I work for Aavid-Thermalloy (formerly Nuventix) and we are currently developing an air-mover which ejects high-momentum puffs of air for the purposes of electronics cooling. Because the jets are highly-turbulent, lower deltaTs can be achieved for the same CFM compared to traditional fans. Details here: http://www.nuventix.com/technology/synjet-flash-demos/
Although our focus is currently on the development of the physical product, I have been building up our CFD modeling capability of these systems for the past few months. I am not interested in fully describing the physics here, as that would require a mesh density and simulation time step which would make any practical system-level model computationally prohibitive. Having built these coolers in a few different sizes and operating points, we have a wealth of empirical data to which we can tune our CFD strategy.
One of the ideas I'm pursuing is whether or not these jets can reproduce experimental thermal test results when modeled as steady flows (by artificially increasing CFM and / or turbulence intensity to match experimentally-measured deltaTs). My free-air system simply consists of a small heat sink and our XF30 cooler (pictured here: http://www.nuventix.com/telecom_cooling_l50-50/). I am currently modeling our cooler as a solid body (NOT modeling the flows inside the cooler!), and applying constant velocity inlet boundary conditions to each of the nozzle exit surfaces (see attached). Using the goal optimization tool (which is awesome, by the way), I've been able to tune the velocity necessary to reproduce the experimentally-measured deltaT. However, most of these coolers will be enclosed in some kind of box with air vents, so I drew a vented box around my free-air model, and here is where I encounter my problem.
I enclose my cooler / heat sink system within a cavity which has an open port at each end. I apply matching outlet boundary conditions near each nozzle to ensure that my boundary conditions are not magically creating or destroying any fluid volume within the enclosure. I have checked and double checked these boundary conditions, and the total inlet CFM matches the total outlet CFM on the cooler body. However, when the simulation converges, I find that the net flows through each end of the enclosure do not equal each other. Is it possible that a converged solution can predict non-zero net flow through a simulation cell, or am I handling my jet boundary conditions incorrectly?
Thanks in advance,