# Size Pumps Like a Pro

## Use a step-by-step process to avoid pitfalls.

My pump sizing calculations were dead-on. Current readings showed the pump was right where it should be in its curve. I followed a procedure I pieced together myself from magazine articles and what I could find in the "Cameron Hydraulic Data Book." It doesn't include the "viscous dissipation factor," which figured in fluid mechanics in school, but never showed up anywhere in these references.

Pump choice should start with the process and fluid. Define elevations, initial and final pressures, equipment pressure drops, minima and maxima, and, especially, fluid physical properties.

For a slurry, you typically need a 3–14 ft/sec (fps) velocity, depending upon the slurry. This affects the net positive suction head available and may require a recirculation line. Don't use orifice plates for slurries! The fluid may be non-Newtonian, necessitating special tests by the pump vendor. Pump speed is another concern with a slurry. Fluid erosion increases exponentially (at a rate related to speed3): high speed means high wear. Centrifugal pumps (CP) generally are available at 1,150, 1,750 and 3,550 revolutions per minute (rpm). You can cut rpm by spending some money for a gearbox and V-belt. Lowering the speed reduces the head available for CP.

The control valve eats the unneeded pressure.

Now, let's move on to selecting the type of pump. Always consider CP first because of their reliability and relatively low cost. However, they don't suit all services. If you're pumping 25 gal/min of slurry to a spray dryer at 2,700 psig, good luck using a centrifugal pump. Viscous non-Newtonian fluids at low flow generally require a piston or diaphragm reciprocating pump or a rotary pump. I favor piston pumps for such applications because their head is independent of flow. Many manufacturers provide charts summarizing pump range.

After selecting pump type, you'll want to define the operating curve. For a CP, that's the arc the pump makes along a curve below its impeller curve. There must be extra head to control flow. Choose a reasonable velocity — I suggest 10 fps — at maximum flow; with the discharge pipe runs estimated, use a table from "Cameron Hydraulic Data Book" to determine line pressure losses on a psi/ft basis. You can make a formal estimate on a spreadsheet.

It's best to rough out the control valve specification at the same time. The control valve will be required to "eat" unneeded pressure when a heat exchanger is clean or a filter is unplugged. The valve pressure drop is the difference between the pump curve and operating curve measured on the pump curve's y-axis. Use maximum process flow at maximum system pressure drop to define the flow coefficient (Cv). Ideally, the pressure drop should at least equal 10% of peak drop with the control valve at about 60% open. Maximum flow is at minimum pressure across the valve. Quarter-turn valves like butterfly and ball valves are practically fully open past 70%. This is a rough guide to keep the valve away from steep or flat portions of the Cv curve.

Explore the entire system curve and the control valve curve (Cv versus % open). The static head, from elevation change, stays the same but velocity head terms for equipment are a function of velocity2/2. A control valve will eat the pressure but flow is another matter; verify its flow characteristics won't allow the control valve to surge back and forth on its curve at this point on the pump operating curve. At minimum flow the danger is you'll deadhead the pump. As flow is decreased from the peak of the pump curve, which is near the best efficiency point (BEP), fluid in the pump may begin to boil (usually this occurs at about 10% of BEP flow). If deadheading is a danger, design a circulation loop or flush; don't turn the pump off and on — the motor and seals will wear quickly. Twenty percent of BEP flow is optimal for recirculation.

Next, formally estimate line losses. For slurries and viscous liquids use the equivalent length (Le) method. With water and similar nonviscous liquids, the traditional K-factor method suffices. The Le method is about 30% more conservative than the K-factor one. "Crane Technical Paper No. 410: Flow of Fluids Through Valves, Fittings and Pipe" and the "Cameron Hydraulic Data Book" fully describe both approaches.

Finally, to guard against common pitfalls, double-check to ensure: 1) the ratio of the NPSH required to the NPSH available doesn't exceed 0.85:1 at maximum flow; 2) the control valve isn't undersized; and 3) fluid properties are carefully defined.

By using tried-and-true techniques, you can size pumps relatively painlessly.

Dirk Willard is a Chemical Processing Contributing Editor. You can e-mail him at: dwillard@putman.net