Вы находитесь на странице: 1из 6

Predicting centrifugal compressor

performance
When a plant revamp is planned, assessing an existing centrifugal compressor
to meet new performance requirements is essential

TEK SUTIKNO
Fluor Enterprises

A
mong various types of
1.20 1.10
compressors commonly Discharge pressure
1.15
used in process plants,

Discharge pressure ratio


Polytropic heads 1.05
1.10
the centrifugal type is widely
Head at design point
Polytropic head

used and suitable in certain 1.05


1.00
applications such as high gas 1.00

flow rates. For some process- 0.95 0.95

ing plants such as hydropro- 0.90


0.90
cessing units, the capital cost 0.85

for the centrifugal compressor 0.80 0.85


recycling the reactor loop gas 0.75
stream is often the highest rel- 0.70 0.80
0.7 0.8 0.9 1.0 1.1 1.2
ative to those for the remain-
Actual inlet flow rate (Q)
ing equipment types or items. Design flow rate
The compressor driver also
requires the highest rate of Figure 1 Ratios of polytropic head and discharge pressure vs flow Q ratio – design
mechanical or electrical energy base case
consumption. When an exist-
ing plant needs to be revamped evaluating overpressure risk the evaluation of the existing
for expansion, repurposing or levels and identifying any addi- compressor. To maximise the
other goals, assessing the capa- tional modification scope neces- potential reuse of an existing
bilities of the existing centrifu- sary for the connected systems. compressor in a new revamp
gal compressor to meet the new While the compressor vendor project, the ability to predict the
performance requirements is may be requested to provide compressor performance can
an essential step for identifying modification options for reus- certainly be essential for assess-
cost-effective revamp options ing the existing compressor to ing or pre-screening the modifi-
without the need for a new meet new performances needed cation options more effectively
compressor and the associated for a revamp project, the ven- and in a more timely fashion,
systems. In cases where the dor typically requires a new especially when several options
compressor needs to be mod- commercial contract and prob- need to be evaluated at a time
ified to meet the new require- ably considers the option for a and additional options may
ments, the performance of this new compressor replacement come up later.
modified compressor at off- is more lucrative than the reuse This article discusses a cor-
design or upset conditions will option, in addition to the time relation method for predicting
also need to be predicted for period required to complete the performance of an existing

www.digitalrefining.com/article/1002245 Revamps 2018 1


1.05 (𝑛𝑛 − 1) (𝑘𝑘 − 1)
= (3)
𝑛𝑛 𝑘𝑘𝑘𝑘
1.03

Polytropic efficiency ratio The compressor discharge


1.01

(η/η at design point) 0.99


pressure Pd in psia at a given
0.97 rotation speed (N) can be esti-
0.95 mated using Equation 4. As
0.93 shown, a number of gas proper-
0.91 ties and suction conditions are
0.89 needed to predict the discharge
0.87 pressure, Pd:
0.85
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 𝑛𝑛 𝐻𝐻 𝑀𝑀𝑀𝑀 (𝑛𝑛 − 1)
𝑃𝑃" = 𝑃𝑃% exp ) 𝐿𝐿𝐿𝐿 0 + 1<=
Actual inlet flow rate (Q) (𝑛𝑛 − 1) 1545 𝑍𝑍789 𝑇𝑇% 𝑛𝑛
Design flow rate

(4)
Figure 2 Polytropic efficiency vs Q ratio – design base case where :
Ps – Suction pressure, psia
centrifugal compressor in new compression path represented n – polytropic exponent
operating conditions and pre- as Equation 1 describing the H – head in lbf-ft/lbm
sents the results from specific pressure (P) and specific vol- MW – molecular weight, lbm/
examples. This method utilises ume (v) relation for this path: lb mole
parameters reported in the lit- Zavg – average compressibility
erature and mainly involves P1 v1n = P2v2n (1) factor
correlating an available set of Ts – suction temperature, °R
performance data to predict the For ideal gas following Cp
performances at new operating – Cv = R, where R is the gas Flow and head coefficients
conditions. constant, Cp heat capacity at The head data in Figure 1 can
constant pressure, and Cv heat be used to generate plots of
Existing performance data capacity at constant volume, head coefficient (ψ) versus flow
The performance curves pro- Equation 2 can be derived to coefficient (φ) and η versus φ.
vided by the vendor for an describe a reversible, adiabatic Equations 5 and 6 define φ and ψ
existing centrifugal compressor process, and kideal equals to Cp/Cv, respectively:
typically include the polytropic or ideal gas heat capacity ratio:
head (H) versus the volume 700 𝑄𝑄 (5)
𝛷𝛷 =
flow rate (Q) at suction and P1v1k = P2v2k (2) 𝑁𝑁 𝑑𝑑 )
polytropic efficiency (η) versus
𝐻𝐻 (1300) )
Q. Figures 1 and 2 show these
𝜓𝜓 =
𝐼𝐼 𝑁𝑁 ) 𝑑𝑑 )
(6)
typical performance curves The polytropic exponent n
provided by the vendor except describing the adiabatic Pvn where:
that heads and flow rates are compression path can be esti- Q – inlet volumetric flow rate,
presented as ratios based on mated by Equation 3 where acfm
the rated or design operating η is the polytropic efficiency N – rotational speed, rpm
points. The left vertical axis of shown in Figure 2 as a function d – impeller diameter (or effec-
Figure 1 shows the polytropic of the inlet volumetric flow tive), IN
head ratios and right verti- rate. k may be estimated from I – number of impellers
cal axis discharge pressures Pd molecular weight and temper- Figures 3 and 4 respectively
ratios. The polytropic heads ature correlations or obtained show the plots of head coeffi-
plotted versus the actual volu- from a commercial simulation cient ψ versus flow coefficient φ
metric flow rate at the suction program. The average of ks at and polytropic efficiency η ver-
condition in Figure 1 are typi- compressor suction and dis- sus φ. These are derived from
cally derived by assuming an charge is generally used for the existing set of curves pro-
essentially adiabatic polytropic estimating the k value: vided by the vendor for the

2 Revamps 2018 www.digitalrefining.com/article/1002245


base case or design case of the
existing compressor. As will be 0.65
discussed next, these figures 0.60
can be used to predict the new

Head coefficient (Ψ)


0.55
performance of the compressor
0.50
when one (or more) of the oper-
0.45
ating parameters has changed.
0.40
Operating conditions 0.35
One of the operating conditions 0.30
affecting compressor perfor- 0.25
mance is the suction pressure.
0.20
Reducing the suction pressure 0.03 0.04 0.05 0.06 0.07 0.08
decreases the gas density and
Flow coefficient (φ)
increases the inlet volumet-
ric flow rate, Q. As indicated FIgure 3 Head coefficient vs flow coefficient – design base case
in Figures 1 and 3, higher flow
rates decrease the polytropic
head and the associated dis- 1.01
charge pressure. For a system
Polytropic efficiency ratio

1.00
where the discharge pressure is
(η/η at design point)

0.99
controlled to stay constant, the 0.98
compressor delivers less mass 0.97
flow rate as the suction pres-
0.96
sure is reduced. Conversely, the
compressor in the same con- 0.95

stant discharge pressure sys- 0.94


tem delivers more mass flow 0.93
rate when the suction pressure 0.92
increases. 0.03 0.04 0.05 0.06 0.07 0.08
To predict the performance Flow coefficient (φ)
curves of the same compressor
at a lower suction pressure and Figure 4 Polytropic efficiency vs flow coefficient – design base case
other variations of parameters
such as the rotational speed,
flow coefficient φ is first calcu- 1.15
lated using Equation 5 for the 1.10
required range of inlet volu- 1.05
metric flow rates, Q. Based on 1.00
Head at design point

the calculated φ values, Figures


Polytropic head

0.95
3 and 4, which are derived
0.90
from the available performance
0.85
curves, are used to predict the
0.80
head coefficient ψ and the poly-
tropic efficiency η. 0.75

As an example, a new oper- 0.70


Predicted from φ and Ψ
ating Case A revises the suc- 0.65
Vendor data
tion pressure from the known 0.60
0.6 0.8 1.0 1.2 1.4
base case value (100%) to 89%
Actual inlet flow rate (Q)
of this value, and the rotational Design flow rate
speed to 99.7%; other parame-
ters remain the same as those Figure 5 Polytropic head ratio vs Q ratio – Case A

www.digitalrefining.com/article/1002245 Revamps 2018 3


ditions and properties – pres-
1.02 sure, temperature, molecular
Predicted
1.00 weight, and compressibility Z.
Polytropic efficiency ratio Vendor data
(η/η at design point) The effects of suction pressures
0.98
and temperatures have been
0.96 discussed in the example Case
0.94 A. As shown in Equation 4, cal-
0.92 culating the discharge pressure
(Pd) requires gas properties –
0.90
molecular weight (MW), Z, and
0.88 polytropic exponent n which
0.86 depends on the specific heat
0.6 0.8 1.0 1.2 1.4 ratio (k) and polytropic effi-
Actual inlet flow rate (Q)
ciency (η). The values for η are
Design flow rate
available from the correlation
Figure 6 Polytropic efficiency vs Q ratio – Case A (see Figure 4); the remaining gas
property values, however, need
to be known to calculate cor-
1.20 rectly the discharge pressure.
For the same compressor dis-
Polytropic head/Rated head

1.15
1.10 cussed so far, Case B, where
1.05 MW and suction pressure dif-
1.00 fer from those in the base case
0.95 or Case A, is another example
0.90
case for predicting the new per-
formance curves including the
0.85
discharge pressure Pd. Figures
0.80 Predicted from φ and Ψ
Vendor data 7, 8, and 9 respectively show
0.75
Coeff. with X the predicted heads, efficien-
0.70
0.6 0.8 1.0 1.2 1.4 1.6
cies, and discharge pressures
Actual inlet flow rate (Q) for Case B with a MW equal to
Design flow rate 115% of that in the base case
and suction pressure at 92.3%
Figure 7 Polytropic head ratio vs Q ratio – Case B of the same base case. Solid
blue lines in these figures show
in the base case. For this Case a system controlled at a con- data provided by the compres-
A, Figures 5 and 6 show respec- stant discharge pressure, how- sor vendor and the dots are the
tively the new heads (H) and ever, a higher Ts results in a values predicted using flow
efficiency (η) taken directly lower Q to increase H for reach- and head coefficients.
from Figures 3 and 4. The solid ing the controlled discharge As Figure 7 shows, the pre-
line in Figure 5 or 6 represents pressure. Similar to suction dicted polytropic head data
data provided by the compres- pressure variation, the flow and (red dots) in the inlet volumet-
sor vendor for this Case A, and head coefficient correlation can ric flow rate range close to the
the dots show the predicted be used to generate the new design or rated flow rate agrees
data which agree with the ven- performance curves at new suc- reasonably well (less than 1.5%
dor data. tion temperatures. difference) with the vendor
The suction temperature data. However, at higher flow
(Ts) relates to the gas density. Gas properties rates, or close to the stone wall
Q increases somewhat as the The inlet volumetric flow rate point, the correlation method
suction temperature becomes Q in Equation 5 for calculating underpredicts the vendor data
higher, and the polytropic head the flow coefficient is the only by about 4.0% maximum.
(H) decreases at higher Qs. For parameter affected by gas con- The polytropic exponent n

4 Revamps 2018 www.digitalrefining.com/article/1002245


values used for calculating the
discharge pressure Pd are based 1.02
on the ideal specific heat ratio
1.00
or kideal which varies somewhat

Polytropic efficiency ratio


with the temperatures. As the

(η/η at design point)


0.98
gas temperature will increase 0.96
through the compressor, kideal at
0.94
the discharge side will not be
the same as that at the suction, 0.92
and the average of suction and 0.90
discharge kideal values is nor- Predicted from φ and Ψ
0.88
mally used for calculating the Vendor data
discharge pressure. For a given 0.86
compression ratio, however, 0.65 0.85 1.05 1.25 1.45 1.65

the gas temperature will also Actual inlet flow rate (Q)
Design flow rate
increase more at lower poly-
tropic efficiencies, and the aver-
age kideal could vary with the Figure 8 Polytropic efficiency vs Q ratio – Case B
flow rate Q. However, using a
fixed value of average kideal for
all flow rates in the operating 1.10

range typically provides rea- 1.05


Discharge pressure (Pd)
PSIA/Pd at design point

sonable results. The average 1.00


value of suction and discharge
compressibility factors, Zavg, 0.95

also varies with pressure ratios 0.90


or Qs, but the variation of Zavg 0.85
in this example is less than that
of kideal. 0.80 Predicted from φ and Ψ
The predicted Pds for all Qs in 0.75 Vendor data
Using X factor
Figure 9 are shown as red dots 0.70
and based on a constant, aver- 0.65 0.85 1.05 1.25 1.45 1.65
age value of kideal. For the new Actual inlet flow rate (Q)
operating condition, Case B, the Design flow rate
values of kideal at suction and dis-
charge can be estimated from a Figure 9 Discharge pressure ratio versus Q ratio – Case B
commercial program but might
vary dependent on the property rection (see Equation 7) can be lar weight (115% of base case).
estimation methods. As Pd for used to calculate the discharge Again, for a new operating
Case B or a new operating case pressure, Pd, and the polytropic case where the new Pd is not yet
needs to be estimated first to heads are calculated from known, iteration of Pd is likely
get the discharge kideal, iteration Equation 4. The green line and needed to calculate the kideal at
of Pd value may be needed until the green dots in Figure 7 show the discharge condition. This
the average kideal value used in the predicted H data using the discharge kideal is needed to cal-
Equation 4 results in the calcu- X factor method with an aver- culate the average kideal (or k2 in
lated Pd matching the trial value age kideal value which is assumed Equation 7) for the new oper-
of Pd. Alternately, Pd can also be constant from one flow rate ating case. The k2 value can be
estimated by correlating the X to another, and the kideal val- used for estimating the Pd pro-
factor discussed next. ues at suction and discharge file when Pd assumed to cal-
Instead of using Equation 4 conditions have been derived culate the discharge kideal value
for calculating the discharge from a simulation program on agrees with Pd calculated from
pressure Pd, the X factor cor- the basis of the new molecu- Equation 7. For this example,

www.digitalrefining.com/article/1002245 Revamps 2018 5


Case B, the green dots in Figure cating these options. Increasing modifications. The predicted
9 show the predicted discharge d or I, for example, requires not performance data agree rea-
pressures using the X factor only extra space in the existing sonably well with the example
method. The predicted or cal- compressor but also evaluation sets of performance data from
culated Pd at the rated flow of the impeller tip speed Mach the vendor. For the example
rate is less than 1% different number, discharge tempera- case involving a 15% change in
from the Pd from the vendor. tures, and other mechanical the gas molecular weight, the
At flow rates close to the end limits. It would likely be worth- correlation method results in
of the curve, the Pd difference while to first consult with the about 2% maximum under-pre-
is about 2% maximum. For this compressor vendor before con- diction of discharge pressure
example, the X method gives sidering the options of increas- at flow rate ranges away from
reasonable agreement with the ing d or I as viable. the design flow rate. As the Cp/
vendor data which are typically For an existing compres- Cv value changes with MW var-
derived from ‘wheel to wheel’ sor equipped with a variable iation, the correlation method
calculations and likely based on speed driver such as a steam might require iteration when
in-house k correlation data: turbine, expander, or motor the new performance require-
with speed-adjustable hydro- ment involves a change in MW.

𝑀𝑀𝑀𝑀" 𝑘𝑘% (𝑘𝑘" − 1)
𝑋𝑋" = 𝑋𝑋%
dynamic coupling, increas- Nonetheless, the example cases
𝑀𝑀𝑀𝑀% 𝑘𝑘" (𝑘𝑘% − 1) ing the rotation speed N can discussed show the correlation
(7) be a more cost effective option method can be considered for
𝑃𝑃&
(*+,)
*
relative to increasing I or d. screening modification options
𝑋𝑋 = $
𝑃𝑃'
( −1 Overall, to predict the new per- at least in the early stages of
Where formance of an existing com- engineering evaluation.
pressor to be modified with an
Tek Sutikno is a Process Engineering
Mechanical parameters increase in d, I, and/or N, the
Manager with Fluor and a Professional
Compressor impeller diame- correlation method presented
Engineer registered in 11 US states with
ter (d), number of impellers (I), earlier may be considered, and over 27 years of experience in the process
and speed of rotation (N) all the economic benefits result- industries. He holds BSc, MSc, and DEngr
greatly affect the performance. ing from the new performance degrees in chemical engineering and a
Increasing any of these param- can be quantitatively evaluated MBA degree, all from the University of
eters will result in higher dis- against the cost of the compres- Kansas.
charge pressures (Pd) and flow sor modification. Email: tek.sutikno@fluor.com
rates (Q). For example, Q var-
ies proportionally with d or Conclusion
N, and Pd varies with the third A correlation method using a LINKS
power of d or N. While the com- set of existing centrifugal com-
pressor capacity can be greatly pressor performance curves More articles from the following
expanded by increasing I, d, has been discussed to predict categories:
and/or N, the existing com- the performance of this com- Condition Monitoring
pressor may have mechanical pressor at new operating con- Rotating Equipment
or hardware limitations compli- ditions or after hardware

6 Revamps 2018 www.digitalrefining.com/article/1002245

Вам также может понравиться