Sie sind auf Seite 1von 2

The Controversy Over HEX Or TET Meshing . . . Which One Is Better?

The advent of automatic tetrahedral (TET) and semi-automatic hexahedral (HEX) meshers in CAD and finite element systems started a recent controversy: should engineers dealing with designs that require solid finite elements model their designs with TET or HEX elements? Solid regions containing thin-walled areas, concentrated loads, connection of assemblies with different materials and different types of elements stretch the capabilities of automatic meshing and the accuracy of the elements, regardless of the type. Such complex designs usually lead to problems with large numbers of DOF. In general, elements that meet important criteria such as shape functions containing complete polynomials, having continuity across boundaries, and that satisfy the patch test, will converge. The real crux of the debate centers on the most efficient way to obtain accurate stresses for complex solid regions- rather than on the muddy waters of the relative merits of TET versus HEX elements. HEX AND TET ARE BOTH BETTER An article titled "A Performance Study of Tetrahedral Elements in 3D Finite Element Structural Analysis," by A.O. Cifuentes and A. Kalbag of IBM Research Division, Yorktown Heights, N.Y., in Finite Elements in Analysis and Desing in 1992, had an interesting conclusion. Its authors concluded, "This study compares the performance of linear and quadratic tetrahedral elements and hexahedral elements in various structural problems. The problems selected demonstrate the different types of behavior, namely bending, sheer, torsional and axial deformations. It was observed that the results obtained with quadratic tetrahedral elements and hexahedral were equivalent in terms of both accuracy and CPU time." The authors also noted that "For simple geometries, or for applications in which it is possible to build a mesh 'by hand,' analysts have relied heavily on the 8-node hexahedral element-commonly known as 'brick'For more complex geometries, however, the analyst must rely on automatic (or semi-automatic) mesh generators. In general, automatic mesh generators produce meshes of tetrahedral elements, rather than hexahedral elements. The reason is that a general 3-D domain cannot always be decomposed into an assembly of bricks. However, it can always be represented as a collection of tetrahedral elements." Structural Research also ran similar studies comparing results of HEX vs. TET elements. The results of our studies were comparable to those of other reputable researches. With either type of element, the fewer the number of nodes, the lower the accuracy. The 4-node TET and the 8-node HEX approximate curved boundaries as straight lines and require many more elements to achieve convergence of curved boundary problems than do 10-node TET or 20-node HEX elements. However, it pays to remember that, because automatic HEX meshers contain so many limitations as to be semi-automatic at best, using HEX elements may be more time-consuming than using 10-node TET elements, which take advantage of the speed of fully automatic meshing. HOW NEW TECHNOLOGY CHANGES THE DEBATE New technology, existing now, can reduce meshing time by starting with 4-node TET and 8-node HEX elements and then appreciably refining from both HEX and TET elements

automatically- from 4- to 10-node TET and from 8- to 20-node HEX elements. Using the p-method, it can also increase the DOF of the 10-node TET so that it matches or exceeds the 20-node HEX in accuracy, without increasing computer resources. This, rather than the old HEX versus TET argument, is the real key to even more accurate and cost-effective analysis results. The new technology blends iterative and sparse matrix techniques to take lower order meshes, either 4-node TET or 8-node HEX, to form selective 10-node TET or 20-node HEX p-elements during the solution stage. Thus, element formulation and assembly takes place after the solid model has been meshed. The solution procedure can solve large numbers of DOF quickly, reducing computer time and storage space significantly. Because of this capability, the new technology can take advantage of the fact that 10-node TET p-elements form system matrices with a smaller bandwidth than 20-node HEX elements, and can be solved faster for comparably accurate results. To avoid a new controversy, this time about hor p-method meshing of HEX or TET elements, we need to look at adaptive meshing. Many engineers propose that adaptive meshing is the only way to insure the accuracy or convergence of stress intensity values. Bothmethod and p-method adaptive meshing are used commonly. The h-method adds elements in high stress regions, while the pmethod increases the order of polynomials describing the shape function of the element. When using the p-element approach, the accuracy of the elements used can be increased significantly by simply increasing p for either HEX or TET elements. If you use a reasonable initial mesh, remeshing should be unnecessary. Selective p-meshing should be used where only the elements with stresses above an accuracy tolerance would need to have higher polynomials added to their shape function. The TET p-element's stiffness matrix is less dense than that of the HEX p-element, yielding a more sparse system matrix that can

be solved faster than the HEX system matrix. Using the p-adaptive approach, only 4-node TET elements need be generated- reducing meshing time. The 10-node TET can be formed during the solution phase, using a 4node ET with the middle nodes following the curved geometry. CONCLUSIONS 20-node HEX and 10-node TET elements provide good stress results for reasonable meshes with comparable DOF, while 4-node TET and 8-node HEX elements require many more elements for solids with curved boundaries to achieve the same geometrical and stress accuracy. P-method HEX or TET elements can approximate curved solid regions accurately, provide good stress results, and selective p-meshing yields converge stress results while reducing computer time. 10-node TET p-elements form system matrices with a smaller bandwidth than 20-node HEX p-elements, and can thus be solved faster, with similar accuracy results. New iterative and sparse matrix technologies can reduce the time for solving 3D solids problems, while providing more accurate solutions for these larger problems-using the PC or engineering workstation. In addition, you should keep in mind that it is not only difficult to use brick meshers with large models, but also that invariably these meshers may have problems with small features or details on a model. Regardless or what a vendor claims, you will indeed spend time refining the mesh created by brick meshers. Therefore, given that both HEX and TET meshers may offer equal accuracy, does it still make sense to buy a system because a vendor claims to have an automatic brick mesher? You be the judge.
Structural Research & Anlaysis Corp. 12121 Wilshire Blvd., 7th Floor Los Angeles, CA 90025 310.207.2800 www.cosmosm.com

Das könnte Ihnen auch gefallen