CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Siemens (https://www.cfd-online.com/Forums/siemens/)
-   -   Wind Tunnel Contraction Modelling (https://www.cfd-online.com/Forums/siemens/74256-wind-tunnel-contraction-modelling.html)

 josh_k March 28, 2010 14:20

Wind Tunnel Contraction Modelling

I'm using STAR CCM+ to validate and optimize the design of the contraction in a wind tunnel, that I'm designing as a Final Year BE project.

The Wind tunnel:
The tunnel is a blower type with a centrifugal fan at the upstream side. A settling chamber with a honeycomb and six screens (last screen is 40 mesh x 38 SWG or 0.1524mm wire diameter).

The contraction curve uses a fifth order Bell-Mehta polynomial biased by an exponent (alpha)to control the curvatures at the inlet and outlet. Ref: "Transformation of a Polynomial for a Contraction Wall Profile" by Daniel Brassard

The test section has a 300 mm square cross section with a maximum velocity of 30 m/s.

Contraction:
Initially, a contraction ratio of 9 (outlet size: 900 mm square) and a length-to-inlet ratio of 0.8 was used with alpha as 1.0 (equal radius at inlet & outlet). A 45 degree corner chamfer of 25 mm reduces the tendency for secondary flows.

A 5% developing length between the last screen & contraction entry along with a 150 mm developing length after the contraction exit was provided. A 1/8th symmetrical model (imagine the volume between the centreline and a diagonal of a square) was exported as a parasolid and imported into STAR CCM.

Models & Boundary condtions:
A "velocity inlet" (30/9=3.33 m/s) with atmospheric "pressure outlet" was used while the wall and fillet boundaries were specified as "wall" type. The two remaining boundaries were set to "symmetry plane".

For simulating an inlet turbulence level of 2%, the k-omega (SST) turbulence model was used after calculating the value for "k" and the length scale equal to the mesh size of the last screen was used for specifying "omega".
This gave k= 0.006 667 and omega = 117.38

Meshing & Results:
Models used: surface remesher, polyhedra with prism layers

Initial attempt:
Base size: 17 mm (The smallest length, viz. fillet size)

No. of prism layers: 10
Prism layer thickness: 15 mm (absolute)

Surface size: 50% (target) & 20% (minimum)

All other parameters were at their default values.

No. of cells: 112261

Surface integral plots for velocity and average velocity at inlet & exit leveled out completely, close to their expected values in about 50 iterations.
Residuals leveled out (with small oscillations) at about 400 iterations.

Second attempt (checking for grid independence):
Base size: 13 mm

No. of prism layers: 12
Prism layer thickness: 15 mm (absolute)

Surface size: 50% (target) & 20% (minimum)

Polyhedra: density: 1.2, growth: 0.8

All other parameters were at their default values.

No. of cells: 298339

Surface integral plots for velocity and average velocity at inlet & exit leveled out completely, close to their expected values in about 50 iterations.
Residuals leveled out (with small oscillations) at about 400 iterations.

The problem: When the axial velocity plots were checked at various planes normal to the axis of the contraction, the velocity at few points near the boundary layer edge had velocities close to zero with their neighbours following the general profile.

Therefore, in the next step, the number of prism layers was increased.

Third attempt:
Base size: 17 mm (The smallest length, viz. fillet size)

No. of prism layers: 24
Prism layer thickness: 15 mm (absolute)

Surface size: 50% (target) & 5% (minimum)
Custom surface size for fillet: 10% (target) & 7% (minimum)

Polyhedra: density: 1.0, growth: 1.0

All other parameters were at their default values.

Surface integral plots for velocity and average velocity at inlet & exit leveled out completely, close to their expected values in about 50 iterations.
Residuals leveled out (with small oscillations) at about 400 iterations.

Same problem reoccurs.

I've described this as well as I can and will post the plots when I get back to my lab. If I have left out any data, please let me know.

I'm a CFD novice and would like to know if the meshing/physical model parameters set for the problem are correct or not.

Also, is wall shear stress and skin friction coefficient a good indicator of separation?

Lastly, how do I compare which design gives the least turbulence at the exit? I tried obtaining the turbulence intensity from the "turbulent kinetic energy(Tke)" using a User-defined field function but could not compile the function.

Your feedback would be deeply appreciated.

Thank you
- Joshua Kanthirao,
Don Bosco Institute of Technology, Mumbai

 CharlieSeviour April 18, 2011 16:21

Hi,
Thanks for writing out your problem so nicely, you have helped me solve my problem.

My solution was to make a scene with showing the turbulent kinetic energy. Then express the turbulence level I was interested in (10%) as a level of turbulent kinetic energy. This is easily done as the turbulent kinectic energy, k, equals:
3/2*(UI)^2
where U is the velocity and I is the intensity.

If your interested in the sorrounding work, then check out my thesis on my web site:

www.charlieseviour.co.uk

 All times are GMT -4. The time now is 23:46.