26 Oct 09 14:09
There is a lot of science that a person could throw at this, but at the end of the day it comes down to comparing pump curves received against the operating range you want to cover.
If asked to be predictive up front, if someone permits me to be scientifically creative, I would start with the assumption that, for centrifugal pumps for example, the characteristic head-capacity curve is a concave downwards parabola of the form:
H = C1 + C2*Q + C3*Q^2
(sometimes I add a fourth term "... + C4*e^Q..." because I have found that equation form is a better fit for most of the curves I have worked with, and it reduces the drop-off and improves continuous rise to shut-off - perhaps coincidentally - near the dead-head range)
I then would arbitrarily pick a "% rise to shut off from BEP", considering how flat a characteristic I want. With that, I can usually get a pretty accurate indication of what C1, C2 and C3 are.
I then would pick an arbitrary pump maximum efficiency at BEP and, again, assume a concave-down parabola with vertex at BEP and Y-intercept at (Q=0, H=0).
Then, when I got the pump curves, I could plot the actual data points from each against the theoretical baseline thus derived and make an assessment as to which comes closest.
Where pump RPM and impeller sizes need to be considered, the affinity laws can easily be applied.
I often take the backwards approach as well: fit equations to actual pump curves and then vary the operating conditions to predict performance. There is some limited curve fitting capability in EXCEL, but I find that there are a few reaaly good - and basically free - Shareware and Freeware utilities that do a much better job. Two that come to mind are the old DOS-platform XYMATH and the somewhat newer CURVE-EXPERT.
The nice thing about the above approach is that once you have built the template the first time you can fiddle with the constants C1, C2, C3, C4 to "shape" the curve you want and then go looking for the closest fit from the vendors, so it's not a "one-time" tool.