Getting the most out of your CFD program [for electronics cooling]

Modern CFD codes dealing with air-cooling of electronics components have evolved to the stage where, properly handled, they can produce useful and fairly reliable estimates of coolant flow rates and component temperatures. For the most part, the outputs of these codes are "maps" of flow and temperature distribution. Useful as they are for identifying potential problems, these maps provide little or no guidance as to how to solve the problems they reveal. This paper proposes a set of diagnostic parameters that should be calculated for each critical component. From these parameters, a heat transfer person can tell not only whether an overheating problem exists but what is the most likely cause: low coolant flow, thermal stratification in the coolant, excessive heat conduction or radiation from hot neighbors, or low heat transfer coefficient. The recommended parameters are: (1) operating temperature, (2) adiabatic temperature, (3) mixed-mean temperature of the coolant, (4) net radiation and conduction input, and (5) the heat transfer coefficient, hadiabatic, and (6) the pressure drop. These can all be obtained without undue effort either by integration of values already available over appropriate areas or by "freezing" the flow calculation and sequentially changing the power distribution (relying on the linearity of the energy equation and the fact that hadiabatic does not change when the temperature distribution changes). The principle computational requirement for implementing this method is to develop a method for "freezing" the flow field and sequentially calculating the operating temperatures for different thermal boundary conditions. Techniques for doing this have been discussed by two previous authors. It remains only to complete the task and use that approach to calculate the diagnostic parameters.