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。     

基于旁压试验土体弹塑性本构模型初探

在大量旁压试验数据分析的基础上,通过曲线拟合,建立了旁压试验弹塑性阶段曲线的椭圆方程,利用土体SMP屈服准则和Rowe流动法则,推导出土体塑性阶段应力增量与应变增量间关系矩阵。在弹性阶段,假设土体应力-应变服从广义虎克定律,建立了基于旁压试验的土体弹塑性本构模型。编制了相应的计算程序,将文中模型计算结果与实际旁压试验曲线进行对比,初步验证了模型的正确性。将本构模型编译为ABAQUS自定义材料子程序UMAT,通过有限元对比分析,文中模型计算变形较弹性模型小,较摩尔-库仑模型大,与模型建立的假设一致。基于旁压试验的土体弹塑性本构模型参数少,且易于获取,便于在实际工程中应用。     

考虑颗粒破碎的粗粒土剪胀性统一本构模型

粗粒土作为无黏性散粒状材料具有状态依赖特性,土体的剪切特性受密度和应力水平影响。易破碎是粗粒土的另一个特点,颗粒破碎影响粗粒土的剪胀、内摩擦角、峰值强度和渗透系数。为了能够准确地描述粗粒土的应力-应变关系,采用初始状态参量描述粗粒土的内部状态,根据三轴试验数据建立考虑颗粒破碎耗能的应力-应变关系,采用相关联流动法则推导考虑颗粒破碎的粗粒土剪胀性“统一本构模型”,并建立初始状态参量与模型参数之间的关系。所建立的统一本构模型既考虑了颗粒破碎对剪胀、内摩擦角的影响,又考虑了剪切特性对土体初始状态的依赖。采用变异粒子群算法拟合试验曲线,确定模型参数。模型计算结果能够很好地拟合试验曲线。采用同一组参数对假定的初始状态进行模拟计算,计算结果表明,模型能够模拟不同初始密度和应力水平下粗粒土变形的一般规律。     

基于CPTu的状态相关边界面模型标定研究

对土工结构进行数值分析时,需要引入土的有效应力本构模型,以预测土体变形或液化破坏。然而,由于土单元初始状态的不确定性和缺乏适当的试验结果,模型参数标定经常成为实际工程应用的障碍。在现有边界面模型中引入状态相关的剪胀性概念,使得土体破坏时处于临界状态。基于现场孔压静力触探试验（piezocone penetration test,简称CPTu）的测试数据,提出了模拟单调和循环荷载下土体液化和破坏的边界面模型的参数标定方法。取现场残余强度为室内试验的临界状态,用于标定临界状态线相关参数。基于现有CPTu数据的抗液化强度图,以及震级比例系数与液化等效循环次数之间的联系,建立了土体液化应力比与等效循环次数的关系,在此基础上简要介绍了标定卸载体积模量参数的方法。最后基于CPTu实测地层数据,具体说明了模型相关参数的标定过程。     

无黏性土压缩曲线的一种数学模式

总结和分析了已有考虑状态相关的无黏性土压缩曲线数学模式做了。进行了不同初始密度砂土的等向压缩试验,提出了一个新的状态相关压缩曲线数学模式。该数学模式能够反映无黏性土不同初始状态等向压缩下的应力-应变关系,参数物理意义明确、试验确定方法简单、方便。利用4种不同砂土的等向压缩试验数据,验证了所提出的数学模式的合理性和有效性。该模式可应用于一般应力、应变条件下状态相关本构模型的建立     

上海土体小应变硬化模型整套参数取值方法及工程验证

小应变硬化(HSS)模型能较合理考虑土体小应变阶段的非线性、应力相关等特性,在隧道和基坑开挖等变形数值计算中得到了广泛的应用,但目前对模型整套参数的合理取值尚缺乏系统地研究。剖析了HSS模型各参数的意义和对应的试验方法,并基于大量室内和现场试验数据建立了上海土体HSS模型主要参数与土体孔隙比的经验关系,为实际工程利用勘察报告方便且较准确地确定HSS模型参数提供了合理途径。通过多个深基坑工程的数值分析表明,采用现场试验确定的小应变初始剪切模量时,围护墙水平位移计算值和实测值基本吻合,表明提出的上海土体HSS模型整套参数取值方法有很好的普适性和工程实用价值,也为其他地区建立类似的HSS模型参数取值方法提供了借鉴。     

基于扰动状态理论的软土压缩变形试验

基于扰动状态理论对漳州和东莞软土扰动样和重塑样单向压缩变形的结构性特征进行了分析,利用重塑样试验数据还原了原状样的压缩曲线,将变形中的土体视为相对完整状态和完全调整状态的混合物,定义原状土样为相对完整状态、重塑土样为完全调整状态,且分别用土弹性模型和修正剑桥模型表达,扰动函数与荷载对数具有较好的幂函数关系,因此建立了土体应力应变关系扰动状态模型,并通过压缩试验结果对模型合理性进行了验证。结果表明,该模型能较好地描述土体压缩变形特性,且参数确定简单易行,为软土地基工程沉降分析提供了一定参考。     

基于量子遗传算法与多输出混合核相关向量机的堆石坝材料参数自适应反演研究

为进一步提高堆石坝材料参数反演模型的计算精度与适用性,建立了基于量子遗传算法（QGA）与多输出混合核相关向量机（MMRVM）的自适应反演模型。通过引入混合核函数,使所构建的MMRVM能够高精度地模拟材料参数与大坝沉降间的复杂非线性关系,从而代替耗时较长的有限元（FEM）计算。通过利用参数较固化的QGA优化确定MMRVM核参数,使反演模型具有自适应性。以实测沉降数据为依据,充分发挥QGA的全局搜索能力反演筑坝材料本构模型参数。在分析模型所需测点个数与信噪比对计算结果影响的基础上,通过公伯峡堆石坝应用实例证明：QGA-MMRVM可快速、精确地反演堆石坝筑坝材料本构模型参数,模型凭借其自适应性在实际工程中具有良好的应用前景和推广价值。     

侧限压缩条件下土体结构性参数的单变量模型

通过分析侧限压缩条件下基于应变考虑的土结构性参数和基于应力考虑的土结构性参数,给出了结构性参数的统一表达式,即采用割线模量表示的土体结构性参数。这种表示的积极意义在于避免了初始状态下基于应变考虑的土结构性参数和基于应力考虑的土结构性参数均无法求解的现象。构建了以主动作用为单变量描述的土体结构性参数模型,即基于相同应力方法的土结构性参数模型和基于相同应变方法的土结构性参数模型,并对杨凌黄土的原状样、重塑样和饱和样进行侧限压缩试验。通过试验值和计算值的比较,检验了模型的合理性。研究结果表明,试验值和计算值能够较好吻合。采用主动作用为单变量描述的土体结构性参数直观地反映主动作用对土体结构性的影响,同时简化结构性参数计算过程,实现数值计算。     

《岩土力学》2012年第4期被EI收录论文(47篇,收录率100%)

序号论文题名作者页码1弹塑性变形条件下围岩-支护相互作用全过程解析侯公羽,李晶晶961-9702裂隙岩体冻融损伤关键问题及研究状况刘泉声,康永水,黄兴,等971-9783利用旁压试验的砂土初始状态反演分析黄文雄,     

Identifying parameters controlling soil delayed behaviour from laboratory and in situ pressuremeter testing

The aim of this paper is to present a methodology for identifying the soil parameters controlling the delayed behaviour from laboratory and in situ pressuremeter tests by using an elasto‐viscoplastic model (EVP‐MCC) based on Perzyna's overstress theory and on the elasto‐plastic Modified Cam Clay model. The influence of both the model parameters and the soil permeability was studied under the loading condition of pressuremeter tests by coupling the proposed model equations with Biot's consolidation theory. On the basis of the parametric study, a methodology for identifying model parameters and soil permeability by inverse analysis from three levels of constant strain rate pressuremeter tests was then proposed and applied on tests performed on natural Saint‐Herblain clay. The methodology was validated by comparing the optimized values of soil parameters and the values of the same parameters obtained from laboratory test results, and also by using the identified parameters to simulate other tests on the same samples. The analysis of the drainage condition and the strain rate effect during a pressuremeter test demonstrated the coupled influence of consolidation and viscous effects on the test results. The numerical results also showed that the inverse analysis procedure could successfully determine the parameters controlling the time‐dependent soil behaviour. Copyright © 2008 John Wiley & Sons, Ltd.     

Determination of silica sand stiffness

Site investigation and evaluation of properties of soil or rock are important aspects of geotechnical design. Determination of the ground stiffness is one of the important parameters in geotechnical engineering. Since the measurement of shear modulus is very sensitive to soil disturbance, especially for sand, determination of the stiffness of soil in the field is more reliable than in laboratory tests on sampled specimens. Measurement of shear modulus is one of the most common applications of self-boring pressuremeter testing. As an in situ device, the pressuremeter provides a unique method for assessing directly the in situ shear modulus of a soil. This paper describes a laboratory study of silica sand stiffness, which varies with stress level and strain amplitude. The results show that the elastic shear modulus value is mainly dependent on the value of the mean effective stress and relative density.     

Computer-aided calibration of a soil plasticity model

A non-linear optimization technique based on the quasi-Newton approach is employed to back-calculate certain model parameters of a simple, bounding surface, soil plasticity model from in situ pressuremeter data. The theoretical response corresponding to a given set of parameters is generated by finite element analysis. A semi-analytical procedure is developed for the accurate and efficient evaluation of the gradient of objective function with respect to the model parameters of interest. The BFGS update is used to update the Hessian. Results of a series of numerical experimentation using artificial pressuremeter responses is first reported and discussed. A set of laboratory cavity expansion data is then used to calibrate the constitutive model.     

Forecast of creep settlements of heavy structures using pressuremeter tests

This paper presents a new method proposed for the prediction of long term settlement of very heavy structures based on a numerical interpretation of long term pressuremeter creep tests. The constitutive model used for the modelling of the soil creep is described. The constants of the constitutive model are determined by fitting a simulated curve, given by a simple Finite Element Method, to the results of a long term pressuremeter creep test. The calculation of long term settlement under nuclear power plants, using the soil parameters defined by this method, is presented and compared with the in situ measured values.     

Numerical simulation of cone penetration testing using a new critical state constitutive model for sand

A new perspective on the numerical simulation of cone penetration in sand is presented, based on an enhanced critical state model implemented in an explicit-integration finite element code. Its main advantage, compared to similar studies employing simpler soil models, is that sand compressibility can be described with a single set of model parameters, irrespective of the stress level and the sand relative density. Calibration is based on back-analysis of published centrifuge experiments, while results of the methodology are also compared against independent tests. Additional analyses are performed to investigate sand state effects on cone penetration resistance, in comparison with empirical expressions from the literature.     

Analysis of pressuremeter geometry effects in clay using critical state models

The conventional interpretation methods of pressuremeter testing effectively approximate pressuremeter membranes as infinitely long. As a result, the effects of the two‐dimensional geometry of pressuremeters are ignored, leading to an overestimation of soil shear strength by pressuremeter testing, as demonstrated in several previous studies. This paper presents results of a numerical study of two‐dimensional geometry effects on self‐boring pressuremeter tests in undrained clay. The results are obtained using critical state soil models with an effective stress formulation. This is in contrast to most (if not all) existing studies on pressuremeter geometry effects, which were based on perfectly plastic soil models (e.g. Yu (Cavity expansion theory and its application to the analysis of pressuremeters. DPhil Thesis, The University of Oxford, 1990), Yeung and Carter (Proc. 3rd Int. Symp. on Pressuremeters, 1990), and Houlsby and Carter (Géotechnique, 1993; 43 (4):567–576)). The present study suggests that the overestimation of soil strength due to the neglect of finite pressuremeter length is significantly affected by the soil model used in the calculations. It is found that for clays with a high overconsolidation ratio (OCR) the strength overestimation predicted using critical state soil models could be considerably smaller than that predicted using perfectly plastic soil models. The main conclusion of this numerical study is that care must be exercised before directly applying any numerically determined pressuremeter geometry correction factors in practice. Copyright © 2005 John Wiley & Sons, Ltd.     

Liquefaction and cyclic mobility model for saturated granular media

A new constitutive law for the behaviour of undrained sand subjected to dynamic loading is presented. The proposed model works for small and large strain ranges and incorporates contractive and dilative properties of the sand into the unified numerical scheme. These features allow to correctly predict liquefaction and cyclic mobility phenomena for different initial relative densities of the soil. The model has been calibrated as an element test, by using cyclic simple shear data reported in the literature. For the contractive sand behaviour a well‐known endochronic densification model has been used, whereas a plastic model with a new non‐associative flow rule is applied when the sand tends to dilate. Both dilatancy and flow rule are based on a new state parameter, associated to the stiffness degradation of the material as the shaking goes on. Also, the function that represents the rearrangement memory of the soil takes a zero value when the material dilates, in order to easily model the change in the internal structure. Proceeding along this kind of approach, liquefaction and cyclic mobility are modelled with the same constitutive law, within the framework of a bi‐dimensional FEM coupled algorithm developed in the paper. For calibration purposes, the behaviour of the soil in a cyclic simple shear test has been simulated, in order to estimate the influence of permeability, frequency of loading, and homogeneity of the shear stress field on the laboratory data. Copyright © 2006 John Wiley & Sons, Ltd.     

中主应力系数影响下的冻结砂土损伤本构模型

人工冻结法是饱水砂层开挖过程中常用的止水和临时支护方法,通过冻土损伤特性研究为冻土力学特性和冻结体稳定性分析奠定基础。为研究冻结砂土的损伤力学特性,在-5℃下进行了不同中主应力系数的冻结砂土三维室内试验。从冻土微元破坏服从Weibull随机分布的特点出发,将Drucker-Prager强度准则作为冻土微元统计分布变量,利用应变等价性假说,建立了三维应力状态下冻结砂土损伤本构模型;在此基础上,讨论模型参数F₀与m和中主应力系数的关系,对模型参数进行合理修正,建立中主应力系数影响下的冻结砂土损伤本构模型,并与试验结果进行对比。分析结果表明:参数F₀和m随着中主应力系数的增大呈现先减小后增大的趋势;参数F₀反映了冻结砂土的强度特性,参数m代表了冻结砂土的延性及脆性特征,考虑中主应力系数影响的冻结砂土损伤本构模型能很好地模拟冻结砂土应力-应变全过程曲线。研究成果为人工冻结法工程设计提供一定的理论依据。移动阅读      

Investigation of Soil Conditioning Tests with Three-Dimensional Numerical Modeling

Due to the diverse and complex structure of soil and the variety of foam-modifier materials that are used, it is difficult to provide a model to predict the laboratory behavior of modified soils. For example, several studies have shown independently that the amount of the foam-modified soil depends on several factors, such as the internal friction angle and normal stiffness. Of late, modeling by numerical methods has become popular in engineering sciences and the modeling of complex material behavior is possible with the help of numerical methods. In this research, the performance and efficiency of the numerical method in the modeling of laboratory tests such as the slump test and the uniaxial compressive strength test were investigated and it was found that numerical modeling performs very well in predicting the results of these tests for foam-modified sand samples. In order to achieve this goal, the slump test and the uniaxial compressive strength test were performed in the laboratory on several modified sand samples in order to obtain the laboratory results for these samples. Then, numerical simulation of these experiments was carried out using PFC^3D software. The results of numerical modeling were compared with the experimental results, and good agreement was observed. Finally, after calibration of the numerical model using the experimental results, the effect of changes in the internal friction angle and the normal stiffness of the modified sand in the amount of the slump was investigated. According to the results of this sensitivity analysis, it was determined that by increasing both effective parameters the amount of the slump of foam-modified sand decreases and that the parameters are the most important factors in controlling the slump value.     

Determining soil permeability from pressuremeter tests

This paper presents a methodology for identifying soil permeability from pressuremeter test. On the first part we present a numerical analysis of the permeability effects on the test results. We demonstrate that different drainage conditions arise during test, as a function of the loading rate and the soil permeability. We also studied the pore pressure dissipation during strain holding stages. Based on this analysis of these tests, we propose a general procedure to identify simultaneously mechanical parameters and permeability from pressuremeter tests with strain holding test stages. This procedure was applied on tests performed on natural Saint–Herblain clay. An apparatus called pressio‐triax was developed for this purpose. The values of the mechanical parameters as well as of the permeability value were found to agree very well with the values of the same parameters obtained from conventional laboratory tests. Copyright © 2002 John Wiley & Sons, Ltd.