I am trying to model a 3D cylindrical fluidized bed. However, due to the dimensions of the bed, 1.5 m x 0.152 m , and the small grid cells to be used (0.008 m), the number of cells is becoming very large, more than 1 lakh. My computer is an old IBM Intellistation, so that the simulations are running very slow. How to make the simulations run faster without compromising the physics...i.e. maintaining the 3D features (no axi-symmetric case). Please suggest .

Since it's cylindrical you could use a 1/4 of your domain (using periodic BC). This way it's still 3D at least, and you cut your mesh by 75%.

Hope this helps, and good luck, Jason

Dear Jason,

Thank you very much for your insights. However, since a fluidized bed is a very transient system with a lot of heterogeneity, can this simulation result be used in confidence to apply for the whole bed ? I mean, Does is it simplify by compromising the physics ? This is important because the acceptibility of the results is an important issue.

I assumed that there wasn't any major difference from one quarter of the model to the next (like there isn't an inlet coming in on one side an not any of the others or anything like that... I simply picked 90° because it's a nice number and easy to model... you could just as easily do 60°, 45°, 30°, etc...). There are no simplifications on the equations or models that you implement in your model when using a periodic (or even symmetric model for that matter). The only assumption in a periodic model is that the flow on one of the periodic source faces is exactly the same as the flow on the other periodic source face and that any flow into one face is a flow out of the other. Another option is the symmetric model where the assumption is that there is no gradient across a symmetric BC (kind of like a wall but without the "no slip" condition). The benefit of a periodic model is that if your flow has a rotational component, the periodic model will capture it (because the flow will enter at one of the periodic BCs and exit at the other).

Basically what I'm trying to say is that the only simplfication of the model is that the "wedge" of the domain you have modeled is going to be the same as every other "wedge" of the domain. Other than that, all equations and models are left in tact (unlike the axisymmetric models which assume all values are constant over a full 2*pi rotation and therefore removes the 3rd dimension (theta in the axisymmetric case) from the equations).

It's a big assumption to assume that your model is periodic. But your other options are cutting down on mesh density, and there you'll have to make some assumptions as to where you can coarsen your mesh. If you expect variation from one quarter to the next, then the periodic assumption won't hold, but if your geometry is periodic, then your flow will most likely be periodic as well.

Hope this helps, and good luck, Jason

Hi Sanjib If the cylinder is curved (not a straight) the assumption of periodic or symmetric B.C. will not be valid because of the secondary flow in bends or curved pipe (cylinder) flow. Also waht about gravity effect, is it considered ?

Hope that can help best regards

Wael Aly

Hi Jason and Wael,

Thanks to both of you for very good insights. I think the best way to go about this is to try and see if things work by simplifying the geometry. As Einstein said, things should be made simple but not simpler. Going by it, if the essential physics is captured (a good example is the simulation of stirred tank with baffles), I don't mind. The geometry is simple straight,cylinder but the flow is highly heterogeneous and intermittent. This makes the application of any boundary condition other than wall, or inlet and outlet a serious look. I will get back to you both very soon about the outcome of the simulations. Thank you very much Jason.

hi try adaptive mesh before final run. it will reduce extra load on CPU.

good luck

Hi,

I have created the 1/4th section of a cylinder. This really have cut down the number of mesh. But,I am having problems withh setting the periodic boundary condition at the ttwo connecting faces. I have specified the velocity inlet, pressure outlet and the wall (the curved section of the cylinder). But the specification of the periodic boundary condition is giving some error. Can you describe how to specify it ?

Thanks for all your help.

Sanjib

Hi,

I would suggest a 2 level solution. On the first level simple model the flow in the surrounding of the bed, replacing the bed with a porous medium (resistance coefficents from the Ergun Eq.).

In a second step, take the results from the first simulation as a stating condition for a detailed bed simulation.

So you save the mesh outside the particles.

Hope it helps, Ralf