Natural convection in a tank ,mass residual, ruler and autosave

TfG · March 31, 2009, 11:15

Hello,
I am new in this forum and in CFD.

I am simulating the unsteady flow of an oil tank submitted to natural convection (using bousinessq aprox.). I am getting different results with different meshes which I think should be fine enough: about 20.000-40.000 elements for a tank of 23x18m. I set 10-3 for residuals of continuity, x and y momentum and 10-6 for energy. I get difficult to reach convergence on each time step ( I set 30 iters per time step as maximum), I have tried with time steps from 1s to 50s. Following fluents manual I get that I should use 65 s or less for my case but I dont see results. Any insight that I could get? I am currently with a laminar model but I may use turbulent later on. I use Pressure-base solver, PISO, Body-weighted, second upwind and default under relax. Initial condition is a fixed temperature of 60ºc for all the tank (ambient is at around 0ºC and U values of walls are around 0.5-1).

I have read some paper that state to set mass residual of x% as a convergence criteria. I was wondering if this means the criteria applied in residuals for continiuty.

I also wanted to ask if there is any way to show any kind of ruler or something similar for contour or vector plotting.

Finally, I was wondering if it is possible to set autosave for depending on the flowtime and not on time steps (useful for adaptive time steps), but i cannot find anything like that. Is not possible to set a list with the flowtimes where you want to autosave the data?

Sorry for the length and thanks in advance.

Laci · March 31, 2009, 18:27

Quote:

Originally Posted by TfG

Hello,
I am new in this forum and in CFD.

I am simulating the unsteady flow of an oil tank submitted to natural convection (using bousinessq aprox.). I am getting different results with different meshes which I think should be fine enough: about 20.000-40.000 elements for a tank of 23x18m. I set 10-3 for residuals of continuity, x and y momentum and 10-6 for energy. I get difficult to reach convergence on each time step ( I set 30 iters per time step as maximum), I have tried with time steps from 1s to 50s. Following fluents manual I get that I should use 65 s or less for my case but I dont see results. Any insight that I could get? I am currently with a laminar model but I may use turbulent later on. I use Pressure-base solver, PISO, Body-weighted, second upwind and default under relax. Initial condition is a fixed temperature of 60ºc for all the tank (ambient is at around 0ºC and U values of walls are around 0.5-1).

I have read some paper that state to set mass residual of x% as a convergence criteria. I was wondering if this means the criteria applied in residuals for continiuty.

I also wanted to ask if there is any way to show any kind of ruler or something similar for contour or vector plotting.

Finally, I was wondering if it is possible to set autosave for depending on the flowtime and not on time steps (useful for adaptive time steps), but i cannot find anything like that. Is not possible to set a list with the flowtimes where you want to autosave the data?

Sorry for the length and thanks in advance.

Hi TfG,

- I would refine the mesh as a first step. Focus on the near wall region!
- 30 iteration isn't too much! If the residuals decrease monotonic, increase the iteration number.
- Contour and vector plotting: you can use Plot/XY plotting. Or Display/Contours, Display/Vectors. I don't know if this is you asked (i don't understand what do you mean about that ruler)

Good luck

TfG · April 1, 2009, 08:02

By ruler I mean some axis where you can read the distance magnitude (to know how big is certain cell for instance).

Other question I have is: when I do report->-fluxes-> Mass flow rate or heat transfer rate, what is the significance of the "default-interior" boundary? can be some how be related to residuals?

For the first iteration (0.5s) I get unscaled residuals on energy of around 10^5 and they dont decrease. Is this normal? I have a heat loss trhourh the boundaries of about 83000 W at the beginning. I have around 10 for continiuty and 1 for x and y (all of them unscaled) I was wondering if for my case it would be better to set convergence based on normalized residuals. What do you think?

March 31, 2009, 11:15	Natural convection in a tank ,mass residual, ruler and autosave	#1
TfG New Member Join Date: Mar 2009 Posts: 17 Rep Power: 17	Hello, I am new in this forum and in CFD. I am simulating the unsteady flow of an oil tank submitted to natural convection (using bousinessq aprox.). I am getting different results with different meshes which I think should be fine enough: about 20.000-40.000 elements for a tank of 23x18m. I set 10-3 for residuals of continuity, x and y momentum and 10-6 for energy. I get difficult to reach convergence on each time step ( I set 30 iters per time step as maximum), I have tried with time steps from 1s to 50s. Following fluents manual I get that I should use 65 s or less for my case but I dont see results. Any insight that I could get? I am currently with a laminar model but I may use turbulent later on. I use Pressure-base solver, PISO, Body-weighted, second upwind and default under relax. Initial condition is a fixed temperature of 60ºc for all the tank (ambient is at around 0ºC and U values of walls are around 0.5-1). I have read some paper that state to set mass residual of x% as a convergence criteria. I was wondering if this means the criteria applied in residuals for continiuty. I also wanted to ask if there is any way to show any kind of ruler or something similar for contour or vector plotting. Finally, I was wondering if it is possible to set autosave for depending on the flowtime and not on time steps (useful for adaptive time steps), but i cannot find anything like that. Is not possible to set a list with the flowtimes where you want to autosave the data? Sorry for the length and thanks in advance.

April 1, 2009, 08:02		#3
TfG New Member Join Date: Mar 2009 Posts: 17 Rep Power: 17	By ruler I mean some axis where you can read the distance magnitude (to know how big is certain cell for instance). Other question I have is: when I do report->-fluxes-> Mass flow rate or heat transfer rate, what is the significance of the "default-interior" boundary? can be some how be related to residuals? For the first iteration (0.5s) I get unscaled residuals on energy of around 10^5 and they dont decrease. Is this normal? I have a heat loss trhourh the boundaries of about 83000 W at the beginning. I have around 10 for continiuty and 1 for x and y (all of them unscaled) I was wondering if for my case it would be better to set convergence based on normalized residuals. What do you think?