The 3D pipe junction

The Cube

the cube was our first contact with ICEM. It enabled us to learn the basic functions of the DDN tool that is used for the definition of the geometry. It also helped us to be familiar with the DDN tetin part. DDN tetin  is a sub program of ICEM where families and domain are defined. The families will then be used as the diffrent types of boundaries, while the domain is the fluid (i.e. the space to be meshed).

The cube is also a good way to check the correct use of the geometry by Fluent and Star-CD. As the geometry is very simple, the solution is known and any problem is easy to detect.

More more details refer to H. Neau's manual.

The 3D pipe junction

The 3D pipe junction is a tutorial of ICEM. The geometry construction, as well as instructions for the various meshes are included in the ICEM manual.

The main dimension are: length !50 m , height 100m, big cylinder 50m radius inlet, small cylinder 20 radius outlet. The elements usually have a dimension of 10m. The number of elements per mesh is around 10 000.

We first build the geometry with DDN.

Once the geometry was completed we used DDN tetin to assign the differents part to families.

Our first mesh was a standard hexa mesh. The first thing to do for an hexa mesh is a blocking. By doing this we define differents rectangular blocks that will coorespond to the geometry. ICEM uses these block to divide the space into various elements and then project the edges of the block to obtain the real geometry. Once the blocking is done an element size can be set. The mesh is then automatically computed to have the following results.

Detail of the inlet or study of the mesh quality show that by using this meshing method some problems might appear.

The idea is to use an O-grid that is more addapted to round shapes.

We also tried the tetra meshing of ICEM. Meshing with tetra elements has as downside that corners are not properly treaten, they are "cut". Besides the mesh did not work with Star-CD.

All meshes except the tetra with Star-CD worked without problem for the computation. We ran a small number of cases. The following images were obtained using the O-grid and our standard computation ( steady-time, k-E turbulence with an inlet velocity of 3 m/s (Re=3.10e8).) with Fluent 5.

This image shows the velocity vectors (color by velocity magnitude, color range 0 to 23m2/s)

We made plane in the middle of the pipe to see the contours:
We take a look at the pressure repartition. Color range is -90 Pa to +400 Pa:

As the model is K-E, we can look at the value of K. Color range is 0 to 1.5m2/s2