United States	National Risk Management

Environmental Protection	Research Laboratory

Agency	Research Triangle Park, NC 27711

Research and Development	EPA/600/SR-97/038	June 1997

vvEPA Project Summary

A Modeling and
Design Study Using
HFC-236ea as an Alternative
Refrigerant in a Centrifugal
Compressor

Predrag Popovic and Howard N. Shapiro

A centrifugal compressor—part of a
chlorofluorocarbon (CFC)-114 chiller
installation—was investigated operat-
ing with the new refrigerant hydrofluo-
rocarbon (HFC)-236ea, a proposed al-
ternative to CFC-114. A large set of
HFC-236ea operating data, as well as a
limited number of CFC-114 data, were
available for this study. It was deter-
mined that the compressor performance
can be successfully described by a rela-
tively simple analytical compressor
model. Two compressor models, the
first of which was obtained trom the
literature, were developed on the basis
of thermodynamic analysis and by uti-
lizing the database. Two empirical rela-
tions were required to predict mass
flow rate and the refrigerant state at
the compressor exit for each model.
The second model is based on empiri-
cal relations derived directly from the
data base rather than the general em-
pirical relations used in the first model.
The literature model had to be opti-
mized for two parameters and corrected
for the influence of the inlet guide vanes
to yield results comparable to the newly
developed model. Both models were
based on the HFC-236ea data, and they
exhibited systematic errors when used
with CFC-114, which indicated the mod-
els' dependence on the refrigerant. Both
models predicted refrigerant state at
the compressor outlet excellently
(±2.8°C), while the mass flow rate was
predicted with larger differences to the
data (±20%). In addition to being quan-
titatively superior, the new model has
more physical relevance; therefore, it

was used for the design analysis. In
the design analysis, the compressor
geometric parameters were varied for
constant operating conditions seeking
trends in compressor performance. It
appeared that the compressor geomet-
ric parameters were appropriately cho-
sen in the compressor design. Also,
the model indicated a valid physical
behavior since all of the trends in the
compressor performance were readily
explainable. The project was sponsored
by the Strategic Environmental Re-
search and Development Program
(SERDP).

This Project Summary was developed
by EPA's National Risk Management
Research Laboratory's Air Pollution
Prevention and Control Division, Re-
search Triangle Park, NC, to announce
key findings of the research project
that is fully documented in a separate
report of the same title (see Project
Report ordering information at back).

Introduction

U.S. Navy surface ships and subma-
rines are equipped with air-conditioning
installations that have centrifugal chillers
operating with CFC-114 refrigerant. The
Environmental Protective Agency (EPA)
in cooperation with the Navy has been
seeking a CFC-114 drop-in replacement.
Therefore, they sponsored a number of
research projects related to the CFC-114
replacement in the mentioned chillers. One
alternative refrigerant that satisfies many
physical and chemical characteristics for
use in the Navy fleet was found to be
HFC-236ea refrigerant. Since HFC-236ea

-------
has very similar thermodynamic charac-
teristics to CFC-114, it was presumed that
the performance of a Navy chiller operat-
ing with this refrigerant would not change
significantly. Hence, this project represents
a part of the investigation directed to evalu-
ate this CFC-114 alternative refrigerant as
a possible drop-in replacement in a Navy
chiller.

The U.S. Navy has an experimental
chiller test facility in Annapolis at the Na-
val Surface Warfare Research Center. This
chiller installation, which is typical for the
Navy fleet, is well instrumented and can
generate experimental data correspond-
ing to different chiller operating conditions.
The data have been gathered over sev-
eral years with several different refriger-
ants. From this large database, the entire
data set of HFC-236ea and several CFC-
114 operating points were accessible to
be investigated in this project. The data
were already taken before this project
started; therefore, the authors neither have
any influence in setting up the experimen-
tal installation, nor any insight in.the qual-
ity of the data recorded.

The objective of this study was to con-
duct a thorough literature review regard-
ing centrifugal compressors and then, on
the basis of the information gathered, build
an accurate but simple compressor model
using the available compressor experimen-
tal data. Further, the developed compres-
sor model would be used to suggest even-
tual design adjustments to enhance com-
pressor performance with the alternative
HFC-236ea refrigerant.

Chiller (Compressor)
Performance

The centrifugal compressor investigated
in this study is an open, single stage com-
pressor with a vaneless diffuser. The com-
pressor performance was controlled by:

•	Using the variable gear box to vary
the compressor shaft speed. The com-
pressor Mach number for the gener-
ated data was between 1.5 and 1.8.

•	Using the inlet guide vanes to fine
control the refrigerant flow rate (ca-
pacity). Also, the guide vanes can
extend the stable compressor operat-
ing range, which is the useful range
of compressor operation between the
surging and choking limits.

•	Bringing a certain amount of the re-
frigerant leaving the compressor to
the compressor inlet through a by-
pass loop where it is mixed with re-
frigerant leaving the evaporator. This
crudely controls the refrigerant flow
rate through the installation (the chiller
capacity).

Studying the chiller operating points, it
was concluded that the available data are
a broad base for compressor model de-
velopment. As an example, the chiller ca-
pacity was varied between 20% and 140%
of the design capacity, while the inlet guide
vane settings were altered between 10
and 90 degrees.

Nevertheless, the data had several limi-
tations, of which the most important con-
cerned the refrigerant flow rate. The flow
rate was not directly measured on the
installation, but it was determined from
the energy balances on the heat exchang-
ers. Also, the total power consumption
and the amount the by-pass valve was
opened were not recorded from the entire
data set. The CFC-114 data were scant,
which prevented a complete comparison
between the two refrigerants.

Compressor Energy Model

The losses occurring in the energy trans-
fer from the compressor shaft to the re-
frigerant are approximately constant.
Quantitatively, these losses are around
1% of the total power delivered to the
compressor shaft. Since the energy losses
are constant, the compressor shaft power
is modeled as a linear function of the
energy transferred to the refrigerant. Thus,
the compressor shaft power can be very
successfully estimated by knowing the
compressor refrigerant mass flow rate and
the enthalpy difference across the com-
pressor.

The CFC-114 data indicated lower en-
ergy losses within the compressor than
the HFC-236ea operating data. Conse-
quently, the HFC-236ea compressor per-
formance line, relating the compressor
shaft power with the amount of energy
transferred to the refrigerant in the impel-
ler, is different from the CFC-114 perfor-
mance line. This was the first indication
that compressor modeling is not invariant
to the operating refrigerant.

Compressor Performance Map

A standard form performance map for
the investigated compressor was provided
by the compressor manufacturer, although
the map was constructed for the com-
pressor operating with a refrigerant differ-
ent than the refrigerants examined in this
study. A feasible relation between the per-
formance map and the actual compressor
operating points was impossible to find.
The experimental data, when plotted in
the form of the compressor performance
map, vaguely resembled a compressor
map. Therefore, it was inferred that the
modeling of the compressor performance
should be sought in some other way.

Compressor Modeling

Based on a thorough literature review,
it was decided to describe the compres-
sor performance thermodynamically by re-
garding the compressor as a control vol-
ume. This modeling approach considers
overall compressor performance charac-
teristics rather than exact flow field infor-
mation within the compressor. Further, a
centrifugal compressor must be consid-
ered to be two entities, the impeller and
the diffuser. Since these compressor parts
are treated as separate control volumes,
the refrigerant state between the impeller
and the diffuser must be determined.

Basically, the model determines the re-
frigerant flow rate and the refrigerant state
at the compressor exit when the following
are entered: the refrigerant state at the
compressor inlet, the inlet guide vane set-
ting, the refrigerant pressure at the com-
pressor exit, and the compressor shaft
speed.

With the model output known, the com-
pressor shaft power can be determined.
The compressor models are based on
empirical relations. For each model output
magnitude, one empirical relation is re-
quired; thus in this model format two em-
pirical relations were required for each
investigated compressor model.

Existing Model

A centrifugal compressor model was
extracted from a chiller simulation that was
found in the literature. The model com-
plied with the desired model input/output
format, and was built on two generic em-
pirical relations based on an extensive
database for similar centrifugal compres-
sors. However, these empirical relations
were based on compressors without inlet
guide vanes.

When initially solved, the model greatly
overestimated the refrigerant flow rate;
therefore, the model was subjected to an
optimization procedure. The flow rate er-
rors were minimized using available data
for two model parameters. When the opti-
mized parameters were incorporated in
the model, the model output yielded rea-
sonable output values. One optimized pa-
rameter, the blockage factor, was found
to be physically unreasonable. In addition
to the doubtful physical validity of the
model, another shortcoming of the exist-
ing model is the requirement to input the
compressor polytropic exponent.

Further, the compressor model flow rate
prediction was corrected for the influence
of the inlet guide vane angles using the
data available. The flow rate was pre-
dicted within ±20% relative difference be-
tween the measured and the modeled flow

2

-------
•4

rates, while the exit state temperature was
estimated within ±2.8°C of the measured
temperatures for most of the HFC-236ea
data points. The CFC-114 data points in-
dicated the presence of a systematic er-
ror, implying that the compressor model is
dependent on the particular refrigerant.

The New Model

The main difference between the new
model and the existing model is that the
available experimental data were used to
generate empirical relations rather than
using the general empirical relations. In
addition, the inlet guide vane settings were
introduced in the compressor model, as
well as, the slip factor.

It was determined that the soundest
empirical relations were those matching
the following model parameters:

•	impeller isentropic efficiency as func-
tion of inlet guide vane position, and

•	dimensionless enthalpy as a function
of flow coefficient.

HFC-236ea data were used to generate
these two empirical relations. The first
empirical relation had a larger variance
than the second relation, and also the
data variations generated in deriving the
empirical relation were proportional to the
difference in model output and the mea-
sured data.

Since the by-pass compressor opera-
tion mode flow rate was successfully cor-
related to the measured flow rate for the
known percentage of the by-pass valve
opening, the new model is feasible for this
operating mode.

A systematic error occurred in the model
estimate for the CFC-114 data points;
hence, the model is a function of the re-
frigerant type.

The blockage factor was optimized to
improve the physical validity of the new
model, and the optimized value has more
physical relevance than the value deter-
mined for the existing model.

Quantitative Comparison
Between Two Models

From the comparison results presented
in Table 1 it can be inferred that the new
model estimates the compressor perfor-
mance parameters better than the exist-
ing model. In addition to the increased
physical validity of the new model, this
improvement in the parameter estimation
definitely classifies the new model as the
better.

The new model is dependent on the
refrigerant type. The model represented
here is developed for HFC-236ea data;
hence, a different set of empirical rela-
tions should be developed for a different

refrigerant operating in the same com-
pressor. Not enough CFC-114 data points
were available in this study to develop a
CFC-114 model.

The new HFC-236ea model can esti-
mate the flow rate within ±20% relative
difference between measured and mod-
eled flow rates, exit state temperature
within ±1 .7°C, and the compressor shaft
power within ±14% relative difference.

In addition to being a quantitatively su-
perior model, the new model also has
more physical validity than the existing
model for the following reasons:

•	The blockage factor in the new model
was derived to be 0.9, while the ex-
isting model blockage factor was op-
timized to the value of 0.24. The ex-
isting model blockage factor must
have accounted for some other com-
pressor performance effects, dimin-
ishing the model's physical relevance.

•	The existing model required the poly-
tropic exponent and the reference
polytropic efficiency to be estimated
as input.

•	The existing model inlet guide vane
setting is introduced in the model
through the flow rate correction for-
mula, while in the new model the inlet
guide vane angle was a fundamental
factor upon which the model was de-
veloped.

Table 1. Quantitative Comparison Between Two Compressor Models

HFC-236ea Data Used to

Comparison Between Measured	Build Empirical Relations	HFC-236ea Additional Data	CFC-114 Data

and Modeled Compressor Parameters Average *	Max.f	Average	Max.	Average Max.

The Existing Model

Flow rate w/o IGV cor. [%] *
Flow rate IGV corrected [%] *
Exit state temperature [°F] ®
Shaft power	[%]*

13.96
6.33
1.22
15.67

36.81
21.85
3.43
21.04

22.27
7.66
3.49
20.27

49.71
15.57
9.07
39.40

30.67
19.22
13.97
33.04

76.25
31.01
47.31
52.99

The New Model

Flow rate prediction [%]*
Exit state temperature [°F]$
Shaft power	[%]*

4.45
0.89
3.13

19.51
2.21
13.38

12.57
2.13
9.26

19.84
6.11
17.22

22.84
6.21
20.54

55.08
17.81
42.61

" The mean value lor a set of absolute differences between measured and modeled particular compressor parameters tor the data set in question,
t The maximum absolute difference between measured and modeled particular compressor parameter for the data set in question.

t The companson results are presented in terms of the absolute relative difference between measured and modeled particular compressor parameter, X, given as percentage;

Ditterence [%]=|(XmMl-	100. (IGV=inlet guide vane.)

§ The companson between measured and modeled compressor exit state temperatures is given in terms of absolute difference between measured and modeled temperatures
in Fahrenheit degrees for the particular data set; Difference [°F]= |tara>„- lamod|* (1°C = 5°F/9.)

3

-------
Refined Model

A more detailed centrifugal compressor
model was built based on the new model.
This refined model needed two empirical
relations, and the best relations appeared
to correlate the same pairs of parameters
as in the new model. The refined model
contains the variable slip factor and a
more detailed diffuser model, which are
the improvements to the new model.

The set of equations of the refined model
never yielded a reasonable solution. Er-
rors generated in the model output solu-
tion were found to be functions of the
sequence in which the equations were
solved. However, the source of such spu-
rious behavior of the system of equations
was never identified. The authors still be-
lieve that the system might be solvable
with good equation-solver software.

Design Analysis

The compressor design parameters in-
cluded in the new model were investi-
gated for the constant compressor operat-
ing input. These design parameters were
the number of impeller blades, the blade
inlet and exit tip angles, the impeller di-
ameters, and the impeller exit axial width.
The compressor performance was char-
acterized with the compressor work coef-
ficient which is proportional to the amount
of energy transferred in the impeller and
the magnitude of the pressure at the im-
peller exit. The other vital parameter in
the compressor operation was the refrig-
erant flow rate.

The effects of the design parameters
on the compressor model results include:

•	The number of impeller blades is di-
rectly related to the slip factor. Impel-
ler performance as a function of the
slip factor has a positive gradient,
while the flow rate as a function of
the slip factor is negatively sloped.
These trends were observed in the
design analysis, and it appears that
the number of blades was chosen
appropriately. Increasing the number
of blades by about 10% might im-
prove compressor performance, but
would result in a pressure drop of
approximately 15%.

•	The blade tip angle is inversely re-
lated to the slip factor, but still directly
related to the compressor perfor-
mance. The trends observed in the
blade tip angle design analysis were
very similar to the trends encountered
in the number of impeller blades de-
sign analysis. The blade tip angle ap-
pears to be well chosen. Enlarging tip

angle by about 5% might improve the
work coefficient by roughly 10% and
reduce the flow rate by about 20%,
with an increase of 6.9-13.8 kPa in
impeller exit pressure.

•	Increasing the inlet impeller diameter
improves compressor performance
and reduces the mass flow rate. The
flow rate variations are more signifi-
cant than the variation in the work
coefficient, which is the primary quan-
titative impeller performance measure.
The pressure at the impeller exit and
the compressor exit enthalpy are not
affected by this parameter. The influ-
ence of inlet diameter on compressor
performance rapidly diminishes as the
flow is throttled with guide vanes. For
these reasons, the inlet diameter
should be excluded from the eventual
impeller design modification.

•	The impeller exit diameter greatly af-
fects the magnitude of the flow rate,
indicating a large gradient as the flow
rate is varied with the size of the
impeller. The compressor perfor-
mance (work coefficient) is a nega-
tively sloped function of the exit im-
peller diameter. The exit diameter
changes should be very limited, since
a 10% increase in the impeller exit
diameter roughly doubles the mass
flow rate.

•	The impeller discharge area is directly
related to the impeller axial width, so
it affects the flow rate considerably.
Impeller performance as a function of
the impeller axial width has a small
negative gradient, while no significant
change is observed in the impeller
exit state pressure. The impeller exit
axial width may be altered to change
flow rate if necessary for a relatively
small change in compressor perfor-
mance.

Although the compressor model indi-
cated reasonable physical behavior, the
design analysis should be taken very cau-
tiously. The intent of the design analysis
was to indicate trends in the compressor
performance with the design input varia-
tions rather then to suggest specific
changes in the compressor design. Since
the compressor model was found to be
sensitive to the refrigerant type, it is rea-
sonable to assume that the model is sen-
sitive to the compressor design modifica-
tions. Also, the assumption of constant
operating input with design input varia-
tions should be regarded with some res-
ervations because of the indicated sensi-
tivity of the compressor model.

Recommendations

This study points to several recommen-
dations:

1.	The refrigerant flow rate was not
measured on the chiller installation,
but rather estimated from the con-
denser and the evaporator energy
balances. Since'the flow rate was
found to be an extremely sensitive
value in the compressor models,
the flow rate should be verified on
the experimental installation. As the
vital value in any refrigeration com-
pressor modeling procedure, the
flow rate should be measured by at
least several flowmeters on a single
experimental refrigeration installa-
tion. At the very least, the flow rates
through the chiller and the com-
pressor by-pass loop should be
measured.

2.	The total compressor power con-
sumption should always be mea-
sured and reported. The total com-
pressor power consumption is an
important parameter to include in
the modeling in order to be able to
predict compressor energy require-
ments for different compressor op-
erating conditions.

3.	The compressor in the Navy chiller
has a vaneless diff user. However,
vaned diffusers are also widely used
in centrifugal compressors. Their
advantage over vaneless diffusers
is in broadening the stable com-
pressor operating range and reduc-
ing the fluid expansion losses at
the diff user inlet. However, this type
of diffuser has greater friction losses
than the vaneless diffuser. The use
of a vaned diffuser should be con-
sidered.

4.	Both models were dependent on
the refrigerant type. The models
presented here were developed
based on HFC-236ea data, hence
a different set of empirical relations
should be developed for a different
refrigerant operating in the same
compressor. This should give in-
sight into the dependence of the
model on the refrigerant type.

5.	A better compressor performance
map than the one provided by the
compressor manufacturer should be
developed, which may require a
wider range of the operating condi-
tions, especially compressor Mach
numbers. Such a performance map
might be a better tool with which to
model compressor performance.

-------
6.	Further attempts should be made
to solve the refined model. It is
anticipated that the refined model,
if a solution to its set of the equa-
tions were found, would improve
the accuracy of the new model.

7.	Increasing the number of blades by
about 10% might improve compres-
sor performance, but would result
in an increased pressure drop of
about 15%.

8.	Enlarging the tip angle by about
5% might improve the work coeffi-
cient by roughly 10% and reduce
the flow rate by about 20%, with an
increase of 6.9-13.8 kPa in the im-
peller exit pressure.

9.	The inlet diameter should be ex-
cluded from the eventual impeller
design modification.

10.	One has to be very careful modify-
ing the impeller diameter, since it
strongly influences compressor per-
formance.

11.	The compressor exit axial width
may be enlarged to increase flow
rate, if necessary, for a relatively
small deterioration in the compres-
sor performance.

5

-------
Predrag Popovic and Howard N. Shapiro are with Iowa State University, Ames,
IA 50011.

Theodore G. Brna is the EPA Project Officer (see below).

The complete report, entitled "A Modeling and Design Study Using HFC-236ea as
an Alternative Refrigerant in a Centrifugal Compressor," (Order No. PB97-
156129; Cost: $35.00, subject to change) will be available only from:

National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:

Air Pollution Prevention and Control Division
National Risk Management Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711

United States

Environmental Protection Agency

Center for Environmental Research Information

Cincinnati, OH 45268

Official Business
Penalty for Private Use $300

BULK RATE
POSTAGE & FEES PAID
EPA

PERMIT NO. G-35

EPA/600/SR-97/038

-------