Learning Forum

[jmwarre2]

Hello all,

I'm having convergence issues when simulating steady-state natural convection between a small cylinder inside of a box and the air inside. The geometry is fairly simple, I'm really doing this as an exercise to see how much energy is transmitted through the different sides of the box and what the steady-state temperature rise in the box is, but I have trouble meeting the convergence criteria for my simulation. My criteria isn't overly restrictive, and when I view results after a non-converged simulation, results are clearly unrealistic. I seem to be making all the appropriate analysis choices, but still have trouble.

How big of a factor is the mesh quality/configuration in my convergence issues? I seem to think my mesh is the issue, but even when I refine cells around the flow boundary layers, it seems there is no difference in results.

Parameters relevant to my issue are as follows:

• Cylinder is 2" long, 0.5" diameter, located inside a 10" x 10" x 10" steel box


• Cylinder generates 17 W of power


• Fluid inside box is air at atmospheric pressure


• Ambient air outside of box is constant 60 deg. F


• Flow is laminar

Thanks for any advice/experience!

-Justin

[Rahul Kumar]

Hi Justin, 

What is the density method you using? 

[jmwarre2]

Hi Rahul,

I'm using a pressure-based solver and a coupled pressure-velocity method with body force weighted selection under spatial discretization. Also I'm selecting the pseudo transient option.

-Justin

[Karthik Remella]

I think what Rahul is asking was - what is your air density? And what are your operating conditions?

Thanks.

Karthik

[jmwarre2]

Apologies Karthik, thank you for clarifying. I'm using an ideal gas density formulation with a specified operating density of 0.

Thanks

-Justin

[Karthik Remella]

Since you say it might be the mesh - what is your mesh quality? 

Also, what do your flux reports show? Are you seeing a good flux balance? Could you share a screenshot?

What are your convergence criteria?

Thanks.

Karthik

[jmwarre2]

Hi Kremella,

My average element quality is 0.48, aspect ratio is between 1.1 and 120 with average of 12. My convergence criteria is the default criteria selected by fluent when initially opened. My flux reports show what I think is a good balance. The % difference between heat generated in the cylinder and heat dissipated through the steel box is very small (< 1%). A couple of screen shots are below.

Thanks,

Justin

[Karthik Remella]

Even though your residuals are not low, if your heat fluxes are balanced and your monitors are steady, you should be able to call the simulation as complete and use the results to analyze the physics.

Is this the case?

Thanks.

Karthik

[Rob]

Also plot monitors of velocity & temperature on points above the hot object.  Natural convection flows tend to be inherently transient so residuals rarely drop below about 5e-3 or so as the buoyant plume moves around very slightly. 

[jmwarre2]

Thank you Kremella, rwoolhou, I believe you are right. As I review my flux reports, percent error is very small (<0.2%). When I change the temperature range on some of my contour plots, results begin to look more realistic.

Thanks for your input!

-Justin

[Karthik Remella]

Excellent!