Definition of NPSH

The margin of pressure over vapor pressure, at the pump suction nozzle, is Net Positive Suction Head (NPSH). NPSH is the difference between suction pressure (stagnation) and vapor pressure. In equation form: NPSH = Ps - Pvap Where: NPSH = NPSH available from the system, at the pump inlet, with the pump running Ps = Stagnation suction pressure, at the pump inlet, with the pump running Pvap = Vapor pressure of the pumpage at inlet temperature Since vapor pressure is always expressed on the absolute scale, suction pressure must also be in absolute terms. In U.S. customary units, both pressures must be in psia. Gauge pressure is converted to absolute pressure by adding atmospheric pressure. In equation form: absolute pressure = gauge pressure + atmospheric pressure The above equation provides an answer in units of pressure (psi). This can be converted to units of head (feet) by the following equation: h = 2.31p/SG Where: h = Head, feet p = Pressure, psi SG = Specific gravity of the liquid Problem No. 1: NPSH Stagnation suction pressure is determined to be 1-psig at a sea level installation. The vapor pressure of the liquid is 8-psia. Calculate NPSH in PSI and feet for a 0.9 SG liquid NPSH = Ps - Pvap = 1 + 14.7 - 8 = 7.7 PSI NPSH = 2.31p/SG = (2.31) (7.7)/0.9 = 19.8 FEET

Velocity Head is Included

Note that suction pressure is stagnation pressure (total pressure); it includes velocity head. Adding velocity head puts all pumps on the same basis, otherwise a pump would require different amounts of NPSH when tested with different sizes of suction lines (assuming that suction pressure is measured in the suction line, which is normally the case). It should also be noted that the velocity head is normally quite small, relative to the NPSH, so it can usually be ignored.
Units of NPSH

For centrifugal pumps, NPSH values are expressed in units of specific energy (equivalent column height) such as feet or meters. For displacement pumps (rotary and reciprocating), NPSH values are normally expressed in pressure units such as pounds per square inch (psi), kilopascals, or bars. NPSH values are neither gauge pressures nor absolute pressures. The g in psig means that the pressure is measured above atmospheric pressure. The a in psia means that the pressure is measured above absolute zero, a perfect vacuum. NPSH is a measurement of pressure above vapor pressure, so the units of NPSH (in the U.S.) are just psi or feet.
What Symbol Should We Use?

Because units of pressure are typically used to express the value of NPSH for a displacement pump, some authors, and companies, use symbols such as NPIP (for net positive inlet pressure) and NIP (for net inlet pressure), because the units are pressure units, not head units. For simplicity, I'll stick with NPSH, regardless of the units used.
Suction Pressure: The First Half of the NPSH Equation

Suction pressure must be determined at the pump suction nozzle when the pump is running. If suction pressure is measured with a gauge, the atmospheric pressure (at the pump location) must be added to the gauge reading to convert the reading to absolute pressure. The elevation of the gauge must also be added (if the gauge is above datum) or subtracted (if the gauge is below datum). Although often negligible, the velocity head in the pipe at the gauge connection should be added to obtain total (stagnation) pressure. For a reciprocating pump (and some rotaries), the acceleration head must be subtracted. (More on acceleration head later.)
Vapor Pressure: The Second Half of the NPSH Equation

Vapor pressure is more difficult to determine than suction pressure. It is a measure of the "desire" of a liquid to boil to a gas. Some liquids, such as butane and ammonia, have high vapor pressures. They must be kept under pressure, or they will boil (flash). An open container of pure ammonia would quickly boil away, filling the area with noxious ammonia gas.

Cool water has a low vapor pressure. An open container of cool water, on the earth's surface, would not boil, but would evaporate slowly over a period of days. The desire of cool water to boil is therefore low. If we were to place that same open container of cool water on the surface of the moon, it would boil away, similar to the ammonia. Why? The atmospheric pressure on the moon is zero, a perfect vacuum. The vapor pressure of pure, air-free water at 80-deg F is about 1/2-psia. This means that if the pressure on the water is reduced below - psia, the water will boil.
A Function of Temperature

Vapor pressure is a function only of temperature. As the temperature of the liquid increases, its vapor pressure increases until the critical temperature is reached. At the critical temperature, vapor pressure vanishes. Above the critical temperature, there is no distinction between a liquid and a gas. It is all fluid.
Boiling Reestablishes Equilibrium Conditions

Any liquid at its vapor pressure is on the verge of boiling (flashing). In such a condition it is said to be in equilibrium, at its bubble point or saturated. If the pressure is reduced slightly, it will start to boil. If the temperature is held constant (which requires heat input) and the pressure held constant (below the vapor pressure), the liquid will continue to boil until it has all flashed to vapor. If heat is not provided to the liquid, the portion flashing to vapor will cool, and will also absorb heat from the remaining liquid, causing the liquid temperature to drop. The lower temperature will result in a lower vapor pressure. The boiling will continue only until the vapor pressure drops to the pressure which is imposed on the liquid. When that vapor pressure is reached, and the boiling stops, the liquid-vapor mixture is again in equilibrium. This is what happens in the suction passage of a pump. Cavitation will cool the liquid and stop the cavitation. Otherwise, all the liquid would flash to vapor.
NPSH Available: A System Characteristic

NPSHA stands for NPSH Available from the system. It can be calculated by measuring suction pressure at the pump suction nozzle, correcting to datum, adding atmospheric pressure, adding velocity head and subtracting vapor pressure. In equation form: NPSHA = Psg + Pz + Patm + Pvel - Pvap Where: NPSHA = NPSH available to the pump, psi Psg = Gauge pressure measured at suction nozzle, psig Pz = Elevation of gauge above pump centerline, converted to pressure units, psi Patm = Atmospheric pressure, psia Pvel = Velocity head, convened to pressure units, psi Pvap = Vapor pressure of the pumpage, at the pump suction nozzle, psia If desired, all units can be convened to head (feet) prior to plugging into the equation.

If the system has not been built, it is necessary to calculate the NPSHA by starting with the pressure in the suction tank. Add atmospheric pressure, add (or subtract) the liquid level above (below) datum, subtract all losses from the tank to the pump and subtract vapor pressure. With reciprocating pumps it is also necessary to subtract acceleration head, a term which will be explained later. In equation form: NPSHA = Pt + Patm + Pzt - Pf - Pvap Where: Pt = Tank pressure, psig Pzt = Elevation of liquid in suction tank, converted to pressure units, psi Pf = Friction losses at tank exit and in suction line, converted to pressure units, psi Problem No. 2: NPSHA A suction gauge with its centerline 2-ft below the centerline of a centrifugal pump reads 152-psig. Atmospheric pressure is 14.0-psia. The pipe is 3-in standard weight steel. Capacity is 100-gpm. Vapor pressure is 163-psia. SG is 0.5. Calculate the NPSHA in feet. Because the desired answer is in feet, rather than PSI, we will convert all pressure units to feet. Flow Area of Pipe = 3.14 x 1.52 = 7.07 square inches VEL = 0.321 x Q/A = 0.321 x 100/7.07 = 4.54 feet/sec Hvel = V2/2G = 4.542/64.4 = 0.3 feet NPSHA = Hsg + Hz + Hatm + Hvel - Hvap = (psg + patm - pvap)(2.31/SG) + Hz + Hvel = (152 + 14.0 - 163)(2.31/0.5) + (-2) + 0.3 = 13.9 - 2 + 0.3 = 12 feet [Note: A new 300-psi gauge, used to measure the suction pressure, normally has an accuracy of 1 percent of full scale, or 3-psi (14ft). Therefore, the error in the gauge could be more than the calculated NPSHA.] [Also note that the velocity head was negligible, and could have been ignored.]

NPSH Reguired: A Pump Characteristic

The letters NPSHR stand for the NPSH required by the pump. This characteristic must be determined by test. Test methods and acceptance criteria for different types of pumps will be discussed later.
System Requirement

For proper operation of the pump, it is necessary that NPSHA > NPSHR. The system must provide more NPSH than the pump requires.
Effect of Gas on NPSH

Effect of Dissolved Gas on Vapor Pressure Determining the vapor pressure of light hydrocarbons is fairly easy. They are usually at their vapor pressure ("bubble point") in the suction vessel. In such a case, the vapor pressure is equal to the vessel pressure. How about cool water? That's easy. A quick reference to the steam tables reveals its vapor pressure, right? Wrong! Steam tables give the vapor pressure of only pure deaerated water. Water has almost always been exposed to air, and cool water has a strong affinity for air. Water will absorb a significant quantity of air, and this air will flash out of solution as the pressure is reduced (such as in the eye of an impeller). Even water pumped from below ground contains a significant amount of air. Therefore, most water has an effective vapor pressure higher than pure water. The dissolved air increases the vapor pressure of the "solution." The air will flash out of solution at a pressure higher than the vapor pressure of pure water. Consequently, we do not have as much NPSH as we calculate using the vapor pressure from the steam tables. How can we compensate for this higher vapor pressure of cool water? Chan (1) has demonstrated that, for water that has been saturated with air in an atmospheric tank the effective vapor pressure can be approximated as the average of atmospheric pressure and the vapor pressure of air-free water to calculate the NPSH available to a centrifugal pump. Heavier hydrocarbons (vapor pressures below atmospheric pressure) have the same problem as water. They are usually stored in atmospheric tanks, and they absorb more air than water does (3). Kerosene is a good example of a heavier hydrocarbon that really soaks up air. Dissolved air has caused problems with pumps handling heavier hydrocarbons (3). Other Sources of Gas (Entrained Gas) Gas can come from sources other than cavitation of the liquid itself, or from dissolved gases. One source can be the suction vessel. If the submergence of the tank outlet to the pump is insufficient to prevent the formation of a vortex, pockets of gas will enter the suction pipe from the tank, eventually flowing into the pump. A vortex breaker in the tank can reduce the tendency of the liquid to form a vortex.

If liquid feeds into the suction vessel above the liquid level, gas bubbles will be carried down into the liquid, and may flow into the suction line. Therefore, the main feed line, and the minimum-flow (by-pass) line from the pump should feed into the vessel below the minimum liquid level. Another source of gas can be the suction line itself. If any portion of the line is under a vacuum, air will leak into the line through gaskets and fittings that are not absolutely tight. A valve in a suction line, under a vacuum, can leak air into the line though the stem packing. The pump itself is also another source of gas. If the stuffing box, or seal chamber, is under a vacuum at any time while starting or running, air can leak through the packing or mechanical seal. Design Systems to Exclude Gas Although a few pumps, such as the minimum-clearance reciprocating and the two-screw rotary, have been adapted to handle some quantities of gas (they are really combination pump-compressor), most pumps are not suitably equipped to handle free gas. Although, as will be discussed in a future article, a small quantity of gas can reduce noise and damage in a centrifugal pump; any significant quantity of gas (more than about 2 percent) will have a detrimental effect on pump performance. For a high-pressure reciprocating pump, any amount of free gas can seriously damage the pump. Operators of pressurized-water reactor nuclear systems can attest to the latter. For these reasons, it is considered good practice to design a system to separate all free gas from the liquid flowing into a pump. This requires careful design of the suction vessel.
Illustration of NPSH

Figure 1 may help clarify your understanding of NPSH. If the lower end of a closed vertical tower is submerged in an open tank of air-free water at 80-deg, and atmospheric pressure is 14.7psia (sea level), the column of water will rise in the tower when a vacuum is applied to the top of the tower. Atmospheric pressure actually pushes the water up into the tower.

Figure 1. Illustration of NPSH As the pressure in the top of the tower is further reduced, the water will rise further, until it is 33ft above the level in the tank. The water will reach this level when the pressure above the water has dropped to 0.5-psia (about 29-in of mercury vacuum). Any attempt to further reduce the pressure and "lift" the water higher will fail. The water column will "break." Actually, the water will boil at the upper surface, pushing water vapor into the cavity above the water and preventing the pressure from dropping below 0.5-psia. Therefore, the vapor pressure of water at 80-deg is 0.5-psia. We will hold the column of water at these conditions, and mount a pump at the side of the tower, taking suction from the tower. If this pump requires 6-ft of NPSH, how high on the side of the tower can the pump be mounted for satisfactory operation? The answer, 6-ft below the surface of the boiling liquid (assuming no friction loss or acceleration head) or 27-ft above the tank surface (position A). If a pump requires 12-ft of NPSH, it should be mounted no higher than position B (12-ft below the surface of the boiling water). A pump operating at position C, with a separate suction line (again assuming no losses), will be subjected to the same suction conditions as the pump in position A. The water above the pump in position A actually has no effect on the pump. Similarly, a pump operating at position D will be subjected to the same suction conditions as the one in position B (again assuming no losses). Any system can be analyzed in a similar fashion if the actual vapor pressure of the liquid and the pump suction pressure are known. The point to remember is that (neglecting velocity head and acceleration head) the amount of NPSH available to a pump is equal to the distance that the liquid would rise above the datum, from the suction line, before it starts to boil.
Example Calculations of NPSHA

The following examples of NPSH calculations are provided by the Center for Professional Advancement from their course Applied Pump Technology, created by the late Igor Karassik.

NPSH = S + (Ps - Pvp) - (hfs + hi) Where: Pvp = Vapor pressure of lquid at pumping temperature hfs = Friction losses in suction losses from A to B

hi = Entrance loss at A Figure 2. How to calculate NPSHA from pressure in a suction vessel (Acknowledgement is made to The Center for Professional Advancement, Applied Pump Technology Course)

NPSHA = ((Ps - Pv ) X 2.31)/SP. GRAVITY + Z - hf Where: Ps = Pressure over liquid surface, PSIA Pv = Vapor pressure of liquid, PSIA Z = Static head, FT hf = Friction losses, FT

A) At Sea Level NPSHA = ((14.7 - 0.5 ) X 2.31)/1.0 - 15 - 3 = 32.8 - 15 - 3 = 14.8 FT B) At 5000-ft above sea level NPSHA = ((12.2 - 0.5 ) X 2.31)/1.0 - 15 - 3 = 27 - 15 - 3 = 9.0 FT

Figure 3. Classical calculations of NPSHA. Note that using 0.5-psia as vapor pressure ignores effect of dissolved air. (Acknowledgement is made to The Center for Professional Advancement, Applied Pump Technology Course)

NPSHA = ((Ps - Pv ) X 2.31)/SP. GRAVITY + Z - hF NPSHA = ((14.7 + 5 - 0.5 ) X 2.31)/1.0 + 10 -4 = 44.3 + 10 - 4 = 50.3 FT Figure 4. Classical calculation of NPSHA with a Pressurized Suction Vessel. Note that using 0.5-psia as vapor pressure ignores effect of dissolved air. (Acknowledgement is made to The Center for Professional Advancement, Applied Pump Technology Course)

NPSHA = ((Ps - Pv ) X 2.31)/SP. GRAVITY + Z - hF NPSHA = ((37.5 + 14.7 - 52.2 ) X 2.31)/0.56 + 10 - 2 = 0 + 10 - 2 = 8 FT Figure 5. Calculation of NPSHA. Because the butane is at its bubble point, NPSHA requires no correction for dissolved gas. (Acknowledgement is made to The Center for Professional Advancement, Applied Pump Technology Course) Note that the calculations illustrated in Figures 3 and 4 use 0.5-psia as the vapor pressure of the pumpage. This is a correct number for pure, deaerated water, but does not allow for any dissolved air. Because it has been exposed to air, the water in these examples contain a significant amount of dissolved air. For Figure 3, if we follow the guideline of the Hydraulic Institute Standards (2), we must add 3psi to the vapor pressure, which would reduce the NPSHA by 7-ft. If we follow Chen's (1) recommendation, we must add 6-to 7-psi (14- to 16-ft) to the vapor pressure. This would reduce the NPSHA to zero or less. If we fail to account for the dissolved air, what will happen? We do not have a pump that sounds like it is pumping gravel, and we do not have an impeller chewed up by cavitation damage. Air does not do that. But the pump will exhibit a reduction in head, and may rumble a bit. A small amount of air injected into the inlet of a centrifugal pump will actually reduce cavitation noise and damage. (More on this in later months.)

NPSHR Shown as a Single Line as a Function of Capacity Only. Figure 1 is an example of single-line performance characteristics of a centrifugal pump. The independent variable capacity is plotted on the horizontal scale. Power, head, efficiency and NPSHR are plotted on vertical scales-each as a function of capacity.

Figure 1. The centrifugal pump family of characteristic curves Note that the NPSH curve is roughly a lazy "U," reaching a minimum value at a capacity Q about 40 percent of the best-efficiency capacity. Although most published curves do not show the increased NPSHR at low flows, all centrifugal pumps exhibit such a characteristic. The NPSHR always rises as the pump capacity approaches shut-off (zero flow rate). Figure 2 is a typical published performance curve. In this example, the NPSH curve is shown as a single line. (Note that it incorrectly shows the NPSHR to be a minimum at shut-off.)

Figure 2. Typical published performance curve. Single-line NPSH curve.

The Effect of Impeller Diameter on NPSHR

Figure 3 shows all parameters in the same fashion except for NPSH. Instead of a single curve that applies to all impeller diameters, lines of "iso-NPSH" are shown. Note that for the same capacity, smaller diameter impellers require more NPSH.

Figure 3. Typical published performance curve. NPSH values increase as impeller diameter decreases. It is normal to experience a rise in NPSHR as the impeller diameter is reduced since a 3 percent drop of a smaller head (resulting from a smaller diameter impeller) is a smaller head drop. By our definition of NPSHR, we are therefore permitting less cavitation with smaller diameter impellers. Currently there is no known way to predict this increase in NPSHR. Testing must therefore be relied on to establish NPSHR at each impeller diameter. Is it fair to penalize the smaller diameter impellers? Some pump vendors reason that it is not-the single-line NPSHR curve, obtained with the maximum-diameter impeller, should apply to all diameters supplied for that pump. What if one of those smaller-diameter impellers is supplied in a smaller-diameter casing (with a duplicate inlet passage)? We would be required to base our NPSHR on a 3 percent drop of a lower pump head, resulting in a higher NPSHR for the same impeller. Such contradictions account for some of the confusion surrounding this confounding subject of NPSH. A Method for Eliminating the NPSH Variation Caused by Different Impeller Diameters Because pump head is a function of the impeller diameter, D2, the traditional 3 percent-head-drop NPSH varies as the impeller diameter varies. With the same degree of cavitation in the impeller eye, a pump with a maximum-diameter impeller will exhibit a smaller percentage of head drop than the same pump with a reduced-diameter impeller. Gongwer (3) recognized this shortcoming in our definition of NPSHR. To normalize the NPSH requirements in his tests, he created a "constant" pseudo-head on which to base his head loss. Rather than use the actual head of the pump, he calculated the head that would be developed with an impeller diameter that was twice the diameter of the impeller eye (D2=2D1). A pump develops a head that is approximately U22/2g. This pseudo-head is therefore (2U1)2/2g, or 2 U12/g. The 3 percent head loss therefore converts to 0.03x2UI2/g = 0.06 UI2/g. In U.S. units, this would convert to a head loss, for the 3 percent loss, of H = (DlxN/5300)2. A 5-in diameter eye, rotating 3,600-rpm, would have an allowable head loss of 11.5-ft (completely independent of impeller diameter and pump head). Adoption by the pump industry of this revised definition of head loss would be a significant step in normalizing NPSH characteristics. It would be of value to pump users and manufacturers. Such an adjustment would result in NPSH3 curves being lifted for all pumps with (discharge) specific speeds less than about 1,500, and a lowering of NPSH3 curves for pumps with (discharge) specific speeds above that value. Prior to the concept of suction specific speed, S, the pump industry attempted to use the Thoma-Moody concept, sigma, which stated that NPSHR was proportional to pump head. It is

not. Numerous authors have so confirmed. S enabled us to begin decoupling NPSHR from H. Adoption of the above suggested criterion would be a further step in decoupling these two independent characteristics. Prior to the development of the NPSH concept, a performance curve would sometimes show the inlet characteristics of a centrifugal pump as lines of constant "maximum suction lift," as can be seen in Figure 4.

Figure 4. An older performance curve that shows maximum suction lift instead of NPSHR (Acknowledgement is made to the Center for Professional Advancement, Applied Pump Technology Course)
Actual NPSH Test of a Centrifugal Pump

NPSHR is a difficult characteristic to measure, and to further complicate matters, pump manufacturers often disagree on the appropriate testing method and acceptance criterion. The most widely-accepted criterion is the 3 percent head drop, as specified by API 610 (1). (There is also disagreement on the definition of "non-cavitating head.") Figure 5 shows the test results obtained during an NPSH test on a centrifugal pump.

Figure 5. Actual NPSH test points for a process centrifugal pump To get an NPSH point, we held the capacity constant while reducing the NPSHA. For each point, we plotted head as a function of NPSHA. We then determined how much NPSH was available when the head dropped 3 percent, which was the NPSHR for that capacity. To obtain a single point for the NPSHR curve, it is necessary to run a pump at a fixed capacity for an extended period of time. Because of the problems of running a centrifugal pump near shut-off for any extended period (heating, radial thrust, recirculation, pulsations and vibrations), we avoid trying to get NPSHR values at low capacities. When we plotted our points and smoothed a curve through them, we got a curve as shown in Figure 1. We heard a loud crackling sound coming from the pump as we reduced the NPSHA. That destructive sound helped us to understand how cavitation could damage pump components. We also noted that the sound level reached a peak, then dropped as we approached the 3 percent point. So here is the question: If you bought this pump, installed it in your plant, and provided the "required" 33-ft of NPSH, would the pump be cavitating? The answer is yes! If pumping cool water, it would cavitate enough to cause the head to be down by 3 percent, which is a lot of cavitation. How much NPSH is required to eliminate all cavitation? If you provided 75-ft, would that stop the cavitation? That would get the head back up to "non-cavitating head," but, unfortunately,

even at that point, there is still a significant amount of cavitation in the impeller eye. But how much more NPSH is required to eliminate all cavitation? To address this question, the Hydraulic Institute formed the NPSHA Margin Work Group. Initial conclusions of this group were published in the August 1997 issue of Plant Services Magazine (2). Key points in the article include the following:

Cavitation exists in all centrifugal pumps. NPSHA at incipient cavitation can be from 2 to 20 times the "3 percent" NPSHR. Achieving full (100 percent) head requires from 1.05 to 2.5 times the "3 percent" NPSHR. (In Figure 5, we can see that this ratio is about 2.3.)

A study by this author revealed that the NPSH required to achieve incipient cavitation for the typical pump at a capacity correspondent with no prerotation at the impeller inlet is a function of the impeller inlet vane angle, and varies from 2 to 10 times the NPSH3. We can conclude that it is normally not practical to try to eliminate cavitation, but what is a reasonable margin that we should strive for in our system designs?

What is a Safe NPSH Margin for a Centrifugal Pump?

The Hydraulic Institute NPSHA Margin Work Group (1) recommended NPSH margins ranging from zero to four times the NPSHR (NPSHANPSHR = 1 to 5). The higher values apply to pumps with high to very high values of suction energy, a term defined as the product of the diameter of the pump suction nozzle, pump RPM and the suction specific speed.

McGuire (2) has provided a similar set of recommendations, seen in Figure 1, based on the service, pump size and peripheral velocity of the impeller eye.

Figure 1. Recommended centrifugal pump NPSH margins (J.T. McGuire [2])

Sloteman, et.al., (7) reported field experience with a four-stage, barrel type, boiler-feed pump, rated at 33,500 hp, 11,000 gpm and 5,800 rpm. The 3 percent NPSHR was 250 ft. The power plant, with a booster pump, provided 500 ft, twice that "required" by the pump. Impellers were made of CA6NM stainless steel. Because of cavitation damage, the first-stage impeller experienced a life of about 12 months. Although a 100 percent NPSH margin was provided, the cavitation damage rate was unacceptable.
Can You Provide Too Much NPSH?

Well, it does seem so. As we have just proven, conventional wisdom is to provide some margin of NPSHA over NPSHR. Unfortunately, the facts indicate that if the pump is operating near the 3 percent head-drop NPSHA, an increase in NPSHA will result in a noisier pump with a higher rate of damage. Grist (3) reported that maximum cavitation-erosion rate corresponded with maximum cavitation noise. Based on the information available, he postulated that this occurs when the NPSHA is about two times the "3 percent" NPSHA. Figure 2 is a reproduction of Figure 2 from Grist's paper (3). Note that this ratio of "two" occurs only at (or near) the BEP. At both lower and higher capacities, the ratio increases. The ratio of two should be taken as typical, for the literature indicates lower ratios for pumps with low suction specific speeds and higher ratios for pumps with high suction specific speeds.

Figure 2. From Grist's "Nett Positive Suction Head Requirements for Avoidance of Unacceptable Cavitation Erosion in Centrifugal Pumps" (3)

Vlaming (4) came to similar conclusions, publishing the graph seen in Figure 3. The major difference is that Vlaming, while acknowledging a reduction in cavitation noise at lower values of NPSHA, did not acknowledge a reduction in cavitation damage at the lower values of NPSHA.

Figure 3. Pump Head & Noise as a Function of NPSHA (D.J. Vlaming [4]) Igor Karassik (5) enjoyed telling the story of his experience with condenser hot well condensate pumps, with the performance seen in Figure 4. These pumps were typically installed in systems that allowed the water level in the hot well to control pump capacity. Water in the hot well, at its bubble point, would readily flash (cavitate) in the impeller eye. A lower water level would result in lower NPSHA, which would result in a lower pump head (more flashing) and lower pump capacity. Pumps operating along a vertical segment of their head curve(s) are said to be operating in the break. Pumps made of standard iron/bronze construction had acceptable component life in this service.

Figure 4. Pump Capacity Controlled by NPSHA (I.J. Karassik [5]) When these systems were changed to control pump capacity by throttling the pump discharge, cavitation damaged the pumps at a higher rate of erosion. The higher NPSHA caused the rate of cavitation erosion to increase. McGuire (2; Figure 6.6) confirms that cavitation erosion occurs at a value of NPSHA above the 3 percent head-drop value. How can a greater amount of cavitation create less noise and erode the metal at a lower rate? A number of theories exist. Some authorities speculate that the greater amount of gas creates a cushion for the collapsing bubbles. Others speculate that the larger bubbles, associated with the greater cavitation, collapse with less intensity. This author speculates that the larger bubbles are swept from the metal surface and collapse in the fluid stream, away from the metal. Nobody really knows. As I think back on the field problems I have seen, I recall a few that left us scratching our heads. There were cases of pairs of apparently identical pumps, where one "cavitated" louder than the other, and the bad actor was the one that appeared to have more NPSHA than the quieter pump. Could it be that the noisier pump was noisier because it had more NPSHA? Would it be heresy to suggest that it might be advantageous, under the right conditions*, to reduce the NPSHA to a "cavitating" pump? Based on the facts, this seems to be a reasonable approach. So what is a safe margin for a centrifugal pump? Figure 2 shows that the NPSHA must be pushed well above the value for maximum erosion rate, and this can be in the range of five to ten

times the NPSHR3, normally a value that is not obtainable. It therefore appears best to keep NPSHA near or slightly above (5 to 10 percent above) NPSHR3. *The right conditions would include current operation with NPSHA in excess of the 3 percent head-drop NPSHA and being able to avoid vortexing in the suction vessel, to control the NPSHA and to accept a slight drop (up to 3 percent) in pump head.
Wear Ring Clearances

Figure 1 plots the clearances specified for wear rings in API 610 (1) and in the Pump Handbook (2) as a function of the ring diameter. To obtain optimum (hydraulic and mechanical) performance, manufacturers want to provide centrifugal pumps with minimum clearance between wear rings. To reduce the chance of galling and seizing between mating rings, users want the clearance to be large. Users often do not fully realize the negative impact of increased ring clearance on the hydraulic and mechanical performance of a centrifugal pump. A plant repair shop will sometimes increase ring clearances the first time a pump is serviced. This usually does reduce the chances for ring seizure, but it has numerous undesirable side effects.

Figure 1. Clearances of wear rings

Hydraulic Effects of Wear Ring Clearances

Tests have shown that wear ring clearances significantly impact the performance of centrifugal pumps, particularly those with lower specific speeds. The effect of wear ring clearance on the performance of a pump is shown in Figure 2. This is a 2x3x13 hydrocarbon process pump, which is built in both horizontal and in-line configurations. The single-suction impeller is equipped with rings at the eye and on the back side (conventional construction). The diameter of the throttling surface of both sets of rings is 4 3/8 in. The lower head curve shows the performance of this pump with a 0.018 in diametral ring clearance is only 0.002 in more than the minimum specified by API 610 (1) for a 4 3/8 in throttling surface. The

upper curve shows pump head with the ring clearance reduced to 0.004 in. (To operate with such a small clearance, special casing rings were designed that are free to float and align themselves with the impeller rings.)

Figure 2. Effect of wear ring clearance on performance of a 2x3x13 process pump With the tighter rings, the head increased about 20 ft (3 percent). The reduction in power was more significant-about 6 bhp (7 percent). Efficiency increased about 5 points (10 percent). The reduction in NPSHR ranged from 0 to about 10 ft. Because we plot the characteristics of a centrifugal pump against capacity, we tend, as we did above, to consider the change in characteristics at each individual capacity. We say that the head increased, the power dropped and the NPSHR dropped. These are all correct statements, but it is easier to understand these changes if we consider the head, power and NPSHR to be constants, and the pump capacity to be increased by the amount that the internal leakage was decreased. With this approach, we can see that the power curve shifted uniformly about 32 gpm to the right, a number that agrees well with the calculated reduction in ring leakage. The head curve also shifted about 32 gpm to the right. The NPSH curve shifted to the right, anywhere from 50 to 75 gpm; the excess over 32 gpm being the compound effect of the capacity change, the reduced energy level of the liquid leaking back into the impeller eye and the reduction in the disturbance of the eye flow pattern. This pump has a specific speed of only about 500, so the proportion of the total flow attributable to ring leakage is greater than for larger pumps with 13 in impellers. The proportional impact on pumps with higher specific speeds would be less than seen above, although the absolute change in power would be larger.
Effect of Other Throttling Devices

Other throttling devices have the same effect on pump efficiency and NPSHR. Balancing Device A horizontal, multistage pump is typically equipped with a balancing device (drum, piston or sleeve) that provides axial hydraulic balance for the pump (to lessen the load on the thrust bearing). For this device to function, it must throttle pumpage from some internal high pressure (half or full pump head, depending on pump style) all the way down to suction pressure. This liquid is then normally piped back to the inlet of the first-stage impeller. Because this "balancing" leakage has been through the pump, then past the balancing drum, it enters the eye of the first-stage impeller warmer than the first time it entered the pump. The higher temperature causes the vapor pressure to be higher, thereby requiring a higher inlet pressure to suppress cavitation. That is higher NPSHR. The higher the leakage through the balancing device, the lower the pump efficiency and the higher the NPSHR. The clearance between the balancing device (the rotating part) and its bushing (the stationary part) must be maintained at the manufacturer's recommended value to retain rated NPSHR. The pump NPSH requirement can be reduced by piping the leakage from the balancing device back to the suction vessel, instead of to the pump suction, but all valves in this line must be locked open. (The closure of any valve in this line, while the pump is running, can over-pressure the seal and wreck the thrust bearing.) Throat Bushing In a vertical, multistage, "turbine" pump, a long throat bushing throttles discharge pressure, down to suction pressure, to reduce the pressure on the packing or mechanical seal. This leakage is normally piped back to the top of the can (into the suction cavity). This pumpage is heated by passing through all impellers and bowls, then heated again as it is throttled to suction pressure. As with the balancing leakage discussed above, when this warmer liquid flows into the eye of the first-stage impeller, the pump NPSHR rises. As the clearance in this bushing increases, the flow though it increases, which causes not only a reduction in pump efficiency, but also an increase in NPSHR. The clearance between the throat bushing and the shaft must be maintained at the manufacturer's recommended value to retain rated NPSHR. The pump NPSH requirement can be reduced by piping the leakage from the throat bushing back to the suction vessel, instead of to the pump suction, but all valves in this line should be locked open, to prevent over-pressuring the seal. Seal Flush

A mechanical seal is usually "flushed" by taking pumpage from the discharge of the pump and throttling it though an orifice. It then passes across the seal, from which it flows through the throat of the seal chamber, into the back ring of the impeller through the impeller "balance holes," and into the impeller eye. This pumpage gets heated by passing through the pump, then through the orifice, then past the seal. All this heat increases the vapor pressure of the liquid. More pressure is therefore required at the pump inlet to prevent cavitation of this flush liquid (i.e., more NPSH is required). The impact of seal flush on NPSHR can be minimized by minimizing the flow rate. Provide no more flush than is required to satisfy the seal naturally with a suitable margin. Excessive flushing can actually damage the seal. The author is aware of a propane pump whose NPSHR was reduced by venting the seal chamber to the suction vessel.