Choosing between first- and second-order elements

In first-order plane strain, generalized plane strain, axisymmetric quadrilateral, hexahedral solid elements, and cylindrical elements, the strain operator provides constant volumetric strain throughout the element. This constant strain prevents mesh locking when the material response is approximately incompressible (see Solid isoparametric quadrilaterals and hexahedra, Section 3.2.4 of the Abaqus Theory Manual, for a more detailed discussion). Second-order elements provide higher accuracy in Abaqus/Standard than first-order elements for smooth problems that do not involve complex contact conditions, impact, or severe element distortions. They capture stress concentrations more effectively and are better for modeling geometric features: they can model a curved surface with fewer elements. Finally, second-order elements are very effective in bending-dominated problems. First-order triangular and tetrahedral elements should be avoided as much as possible in stress analysis problems; the elements are overly stiff and exhibit slow convergence with mesh refinement, which is especially a problem with first-order tetrahedral elements. If they are required, an extremely fine mesh may be needed to obtain results of sufficient accuracy.

Choosing between full- and reduced-integration elements

Reduced integration uses a lower-order integration to form the element stiffness. The mass matrix and distributed loadings use full integration. Reduced integration reduces running time, especially in three dimensions. For example, element type C3D20 has 27 integration points, while C3D20R has only 8; therefore, element assembly is roughly 3.5 times more costly for C3D20 than for C3D20R. In Abaqus/Standard you can choose between full or reduced integration for quadrilateral and hexahedral (brick) elements. In Abaqus/Explicit you can choose between full or reduced integration for hexahedral (brick) elements. Only reduced-integration first-order elements are available for quadrilateral elements in Abaqus/Explicit; the elements with reduced integration are also referred to as uniform strain or centroid strain elements with hourglass control. Second-order reduced-integration elements in Abaqus/Standard generally yield more accurate results than the corresponding fully integrated elements. However, for first-order elements the accuracy achieved with full versus reduced integration is largely dependent on the nature of the problem.

Hourglassing
Hourglassing can be a problem with first-order, reduced-integration elements (CPS4R, CAX4R, C3D8R, etc.) in stress/displacement analyses. Since the elements have only one integration point, it is possible for them to distort in such a way that the strains calculated at the integration point are all zero, which, in turn, leads to uncontrolled distortion of the mesh. First-order, reduced-integration elements in Abaqus include hourglass control, but they should be used with reasonably fine meshes.

Hourglassing can also be minimized by distributing point loads and boundary conditions over a number of adjacent nodes. In Abaqus/Standard the second-order reduced-integration elements, with the exception of the 27node C3D27R and C3D27RH elements, do not have the same difficulty and are recommended in all cases when the solution is expected to be smooth. The C3D27R and C3D27RH elements have three unconstrained, propagating hourglass modes when all 27 nodes are present. These elements should not be used with all 27 nodes, unless they are sufficiently constrained through boundary conditions. First-order elements are recommended when large strains or very high strain gradients are expected.

ALE: it is a mixture of Lagrangian and Eulerian discretization. Lagrangian is when the mesh deforms with the matertial, and Eulerian is when you have a fixed mesh in the space, adn the material flows from one cell to another. Lagrangian is easier to handle (particularly the definition of boundary conditions) but in cases with large deformation the mesh could be highly distorted, and the calculation becomes inaccurate, or even fail. In Eulerian the material flows within the cells, hence the accuracy is better for large deformation, because there is no mesh disortion at all. However, the treatment of BC is not an easy task, and the flow of material requires convection algorithms. Traditionally Lagrangian is more suited for solid mechanics, and Eulerian for fluid mechanics, but recently Eulerian is more involved in large deformation solid mechanics. ALE (Arbitrary LagrangianEulerian) is a mixture of both, where after a lagrangian deformation the boundaries of the elements (within an ALE-region) are moved - so there is material transport across the boundaries (eulerian) and the distortion of the elements does not become extensive. This procedure is made in every 1-10 timestep, depending on the simulation. ALE combines the advantage of both methods: you can use the easy BC-definition on the outer edge of the ALE-region (which is lagrangian), and the moving element edges within the ALE-region helps to provide reasonable accuracy. Ideally suited for metal forming, where the large deformation could cause severe distortion. Hourglassing: the standard numerical integration (e.g. 2x2 Gaussian quadrature for a bilinear quad, 2x2x2 for a trilinear hexa) has some flaws when combined with incompressible material. The displacements in the mesh are orders of magnitude smaller than in the reality. This is called volumetric locking. Incompressibility is not as rare as one would think, since e.g. most plastic matrial models or many hyperelastic material models assume isochoric deformation. To cure this overstiffening we use reduced integration (for 4-node quads and 8-node hexas 1 integration point in the middle). Theory and tests show that reduced integration solves the volumetric lockin. But with this procedure we introduce another problem: consider a 4-node quad. Move the nodes on the lower edge towards each other, and on the upper edge in the other direction, on both side with the same amount. What you now have is a trapezoid, so your element is deformed. But your integration point (where your starins are measured) is in the middle of the element, and it did not feel anything of this deformation. Neither in the vertical nor in the horizontal direction changed the length and angle of your middle lines! So you have a deformation which produces no strains, hence no forces to resist. This pattern can grow unbounded, and easily destroy your whole simulation. This deformation pattern called hourglass mode, or zero energy mode, or kinematic mode etc. Bottom line is you have to stabilize your element against hourglassing, so you have to build in some artificial stiffness to

prevent this kind of deformation - up to a certain level, because beam bending uses this deformation. There are several methods for this, you should check some textbooks. Both topics are thoroughly discussed in Belytschko's book (Nonlinear Finite Elements for Continua and Structures), far better than I have described them. However I hope, I gave you some insight.

Shear and volumetric locking

Fully integrated elements in Abaqus/Standard and Abaqus/Explicit do not hourglass but may suffer from locking behavior: both shear and volumetric locking. Shear locking occurs in first-order, fully integrated elements (CPS4, CPE4, C3D8, etc.) that are subjected to bending. The numerical formulation of the elements gives rise to shear strains that do not really exist the so-called parasitic shear. Therefore, these elements are too stiff in bending, in particular if the element length is of the same order of magnitude as or greater than the wall thickness. See Performance of continuum and shell elements for linear analysis of bending problems, Section 2.3.5 of the Abaqus Benchmarks Manual, for further discussion of the bending behavior of solid elements. Volumetric locking occurs in fully integrated elements when the material behavior is (almost) incompressible. Spurious pressure stresses develop at the integration points, causing an element to behave too stiffly for deformations that should cause no volume changes. If materials are almost incompressible (elastic-plastic materials for which the plastic strains are incompressible), secondorder, fully integrated elements start to develop volumetric locking when the plastic strains are on the order of the elastic strains. However, the first-order, fully integrated quadrilaterals and hexahedra use selectively reduced integration (reduced integration on the volumetric terms). Therefore, these elements do not lock with almost incompressible materials. Reduced-integration, second-order elements develop volumetric locking for almost incompressible materials only after significant straining occurs. In this case, volumetric locking is often accompanied by a mode that looks like hourglassing. Frequently, this problem can be avoided by refining the mesh in regions of large plastic strain. If volumetric locking is suspected, check the pressure stress at the integration points (printed output). If the pressure values show a checkerboard pattern, changing significantly from one integration point to the next, volumetric locking is occurring. Choosing a quilt-style contour plot in the Visualization module of Abaqus/CAE will show the effect.