基于结构化网格自动划分的汽车发动机强度分析

使用三维设计软件CATIA设计了汽车发动机的复杂结构,然后在网格划分软件HARPOON中将设计的发动机零件自动划分为以结构化六面体网格为主的有限元网格,最后在大型非线性有限元软件ABAQUS中,分析了发动机在静力和动力作用下的应力分布.计算结果表明,采用结构化六面体网格,单元数量少,计算速度快,结果可靠;而采用四面体网格,单元数量大,对于同样配置的计算机,无法进行四面体网格的模型的计算.网格尺寸增加后,单元数量减少,计算机可以计算四面体网格模型,但计算结果与六面体网格偏差比较大.     

基于投影法三维模型全六面体网格的划分方法

针对六面体网格自动划分的难度远高于四面体网格的问题,用投影法对简单形状的初始网格进行投影变换得到每个块体的实际网格,用节点合并算法和再分割技术实现不同块体之间的不同密度网格的过渡,从而生成复杂三维几何形体的全六面体网格.算例表明：该方法生成的网格质量很好,易于实施,适用性广.     

扫掠法有限元网格生成方法

为了提高有限元网格的生成质量,扫掠法生成六面体网格过程中内部节点定位成为关键一步,在研究复杂扫掠体六面体有限元网格生成算法过程中,提出了一种基于扫掠法的六面体网格生成算法,算法利用源曲面已经划分好的网格和连接曲面的结构化网格,用仿射映射逐层投影,生成目标曲面,提出基于Roca算法的内部节点定位的新算法,运用由外向内推进的波前法思想,生成全部的六面体网格。通过实例表明,该算法快速,稳定,可靠,可处理大量复杂2.5维实体六面体网格生成问题。     

一种适用于任意复杂结构的曲六面体网格生成算法

曲六面体网格自动生成是时域谱元法发展的瓶颈问题。本文采用基于体基函数的协变投影算法,将一个10节点曲四面体单元分解为四个20节点曲六面体单元,从而形成一种适用于任意复杂结构的曲六面体网格生成算法。使用该算法能够将曲四面体单元均匀分割,且确保新形成的曲六面体单元的翘曲最小。典型算例表明,使用该算法建立的曲六面体离散模型,完全满足时域谱元法的计算精度要求。     

子域约束扫掠体的六面体网格生成方法

为了保证扫掠法生成六面体网格的效率,必须将扫掠体中的扫掠面划分为结构网格.受到连接体在扫掠面上形成的子域影响,扫掠面的网格划分会出现局部的非结构网格,阻碍扫掠法的应用,为此提出一种新的扫掠方法.该方法将扫掠体中含有子域约束的面网格进行分割,将分割出的结构网格与非结构网格重组为新的扫掠组;在各扫掠组内补充边界点,在边界点内插值生成内部节点,最终完成整个扫掠体六面体网格节点的生成.实例结果表明,文中方法稳定、可靠,可处理复杂2.5D实体六面体网格生成问题.     

非结构化四边形网格生成新算法

改进了一类基于递归区域分解过程的四边形网格生成算法。引入一套健壮的网格模板,为子域的网格剖分提供统一的处理方案,不再限制最终子域为4节点、6节点或8节点子域,提高了算法的时空效率。结合新的子域网格生成过程和自动区域分解算法,利用背景网格和网格源控制分解线上点的布置,得到一个全自动的非结构化四边形网格生成算法。最后通过网格及数值模拟实例验证了算法性能和实用性。     

四面体网格生成方法的研究与实现

针对四面体网格生成过程中需要人工构造背景网格和指定尺寸信息的问题,提出了一种自动计算网格尺寸的方法。该方法通过按层次推进产生四面体网格,根据周围前沿面的尺寸和法线信息,计算新生成四面体的尺寸,使四面体网格在尺寸上能够均匀分布。在网格生成过程中,无需人工指定相关信息,并能保证新生成四面体的质量。联立直线和平面的参数方程,根据方程组解的情况判断线段和三角形是否相交,并对相交的条件作了详细的分析。使用空间八叉树管理前沿面,降低与前沿面相关操作的复杂度。数值算例表明,该方法能够生成较高质量的四面体网格。     

基于医学体数据生成四面体网格的方法

为了从医学体数据直接构造四面体网格,提出一种基于栅格的网格生成算法.该算法的主要思想是从背景栅格中提取并填充代表区域边界的等值面.首先,对医学体数据进行预处理与采样,构建一个背景栅格.其次,用对偶方法从栅格提取三角表面网格,用于分段线性逼近等值面.然后,对栅格中所有位于等值面之内或与等值面相交的立方体,用预定义的模板分解成四面体单元.最后,用Laplacian平滑技术优化四面体网格.在均匀网格的基础上,研究了自适应网格生成算法,在保持网格几何精度的同时精简单元数量,以提高有限元计算效率.给出了从CT数据生成人体股骨远端四面体网格的实例,该网格模型被用于虚拟膝关节镜手术.     

三维空间非结构网格生成方法研究

随着计算流体力学领域待解决问题复杂程度的不断提高,传统的统一贴体结构网格已不能很好地满足针对复杂外形的高精度网格生成需求,而非结构网格以其独特优势受到CFD工作者的普遍关注。带有附面层的非结构网格是非结构网格生成的难点。进行了非结构四面体网格的生成方法研究,同时结合Spider软件平台中结构网格参数化附面层推进的技术优势,进行了基于Spider软件平台中非结构网格生成模块“UGCS”的开发。通过分析大量网格生成实例中网格质量和数值计算结果,验证了算法的可靠性与鲁棒性。     

子分元素法:一种体网格压缩算法

提出了一个包含六面体,四面体,金字塔以及三棱柱单元的混合体网格的压缩与解压算法。首先对非四面体单元按照最小节点标号策略进行子分,然后利用修改的增长缝合算法压缩子分后的四面体网格,解压阶段再通过面删除操作来恢复原始网格。压缩后每个四面体约需10bits的存储,初步试验表明,对于通常的六面体网格,能将数据模型压缩至原先的1／4。     

3D conforming power diagrams for radial LGR in CPG reservoir grids

Mesh generation becomes a crucial step in reservoir flow simulation of new generation. The mesh must faithfully represent the architecture of the reservoir and its heterogeneity. In (Flandrin et al. in IJNME 65(10):1639–1672, 2006) a three-dimensional hybrid mesh model was proposed to capture the radial characteristics of the flow around the wells. In this hybrid mesh, the reservoir is described by a non-uniform Cartesian structured mesh and the drainage areas around the wells are represented by structured radial circular meshes. Unstructured polyhedral meshes are used to connect these two kinds of structured grids. The construction of these transition meshes is based on 3D power diagrams (Aurenhammer in SIAM J Comput 16(1):78–96, 1987) to ensure finite volume properties such as mesh conformity, dual orthogonality and cell convexity. In this paper, we propose an extension of this hybrid model to the case where the reservoir is described by a corner point geometry (CPG) grid. At first, the CPG grid is mapped, in a reference space, into a non-uniform Cartesian grid by minimizing the mapping deformation. Then, a hybrid mesh is generated in this reference space using the previous method. Finally, this mesh is mapped back into the real space. Some quality criterions are introduced to measure and improve the quality of the polyhedral transition mesh.     

多边形单元网格自动生成技术

近年来兴起的多边形有限元方法,在有限元计算中采用多边形单元划分网格,不仅可以更好地适应求解区域的几何形状,而且增加了网格划分的灵活性。为了更方便有效地生成多边形单元网格,在Delaunay三角形的基础上,通过将共圆Delaunay三角形合并为一个圆内接多边形,首先提出了Delaunay多边形的概念,进而提出了一种多边形网格自动生成的Delaunay多边形化算法。利用该Delaunay多边形化技术,对工程中常见的几何形状进行网格划分的具体算例表明,Delaunay多边形化方法可以生成性质优良的多边形单元网格。     

三维实体仿真建模的网格自动生成方法

有限元网格模型的生成与几何拓扑特征和力学特性有直接关系。建立网格模型时，为了更真实地反映原几何形体的特征，在小特征尺寸或曲率较大等局部区域网格应加密剖分；为提高有限元分析精度和效率，在待分析的开口、裂纹、几何突变、外载、约束等具有应力集中力学特性的局部区域，网格应加密剖分。为此，该文提出了基于几何特征和物理特性相结合的网格自动生成方法。该方法既能有效地描述几何形体，又能实现应力集中区域的网格局部加密及粗细网格的均匀过渡。实例表明本方法实用性强、效果良好。     

边界简化与多目标优化相结合的高质量四边形网格生成

目的 高质量四边形网格生成是计算机辅助设计、等几何分析与图形学领域中一个富有挑战性的重要问题。针对这一问题,提出一种基于边界简化与多目标优化的高质量四边形网格生成新框架。方法 首先针对亏格非零的平面区域,提出一种将多连通区域转化为单连通区域的方法,可生成高质量的插入边界;其次,提出"可简化角度"和"可简化面积比率"两个阈值概念,从顶点夹角和顶点三角形面积入手,将给定的多边形边界简化为粗糙多边形;然后对边界简化得到的粗糙多边形进行子域分解,并确定每个子域内的网格顶点连接信息;最后提出四边形网格的均匀性和正交性度量目标函数,并通过多目标非线性优化技术确定网格内部顶点的几何位置。结果 在同样的离散边界下,本文方法与现有方法所生成的四边网格相比,所生成的四边网格顶点和单元总数目较少,网格单元质量基本类似,计算时间成本大致相同,但奇异点数目可减少70% 80%,衡量网格单元质量的比例雅克比值等相关指标均有所提高。结论 本文所提出的四边形网格生成方法能够有效减少网格中的奇异点数目,并可生成具有良好光滑性、均匀性和正交性的高质量四边形网格,非常适用于工程分析和动画仿真。     

A Knowledge-Based Mesh Generation System for Forging Simulation

We developed a knowledge-based system GENMAI (Artificial Intelligence Mesh GENerator) to auto-generate two-dimensional structured meshes. GENMAI is easily applicable to various kinds of application domains. Mesh generation is one of the major tasks confronted in computational simulation. The quality of generated meshes affects computational accuracy and computing time. Since various kinds of domain knowledge are needed to generate high quality structured meshes, the knowledge-based approach has been found effective and successful. Before designing GENMAI, we analyzed mesh generation jobs in plastic deformation analysis and computational fluid dynamics. Then, we formulate GENMAI so that it searches feasible plural divided patterns combinatorially and selects the best pattern. The characteristics of GENMAI are as follows: the meta-inference mechanism and its knowledge representation are widely applicable to various kinds of application domains; and plural patterns can be efficiently obtained at the same time by a search technique based on global dependency and local dependency. We applied GENMAI to forging simulation and developed AI-FESTE, which integrated a rigid-plastic deformation analysis program and GENMAI. Forging designers can easily decide shapes of a forging product and dies and also plan the forming sequence using AI-FESTE. AI-FESTE automates a series of forging analysis operations and shortens the execution time from 1 or 2 day(s) to a few hours. As a result, not only can AI-FESTE shorten the turn-around time, but it can improve the quality of product and die design.     

Three dimensional discontinuous Galerkin methods for Euler equations on adaptive conforming meshes

In the numerical simulation of three dimensional fluid dynamical equations, the huge computational quantity is a main challenge. In this paper, the discontinuous Galerkin (DG) finite element method combined with the adaptive mesh refinement (AMR) is studied to solve the three dimensional Euler equations based on conforming unstructured tetrahedron meshes, that is according the equation solution variation to refine and coarsen grids so as to decrease total mesh number. The four space adaptive strategies are given and analyzed their advantages and disadvantages. The numerical examples show the validity of our methods.     

Using sequential NETGEN as a component for a parallel mesh generator

A parallel tetrahedral mesh generator is developed using the existing sequential NETGEN mesh generator. Mesh generation algorithms developed decompose the geometry into multiple sub-geometries sequentially on a master node and then mesh each sub-geometry in parallel on multiple processors. Two methods are implemented. The first decomposes the geometry and produces conforming surface sub-meshes from which volume meshes can be generated in parallel. A second refinement based method also makes use of the CAD geometry information. A scalable mesh migration algorithm that utilizes “owner updates” rule is implemented. Results show that using the refinement based method, a mesh with a billion elements can be generated in about a minute.     

Large-eddy simulation of the flow around a freight wagon subjected to a crosswind

The purpose of the research reported in this paper is to investigate the aerodynamic response of a freight train subjected to a crosswind at 90° yaw angle using large-eddy simulation (LES). The freight train under investigation consists of many identical wagons. The wagons of the train are of the single-stacked container type that is having a box-like shape. In order to reduce the computational cost, attention is confined to the aerodynamic loads and flow structures around a selected single wagon from the middle of the train. The influences of the neighbouring wagons are simulated by imposing spanwise periodicity. The Reynolds number of the flow, based on the time-averaged speed of the crosswind and the height of the wagon from the ground, is 300,000. The standard Smagorinsky model with model constant of 0.1 and the Van Driest damping function, implemented in the commercial CFD package CFX, are used. The assessment of LES accuracy was carried out by performing three different computations using three different meshes; coarse, medium and fine. The fine mesh simulation gives results similar to that of the medium mesh simulation. The LES results showed that the flow moves in the gaps between the containers and under the wagon with high speed resulting in a much more complex flow topology in the wake of the wagon compared to that previously published around passenger trains. The flow separates at the sharp windward edge of the container to form a large separated flow region on the roof of the container. Aerodynamic forces on the container and freight wagon were computed and their time history was used to reveal the characteristic frequencies of the flow motion around the body. Further, the effect of the moving ground on aerodynamic coefficients is investigated by performing simulation on a moving ground. As a result of the floor motion, the side force coefficient was reduced by about 2.5% while the lift force was increased by about 11%.     

Light mesh: soft shadows as interpolation of visibility

The light meshes (LM) method is a modification of the recursive ray tracing algorithm (RRTA). In this paper we consider an original procedure of shadow calculations, which allows soft shadows to be obtained. The light meshes method can be integrated easily into existing implementations of RRTA as a supplementary technique or plug-ins. In comparison with conventional RRTA the suggested technique allows us to: (a) build soft shadows, (b) speed up a process of shadow calculations, (c) simulate main effects of diffuse interreflections, (d) omit the ambient term. Obviously, visual results of the LM method differ slightly from those calculated with the help of the basic RRTA.

A light mesh is a point set in the 3D space of a scene. In practice, it is approximated by a finite set of points, as a rule, by evenly spaced grids. At each mesh point we compute illuminance, which characterizes some kind of “light field intensity” in the scene space. Then an important part of resulting illumination of a scene object point is obtained by the interpolation of the stored values of nearby mesh points.

Experiments show that the approach proposed reduces computational time with respect to the ray tracing algorithm if a scene contains a lot of light sources and/or an image has a large resolution.     

Implementation and evaluation of automated tetrahedral–prismatic mesh generation software

An open-source implementation of an efficient mesh generation procedure for hybrid prismatic–tetrahedral meshes intended for use in Reynolds-averaged Navier–Stokes solutions is presented. The method employed combines the established, and very fast, Delaunay-based tetrahedral mesh generator TetGen with a novel technique for the creation of a prismatic layer, where constrained global optimization of the envelope is employed. Once a well-shaped envelope is thus obtained, a semi-structured layer of pentahedral elements is extruded between wall and envelope surface. Satisfactory mesh quality is demonstrated by comparing solutions obtained using the new meshes with reference data computed on high-quality advancing-front grids. Mesh generation time is shown to be substantially smaller than with many other methods. Overall, the presented implementation is deemed a valuable tool for cases where many meshes need to be generated for routine analyses and turnaround time is critical.This is an extended version of the paper presented at the 23rd International Meshing Roundtable in London, October 2014.