一种结构动力响应分析的快速高阶算法

为了提高基于高阶格式的结构动力响应微分求积分析方法的计算效率,发展了一种求解动力方程的快速算法.利用微分求积原理将结构动力方程转化为标准Sylvester方程的形式,通过对系数矩阵进行矩阵分解,进而将动力响应Sylvester方程化为一系列标准线性方程组,采用相关成熟算法求解这些线性方程组后即可获得结构动力时程响应的全...     

A PRESSURE-BASED ALGORITHM FOR HIGH-SPEED TURBOMACHINERY FLOWS

The steady state Navier–Stokes equations are solved in transonic flows using an elliptic formulation. A segregated solution algorithm is established in which the pressure correction equation is utilized to enforce the divergence-free mass flux constraint. The momentum equations are solved in terms of the primitive variables, while the pressure correction field is used to update both the convecting mass flux components and the pressure itself. The velocity components are deduced from the corrected mass fluxes on the basis of an upwind-biased density, which is a mechanism capable of overcoming the ellipticity of the system of equations, in the transonic flow regime. An incomplete LU decomposition is used for the solution of the transport-type equations and a globally minimized residual method resolves the pressure correction equation. Turbulence is resolved through the k–ε model. Dealing with turbomachinery applications, results are presented in two-dimensional compressor and turbine cascades under design and off-design conditions. © 1997 John Wiley & Sons, Ltd.     

Nodal operator splitting adaptive finite element algorithms for the Navier–Stokes equations

Solution algorithms for solving the Navier–Stokes equations without storing equation matrices are developed. The algorithms operate on a nodal basis, where the finite element information is stored as the co-ordinates of the nodes and the nodes in each element. Temporary storage is needed, such as the search vectors, correction vectors and right hand side vectors in the conjugate gradient algorithms which are limited to one-dimensional vectors. The nodal solution algorithms consist of splitting the Navier–Stokes equations into equation systems which are solved sequencially. In the pressure split algorithm, the velocities are found from the diffusion–convection equation and the pressure is computed from these velocities. The computed velocities are then corrected with the pressure gradient. In the velocity–pressure split algorithm, a velocity approximation is first found from the diffusion equation. This velocity is corrected by solving the convection equation. The pressure is then found from these velocities. Finally, the velocities are corrected by the pressure gradient. The nodal algorithms are compared by solving the original Navier–Stokes equations. The pressure split and velocity–pressure split equation systems are solved using ILU preconditioned conjugate gradient methods where the equation matrices are stored, and by using diagonal preconditioned conjugate gradient methods without storing the equation matrices. © 1998 John Wiley & Sons, Ltd.     

A comparison of coarse and fine grain parallelization strategies for the simple pressure correction algorithm

The primary aim of this work was to determine the simplest and most effective parallelization strategy for control-volume-based codes solving industrial problems. It has been found that for certain classes of problems, the coarse-grain functional decomposition strategy, largely ignored due to its limited scaling capability, offers the potential for significant execution speed-ups while maintaining the inherent structure of traditional serial algorithms. Functional decomposition requires only minor modification of the existing serial code to implement and, hence, code portability across both concurrent and serial computers is maintained. Fine-grain parallelization strategies at the ‘DO loop’ level are also easy to implement and largely preserve code portability. Both coarse-grain functional decomposition and fine-grain loop-level parallelization strategies for the SIMPLE pressure correction algorithm are demonstrated on a Silicon Graphics 4D280S eight CPU shared memory computer system for a highly coupled, transient two-dimensional simulation involving melting of a metal in the presence of thermal-buoyancy-driven laminar convection. Problems requiring the solution of a larger number of transport equations were simulated by including further scalar variables in the calculation. While resulting in slight degradation of the convergence rate, the functional decomposition strategy exhibited higher parallel efficiencies and yielded greater speed-ups relative to the original serial code. Initially, this strategy showed a significant degradation in convergence rate due to an inconsistency in the parallel solution of the pressure correction equation. After correcting for this inconsistency, the maximum speed-up for 16 dependent variables was a factor of 5·28 with eight processors, representing a parallel efficiency of 67%. Peak efficiency of 76% was achieved using five processors to solve for 10 dependent variables.     

Reliability Index and Failure Probability

Abstract

Numerical algorithms for the solution of nonlinear algebraic equation systems are discussed. Special application to the mechanism and multibody system kinematic analysis, as well as to the problems of constraint stabilization during dynamics simulation is regarded. Special attention is paid to the approaches of a separate solution of the differential equations and constraint stabilization. Numerical procedures that are effective additions to the well-known algorithms based on the Newton-Raphson method are presented. The problems of loss of precision and achievement of large unreal increments of the varying parameters are discussed. The traditional Newton-Raphson method is modified by applying a step reduction procedure that is developed numerically for the symbolic form of kinematic and dynamic equations. An optimization method for stabilization of constraints using the mass matrix of dynamic equations is suggested. According to the objective function defined the stabilization procedure provides minimal deviations of the parameters and their velocities with respect to the solution of the differential equations. No generalized coordinate partitioning is required either for solution of the dynamic equations or for stabilization of the constraints. Several examples of kinematic analysis of single and four contour plane mechanisms and constraint stabilization are solved, and the results are compared. The advantages of the algorithms developed are tested with a high-degree of initial deviation from the real solution. It is also shown that the step correction algorithm could provide admissible solution even when, in many cases, the classical approaches are not reliable. An example of the direct and inverse kinematic problem solutions of the four-degrees-of-freedom spatial platform is presented.     

Finite element solution of the incompressible Navier-Stokes equations by a Helmholtz velocity decomposition

Finite element solution methods for the incompressible Navier-Stokes equations in primitive variables form are presented. To provide the necessary coupling and enhance stability, a dissipation in the form of a pressure Laplacian is introduced into the continuity equation. The recasting of the problem in terms of pressure and an auxiliary velocity demonstrates how the error introduced by the pressure dissipation can be totally eliminated while retaining its stabilizing properties. The method can also be formally interpreted as a Helmholtz decomposition of the velocity vector. The governing equations are discretized by a Galerkin weighted residual method and, because of the modification to the continuity equation, equal interpolations for all the unknowns are permitted. Newton linearization is used and at each iteration the linear algebraic system is solved by a direct solver. Convergence of the algorithm is shown to be very rapid. Results are presented for two-dimensional flows in various geometries.     

A comparative study of high‐order variable‐property segregated algorithms for unsteady low Mach number flows

In this paper, five different algorithms are presented for the simulation of low Mach flows with large temperature variations, based on second‐order central‐difference or fourth‐order compact spatial discretization and a pressure projection‐type method. A semi‐implicit three‐step Runge–Kutta/Crank–Nicolson or second‐order iterative scheme is used for time integration. The different algorithms solve the coupled set of governing scalar equations in a decoupled segregate manner. In the first algorithm, a temperature equation is solved and density is calculated from the equation of state, while the second algorithm advances the density using the differential form of the equation of state. The third algorithm solves the continuity equation and the fourth algorithm solves both the continuity and enthalpy equation in conservative form. An iterative decoupled algorithm is also proposed, which allows the computation of the fully coupled solution. All five algorithms solve the momentum equation in conservative form and use a constant‐ or variable‐coefficient Poisson equation for the pressure. The efficiency of the fourth‐order compact scheme and the performances of the decoupling algorithms are demonstrated in three flow problems with large temperature variations: non‐Boussinesq natural convection, channel flow instability, flame–vortex interaction. Copyright © 2010 John Wiley & Sons, Ltd.     

Local and parallel finite element algorithms based on two-grid discretization for steady Navier-Stokes equations

Local and parallel finite element algorithms based on two-grid discretization for Navier-Stokes equations in two dimension are presented. Its basis is a coarse finite element space on the global domain and a fine finite element space on the subdomain. The local algorithm consists of finding a solution for a given nonlinear problem in the coarse finite element space and a solution for a linear problem in the fine finite element space, then droping the coarse solution of the region near the boundary. By overlapping domain decomposition, the parallel algorithms are obtained. This paper analyzes the error of these algorithms and gets some error estimates which are better than those of the standard finite element method. The numerical experiments are given too. By analyzing and comparing these results, it is shown that these algorithms are correct and high efficient.     

A stable unstructured finite volume method for parallel large-scale viscoelastic fluid flow calculations

A new stable unstructured finite volume method is presented for parallel large-scale simulation of viscoelastic fluid flows. The numerical method is based on the side-centered finite volume method where the velocity vector components are defined at the mid-point of each cell face, while the pressure term and the extra stress tensor are defined at element centroids. The present arrangement of the primitive variables leads to a stable numerical scheme and it does not require any ad-hoc modifications in order to enhance the pressure–velocity–stress coupling. The log-conformation representation proposed in [R. Fattal, R. Kupferman, Constitutive laws for the matrix–logarithm of the conformation tensor, J. Non-Newtonian Fluid Mech. 123 (2004) 281–285] has been implemented in order improve the limiting Weissenberg numbers in the proposed finite volume method. The time stepping algorithm used decouples the calculation of the polymeric stress by solution of a hyperbolic constitutive equation from the evolution of the velocity and pressure fields by solution of a generalized Stokes problem. The resulting algebraic linear systems are solved using the FGMRES(m) Krylov iterative method with the restricted additive Schwarz preconditioner for the extra stress tensor and the geometric non-nested multilevel preconditioner for the Stokes system. The implementation of the preconditioned iterative solvers is based on the PETSc library for improving the efficiency of the parallel code. The present numerical algorithm is validated for the Kovasznay flow, the flow of an Oldroyd-B fluid past a confined circular cylinder in a channel and the three-dimensional flow of an Oldroyd-B fluid around a rigid sphere falling in a cylindrical tube. Parallel large-scale calculations are presented up to 523,094 quadrilateral elements in two-dimension and 1,190,376 hexahedral elements in three-dimension.     

The Numerical Technique for the Landslide Tsunami Simulations Based on Navier–Stokes Equations

The paper presents an integral technique simulating all phases of a landslide-driven tsunami. The technique is based on the numerical solution of the system of Navier–Stokes equations for multiphase flows. The numerical algorithm uses a fully implicit approximation method, in which the equations of continuity and momentum conservation are coupled through implicit summands of pressure gradient and mass flow. The method we propose removes severe restrictions on the time step and allows simulation of tsunami propagation to arbitrarily large distances. The landslide origin is simulated as an individual phase being a Newtonian fluid with its own density and viscosity and separated from the water and air phases by an interface. The basic formulas of equation discretization and expressions for coefficients are presented, and the main steps of the computation procedure are described in the paper. To enable simulations of tsunami propagation across wide water areas, we propose a parallel algorithm of the technique implementation, which employs an algebraic multigrid method. The implementation of the multigrid method is based on the global level and cascade collection algorithms that impose no limitations on the paralleling scale and make this technique applicable to petascale systems. We demonstrate the possibility of simulating all phases of a landslide-driven tsunami, including its generation, propagation and uprush. The technique has been verified against the problems supported by experimental data. The paper describes the mechanism of incorporating bathymetric data to simulate tsunamis in real water areas of the world ocean. Results of comparison with the nonlinear dispersion theory, which has demonstrated good agreement, are presented for the case of a historical tsunami of volcanic origin on the Montserrat Island in the Caribbean Sea.     

THREE-DIMENSIONAL VISCOUS FLOW THROUGH A ROTATING CHANNEL: A PSEUDOSPECTRAL MATRIX METHOD APPROACH

A Fourier–Chebyshev pseudospectral method is used for the numerical simulation of incompressible flows in a three-dimensio nal channel of square cross-section with rotation. Realistic, non-periodic boundary conditions that impose no-slip conditions in two directions (spanwis e and vertical directions) are used. The Navier–Stokes equations are integrated in time using a fractional step method. The Poisson equations for pressure and the Helmholtz equation for velocity are solved using a matrix diagonalization (eigenfunction decomposition) method, through which we are able to reduce a three-dimensional matrix problem to a simple algebraic vector equation. This results in signficant savings in computer storage requirement, particularly for large-scale computations. Verification of the numerical algorithm and code is carried out by comparing with a limiting case of an exact steady state solution for a one-dimensional channel flow and also with a two-dimensional rotating channel case. Two-cell and four-cell two-dimensional flow patterns are observed in the numerical experiment. It is found that the four-cell flow pattern is stable to symmetri cal disturbances but unstable to asymmetrical disturbances.     

A method for finite element parallel viscous compressible flow calculations

This paper presents the parallelization aspects of a solution method for the fully coupled 3D compressible Navier-Stokes equations. The algorithmic thrust of the approach, embedded in a finite element code NS3D, is the linearization of the governing equations through Newton methods, followed by a fully coupled solution of velocities and pressure at each non-linear iteration by preconditioned conjugate gradient-like iterative algorithms. For the matrix assembly, as well as for the linear equation solver, efficient coarse-grain parallel schemes have been developed for shared memory machines, as well as for networks of workstations, with a moderate number of processors. The parallel iterative schemes, in particular, circumvent some of the difficulties associated with domain decomposition methods, such as geometry bookkeeping and the sometimes drastic convergence slow-down of partitioned non-linear problems.     

A fast riemann solver with constant covolume applied to the random choice method

The Riemann problem for the unsteady one-dimensional Euler equations together with the constant-covolume equation of state is solved exactly. The solution is then applied to the random choice method to solve the general initial-boundary value problem for the Euler equations. The iterative procedure to find p*, the pressure between the acoustic waves, involves a single algebraic (non-linear) equation, all other quantities follow directly throughout the x–t plane, except within rarefaction fans where an extra iterative procedure is required. The solution is validated against existing exact results both directly and in conjunction with the random choice method.     

An application of Roe's flux-difference splitting for k-ϵ turbulence model

In this paper Roe's flux-difference splitting is applied for the solution of Reynolds-averaged Navier-Stokes equations. Turbulence is modelled using a low-Reynolds number form of the k-? tubulence model. The coupling between the turbulence kinetic energy equation and the inviscid part of the flow equations is taken into account. The equations are solved with a diagonally dominant alternating direction implicit (DDADI) factorized implicit time integration method. A multigrid algorithm is used to accelerate the convergence. To improve the stability some modifications are needed in comparison with the application of an algebraic turbulence model. The developed method is applied to three different test cases. These cases show the efficiency of the algorithm, but the results are only marginally better than those obtained with algebraic models.     

An efficient parallel and fully implicit algorithm for the simulation of transient free-surface flows of multimode viscoelastic liquids

The parallelization of a fully implicit and stable finite element algorithm with relative low memory requirements for the accurate simulation of time-dependent, free-surface flows of multimode viscoelastic liquids is presented. It is an extension of our multi-stage sequential solution procedure which is based on the mixed finite element method for the velocity and pressure fields, an elliptic grid generator for the deformation of the mesh, and the discontinuous Galerkin method for the viscoelastic stresses [Dimakopoulos and Tsamopoulos [12], [14]]. Each one of the above subproblems is solved with the Newton–Rapshon technique according to its particular characteristics, while their coupling is achieved through Picard cycles. The physical domain is graphically partitioned into overlapping subdomains. In the process, two different kinds of parallel solvers are used for the solution of the distributed set of flow and mesh equations: a multifrontal, massively parallel direct one (MUMPS) and a hierarchical iterative parallel one (HIPS), while viscoelastic stress components are independently calculated within each finite element. The parallel algorithm retains all the advantages of its sequential predecessor, related with the robustness and the numerical stability for a wide range of levels of viscoelasticity. Moreover, irrespective of the deformation of the physical domain, the mesh partitioning remains invariant throughout the simulation. The solution of the constitutive equations, which constitutes the largest portion of the system of the governing, non-linear equations, is performed in a way that does not need any data exchange among the cluster's nodes. Finally, indicative results from the simulation of an extensionally thinning polymeric solution, demonstrating the efficiency of the algorithm are presented.     

基于神经网络的差分方程快速求解方法

近年来, 人工神经网络(artificial?neural?networks, ANN), 尤其是深度神经网络(deep?neural?networks, DNN)由于其在异构平台上的高计算效率与对高维复杂系统的拟合能力而成为一种在数值计算领域具有广阔前景的新方法. 在偏微分方程数值求解中, 大规模线性方程组的求解通常是耗时最长的步骤之一, 因此, 采用神经网络方法求解线性方程组成为了一种值得期待的新思路. 但是, 深度神经网络的直接预测仍在数值精度方面仍有明显的不足, 成为其在数值计算领域广泛应用的瓶颈之一. 为打破这一限制, 本文提出了一种结合残差网络结构与校正迭代方法的求解算法. 其中, 残差网络结构解决了深度网络模型的网络退化与梯度消失等问题, 将网络的损失降低至经典网络模型的1/5000; 修正迭代的方法采用同一网络模型对预测解的反复校正, 将预测解的残差下降至迭代前的10?5倍. 为验证该方法的有效性与通用性, 本文将该方法与有限差分法结合, 对热传导方程与伯格方程进行了求解. 数值结果表明, 本文所提出的算法对于规模大于1000的方程组具有10倍以上的加速效果, 且数值误差低于二阶差分格式的离散误差.      

Hybrid Parallel Linear System Solvers

This paper presents a new approach to the solution of nonsymmetric linear systems that uses hybrid techniques based on both direct and iterative methods. An implicitly preconditioned modified system is obtained by applying projections onto block rows of the original system. Our technique provides the flexibility of using either direct or iterative methods for the solution of the preconditioned system. The resulting algorithms are robust, and can be implemented with high efficiency on a variety of parallel architectures. The algorithms are used to solve linear systems arising from the discretization of convection-diffusion equations as well as those systems that arise from the simulation of particulate flows. Experiments are presented to illustrate the robustness and parallel efficiency of these methods.     

求解三维流固耦合问题的一种全隐全耦合区域分解并行算法

三维流固耦合问题的非结构网格数值算法在很多工程领域都有重要应用,目前现有的数值方法主要基于分区算法,即流体和固体区域分别进行求解,因此存在收敛速度较慢以及附加质量导致的稳定性问题,此外,该类算法的并行可扩展性不高,在大规模应用计算方面也受到一定限制．本文针对三维非定常流固耦合问题,提出一种基于区域分解的全隐全耦合可扩展并行算法．首先基于任意拉格朗日-欧拉框架建立流固耦合控制方程,然后时间方向采用二阶向后差分隐式格式、空间方向采用非结构稳定化有限元方法进行离散．对于大规模非线性离散系统,构造一种结合非精确Newton法、Krylov子空间迭代法与区域分解Schwarz预条件子的Newton-Krylov-Schwarz (NKS)并行求解算法,实现流体、固体和动网格方程的一次性整体求解．采用弹性障碍物绕流的标准测试算例对数值方法的准确性进行了验证,数值性能测试结果显示本文构造的全隐全耦合算法具有良好的稳定性,在不同的物理参数下具有良好的鲁棒性,在“天河二号”超级计算机上,当并行规模从192增加到3072个处理器核时获得了91%的并行效率．性能测试结果表明本文构造的NKS算法有望应用于复杂...     

Numerical simulation of viscous flow around unrestrained cylinders

A 2-D analysis is made for the dynamic interactions between viscous flow and one or more circular cylinders. The cylinder is free to respond to the fluid excitation and its motions are part of the solution. The numerical procedure is based on the finite volume discretization of the Navier–Stokes equations on adaptive tri-tree grids which are unstructured and nonorthogonal. Both a fully implicit scheme and a semi-implicit scheme in the time domain have been used for the momentum equations, while the pressure correction method based on the SIMPLE technique is adopted to satisfy the continuity equation. A new upwind method is developed for the triangular and unstructured mesh, which requires information only from two neighbouring cells but is of order of accuracy higher than linear. A new procedure is also introduced to deal with the nonorthogonal term. The pressure on the body surface required in solving the momentum equation is obtained through the Poisson equation in the local cell. Results including flow field, pressure distribution and force are provided for fixed single and multiple cylinders and for an unrestrained cylinder in steady incoming flow with Reynolds numbers at 200 and 500 and in unsteady flow with Keulegan–Carpenter numbers at 5 and 10.