机器学习在求解一维双曲守恒律方程中的应用

双曲守恒律方程对空气动力学、物理学和海洋学等众多领域问题的计算有着重大意义,本文应用机器学习框架下的BP神经网络对双曲守恒律方程近似求解.首先,采用熵稳定格式及基于自适应移动网格的熵稳定格式所得多个时间层的数值解构造网络输入,采用高分辨率熵稳定格式所得对应的多个时间层的数值解构造网络输出,并对数据集作归一化处理.随后,...     

求解双曲守恒律方程的熵相容格式

为了更好地求解流体动力学中的双曲守恒律方程,本文提出了一种熵相容格式。通过分析单元内跨越激波时熵的产生情况,得到熵产的显式表达式。在熵守恒通量中加入耗散项与熵增项获得熵相容格式的通量,并在此基础上加入限制器构造出高分辨率熵相容格式。在一维浅水波方程与一维相对论力学方程基础上对新格式进行检验,数值模拟结果表明:这种新的格式能准确捕捉解的结构,具有稳定、无振荡、高分辨率特性。因此,本文方法是求解双曲守恒律方程的较为理想的方法。     

高效时空同步四阶熵稳定格式

求解双曲守恒律方程的高阶半离散熵稳定方法在时间方向上采用Runge-Kutta型方法时,算法的计算效率较低,Lax-Wendroff型时间离散方法为这一问题的解决提供了新思路．将WENO (Weighted Essentially NonOscillatory)型四阶熵稳定格式与Lax-Wendroff型两步四阶时间离散方法相结合求解双曲守恒律方程,时空同步可达到四阶精度．相较于流行的Runge-Kutta型时间离散方法,Lax-Wendroff型两步四阶方法只需两步就可以达到四阶精度,从而可提高计算效率．多个不同类型双曲型方程数值结果表明：新的耦合算法计算效率有明显提高,一维问题计算效率至少提高35%,二维问题计算效率至少提高39%,且新算法依旧具有熵稳定性,数值结果分辨率高．     

界面不稳定性的自适应欧拉数值计算

采用欧拉网格自适应算法数值模拟Richtmyer Meshkov和Rayleigh Taylor不稳定多介质流界面,获得了高精度界面特征。对不同流体引入不同位标函数跟踪界面运动,将位标函数方程与流体动力学方程耦合求解,在笛卡儿坐标系中运用二阶精度有限体积算法,保持流场守恒条件下,通过采用多层网格级对笛卡儿网格嵌套细化,从而实现多介质流体界面的高精度自适应跟踪。给出的方法逻辑简单,可以大大节省CPU时间。     

一种新的求解对流占优问题的自适应网格细化方法

在均匀网格上求解对流占优问题时,往往会产生数值震荡现象,因此需要局部加密网格来提高解的精度。针对对流占优问题,设计了一种新的自适应网格细化算法。该方法采用流线迎风SUPG(Petrov-Galerkin)格式求解对流占优问题,定义了网格尺寸并通过后验误差估计子修正来指导自适应网格细化,以泡泡型局部网格生成算法BLMG为网格生成器,通过模拟泡泡在区域中的运动得到了高质量的点集。与其他自适应网格细化方法相比,该方法可在同一框架内实现网格的细化和粗化,同时在所有细化层得到了高质量的网格。数值算例结果表明,该方法在求解对流占优问题时具有更高的数值精度和更好的收敛性。     

高精度迎风紧致差分格式与热流计算研究的注记

利用高精度差分格式求解了可压缩 N-S方程球头热流问题。分析了不同差分格式在对球头粘性绕流热流计算中存在的问题 ,并分析了相应的网格雷诺数。在利用高精度迎风紧致 [1 ] 格式求解粘性绕流热流问题时 ,采用 Steger-Warming[2 ]的通量分裂技术将守恒型方程中的流通向量分裂成两部分 ,在此基础上据风向构造逼近于无粘项的高精度迎风格式。对方程中的粘性部分采用中心差分格式。数值结果表明 :高精度差分格式能在较大的网格雷诺数下较好地计算球头驻点热流     

基于ALE有限元法的流固耦合强耦合数值模拟

针对不同流固耦合问题,提出一种基于任意拉格朗日-欧拉(ALE)有限元技术的分区强耦合算法.运用半隐式特征线分裂算法求解ALE描述下的不可压缩黏性流体Navier-Stokes方程.分别考虑一般平面运动刚体和几何非线性固体,采用复合隐式时间积分法推进结构运动方程,故可选用较大时间步长;进一步应用单元型光滑有限元法求解几何非线性固体大变形,获得更精确结构解且不影响计算效率.运用子块移动技术结合正交-半扭转弹簧近似法高效更新流体动网格;同时将一质量源项引入压力泊松方程满足几何守恒律,无需复杂构造网格速度差分格式.采用简单高效的固定点法配合Aitken动态松弛技术实现各场耦合,可灵活选择先进单场求解技术,具备较好程序模块性.运用本文算法分别模拟了H型桥梁截面颤振问题和均匀管道流内节气阀涡激振动问题.研究表明,数值结果与已有文献数据吻合,计算精度和求解效率均令人满意.     

爆炸冲击波强间断问题的高阶伪弧长算法研究

为了提高对冲击波强间断处的分辨率,通过引入弧长参数,使网格自适应地朝着间断处移动,并结合高精度WENO数值格式,进而达到了对大梯度物理量的高分辨率捕捉。针对网格移动造成的非均匀和非正交现象,通过坐标变换,使得计算过程在均匀正交的计算空间中进行。通过和有限体积下的数值结果对比,结合数值误差分析,可以看到高阶伪弧长数值算法不仅保证了高精度而且对间断的捕捉更加明显,在间断附近解的整体光滑性较好,网格的自适应移动使得解的奇异性得到了削弱,因此可以削弱高阶格式容易引起数值振荡这个缺点。最后采用高阶伪弧长算法计算了化学反应流问题,结果表明高阶伪弧长算法有着较快的收敛率,对于解决爆炸与冲击强间断问题有着较为明显的优势。     

基于ALE有限元法的流固耦合强耦合数值模拟

针对不同流固耦合问题,提出一种基于任意拉格朗日--欧拉(ALE)有限元技术的分区强耦合算法. 运用半隐式特征线分裂算法求解ALE描述下的不可压缩黏性流体Navier-Stokes方程. 分别考虑一般平面运动刚体和几何非线性固体,采用复合隐式时间积分法推进结构运动方程,故可选用较大时间步长;进一步应用单元型光滑有限元法求解几何非线性固体大变形,获得更精确结构解且不影响计算效率. 运用子块移动技术结合正 交--半扭转弹簧近似法高效更新流体动网格;同时将一质量源项引入压力泊松方程满足几何守恒律,无需复杂构造网格速度差分格式. 采用简单高效的固定点法配合Aitken动态松弛技术实现各场耦合,可灵活选择先进单场求解技术,具备较好程序模块性. 运用本文算法分别模拟了H型桥梁截面颤振问题和均匀管道流内节气阀涡激振动问题. 研究表明,数值结果与已有文献数据吻合,计算精度和求解效率均令人满意.      

不可压N-S方程高效算法及二维槽道湍流分析

构造了基于非等距网格的迎风紧致格式,并将其与三阶精度的Adams半隐方法相结合,构造了求解不可压N－S方程高效算法。该算法利用基于交错网格的离散形式的压力Poisson方程求解压力项,解决了边界处的残余散度问题;同时还利用快速Fourier变换将方程的隐式部分解耦,离散后的代数方程组利用追赶法求解,大大减少了计算量。通过对二维槽道流动的数值模拟,证实了所构造的数值方法具有精度高,稳定性好,能抑制混淆误差等优点,同时具有很高的计算效率,是进行壁湍流直接数值模拟的有效方法。在数值模拟的基础上对二维槽道流动进行了分析,得到了Reynolds数从6000到15000的二维流动饱和态解（所谓“二维槽道湍流”）;定性及定量结果均与他人的数值计算结果吻合十分理想。对流场进行了分析,指出了“二维湍流”与三维湍流统计特性的区别。     

A divergence-free semi-implicit finite volume scheme for ideal,viscous, and resistive magnetohydrodynamics

In this paper, we present a novel pressure-based semi-implicit finite volume solver for the equations of compressible ideal, viscous, and resistive magnetohydrodynamics (MHD). The new method is conservative for mass, momentum, and total energy, and in multiple space dimensions, it is constructed in such a way as to respect the divergence-free condition of the magnetic field exactly, also in the presence of resistive effects. This is possible via the use of multidimensional Riemann solvers on an appropriately staggered grid for the time evolution of the magnetic field and a double curl formulation of the resistive terms. The new semi-implicit method for the MHD equations proposed here discretizes the nonlinear convective terms as well as the time evolution of the magnetic field explicitly, whereas all terms related to the pressure in the momentum equation and the total energy equation are discretized implicitly, making again use of a properly staggered grid for pressure and velocity. Inserting the discrete momentum equation into the discrete energy equation then yields a mildly nonlinear symmetric and positive definite algebraic system for the pressure as the only unknown, which can be efficiently solved with the (nested) Newton method of Casulli et al. The pressure system becomes linear when the specific internal energy is a linear function of the pressure. The time step of the scheme is restricted by a CFL condition based only on the fluid velocity and the Alfvén wave speed and is not based on the speed of the magnetosonic waves. Being a semi-implicit pressure-based scheme, our new method is therefore particularly well suited for low Mach number flows and for the incompressible limit of the MHD equations, for which it is well known that explicit density-based Godunov-type finite volume solvers become increasingly inefficient and inaccurate because of the more and more stringent CFL condition and the wrong scaling of the numerical viscosity in the incompressible limit. We show a relevant MHD test problem in the low Mach number regime where the new semi-implicit algorithm is a factor of 50 faster than a traditional explicit finite volume method, which is a very significant gain in terms of computational efficiency. However, our numerical results confirm that our new method performs well also for classical MHD test cases with strong shocks. In this sense, our new scheme is a true all Mach number flow solver.     

Hall效应对三维磁流体发生器的影响

应用三维非理想低磁雷诺数磁流体五方程模型发展了对带有强制项的Navier-Stokes方程组采用熵条件格式, 对椭圆型电势方程采用SOR进行迭代的数值方法,研究了Hall效应对磁流体旁路超燃冲压发动机中磁流体发生器流动及性能的影响.磁流体发生器采用电子束获得有效可靠的电导率. 计算结果表明,Hall效应可引起流场和电场的扭曲, 从而诱导出不稳定二次流的发展与演变,并破坏Joule热的分布. 对这些磁流体现象作出了较详细的分析.最后计算了磁流体发生器的性能参数, 说明Hall效应将导致磁流体发生器的性能下降.      

Solution of the shallow‐water equations using an adaptive moving mesh method

This paper extends an adaptive moving mesh method to multi‐dimensional shallow water equations (SWE) with source terms. The algorithm is composed of two independent parts: the SWEs evolution and the mesh redistribution. The first part is a high‐resolution kinetic flux‐vector splitting (KFVS) method combined with the surface gradient method for initial data reconstruction, and the second part is based on an iteration procedure. In each iteration, meshes are first redistributed by a variational principle and then the underlying numerical solutions are updated by a conservative‐interpolation formula on the resulting new mesh. Several test problems in one‐ and two‐dimensions with a general geometry are computed using the proposed moving mesh algorithm. The computations demonstrate that the algorithm is efficient for solving problems with bore waves and their interactions. The solutions with higher resolution can be obtained by using a KFVS scheme for the SWEs with a much smaller number of grid points than the uniform mesh approach, although we do not treat technically the bed slope source terms in order to balance the source terms and flux gradients. Copyright © 2004 John Wiley & Sons, Ltd.     

飞行器跨声速气动弹性数值分析

将流体和结构运动方程分别构造为含子迭代的计算格式,发展了一种紧耦合气动弹性分析方法.其中流体计算的空间离散采用改进的HLLEW(Harten—Lax-van Leer-Einfeldt-Wada)格式. TFI(transfinite inter- polation)方法用于生成随结构变形的自适应多块动网格.利用所发展的方法,对-翼-身-尾气动外形,数值预测了马赫数在0.3-1.3范围内的气动颤振边界.并详细研究了时间步长、子迭代步数、初始流场、耦合方法、疏密网格对颤振计算结果的影响.     

Computations using the preconditioning BI-CGSTAB algorithm in chemical non-equilibrium problems

A robust method for solving the chemical non-equilibrium Navier–Stokes equations, including all of the species conservation and energy production equations, is developed. The algorithm is embodied in a fully coupled, implicit, large block structure. Van Leer flux splitting for inviscid terms and central differencing for viscous terms in the explicit operators are applied in the numerical algorithm. The fully-coupled system is solved implicitly and the bi-conjugate gradient stable (Bi-CGSTAB) method with a preconditioner of incomplete lower–upper (LU)-factorization (ILU) is used for solving large block structure and diagonal dominate matrix equations. The computations are performed for the hypersonic inflow over blunt bodies including half cylinder, double ellipse and blunt nose. The adaptive grid constructed by moving grid method is employed to capture the shock location. Computational results in the present study are compared with other calculated data and exhibit good agreement. Convergence histories of the mean flow variables and species equations demonstrate that the fast convergent rate can be achieved by the preconditioned Bi-CGSTAB method. © 1998 John Wiley & Sons, Ltd.     

Performance study of the multiwavelet discontinuous Galerkin approach for solving the Green-Naghdi equations

This paper presents a multiresolution discontinuous Galerkin (DG) scheme for the adaptive solution of Boussinesq-type equations. The model combines multiwavelet (MW)–based grid adaptation with a DG solver based on the system of fully nonlinear and weakly dispersive Green-Naghdi (GN) equations. The key feature of the adaptation procedure is to conduct a multiresolution analysis using MWs on a hierarchy of nested grids to improve the efficiency of the reference DG scheme on a uniform grid by computing on a locally refined adapted grid. This way, the local resolution level will be determined by manipulating MW coefficients controlled by a single user-defined threshold value. The proposed adaptive multiwavelet DG solver for GN equations is assessed using several benchmark problems related to wave propagation and transformation in nearshore areas. The numerical results demonstrate that the proposed scheme retains the accuracy of the reference scheme, while significantly reducing the computational cost.     

采用自适应直角网格计算三维增升装置绕流

针对三维增升装置绕流，对存在剪刀叉的不连续外形，基于自适应直角网格，提出并介绍了分区和面搭接技术，采用变长宽比网格，进行了直角网格生成和流场Euler方程数值计算. 根据几何外形的特点，在直角网格生成过程中，以外形不连续面作为分区边界，对初始``根'网格实施分区处理，降低了整个网格的生成难度. 通过基于外形的自适应网格加密，详细描述了剪刀叉外形和缝道，提高了网格质量. 在分区边界面上，基于面搭接技术，构造重叠面积切割算法，实现边界两侧网格间的流场信息传递，保证流场计算中的通量守恒. 采用中心有限体积方法，结合双时间推进算法，完成了两段机翼、带增升襟翼翼身组合体绕流流场的Euler方程数值模拟，对计算结果与实验数据进行了对比，验证了所提方法、算法的合理性和实用性.     

A pressure‐based unstructured grid method for all‐speed flows

An all‐speed algorithm based on the SIMPLE pressure‐correction scheme and the ‘retarded‐density’ approach has been formulated and implemented within an unstructured grid, finite volume (FV) scheme for both incompressible and compressible flows, the latter involving interaction of shock waves. The collocated storage arrangement for all variables is adopted, and the checkerboard oscillations are eliminated by using a pressure‐weighted interpolation method, similar to that of Rhie and Chow [Numerical study of the turbulent flow past an airfoil with trailing edge separation. AIAA Journal 1983; 21 : 1525]. The solution accuracy is greatly enhanced when a higher‐order convection scheme combined with adaptive mesh refinement (AMR) are used. Copyright © 2000 John Wiley & Sons, Ltd.     

On the accuracy and efficiency of CFD methods in real gas hypersonics

A study of viscous and inviscid hypersonic flows using generalized upwind methods is presented. A new family of hybrid flux-splitting methods is examined for hypersonic flows. The hybrid method is constructed by the superposition of the flux-vector-splitting (FVS) method and second-order artificial dissipation in the regions of strong shock waves. The conservative variables on the cell faces are calculated by an upwind extrapolation scheme to third-order accuracy. A second-order-accurate scheme is used for the discretization of the viscous terms. The solution of the system of equations is achieved by an implicit unfactored method. In order to reduce the computational time, a local adaptive mesh solution (LAMS) method is proposed. The LAMS method combines the mesh-sequencing technique and local solution of the equations. The local solution of either the Euler or the NAVIER-STOKES equations is applied for the region of the flow field where numerical disturbances die out slowly. Validation of the Euler and NAVIER-STOKES codes is obtained for hypersonic flows around blunt bodies. Real gas effects are introduced via a generalized equation of state.     

Central finite volume methods for ideal and shallow water magnetohydrodynamics

We propose two‐dimensional central finite volume methods based on our multidimensional extensions of Nessyahu and Tadmor's one‐dimensional non‐oscillatory central scheme and a constrained transport‐type method to solve ideal magnetohydrodynamic problems (MHD) and shallow water magnetohydrodynamic problems (SMHD). The main numerical scheme is second‐order accurate both in space and time and uses an original Cartesian grid coupled to a Cartesian‐ or diamond‐staggered dual grid to by‐pass the resolution of the Riemann problems at the cell interfaces. To treat the non‐vanishing magnetic field/flux divergence we have constructed an adaptation of Evans and Hawley's constrained transport method specifically designed for central schemes. Our numerical results show the efficiency and the potential of the scheme. Copyright © 2010 John Wiley & Sons, Ltd.