Index

Interpolation

Ver: 2
 
 

DESCRIPTION

Allows the user to modify the discretization scheme (the Finite Element method) implemented for various types of problems. Polydata selects default interpolations which are most commonly used. In some cases however, it is useful to change the interpolation in order to reduce the computer time or to increase the stability and the accuracy.   OPTIONS
 
-1 Upper level menu
1' Quadratic coordinates
 Not all remeshing rules are available with Quadratic coordinates.
1" Linear coordinates
Linear coordinates usually work better for problems having free surfaces or interfaces, in the absence of surface tension.  Otherwise, use Quadratic coordinates. 
2 Quadratic element for stresses
Available for sub-tasks of the differential viscoelastic type.
By default, the 4x4 SU element is used for differential viscoelastic simulations. The 4x4 SU element has proven to be stable at high We numbers; its main drawback is the cost in CPU time and memory.
For low We numbers, the user might want to use quadratic elements in stress and velocity and linear elements in pressure (with no upwinding).
Ref :     A new mixed finite element for calculating viscoelastic flows. 
            J.M.Marchal, M.J.Crochet, J.Non-Newt. Fluid. Mech.
            Vol. 26, 77 (1987)
3 4x4 SU element for stresses
Available for sub-tasks of the differential viscoelastic type.
Enables the use of the 4x4 sub-interpolation for stresses with Streamline Upwinding.
4 4x4 SUPG element for stresses
Available for sub-tasks of the differential viscoelastic type.
Enables the use of the 4x4 sub-interpolation for stresses with Streamline Upwinding Petrov-Galerkin.
5 EVSS for stresses
Available for sub-tasks of the differential viscoelastic type in steady-state or evolution mode in 2D and 3D.
Enables the use of the 'Elastic Viscous Split System' method that is much cheaper than the 4x4 sub-interpolation.
6 EVSS SU for stresses
Available for sub-tasks of the differential viscoelastic type in steady-state or evolution mode in 2D and 3D.
Enables the use of the 'Elastic Viscous Split System' method with Streamline Upwinding.
7 EVSS SUPG for stresses
Available for sub-tasks of the differential viscoelastic type in steady-state or evolution mode in 2D and 3D.
Enables the use of the 'Elastic Viscous Split System' method with Streamline Upwinding Petrov-Galerkin.
8 Quadratic velocities, linear pressure
The pressure field is linear and continuous  (with pressure stabilization term).
9 Mini-element for velocities, constant pressure
Available for sub-tasks of the Gen. Newtonian type (isothermal and non-isothermal) and differential viscoelastic flow problems (isothermal and non-isothermal, EVSS only)
This is the simplest (known) velocity-pressure element that satisfies the LBB condition. It is based on quadratic normal velocities on the mid-segments (2-D) or mid-faces (3-D) nodes. The pressure field is constant per element and discontinuous.
Ref : Old and New Finite Elements for Incompressible Flows
M. Fortin, Int. J. Numerical Methods Fluids, Vol. 1, 347-364 (1987)
10 Linear velocities, constant pressure
The pressure field is constant per element and discontinuous
11 Quadratic velocities, linear discontinuous pressure
12 Picard iterations on viscosity(g)
Available for sub-tasks of the Gen. Newtonian type (isothermal and non-isothermal).
This implements Picard (fixed point) iterations instead of Newton-Raphson iterations for the viscosity dependence upon shear-rate. This choice is recommended for power-law or Bird-Carreau fluids with a power law index lower than 0.7.
A Picard iteration typically requires 20 to 30 iterations to converge, while a Newton-Raphson scheme requires 4 to 5 iterations. However, the radius of convergence of the N-R technique is small for power-law fluids.
For low power-law indices, a N-R scheme can be used within an evolution scheme which decreases the power-law index.
Option 3 acts on the current sub-task only (i.e. a power-law flow defined in another sub-task will still be calculated by means of a Newton-Raphson scheme, by default).
13 Linear element for temperature
Available for sub-tasks of the Gen. Newtonian type (non-isothermal).
The linear element for temperature is recommended for low Pe numbers in 3D simulations where the CPU time and/or problem size must be reduced.
14 Quadratic element for temperature
Available for sub-tasks of the Gen. Newtonian type (non-isothermal).
The quadratic element for temperature is the default choice, and is recommended for low Pe numbers.
15 2x2 element for temperature
Available for sub-tasks of the Gen. Newtonian type (non-isothermal).
This choice implements a sub-divided (2x2) linear element for temperature with a Galerkin formulation of the energy equation.
It is recommended for moderate Pe numbers.
16 4x 4 element for temperature
Available for sub-tasks of the Gen. Newtonian type (non-isothermal); not available in 3-D.
This choice implements a sub-divided (4x4) linear element for temperature with a Galerkin formulation of the energy equation.
This is recommended for high Pe numbers, because it refines the mesh for temperature without increasing the cost of the velocity-pressure calculation.
17 Upwinding in momentum equations
Available for sub-tasks of the Gen. Newtonian (isothermal or not) type; not yet implemented in 3-D.
This choice implements a sub-divided (2x2) linear element for the velocity field, a linear element for the pressure, with a Streamline Upwind formulation of the momentum equation.
(The SUPG formulation is not available because troubles arise in the case of a non-zero gravity term with some types of boundary conditions).
18 Sub-interpolation
For stress and temperature fields, this option allows the user to enter a sub-menu which gives the opportunity to define different interpolations on the different sub-domains that constitute the domain of the sub-task
LIMITATIONS The 4x4 element is not implemented in 3-D because of its prohibitive cost; in case of a high Pe number in 3-D, select "linear velocities-constant pressure" together with the 2x2 formulation for temperature.

The use of the EVSS method is mandatory when the viscoelastic model includes different relaxation times,  and when it is non-isothermal.

 
SEE ALSO