基于机器学习的非结构网格阵面推进生成技术初探

网格生成和自适应是制约计算流体力学未来发展的瓶颈问题之一,网格生成自动化和智能化仍是一个需要持续研究的领域.随着高性能计算算力的提升和大数据时代的到来,以机器学习为代表的人工智能方法已经成功应用于包括流体力学在内的多个领域,革命性地推动了这些领域的发展.本文首先简要综述机器学习方法在非结构网格生成领域的研究进展,分析基于机器学习进行非结构网格生成的关键问题;其次,设计非结构网格样本数据格式并实现了样本数据集的自动提取,通过结合人工神经网络和阵面推进法,初步发展了一种基于人工神经网络的二维非结构网格阵面推进生成方法;最后,采用新发展的方法生成了几个典型二维各向同性非结构三角形网格(二维圆柱、二维NACA0012翼型和30p30n三段翼型),进一步采用合并法生成了相应的三角形/四边形混合网格,并测试了网格质量和生成耗时,结果显示本文方法生成的网格质量可以达到商业软件的水平,且生成效率较传统阵面推进法提高30%.      

嵌套网格计算的一种实用初始洞面构造方法及应用

针对结构嵌套网格计算,提出一种实用的初始洞面构造方法,成功应用于自动化嵌套网格生成过程。首先采用封闭物面作为嵌套网格"挖洞"的初始洞边界,然后通过分别给定期望的壁面距离及初始洞间距比例因子的方法进行优化,得到真正的挖洞曲面。通过对NACA4412单段翼型、30P30N三段高升力翼型等算例计算,结果表明:本文提出的方法能有效解决嵌套网格计算中初始洞边界自动构造问题,并能有效避免在近壁面区域进行流场传值,在不需要额外采用洞面优化技术的情况下解决壁面狭缝网格重叠问题,获得更优的嵌套网格流场计算结果。结合本文方法对传统嵌套网格方法进行改进,可为自动化嵌套网格生成提供有效的途径。     

一种基于径向基函数的两步法网格变形策略

基于径向基函数(RBF) 的网格变形方法是一种可靠的网格变形技术,对于任意拓扑的网格都能获得高质量的变形网格. 缩减控制点的RBF 网格变形方法可以大幅提高网格变形效率,但也存在变形后物面误差较大、边界层网格交错的问题. 在缩减控制点方法的基础上,提出了一种适合于带有边界层的黏性网格变形的方法,该方法从物面中选择两组控制点,利用其中一组控制点粗略计算变形后网格位置及变形误差,利用第二组控制点与变形误差插值得到更为精确的变形网格. 利用该方法完成带襟翼的NLR 7301 二维构型和带发动机短舱的DLR F6 翼身组合体的网格变形问题,结果表明该方法可以较大幅度降低变形网格的物面误差,并且有效避免边界层网格交错问题.      

一种新的准结构网格生成方法

在有限元分析中，高质量的结构网格可以有效地提高有限元分析的精度，但结构网格的几何适应性差，针对复杂边界的二维计算模型，现有的方法很难自动生成高质量的结构网格；而非结构网格几何适应性很好，但存在计算效率低和精度差等问题。提出了一种新的准结构网格生成方法，能够实现复杂区域的网格自动生成并且具有高网格质量。该方法首先对计算区域运用Delaunay三角剖分技术生成粗背景网格；然后利用背景网格，使用优化的Voronoi图生成过渡的蜂巢网格；最后，通过中心圆方法对蜂巢网格单元进行结构网格剖分。分析NACA0012翼型数值模拟结果表明，提出的新准结构网格生成方法能够对边界复杂的模型自动生成高质量的网格，并且通过三种不同拓扑类型网格计算结果相互对比及与实验结果对比，证明准结构网格具有高计算精度。     

基于背景网格变形的动态网格移动方法

基于Delaunay背景网格插值技术的动态网格生成方法无需迭代计算,效率较高。但对复杂构形大幅运动的动边界问题,尤其当边界大幅转动时,背景网格极易交叉重叠。重新生成背景网格和重新定位网格节点信息不仅费时而且会导致网格质量的严重下降。本文提出改进的基于背景网格的动态网格变形方法,通过在初始Delaunay背景网格中添加辅助点,生成一层新的背景网格和新的映射关系;采用ball-vertex弹簧法驱动新背景网格的变形,进而牵动目标网格的变形。算例表明,本文提出的动态网格变形方法对所关心区域的网格具有良好保形性,边界可转动更大角度而不会出现网格交叉重叠问题,总体上提高了动态网格更新的效率和质量。     

一种简单快速的多块网格生成方法

利用有限元中的形函数进行网格插值映射分块生成网格,构造了一种块网格拓扑结构,使得块与相邻块之间自动组合,并利用该方法对二维、三维复杂区域进行分块网格生成,说明该方法生成的网格质量较好,适合于复杂区域分块求解。     

用逐点插入法生成Delaunay四面体自适应网格

介绍一种基于Delaunay算法的四面体自适应网格的自动划分方法。该方法用单元尺度场控制生成网格的疏密分布,在不满足尺度场要求的单元面形心处插入新节点,同时计算新节点单元尺寸参数,实现三维实体的Delaunay四面体自动划分。此方法具有几个特点:一是表面网格与体内网格同步划分,无需区分两者;二是结点与单元同时生成;三是生成网格自适应性好,疏密分布任意。另外,还介绍了三维网格划分中两个相关算法:一个是约束面恢复算法,该算法基于约束面不允许有单元边与之相交的性质而提出的;另一个是将二维射线法推广至三维空间,判断一个点是否在一多面体内,实现了凹多面体的划分。最后通过算例对单元质量进行了评价。本文所述方法是一种有效的四面体自适应单元生成算法。     

动态混合网格生成及隐式非定常计算方法

建立了一种基于动态混合网格的非定常数值计算方法. 混合网格由贴体的四边形网格、外场 的多层次矩形网格和中间的三角形网格构成. 当物体运动时，贴体四边形网格随物体运动而 运动，而外场的矩形网格保持静止，中间的三角形网格随之变形；当物体运动位移较大，导 致三角形网格的质量降低，甚至导致网格相交时，在局部重新生成网格. 新网格上的物理量 由旧网格上的物理量插值而得. 为了提高计算效率，采用了双时间步和子迭代相结合的隐式 有限体积格式计算非定常Navier-Stokes方程. 子迭代采用高效的块LU-SGS方法. 利用该 方法数值模拟了NACA0012振荡翼型的无黏和黏性绕流，得到了与实验和他人计算相当一致 的结果.     

一种基于增量径向基函数插值的流场重构方法

由于流场参数重构中, 用于重构的基网格单元的物理参数波动量相对于均值较小, 径向基函数（RBF） 直接插值方法重构会产生较大的数值振荡, 论文提出了一种增量RBF 插值方法, 并用于有限体积的流场重构步, 明显改善了插值格式的收敛性和稳定性. 算例首先通过简单的一维模型说明该方法的有效性, 当目标函数波动量相对于均值为小量时, 增量RBF 插值能够抑制数值振荡; 进一步通过二维亚音速、跨音速定常无黏算例、静止圆柱绕流非定常算例以及超音速前台阶算例来说明该方法在典型流场数值求解中的通用性和有效性. 研究表明增量RBF 重构方法可陡峭地捕捉激波间断, 可有效改善流场求解的收敛性和稳定性, 数值耗散小, 计算效率高.      

A RECONSTRUCTION METHOD FOR FINITE VOLUME FLOW FIELD SOLVING BASED ON INCREMENTAL RADIAL BASIS FUNCTIONS

由于流场参数重构中, 用于重构的基网格单元的物理参数波动量相对于均值较小, 径向基函数（RBF） 直接插值方法重构会产生较大的数值振荡, 论文提出了一种增量RBF 插值方法, 并用于有限体积的流场重构步, 明显改善了插值格式的收敛性和稳定性. 算例首先通过简单的一维模型说明该方法的有效性, 当目标函数波动量相对于均值为小量时, 增量RBF 插值能够抑制数值振荡; 进一步通过二维亚音速、跨音速定常无黏算例、静止圆柱绕流非定常算例以及超音速前台阶算例来说明该方法在典型流场数值求解中的通用性和有效性. 研究表明增量RBF 重构方法可陡峭地捕捉激波间断, 可有效改善流场求解的收敛性和稳定性, 数值耗散小, 计算效率高.     

A large-scale parallel hybrid grid generation technique for realistic complex geometry

High-Performance Computing (HPC) systems and Computational Fluid Dynamics (CFD) have made significant progress in recent years; however, as the basis of the large-scale parallel computing, the massive grid generation of billions of cells has become a bottleneck problem. In this study, a parallel grid generation technique is proposed to generate large-scale mixed grids with arbitrary cell types and scales. The basic idea of our method is analogous to the global mesh refinement technique. An initial coarse grid with arbitrary cell types is regarded as a background mesh which is partitioned into subzones, and subzones are assigned onto different CPU cores. After the cells and faces in each subzone are split, the inserted new points of the solid wall are projected onto the original CAD entities to preserve the geometry accurately. Finally, the tangled cells caused by the projection in the boundary layer are untangled by a local Radial Basis Function mesh deformation technique. Furthermore, a parallel partition approach and an efficient wall distance computing technique for massive grids are developed also to shorten the preprocessing time. The tests show that the preprocessing efficiency has been increased by two or three orders compared with traditional methods. Billions of grids are generated for the AIAA JSM high-lift model and the Chinese CHN-T1 transport model to test the ability of the parallel grid generation technique. The maximum scale up to 19 billion mixed elements is generated using 16 384 CPU cores in parallel, and the mesh quality is acceptable for CFD simulations.     

非结构混合网格在气动热标模模拟中的应用

非结构混合网格消除了结构网格节点的结构性限制,可以较好地处理边界,同时兼顾了粘性边界层模拟的需求,具有灵活性大、对复杂外形适应能力强和生成耗时短等优点,在飞行器气动特性模拟中得到广泛应用.本文针对非结构混合网格的特点,把前期针对非结构混合网格气动力高精度模拟发展改进的梯度计算方法和Roe格式熵修正方法推广应用到气动热流的数值模拟.以典型钝锥标模外形的高超声速绕流为研究对象,开展了不同网格形式和第一层网格不同间距的影响研究.结果 表明,热流计算时,头部物面网格最好采用四边形或四边形交叉剖分得到的三角形网格,物面法向的网格雷诺数取20左右,为热流计算时非结构混合网格的生成提供了指导,同时验证了计算方法的有效性和可靠性.     

Anisotropic mesh adaptation: towards user‐independent,mesh‐independent and solver‐independent CFD. Part II. Structured grids

The present paper is the second article in a three‐part series on anisotropic mesh adaptation and its application to (2‐D) structured and unstructured meshes. In the first article, the theory was presented, the methodology detailed and brief examples given of the application of the method to both types of grids. The second part details the application of the mesh adaptation method to structured grids. The adaptation operations are restricted to mesh movement in order to avoid the creation of hanging nodes. Being based on a spring analogy with no restrictive orthogonality constraint, a wide grid motion is allowed. The adaptation process is first validated on analytical test cases and its high efficiency is shown on relevant transonic and supersonic benchmarks. These latter test cases are also solved on adapted unstructured grids to provide a reference for comparison studies. The third part of the series will demonstrate the capability of the methodology on 2‐D unstructured test cases. Copyright © 2002 John Wiley & Sons, Ltd.     

动网格生成技术及非定常计算方法进展综述

对应用于飞行器非定常运动的数值计算方法(包括动态网格技术和相应的数值离散格式)进行了综述.根据网格拓扑结构的不同,重点论述了基于结构网格的非定常计算方法和基于非结构/混合网格的非定常计算方法,比较了各种方法的优缺点.在基于结构网格的非定常计算方法中,重点介绍了刚性运动网格技术、超限插值动态网格技术、重叠动网格技术、滑移动网格技术等动态结构网格生成方法,同时介绍了惯性系和非惯性系下的控制方程,讨论了非定常时间离散方法、动网格计算的几何守恒律等问题.在基于非结构/混合网格的非定常计算方法中,重点介绍了重叠非结构动网格技术、重构非结构动网格技术、变形非结构动网格技术以及变形/重构耦合动态混合网格技术等方法,以及相应的计算格式,包括非定常时间离散、几何守恒律计算方法、可压缩和不可压缩非定常流动的计算方法、各种加速收敛技术等.在介绍国内外进展的同时,介绍了作者在动态混合网格生成技术和相应的非定常方法方面的研究与应用工作.     

Mass conservation in computational morphodynamics: uniform sediment and infinite availability

Computational morphodynamics in finite volume methods are based on the evaluation of the rate of bed level change in the vertices on the deforming bed. With the use of finite volume methods on collocated (unstructured) grids, the rate of bed level change needs to be interpolated from the mesh faces to the vertices. First, this work reviews two methods based on a vectorial shape of the bed evolution equation (no scalar contributions from storage, erosion and deposition) in terms of their mass conserving properties. Second, a method that allows for scalar contributions in the bed evolution equation (the Exner equation) is proposed for general, unstructured meshes, and an analytical derivation for the simple one‐dimensional problem on a non‐equidistantly discretised grid is considered. The solution is compared with the general two‐dimensional formulation. The two‐dimensional formulation leads to the formulation of a geometric sand sliding routine on unstructured grids. The newly proposed interpolation method and the sand sliding routine are tested, and mass conservation of the sediment is considered with special emphasis on the effect of the solution accuracy for the suspended sediment transport. Discussions on other interpolation methods and their mass conserving properties are given with a special focus of the distance weighted interpolation method directly available and easily applied in O penFOAM . Furthermore, effects from horizontal displacements of the vertices, explicit filtering of the evolving bed and morphological acceleration on global mass conservation, are discussed. Copyright © 2015 John Wiley & Sons, Ltd.     

High-order curvilinear mesh generation technique based on an improved radius basic function approach

A high-order curvilinear hybrid mesh generation technique is developed for high-order numerical method (eg, discontinuous Galerkin method) applications to improve the accuracy for problems with curve boundary. The grid generation technique is based on an improved radius basic function (RBF) approach by which the straight-edge mesh is converted into high-order curve mesh. Firstly, an initial straight-edge mesh is prepared by traditional grid generation software. Then, high-order interpolation points are inserted into the mesh entities such as edges, faces, and cells according to the final demand of mesh order. To preserve the original geometry, the inserted points on solid wall are then projected onto the CAD model using an open source tool “Open Cascade.” Finally, other inserted points in the field near the solid wall are moved to appropriate positions by the improved RBF approach to avoid tangled cells. If we use the original RBF approach, then the inserted points on the edge and face entities normal to the solid boundary in the region of boundary layer will move to improper positions. To overcome this problem, a weighting based on the local grid aspect ratio between normal direction and tangential direction is introduced into the baseline RBF approach. Three typical configurations are tested to validate the mesh generator. Meanwhile, a third-order solution of subsonic flow over an analytical 3D body of revolution in the second International Workshop on High-Order CFD Methods is supplied by a discontinuous Galerkin solver. These numerical tests demonstrate the potential capability of present technique for high-order simulations of complex geometries.     

复杂无粘流场数值模拟的矩形/三角形混合网格技术

建立了一套模拟复杂无粘流场的矩形／三角形混合网格技术，其中三角形仅限于物面附近，发挥非结构网格的几何灵活性，以少量的网格模拟复杂外型；同时在以外的区域采用矩形结构网格，发挥矩形网格计算简单快速的优势，有效地克服全非结构网格计算方法需要较大内存量和较长ＣＰＵ时间的不足．混合网格系统由修正的四分树方法生成．将ＮＮＤ有限差分与ＮＮＤ有限体积格式有机地融合于混合网格计算，消除了全矩形网格模拟曲边界的台阶效应，同时保证了网格间的通量守恒．数值实验表明本方法在模拟复杂无粘流场方面的灵活性和高效性．     

A gas‐kinetic scheme for the simulation of turbulent flows on unstructured meshes

This paper presents a numerical method for simulating turbulent flows via coupling the Boltzmann BGK equation with Spalart–Allmaras one equation turbulence model. Both the Boltzmann BGK equation and the turbulence model equation are carried out using the finite volume method on unstructured meshes, which is different from previous works on structured grid. The application of the gas‐kinetic scheme is extended to the simulation of turbulent flows with arbitrary geometries. The adaptive mesh refinement technique is also adopted to reduce the computational cost and improve the efficiency of meshes. To organize the unstructured mesh data structure efficiently, a non‐manifold hybrid mesh data structure is extended for polygonal cells. Numerical experiments are performed on incompressible flow over a smooth flat plate and compressible turbulent flows around a NACA 0012 airfoil using unstructured hybrid meshes. These numerical results are found to be in good agreement with experimental data and/or other numerical solutions, demonstrating the applicability of the proposed method to simulate both subsonic and transonic turbulent flows. Copyright © 2016 John Wiley & Sons, Ltd.