‘Compeossor Controls Cor, Buletn E94
‘A McGRAW- HILL PUBLICATION
Improved surge control for
centrifugal compressors
Surge may occur without being detected by conventional control
systems, cutting compressor and process efficiency and causing
compressor damage. Excess flow recirculation or blowoff can help
avoid some surges, but energy costs are high. Here is a control system
that copes with all surges without incurring such penalties.
Naum Staroselsky and Lawrence Ladin, Compressor Controls Corp.
CONTENTS:
1. INTRODUCTION 175
4. Improved Surge Control Considerations
b. Surge Characteristics
Il. SURGE LIMIT LINE 176
The Theory and Equations Showing influence
of Temperoture, Molecular Weight, Ete
|. SURGE CONTROL LINE - The Recommended Equation 179
IV. INSTRUMENTATION AND VALVES - Required Speed 179
V. ANTISURGE VALVE AND Selection Criter 80
vi (TROL: ISOLAT 180
Functions of an Improved Controller
Vil. INTERACTION WITH PROCESS CONTROL 83
VIII. INTEGRATED SYSTEM OF PROCESS AND ANTISURGE CONTROL 184
IX SETTING AND TESTING EQUIPMENT 184
X. EXAMPLE OF ENERGY SAVINGS 184Improved surge control for
centrifugal compressors
Surge may occur without being detected by conventional control
systems, cutting compressor and process efficiency and causing
compressor damage. Excess flow recirculation or blowoff can help
avoid some surges, but energy costs are high. Here is a control system
that copes with all surges without incurring such penalties.
au Staroselsky and Laucrence Ladin, Compressor Controls Cop.
Ci The operation of a centrifugal compressor can be-
‘come unstable due to changes in many conditions such
as flowrate, pressure, and the molecular weight of gas.
‘This causes rapid pulsations in flow, called surge. No
system is immune to sudden upsets at one time or
another.
We have surge-tested air compressors, gas compres-
sors, centrifugal compressors and axial compressors, at
high and low pressures. The amplitude of the low drop
and the frequency of surge cycles vary. However, there
are certain common results: the pressure change is less
than the flow change in every case that we have re-
corded; flow usually drops extremely fast just before
surge, and then always drops precipitously during
surge, at least in the authors’ esperience—a time of
0.05 s from setpoint flow to reverse flow being common.
‘The speed of some surges is such that conventional
control systems cannot detect them, let alone react to
them. So, often even when records indicate that no
surge has taken place, stripping down a compressor will,
reveal surge damage, ranging from changes in clea
ances, which exact a penalty in compressor efficiency, t0
destruction of parts
‘Unstable operation—whether itis detected or not—
affects the operation of the process that relies on the
‘Typically, surge is prevented by recirculating some of
the flow or by blowing off the excess flow. Constant
recirculation af flow—often at 30-40% of the amount
needed by the process—is common. For instance, this is,
found frequently in units used to compress chlorine and
wet gases, among others. Obviously, this excess flow
‘costs energy. And it does nat necessarily avert all surges.
Improved surge control
Here, we will present an improved surgecontol
strategy and instrumentation system, This system copes
vith ven very rapid upsets minimizing disturbances o
proce operations. The amount of recreation re
Chced sharply, saving sgncane amounts of energy.
Furthermore the sytem lends el to actual Bld a
calation of the suge lit
With conventional cweloop control schemes, once
suige begin the comprar cmtolysten "cx
Iate—and the only way to sop this is by manual
reid Hr if marge occur he comeprenc wl atr
tomatcally be brought out off ring the fst ge
Spel, and the contrl stem il et fel to prvest
Sch collation from Fecuring.
Conventional syutens can be at ito otilation by
operator cor nich ero are minimized by this
{Gn Alto, Auctatons inthe pressure and owrae of
the comprewed gn are kept to. minimum, resulting in
Tntvely mines dnurbunee tothe proce (The fon
Sons forthe advantages wll be explained later oo)
Tn order to develop this contol system, we wil rst
relate sarge to compres operation, then determine
tthe surge begin and where to set te control eater
to prevent it Next, we wil dacs instrumentation,
‘contol strategies and calibrating the equipment, Com:
presor in seres and parallel will not be dined here,
Ste to Tiniations of space.
Surge characteristics
ig, 1a shows characteristic curves for a typical sin-
sl-stage centrifugal compresor. Each rotational speed
(N,, Na, Ny, Ny) has its own characteristic curve.
‘the compresor is operating at speed Nat Point A
(mass flowrate W,) and the flow is dereased, surge will
‘oorur at Wp. The pressure has increased from Fog to
Pg. Point B les on the surge limit line, a naturally
‘occurring line that is peculiar to each compressor ge-
‘om etry. The area tothe left ofthis line is the surge zone,
where there are oscillations of flow and pressure
‘int y pci non Chil ageing News May, 19 Copa 99 Mei le, HL Av the Amin No Yar NY.Som ot pa
(et
i
I
i
7 7
Characteristic curves and surge control lines define regions of operation for compressor
‘The line tothe right of the surge limit line isthe surge
control line, which is an artificial line set by an engi-
neer. These lines are the boundaries ofthe recirculation
zone, where recirculation or blowoff is used to prevent
surge. To the right of the control line is the safe operat-
ing zone, where recirculation is considered unecessary
In order to develop the control strategy, we must frst
develop improved equations for the control and limit
Hines
Surge limit line
Here, we will formulate an equation for the surge
ine that does mw assume constant inlet tempera-
re and molecular weight. Typical equations assume
that these parameters are constant, Corrections must be
applied when conditions change. Such is the case with
White [/], whose method is widely used.
‘The head ofa centrifugal compressor isa function of
angular velocity, speed of rotation, inlet volumetcic
flowrate and impeller diameter. Using a form of the
kinetic energy of the comprestor, given by Davis and
Corripio (2, and performing a dimensional analysis:
a)
Fi o
‘At the high velocities at which centrifugal compres
sors operate, the variation of Reynolds number, Re,
with the velocity of the gas is negligible, The variation
of Mach number, Ma, with velocity also is negligible
Mach number can be assumed constant as long as the
2s velocity does not approach the speed of sound, We
will assume that operation is well below this speed in
the region close to surge
For a given compresor geometry, under these cond
tions, if H,/N? is plotted against Q,/N, then the per-
formance Curves for different speeds will reduce to one
curve, called the universal performance curve (Fig. 2)
176
|
|
Fig.
‘The surge limit line is reduced to a single point, the
surge limit. Since the surge limit is fixed fora particular
geometry, the values of this point are constant
ae @
Ge = Cy 8
Thus, the curves ofthe surge limit line follow the fan
law, which states that inlet volume is proportional to
speed, and head is proportional to the square of speed.
‘The surge limit may be defined by either Eq, (2) oF (3)
These equations apply not only to single-stage centrif
lugals but also to multistage units without intercoolers.
‘While Eq. (2) and (3) can be used to set the surge
limit (and the surge limit line), this is impractical since
1H, and Q, depend upon a measurement of molecular
weight
Variations in molecular weight cannot be continuc
‘ously measured at the high speeds present in compres:
sors, and therefore must be eliminated from the equa
tion of the surge line. Also, H, and Q, depend upon
variations in inlet temperature and pressure. We will
climinate molecular weight and temperature from the
‘equation of the surge limit line. Eq, (2) and (3) are
combined to eliminate N
CoH, = 2 “
Polytropic head is given by
°
Substituting Min Eq. (4)
C2uT (R=) _ 9
sao ehNomenclature
based
Distance between surge contal line
and surge limit line (n= 12,3), in. HO
‘ = Unita vay rom
>
4 Bias aed by summing device in No 5,
Big 7
4 Bias added by srmingdevie in No.7,
ig, pa
& [Aceleration of gravity, 6/5
i, Polyropi hed lb
|e Ratio of specie eat at constant pre
| ture to pele heat at conta volume
4 Slope of surge ental ine, dimensiones
| a ‘Molecular weight
| Me Mach number
y Speed of rotation, rm
ae Diferenalprenure acrs compresin,
i
| sty Bfhue Once prure a suction: char ps
| Re" Section paar: dicharye prs
, Yolomeire Howrate fn suction, etal
oe
Rn ocr ees
i eesti
TT, Temperate at nction ich
¥ Mase lwrate fm
ze Average cmpeeniiy = (Z, + 23/2
Grok ts
» Poisons eicienes
Votumetrie flowrate equals mass Havstate divided by:
“7PM 2,1)
Substinaing Q, in Eq, 6
Using an equation tor gas low across an orifice ithe
suction of the compresir
PPM
W a
Tey
\
Eq, i becomes
or
a
ig, H1 may be written for the discharge side of the
compressor. Since the mass flowrate at any instant in
|
4 |
sarge iit
|| Drivers perform
the surge limit fine toa single point
Os 1s 30
Fig. 3Bee
the compressor the same at suction and discharge, Eq
(9) can be written as:
[PenaPM
Wwe Feral 2)
V u
Or
Mone = Moe GTZ as)
Eq, (11) becomes:
ZTh (Re
ae MPa (14)
small, then Z,
bbe approximated by a
less than 2.5, the approximation C
used. Eq. (11) and (14) become:
GylPy — PY = Cad, a5)
Ply
ane Myeg (16)
Eq, (15) isthe same as one developed by White [J].
However, White arrived atthe same result by assuming
that inlet temperature and molecular weight were con
‘Surge, detected by diffused silicon device, occurs too fast for pneumatic transmitter Fig
stant; he later offered corrections for these. Here, s
hhave shown that such corrections sve unnecessary” As
White's case, here the surge limit Line is unaffected by
variations ip suction peessare
IA, is greater than 2.5, (K,? — 1) @ may be approw
mated by CAP, — 1) + and the surge lirat line
becomes:
CyAP +4 7 a7
When changes in inlet presure are negligible:
CMe + by = MPa, us)
where b, is a constant
Note that the above equations apply to centrifugal
compressors without inereoolers, With mntereoolers, iis
necessary to consider each section as a separate com.
pressor, or use an approximation chat considers «le
tunic asa shal
Eq, (15), (17) and (18) are recommended for eale
lating the siege limit line, as thes contain the fe
umber of varables, all of which an be easily
tired. Sometimes, the surge limit ve i obtaaned
Eq, (3), When this is done, it is nevesary to negle
vatiations in molecular weight, and this can lead 10
Inaccuracies, even on aie compressors, since the humid
ity of ambient air varies,
178 ‘HERES ENGSEERIRE RAVEFlow drops precipitously before surge cycles begin, then reverses quickly Fig. 5
Surge control line
‘The three comoion forms of the surge control Tine are
shown in Fig, 1. The optimal position of this fine is
parallel to the surge limit line (Fig, 1a). To minimize
recirculation, the surge conttol line should be set as
close to the surge limit line as posible. Setting the
control line with a slope less than that of the limit fine
(Fig, tb) can lead to excess recirculation at high pres-
sures, and surge at low pressures during stopping and
startup. The third method isto select a minimum safe
volumetric flow, and set a vertical control line (Fig, Ic).
‘This ean lead to excess recteulation at low pressures,
and surge at high pressures, Many systems measire flow
in the discharge without correcting for suction condi-
tions. Ths ives maximum recieulation with minimum,
singe protection and is not recommended
It itis desied to keep the control line parallel co the
limit Fine, how close together can the t xo lines be?
“This depends upon how accurately the surge control
Tine is set and how wel i¢ accounts for changes in inlet
temperature and pressure, and molecular weight
‘Also critical to the location of the control ine isthe
effectiveness of the antisurge system in handling upsets
‘This eflectiveness depends on the control strategy cho
sen and the nature of the antsurge system, its transmit
ters, controller and antisurge valve.
"Fo set the surge eonteol line equidistant from the
limit line, Eq. (15) can be used. The conteol line is
displaced to the right by some fixed amount, 5,
CyyMP, +b, = Mpg a9)
Eq, (19) applies when R, is less than 2.5. For values of
2.5 and greater, Eq. [17) may be used. The conteo line
is displaced by some fixed amount,
CishP + aah, + by = APoyy (20)
Instrumentation and valves
How fast must the antisurge system be to detect the
onset of surge and effectively stop it?
Tis often thought that a very rapid response is not
raecesary: alo, many believe that protection against
“arge upsets is unnecessary, However, upsets that are
oth fast and large ate not rare, and these can be caused
by factors such as jammed check valves, operator errors
and shutdowns of process equipment triggered rapidly
by protective systems
We have found that surge cycles can occur faster than
is thought posible, and that often surge goes unde-
tected. This is because conventional pneumatic control-
lers are to0 slow to detect this phenomenon, Only fast
transmitters can cope with the high speeds of compres-
sor transients, We recommend a diffusedsilicon elec-
{onic transmitter, Wealsorecommend pressure-differen-
179sal ransters with lag not greater than 0.0385 sas | sometimes complet blockage ofthe system ours, and
necesary to mainiain surge contulandtoeaibrate tie | allo the fow mas be recreulated
Surgelimiine. Also, recorders (wed forcalirating and | Field wets show thatthe full ake of the aniurge
testing the anisurge system) must have a char speed | valve shouldbe fom 0. to 1.5 « New valves can meet
tot les than 25 mm/s thi retreat; exizing valves can be pended up
P Irate) Ueseavareal Gerri oe net tenet ieee a [i i cee Beenie incase ie asset esos ea
transmitr with that ofa pneumatic device Presure_| Increasing the preumati signal tothe valve actuator
(Gereial aa menarel trou a ventrt At bout | lWhy should the uanssiter bara a exten
32 5 surge begac, god the pocnosaticuanmniter was | spore de of 0038 0 whdle thc enaiurge al Kast
too Hog to date Gow fachurdoas Risin tine of 13 4? The urge coal it
"The slowness of mow transite and recorders ex- | lve to the mrge Hite lineal, the diane
lain yy records wll show no surge yes opontmper- | berwen theat Wet at 15% of Sign flow, Tha,
thn, compremors reveal changed examen or dam | paral salve tovks wil be enough te nop reovemeae
agedimpalry seals bearings Soch undctctcdsurges | foward sage fuer opening ofthe valve wil reore
wil gradualy ater the conpresor, radting in | the operator poise to the surge contol Ic
trewed ficiency” Some ceetonic’ tammiier are | "The conweler must be fast at wel, Preurate coe
dampened and ae too sow for ui here; others a not_| troller ae two low to reliably prevent surge. Digital
beat ugh joes frurliy doald be usd wi esuton, Soe the
Figs 3 thw ely surge ean bey Tet were) | devices ol la equenceat ach contol lop, eamaing
miade on a cafuga ai axvpremor Git supped =. | taam may not be tort enough to dete the ancien
blast furnace Here marge occanelat about 1s, Osa | typical of eameprewor surge cyclen Micoproconor
Sonali there are prosunge cacllatoms (acrent | cron have « scan tne fm than
reat errare gus Ssseenigtnceny Acie
begins, flow drops precipitously —typically from a set- | Control: isolated antisurge loop
point level to a revere condition hr about 008 » re | Fine, we will look at the satnunge comtol sytem
gardless of pressure or compressor speed (see Fig: 4). | without considering its interaction with the process
: eet Tg val atgrat tes bro sya Th
Antisurge valve and controller eee ae ee apace edie
eee Cooter et ell We at rca pope onal paren caellanaed
large and fast-acting I shouldbe capabcofeceula_ | & relay (oncoff)conteller
ing 100% of design Borate sees accra anaky eater
ioc nee essere te tear met | tinea Groperenst plete rset wf
PP ctor tongs itis ecemary totes quick incearein | antivindup devi (Fig. 6x),
fon ‘The ate at wich ow cane ineaed depends | "Th anindp dee orrry doh tare
tp the repens ofthe valve and iuise! The larger | ofthe proportion rock functions Normally the
thevalv the greaer theft ofits opening, The valve | compresor operas in an area sme distance fom the
should bc abletoveciclat teenie rate because | Conta in, reslting in an offer between the mes
180 neta
ee‘ment and the setpoint of the controller. As a result, the
‘output signal winds up to its high or low limit
‘The proportional band and the operating point will
be on opposite sides of the setpoint or control line—
‘control will nt be affected until the measured operat-
ing point reaches the contro line. Surge can result ifthe
measurement approaches the control line quickly, since
‘overshoot will occur before the controller ean unwind.
Antiwindup adjusts the reset loading to shift the
proportional band to the same side of the control line
that the measurement ison when the controller reaches
its output limit. Then, ifthe control lin is approached
rapidly, the measurement enters the proportional band,
and control starts before the value reaches the control
line. Overshoot is reduced.
Derivative control is not recommended, for it can
‘open the antisurge valve far from the surge control line
and can cause system oscillations, Rapid orillations in
flow, even in the safe operating zone, can cause the
valve to open because of the nature of the derivative
response,
The response speed of this controller depends upon
the proportional-band width and reset time. These pa
rameters influence the stability ofthe system, Decreas-
ing the proportional band or increasing the reset time
Increases the speed of the controllers response, but past
4 certain point system stability will be disturbed —all
closed-loop control systems have a stability lim
‘The speed of an antisurge controller's response is
limited mainly by the inertia of the compressor and its
networks, the transmitters, and the ancsurge valve and
The slower these elements ate, the slower the control:
ler must be set. The inertia of the compressor and its
piping network cannot be changed, but the inertia of
the transmitter and the valve and its actuator can be
decreased by a proper selection of faster elements
‘Once fast transmitters and an antisurge valve are
selected, the proportional band level and reset time are
set, based on the parameters of the compressor and its
network. These parameters include the volume of the
system, length of the pipes, and inertia of the compres-
sor. This limitation is common to all closed control
loops with feedback, Thus, modulated control responds
‘well 0 slow upsets and give good control, However, the
valve cannot be opened quickly for fast upsets, de to
Timitations of system stability
"Now consider using a relay for control. ‘The relay is
part of an open-loop system that opens the antisurge
valve to a pre-established level after the compressors
‘operating point reaches the surge control line. The
‘output of a relay device can change from minimum to
maximum at any speed, without upsetting the system's
stability. This is because (1) the relay isan on-off device
and there is no limit as to how fast the device can be
turned on and off and (2) the system is open-loop and
ccillaions cannot be set up in the control system,
However, the relay syatem has certain disadvantages
for antisurge controller. The preadjusted output level
may be either lower or higher than that required for
protection. If it is lower, the compressor’s operating
point will cross the surge limit; if it i higher, the com>
presor will be operated with large amounts of recincu
a [ae
'
emi
| : z ne
P g
2 i
peeea | ieeed oh Cm
ct 24 om
[iiconeteie pina
! ol
on
= oe
ser] ohn
aeepcen el ree baer wo a
eeepc tert ca ney
lation or blowing off when the antisurge valve is
‘opened. Also, the operator can be unsure as to when 0
reset the relay without endangering the compressor.
Comparing the closed-loop system to the open-loop
system has led to a patented method that combines the
advantages of both, while limiting their shortcomings
Fig. 7 shows the controller with inputs for Eq, (18),
‘one form of the surge control line. The input corre
sponding to AP. passes through a sealer (No. 1) and
summing unit (No. 2) that transform it to the required
form:
byAP. + by ay
where fy isthe slope of the surge limit ines and 8 is the
normal distance between the surge limit line and the
sarge control line.
‘The value of £2, + 6, is compared with AP, b
the proportional-plus-reset controller (No. 3) IAP, is
greater, then the controllers output i seat zero:
When S2,,, i less than the signal given by Eq. (21),
the controller begins producing a signal that increases
vunti KAP, +, becomes equal to AF,
‘A: a result, the compressors operating point will be
restcred to the surge control line. This system will work
‘well, disturbances are small and slow, but cannot open,
the antisurge valve quickly enough to prevent surge
wher disturbances are large or fast
‘Tr. handle large disturbances, the output from sum:
min, device (No, 2) is deers~sed by a fixed amount, d,
‘which is typically et ato. . half the value of by. This
addi ion is done by the summing device in element 5
‘The comparator in element 5 compares the value of
APyag with ESP. + by — dy, When AP,,, is smaller, a
signal is sent to the curve generator. The output of the
‘curve generator follows the shape shown in Fig. 7—it
181Sooed setgoin.
isolated contol Antsure
‘opened a speed setpoint is lowered
Fig 8
‘Seed stint,
Integrated system. Compared with Fig. 8, outputs
of controller and speed setpoint vary Fig. 9
increases nearly instantaneously, then exponentially
decreases to zero, This signal goes to a summer ‘No. 4).
overriding the signal from the proportional-plus-reset
controller.
‘Thus, for slow upsets, the valve will be opened slowly
182 SHEET RE
by the action of the proportional-plus-eset controller;
for fast upsets, No.5 will quickly open the valve, over:
riding the slow response of the contoller.
‘The output from No. 5 decreases exponentially, al-
lowing the proportional:plus-reset element to take over.
—“Thus, the operating point is restored to the surge con- “=
trol line asin any other antisurge closed loop. After the 409)
‘output of No. 3 decreases to zero, No. 6 automatically
resets the override system,
Field tests have shown that, ifthe surge control line is
sec far enough from the surge limit line, then the com-
bination of proportional-plus-reset response with the
relay override signal is sufficient to prevent surge
However, ifthe surge control line is set too close 10
the limit line, the operating point of the compressor can
‘ross the limit line before the override signal appears. As
‘result, surge may begin, causing an almost instanta-
neous flow decrease. Under such circumstances, the
‘deviation of the comprestor’s operating point from the
8
Dieter pres Py
Here itis necessary to move the surge control line to
the right-—the task of No. 7. This element, like No. 5,
subtracts fixed amount (¢,) fom the signal from No.2
and compares the sum with AP,,,. When the sum is e
greater than AP, a preselected bias, by, is added:
Tah th+h=ah, @) aa
ee eve ea eel etek ara
ried ester Real fuer
Pee ent cae soca
ja cine rcs sarge ais eve gc
Bp ccaivete ieee genes aceon |p
fet ea tamnsaeerre ter pecan eee
ecore ee een eel ea ce
reaped a etter aoe |
ee eieas a et artie mre arcs
Pate ih setae a eet ta rg eee mec
Pen eieteee red eU ogs (etpeeecatian rae a a a
Fees paar ar el ee Te oe son
the antsurge mucm seen the graph of he | | —
‘new antisurge controller and reduced
antisurge controller output. The horizontal part corre- |_| A new anusurge contaliey and 7
3
sponds to the antisurge valve’s being closed; the nega
Lively sloped line, to the opening of the valve by the
proportional-plusteset controller (No. 3 in Fig. 7) the
Yertical drop, to the action of the override controller
(No. 5 in Fig. 7), oe)
Interaction with process control |
A process control system is shown in Fig, 6a, This
consists of a pressure transmitter, pressure contolle,
And buttery valve with actuator
By closing the butterfly valve, the process controller
Forces the compressors operating point to move toward
surge. Ifthe point croses the surge control line, the
Anuisurge controller will open the amtisurge valve,
‘With both control systems, the deviation of the com-
presor's operating point from the surge control line
{depends on the speed of response of the antisurge con-
troller and of the process controller, and on the dead |
times of the antisurge valve and its actuator.
Ifthe process controller is faster than the antisurge
controller, then this difference may interleee with surge 03,
‘conteol. The dead time of the amtisurge valve can fur- 8000800 Te a0) Tp ~
ther add to this problem. Until the output of the anti Noss flowrate, sm
surge contoller overcomes the dead ime, the process || Se
compressor reflects less recirculation Fig. 11
Soscitic power required lst
controller may push the campressor’s operating point
lover the surge limit line
TT 183‘This interaction between an isolated surge-control
system and the process controller occurs not only with
control of discharge pressure but also with control of
suction pressure and Row.
Integrated system
System performance can be improved by changing
from an iacatedantisurge lop to an integrated proces
control and protective system (Fig. 6)
Tn this system, the buttery valve is controled by the
output of the summing device. Ths applies equally toa
compresor that is under variable speed control The
output of this device depends on the output of the
Ancsurge controller, as well as on tht of the process
controler. The operating point wil, after reaching the
surge control in, fellow this line rather than the line of
With integrated proces control and andsurge pro-
tection, the effet ofthe dead time ofthe antiurge valve
‘will be minimized, since two devices are used to protect
the compressor from surge —the antisurge valve and the
butterfly valve. Tet rents ofthe integrated system are
shown in Fig 9, with a variable-speed compresor.
Setting and testing equipment
‘The surge limit ine can be located precisely, since the
control system stops surge during its frst cycle
To calibrate the surge limit line, the slope ofthe surge
control line, fy, is vet at zero t0 avoid response by the
proportionai-pius-reset controller, The distance between
the surge limit and surge control lines, by, i set at 15%
of flowrate, and bis set at 75%. This 1s because the
control line is vertical and a vaiue of b, = 15% will
induce surge. The distance between the surge control
Tine and the control line of the override antisurge
relay-clement set at zero, so that override wll tart
immediately. The level of the override signal opening
the antisurge valve raust be set at 100%; the time ofits
‘exponential decrease must be not less than 3 min. This,
‘ensures that the valve is kept open.
Ii the discharge line ofthe compressor is closed com:
pletely, surge will be induced. At the beginning of the
first surge cycle, the override antisurge relay-lement
opens the antsueye valve at maximum speed. At the
same time, the comparator (No. 7; Fig. 1) moves the
surge control line to the right, making the distance
between the surge control and surge limit lines 5, +,
‘equal to 90%, which is safely out ofthe surge zone. This
will keep the antisurge valve open, and surge will be
Stopped. The outputsof the transmitters, SP. and MP,
will be obtained at the moment surge occu,
‘Theratio MP,,/AP, yields fy. Ifthe contol lines not
a straight line,"this test can be repeated to generate
diferent values of fy
Setting dy depends on the spoed of response of the
proportonal-plusintegral part of the antisurge com
troller. IF the discharge line is closed slowly, the override
antisurge relay should not operate. Experience indicates
that 4, may be between 2% and 5% of AP,
“The distance, b, between the surge limit"and surge
‘contol line i selected so that the severest disturbances
(forexample, the fastest posible complete closing ofthe
discharge line) will not cause surge. This distance can
be narrow, because the improved antisurge controller
‘operates at such high speed. In most cases, 5, is set
between 5% and 10% of AP... Careful selection of ky
and b, will reduce the energy expense for any compres:
sor that operates with recirculation or blowoff.
Example of energy savings
‘A-4000-hp air compresor was being operated at
constant 40% blowoll sce Fig. 10) The specie power
Fequired by this machine is shown in Fig. Ie The
flowrate was 9.40 acim at 310 pu
“The unit had a pneumatic suge-control system that
was supplied by the manufacturer. In aditon to cone
Sant blowoff, damage was frequent~the machine hed
to be rebuilt twice during one year The contol system
desribed here was installed ata cost of about 813,000
About half the con was for instruments, the othe half
for installation. After installation, the antinurge valve
vas closed and recreation was kept toa minimum. A
valve booster was not needed
Assuming operation on a yearly basis the savings in
energy cons are calculated, with electricity cost of
50024 Wh
kw b
Savings = 0.7465 x 0.760% x
> ye
5) be, $0.024
5 — 0.3825) Px SO 9.400 scfm
(0.4225 ~ 0.3825) ZP- x ES x 9.400 sf
$59,000/yr.
Richard Grane, Bite
References
peeeceieaans
‘The authors
184 SRE RENT RATT
TTT