Anda di halaman 1dari 4

Reprinted from:

May 1996 issue, p. 63-65. Used with permission.

Pumps and NPSH: avoid


problems and improve reliability
Process engineers can use these guidelines to improve equipment selection

Vessel

G. R. Martin, Process Consulting Services Inc.,


Bedford, Texas
rocess engineers routinely specify and check pumps
even though their understanding of the subject is
often rudimentary compared to a machinery specialist. In the standard process plant, machinery specialists
(if any) usually focus on the delicate, expensive and critical equipment: typically compressors, expanders, centrifuges, rotary driers, etc. Hence, despite the process engineers lack of appropriate knowledge and experience, the
pump specification often goes directly to the pump vendor
without the information required to select proper equipment.
Proper equipment specification refers to more than the
simple volume and head requirements of the service. Reliability, onstream time, maintenance manpower, maintenance cost and safety can be compromised by not understanding the criteria for pump selection.
The most commonly used pump type in process plants is
the centrifugal. Criteria vital for proper centrifugal pump
selection include net positive suction head (NPSH) available,
range of operation, fluid properties, head required, pump
performance characteristics and variability in each of these
areas. Process engineers understand that the interaction
between suction specific speed and seal thermodynamics;
pump reliability; and the impact of process variability from
fluid flowrates, properties, and conditions (to mention a few
items) may be outside their expertise. Here, they should
solicit advice from a competent mechanical consultant. Misunderstanding even an apparently simple concept such as
NPSH can often lead to unfortunate consequences.

P abs
Reference
liquid level

Net positive suction head. Simply, NPSH is the minimum


suction condition (pressure) required to prevent pump cavitation.1 Conceptually, NPSH can be imagined as the pressure
drop between the pump inlet flange and the point inside the
pump where the fluid dynamic action, as it leaves the impeller,
causes a pressure rise. Sufficient NPSH allows for pumping
without liquid vaporizing in the pump first-stage impeller eye
as the fluid pressure drops due to pump suction losses.
NPSH required is reported in head of fluid (being pumped)
required at the pump inlet. As such, NPSH required has units
of length (height). Usually the datum line for pump NPSH
is the centerline of the inlet. This is satisfactory for small
pumps. For larger pumps, the NPSH requirements reported
by the manufacturer should be verified for the datum line
being used. A review of NPSH analysis and pump mechanical details and performance can be found in Karassik, et al.2

Liquid at
V=0
H
L friction loss
NPSH available
calculated at end
of pipe flange

NPSH required
specified at
pump flange
Pump

Evaluate Pv

Fig. 1. The elements of Eq. 2 are illustrated with a pump taking suction from a tower.

NPSH available differs from NPSH required. NPSH


required deteremined during the manufacturers test and
shown on the vendors pump curve is based upon a 3% head
pump differential loss. NPSH available must be large enough
to eliminate head loss. NPSH available is the excess of pressure over the liquids vapor pressure at the pump suction
flange.3 Except in rare circumstances,4 centrifugal pumps
require the NPSH available to be greater than the NPSH
required to avoid cavitation during operation. Determining the
NPSH available is the process engineers job. Determining
the NPSH required is the job of the pump vendor.5
Our concern here is the process system side of determining what NPSH is available. Pressure balance and
NPSH available derive from Bernoullis equation for a stationary conduit where the total energy (head) in a system
is the same for any point along a streamline (assuming
no friction losses). Incorporating friction losses and restating the formula in a form familiar to process engineers,
the NPSH available in a system can be expressed as:6

NPSHa =

2.31 ( P + Pa Pv )
V2
+ S B L +
(1)
sg
2 g

where:
NPSHa = net positive suction head available, ft
P = pressure above liquid, psi gage
Pa = atmospheric pressure, psi
Pv = vapor pressure of liquid at pumping
conditions, psia
sg = specific gravity of liquid at pumping
conditions

HYDROCARBON PROCESSING / MAY 1996

From main fractionator


overhead condenser

Wet gas
compressor

Two-phase

Vapor

30 ft
110F

10 ft

F 20.6 psia(5.9 psia)

3 ft (hydrocarbon)

(Minimum liquid
levels shown)

1 ft (water)

2 ft
L=1.8 ft
1 ft
15 ft

L=0.2 ft

Product
recovery

1.5 ft
Two parallel
water pumps
(vertical
centrifuge)

Two parallel
hydrocarbon pumps
(horizontal centrifugal)

Water treating

Fig. 2. During a revamp project several problems were found with


this installation.

S = static height of liquid from grade, ft


B = distance of pump centerline (suction
nozzle centerline for vertical pumps)
above grade, ft
L = suction system friction losses, ft of pumping
liquid
V = average liquid velocity at pump suction
nozzle, ft/sec
Converting to absolute pressures, fluid density and resetting the datum line to the pump centerline results in:

NPSHa =

144 ( Pabs Pv )
V2
+H L+
2g

( 2)

where:
Pabs = pressure above liquid, psia
H = static height of liquid between liquid level and
pump suction centerline (datum line), ft
= fluid density, lb/ft3
Fig. 1 shows a pump taking suction from a tower bottoms with the elements of Eq. 2 illustrated.
In most applications, pump inlet velocity is low to minimize pressure drop between the liquid level upstream and
the pump suction. For example, at a fluid velocity of 5 ft/sec,
the V 2/2g term reduces to 0.4 ft. Therefore, this term is usually ignored. If the pumped liquid is at its bubble point (Pabs
= Pv) in the upstream vessel, the term 144(Pabs Pv)/
reduces to zero. This gives the most familiar form of the
NPSH equation suitable for a draw from an upstream vessel (tower bottoms, condenser drum) to a pump:
NPSHa = H L

(3)

While Eq. 3 shows the form most commonly understood


by process engineers, the following case histories illustrate why Eq. 2 and the importance of its terms should
never be forgotten. Special cases where the NPSH requirement should be modified to account for inert7 or dissolved8
gases can also be accommodated by modifying Eq. 2.
Main fractionator overhead drum. Fig. 2 shows the overhead drum for a refinery main fractionator overhead system. Mixed phase material enters from the main fractionator overhead condensers. The drum separates the liquid into

a hydrocarbon and a water phase. The deentrained vapor


leaves the drum and goes to the charge gas machine for compression to the product recovery section.
The water is pumped from the bottom of the drum and
sent to water treating. The hydrocarbon is sent to further
product separation units. An internal standpipe in the drum
allows taking a hydrocarbon-only stream to the hydrocarbon
pumps. The drum diameter is 10 ft. The bottom drum level
stands 2 ft above grade. The minimum water level is 1 ft and
the minimum hydrocarbon level is 3 ft. The hydrocarbon specific gravity is approximately 0.7 at conditions. The operating
conditions above the liquid are 20.6 psia (5.9 psig) at 110F.
During a revamp project, several interesting things were
found and corrected. First, use of a vertical centrifugal pump
in such a deep sump for the water phase is very unusual. By
Eq. 2, the NPSH available for the pump is:

NPSHa =

144 ( 20.6 1.28 )


+ ( 3 1)
62.4
+ ( 3 1) * 0.7 1.8

( 4)

V 2/2g

has been set to zero and that


Note that the term
H has been modified to account for the difference in the
hydrocarbon and water densities. H is now:
H = (Waterlevel pumpsuction)
+ (HClevel waterlevel) * HCsg

(5)

NPSHa = 46.2 ft

(6)

The vertical centrifugal pump had an NPSH required of 1 ft


at the flowing rates. Why would such an expensive pump be
used when an ordinary horizontal centrifugal pump would
have been suitable? Examining the specification sheets for the
original pump purchase over 20 years ago, showed that the
NPSH available was listed at 1 ft. Given a safety margin for
pressure drop in the line to the pump, the only way to calculate
1 ft of available NPSH is to assume that the pumped fluid is at
its bubble point in the upstream vessel. This is the assumption for Eq. 3. In this case, the water was not at its bubble point.
The more rigorous form of the NPSH calculation in Eq. 2 should
have been used. To accommodate the specifications, the vendor
offered a pump with a very low NPSH requirement.
The water phase has a vapor pressure (from steam tables)
of 1.28 psia. Failure to account for the separation between the
water and hydrocarbon phases and use of the wrong vapor
pressure greatly increased initial equipment, installation, and
maintenance costs. Any thoughtful inspection of the pump
installation should have immediately raised the question, Why
do we have a centrifugal pump (with inducer) on one service in
this drum and a vertical centrifugal on the other? The place to
correct errors of this type is during P&ID review before equipment is ordered. For this, the review engineer needs a sound
understanding of pump basics plus the willingness to question and think about every part of the proposed layout.
Solvent recovery system. Fig. 3 shows the bottoms pump
of a sulfolane unit solvent recovery tower. The system design
basis assumed the exchanger would reduce the fluid vapor
pressure enough to overcome the exchanger pressure loss
and provide available NPSH higher than the pump requirements. This would prevent cavitation. Additionally, the colder
sufolane temperature would improve pump seal life.
During the evaluation, the plant was running at 50%
above design. The pump was vibrating noisily with all the

HYDROCARBON PROCESSING / MAY 1996

signs of cavitation. Maintenance records showed that the


pump had continuous seal and impeller problems. Unquestionably, cavitation was occurring. The solution was to
move the exchanger downstream of the pump.
When considering putting an exchanger upstream of a
pump to reduce the pumping temperature, the system
hydraulics at off-design operation must be understood.
Invariably, plants are pushed to their maximum limits.
Basically, this means that engineers specifying exchangers upstream of pumps must understand the empirical
relationships between exchanger pressure drop and heat
transfer. These can be summarized as:9

Sulfolane solvent
recovery tower

25 ft

Lean sulfolane

hshell
= 0.20 Re 0.4 Pr 0.67
cG

(7 )

htube
= 0.023 Re 0.2 Pr 0.67
cG
0.326 2 Lo
Pshell =
Wo 2
gc
PB Do
1.8

Ptube

Z 0.2 W
= i i
si n

NPT

Rich sulfolane

(8 )
(9 )

Lo

+ 25
di

(5.4 di )3.8

(10 )

c = specific heat, Btu/lbF


di = tube diameter, in.
Do = inside shell diameter, in.
G = mass flux, lb/hr ft2
gc = gravitational constant, 4.18e08 ft hr2
h = heat transfer coefficient, Btu/hr ft2F
Lo = length of shell, ft
n = number of tubes per pass, dimensionless
NPT = number of tube passes per shell, dimensionless
PB = baffle spacing, in.
Pr = Prandtl number
Re = Reynolds number
si = specific gravity, dimensionless
W = flowrate, 1,000 lb/hr
Z = viscosity, cp
P = pressure drop, psi
These equations are for fully developed turbulent flow
with a Reynolds number greater than 10,000, with sensible heat transfer on both sides of the exchanger, and
crossflow on the shell side. This is the most common type
of shell-and-tube exchanger configuration.
Given that: 1) the overall fluid properties on each side
of the exchanger do not change, 2) the rate change on each
side of the exchanger is proportional, 3) the exchanger
geometry does not change, and 4) substituting the definitions for the Reynolds number and the Prandtl number, we
find that the heat transfer coefficients and pressure drop
of the exchanger varies with flow by:
hshell G0.6

(11)

htube G0.8

(12)

Pshell

W2

(13)

Ptube

W1.8

(14)

These relations show that for a given flowrate increase,


the increase in heat transfer coefficient (and overall heat
F/5M/8-2000

P=20 psi
Pump

Fig. 3. The intent of this arrangement was to cool the bottoms sufficiently to prevent cavitation.

transfer) will be proportionately far less than the pressure drop increase. High fouling services, such as sulfolane, accentuate this. The overall heat transfer coefficient, Uo, is related to the coefficient on each side by the
equation (ignoring tube resistance):

1
1
1
=
+
+ fouling factor
Uo htube hshell

(15)

If the fouling factor is high, it dampens the overall heat


transfer coefficient rise with increasing flowrate.
Based on the result that the pressure drop through an
exchanger between the tower and the pump suction will
rise at a minimum 1.2 to 2.5 power faster than the
exchanger heat transfer capability, exchangers between
the bubble point liquid and a pump suction should be
avoided. If absolutely required, the engineer should carefully examine the system hydraulics and heat transfer
characteristics at expanded, as well as design, rates.
1
2
3
4
5
6
7
8
9

LITERATURE CITED
Karassik, I. J., W. C. Krutzsch, W. H. Fraser, and J. P. Messina, Pump Handbook, Second Edition, 1985, McGraw-Hill Book Company, New York, p. 2.215.
Ibid., pp. 2.215229.
Benedict, R. M., NPSH and Centrifugal Pumps, Chemical Engineering Progress, Vol.
66, No. 5, May 1970, pp. 5859.
Shah, G. C., Inert Gas Injection Reduces Pump Noise, Chemical Engineering, July 8,
1974, pp. 93.
Sherzer, A. F., Net Positive Suction Head, Chemical Engineering Progress, Vol. 55, No.
9, September 1959, pp. 7984.
Dufour, J. W. and W. E. Nelson, Centrifugal Pump Sourcebook, 1993, McGraw-Hill, Inc.,
New York, p. 36.
Penney, W. R., Inert Gas in Liquid Mars Pump Performance, Chemical Engineering,
July 3, 1978, pp. 6368.
Tsai, M. J., Accounting for Dissolved Gases in Pump Design, Chemical Engineering,
July 26, 1982, pp. 6569.
Lord, R. C., Minton, P. E., and Slusser, R. P., Design of Heat Exchangers, Process Heat
Exchange, Cavaseno, V. editor, McGraw-Hill Publications Co., New York, 1979, pp. 1436.

The author
Gary R. Martin is a chemical engineer with Process Consulting Services, Inc., Bedford, Texas. His
responsibilities include revamping and troubleshooting refinery processes. Mr. Martin specializes in improving refinery profitability by troubleshooting, optimizing and revamping refinery
units. He previously worked as a refinery process
engineer and distillation system troubleshooter. Mr.
Martin holds a BS degree in chemical engineering from Oklahoma
State University. He is the author of more than 40 revamp and troubleshooting technical papers.
Updated 6-8-2000

Copyright 2000 by Gulf Publishing Company. All rights reserved. Printed in USA.

Anda mungkin juga menyukai