[DanielMarques]

Hi

I am simulating the use of a phase change material with low temperature melting point. The idea is to place a slab of pcm around a pipe of an heat exchanger so i did a 2D model with a circle inside a rectangular shape of PCM.

For the PCM i used piecewise-linear properties at density, thermal conductivity and specific heat. For the liquidus and solidus temperature i inserted a 0.5 tolerance from the real point. For example 251.5 and 252.5 Kelvin for a melting point of -21 degrees.

On boundary conditions I used adiabatic walls at top and bottom of the rectangle. At the outside wall I placed temperature fixed at 298K since the material is polyurethane and I added a thickness of 0.09m. At the inside wall of the rectangle I used convection BC with 255 Kelvin, an h value of 5 and a very small thickness of aluminium. The pipe of the heat exchanger BC was coupled.

When simulating for a time step of 1s I have sometimes the problem of floating point exception. And the simulation is already taking a lot of time.
I also tried adaptative time step using 0.01 of tolerance with fmin=0.9 and fmax=1.5 but the simulation is taking too long...always using time-steps around 10-3. I will need months for a single simulation without beeing certain if the floating pointe exception error will happen again or not.

I tried to use a refined enough mesh and the levels of skewness on the metrics are good i guess since the max value is 0.6 and the average is very low.

Many times this model is reaching a situation of floating point exception. Can anybody help?

[vinerm]

What is the domain size and mesh resolution? 1 s could be very large time-step until the flow is laminar. Furthermore, a finer spatial resolution requires a finer temporal resolution. So, if you wish to use larger , increase the size of the cell in mesh, i.e., reduce the mesh resolution as well. Furthermore, since there is no inlet or outlet, you need to enable gravity and correct value of operating density. Else, there will only be conduction.

[DanielMarques]

The domain size is a rectangle of 100 mm * 7 mm. In the center there is a circle meaning the pipe of the heat exchanger. It is tangent to the rectangle. Usually i consider it at a fixed temperature below the melting point (if solidifying) or I just let it be a result of the heat conduction when simulating melting.

The mesh size is

Level Cells Faces Nodes Partitions
0 63749 128703 64953 26

Gravity is on.
And the model of viscous is laminar

So your advice would be to reduce the number of elements? having a rude(gross) mesh won't I have the same problem of floating point exception?

[vinerm]

It appears to be rather fine. Floating point exception might be because of precision. Are you using double precision or single precision? Use double precision.

[DanielMarques]

Double precision.

And I am also using parallel to use the maximum number of cores possible of the workstation.

[vinerm]

Then, the error is related to numerics. Somewhere some value is becoming very small. Does Fluent report hitting any limits, such as temperature of 1 K or 5000 K?

[DanielMarques]

I guess not. At least I never noticed something like that.

Someone once told me something about interfaces between the heat source inside the circle and the area of PCM. Something that I should account when doing the mesh and geometry. Consider one part as a solid and other as a fluid, have an interface on the wall between those two regions, would allow to fluent a better calculation of the heat process.

I also saw something about lefting some space for PCM to expand since I am considering the change of density value when the phase changes and the volume will change.

Those are two things I never explored too much. Does any of it makes sence for you?

One thing is for sure...floating point exception error is always occuring when some of the material is in the range of temperature where change of phase occurs. When completly melted or solidified, usually works fine.

[vinerm]

That's correct. If the density is being modeled using any method other than Boussinesq, then you need to provide extra space for expansion and extra material for contraction. Else, it will lead to divergence.