The centrifugal pump is one of the most popular members of the turbo-machinery family. It has applications in the areas of potable water distribution, wastewater transfer, the pumping of chemicals, slurries, pulp and seawater. They pressurize sprinkler and standpipe systems. They are used to circulate purified water for pharmaceutical applications. They deliver feed water to boilers and pump steam condensate, and so on. Every engineer and designer in the mechanical, plumbing and fire protection disciplines must be a master of centrifugal pump selection. Unfortunately, as a dynamic machine involving the interplay between the physics of momentum, thermodynamics and fluid dynamics, the centrifugal pump can be a potential source of confusion and misunderstanding. Far too many designers and engineers proceed with pump selections unfettered by the facts. This practice of "eyeball" engineering leads to gradual degradation of technical knowledge among practicing engineers and designers. It is worthwhile to review pump fundamentals every so often. It is equally important to keep these fundamentals in perspective when considering practical applications.
The field of centrifugal pump design and application is a specialized one. It has its own set of terminology that, if not
completely understood by a designer, can affect the performance of a centrifugal pump installation. Complete familiarity of this
vocabulary as well as a solid working knowledge of fluid dynamics and pump design is essential to being a member of the
pumpology clergy. To narrow the scope here, let's consider water at 70° F at sea level. To approach some centrifugal pump
fundamentals, let's consider a few common myths:
Fallacy #1A pump requiring a net positive suction head of, say, 10 feet must take suction from a reservoir with a free surface at least 10 feet above the pump suction flange (or a minimum suction inlet pressure of 4.3 psig).
A corollary to this myth is that a centrifugal pump must have a "flooded suction," implying that there must always be water above the pump inlet or a positive gauge pressure at the pump inlet. Yet another myth is the mistaken notion that a centrifugal pump sucks water into its casing. It is absolutely true that there must always be water in the pump casing and suction pipe and there must be a positive suction head. If air is trapped in the casing, the impeller will spin freely with utterly no possibility of priming itself. The fact is, however, centrifugal pumps are capable of "lifting" water from a reservoir with a free surface below the pump suction with the assistance of atmospheric pressure-as long as the pump casing and suction pipe are pre-primed.
At no-flow conditions, the vertical distance (in feet) below the centerline of the pump impeller to the free surface of the reservoir is called the static suction lift (hs). If the free surface is above the inlet, it is called the static suction head. The theoretical maximum elevation a column of fresh water can be lifted above the level of its reservoir at standard atmospheric conditions (14.7 psia and 70° F) is about 33.1 feet. (See the apparatus illustrated at right.) Most people say the maximum height water will go is 33.95 feet, the value of the atmospheric head (ha) at standard atmospheric conditions. However, before a column of water can reach this height, the pressure at the top of the column approaches the vapor pressure of water or vapor pressure head at that temperature (about 0.84 feet). Once the vapor pressure is reached at the top of the column, water will vaporize into the vacuum that the column of water is trying to pull. The liquid water level will never go above 33.1 feet, no matter how high the tube is lifted.
The energy loss of water flowing through the suction pipe due to pipe resistance, fittings, valves and entrance conditions is referred to the friction head. The actual total head pushing the water into (or flooding) the impeller consists of the atmospheric head plus the static suction head (or minus the static suction lift) minus the vapor pressure head and the friction head. The sum is the Net Positive Suction Head Available (NPSHA).
The Net Positive Suction Head Required (NPSHR) is the minimum head needed by a given pump and is a function of the pump
design. The pump manufacturer specifies this value. For example, a pump at sea level with a NPSHR of 10 feet is drawing 20
gallons per minute from a reservoir of water with a free surface 12 feet below the centerline of the impeller. If a total of 3 feet is
deducted for friction head and vapor pressure, the NPSHA is 33.95 ft. - 12 ft. - 3 ft. or 18.95 feet. Note how significantly absolute
atmospheric pressure contributes to the flooding head. As long as the connected pump has a NPSHR lower than 18.95 feet, the
pump will operate properly. If, however, the NPSHA is lowered below the NPSHR, water will spontaneously vaporize in the pump
casing, causing violent and damaging vibration and a sharp drop in discharge pressure.
Fallacy #2In a duplex pump arrangement (two identical centrifugal pumps piped in parallel), the relative pumping capacity is doubled when both pumps run as compared to just one pump operating. When designing a system, safety lies in numbers.
Another source of misunderstanding is the notion that a pump will automatically operate at its rated capacity. In other words, a 100 gpm pump will operate at 100 gpm, and two such pumps in parallel will operate at 200 gpm. This is not necessarily so. Most pump performance curves offered by manufacturers will list the rated capacity and head. But these specifications only indicate the flow and head where the pump will operate at optimal efficiency. They in no way suggest the actual operating condition you will see when the pump is installed in a system.
The actual operating point of a system is a balance between the friction head (hf), the static discharge head (hsd) and the performance curve of the pump. The friction head is the energy loss through the discharge piping system due to friction, fittings, valves and entrance/exit conditions. One method of estimating friction head for water under turbulent flow conditions and at 60° F is the Hazen and Williams Empirical Formula.
In the Hazen and Williams Formula, Le is the equivalent length of pipe in feet (including fittings, valves and other components), C is a friction factor found in engineering references, and d is the pipe inner diameter expressed in inches. If this function were to be graphed out for a given length of pipe, inner diameter and friction factor, it would look like Curve A shown in Figure 2.
In Figure 2, Curve B is the same system curve (i.e. the same piping system) but with a static head condition as illustrated in Figure 3.
The curves in Figure 2 represent the head or pressure required at the discharge flange of the pump to achieve a given flow. Note that at zero flow, Curve B shows that the pump discharge flange experiences the static discharge head. Flow-induced losses are superimposed onto the static head resulting in a steady increase in required discharge pressure.
Every centrifugal pump has a characteristic performance curve at a fixed impeller diameter and speed. A sample performance curve represented in red is shown in Figure 4. The operating point is where the system curve intersects the red pump performance curve (Point A).
If Figure 4 above is the performance of one pump running, then the total performance curve of two such pumps is represented by the blue curve in Figure 5. Each pump will carry one half of the total flow (assuming they are piped symmetrically). If one pump running results in a flow of Q1, then the operation of two pumps against the same system curve will result in a flow of Q2. In this example, the total flow is increased by only 22 percent, which is far from the 100 percent increase that might be assumed. Depending on the steepness or shallowness of the system curve, this percentage will decrease or increase.
The only proper way to select a centrifugal pump is to estimate the system curve by whatever method appropriate and by selecting a pump performance curve with the optimal efficiency at the design flow and head. You cannot simply pick a "100 gpm pump" for a design flow of 100 gpm and automatically duplex a pump set expecting to cover yourself fully in case you get into a jam. If you painstakingly compute the system curve taking into consideration every length of pipe, fitting, valve and transition, you are at a good starting point. However, if you select a pump impeller and casing that fits the curve exactly down to the decimal point believing that the installed system will work the same way it does on paper, you might as well take that system curve, tie it into a noose and hang yourself from the rafters.
Always leave yourself room for error. A given pump casing will accept a range of impeller sizes. Select a casing with an impeller that falls in the middle range. Never select the largest impeller that can fit into the casing. It is easy and cheap to change out an impeller. It isn't so cheap to change out an entire pump skid if the performance falls short of your calculations. If you're working with a variable frequency drive or a belt drive, you're golden. Adjustments can be made by varying the drive speed either electronically or by sheaving the pulleys.
The advantage of multiplexing pumps comes into play when the system curve itself can vary over a wide range. A system curve
can be altered in a system that has an automatic flow or temperature control valve that modulates open and closed. A system
curve can vary widely in a domestic water distribution system from no flow with all fixtures closed to some maximum expected flow
with many fixtures in use simultaneously. In these cases, a multiplexed pump system can provide optimal performance over the full
flow regime. At minimal- to no-flow periods, one pump running in its highest efficiency range works well. As the demand increases
and the system curve slides to the right as shown in Figure 6, one or more pumps cascade on to provide more flow without
exceeding the flow capacity of any one pump. The arrangement also provides a relatively constant discharge pressure. The points
A, B and C in Figure 6 represent the operating conditions when one, two and three pumps operate.
Fallacy #3Increasing the horsepower of a pump's drive will increase capacity.
This is the statement that sparked the magazine controversy. The required horsepower to turn a centrifugal pump under load (referred to as brake horsepower, bhp) is a function of the mass flow rate across the impeller, the pump head, and the pump's inefficiency. As any one of these factors is increased, the required horsepower goes up. However, increasing the horsepower while keeping the pump impeller and speed constant will not result in a corresponding increase in discharge head. The only way to increase head is to increase speed, increase the impeller diameter, or use a more efficient impeller design.
Figure 7 illustrates how, by varying the impeller diameter of a centrifugal pump, you get a corresponding increase in discharge pressure and flow. Each successive performance curve establishes a new operating point along the system curve. The dotted lines of constant horsepower show that each successive operating point demands a larger brake horsepower in order for the pump to achieve that operating point. It is not correct to say that raising the horsepower alone will draw the operating point upwards and to the right, away from the impeller's performance curve.
These are just three mistaken notions that engineers and designers may be faced with. It is crucial to stay on top of centrifugal
pump fundamentals in order to properly size and specify pumps for any application. It is so important, in fact, that aspiring
Professional Engineers can count on facing more than one centrifugal pump problem on the Principles and Practices Exam. The
references listed below are excellent sources of centrifugal pump guidance.
- Lindeburg, PE, Michael R., Mechanical Engineer Reference Manual, 9th Ed., Professional Publications, Inc, 1994.
- Heald, C.C., Cameron Hydraulic Data, Ingersoll-Rand, 1992.
- Pump Handbook, 2nd Ed., McGraw-Hill, 1986.