Learning Forum

[Greg Janes]

Hello,

I am trying to simulate the transient cooling of a flate aluminum plate from an initial temperature of 800K to the DPM liquid temperature around 300K using a flat fan atomizer dpm simulation. I am running into an issue around 1.6 seconds into the 5 second simulation, ~9250 iterations with 50 iterations per time step, the residual values explode causing a floating point error and a bunch of cells limited at 1K or 5000K. I have included a few figures which show the locations of the 5000K and 1K cells and the temperature gradient on the surface at the last time step. Up until this last time step the simulation is able to run and shows the plate cooling a rate that is expected. I have a minimum orthogonal value of 0.1 and average of around 0.8. I can answer any questions you have for me, but any help would be greatly appreciated.

[Greg Janes]

[jcooper]

 

Hi Greg:

I assume that this is a transient run?  If it is, the failure point will coincide with a certain point in the analysis, such as the spray just arriving at the hot plate.  This is the first thing I would try to pin down.  If the dpm particles are not at the plate yet, then the failure may be due to cht numerics. 

 

It looks like the environment is getting up to 5000 K.  This is very odd and may be related to the boundary conditions placed on the cube of air around the plate, or something to do with particles Something is producing this heating. Some questions to ask:

 

 

1. What is the initial temperature of the plate and environment?
2. What is the assumed backflow temperature of the boundaries of the cube of fluid around the plate?
3. What are the droplet temperatures as they move toward the hot plate?
4. What is the Cp property of fluid, particles and plate? (If any of these values are very low, even a little energy will bring things up to really high temperatures quickly.)
5. What is the solid time step?  (This is automated, but often defaults to a very high value that is not always suitable.)
6. As you run, do any of the equations show very high residuals?
7. Are there any turbulence, backflow at outlet or other warnings?

 

I would set up a solution animation that shows a temperature coutour of the particle plume and plate in cross section.  That way you can view the development of the solution in real time and get some clues as to the cause of the high temperatures.

 

Best Regards,

Judy

 

[Greg Janes]

Hi Judy,

Thanks for getting back to me.

1) The initial temperature of the plate is 800K with the environment at 300K

2) I just have pressure outlets around the cube of fluid, I didn't assign any backflow temperatures.

3) The droplet temp is 300K as well.

4) I am using an inert water liquid droplet, an aluminum plate, and standard air all basic for ansys fluent, I will have to check for these values if they are critical. I will share a link to a youtube video of the first 1.8s of the simulation, you will see that this isn't an issue at all until this random iteration. 

5) The time step is 0.01s, as I am trying to run it for 5s. 

6) From the figures shown in the post, the continuity equations are always greater than 1, never converging under that. Then at the 8000 iteration it blows up.

7) I am getting backflow warnings at the outlet bc and temperature warnings at the localized cells, I included an image of this.

 

Here is the link to the video of the spray quench for the first 1.8s

Thank you,

[jcooper]

Hi Greg:

If it is blowing up at a certain iteration, I would try to stop the run just before then and take a look at the solution in detail.  With particles, the isues can come from many things.  Unfortunately, with transient operation, there are not as many underrelaxations that you can do, so your only option is generally to reduce the timestep.  If it is possible to inject more representative particles for the spray, this may be beneficial too.

Expanding the box around the plate may help to stabilize the boundaries because conditions there are more likely to be predictable, (ie: ambient temperature with little flow).  A hemisphere type boundary shape may help to give a better mesh and prevents directional wierdness at corners.  Flow at the boundaries should come back in from nearest cell and not normal to the boundary.

I hope this helps,

Best Regards,

 

Judy