How do I include minor losses in the system curve for a centrifugal pump?

[Scott27]

I'm trying to size a centrifugal pump for a closed-loop cooling system, but I'm getting stuck on the system curve calculation. My friction loss estimates from the Darcy-Weisbach equation seem okay, but I'm unsure how to properly account for the minor losses from all the elbows and the control valve in a way that feels practical for the final operating point.

[Savannah_H]

From my years on the shop floor I treat every elbow and the valve as a single K. I add those Ks to the friction head and solve h = fL/D * (V^2/2g) plus sum(K)*(V^2/2g). The trick is to pick the velocity that gives the target flow and then read off the pump head from the curve; the system curve starts to look sensible once you’ve tallied a few Ks from tables or from the valve data. I keep a small table of K values for common fittings and use an iterative loop to converge on Q and H.

[VioletYW]

I chased this once and nailed nothing for days. I assumed a big K for minor losses and the head predicted was way higher than the loop could deliver; once I split the losses into separate pieces and calibrated the valve K from Cv data, the numbers lined up a bit better, but I still had to throw away a couple of fittings that seemed to overkill the head.

[AndrewM]

Are you sure the real problem isn’t something more mundane like the valve Cv or the coolant viscosity changing with temperature, which would shift the curve more than the elbows?