A comparison of preconditioners for incompressible Navier–Stokes solvers

We consider solution methods for large systems of linear equations that arise from the finite element discretization of the incompressible Navier–Stokes equations. These systems are of the so‐called saddle point type, which means that there is a large block of zeros on the main diagonal. To solve these types of systems efficiently, several block preconditioners have been published. These types of preconditioners require adaptation of standard finite element packages. The alternative is to apply a standard ILU preconditioner in combination with a suitable renumbering of unknowns. We introduce a reordering technique for the degrees of freedom that makes the application of ILU relatively fast. We compare the performance of this technique with some block preconditioners. The performance appears to depend on grid size, Reynolds number and quality of the mesh. For medium‐sized problems, which are of practical interest, we show that the reordering technique is competitive with the block preconditioners. Its simple implementation makes it worthwhile to implement it in the standard finite element method software. Copyright © 2007 John Wiley & Sons, Ltd.     

Sequential Upscaling Method

This paper presents a new method for scaling up multiphase flow properties which properly accounts for boundary conditions on the upscaled cell. The scale-up proposed does not require the simulation of a complete finely-gridded model, instead it calls for assumptions allowing the calculation of the boundary conditions related to each block being scaled up. To upscale a coarse block, we have to assume or determine the proper boundary conditions for that coarse block. To date, most scale-up methods have been based on the assumption of steady-state flow associated with uniform fractional flows over all the boundaries of the coarse block. However, such an assumption is not strictly valid when we consider heterogeneities. The concept of injection tubes is introduced: these are hypothetical streamtubes connecting the injection wellbore to all inlet faces of the fine grid cells constituting the block to be scaled up. Injection tubes allow the capturing of the fine-scale flow behavior of a finely-gridded model at the inlet face of the coarse block without having to simulate that fine grid. We describe how to scale up an entire finely-gridded model sequentially using injection tubes to determine the boundary conditions for two-phase flow. This new scale-up method is able to capture almost exactly the fine-scale two-phase flow behavior, such as saturation distributions, inside each isolated coarse-grid domain. Further, the resultant scaled-up relative permeabilities reproduce accurately the spatially-averaged performance of the finely-gridded model throughout the simulation period. The method has been shown to be applicable not only to viscous-dominated flow but also to flow affected by gravity for reasonable viscous-to-gravity ratios.     

A multiblock/multilevel mesh refinement procedure for CFD computations

A multiblock/multilevel algorithm with local refinement for general two‐ and three‐dimensional fluid flow is presented. The patched‐based local refinement procedure is presented in detail and algorithmic implementations are also presented. The multiblock implementation is essentially block‐unstructured, i.e. each block having its own local curvilinear co‐ordinate system. Refined grid patches can be put anywhere in the computational domain and can extend across block boundaries. To simplify the implementation, while still maintaining sufficient generality, the refinement is restricted to a refinement of the grid successively halving the grid size within a selected patch. The multiblock approach is implemented within the framework of the well‐known SIMPLE solution strategy. Computational experiments showing the effect of using the multilevel solution procedure are presented for a sample elliptic problem and a few benchmark problems of computational fluid dynamics (CFD). Copyright © 2001 John Wiley & Sons, Ltd.     

CFD多块网格生成新进展

网格生成是计算流体力学的重要组成部分,多块网格在CFD实践中获得了广泛的应用．结合对网格生成技术规范和网格生成系统的讨论,综述了多块网格近年来的新进展,重点评述了网格拓扑和网格拼接技术（包括所谓的连续拼接、非结构拼接和广义拼接）,完整飞机外形多块网格生成策略,自动分块技术以及相应的块合井技术,CAD和CFD之间的数据交换技术和基于NURBS的曲面网格生成技术,网格质量分析和控制技术,若干网格生成新方法,以及多块网格在航空气动力数值模拟中的应用．      

Simulation of Gas Dipole Tests in Fractures at the Intermediate Scale Using a New Upscaling Method

A high-level radioactive waste disposal site may lead to gas generation by different physical mechanisms. As these sites are to be located in areas with low water flow, any small amount of gas can lead to relative high gas pressures, so that multiphase flow analysis becomes relevant. The movement of gas and water through the system has two important implications. Firstly, water flow takes place in unsaturated conditions, and thus travel times of the radioactive particles transported are affected; and secondly, gas can also carry radioactive particles. Therefore, one of the key points in such studies is the time when gas would break through the biosphere under a number of different flow conditions. In fractured zones, gas would flow preferentially through the most conductive features. We consider a two-dimensional system representing an isolated fracture. In each point we assign a local porosity and permeability and a local pressure-saturation relationship. A dipole (injector-producer) gas flow system is generated and the variation in water saturation is studied. A simple method is proposed for obtaining upscaled values for several parameters involved in two-phase flow. It is based on numerical simulation on a block scale assuming steady-state conditions and absence of capillary pressure gradients. The proposed method of upscaling is applied to simulate a dipole test using a coarser grid than that of the reference field. The comparison between the results in both scales shows an encouraging agreement.     

Influence of capillary pressure,adsorption and dispersion on chemical flood transport phenomena

This is the second of two joint papers which study the influence of several physical properties on the transport phenomena in chemical flooding. To that aim, we use a previously reported ternary two-phase model into which representative physical properties have been incorporated as concentration-dependent functions. Physical properties such as phase behavior, interfacial tensions, residual saturations, relative permeabilities, phase viscosities and wettability have been analyzed in the first paper.

In this paper, we discuss the influence of capillary pressure, adsorption of the chemical component onto the rock and dispersion. Although arising from different phenomenological sources, these transport mechanisms show some similar effects on concentration profiles and on oil recovery. They are studied for systems with different phase behavior. A numerical analysis is also presented in order to determine the relevance of the number of grid blocks taken in the discretization of the differential equations. This numerical analysis provides useful guidelines for the selection of the appropriate numerical grid in each type of displacement.

     

Improved Lattice Boltzmann Models for Precipitation and Dissolution

A challenge when modeling mineral growth inside the pore space of a porous media is to minimize the effect of the computational grid on the shape of the minerals being formed. Pore surface area and volume are important quantities in estimating upscaled permeability and effective rate equations, which emphasize the importance of models that minimize or completely eliminate grid effects. In this paper, we study how the initial orientation of the solid structure on the numerical grid affects the growth pattern due to precipitation in a lattice Boltzmann model. We have implemented a volume of fluid method to represent the solid interface, and we introduce a surface tension term that extensively reduces the dependency on the underlying numerical grid. We study both diffusion-limited and reaction-limited precipitation. In the diffusion-limited case, instabilities will develop on small scales. The surface tension term effectively introduces a short wavelength cut off which limits the unstable precipitation and reduces grid effects. We argue that the surface tension term is needed to obtain a growth pattern independent of the initial orientation on the underlying grid in the diffusion-limited case, and that simpler models can be used in the reaction-limited case.     

两级分块结构化网格在SK型静态混合器中的应用

给出一种适用于SK型静态混合器流场数值模拟的两级分块结构化网格划分方案和分块圆弧区点的极坐标方程。与常用的非结构四面体网格划分对比,两种方法划分网格数量相近时,两级分块结构化网格达到的迭代精度高于非结构化网格一个等级,而非结构化网格收敛所用时间约为结构化网格收敛所用时间的1.25倍;运用控制体积容积变化值对两种方法划分网格进行质量评估,结构化网格划分方法接近99%网格质量指标在1.0~2.0范围内,而非结构化网格仅接近60%网格质量指标在1.0~2.0范围内。通过实验测量对两种方法划分网格数值模拟结果进行对比验证,结构化网格计算结果与实验值更为接近,表明其可行性和准确性均优于非结构四面体网格划分。     

Grid Orientation Revisited: Near-well, Early-time Effects and Solution Coupling Methods

The grid orientation effect is a phenomenon which leads to the computation of fundamentally different solutions on grids oriented diagonal and parallel to the principal flow direction. Grid orientation remains an important consideration for many practical simulation studies, and renewed interest in gas injection processes motivates the revisiting of this classical problem. In this article, we show that there are aspects of the grid orientation effect that can be traced back directly to the treatment of early-time, near-well flow and therefore have a major impact on adverse mobility ratio displacements such as miscible or immiscible gas injection. The details of this effect mean that any uncertainty quantification study should account for the interaction of the near-well heterogeneity and the grid orientation effect. We also show how two possible methods—a well-sponge method and a local embedding technique—are able to produce a solution largely independent of grid orientation for single phase two-component miscible flow. Both methods are versatile in that they can be implemented on general grid topologies. They are illustrated on Cartesian grids for both the standard quarter five spot problem with two different grid orientations, and a problem with a single injection well and two producing wells at different angles to the grid lines. Our results show that it is possible to reduce grid-orientation effects for challenging adverse mobility ratio miscible displacements with only local treatments around the injection wells.     

Estimates of accuracy of approximate solutions of some combined modular elastic continuum systems with discrete interaction

At present, it happens increasingly often that one needs to study complicated modular systems with discrete interaction between the units. If a system contains a distributed unit, then it is said to be combined. One method for studying such systems is to analyze their frequency models; in this case, it is sometimes required to calculate the Green functions of the distributed unit and study the block structures generated by a model of a system with superelements. A technique for studying such systems was developed in [1] and is called the factorized perturbation method; in foreign literature, this technique is called the Green function method (e.g., see [2]).The main idea of such methods is first to construct an equivalent system in the characteristic space that arises in the discrete macrostructure of the original system as a result of the discrete interaction between the units of the latter and then to reduce the characteristic solution to the original solution by using a simple transfer relation. A typical characteristic of structural methods is their algorithmic universality, which is independent of the special form of the discrete microstructure and requires solving a characteristic matrix equation. To obtain approximate solutions by these methods, one should approximately calculate the Green functions of the distributed units and explicitly indicate their operating frequency ranges ( see [3, 4]). The concept of structural methods is to obtain a system solution by using a priori studied elements.There are a variety of methods for studying combined modular systems. In view of this, the accuracy of the approximate solutions, which are also obtained by numerous methods, should be estimated from the algorithmic standpoint. At present, this is done at the level of practical-empirical considerations like doubling the number of grid points for the analysis of the distributed unit, taking into account additional vibration modes in the Green functions, etc.In the present paper, we describe a structural method for solving elastic one-dimensional distributed systems with discrete interaction and present an efficient scalar a priori estimate and a universal a posteriori estimate for the accuracy of an approximate solution, which can be used in many approximate methods, e.g., FEM [5].     

Experimental displacement patterns in a 2 × 2 quadrant block with permeability and wettability heterogeneities—problems for numerical modelling

The effect of heterogeneities on miscible and immiscible flood displacements in 2D bead packs in quadrant form, 2 × 2 block heterogeneity, with either a permeability or a wettability contrast is the subject of this paper. The physical processes occurring during miscible and immiscible flow and displacement within permeability and wettability quadrant bead pack models have been studied experimentally. This geometry occurs in a number of situations relevant to hydrocarbon production: particularly faults where adjacent rocks have large permeability contrasts with rapid changes, in the laboratory with core butting, in reservoir simulation where grid blocks have different permeability and in reservoirs having near-wellbore damage problems. The model quadrants 1–4, had 1 and 4 and 2 and 3 with identical properties, either in permeability or wettability. Reported are complete unit mobility miscible displacements, then the effects of viscosity differences (mobility modifiers) and finally immiscible displacements on displacement patterns for initial linear injection. The experiments demonstrate that nodal flow occurs for both miscible and immiscible flow, but for immiscible flow there are boundary effects due to capillary pressure differences created by water saturation changes or wettability contrasts which can leave patches of isolated fluid within a quadrant. The displacement patterns for the different models and fluids change significantly with the viscosity and wettability changes, particularly for the immiscible displacements. This is due to the changing capillary pressure between the quadrant blocks as the water saturation change. These are difficult to address in numerical modelling but should be accounted for. Other effects include coupling of all physical processes governing the flow through the node and creations of microzones of trapped residual oil. Our displacement patterns can therefore be a valuable verification benchmark tool for numerical modelling and a calibration data source for those wishing to simulate the effects of capillary pressure under differing wettability conditions and for those investigating upscaling modelling procedures. However, the possible loss of physical reality when averaging must always be considered.     

Mechanisms of prestressed reticulate systems with unilateral stiffened components

The stability of space reticulate systems is dependent on the existence of mechanisms. The methods that have been developed to determine them are mainly based on the calculation of a basis of the mechanisms' vectorial subspace by computing the kernel of the transpose equilibrium matrix of the structure. However, they can only consider a bilateral stiffness of the members, which applies to the case for systems composed of bars with traction and compression stiffness. Nevertheless, some classes of reticulate systems, like tensegrity systems, use unilateral rigidity components such as cables. The objective of this paper is to develop a method for calculating the mechanisms which can take into consideration the presence of components with unilateral rigidity. In this case, specific mechanisms associated with these elements may appear; these are referred to as “unilateral mechanisms”. An approach is therefore proposed in order to write a basis of their vectorial subspace. It is included in a methodology devoted to the analysis of space structures with initial stresses. The process is based on the identification of the possible prestress states and of the bilateral mechanisms and, next, to the characterization of the unilateral mechanisms.     

一种快速稳健的并行多块结构动网格方法

为解决传统网格处理方法不能满足复杂外形在大设计空间内进行优化时对网格质量的要求的问题,提出了一种并行多块动网格方法,该方法基于初始外形的多块结构网格,根据优化过程中个体外形与初始外形拓扑结构相近的特性,利用体样条插值方法来拟合多块结构网格各块顶点的位移,得到几何外形变化后的拓扑结构,再利用无限插值方法并行地移动初始外形多块结构网格的边、面和块内的网格点,进行光顺处理后得到变形后几何外形的空间网格;该方法在保证网格质量的同时,可以极大地提高网格生成效率,本文以某翼身组合体为例结果表明,该方法在大设计空间的复杂外形设计问题中具有很强的实用性。     

Global convergence of inexact Newton methods for transonic flow

In computational fluid dynamics, non-linear differential equations are essential to represent important effects such as shock waves in transonic flow. Discretized versions of these non-linear equations are solved using iterative methods. In this paper an inexact Newton method using the GMRES algorithm of Saad and Schultz is examined in the context of the full potential equation of aerodynamics. In this setting, reliable and efficient convergence of Newton methods is difficult to achieve. A poor initial solution guess often leads to divergence or very slow convergence. This paper examines several possible solutions to these problems, including a standard local damping strategy for Newton's method and two continuation methods, one of which utilizes interpolation from a coarse grid solution to obtain the initial guess on a finer grid. It is shown that the continuation methods can be used to augment the local damping strategy to achieve convergence for difficult transonic flow problems. These include simple wings with shock waves as well as problems involving engine power effects. These latter cases are modelled using the assumption that each exhaust plume is isentropic but has a different total pressure and/or temperature than the freestream.     

Simulation of Richtmyer–Meshkov instability by sixth-order filter methods

Simulation of a 2-D Rightmyer–Meshkov instability (RMI), including inviscid, viscous and magnetic field effects was conducted comparing recently developed sixth-order filter schemes with various standard shock-capturing methods. The suppression of the inviscid gas dynamics RMI in the presence of a magnetic field was investigated by Samtaney and Wheatley et al. Numerical results illustrated here exhibit behavior similar to the work of Samtaney. Due to the different amounts and different types of numerical dissipation contained in each scheme, the structures and the growth of eddies for the chaotic-like inviscid gas dynamics RMI case are highly grid size and scheme dependent, even with many levels of refinement. The failure of grid refinement for all studied numerical methods extends to the viscous gas dynamics case for high Reynolds number. For lower Reynolds number, grid convergence has been achieved by all studied methods. To achieve similar resolution, standard shock-capturing methods require more grid points than filter schemes and yet the CPU times using the same grid for all studied methods are comparable. This paper is based on work that was presented at the 17th International Shock Interaction Symposium (ISIS17), Rome, Italy, 4–8 September 2006.     

Multi‐stage high order semi‐Lagrangian schemes for incompressible flows in Cartesian geometries

Efficient transport algorithms are essential to the numerical resolution of incompressible fluid‐flow problems. Semi‐Lagrangian methods are widely used in grid based methods to achieve this aim. The accuracy of the interpolation strategy then determines the properties of the scheme. We introduce a simple multi‐stage procedure, which can easily be used to increase the order of accuracy of a code based on multilinear interpolations. This approach is an extension of a corrective algorithm introduced by Dupont & Liu (2003, 2007). This multi‐stage procedure can be easily implemented in existing parallel codes using a domain decomposition strategy, as the communication pattern is identical to that of the multilinear scheme. We show how a combination of a forward and backward error correction can provide a third‐order accurate scheme, thus significantly reducing diffusive effects while retaining a non‐dispersive leading error term. Copyright © 2016 John Wiley & Sons, Ltd.     

基于建筑风场数值模拟的网格层块加密法

在建筑风场的数值模拟中,当前普遍采用的离散网格多是计算前一次性布置的固定网格,通常很难适应实际流场变量的变化要求.为提高数值模拟的精度,基于结构化同位网格系统及控制容积离散微分方程的方法,将适应性网格局部加密(AMR)的思想引入到采用压力校正迭代算法的建筑风场模拟中,提出了一种半自适应的层块网格加密方法.该方法可结合误差分析对误差较大的区域网格实行自动判别并实施逐层块状加密.算例分析表明,该方法能在较高的效益下提高数值解的精度.     

Measurement of residual-stress distribution by the incremental hole-drilling method

The hole-drilling method is widely used to measure residual stresses in mechanical components. Recent developments have shown that strains measured on the surface during an incremental drilling can be related to residual-stress distribution. Researchers throughout the world have proposed different calibration methods which lead to more or less accurate results.The present paper discusses different approaches used. A new calibration method is proposed. We also show how finite-element analysis can be used to determine the correlation coefficients. The variation of the strains measured on the surface for each increment is due to, first, the residual stresses in the layer and, second, the change of the hole geometry. Most authors do not consider the latter aspect. Our results show that this causes a significant error in the experimental data. The finite-element method has been used to compute the coefficients for all types of strain-gage rosettes when the hole diameter is predetermined.Another problem of the hole-drilling method is the selection of the drilling tool. Two systems have been studied: ultra-high-speed air turbine and conventional milling machine. The method has been applied on both shot-peened and water-quenched test specimens. The results are successfully compared with the bending-deflection and the X-ray method.     

On the study of waveguide properties of a stack of elastic layers with a set of rigid inclusions

Although there are quite a few papers dealing with dynamical processes in semibounded media containing a set of inhomogeneities of various nature, these process are nowadays far from being described completely. Since the stress-strain states of such mechanical systems depend on many parameters, the traditional analytic and numerical methods for their analysis are inefficient even for a small number of imperfections, and some of them cannot be applied at large vibration frequencies and in large domains. In this connection, it is of interest to study this class of problems in a new setting and develop new numerical-analytic methods for solving these problems. It is especially important to develop methods for studying the resonance properties of mechanical systems.In the present paper, we consider a set of imperfections of the simplest type, namely, plane rigid inclusions lying in planes parallel to the interfaces in a multilayered semibounded medium. In this case, attention is mainly paid to the analysis of singular sets of the determinants for the symbols of the kernels of systems of integral equations generated by the boundary value problems in the class under study.