Sebastien MASSART



Theorical aspects


Return to all reports


1.1. Physical problem.

Let's consider a viscous fluid which can transfer heat contained between two plane surfaces which are at different temperatures.

We will try to find what kind of motion can appear.

back to index

1.2. Condition of motion.

If TUp>TDown , then there is no motion and the temperature distribution is linear.

If TUp<TDown, then 2D stationary contra-rotative rolls appear.

back to index

1.3. Mecanisms and criteria of instability.

Let's take the example of a fluide which is heat by the down plane.

There are three different forces which act on particules :

The balance between this three forces introduces the stability condition on the Rayleigh number :

back to index


2.1 Geometry

We choose to illustrate the Rayleigh Benard instability in a spherical box.

The dimensions of our cylindrer are a diameter of 4 cm and a hight of 1cm.

back to index

2.2 Different kinds of Mesh

We wanted to know the influence of the mesh because there are many difficulty to adapt a grid to a circular aera.

First we used a BFC mesh without adjusting mapped nodes.

Then we tried to specify the weight of the eight corners of the initial cube.

We began with increasing the weight, and then with decreasing it.

Finally we choosen an unstructured mesh.

Normal BFC mesh

BFC mesh constrict in the boundary

BFC mesh constrict in the center

Unstructured mesh

back to index

2.3 Mesh dependance

We have done the same experience for each mesh at Ra=8021.7.

2.3.1 Normal BFC mesh

This mesh does not converge. So we have to play whith underrelaxation

So we can obtain the following results :

X-plane velocity

X=Y-plane velocity

Z=h/2-plane velocity

Z=h/2-plane temperature

back to index

2.3.2 BFC mesh constrict in the boundary

The problem of this mesh is exactly the same as the last case.

X-plane velocity

X=Y-plane velocity

Z=h/2-plane velocity

Z=h/2-plane temperature

back to index

2.3.3 BFC mesh constrict in the center

This mesh is the most adapted because the solution can converge without any particular condition.

X-plane velocity

X=Y-plane velocity

Z=h/2-plane velocity

Z=h/2-plane temperature

In fact the results are non axisymetric because of the BFC mesh which is based on a tranformation of a cube .

back to index

2.3.4 Unstructured mesh

This mesh can resolve the problem of the last meshs.

    X-plane velocity

X=Y-plane velocity

Z=h/2-plane velocity

Z=h/2-plane temperature

This mesh make the solution converge but it does not resolve the problem of an axisymetric results.

back to index

2.4. Critical Rayleigh number

We tried to determine the relation that exists between the Rayleigh number and the geometrical factor radius/hight in the case of the cylinder.

A experimental graph gives a relation we wanted to find with Fluent.

First, we determined the value of the critical Rayleigh for h=2, the factor used in the report (0.02/0.010).

We found Rc~1765. This value is not so different of the theorical 1708...

The following pictures are the result of many different simulations for many different values of the temperature gradient.

You can easily notice the value of the temperature gradient significant : the gradient of temperature becomes non linear.

h=2 dT=0.2oC

h=2 dT=0.21oC

h=2 dT=0.22oC

h=2 dT=0.3oC

h=2 dT=0.35oC

back to index

2.5. Example of 3D geometry : a piece of cylinder.

The aim is to observe the reaction of FLUENT with such a geometry.

We will be able to compare with the results we have in the case of the whole cylinder.

(the caracteristics are developped in the precedent parts of this report).

2.5.1. The GEOMETRY:

You can discover on the first picture the geometry of the "system" used for the simulation .

This is a simple part of a cylinder.

It is made of six surfaces, in order to simplify the construction.

This first picture comes from PreBFC.

back to index

2.5.2. The MESH :

You can see on these two others pictures the mesh of the geometry.

The particularity of this mesh : there is only one cell in angle-direction (the Y- direction in rectangular representation).

back to index

2.5.3. Results of some different simulations:

This image gives you a vision of the 3D results you can obtain with some degrees between the two plane surfaces.

TEMPERATURE vectors in a 2oC simulation:

We find nearly the same results as the whole cylinder. There are two contra-rotative rolls really significant.

This result is the proof of a realistic simulation according to the second part of this report.

VELOCITY in a 2oC simulation:

VELOCITY in a 2oC simulation:

VELOCITY vectors in a 5oC simulation:

With an increase of the difference of temperature, we observe a modification of the structure of the rolls.

In conclusion of these simulations with a part of a cylinder, we can say it is simple to simulate the Rayleigh-Benard instabilities with FLUENT, in many different geometries.

And the most important results are the same : we can easily observe the contra-rotative rolls...

back to index

2.6 Relation between Nusselt and Rayleigh numbers

We wanted to know if the relation between the Nusselt number and the Rayleigh number remains valid in this particular geometry.

This last image show that we have to separate two cases:

back to index

2.7. Determination of the critical number of Rayleigh depending on the factor radius/hight








Conclusion : in fact, the results are not precise because of the mesh which was not adapted to the geometrical dimensions.

Others simulations must be done to really evaluate the dependance of the Rayleigh with the geometrical caracteristics in the case of a cylinder...

Because of the time it takes to simulate the different situations, it has not been possible before the dead-line, but it would be very interesting to verify the experimental results.

back to index

2.8. Conclusion

It seems that Fluent can modelise 3D instabilities in a geometry more complexe than a simple cube.

However, the structured mesh necessitates an adapted geometry. Indeed we observed deformations of the rolls on the boundaries walls.

A larger geometry could resolve this problem.

Moreover, the mesh must be modified to make the solution converge.

This kind of study opens a large view of research.

back to index