Natural convection occurs as a result of buoyancy-driven flow caused by density gradients due to temperature variations. Typical applications of natural convection include electronic systems that are either vented or completely sealed. These devices generally do not have fans or blowers. Instead, they are cooled by buoyancy-driven flow that convects heat from heated components and conducts through the outer casing.
Natural convection within a sealed device is considered Internal. Natural convection around a device in a large enclosure or open environment is considered External. The analysis techniques for these two physical situations are a little different, and are explained in the Related Topics, below.
Natural and free convection flows are largely dominated by buoyancy forces. The buoyancy forces are generated by density gradients which vary primarily with temperature since pressure gradients are relatively small in these flows. Natural convection flows may be laminar or turbulent.
It is very important that a temperature be specified somewhere in the model (in addition to the known heat loadings). This can be an applied temperature boundary condition, but can also be the reference temperature for a film coefficient or radiation boundary condition. Without a specified temperature somewhere in the model, the temperature solution will not converge.
When defining the mesh for buoyancy-driven analyses, more elements will be required in the interior of the domain (away from the solid boundaries) than for a pressure driven flow. The reason is that accurate representation of the small density gradients is critical to computing the driving buoyancy forces correctly.
Use Mesh Refinement Regions to focus the mesh around critical areas. This is a very convenient way of transitioning the mesh from finer density to a more coarse density toward the flow region.
It is always good practice to ensure that a midpoint node exists on all objects with heat generation boundary conditions. This is most critical on thin objects such as heat sink fins and chips.
Some basic guidelines for setting up a natural convection analysis include:
Note that both flow and heat transfer should be enabled because the flow and thermal physics are coupled. The physics are considered coupled when the flow solution depends on the thermal solution. This occurs in natural convection because the density varies with the temperature. As the temperature changes, the density of the fluid changes, which affects the flow solution.
For natural convection analyses, property values must be allowed to vary (by selecting Variable) from the Material Environment dialog.
By default, initial fluid (air) property values are computed based on the lowest specified temperature boundary condition in the analysis. This is the desired behavior for simulating cooling of a device by natural convection.
For situations in which the device is heated by natural convection, it is better to initialize the fluid (air) properties based on the Environment temperature. To do this, enable the following flag in the Flag Manager:
Because a larger temperature gradient is required to cause buoyancy-driven movement in liquids, overall solution times can be reduced by first inducing a temperature gradient through the fluid prior to running the flow and thermal analysis. Do this by running 10 iterations thermal only (without flow). After a thermal gradient is achieved, flow and thermal should be run simultaneously.
While an external natural convection analysis is running, the temperatures will often initially climb quite high (because the air is still moving very slowly) and then will settle back down as the flow field develops. Natural convection analyses usually require more iterations than internal flow problems to reach a steady-state solution. The number of iterations required, and hence the total solution time, will be longer for a natural convection than for a pressure-driven flow analysis. Solution progression is slowed by the fact that buoyancy forces are generally significantly larger than pressure forces.
A flat-line convergence may not always be reached in natural convection analyses due to their inherently transient nature. Chaotic perturbations in the system will sometimes prevent “perfect” numerical convergence, but the trends should settle to within 5% change of the parameters of interest (velocity, pressure, temperature) over the last 20% of the analysis iterations.
Some techniques to follow if the solution slows or diverges include:
Review the “.sol” file found in your analysis directory to locate the problem.
Related Topics:
Electronics Cooling Best Practices
LED and Fluorescent Lighting Best Practices
Example of External Natural Convection