[Pigrenok]

Dear community,

I am completely stuck here and would appreciate any help.

I am supporting a student, who is trying to simulate the hyperloop train with the fan, which is "sucking" at the front of the pod and throwing it behind the train. He is using Star-CCM+ v14.

The geometry is as follows:


At the moment the pod is modelled as a simple conical frustum. The larger front circle is supposed to be the fan inlet and the smaller back circle is supposed to be fan outlet. The pod is located inside the large tunnel.

Due to the very restrictive timescale of the project, it is infeasible to model actual fan and pipework between the inlet in front of the pod and outlet in the back. So, the idea was to model the "sucking" effect in front. Initial analysis of the documentation has shown that the only realistic option without modelling actual fan is to use Blower Interface between the fan inlet and fan outlet.

The fan curve is a straight line with stagnation pressure at 0 volumetric flow rate and just around zero roughly at the target volumetric flow rate (with the negative slope).

Unfortunately, neither the student nor myself cannot get the blower interface to work even for fairly low ambient velocity. Instead of actually stabilising at specific volumetric flow rate and pressure rise, the volumetric flow rate jumps from -50 to 200 m^3/s with target volumetric flow rate at around 100 m^3/s. The pressure rise does not follow the provided fan curve polinomial. And we end up with uncontrolled rise of the outlet velocity (to hundred thousands m/s in a single cell and with back flow at the fan inlet.

Search on the Internet did not bring anything at all, StarCCM+ documentation describes the blower interface very vaguely, and there not a single tutorial regarding blower interface.

We would appreciate any help. Any more detailed documentation, tutorial or personal experience with blower interface would be highly appreciated. If you know exactly what we are doing wrong, that would be even better.

Thank you very much in advance for any help in this matter.

Best regards,

Vitaly.

[fluid23]

I am confused by the description and image shown. The pod is the gray tapered cone and the blower interface is where, exactly? You talk about fan inlet/outlet on opposite sides, but a blower interface exists as a plane.

If you can describe your domain and boundary conditions a little better someone will be able to help. I have some thoughts, but its still not entirely clear what you are trying to accomplish with this.

[Pigrenok]

Originally Posted by fluid23

I am confused by the description and image shown. The pod is the gray tapered cone and the blower interface is where, exactly? You talk about fan inlet/outlet on opposite sides, but a blower interface exists as a plane.

If you can describe your domain and boundary conditions a little better someone will be able to help. I have some thoughts, but its still not entirely clear what you are trying to accomplish with this.

fluid23, thank you very much for pointing at some lack of clarity.

I am sorry that the description turned out to be confusing. I believed I explained it fairly clearly. Apparently, I was wrong.

The pod is the grey cone, and the tunnel outline is shown as the black lines above and below the cone. It is going about 3 pod lengths upstream (to the left) and about 7 pod lengths downstream (to the right). The tunnel is a straight cylinder.

Star CCM+ has two fairly similar interfaces: fan and blower. Fan interface is a contact interface, it exist as a plane and model axial fan. Blower interface simulates radial fan and is an indirect interface, which means the inlet and outlet should not actually form a contact.

So, the upstream (left) end of the cone is boundary 0 (blower inlet) and the downstream (right) end of the cone is boundary 1 (blower outlet).

The tunnel upstream (left, not visible) end is a velocity inlet with currently 50 m/s set velocity. The downstream (right, not visible) end is the pressure outlet with set 0 Pa gauge pressure. Reference pressure is set to 100 Pa.

In the model stack coupled flow and coupled energy models are used along with ideal gas (compressible) model and RANS solver with K-epsilon turbulent model.

The fan curve is set by the polynomial: straight line with stagnation pressure intercept and negative slope (so that at around target volumetric flow rate the pressure rise will be approximately zero). Although we tried various slope values.

The simulation is steady (although we also tried to switch to implicit unsteady).

The problem is that initial several iterations the interface shows sensible volumetric flow rates, and the flow itself seems reasonable. But then at some point flow rate through the interface starts to drop dramatically, reaches negative values (showing at the same time negative pressure rise) then jumps back to positives, and pressure rise suddenly jumps as well. Then it starts to accelerate uncontrollably. At some point, pressure correction limiting kicks in. At this point pressure at the blower outlet is also limited to 1 Pa (minimum allowable pressure), and the velocity at blower outlet reaches values of about 7-8 km/s (kilometers per second) and then one of the cells in the region go unstable and reach local velocity of 100s. km/s. At the same time the flow at the blower inlet reverses and starts to go against the main tunnel flow.

Obviously, solution just becomes unstable and diverges. What we cannot do at the moment is to find the source of this instability and thus cannot eliminate it.

If anybody can point us in the right direction (either our error with this setup or better setup) or will show us any documentation (any better than the official one) or good tutorial on the blower interface (or on situation similar to ours) we will really appreciate it.

I hope it is a bit more clear now. Thank you very much in advance for any help and advice.

[fluid23]

Yes, that is clear. Thank you.

First things first... have you tried changing the pressure jump relaxation factor? That will improve stability with the the blower interface. You might also try a courant ramp if you haven't already.

Also, its a long shot, but you might try specifying a table rather than polynomial. They should behave the same numerically, but there could be a bug in how the operating point is set at the limits of the definition. Both should assume the closest value outside of the range you defined, but you never know...

If you have a fixed operating point and don't need the blower to vary its response to system pressure rise then you can try placing a target mass flow pressure outlet at the blower inlet and a mass flow inlet at the blower outlet with appropriate properties defined. Its a little less robust than the blower interface but will be more stable in a fixed condition.

[Pigrenok]

Thank you very much for your suggestions.

I tried to reduce pressure jump relaxation factor to even very low values (about 0.05). It does keep the simulation from diverging much longer, but it is still slowly accelerates at some point and it definitely do not even try to converge.

Courant ramp cannot help because first several iterations (20-100) the solution stays put (although not converging) and then starts rapidly diverging. Courant number ramp keeps the courant number low in the beginning and then starts to slowly increase it. So, it will get the courant number to the large number by the time it starts to diverge. But I tried it anyway just to exclude this factor and behaviour did not change. Interestingly, Star CCM+ itself tried to reduce Courant number when the pressure correction limiting kicks in.

Regarding fixed mass flow rate, it is obviously an option, but I would prefer not to do it because we would prefer to have calculated pressure rise and mass flow rate. The fan inlet is not receiving exactly far field evenly distributed inlet velocity flow, so, calculating the mass flow rate manually is an option, but we would leave it to the last resort.

With changing the polynomial to the table is an interesting idea. It felt like the system does not properly take information from the polynomial. I just did not finish this logical chain to the point that it is polynomial is glitchy, but possibly not table. I will definitely give it a shot and report the results.

Again, thank you very much for your help and suggestions. fluid23, I am wondering whether you used blower interface yourself and whether you experienced this kind of issues?

[fluid23]

Unfortunately, no. I do not use the blower interface. I use the fan interface and source (especially virtual disk) frequently, but not blower interface.

In fact, I didn't realize it existed until after my first reply. I thought you meant fan interface, which is very similar, but different topology. After you corrected me on the interface topology, I had to go back and educate myself on this one.

Fan interface will not do indirect topology, hence the early confusion on the interface existing on a plane.

I will be interested to see how you resolve this. Please report back.

One last thought... if a courant ramp will not work, what about just reducing courant? slow the calculation down to avoid the instability.

[fluid23]

Oh! Another thought...

How are you initializing this flow? If you aren't or haven't tried it, consider expert initialization. You are using a coupled solver, so you should be able to activate this option. I have successfully used it to get around early divergence issues in the past. It may be may favorite feature in Star-CCM.

That being said, it doesn't always play nicely with fan/momentum sources. Not sure if that would extend to blower interface, but its worth a shot if nothing else is working.

Last resort, if nobody here can help, is to contact your support engineer and open a help ticket through the Steve Portal.