Towards petascale spectral simulations for transition analysis in wall bounded flow

An efficient parallel spectral method for direct numerical simulations of transitional and turbulent flows is described in this paper. The parallelization is classically based on a bidimensional domain decomposition, but has been specifically developed for a solenoidal Fourier–Chebyshev spectral approximation where in one Fourier direction, the number of modes is very large compared with the two other directions. The approach therefore differs from classical libraries developed for cubic Fourier boxes. The strategy uses message‐passing interface (MPI) for message‐passing among nodes and is fairly portable. One of the originalities of this paper is the use of an efficient hybrid programming with MPI for internodes communications and a coarse grain parallelism using OpenMP for core shared‐memory computation, instead of the classical hybrid programming with MPI and a fine granularity parallelism at the loop level with OpenMP directives. This hybrid parallelism has been tested on the recent generation of high‐performance parallel supercomputers involving a few tens of cores per node. Performances are evaluated on different low‐frequency and high‐frequency processors massively parallel platforms. We demonstrate that spectral methods, which are known to be inherently ill‐fitted for the new generation of high‐performance distributed‐memory computers, can be implemented efficiently using this hybrid programming with good scalability and a very fast wall‐clock time per iteration. New numerical experiments are therefore now accessible on petascale computers, while keeping the attractive features of spectral methods such as accuracy, exponential convergence, computational efficiency and conservative properties. This is illustrated by a direct numerical simulation of the transition of the boundary layers developing from the entrance section of a plane channel and interacting to merge into a fully turbulent flow. Copyright © 2012 John Wiley & Sons, Ltd.     

Quasiperiodic solutions of the Navier-Stokes equations induced by the quasiperiodic shape of the boundaries of a two-dimensional flow domain

Steady quasiperiodic solutions of the Navier-Stokes equations in an infinite two-dimensional layer with quasiperiodic boundaries are obtained on the Reynolds number range 0 < Re* < 200. The calculations are performed using a spectral-difference method based on the representation of the quasiperiodic solutions in the form of convergent double Fourier series. The properties of these solutions and the distinctive features of their spectra are studied and their fundamental differences from periodic solutions are shown. The possibility of applying the quasiperiodic solutions for modeling flows in fractal layers is discussed.     

An implicit pseudo‐spectral multi‐domain method for the simulation of incompressible flows

A pseudo‐spectral method for the solution of incompressible flow problems based on an iterative solver involving an implicit treatment of linearized convective terms is presented. The method allows the treatment of moderately complex geometries by means of a multi‐domain approach and it is able to cope with non‐constant fluid properties and non‐orthogonal problem domains. In addition, the fully implicit scheme yields improved stability properties as opposed to semi‐implicit schemes commonly employed. Key components of the method are a Chebyshev collocation discretization, a special pressure–correction scheme, and a restarted GMRES method with a preconditioner derived from a fast direct solver. The performance of the proposed method is investigated by considering several numerical examples of different complexity, and also includes comparisons to alternative solution approaches based on finite‐volume discretizations. Copyright © 2003 John Wiley & Sons, Ltd.     

Aerodynamic and aeroacoustic analysis of a pulsating internal flow by a multidomain weak collocation spectral method

A numerical methodology is presented for the coupled aerodynamic and aeroacoustic analysis of time dependent laminar viscous flows in general two-dimensional geometries. The overall procedure is constituted firstly by the solution of the incompressible Navier-Stokes equations and secondly by the solution of a linear evolution equation of second order in space and time which originates from Lighthill's acoustic analogy. A semi-implicit projection method is utilized for the time integration of the incompressible Navier-Stokes system, while a choice between a second order implicit Newmark scheme and a fourth order explicit Runge-Kutta-Nyström method is offered for the temporal discretization of the wave equation. The multidomain weak Legendre collocation spectral method on quadrilateral subdomain topologies is employed for the spatial approximation of both the fluid dynamic and the acoustic problems. A specific pulsating internal flow inside a plane constricted channel is selected as a representative application for the assessment of the capabilities of the proposed discrete algorithm. Numerous results are presented and discussed, so as to thoroughly demonstrate the behavior of the numerical procedure.     

Computations of a laminar backward‐facing step flow at Re=800 with a spectral domain decomposition method

The two‐dimensional laminar incompressible flow over a backward‐facing step is computed using a spectral domain decomposition approach. A minimum number of subdomains (two) is used; high resolution being achieved by increasing the order of the basis Chebyshev polynomial. Results for the case of a Reynolds number of 800 are presented and compared in detail with benchmark computations. Stable accurate steady flow solutions were obtained using substantially fewer nodes than in previously reported simulations. In addition, the problem of outflow boundary conditions was examined on a shortened domain. Because of their more global nature, spectral methods are particularly sensitive to imposed boundary conditions, which may be exploited in examining the effect of artificial (non‐physical) outflow boundary conditions. Two widely used set of conditions were tested: pseudo stress‐free conditions and zero normal gradient conditions. Contrary to previous results using the finite volume approach, the latter is found to yield a qualitatively erroneous yet stable flow‐field. Copyright © 1999 John Wiley & Sons, Ltd.     

Analytical solutions for non-isothermal viscoelastic torsional flow in a bounded domain

The effects of viscous heating on torsional flow of viscoelastic fluids in the parallel-plate viscometer are considered. The fluid is modelled by the Oldroyd-B constitutive equation with temperature-dependent viscosities and relaxation time which obey a Nahme type law. Analytic solutions are obtained in the limit of small Nahme–Griffith number which are uniformly valid up to the fluid–air interface. If the aspect ratio, α is small then there is a thermal boundary layer of width α at the free surface.     

不可压缩机翼绕流的有限谱法计算

结合有限谱QUICK格式求解不可压缩粘性流问题。这一格式用于模拟不同攻角下的NACA1200机翼绕流问题。利用体积力,提出了将流场速度从0加速到来流速度的方法。区别于传统的压力梯度为零的边界条件,推导出一个更精确的压力边界条件。为使速度散度保持为零,在泊松方程中给速度散度一个特殊的处理。这一成果说明了有限谱法不但具有很高的精度,而且能灵活地和其他格式一起构造出新的格式,从而成功地应用到复杂流场不可压缩流动的数值计算中。     

Comparing vortex methods and finite difference methods in a homogeneous turbulent shear flow

The vortex method is applied to the calculation of a homogeneous shear turbulence, and compared with a finite difference code using identical calculation conditions. The core spreading method with spatial adaptation is selected as the viscous diffusion scheme of the vortex method. The shear rate is chosen so that it matches the maximum value observed in a fully developed channel flow. The isosurface, anisotropy tensors, and joint probability density functions reflect the ability of the present vortex method to quantitatively reproduce the anisotropic nature of strongly sheared turbulence, both instantaneously and statistically. Copyright © 2009 John Wiley & Sons, Ltd.     

A spectral boundary integral method for inviscid water waves in a finite domain

In this paper, we show how the spectral formulation of Baker, Meiron and Orszag can be used to solve for waves on water of infinite depth confined between two flat, vertical walls, and also how it can be modified to take into account water of finite depth with a spatially varying bottom. In each case, we use Chebyshev polynomials as the basis of our representation of the solution and filtering to remove spurious high‐frequency modes. We show that spectral accuracy can be achieved until wave breaking, plunging or wall impingment occurs in two model problems. Copyright © 2016 John Wiley & Sons, Ltd.     

Spectral analysis of flux vector splitting finite volume methods

New results are presented here for finite volume (FV) methods that use flux vector splitting (FVS) along with higher‐order reconstruction schemes. Apart from spectral accuracy of the resultant methods, the numerical stability is investigated which restricts the allowable time step or the Courant–Friedrich–Lewy (CFL) number. Also the dispersion relation preservation (DRP) property of various spatial and temporal discretization schemes is investigated. The DRP property simultaneously fixes space and time steps. This aspect of numerical schemes is important for simulation of high‐Reynolds number flows, compressible flows with shock(s) and computational aero‐acoustics. It is shown here that for direct numerical simulation applications, the DRP property is more restrictive than stability criteria. Copyright © 2001 John Wiley & Sons, Ltd.     

An analysis of perturbation based methods for the treatment of parameter uncertainty in numerical groundwater models

Monte Carlo methods are robust approaches for the estimation of prediction uncertainty in groundwater flow and transport modelling under uncertain model parameters. However Monte Carlo procedures estimate the prediction statistics by generating a population of solutions from random realisations of the model parameters which are consistent with the parameter statistics, and as a result are computationally demanding. Taylor series based procedures offer an alternative to Monte Carlo methods for calculating the prediction statistics. Two such approaches, the first-order second moment and McLaughlin and Wood's perturbation method, are based on using a Taylor series to derive approximate expressions for the model predictions first and second statistical moments. In this paper the perturbation method presented by McLaughlin and Wood is rederived using Vetter matrix notation. This is compared with the first-order second moment (FOSM) method and while the steady state expressions for these two approaches are shown to be equivalent, the transient forms are considerably different. A new form of the FOSM is derived, which is simpler and has a lower computational burden. However, the transient McLaughlin and Wood expression is found to have a significantly lower computational overhead than either of the FOSM methods presented.     

Fourier analysis of Schwarz domain decomposition methods for the biharmonic equation

Schwarz methods are an important type of domain decomposition methods. Using the Fourier transform, we derive error propagation matrices and their spectral radii of the classical Schwarz alternating method and the additive Schwarz method for the biharmonic equation in this paper. We prove the convergence of the Schwarz methods from a new point of view, and provide detailed information about the convergence speeds and their dependence on the overlapping size of subdomains. The obtained results are independent of any unknown constant and discretization method, showing that the Schwarz alternating method converges twice as quickly as the additive Schwarz method.     

The application of evolutionary and maximum entropy algorithms to photoelastic spectral analysis

Over the past 10 years, spectral analysis has been shown to have the potential to be a reliable means of automating photoelasticity. However, the four methods of analyzing the spectra that have previously been proposed are slow and, in some cases, inaccurate. This paper describes three new methods for spectral analysis based on the maximum entropy method, a genetic algorithm and a memetic algorithm. Thirty-five spectra for known fringe orders were recorded and used in testing the four existing methods and the three new ones. It was found that the new methods were all considerably faster than the existing methods, although less accurate than the best existing method. By combining the maximum entropy method with either the genetic algorithm or the memetic algorithm, spectra could be analyzed up to 30 times as fast as they could with any of the existing methods and with comparable accuracy.     

Modified Laguerre spectral and pseudospectral methods for nonlinear partial differential equations in multiple dimensions

The Laguerre spectral and pseudospectral methods are investigated for multidimensional nonlinear partial differential equations. Some results on the modified Laguerre orthogonal approximation and interpolation are established, which play important roles in the related numerical methods for unbounded domains. As an example, the modified Laguerre spectral and pseudospectral methods are proposed for two-dimensional Logistic equation. The stability and convergence of the suggested schemes are proved. Numerical results demonstrate the high accuracy of these approaches.     

Calculating response spectral moments by complex modal analysis

Response spectral moments are useful for system reliability analysis. Usually, spectral moments are calculated by the frequency domain method. Based on the time domain modal analysis of random vibrations, the authors present a new method for calculating response spectral moments through response correlation functions. The method can be applied to both classical and non-classical damping cases and to three kinds of random excitations, i.e., white noise, band-limited white noise, and filtered white noise. Project supported by the National Natural Science Foundation of China.     

广义KdV方程Fourier谱逼近的最优误差估计

A Fourier spectral method for the generalized Korteweg-de Vries equation with periodic boundary conditions is analyzed, and a corresponding optimal error estimate in L^2-norm is obtained. It improves the result presented by Maday and Quarteroni. A modified Fourier pseudospectral method is also presented, with the same convergence properties as the Fourier spectral method.     

Numerical study on heavy rigid particle motion in a plane wake flow by spectral element method

Experimental particle dispersion patterns in a plane wake flow at a high Reynolds number have been predicted numerically by discrete vortex method (Phys. Fluids A 1992; 4 :2244–2251; Int. J. Multiphase Flow 2000; 26 :1583–1607). To address the particle motion at a moderate Reynolds number, spectral element method is employed to provide an instantaneous wake flow field for particle dynamics equations, which are solved to make a detail classification of the patterns in relation to the Stokes and Froude numbers. It is found that particle motion features only depend on the Stokes number at a high Froude number and depend on both numbers at a low Froude number. A ratio of the Stokes number to squared Froude number is introduced and threshold values of this parameter are evaluated that delineate the different regions of particle behavior. The parameter describes approximately the gravitational settling velocity divided by the characteristic velocity of wake flow. In order to present effects of particle density but preserve rigid sphere, hollow sphere particle dynamics in the plane wake flow is investigated. The evolution of hollow particle motion patterns for the increase of equivalent particle density corresponds to that of solid particle motion patterns for the decrease of particle size. Although the thresholds change a little, the parameter can still make a good qualitative classification of particle motion patterns as the inner diameter changes. Copyright © 2008 John Wiley & Sons, Ltd.     

Evaluation analysis of prediction methods for two-phase flow pressure drop in mini-channels

Two-thousand and ninety-two data of two-phase flow pressure drop were collected from 18 published papers of which the working fluids include R123, R134a, R22, R236ea, R245fa, R404a, R407C, R410a, R507, CO₂, water and air. The hydraulic diameter ranges from 0.506 to 12 mm; Re_l from 10 to 37,000, and Re_g from 3 to 4 × 10⁵. Eleven correlations and models for calculating the two-phase frictional pressure drop were evaluated based upon these data. The results show that the accuracy of the Lockhart–Martinelli method, Mishima and Hibiki correlation, Zhang and Mishima correlation and Lee and Mudawar correlation in the laminar region is very close to each other, while the Muller-Steinhagen and Heck correlation is the best among the evaluated correlations in the turbulent region. A modified Chisholm correlation was proposed, which is better than all of the evaluated correlations in the turbulent region and its mean relative error is about 29%. For refrigerants only, the new correlation and Muller-Steinhagen and Heck correlation are very close to each other and give better agreement than the other evaluated correlations.     

A spectral solution of the magneto‐convection equations in spherical geometry

A fully three‐dimensional solution of the magneto‐convection equations—the nonlinearly coupled momentum, induction and temperature equations—is presented in spherical geometry. Two very different methods for solving the momentum equation are presented, corresponding to the limits of slow and rapid rotation, and their relative advantages and disadvantages are discussed. The possibility of including a freely rotating, finitely conducting inner core in the solution of the momentum and induction equations is also discussed. Copyright © 2000 John Wiley & Sons, Ltd.     

Demonstration of ultra hi-fi (UHF) methods

Computational aero-acoustics (CAA) requires efficient, high-resolution simulation tools. Most current techniques utilize finite-difference approaches because high order accuracy is considered too difficult or expensive to achieve with finite volume or finite element methods. However, a novel finite volume approach i.e. ultra hi-fi (UHF) which utilizes Hermite fluxes is presented which can achieve both arbitrary accuracy and fidelity in space and time. The technique can be applied to unstructured grids with some loss of fidelity or with multi-block structured grids for maximum efficiency and resolution. In either paradigm, it is possible to resolve ultra-short waves (defined as waves having wavelengths that are shorter than a grid cell). This is demonstrated here by solving the 4th CAA workshop Category 1 Problem 1.