Hello, I have only started using GAMBIT and FLUENT for around 2 weeks now. Forgive me if this question may seem irrelevant but I was constructing a model consisting of several volumes (basically, a complete ventilation unit with individual vents, louvres, partitions etc.) and the objective was to analyse a 3-D flow around and inside the unit. According to the help guide found in FLUENT, it is normally the case to make everything as one volume (is this really true?).

However, once I ended up with a single overall volume (by means of the unite, subtract, intersect functions), meshing was almost impossible due to the fact that the previous individual volumes having such a complex geometry hence producing meshes within a figure of approximately 5 million cells (which is the maximum for the server here). This is what I think is the cause; if this is not the case, I appreciate it if someone could correct me.

I have been asking around some seniors and one method they have told me to try is to just leave all the individual volumes as they are, mesh them as normal and then when the time comes to export to FLUENT, define all of them as WALLS within the overall 'big' volume. The reason for doing this, which I have been told is that the software will interpret these walls as solid/obstructions and so force the flow to go around them rather than through the individual volumes.

This is where I find myself confused. I did exactly that but the flow just passes through as if the 'big' volume was empty to begin with. Is this method valid to begin with? Is there another method besides this and subtracting. I have reconstructed and simplified the model several times but leaving it as a single volume almost yields no favourable result as all sorts of errors pop up during the meshing phase due to the complex and non-uniform geometries being cut out.

I appreciate help or advise if any. Many thanks.

You write: everything has to be one volume. What do you imagine instead? Seperate disconnected volumes? How should the flow be related between them? I would say: yes, your right, only one volume.

Concerning number of cells and complexity of geometry: it is a fact that there are limits as to what can me modelled with limited computational resources. Maybe you really touch or (try to) cross that limit.

Some general remarks about meshing that you probably already know: Gambit first meshes edges, then faces, then volumes. To represent the geometry to some degree, there have to be at least some mesh nodes on each edge. So if you have a wide variety in the distribution of lenghts of your edges, I would expect some problems with mesh quality, not with the possibility to mesh the model at all.

What I would recommend is to consider 'size functions'. They give you the possibility to make the mesh finer in areas where the geometry has finer length scales and coarser in areas of wider space. By this you can possibly save some cells somewhere. Of course this has to be balanced with the problem definition. Maybe you really need that big number of cells to answer your question. I am only talking about meshing here.

Another point to consider is the meshing 'scheme'. For 3D models what meshes almost everything is Elements=Tet/Hybrid, Type=TGrid

I hope I did not confuse you too much. It is not long ago that I myself also started with Gambit/Fluent. Good luck.

One of the benefits of modern 3D pre-processors is volume composition and decomposition.

The decomposition is easy to explain... you have a large complicated volume, and you break it apart into smaller, more manageable volumes. In Gambit, use the "split" function with the "connected" option turned on (and turn off bi-directional and retain if they are on). You can split a volume with faces or other volumes. Make sure the connected option is on though... lets say you split a cube with a smaller cube centered in it... it will create two volumes (a small cube, and a large cube with a cube shaped void in it). Instead of creating 6 faces for each volume though, it will only create 6 faces that are shared by both volumes. This way, when you mesh the face, it's now a source face for both volume meshes. And when Gambit exports this mesh, it will ignore "connected" faces, so that you get a continuous volume mesh and those "connected" faces won't even show up when you import the mesh into Fluent.

Now volume composition is related, but a little harder to explain. It's a way of combining the volumes to create a continuous control volume, without having to unite the volumes, which tends to create large, unwieldly volumes. You have to manually "connect" the faces between the volumes. You have two volumes that are coming together, use the "connect" function under the Surface Geometry Tools (it looks like two ends of a black plug being put together) on the two faces where the volumes touch. This will create a single "connected" face that is shared by both volumes, and it will follow the same meshing and export rules as if you had split the volume with the "connected" option turned on. One word of caution though... when you "connect" two faces, they have to have complimentary geometry... which means they have to have the same number of vertices and edges that match up within Gambits tolerance (of 1e-06). If one face is larger than the other, then you're going to have to split the larger face so that it is made of two faces (one of which matches the face on the other volume). If the faces are the same size/shape but there are a different number of vertices, then you'll have to split edges to get the right geometry.

The split and connect functions are very powerful tools. To be honest, Gambit is such a weak meshing tool that I would've probably ripped out my hair (well, what's left of it) a long time ago without these functions. It seems to be a sticking point when people are learning to use Gambit, but it's very important to learn, and I highly recommend playing around with them to make sure you understand what is going on, and what the different options do.

Once you have your volumes set up properly, then sizing functions help you control your mesh. The tgrid meshing scheme in fluent is very questionable in my mind unless it's being controlled by the sizing functions. Also, if you can, use structured mesh (quad mesh) as much as possible. For the same edge length, a quad element uses more of your volume than a tet element, so it typically takes less quad elements to fill a volume (I say typically because it depends on meshing technique if there are large mesh size differences withing the volume). Also, you can usually get away with a slightly larger (longer edge length I mean) quad element than tet element because they are more reliable and accurate in the solution. Structured meshes take longer to set up, but if you plan your mesh in advance, you can usually pull it off by decomposing your volumes properly. Also, you can have quad in one volume, and tet in another. Mesh the quad volumes first, and in the tet volumes Gambit will create pyramidal elements to make the transition.

Hope this helps, and good luck, Jason