To many professional analysts with a background in Computational Fluid Dynamics, most electronics cooling applications look pretty straightforward. There aren’t any complex physical processes to model like combustion (well hopefully not), high pressures or high temperatures (again, hopefully not) or high speed flow. Typically there are no free surfaces or multiphase flows, and no moving geometry. So, nothing really challenging from a CFD perspective then?
Um… well not exactly. All of the above are true. Well, OK some are only partly true – fans rotate and heat pipes involve phase change, but these can be handled without modeling the fan’s rotation or the phase change directly. So why not use a general-purpose CFD tool? Should be a piece of cake, right?
In theory general-purpose software can be used, but as the saying goes:
“In theory there’s no difference between theory and practice. But in practice there is”
The short answer is ‘productivity’. Electronics products are characterized by extremely short design cycles, requiring exceptional productivity across the CFD process. The challenge is to compress the most time consuming steps in the traditional CFD process – typically geometry preparation and meshing.
General-purpose CFD tools are often used late in the design process. The starting point is to prepare manufacturing CAD geometry for the analysis. As an aside, such CAD geometry is sometimes wrongly referred to as ‘dirty’ and so needs to be ‘cleaned’. Actually it was just created for a different purpose and so needs to be re-worked to make it suitable for analysis.
Once prepared, the geometry has to be meshed. Almost all general-purpose CFD tools use a body-fitted mesh. This is created by first generating a 2D mesh on the surfaces in the model, from which a fully unstructured volume mesh is generated. This is easiest when dealing with relatively few non-abutting objects. Meshing is often the most time-consuming part as both mesh quality and density have to be controlled. Meshing is a BIG subject in CFD. There many techniques and even more software solutions out there, which (rightly) suggests it’s not trivial.
The process above often uses different tools for geometry preparation and meshing. Once created, the mesh is imported into the CFD pre-processor. The pre-processor applies boundary conditions to cell faces and volumes within the mesh to define walls, inlets, outlets, heat sources etc. The process works well provided the geometry is not expected to change, or not change significantly, as small changes can often be accommodated by distorting the mesh. Otherwise, every time the geometry changes the mesh has to be reconstructed and the boundary conditions reapplied. Ouch! – geometry change is generally what product design is all about.
From the outset, electronics cooling CFD tools took a radically different approach to the way geometry is handled and the way the model is meshed – more on that in later posts…