Differential-equation-based representation of truncation errors for accurate numerical simulation

High-order compact finite difference schemes for two-dimensional convection-diffusion-type differential equations with constant and variable convection coefficients are derived. The governing equations are employed to represent leading truncation terms, including cross-derivatives, making the overall O(h⁴) schemes conform to a 3 × 3 stencil. We show that the two-dimensional constant coefficient scheme collapses to the optimal scheme for the one-dimensional case wherein the finite difference equation yields nodally exact results. The two-dimensional schemes are tested against standard model problems, including a Navier-Stokes application. Results show that the two schemes are generally more accurate, on comparable grids, than O(h²) centred differencing and commonly used O(h) and O(h³) upwinding schemes.     

A critical evaluation of seven discretization schemes for convection–diffusion equations

A comparative study of seven discretization schemes for the equations describing convection-diffusion transport phenomena is presented. The (differencing) schemes considered are the conventional central- and upwind-difference schemes, together with the Leonard,¹ Leonard upwind¹ and Leonard super upwind difference¹ schemes. Also tested are the so called locally exact difference scheme² and the quadratic-upstream difference scheme.^3,4 In multidimensional problems errors arise from ‘false-diffusion’ and function approximations. It is asserted that false diffusion is essentially a multidimensional source of error. No mesh constraints are associated with errors in function approximation and discretization. Hence errors associated with discretization only may be investigated via one-dimensional problems. Thus, although the above schemes have been tested for one- and two-dimensional flows with sources, only the former are presented here. For 1D flows, the Leonard super upwind difference scheme and the locally exact scheme are shown to be far superior in accuracy to the others at all Peclet numbers and for most source distributions, for the test cases considered. Furthermore, the latter is shown to be considerably cheaper in computational terms than the former. The stability of the schemes and their CPU time requirements are also discussed.     

HIGH-ORDER-ACCURATE DISCRETIZATION STENCIL FOR AN ELLIPTIC EQUATION

The coefficients for a nine-point high-order-accurate discretization scheme for an elliptic equation ∇²u− γ²u=r₀ (∇² is the two-dimensional Laplacian operator) are derived. Examples with Dirichlet and Neumann boundary condtions are considered. In order to demonstrate the high-order accuracy of the method, numerical results are compared with exact solutions.     

A three-level time-split MacCormack method for two-dimensional nonlinear reaction-diffusion equations

A three-level explicit time-split MacCormack method is proposed for solving the two-dimensional nonlinear reaction-diffusion equations. The computational cost is reduced thank to the splitting and the explicit MacCormack scheme. Under the well-known condition of Courant-Friedrich-Lewy (CFL) for stability of explicit numerical schemes applied to linear parabolic partial differential equations, we prove the stability and convergence of the method in L^∞(0,T;L²)-norm. A wide set of numerical evidences which provide the convergence rate of the new algorithm are presented and critically discussed.     

An evaluation of eight discretization schemes for two-dimensional convection-diffusion equations

A comparative study of eight discretization schemes for the equations describing convection-diffusion transport phenomena is presented. The (differencing) schemes considered are the conventional central, upwind and hybrid difference schemes,^1,2 together with the quadratic upstream,^3,4 quadratic upstream extended⁴ and quadratic upstream extended revised difference⁴ schemes. Also tested are the so called locally exact difference scheme⁵ and the power difference scheme.⁶ In multi-dimensional problems errors arise from ‘false diffusion’ and function approximations. It is asserted that false diffusion is essentially a multi-dimensional source of error. Hence errors associated with false diffusion may be investigated only via two- and three-dimensional problems. The above schemes have been tested for both one- and two-dimensional flows with sources, to distinguish between ‘discretization’ errors and ‘false diffusion’ errors.⁷ The one-dimensional study is reported in Reference 7. For 2D flows, the quadratic upstream difference schemes are shown to be superior in accuracy to the others at all Peclet numbers, for the test cases considered. The stability of the schemes and their CPU time requirements are also discussed.     

Probabilistic Analysis of the Upwind Scheme for Transport Equations

We provide a probabilistic analysis of the upwind scheme for d-dimensional transport equations. We associate a Markov chain with the numerical scheme and then obtain a backward representation formula of Kolmogorov type for the numerical solution. We then understand that the error induced by the scheme is governed by the fluctuations of the Markov chain around the characteristics of the flow. We show, in various situations, that the fluctuations are of diffusive type. As a by-product, we recover recent results due to Merlet and Vovelle (Numer Math 106: 129–155, 2007) and Merlet (SIAM J Numer Anal 46(1):124–150, 2007): we prove that the scheme is of order 1/2 in L^￥([0,T],L¹(\mathbb R^d)){L^{\infty}([0,T],L^1(\mathbb R^d))} for an integrable initial datum of bounded variation and of order 1/2−ε, for all ε > 0, in L^￥([0,T] ×\mathbb R^d){L^{\infty}([0,T] \times \mathbb R^d)} for an initial datum of Lipschitz regularity. Our analysis provides a new interpretation of the numerical diffusion phenomenon.     

Efficient higher-order backward characteristics schemes for transient advection

The use of the highest-order ((N – 1)th-order) Lagrangian interpolation Polynomial for the approximation of the exact solution in the backward characteristics scheme with N nodes is inefficient owing to the excessive number of terms in the polynomial. New schemes based on a combination of lower-order polynomials to approximate the exact solution are developed, with the relative weighting of the polynomials determined by Fourier mode analysis. With the addition of a flux limiter and a modified discriminator, the resulting schemes are oscillation-free, highly accurate, efficient and more cost-effective as compared with those schemes using the highest-order Lagrangian polynomial.     

Accuracy of the Renormalization Method for Computing Effective Conductivities of Heterogeneous Media

The accuracy of the renormalization method for upscaling two-dimensional hydraulic conductivity fields is investigated, using two canonical 2 × 2 blocks: a checkerboard geometry and a geometry in which three of the cells have conductivity K ₁ and the other has conductivity K ₂. The predictions of the renormalization algorithm are compared to the arithmetic, harmonic and geometric means, as well as to theoretical predictions and finite element calculations. For the latter geometry renormalization works well over the entire range of the conductivity ratio K ₂/K ₁, but for the checkerboard geometry the error becomes unbounded as the conductivity ratio grows.     

An improved third-order finite difference weighted essentially nonoscillatory scheme for hyperbolic conservation laws

In this article, we present an improved third-order finite difference weighted essentially nonoscillatory (WENO) scheme to promote the order of convergence at critical points for the hyperbolic conservation laws. The improved WENO scheme is an extension of WENO-ZQ scheme. However, the global smoothness indicator has a little different from WENO-ZQ scheme. In this follow-up article, a convex combination of a second-degree polynomial with two linear polynomials in a traditional WENO fashion is used to compute the numerical flux at cell boundary. Although the same three-point information is adopted by the improved third-order WENO scheme, the truncation errors are smaller than some other third-order WENO schemes in L_∞ and L₂ norms. Especially, the convergence order is not declined at critical points, where the first and second derivatives vanish but not the third derivative. At last, the behavior of improved scheme is proved on a variety of one- and two-dimensional standard numerical examples. Numerical results demonstrate that the proposed scheme gives better performance in comparison with other third-order WENO schemes.     

Spatial structure of the flow through an axisymmetric sudden expansion

Spatio-temporal velocity fields of an axisymmetric sudden expansion were measured using an ultrasonic velocity profiler and analyzed to investigate the transitional scheme of the spatial structure using two-dimensional Fourier transform and proper orthogonal decomposition techniques. The variation of the zero-crossing point, the fluctuation energy directed upstream and the eigenmode spectrum all have the same transitional scheme as a function of the Reynolds number. The transitional scheme can be classified Re _d<1,000 for the laminar regime, Re _d=1,000–3,000 for the transitional regime and Re _d>3,000 for the turbulent regime. Especially, in the transitional regime, we found large changes in the flow structure at Re _d=1,500 and 2,000. The jump at Re _d=2,000 is caused by the change in the flow condition upstream. The jump at Re _d=1,500 clearly shows a change in the spatial structure of the flow.     

New numerical schemes based on a criterion for constructing essentially stable and accurate numerical schemes for convection-dominated equations

In order to obtain stable and accurate numerical solutions for the convection-dominated steady transport equations, we propose a criterion for constructing numerical schemes for the convection term that the roots of the characteristic equation of the resulting difference equation have poles. By imposing this criterion on the difference coefficients of the convection term, we construct two numerical schemes for the convection-dominated equations. One is based on polynomial differencing and the other on locally exact differencing. The former scheme coincides with the QUICK scheme when the mesh Reynolds number (Rm) is $\mathop \[{\textstyle{{\rm 8} \over {\rm 3}}}\] $, which is the critical value for its stability, while it approaches the second-order upwind scheme as Rm goes to infinity. Hence the former scheme interpolates a stable scheme between the QUICK scheme at Rm = $\mathop \[{\textstyle{{\rm 8} \over {\rm 3}}}\] $ and the second-order upwind scheme at Rm = ∞. Numerical solutions with the present new schemes for the one-dimensional, linear, steady convection-diffusion equations showed good results.     

A further assessment of interpolation schemes for window deformation in PIV

We have evaluated the performances of the following seven interpolation schemes used for window deformation in particle image velocimetry (PIV): the linear, quadratic, B-spline, cubic, sinc, Lagrange, and Gaussian interpolations. Artificially generated images comprised particles of diameter in a range 1.1 ≤ d _p ≤ 10.0 pixel were investigated. Three particle diameters were selected for detailed evaluation: d _p = 2.2, 3.3, and 4.4 pixel with a constant particle concentration 0.02 particle/pixel². Two flow patterns were considered: uniform and shear flow. The mean and random errors, and the computation times of the interpolation schemes were determined and compared.     

High‐order k‐exact WENO finite volume schemes for solving gas dynamic Euler equations on unstructured grids

This paper presents a family of High‐order finite volume schemes applicable on unstructured grids. The k‐exact reconstruction is performed on every control volume as the primary reconstruction. On a cell of interest, besides the primary reconstruction, additional candidate reconstruction polynomials are provided by means of very simple and efficient ‘secondary’ reconstructions. The weighted average procedure of the WENO scheme is then applied to the primary and secondary reconstructions to ensure the shock‐capturing capability of the scheme. This procedure combines the simplicity of the k‐exact reconstruction with the robustness of the WENO schemes and represents a systematic and unified way to construct High‐order accurate shock capturing schemes. To further improve the efficiency, an efficient problem‐independent shock detector is introduced. Several test cases are presented to demonstrate the accuracy and non‐oscillation property of the proposed schemes. The results show that the proposed schemes can predict the smooth solutions with uniformly High‐order accuracy and can capture the shock waves and contact discontinuities in high resolution. Copyright © 2011 John Wiley & Sons, Ltd.     

MOSQUITO: An efficient finite difference scheme for numerical simulation of 2D advection

An explicit finite difference method for the treatment of the advective terms in the 2D equation of unsteady scalar transport is presented. The scheme is a conditionally stable extension to two dimensions of the popular QUICKEST scheme. It is deduced imposing the vanishing of selected components of the truncation error for the case of steady uniform flow. The method is then extended to solve the conservative form of the depth‐averaged transport equation. Details of the accuracy and stability analysis of the numerical scheme with test case results are given, together with a comparison with other existing schemes suitable for the long‐term computations needed in environmental modelling. Although with a truncation error of formal order 0(ΔxΔt, ΔyΔt, Δt²), the present scheme is shown actually to be of an accuracy comparable with schemes of third‐order in space, while requiring a smaller computational effort and/or having better stability properties. In principle, the method can be easily extended to the 3D case. Copyright © 1999 John Wiley & Sons, Ltd.     

A cross-correlation technique for velocity field extraction from particulate visualization

A rapid time series of photographs of the horizontal cross-sections of several y ⁺ locations were taken of a turbulent open-channel water flow with Re _d = 3,900. A pair of photographic images were obtained with a time difference of 1.3 v/u ² at each y ⁺ locations. The pictures were digitized into 8 bit data with a spatial resolution of 2.5 viscous scales. Instead of identifying discrete particles, a variable interval spatial correlation technique was used to extract the velocity components. With this technique, two-dimensional spatial cross-correlations of the illumination intensities were taken between a pair of picture images. The correlations were taken over small areas and the peak of the correlation coefficients were used to obtain the convection velocity yielding the u and w components of velocity. Some statistical properties were calculated and are shown to be comparable with previous data. Spatial correlations of the velocity components revealed some unique characteristics related to the structure of turbulence.     

Least-squares finite element methods for compressible Euler equations

A method based on backward finite differencing in time and a least-squares finite element scheme for first-order systems of partial differential equations in space is applied to the Euler equations for gas dynamics. The scheme minimizes the L²-norm of the residual within each time step. The method naturally generates numerical dissipation proportional to the time step size. An implicit method employing linear elements has been implemented and proves robust. For high-order elements, computed solutions based on the L²-method may have oscillations for calculations at similar time step sizes. To overcome this difficulty, a scheme which minimizes the weighted H¹-norm of the residual is proposed and leads to a successful scheme with high-degree elements. Finally, a conservative least-squares finite element method is also developed. Numerical results for two-dimensional problems are given to demonstrate the shock resolution of the methods and compare different approaches.     

Role of the inclination of an inverted‐V upper plate on the heat and flow behavior of trapped gases in a modified Rayleigh–Bénard cavity

Thermal buoyant air inside a modified Rayleigh–Bénard (RB) cavity bounded by a lower flat plate and an inverted‐V upper plate has been investigated numerically using the finite‐volume method. The second‐order‐accurate QUICK and SIMPLE schemes were used for the discretization of the convective terms and the pressure–velocity coupling in the set of conservation equations, respectively. The problem under study is controlled by two parameters: (1) the Rayleigh number ranging from 10³ to 10⁶ and (2) the relative height of the vertical sidewalls d. In reference to the latter, it varies from one limiting case corresponding to the standard RB cavity (a rectangle with d = 1) to another limiting case represented by an isosceles triangular cavity where d = 0. The numerical results for the velocity and temperature fields are presented in terms of streamlines, isotherms, local and mean heat fluxes. An additional effort was devoted to determine the critical Ra values characterizing the transition from symmetrical to asymmetrical buoyant airflow responsive to incremental changes in Ra. For purposes of engineering design, a general correlation equation for the Nusselt number in terms of the pertinent Ra and d was constructed using nonlinear multiple regression theory. Copyright © 2007 John Wiley & Sons, Ltd.     

Three-dimensional wake structure measurement using a modified PIV technique

Three-dimensional vortical structures have been measured in a circular-cylinder wake using particle imaging velocimetry (PIV) for the Reynolds number range of 2×10³ to 1×10⁴. The PIV was modified, compared with the conventional one, in terms of its light sheet arrangement to capture reliably streamwise vortices. While in agreement with previous reports, the presently measured spanwise structures complement the data in the literature in the streamwise evolution of the near-wake spanwise vortex in size, strength, streamwise and lateral convection velocities, shedding new light upon vigorous interactions between oppositely signed spanwise structures. The longitudinal vortices display mushroom patterns in the (x, z)-plane in the immediate proximity to the cylinder. Their most likely inclination in the (x, y)-plane is inferred from the measurements in different (x, z)-planes. The longitudinal vortices in the (y, z)-plane show alternate change in sign, though not discernible at x/d > 15. They decay in the maximum vorticity and circulation rapidly from x/d = 5 to 10 and slowly for x/d > 10, and are further compared with the spanwise vortices in size, strength and rate of decay.     

Computation of buoyancy-driven flow in an eccentric centrifugal annulus with a non-orthogonal collocated finite volume algorithm

A computational study is performed on two-dimensional mixed convection in an annulus between a horizontal outer cylinder and a heated, rotating, eccentric inner cylinder. The computation has been done using a non-orthogonal grid and a fully collocated finite volume procedure. Solutions are iterated to convergence through a pressure correction scheme and the convection is treated by Van Leer's MUSCL scheme. The numerical procedure adopted here can easily eliminate the ‘Numerical leakage’ phenomenon of the mixed convection problem whereby strong buoyancy and centrifugal effects are encountered in the case of a highly eccentric annulus. Numerical results have been obtained for Rayleigh number Ra ranging from 7×10³ to 10⁷, Reynolds number Re from 0 to 1200 and Prandtl number Pr from 0.01 to 7. The mixed rotation parameter σ (=Ra/PrRe²) varies from ∞ (pure natural convection) to 0.01 with various eccentricities ε. The computational results are in good agreement with previous works which show that the mixed convection heat transfer characteristics in the annulus are significantly affected by σ and ε. The results indicate that the mean Nusselt number Nu increases with increasing Ra or Pr but decreases with increasing Re. In the case of a highly eccentric annulus the conduction effect becomes predominant in the throat gap. Hence the crucial phenomenon on whereby Nu first decreases and then increases can be found with increasing eccentricity. © 1998 John Wiley & Sons, Ltd.     

AN ACCURATE FINITE DIFFERENCE SCHEME FOR SOLVING CONVECTION-DOMINATED DIFFUSION EQUATIONS

Approximating convection-dominated diffusion equations requires a very accurate scheme for the convection term. The most famous is the method of backward characteristics, which is very precise when a good interpolation procedure is used. However, this method is difficult to implement in 2D or 3D. The goal of this paper is to show that it is possible to construct finite difference schemes almost as accurate as the method of characteristics. Starting from a family of second- and third- order Lax–Wendroff-type schemes, a TVD and L^∞- stable scheme that is easy to implement in higher dimensions is constructed. Numerical tests are performed on various model problems whose solution is known and on classical problems. Comparisons with some other limiter schemes and the method of characteristics are discussed. © 1997 by John Wiley & Sons, Ltd.