Although a preliminary discussion of this topic cam up in another discussion, it was somewhat off-post and I thought it useful to focus the discussion on the best means to model enhanced heat exchange tubes (such as Turbo-Chill manufactured by Wolverine www.wlv.com) in Flow. Analysis of heat exchange in bare tubes without any internal or external enhancements is very straightr forward. The below example illustrates the effect of a bare tube with 2.5 GPM of water at 40 F. passing 1200 mm (47 in.) through a "tank" full of water at 100 F. :
With a bare 3/4" copper tube, as shown in the left image, chilled water increased in temperature from 40 F. to an average temperature of 44.8 F. The image on the right depicts a 3/4" copper tube that is enhanced with twelve internal spirals and "fins" spaced at approximately 24 per inch of tube length. With these enhancements the water heats up from 40 F. to 46.5 F. The simulation on the left takes about 10 minutes of processing time utilizizng four cores. The simulation on the right takes a little over ten hours of processing time utilizing 4 cores (initial mesh of 6). Typical commercial heat exhangers such as this utilize 200 to 500 or more such tubes and are arranged so that they will pass through the liquid 2-4 times or more, and with much less tempere difference than is shown above. To simulate a four pass arrangement there would be four such tubes connected in a serpentine manner, and each tube would be at least twice as long. I wouold estimate processsing time to be about 80 hours if enhanced tubes are used. Not insignificant when it is desired ti investigate a number of tube sizes and fluid variables.
Jared Conway suggested using a porous media to replace the fins on the tubes. To do this I modeled the porous media with a thermal conductivity about 4 time that of copper for starters. I did obtain a result that closely matched the result obtained witht he enhanced tubes:
I was able to converge the leaving water temperature of 46.5 F.so it would appear that the porous is a very good approximation.
thanks, Jared. The simulation time reduces to about four hours which is also helpful, but still scales to a large number.
Perhaps some of you might share the extent that you have scaled your results in Flow and still get a result that can be trusted.