The U.S. Environmental Protection Agency’s (EPA) Water Infrastructure Resiliency and Finance Center, in collaboration with the...
Suction and NPSH
One of the claimed advantages of centrifugal pumps over
positive displacement pumps is their ability to operate over a wide range of
flow. Since a centrifugal pump operates at the intersection of a pump curve and
a system curve, by varying the system curve the operating point of the pump
easily is changed. (See Figure 1.)
The convenience and simplicity of such flow control by the
discharge valve throttling comes at a price because a pump is forced to run
either to the left or right of its best efficiency point (BEP). However, the
real danger of operating the pump too far off peak comes from the suction side
considerations. Too far to the right and you easily are risking to run out of
the available net positive suction head available (NPSHa), causing cavitation
problems. Too far to the left and flow recirculation at the impeller eye will
let itself be known through the noise, vibration and damage. Thus, the flow
must be limited on both sides of the BEP. (See Figure 2.)
Consider the first limitation: high flow. The centrifugal
pump stops pumping when liquid turns to vapor. This happens when the pressure
somewhere inside the pump drops below liquid vapor pressure. Vapor pressure
depends on the temperature as well as a few other things. Water turns to vapor
at 212° F at atmospheric pressure, when you boil water in the open pot. If
the pot were closed, the water would reach higher pressure before it boils.
Conversely, if the pressure were reduced (vacuum), water would boil at a lower
temperature. It will boil at room temperature, if the absolute pressure is less
than approximately 0.4 psia. Water has low vapor pressure, but other substances
may have very high value.
Freon, for example, has vapor pressure of about 90 psia, and
ethane value of vapor pressure is about 700 psi at 80° F. Knowing vapor
pressure without relating it to a corresponding temperature is meaningless. The
higher the temperature, the higher the vapor pressure is.
A centrifugal pump is a "pressure generator"
producing pressure by the centrifugal force of its rotation impeller. The
pressure rises as flow progresses from the suction to discharge. This is why
vaporization of liquid is most likely to happen in the inlet (suction) region
where the pressure is lowest. In practice, it is difficult to know exactly when
vaporization (cavitation) happens, so it is wise to keep some margin of
available pressure over vapor pressure. Pressure is expressed in
"psi," but also can be expressed in feet of water. The conversion
formula is as follows.
FT = psi x 2.31 / SG, where SG is specific gravity
This pressure, expressed in feet of water, is called
discharge head at the pump exit side or suction head on the inlet side. The
difference is a pump-developed head, also called a total dynamic head (TDH).
These heads must include both static and dynamic components. Static part is
what we measure by the gage in front of a pump and dynamic, according to
Bernoulli, is velocity head V2/2g.
For example, suppose an inlet pressure gage installed in a
2-inch pipe directly in front of a pump delivering 100 gpm oil with specific
gravity SG = 0.9 reads 10 psig. To calculate velocity head, find the pipe net
area, which is A = 3.14 x d2 / 4 = 3.14 x 22 / 4 = 3.1 in2.
The velocity can be calculated by the following formula .
V = (Q x 0.321) / A = (100 x 0.321) / 3.1 = 10.4 ft./sec.
Then, the velocity head is
V2 / 2g = 10.42 / (2 x 32.2) = 1.7 ft. or, converted to psi,
1.7 x 0.9 / 2.31 = 0.7 psi.
The total suction pressure then is 10 + 0.7 = 10.7 psi or,
if expressed in feet of water,
10.7 x 2.31 / 0.9 = 27.5 feet.
It is best to have gages as close as possible to the pump on
the suction and discharge sides. Unfortunately, these gages often are not
installed, which somehow happens more often on the suction side, and suction
head in front of the pump is estimated by calculations, by subtracting the
pressure (head) losses from the known value of head upstream and adjusting by
elevation correction, according to Bernoulli. In many cases, the upstream datum
is a known liquid level in a suction tank.
tank (Figure 3a),
tank (Figure 3b), and
under vacuum (Figure 3c).
For water and similarly low viscosity liquids, suction
losses usually are low and often are disregarded. However, for more viscous
substances such as oils, these losses can be substantial and may cause the
pressure in front of the pump to drop below the vapor pressure, causing
cavitation. This is why the inlet velocity must be minimized as the losses
depend on velocity squared.
Longer pipe runs, bends, turns and other restrictions add to
inlet losses, leading to further pressure reduction in front of a pump. As a
quiz, using the examples above, see if you can figure out what happens to inlet
pressure if the pipe diameter is doubled? What if the diameter is halved?
To avoid cavitation, what matters is not the suction
pressure but how much higher it is than the vapor pressure of the liquid being
pumped. This is where a concept of NPSH comes in handy. The available NPSHa
simply is the difference between this total suction head (as discussed
previously) and vapor pressure expressed as head in feet.
Pump manufacturers conduct tests by gradually lowering
suction pressure and observing when things begin to get out of hand. For a
while, as pressure decreases (i.e., NPSHa gets smaller), nothing happens, at
least nothing obvious. A pump operating at a set flow keeps on pumping and
develops constant head. At some point, when the value of suction pressure (and
corresponding NPSHa) reaches a certain value, a pump head begins to drop, which
typically happens suddenly. (See Figure 4.)
Actually, things are happening inside the pump well before
the sudden drop of head, but they are not as obvious. First, at still
substantial suction pressure, small bubbles begin to form. This is called
incipient cavitation--tiny bubbles in your water kettle that begin to
percolate before water is fully boiling. These small bubbles are formed and
collapse at very high frequency and only can be detected by special instrumentation.
As pressure is decreased further, more bubbles are formed and, eventually,
there are so many of them that the pump inlet becomes "vapor
locked" so that no fluid goes through and the pump stops pumping; the
head drops and disappears quickly. It would be nice if enough pressure always
was available at the suction so that no bubbles were formed whatsoever.
However, this is not practical, and some compromise must be reached. The
Hydraulic Institute (HI) has established a special significance to a particular
value of NPSHa at which the pump total developed head drops by 3 percent. The
value of this NPSHa at which a pump loses 3 percent TDH over (i.e., in access
of) vapor pressure is called net positive suction head required (NPSHr) in
order to maintain 3 percent TDH loss.
NPSHr = (Hsuction - Hvapor), required to maintain 3
percent TDH loss
Therefore, NPSHr is established by an actual test and may
vary from one pump design to another.
In contrast, the available NPSHa has nothing to do with a
pump but strictly is a calculated value of total suction head over vapor
pressure. Clearly, NPSHa must be greater than NPSHr in order for a pump to make
its performance (i.e., to deliver a TDH) at a given flow.
It is easy to know when a NPSH problem is obvious--a
pump just stops pumping--but the vapor bubbles do not need to be so
dramatically developed to cause TDH drop because even smaller bubbles can cause
problems. The evolved bubbles get carried on through the impeller passage at
which pressure is rising from inlet to exit of the blade cascade. This
increased pressure causes the reverse of what happened to a bubble
"awhile back," when it first became a bubble formed from a liquid
particle during phase transformation (boiling). Now, the bubble is at the
somewhat higher pressure, which tries to squeeze it against the vapor surface
tension that keeps the bubble a bubble. The bubble collapses (implodes) with a
sudden in-rush of surrounding liquid into a vacuum space previously occupied by
the bubble. The in-rush is accompanied by a tremendous but localized pressure
shock, which, if imploded in the vicinity of the metal (impeller blade), would
cause a microscopic hammer-like impact, eroding a small particle of metal. With
enough bubbles and enough time, the impeller vanes can be eroded away quickly,
a phenomenon known as cavitation damage.
This is why an NPSHa margin (M=NPSHa-NPSHr) is
important. This margin typically is at least 3-5 feet and preferably should
be even more, if possible.
The NPSHr, discussed above, was so far limited to a
particular flow on a pump performance curve. At higher flow, the internal fluid
velocities are higher and, according to Bernoulli, the static pressure (or
static head) part becomes less (i.e., closer to vapor pressure). The static
pressure, therefore, must be increased externally (i.e., a higher value of
NPSHR is needed for higher flows). This is why the NPSHR curve shape looks like
It is important to specify an ample margin of NPSHa over the
pump NPSHr for a complete range of operation and not just at a single-rated
flow point. The following example illustrates a common mistake, leading to the
NPSH problem. The pump was procured with the intent to deliver between 350 and
500 gpm, and the manufacturer quotation indicated 16 feet required NPSHr at 500
gpm. As a process later changed, more flow was required, and the discharge
valve was opened to allow the pump to deliver more flow, 750 gpm. However, as
can be seen from Figure 5, at about 700 gpm, the NPSHr exceeded the NPSHa
available at the installation, and the pump started to experience typical NPSH
problems-: noise, loss of performance and impeller cavitation damage.
An instinctive thought to address the issue of cavitation
due to flow run-out operation is to "overkill" on a pump size and,
therefore, always stay to the left of the BEP. In this example, a larger pump
having some 16 feet NPSHr but at 750-800 gpm, would never run out of the
NPSHa. That is true and, in fact, this is exactly what has been a common
practice in the past where an oversized (and more expensive) pump would be
specified "to make sure" to discover other, just-as-severe
When a centrifugal pump operates below a certain flow point,
a phenomenon known as flow recirculation in the impeller eye starts. This
depends on several design factors such as suction-specific speed but,
generally, recirculation begins below 60-80 percent flow and becomes
quite severe below 20-40 percent. At even lower flows, recirculation may
become especially severe and is known as surge--violent, low-frequency
sound accompanied by strong low-frequency vibration of the pump and piping.
(See Figure 6.)
In addition to obvious mechanical problems with
recirculation, the flow undergoes a complex vortexing motion at the impeller
inlet (eye) with localized high velocities of the vortex causing
horseshoe-looking cavitation damage, usually on the "blind" side of
the blade, as compared to high-flow cavitation. Other problems add oil to the
fire: radial thrust, which sky-rockets at low flow, causes deflections of the
shaft, leading to seal leaks, bearings life reduction and even shaft breakage.
Troubleshooting methods and failure analysis techniques help
to pinpoint a cavitation problem with a particular pump. The indications of the
high flow cavitation are different from the low flow recirculation damage. The
side of the blades, the extend and shape of the cavitation trough can be
helpful in determining the causes of each individual problem.