应用并行PEST算法优化地下水模型参数

基于列文伯格-马夸尔特（Levenberg-Marquardt）算法的PEST参数优化程序具有寻优速度快、健壮性好的优点,在地下水模型参数优化研究中有许多成功的应用实例。但是,对于大尺度、高精度和高复杂性的大规模地下水模拟,使用PEST进行参数优化需要大量的计算时间,优化效率较低。本文应用OpenMP并行编程方法对PEST算法进行了并行化,使之可以在共享存储并行计算机上进行参数优化的并行计算。并将此方法应用于甘肃北山区域地下水模型的参数优化中,并行实验表明,使用并行化的PEST可以将地下水模型参数优化效率提高3.7倍。      

地下水数值计算与应用研究进展综述

地下水数值计算是研究分析地下水各种问题的重要手段。文中对近年来地下水数值模拟计算相关方法研究进展、地下水数值模拟工作程序的方法论、地下水数值模拟的参数反演方法、国际流行的地下水数值模型及软件平台、目前国内地下水数值模型的相关应用等几个方面进行了综述,力图从相对全面的角度认识当前地下水计算技术的发展。在回顾地下水以上研究进展的基础上,对现代信息技术在地下水数值计算技术中的积极促进作用、地下水更深入和复杂规律的研究、地下水研究与其他学科之间的综合集成和协同互补等发展趋势进行了探讨,为相关研究人员提供一定的参考。     

iTOUGH2反演模型在地下水模拟中的应用

模型参数快速校准是地下水数值模型应用中非常重要和困难的工作,反演模型提高了该工作的效率。在阐述i TOUGH2反演模型流程和TOUGH2/EOS9模块原理的基础上,介绍了i TOUGH2关于敏感性分析、参数估计和不确定性分析的原理。以FEFLOW软件中含三个观测井信息的Breyell抽水试验为例,利用i TOUGH2进行了敏感性分析、参数估计和不确定性分析,发现渗透率和孔隙度为敏感参数,而且通过加权最小二乘法的目标函数和优化算法获得了参数的估计值,最后使用一次二阶矩法和蒙特卡罗法分别进行了不确定性分析。将i TOUGH2反演的参数估计值与FEFLOW反演结果对比,发现两者接近,说明i TOUGH2的反演结果是可靠的。i TOUGH2对应的功能全,包括参数反演,敏感性分析,不确定性分析,可作为地下水模型反演模型的选择之一。     

VG模型参数选择对咸水层CO_2封存的影响研究

在使用TOUGH2程序模拟咸水层CO2封存时,常使用VG模型计算岩石表面毛细压力,而岩石表面毛细压力的大小则会直接影响CO2在咸水层中运移形式和储存量。总结了不同类型常见砂泥岩的VG模型参数取值范围,基于松辽盆地典型储盖地质结构特征,运用TOUGH2构建了二维CO2封存模型,设计3组算例对比分析了VG参数取值对咸水层CO2封存的影响。研究结果表明:VG模型参数值大小取决于岩石的性质,TOUGH2使用的VG模型缺省值仅适用于小部分岩石;盖层的3个VG模型参数取值对咸水层CO2封存效果影响很大,单纯微调部分VG参数仍不能提高模拟可信度,而对于高、中渗透率储层,VG模型参数取值对模拟结果影响较小。因此,咸水层CO2封存模拟时,应结合储盖层岩性特征赋VG模型参数值。     

基于地下水数值模拟的地球物理反演解释优化方法研究

地球物理的多解性无处不在,面对这种情况我们往往结合地质资料来对反演结果进行合理的地质解释。然而在偏远地区地质、深部钻探资料非常缺乏,野外已有的井孔也往往难以获取岩性编录资料。当这些资料比较少时,反演结果依然存在难以解释或者不确定等问题。是否可以借助于地下水位与水质信息包含的地质体水力特性,来辅助地球物理反演过程？基于此思路探索了基于地下水模型的地球物理反演解释优化方法:以大地电磁法为实现手段,通过研究已知理想地质模型和多个基于反演结果解释得到的地质模型,建立地下水数值模型,对比地下水水流和溶质运移数值模拟结果,最终确定相对最佳地质模型。并通过敏感性分析验证了运用地下水数值模拟搜索相对最佳地质模型的可行性。     

模拟退火算法反演水文地质参数算例研究

基于模拟退火算法(SA)这一全局优化技术,耦合地下水模拟的有限元模型,给出水文地质参数的反演方法.并用水文地质参数反演算例来验证反演模型的可用性和适用性.结果表明,用"变尺度"产生新的渗透系数反演结果明显优于"随机法",当初始温度在500～650、降温系数在0.92～O.95时,模拟退火法在正常情况获得了很好的反演结果,在考虑减少观测孔个数和观测水头值有误差的情况下,也取得了较为满意的反演结果.为了提高运算效率,可考虑与传统反演方法或其它人工智能算法相结合.     

四川南江县窑厂坪滑坡地下水数值模拟

地下水流场的分布是评价滑坡体稳定性的关键因素。使用Visual MODFLOW地下水数值模拟软件,建立窑厂坪滑坡在天然、暴雨和排水条件下的三个模型。通过参数反演提高天然模型的精度。分析暴雨和排水模型表明,强降雨抬升滑坡区地下水位,并在部分区域形成雍高水位,而廊道地下排水工程降低地下水水位,工程效果明显。     

基于实码遗传算法的地下水污染物运移参数反演

介绍了采用实数编码遗传算法优化反演地下水污染物运移参数的基本步骤,不仅求解了弥散系数,而且反演得到了地下水运动速度。算例表明,采用遗传算法———特征有限元离散方法反演地下水污染物运移参数是可行的。     

基于基坑降水优化设计的地面沉降数值模拟

深基坑降水引起的地面沉降给工程和周边环境带来很大的危害,以天津地铁 XX站深基坑为例,通过使用有限差分模拟软件Processing Modflow建立三维地下水渗流场与地面沉降耦合模型,利用抽水试验数据反演识别场区的水文地质参数,对基坑降水对周边环境的影响进行预测。结合施工要求,本着在规定时间内用最少的抽水量完成降水要求,并对基坑周围造成的环境影响最小的原则,对基坑降水进行优化设计。     

基于粒子群算法优化小波支持向量机的岩土力学参数反演

常用的确定岩土力学参数的方法有原位测试和室内试验两种,但都存在一定的局限性,参数选择的合理与否,对设计计算及数值模拟分析结果的有效性影响很大。支持向量机法在理论基础和求解算法方面都具有明显优势,为确保岩土力学参数取值的合理性,采用支持向量机法对岩土力学参数进行反演。先通过小波分析理论构造出支持向量机的核函数,再用粒子群算法(PSO)分别优化Morlet小波、Mexico小波和RBF函数的支持向量机模型参数,通过小波支持向量机模型建立反演参数与沉降值间的非线性映射关系。根据正交试验和均匀试验对需反演的岩土力学参数进行设计,结合有限元软件进行计算分析,得到学习样本和测试样本。分别采用Morlet小波、Mexico小波和RBF函数得出的预测结果和原始数据进行对比分析,发现采用Morlet小波核函数预测效果更佳。使用Morlet小波核函数预测的参数输入到Midas模型中计算建筑物最终沉降量,比较计算值与实际监测值,其相对误差不超过8.1%。研究结果表明,该方法在岩土工程参数的反演中具有良好的应用价值,对今后岩土力学参数的确定及校核提供了一种新方法。     

新安江模型参数的线性化率定

为了解决一般概念性水文模型参数率定结果不稳定的问题,以新安江模型为例,提出了新安江模型日模参数的线性化率定方法。首先通过理想模型将该方法与SCE-UA方法及单纯形方法进行对比研究。率定结果中3种算法所获得的平均目标函数值分别为0.02、0.10、8.39 m³/s,平均循环次数分别为8、637、327,而且由线性化率定方法所获得的各参数值方差也要比其他2种方法小得多;说明线性化率定方法能够找到参数真值,计算精度和效率更高,率定结果更稳定。然后采用建阳和长滩河2个流域的实测资料对该方法进行应用检验,结果表明,同样可以较快地率定出稳定的模型参数优值。建阳流域的10组率定结果的目标函数值皆为100.35 m³/s,循环次数也皆在8次以内,而且2个流域检验期的径流深相对误差皆在9.68%以内,确定性系数皆在0.819以上。因此,线性化参数率定方法确实能够解决非线性模型参数率定结果不稳定的问题,不会产生不相关的局部参数优值,并且不受参数初值影响,计算精度高,循环次数少,是一种可行有效的全局优值参数优选方法。     

The role of observations in the inverse analysis of landslide propagation

Model calibration is usually based on trial-and-error procedures that, in turn, rely on expert judgment or previously acquired experiences for similar phenomena. Efficient and reliable procedures for model calibration of the propagation stage of landslides are still needed. This paper addresses this issue by proposing an inverse analysis procedure and applying it to the case history of a short run-out landslide triggered by a rising perched water table after a heavy rainfall. It focuses on the key role played by the field observations used to set up the inverse analysis, and evaluating the reliability of the numerical simulations. It also investigates the effect of different types of optimization parameters on the inverse analysis results, referring to a mixed-phase model or to a two-phase model for the propagating soil. Several sets of observations are used; all of them refer to the soil deposit thickness at the end of propagation, but differ in both location and number of the adopted values. The numerical analysis of the case history is performed through the academic “GeoFlow_SPH” model, and model calibration by inverse analysis is conducted using the “UCODE” software. The results obtained are discussed with the aim to provide practical criteria to identify the minimum amount of information required for a satisfactory model calibration.     

水文模型参数自动优选方法的比较分析

模型参数的识别是模糊研制与应用成功与否的关键。介绍了三个自动优选模型参数的方法，以新安江模型为例，应用１４个流域的资料，对罗森布郎克法、改进的单纯形法和基因算法优算法优选模型参数的效果，优化方法和收敛速度及参数初值对优选效果的影响进行了比较分析。     

Simulation-based calibration of geotechnical parameters using parallel hybrid moving boundary particle swarm optimization

Simulation-based optimization methods have been recently proposed for calibrating geotechnical models from laboratory and field tests. In these methods, geotechnical parameters are identified by matching model predictions to experimental data, i.e. by minimizing an objective function that measures the difference between the two. Expensive computational models, such as finite difference or finite element models are often required to simulate laboratory or field geotechnical tests. In such cases, simulation-based optimization might prove demanding since every evaluation of the objective function requires a new model simulation until the optimum set of parameter values is achieved. This paper introduces a novel simulation-based “hybrid moving boundary particle swarm optimization” (hmPSO) algorithm that enables calibration of geotechnical models from laboratory or field data. The hmPSO has proven effective in searching for model parameter values and, unlike other optimization methods, does not require information about the gradient of the objective function. Serial and parallel implementations of hmPSO have been validated in this work against a number of benchmarks, including numerical tests, and a challenging geotechnical problem consisting of the calibration of a water infiltration model for unsaturated soils. The latter application demonstrates the potential of hmPSO for interpreting laboratory and field tests as well as a tool for general back-analysis of geotechnical case studies.     

缺资料流域水文模型参数区域化研究进展

缺资料流域由于缺乏历史径流资料无法进行水文模型参数率定,因此模型参数识别具有很大的难度和不确定性。目前国内外学者对缺资料流域水文模型参数识别一般采用区域化方法,即通过某种途径,利用有资料流域的模型参数推求缺资料流域的模型参数,从而对缺资料流域进行预报。文章总结分析了缺资料流域水文模型参数区域化方法中的参数移植法和回归法的研究进展,对存在的方法选择问题、尺度问题等进行了讨论,并指出在参数不确定性、尺度转换以及多种信息源利用等方面还有待于进一步研究。     

基于变动饱和带的产汇流模型及其参数确定方法

当前分布式水文模型的参数确定仍主要依赖率定方式,在缺资料地区应用受到限制。建立一种基于变动饱和带产流模式和网格水滴汇流方法的分布式产汇流模型,提出利用下垫面特征来确定模型参数的方法。结合野外入渗试验和参数敏感性分析,建立地表饱和水力传导度(K_0z)和饱和水力传导度随深度衰减系数(f)2个敏感性产流参数与地形参数、土壤类型数据的定量统计关系,利用野外坡面流观测试验确定坡面汇流参数,并在多个实际流域进行应用验证。结果表明：(1)利用地形参数确定K_0z与使用遥感资料确定K_0z的模型精度进行对比,在姜湾实验流域场次洪水模拟的平均确定性系数从0.82提高至0.86,洪峰与洪量误差的平均绝对值分别降低了2.2%和0.95%,但峰现时间误差平均绝对值增大了4%(仍控制在2 h内)。(2)建立姜湾等14个流域参数f率定值与不同深度土壤类型数据的定量关系,移用至七邻等6个流域进行验证,表明参数f关系式与模型率定的精度非常接近,相对误差的平均绝对值为2.8%,场次洪水模拟的平均确定性系数为0.83,洪峰与洪量误差的平均绝对值为10.07%和...     

断裂韧度试样CCNBD宽范围应力强度因子标定

国际岩石力学学会（ISRM）在1995年提出了一种新型岩石断裂韧度试样--人字形切槽巴西圆盘试样CCNBD,但是,其断裂韧度计算公式中的重要参数（即无量纲应力强度因子的标定）仍存在问题。采用一种新的分片合成方法并结合有限元法,参照ISRM给出的CCNBD试样的尺寸限制,对该范围试样的应力强度因子进行了宽范围的标定,以便在试验中能因地制宜地选用不同几何参数的CCNBD试样。结果表明：分片合成方法的计算值有很高的精度,不但减少了工作量,也使标定的无量纲应力强度因子比现有文献值更加准确、可靠。     

Experience in calibrating the double‐hardening constitutive model Monot

The Monot double‐hardening soil model has previously been implemented within a general purpose finite element algorithm, and used in the analysis of numerous practical problems. This paper reviews experience gained in calibrating Monot to laboratory data and demonstrates how the calibration process may be simplified without detriment to the range of behaviours modelled. It describes Monot's principal features, important governing equations and various calibration methods, including strategies for overconsolidated, cemented and cohesive soils. Based on a critical review of over 30 previous Monot calibrations, for sands and other geomaterials, trends in parameter values have been identified, enabling parameters to be categorized according to their relative importance. It is shown that, for most practical purposes, a maximum of only 5 parameters is needed; for the remaining parameters, standard default values are suggested. Hence, the advanced stress–strain modelling offered by Monot is attainable with a similar number of parameters as would be needed for some simpler, less versatile, models. Copyright © 2003 John Wiley & Sons, Ltd.     

Inverse modeling of natural tracer transport in a granite massif with lumped-parameter and physically based models: case study of a tunnel in Czechia

This study deals with numerical modelling of hydraulic and transport phenomena in granite of the Bohemian massif in Bedrichov, Czechia (Czech Republic). Natural tracers represented by stable isotopes δ¹⁸O and δ²H were collected at the tunnel outflow points and nearby catchment and their concentrations were monitored for seven years. The study compared transport simulations by a two-dimensional (2D) physically based model (advection-dispersion) developed in Flow123d software and a simpler lumped-parameter model, calculated with FLOWPC. Both variants were calibrated with UCODE software, either fitting the concentration data alone, or including the tunnel inflow rates in the case of the 2D model calibration (either in separate steps or within a single optimization problem). Since each of the models describes the tracer transport with different parameters, the models were compared based on the mean transit time as a postprocessed quantity. Besides this, two different options for processing the recharge data (input for both models) were evaluated. Calibration and data interpretation were possible for three of the four observed places in the tunnel, thus determining the depth limit of applicability of the stable isotopes. The estimates for discharge sampling at 25–35 m depth based on inverse modelling provide reasonable values of mean transit time (20–40 months) for the lumped parameter models, little revising the results of previous studies at the site. The resulting transport parameters of the advection-dispersion model (porosity and dispersivity) are in accordance with the hydrogeological structures present at the sampling sites.

     

r.massmov: an open-source landslide model for dynamic early warning systems

This paper illustrates the main characteristics of the newly developed landslide model r.massmov, which is based on the shallow water equations, and is capable of simulating the landslide propagation over complex topographies. The model is the result of the reimplementation of the MassMov2D into the free and open-source GRASS GIS with a series of enhancements aiming at allowing its possible integration into innovative early warning monitoring systems and specifically into Web processing services. These improvements, finalized at significantly reducing computational times, include the introduction of a new automatic stopping criterion, fluidization process algorithm, and the parallel computing. Moreover, the results of multi-spatial resolution analysis conducted on a real case study located in the southern Switzerland are presented. In particular, this analysis, composed by a sensitivity analysis and calibration process, allowed to evaluate the model capabilities in simulating the phenomenon at different input data resolution. The results illustrate that the introduced modifications lead to important reductions in the computational time (more than 90 % faster) and that, using the lower dataset resolution capable of guaranteeing reliable results, the model can be run in about 1 s instead of the 3.5 h required by previous model with not optimized dataset resolution. Aside, the results of the research are a series of new GRASS GIS modules for conducting sensitivity analysis and for calibration. The latter integrates the automated calibration program “UCODE” with any GRASS raster module. Finally, the research workflow presented in this paper illustrates a best practice in applying r.massmov in real case applications.