Sizing a heat exchanger does not require CFD. The effectiveness-NTU (ε-NTU) and LMTD methods give overall performance quickly and reliably once the surface heat transfer and friction characteristics are known [1]. These methods rest on experiment and, within their range of validity, are more reliable than CFD. See the note on when CFD is necessary.
The real question is where that range ends.
The silent assumptions behind ε-NTU
The method rests on a few assumptions, and in industrial equipment they frequently do not hold:
- The flow divides equally between channels. In reality the header design makes the distribution uneven.
- The heat transfer coefficient is uniform over the surface. In the entrance region and at reattachment points the local value is a multiple of the mean.
- The surface family matches the one the correlation was derived from. A new fin geometry or an unusual pitch-to-height ratio breaks this.
- Fouling is evenly distributed. It is not; it accumulates in the low-velocity regions.
Where calling on CFD is worthwhile
The distribution problem. Seeing how the flow divides between channels. This directly reduces overall performance and is something the correlation cannot see at all. Changing the header geometry to improve distribution buys performance far more cheaply than adding surface area.
New surface geometry. Extracting the j and f characteristics of a fin that has no correlation, by solving a periodic unit cell. The whole device need not be solved; a single repeating unit suffices, and that makes the computation an order of magnitude cheaper.
Local problems. Dead volumes, freezing risk, temperature gradients that cause thermal stress, vortex shedding that excites vibration. None of these appears in a mean value.
Conjugate behaviour. In compact devices where solid conduction competes with the resistance on the fluid side. See the conjugate heat transfer note.
A practical workflow
The order is: size it with ε-NTU first, then test that sizing's assumptions with CFD. If the two differ by an order of magnitude, understand that first — in our experience the error is usually in the CFD setup, not in the correlation.
If you are extracting a new surface characteristic, that result is not a calculation but data, and it has to be validated; at the very least, run the same method on a known surface and compare against a published j–f curve. See the verification and validation note.
References
- Shah, R. K. & Sekulić, D. P. (2003). Fundamentals of Heat Exchanger Design. John Wiley & Sons. DOI: 10.1002/9780470172605
- Celik, I. B., Ghia, U., Roache, P. J., Freitas, C. J., Coleman, H. & Raad, P. E. (2008). Procedure for estimation and reporting of uncertainty due to discretization in CFD applications. Journal of Fluids Engineering, 130(7), 078001. DOI: 10.1115/1.2960953