[jayorayo]

Hi,

At the moment I'm trying to model the flow in a hydrocyclone but I'm having trouble resolving the air core in the centre. I was wondering if anyone has any suggestions in order to solve the flow correctly.

I'm using the Implicit VOF multiphase model (with Implicit Body Force) with the RSM model. The inlet to the cyclone is set as velocity inlet with the overflow and spigot set to pressure outlet with 0kpa gauge pressure (the operating pressure is set to 101325 Pa). The discretization schemes are as follows: Pressure=PRESTO, Volume Fraction=Modified HRIC, Momentum and Turbulence = First Order initially, switched to Second Order and then QUICK as solution progresses. The model contains approx 400k cells.

Initially I get a steady solution for just the water phase until a negative pressure air core has developed in the centre. Then I specify the back flow air volume fractions at both outlets as 1 and begin a transient solution in order to resolve the water and air volume fractions. The timestep is set to 1e-5 secs with the convergence criteria set to 1e-4.

I was expecting to see the air core develop gradually from the outlets until the met in the centre of the cyclone and a fully resolved air core was present. Unfortunately the air core only developed about a 1/3 of the distance from both outlets after solving the flow for 5 secs whereas previous studies have reported a fully developed flow under the same conditions in less than 1 sec.

Any suggestions or tips would be much appreciated.

Thanks

[Trev]

I've had similar problems to this before and with some cases the air core doesn't seem to progress much into the chamber no matter how long I leave it. Under certain conditions I have observed this experimentally aswell as the flow adapts to variations in the inlet velocity. Unfortunately the time scale required for this in combination with relatively low time steps means that the time required for the core to be fully developed will take an unfeasible amount of time.

Despite this it can be easily remedied by just making a register for the negative pressure core region then patching this with the air volume fraction. When you are changing from single to multiphase flow during the solution.

You can probably save time aswell by shifting the discretisation methods for momentum and turbulence from 1st order straight to QUICK or even start with QUICK if your solution is stable. This of course is going to be dependant on how stable the Reynolds Stresses residuals are as they can be troublesome when on 2nd order and greatly affect the stability of the solution.

Apart from that you have everything setup correctly by the sounds of it. Although the modified HRIC discretisation method for the VOF equation doesn't give the best possible interface resolution I've found it to be more than adequate for this kind of simulation so would stick with it.

Just out of interest are you using the pressure based or density based solver? Its just that I have done simulations like this using the pressure based solver and had to run the single phase flow as transient for the negative pressure core region to develop properly. But have done similar things for compressible air cyclones using the density based solver under steady state conditions and it was able to calculate the negative pressure core.

Neil

[jayorayo]

Thanks for the advice Neil, I'll try that out and let you know how I get on. To answer your own question, I've been using the pressure based solver so far but I might use the density based one and see what difference it makes to the results.

[jayorayo]

I managed to resolve the aircore successfully and produce some decent results. The problem turned out to be the quality of the mesh. Initially I had an a fine (~1M cells) unstructured tetrahedral mesh but as the cells were not properly aligned with the flow the model was unable to resolve the air core. I changed the mesh to a structured quadrilateral mesh (500k cells) in the body of the cyclone and made a vast improvement in the results. All other set up parameters were kept the same (turbulence model, discretisation, timestep size etc).