球产生的高频信息。在使用高阶次球谐系数模型获取大范围高分辨率的扰动重力梯度数据时,存在重复运算多、计算效率低下、耗时较长的问题。针对该问题,推导了简化计算公式,将中间变量提取出来作为全局参数和局部参数单独进行计算、存储,从而有效减少重复运算;并在简化公式的基础上,提出了扰动重力梯度张量快速异构并行算法,利用CUDA（compute unified device architecture）实现了梯度全张量在GPU端的并行计算。根据Txx、Tyy、Tzz三个分量满足Laplace条件验证了算法可靠性,并与传统串行算法进行了计算效率对比,实验结果表明,相较于串行算法,所提算法可减少90%以上计算耗时,可将计算效率提高60倍以上。最后利用该算法基于2 190阶EIGEN6C4模型快速构建了5′×5′分辨率的全球扰动重力梯度全张量图,计算结果显示了扰动重力梯度同地形、地球质量分布变化的相关性及其在全球范围内的数值特征。