Remarks on the TAU grid adaptation

Thomas Gerhold

Concept – initial idea for adaptation tool

Introduction

A short presentation of the basic concept & restrictions for the local

grid refinement is given, because it might be useful for discussions

about the adaptation tool.

concept for grid refinement

restrictions

summary

Concept – initial idea for adaptation tool

Unstructured grids allows for local refinement

Flow fields are dominated by local phenomena

(stagnation point/line, expansion, shock,

separation, wake, vortices)

Flow phenomena change with flow conditions

One initial grid and local refinement allows to resolve

flow phenomena efficiently for varying flow

conditions (e.g. for a polar)

Concept - automated chain from CAD to solution

Initial grids can be generated (more or less) automatically,

e.g. by refinement based on surface curvature and other

parameters

Local refinement based on indicators for flow phenomena allows

for an automated process to obtain a final grid for the case-

specific flow conditions

- overhead for a simulation with local refinement

+ saving time in initial grid generation (less user-interaction)

+ improved resolution for the actual flow conditions for a given

number of points

Concept – local refinement

Local grid refinement based on the equi-distribution principle:

insert points to equalize maximum differences of the indicator

function between neighboured cells

Equi-distribution provides the best spatial resolution for a target

number of points for a given initial grid and a given indicator

function

initial grid points

inserted points

maximumdifference

Concept – required functionality

Handle grid refinement

Provide a tool which refines unstructured grids locally based on

edges marked for refinement.

Define refinement regions

Offer build-in indicators for marking edges: default: Pt, Ht, V,

Rho

Allow the user alternatively to select the refinement locations

(e.g. read user-defined sensor-variables, lists for selected

elements, adjoint based selection in future)

concept – offer choices for the indicator setting

Final grid depends on the choice of

the indicator (weights) and other

parameter settings.

Thus, experience to be gained by the

users, which settings to use for

which use cases. Default setting

are certainly not optimal for each

case.

With the user experiences feedback

will show the need for other

additional indicators

Indicator Pt

Indicator V

Proof of concept

It has been shown for cruise

conditions that local grid

refinement (using build-in

Indicators) lead to an improved

solution in comparison to a

reference solution

Turbulent F6 simulation with 2

adaptation steps from 2 mio to 3.5

mio points (Stab-Paper 1998)

Convergence history:

restrictions

Sizes of refined cells depend on

the initial grid and the number

of refinement steps.

Thus, to resolve a wake far

downstream needs several

adaptation steps (eventually

more than 10), because

farfield cell sizes are often 5

orders larger than nearfield

cell sizes

restrictions (in semi structured layers)

No refinement under chopping

regions!

Indicators not evaluated on wall

normal prism-edges: no wall-

tangential refinement!

Thus, some regions of interest

(especially for high lift

configurations) can not be refined

or not be refined locally!

Conclusion

summary:

Basic concept for grid refinement & restrictions have been presented

Experiences:

Many applications has shown that adaptation improve results compared to the

initial coarse grid solution

Potential:

Improved indicators, e.g. ongoing work based on adjoint-solutions can make

the adaptation more efficient, using the existing infrastructure for local

refinement.