三维大地电磁自适应正则化有限内存拟牛顿反演

有别于传统基于梯度信息的反演方法在正则化约束中用总梯度逼近海塞逆矩阵的技术,本文将正则化约束问题的数据拟合项和模型光滑项分开考虑,只利用数据拟合函数的梯度信息对数据拟合项的海塞矩阵进行逼近,通过求解类高斯牛顿下降方向方程得到不依赖前几次迭代正则化因子的更精确下降方向,在求解当前迭代下降方向的过程中,通过保证右端项中两个向量的二范数在同一数量级的原则,实现了正则化因子的自动更新.对理论模型的试算表明这种自适应正则化反演方案可以在拟牛顿反演框架下基本达到OCCAM的算法稳定性,反演结果对初始模型依赖性较小,同时又无需在一次迭代中多次搜索最佳正则化因子.本文还基于此算法讨论了大地电磁各参数对于反演结果的影响,由于本文的反演结果能得到充分的正则化约束,因而在此框架下讨论阻抗和倾子在反演中的作用相对更为客观.     

基于正则化方法的高频地波雷达海浪方向谱反演

提出了一种从高频地波雷达海面回波谱中提取海浪方向谱的新方法:基于高频电波海洋探测基本原理将线性化后的非线性积分方程离散为矩阵方程组,通过引入正则化数学模型将不适定方程转化为正规方程,然后采用奇异值分解法求解.对于正则化方法中正则化参数,采用L曲线法来确定.为了验证本文算法的有效性,分别在单部雷达和双部雷达探测下进行了不同噪声水平下的数值模拟,结果表明了该正则化反演方法的有效性.文中还给出了实测数据分析初步结果.该算法的工程性应用还需进一步研究.对于实际的地波雷达海浪反演具有良好的应用前景.     

基于Tikhonov正则化的双频电磁波电导率成像反演

本文将Tikhonov正则化方法与active-set算法相结合,利用双频电磁波电导率成像原理,求解其反演成像方程.不仅对现有算法进行了改进,也促进了算法的实际应用.本文研究了在双频电磁波电导率成像方程建立后,如何根据其严重病态性质,选择合适的算法求解矩阵成像方程.针对电导率非负的特性,引入正则化参数,将问题转化为一个非负最小二乘问题,并用active-set算法求解.采用改进后的迭代算法对理论模型进行了数值模拟计算,验证了该方法的有效性.应用到实际电导率成像反演,与常规的LSQR、SP-LSQR、Tikhonov正则化等算法进行比较,取得了满意的结果.     

利用加速差分进化算法反演非均匀介质电磁成像

本文研究了差分进化算法在地球物理反演中的几种应用.利用双频电磁波电导率成像原理建立成像方程后,根据其严重病态性质,将Tikhonov正则化方法与差分进化算法结合,反演其成像方程.为加速差分进化算法的收敛速度,提出了将种群熵的自适应差分进化(ARDE)算法以及粒子群差分进化混合(PSODE)算法分别与Tikhonov正则化方法结合.在大型反演计算中,这两种方法可以在不影响反演效果的前提下,不同程度地提高收敛速度,降低时间成本.适宜于在正则化参数选取困难情况时的地球物理反演问题的求解.     

时间推移地震反演的连续模型与算法

针对时间推移地震反演问题建立了依赖勘探时间的数学模型,并在此基础上设计了全局收敛的连续型及逐次递归算法.将同伦方法与Tikhonov正则化方法相结合,设计了初次勘探反演的同伦-正则化方法.以三维时间推移地震为例,在连续型及逐次递归算法的基础上构造了快速的局域化反演算法.数值模拟结果说明了上述方法都是有效的.     

基于Bayes算法的Abel积分反演折射指数的研究

本文将Bayes算法应用到Abel积分方程中,利用End-to-End Generic Occultation Performance Simulation and Processing System(EGOPS)软件仿真出的弯角数据来反演大气折射指数,并与Tikhonov正则化的反演结果进行对比.当直接利用仿真所得的弯角数据(认为它是不含误差的)时,Bayes算法与Tikhonov正则化的反演结果具有很好的一致性,二者的均方根误差均为2.5470×10~(-8);但在实际观测过程中,由于电离层以及水汽等方面的影响,所得到的弯角数据不可避免地会含有误差,可能产生高频成分,甚至存在间断的现象,因此我们在仿真的弯角数据中加入了满足高斯型分布的随机噪声,然后进行折射指数的反演试验.结果表明,相比于Tikhonov正则化技术,Bayes算法具有更高的反演精度.     

数据驱动的块排列正则化多道叠前地震反演

叠前地震反演充分利用地震记录振幅随偏移距变化的信息,在储层预测中发挥着重要作用.由于地震观测数据的带限性和其中所含有的噪声,地震反演是一个典型的不适定问题.目前,提出的各种正则化方法通过对反演参数施加约束,使反演结果具备期望的特征.然而,传统的反演方法忽略了地下构造的空间连续性,当地震数据信噪比较低时,单道反演方法无法得到稳定、准确的反演结果.本文提出了一种基于自适应块排列正则化的多道叠前地震反演方法,利用叠后地震资料高信噪比的特点,提取了一个块排列矩阵来记录地下构造信息.该方法通过将地震剖面分解为地震记录块,利用地震数据的局部相似性来记录相邻采样点的位置.在此基础上,构造了多道块排列正则项,建立了多道叠前地震反演的目标函数.在多道块排列正则项中,可以自适应调整权重系数,在增强空间连续性的同时保护边缘.利用L-BFGS算法可以有效地求解多道叠前反演目标函数.合成数据和实际数据测试均表明,当观测记录信噪比较低时,该方法的反演结果优于传统的基于单道模型约束的反演结果.     

储层重力密度反演后验约束正则化方法

本文针对蒸汽辅助重力泄油(SAGD)生产中开发监测问题,发展了综合应用地震及重力数据反演储层密度的联合反演算法.通过测井数据建立纵波阻抗与密度的直接关系,并推导出这种关系下重力与纵波阻抗数据联合反演的计算方法,从而计算出蒸汽腔体密度分布规律.文中应用密度反演后验约束正则化方法,采用Tikhonov正则化模型,通过波阻抗数据作为约束进行联合反演,在算法上提高了稳定性,同时得到较高的反演精度.文中对SAGD生产中的理论模型进行了方法试算,并分析了算法的误差,最终应用于SAGD生产的实际数据中,通过最终反演结果分析,该方法取得了很好的应用效果.     

正则化等效层重力向下延拓方法

近年来,利用时移微重力技术进行储层开发监测受到国内外学者广泛关注.时移微重力观测数据存在信噪比低,信号弱的问题,难以实现储层内物质运移的定量解释.为压制数据噪声,增强有效弱信号,本文研究了利用Tikhonov正则化方法反演等效层（源）,并由等效源实现重力场向下延拓的方法;在此基础上,本文推导了波数域正则化等效源向下延拓算子.针对向下延拓场幅值衰减问题,提出了正则化等效源迭代补偿算法.通过模拟数据实验研究了不同深度正则化等效源滤波算子及向下延拓算子的波数响应;与波数域Tikhonov正则化向下延拓方法相比,正则化等效源向下延拓方法的延拓精度更高、更稳定.最后,将基于迭代补偿的正则化等效源向下延拓技术应用于实测时移微重力数据证实了该方法能够有效增强局部异常,实现时移微重力数据大深度稳定向下延拓.     

基于初至波走时层析成像的Tikhonov正则化与梯度优化算法

初至波走时层析成像是利用地震初至波走时和其传播的射线路径来反演地下介质速度的技术.该问题本质上是一个不适定问题,需要使用正则化方法并辅之以适当的最优化技巧.本文从数值优化的角度介绍了初至波走时层析成像的反演原理,建立了Tikhonov正则化层析成像反演模型并提出求解极小化问题的加权修正步长的梯度下降算法.该方法可以从速度模型的可行域中迭代找到一个最优解.数值试验表明,该方法是可行和有应用前景的.     

Analysis of biodegradation in a 3-D heterogeneous porous medium using nonlinear stochastic finite element method

Probabilistic analysis by Monte Carlo Simulation method (MCSM) is a computationally prohibitive task for a reactive solute transport involving coupled PDEs with nonlinear source/sink terms in 3-D heterogeneous porous media. The perturbation based stochastic finite element method (SFEM) is an attractive alternative method to MCSM as it is computationally efficient and accurate. In the present study SFEM is developed for solving nonlinear reactive solute transport problem in a 3-D heterogeneous medium. Here the solution of the biodegradation problem involving a single solute by a single class of microorganisms coupled with dynamic microbial growth is attempted using this method. The SFEM here produces a second-order accurate solution for the mean and a first-order accurate solution for the standard deviation of concentrations. In this study both the physical parameters (hydraulic conductivity, porosity, dispersivity and diffusion coefficient) and the biological parameters (maximum substrate utilization rate and the coefficient of cell decay) are considered as spatially varying random fields. A comparison between the MCSM and SFEM for the mean and standard deviation of concentration is made for 1-D and 3-D problem. The effects of heterogeneity on the degradation of substrate and growth of biomass concentrations for a range of variances of input parameters are discussed for both 1-D and 3-D problems.     

Using data assimilation method to calibrate a heterogeneous conductivity field and improve solute transport prediction with an unknown contamination source

Hydraulic conductivity distribution and plume initial source condition are two important factors affecting solute transport in heterogeneous media. Since hydraulic conductivity can only be measured at limited locations in a field, its spatial distribution in a complex heterogeneous medium is generally uncertain. In many groundwater contamination sites, transport initial conditions are generally unknown, as plume distributions are available only after the contaminations occurred. In this study, a data assimilation method is developed for calibrating a hydraulic conductivity field and improving solute transport prediction with unknown initial solute source condition. Ensemble Kalman filter (EnKF) is used to update the model parameter (i.e., hydraulic conductivity) and state variables (hydraulic head and solute concentration), when data are available. Two-dimensional numerical experiments are designed to assess the performance of the EnKF method on data assimilation for solute transport prediction. The study results indicate that the EnKF method can significantly improve the estimation of the hydraulic conductivity distribution and solute transport prediction by assimilating hydraulic head measurements with a known solute initial condition. When solute source is unknown, solute prediction by assimilating continuous measurements of solute concentration at a few points in the plume well captures the plume evolution downstream of the measurement points.     

A New Parameter Estimation Method for Solute Transport in a Column

To study contaminant transport in groundwater, an essential requirement is robust and accurate estimation of the transport parameters such as dispersion coefficient. The commonly used inverse error function method (IEFM) may cause unacceptable errors in dispersion coefficient estimation using the breakthrough curves (BTCs) data. We prove that the random error in the measured concentrations, which might be described by a normal distribution, would no longer follow the normal distribution after the IEFM transformation. In this study, we proposed a new method using the weighted least squares method (WLSM) to estimate the dispersion coefficient and velocity of groundwater. The weights were calculated based on the slope of the observed BTCs. We tested the new method against other methods such as genetic algorithm and CXTFIT program and found great agreement. This new method acknowledged different characteristics of solute transport at early, intermediate, and late time stages and divided BTCs into three sections for analysis. The developed method was applied to interpret three column tracer experiments by introducing continuous, constant‐concentration of sodium chloride (NaCl) into columns filled with sand, gravel, and sand‐gravel media. This study showed that IEFM performed well only when the observed data points were located in the linear (intermediate time) section of BTCs; it performed poorly when data points were in the early and late time stages. The new WLSM method, however, performed well for data points scattering over the entire BTCs and appeared promising in parameter estimation for solute transport in a column.     

Uncertainty propagation of hydrodispersive transfer in an aquifer: an illustration of one-dimensional contaminant transport with slug injection

It is evident that the hydrodynamic dispersion coefficient and linear flow velocity dominate solute transport in aquifers. Both of them play important roles characterizing contaminant transport. However, by definition, the parameter of contaminant transport cannot be measured directly. For most problems of contaminant transport, a conceptual model for solute transport generally is established to fit the breakthrough curve obtained from field testing, and then suitable curve matching or the inverse solution of a theoretical model is used to determine the parameter. This study presents a one-dimensional solute transport problem for slug injection. Differential analysis is used to analyze uncertainty propagation, which is described by the variance and mean. The uncertainties of linear velocity and hydrodynamic dispersion coefficient are, respectively, characterized by the second-power and fourth-power of the length scale multiplied by a lumped relationship of variance and covariance of system parameters, i.e. the Peclet number and arrival time of maximum concentration. To validate the applicability for evaluating variance propagation in one-dimensional solute transport, two cases using field data are presented to demonstrate how parametric uncertainty can be caught depending on the manner of sampling.     

An integrated approach for modeling solute transport in streams and canals with applications

A new modeling approach for solute transport in streams and canals was developed to simulate solute dissolution, transport, and decay with continuously migrating sources. The new approach can efficiently handle complicated solute source feeding schemes and initial conditions. Incorporating the finite volume method (FVM) and the ULTIMATE QUICKEST numerical scheme, the new approach is capable of predicting fate and transport of solute that is added to small streams or canals, typically in a continuous fashion. The approach was tested successfully using a hypothetical case, and then applied to an actual field experiment, where linear anionic polyacrylamide (LA-PAM) was applied to an earthen canal. The field experiment was simulated first as a fixed boundary problem using measured concentration data as the boundary condition to test model parameters and sensitivities. The approach was then applied to a moving boundary problem, which included subsequent LA-PAM dissolution, settling to the canal bottom and transport with the flowing canal water. Simulation results showed that the modeling approach developed in this study performed satisfactorily and can be used to simulate a variety of transport problems in streams and canals.     

Modelling soil solute release and transport in run‐off on a loessial slope with and without surface stones

Stone covers on loessial slopes can increase the time of infiltration by slowing the velocity of the overland flow, which reduces the transport of solutes, but few mechanistic models have been tested under water‐scouring conditions. We carried out field experiments to test a previously proposed, physically based model of water and solute transport. The area of soil infiltration was calculated from the uncovered surface area, and Richards' equation and the kinematic wave equation were used to describe water infiltration and flow along slopes with stone covers. The transport of chemicals into the run‐off from the surface soil, presumably by diffusion, and their movement in the soil profile could be described by the convection–diffusion equations of the model. The simulated and measured data correlated well. The stones on the soil surface reduced the area available for infiltration but increased the Manning coefficient, eventually leading to increased water infiltration and decreased solute loss with run‐off. Our results indicated that the traditional model of water movement and solute migration could be used to simulate water transport and solute migration for stone‐covered soil on loessial slopes.     

Estimation of Degradation Rates by Satisfying Mass Balance at the Inlet

Using a steady-state mass conservative solute transport analytical solution that is based on the third-type (or flux-type or Cauchy) source condition, a method is developed to estimate the degradation parameters of solutes in groundwater. Then, the inadequacy of the methods based on the first-type source-based analytical solute transport solution is presented both theoretically and through an example. It is shown that the third-type source analytical solution exactly satisfies the mass balance constraint at the inlet location. It is also shown that the first-type source (or constant source concentration or Dirichlet) solution fails to satisfy the mass balance constraint at the inlet location and the degree of the failure depends on the value of the degradation as well as the flow and solute transport parameters. The error in the first-type source solution is determined with dimensionless parameters by comparing its results with the third-type source solution. Methods for estimating the degradation parameter values that are based on the first-type steady-state solute transport solution may significantly overestimate the degradation parameter values depending on the values of flow and solute transport parameters. It is recommended that the third-type source solution be used in estimating degradation parameters using measured concentrations instead of the first-type source solution.     

MODFLOW/MT3DMS-based simulation of variable-density ground water flow and transport

This paper presents an approach for coupling MODFLOW and MT3DMS for the simulation of variable-density ground water flow. MODFLOW routines were modified to solve a variable-density form of the ground water flow equation in which the density terms are calculated using an equation of state and the simulated MT3DMS solute concentrations. Changes to the MODFLOW and MT3DMS input files were kept to a minimum, and thus existing data files and data files created with most pre- and postprocessors can be used directly with the SEAWAT code. The approach was tested by simulating the Henry problem and two of the saltpool laboratory experiments (low- and high-density cases). For the Henry problem, the simulated results compared well with the steady-state semianalytic solution and also the transient isochlor movement as simulated by a finite-element model. For the saltpool problem, the simulated breakthrough curves compared better with the laboratory measurements for the low-density case than for the high-density case but showed good agreement with the measured salinity isosurfaces for both cases. Results from the test cases presented here indicate that the MODFLOW/MT3DMS approach provides accurate solutions for problems involving variable-density ground water flow and solute transport.     

煤矿井下微震震源高精度定位研究

煤矿井下微震震源准确定位,对于动力灾害监测预警具有重要意义.由于微震震源需要通过井下传感器接收信息反演确定,传感器的安装位置限制于煤矿井下巷道周围,传感器沿巷道近平面的不合理布置将大大降低震源定位精度.针对由传感器信息反演震源位置引起的病态问题,本文提出了基于微震监测测点优化布置的震源高精度定位算法.首先通过计算系数矩阵条件数,判定病态问题;然后利用中心化法和行平衡法联合进行病态矩阵预处理.对预处理后的矩阵A、b利用L曲线法计算正则参数,结合Tikhonov正则化算法计算得到震源坐标正则解.研究结果表明,中心化法有效降低了矩阵数量级,行平衡预处理降低了病态条件数,预处理后Tikhonov正则解的震源坐标误差最小可以达到3.09m,与预处理前的高斯消去解相比误差大大降低.通过上述优化处理,实现了井下受限空间微震监测震源高精度定位.     

Modeling contaminant transport in homogeneous porous media with fractional advection-dispersion equation

The newly developed Fractional Advection-Dispersion Equation (FADE), which is FADE was extended and used in this paper for modelling adsorbing contaminant transport by adding an adsorbing term. A parameter estimation method and its corresponding FORTRAN based program named FADEMain were developed on the basis of Nonlinear Least Square Algorithm and the analytical solution for one-dimensional FADE under the conditions of step input and steady state flow. Data sets of adsorbing contaminants Cd and NH4+-N transport in short homogeneous soil columns and conservative solute NaCI transport in a long homogeneous soil column, respectively were used to estimate the transport parameters both by FADEMain and the advection-dispersion equation (ADE) based program CXTFIT2.1. Results indicated that the concentration simulated by FADE agreed well with the measured data. Compared to the ADE model, FADE can provide better simulation for the concentration in the initial lower concentration part and the late higher concentration part of the breakthrough curves for both adsorbing contaminants. The dispersion coefficients for ADE were from 0.13 to 7.06 cm2/min, while the dispersion coefficients for FADE ranged from 0.119 to 3.05 cm1.856/min for NaCI transport in the long homogeneous soil column. We found that the dispersion coefficient of FADE increased with the transport distance, and the relationship between them can be quantified with an exponential function. Less scale-dependent was also found for the dispersion coefficient of FADE with respect to ADE.