OpenFOAM: COARSER MESH AND ADJUST TIME STEP

OpenFOAM: COARSER MESH AND ADJUST TIME STEP

OpenFOAM version: 9

Case: clamp_09

Points of note: new coarser mesh, adjust time step implemented, 60 minute simulation.

NEW MESHES

There is a need for more speed in this simulation. It is currently moving at about 0.1 seconds per second. So a 3 day simulation would take 30 days. To achieve this increase in speed, a coarser mesh will be used. The cover has been increased to 150 mm in thickness and the number of cells in cover reduced to one. The cells in this area are then widened so that they are basically square. The from that, the surrounding inner domain and the internal clamp area have also been modified to ensure there are no split cells at the cover interfaces. The outer domain was also reduced to be just 8m above the clamp (previously 10m). The final meshes are ca. 5000 cells for the coarse mesh, and ca 2800 for the super coarse mesh. These are down from ca. 16000.

15cm is not far off the average diameter of a fully grown sugar beet, so it feels a good cell size. There are still some smallish cells inside the clamp area that might be causing problems, but I don’t think so – the highest Courant numbers are probably in the area outside of the clamp, where the wind is blowing unrestricted. There are also still some elongated / higher aspect ratio cells, but again, probably not an issue.

The cost to accuracy from this increased coarseness is not known yet, but the model needs to be able to run a simulation of around 3 days of weather. If this takes 3 week of simulation time, then it is of no value.

RESULTS

These are the same uncovered clamp (9 m wide, 3 m high), with 5 m/s inlet velocity (from left). The inlet/ wind temperature is constant at 300 K, and the clamp is initially 274 K. The only difference is the mesh.

But I’m not posting any animations: something is wrong.

The results seem pretty similar on the two meshes, which is good. But the problem is that the fields look the same at 3600 seconds as they did at 60 seconds, which is bad. With such a big temperature difference between the two phases, I would expect that they would become much more alike over an hour. Or, more likely given the inlet temperature is constant, that the porous sugar beet region would be a lot, lot warmer at 3600 seconds than at 60 seconds. With the previous simulation of 30 or 60 seconds, there were hints that was an issue with the heat transfer, it was not possible to judge properly given the short time span. But now it appears very much so that there is an issue.

On the bright side, things have sped up a lot. On the rather slow desktop I’m using, the 3600 second simulations took 8050 seconds for the coarse mesh, and 3435 seconds for the super coarse mesh – so better than real-time. With max Courant number set to 0.6, the coarse mesh was proceeding at a delta t of 0.00676, and the super coarse mesh at a delta t of around 0.011 seconds.

My first guess at the probable source of the issue is the huge difference in density x specific heat of air verse the beet mass. If div.T = 26, the specific surface area is 30, and the convective heat coefficient is taken as 10, then we have transfer of 7800 W/m3. The change in temperature is then this value divided by (density x specific heat). Which (density x specific heat) is it meant to be? Those specific to each phase? That’s what I’m using. (density x specific heat)f = 1253, (density x specific heat)s = 4.145 e06. So, with the 7800 W/m3, the temperature changes are, fluid = 6.23 degrees per second, and solid = 0.00188 C/s. The air is cooling instantly when hitting the porous zone, while the beets aren’t heating at all, so there just isn’t the expected changes. Is this the problem?

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.