9.2          Continuum Elements

 

9.2.0       Analysis Options

 

            In one dimension the element is defined by 2 nodes.  If the number of spatial dimension NSD = 1 (defined in Section 2.1), usual rod theory is used, and the element is assumed to have axial kinematics only.  Otherwise, i.e. if NSD > 1, the element is assumed to be aligned with the -axis in the global reference frame (), and NSD kinematics in directions etc., are assigned to each node.

 

            In two dimensions the element may be used in triangular (3 node or 6 node) or quadrilateral (4 node, 8 node or 9 node) form for plane and axisymmetric analysis.  The nodes of the element must be input in counterclockwise order in the order shown in Figure 9.2.0.1.  The plane of analysis is assumed to be the  plane, and the element is assumed to have unit thickness in the plane option.  In axisymmetric analysis the radial direction is specified as the -axis. Reduced / selective numerical integration and the mean dilatational formulation may be employed to improve element behavior in various situations.  These options should be activated only by users fully knowledgeable in their use.

 

            In three dimensions the element may be used in tetrahedral (10 or 4 node), wedge (6 node or 15 node) or brick (8 node or 20 node) form.  The nodes of the element must be input in the order shown in Figure 9.2.0.1.  Reduced/selective numerical integration and the mean dilatational formulation may be employed to improve element behavior in various situations.  These options should be activated only by users fully knowledgeable in their use.

 

            Stresses/strains in the global coordinate system, principal stresses/strains, maximum shear stress/strain and angle of inclination, in degrees, of principal states are output at the element centroid, which is generally the point of optimal accuracy.  All shear strains are reported according to the "engineering" convention (i.e. twice the value of the tensor components).

 

            In the following, NSD = number of spatial dimension; NDOF = total number of degrees of freedom; and NED = element nodal local degrees of freedom.

 

 


 

Figure 9.2.0.1 Continuum Elements

 


 

9.2.0.1       Solid Equation

 

QDC_SOLID

 

 

            Element_name = QDC_Solid, etc…

                  < stress material data >

                  < body force data >

                  < connectivity data >

                  < field output data >

 

 

            The element is used for solution of the following equations:

 

                            

 

where  solid stress,  solid acceleration,  body force (per unit mass) and  mass density.  NSD solid kinematic degrees of freedom are assigned to each node, in the  directions, respectively.

 

·         For saturated porous media applications

where solid effective stress,  pore fluid pressure;  ; Cs = solid grains compressibility; Cm = solid matrix compressibility; total mass density,  solid mass density,
 fluid mass density and  porosity.  For fully undrained (viz., no diffusion) cases, the pore fluid pressure is determined from the computed solid velocities through the following equation:

where ;  solid grains modulus, fluid bulk modulus

 

·         For multi-phase fluid flow applications:

 

 

where  solid effective stress;  fluid phase number ( 1, number_of_phases);  degree of saturation;  fluid pressure;

 total mass density;  solid mass density and  = porosity;  fluid phase mass density.

 

·         For thermo solids:

 

where  solid effective stress,  temperature and  thermal moduli (second-order tensor).

 

·         For piezoelectric solids:

 

where  solid effective stress,  electric field (viz.,  where  electric potential), and  piezoelectric constants (third-order tensor, viz., ).

 

 

Implementation Issues

 

            For coupled problems, in the implementation we adopted, the dependent variables are the velocity and the pressure and/or temperature, and a Petrov-Galerkin formulation is used to circumvent restrictions of the Babuska-Brezzi condition.  In particular, equal-order interpolations are used for both the velocity and the pressure.

 

 

References / Bibliography

 

1.      Babuska, I., "Error Bounds for Finite Element Method," Numer. Math, 16 (1971), 322-333.

 

2.      Brezzi, F..,"On the Existence, Uniqueness and Approximation of Saddle-Point Problems Arising from Lagrange Multipliers," Rev. Francaise d'Automatique Inform. Rech. Oper., Ser. Rouge Anal. Numer. 8, R-2 (1974) 129-151.

 

3.      Hughes, T.J.R, Franca, L.P. and Balestra, M., "A New Finite Element Formulation for Computational Fluid Dynamics: V. Circumventing the Babuska-Brezzi Condition: A Stable Petrov-Galerkin Formulation of the Stokes Problem Accommodating Equal-Order Interpolations," Comp. Meth. Appl. Mech. Eng., Vol. 59 (1986) 85-99.

 

 

 


 

9.2.0.2       Fluid Equation

 

QDC_FLUID

 

 

            Element_name = QDC_Fluid ,  etc…

                  < stress material data >

                  < body force data >

                  < connectivity data >

                  < field output data >

 

 

            The element is used for solution of the following equations:

 

 

where  stress,  velocity,  body force (per unit mass),  mass density.  The fluid stress is given by the following equation:

 

 

where  is the pressure, and  the viscous stress tensor.  For isotropic fluids the following viscous relation is used:

 

where

 

symmetric part of the velocity gradient, i.e., in components form:

and

 

 

In the above equation  and  are the viscosity coefficients (also referred to as the Lamé coefficients).  For incompressible flows, .  In our implementation, isothermal incompressible and "slightly compressible" flows are considered.

 

            For incompressible flows, the pressure is determined from the computed velocities through the following equation:

 

 

where  a penalty parameter.  Clearly,  must be large enough so that the compressibility and pressure errors are negligible, yet not so large that numerical ill-conditioning ensues.  The criterion used is (see [1]):


 

 

where  is a constant which depends only on the computer word length (it is independent of the mesh parameter ) and  is the Reynolds number.  Numerical studies reveal that for floating-point word lengths of 60 to 64 bits, . The Reynolds number is computed as

 

 

where  and  are "characteristic" velocity and length, respectively.  The characteristic velocity  is usually taken to be the maximum expected velocity in the flow, and  is taken as a major dimension of the problem (e.g., the diameter).

 

            For "slightly compressible" flows, the pressure is determined through the following equation:

 

 

where

 

Implementation Issues

 

            In the implementation we adopted, the convective force is treated explicitly (to avoid non-symmetric matrices).  This engenders some stability restrictions.  Stability analyses indicate that if no iterations are performed, the upwind scheme used (see Section 9.2.1) for the convection are stable if  satisfies a Courant condition [1], viz.,

 

 

where  = mesh size parameter, and  velocity magnitude for the element.  The above inequality must be satisfied for each finite element in the mesh.  (Note it is solely a convection condition and in particular is independent of the Reynolds number.)

 

            For fluid applications, NSD fluid kinematic degrees of freedom are assigned to each node, in the  directions, respectively.  The time integration must be performed with a hyperbolic time integrator (see Section 12.2) as the resulting initial boundary value problem is treated as parabolic-hyperbolic.

 

Remark:  For isothermal incompressible flows, there is no energy balance or temperature equation.  Thus the pressure variable enters the system of equations only through its derivative and therefore  can be determined only up to an arbitrary constant. This means that the pressure must be specified externally somewhere in the flow field.

 


 

References / Bibliography

 

1.      Hughes et al., "Review of Finite Element Analysis of Incompressible Viscous Flows by the Penalty Function Formulation," J. Comp. Phys., 30, No. 1, (1979), 1-60.

 

2.      Hughes, T.J.R., et al.,"Lagrangian-Eulerian Finite Element Formulation for Incompressible Flows," Comp. Meth. Appl. Mech. Eng., Vol. 29, (1981), 329-349.

 

3.      Prevost, J.H., and Hughes, T.J.R.,"Dynamic Fluid-Structure-Soil Interaction," ASCE Publication on Geotechnical Practice in Offshore Engineering, (1983),  133-143.

 

 

 


 

9.2.0.3       Stokes Flow Equation

 

QDC_STOKES

 

 

            Element_name = QDC_Stokes ,  etc…

                  < stress material data >

                  < body force data >

                  < connectivity data >

                  < field output data >

 

 

            The element is used for solution of the following equations:

 

                                               (momentum balance)

                                                          (incompressibility condition)

 

where  Cauchy stress,  = body force (per unit mass),  mass density.  The fluid stress is given by the following equation:

 

 

where  is the pressure, and  is the dynamic viscosity; is the velocity vector, and  is the symmetrical part of the velocity gradient.

 

Implementation Issues

 

            In the implementation we adopted, the dependent variables are the velocity and the pressure, and a Petrov-Galerkin formulation is used to circumvent restrictions of the Babuska-Brezzi condition.  In particular, equal-order interpolations are used for both the velocity and the pressure.

 

 

References / Bibliography

 

1.      Babuska, I., "Error Bounds for Finite Element Method," Numer. Math, 16 (1971), 322-333.

 

2.      Brezzi, F..,"On the Existence, Uniqueness and Approximation of Saddle-Point Problems Arising from Lagrange Multipliers," Rev. Francaise d'Automatique Inform. Rech. Oper., Ser. Rouge Anal. Numer. 8, R-2 (1974) 129-151.

 

3.      Hughes, T.J.R, Franca, L.P. and Balestra, M., "A New Finite Element Formulation for Computational Fluid Dynamics: V. Circumventing the Babuska-Brezzi Condition: A Stable Petrov-Galerkin Formulation of the Stokes Problem Accommodating Equal-Order Interpolations," Comp. Meth. Appl. Mech. Eng., Vol. 59 (1986) 85-99.

 

 

 


 

9.2.0.4       Scalar Convection-Diffusion Equation

 

QDC_TRANSPORT

 

 

            Element_name = QDC_Transport ,  etc…

                  < scalar diffusion material data >

                  < body force data>

                  < connectivity data >

 

 

            The element is used for solution of the following scalar convection-diffusion equation:

 

 

where  concentration,  mass density,  diffusivity = diffusion/dispersion coefficient matrix,  body force, and  given flow velocity.  The Peclet number  is defined as follows:

 

 

where  characteristic length. For  a purely diffusive solution is obtained, whereas for  solution to the first-order wave equation is obtained. For convection-diffusion (parabolic mode) and advection-diffusion (elliptic mode) one degree of freedom is assigned to each node for the concentration .  In the implementation we adopted, the connective force is treated implicitly and a non-symmetric linear-solver must therefore be used (see Section 12.4).  A stabilized SUPG formulation is used.

 

References / Bibliography

 

1.      Brooks, A.N. and Hughes, T.J.R., “Streamline Upwind/Petrov-Galerkin Formulations for Convection Dominated Flows with Particular Emphasis on the Incompressible Navier-Stokes Equations,” Computer Methods in Applied Mechanics and Engineering, Vol. 32, (1982), pp. 199-259.

 

 

 


 

9.2.0.5       Helmoltz/Laplace Equation

 

QDC_HELMOLTZ

 

 

            Element_name = QDC_Helmoltz ,  etc…

                  < material data >

                  < connectivity data >

 

 

            The element is used for solution of the following scalar equation:

 

 

where p = pressure and c = wave speed .

One degree of freedom is assigned to each node for the pressure.  The element may be used to solve the Laplace equation (elliptic mode) and the Helmoltz equation (hyperbolic mode).

 

 

 


 

9.2.0.6       Mesh Motion Equation

 

QDC_ALE

 

 

            Element_name = QDC_ale ,  etc…

                  < connectivity data >

 

 

            The element is used to compute the mesh displacement field in arbitrary Lagrangian-Eulerian (ALE) models (see Section 5.4), by solving the following vector Laplace equation:

 

 

subject to appropriate prescribed displacement boundary conditions at the moving boundary.  The moving boundary is composed of the moving fluid-solid interfaces as well as the oscillating free surfaces.  The parameter  is a bounded, nondimensional function designed to prevent the inversion of small elements (see e.g. [1]) as follows.  For each element  is defined as:

 

                                           (1)

 

where  area (or volume in 3D) of the current element,

 

                                                                   (2)

 

                                                                     (3)

 

Remarks:  For the degenerate case of a uniform mesh,  and the Laplace equation works well without the additional constraint equation over the element, and .

 

 

References / Bibliography

 

1.      Masud, A. and T.J.R. Hughes, "A Space-time Finite Element Method for Fluid-structure Interaction," SUDAM Report No. 93-3, Stanford University, Stanford, CA, (1993).

 

 


 

 

Notes . .