Accepted Manuscript

Title: Thermodynamic model for reciprocating compressors with the focus on

fluid dependent efficiencies

Author: Dennis Roskosch, Valerius Venzik, Burak Atakan

PII: S0140-7007(17)30330-4
DOI: http://dx.doi.org/doi: 10.1016/j.ijrefrig.2017.08.011
Reference: JIJR 3729

To appear in: International Journal of Refrigeration

Received date: 30-3-2017

Revised date: 21-8-2017
Accepted date: 23-8-2017

Please cite this article as: Dennis Roskosch, Valerius Venzik, Burak Atakan, Thermodynamic
model for reciprocating compressors with the focus on fluid dependent efficiencies, International
Journal of Refrigeration (2017), http://dx.doi.org/doi: 10.1016/j.ijrefrig.2017.08.011.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will
undergo copyediting, typesetting, and review of the resulting proof before it is published in its
final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.
 Thermodynamic model for reciprocating compressors with the focus on fluid
dependent efficiencies

Dennis Roskosch, Valerius Venzik and Burak Atakan

Thermodynamics, IVG, Faculty of Engineering, University of Duisburg-Essen, 47057 Duisburg, Germany
Dennis.Roskosch@Uni-DuE.De

Highlights
 Reciprocating compressor model for fluid replacement in existing compressors
 Efficiency predictions based on only two fitting parameters
 Considering in-cylinder energy and mass balances, heat transfer and valve flow
 Semi-physical correlations for the valve flows
 Model fitting and validation based on numerous fluids
 Good agreement with experiments, for different fluids and compressors

Abstract
Fluids with high global warming potential, which are used in existing refrigeration cycles and heat pumps will
have to be replaced soon by less harmful fluids, but the fluid selection is difficult especially due to the
unknown compressor performance. In this work a differential compressor model for reciprocating
compressors is introduced which predicts volumetric and isentropic efficiencies quickly and can be easily
fitted with measured data at only one operation point of an existing compressor. In order to characterize the
influence of different fluids two semi-physical correlations for the valve flows are fitted here, and a procedure
of transferring them to different compressors is shown. The model is validated on, in total, 63 measured points
based on numerous fluids from one semi-hermetic reciprocating compressor which is part of a heat pump
cycle. The calculations lead to mean prediction errors of 3.0 % for the isentropic and 2.3 % for the volumetric
efficiency.
Keywords
reciprocating compressors, thermodynamic model, semi-physical model, heat pumps, refrigerators.
Nomenclature
A area, m²
a crank length, m
COP coefficient of performance

1
Page 1 of 29
ccl relative clearance volume
c specific heat capacity, J·kg-1·K-1
D piston diameter, m
f frequency, Hz
mass flux, kg·m-2·s-1
H stroke, m
h specific enthalpy, J·kg-1·K-1
L rod length, m
m mass, kg
mass flow rate, kg·s-1
p pressure, kPa
Q heat, J
heat flow rate, W
R2 coefficient of determination
Ri specific gas constant, J·kg-1·K-1
rpm compressor rotation speed, min-1
T temperature, K
t time, s
u specific internal energy, J·kg-1
V volume, m³
W work, J
w specific work, J· kg-1·K-1
x axial piston position, m
Greek letters
α convection coefficient, W·m-2·K-1
ζ flow resistance coefficient
η efficiency
ϴ crank angle, rad
ν velocity, m·s-1
ρ density, kg·m-3
ω angular speed, rad·s-1
Acronyms
BDC bottom dead center

2
Page 2 of 29
CFD computational fluid dynamics
DME dimethyl ether
HCFC hydrochlorofluorocarbons
HFC hydrofluorocarbons
TDC top dead center
Subscripts and superscripts
cl clearance
cyl cylinder
d discharge
eff effective
env environment
fr friction
i various component / state
is isentropic
nat natural
p piston
re refrigerant
rev reversible
s suction
vol volumetric
w wall
wg wall-gas
0 initial

1 Introduction
Today most thermodynamic cycles like heat pumps or refrigerators work with fluids from the group of the
HFCs (hydrofluorocarbons), such as R134a or R410A. Contrary to HCFCs (hydrochlorofluorocarbons) HFCs
do not have an ozone depletion potential, but some of them, however, possess an extremely high global
warming potential, and thus, have to be replaced in future. In this context, it is also necessary to find
individual replacement working fluids for existing systems. Fluid selection is an actual topic and many
advanced methods are developed with respect to systematize the selection in heat pumps and also in the
related Organic Rankine Cycles (ORC) (Lampe et al., 2014; Roskosch and Atakan, 2015; Schilling et al.,
2017). In many of these cycles, especially for heat pumps and refrigerators, the exergetic losses of the
compressor are quite large and therefore the compressor strongly influences the coefficient of performance
3
Page 3 of 29
(COP, useful heat flow rate divided by compressor power consumption) of the cycle (Venzik et al., 2017).
Many smaller thermodynamic cycles often work with reciprocating compressors and the efficiency of the
entire process strongly depends on the compressor efficiency. Thus, finding suitable and high efficient fluids
would strongly be facilitated, if the volumetric and isentropic efficiencies of the already used compressors as a
function of the fluid and the operating conditions could be predicted. Existing plants commonly operate in
steady state and thus, data at one operation point with a single fluid should generally be available. With this
data point and some compressor specific parameters, it is aimed to obtain a simple model which is accurate
enough to be used in cycle modelling and would help engineers to predict reliable COPs for different fluids in
a cycle and thus, leading to reliable fluid recommendations for a specific process.
There are several publications about models for reciprocating compressors. These models can generally be
divided into global models and differential models. The global models (Cavallini et al., 1996; Navarro et al.,
2007; Navarro-Peris et al., 2013; Pérez-Segarra et al., 2005; Stouffs et al., 2001) describe the thermodynamic
behaviour, based on a set of empirical equations, which partly do not have a well-grounded physical
background. Global models require less computational time, and work well for a specific fluid-parameter
range, but cannot be extrapolated. As an example, Navarro-Peris et al. (2013) developed an empirical model
which bases on the ARI standard 540 (1999). They introduce two new performance parameters called non-
dimensional mass flow rate and non-dimensional power consumption. The non-dimensional mass flow rate
differs from the well-known volumetric efficiency; it relates the actual mass flow rate to the ideal mass flow
rate, not at suction conditions, but for saturated vapour at the same pressure. The non-dimensional power
consumption is defined as the ratio between the compressor power and the product of pressure and volumetric
flow rate at suction conditions. The authors derived two exponential functions with five fitting parameters
overall to correlate the new performance parameters. These correlations were tested for six different
reciprocating compressors; however, it was not focussed on the prediction of the behaviour of different fluids.
Five of the compressors were working with propane (R290) and one was running with R407C. They found
accurate results and stated that their correlations are suitable to describe all types of piston compressors with
an error lower than 5 %. However, the model of Navarro-Peris et al., as well as further global models, share
the need of several data points from different operating conditions to fit the various input parameters. These
are generally not available, normally never for the compressor running with the replacing fluid, and thus, they
are not suitable for fluid replacement applications.
Differential models generally solve the time (or crank angle) dependent energy balances for the in-cylinder
gas and partly also for the compressor shell. The complexity of the published models differ. The most
extensive ones (Dutra and Deschamps, 2015; Farzaneh-Gord et al., 2015), include the modelling of the
geometry (including the clearance volume), physical equations for the heat transfer between gas and
compressor shell as well as between shell and environment. Also, friction of the piston, leakage flows, the
4
Page 4 of 29
mechanical valve behaviour and the charge changing flows are modelled. Such extensive models lead to
several additional parameters, which are generally not known to the user: for example, the spring rate of the
valves and the diameters of the valves have to be either known or fitted to describe the compressor behaviour
accurately. Castaing-Lasvignottes and Gibout (2010) developed a differential model only based on the
clearance volume and a friction parameter as fitting parameters. They compared the modelled isentropic,
volumetric and a so-called effective efficiencies with experimental results of different cycles using R134a as
working fluid. They found that the model reproduces the experiments quite closely, but the authors did not
state whether the model could also be applied accurately to other fluids. Farzaneh-Gord et al. (2015)
considered also the valve motion in their compressor model; furthermore, the heat transfer between gas and
compressor shell as well as the flow through the valves (as incompressible flow) were included in their model.
The validation of their model is mainly qualitative; they do not compare with measurements but investigate
the influence of real gas effects in detail. Most authors assume constant flow resistances for the valve flows
(Castaing-Lasvignottes and Gibout, 2010; Dutra and Deschamps, 2015; Navarro et al., 2007). The valve flows
have a strong impact on the isentropic and volumetric efficiencies and the flow resistance strongly depends on
the fluid and the flow rate through the valve. Therefore, setting it constant is expected to lead to inaccurate
results, especially if the fluid is varied. Commonly, differential models only take the cylinder itself into
account and neglect the mechanical parts (crankshaft, bearings) or the electric motor; thus, the calculated
power does not fit to the actual electrical power input. In order to improve this, Dutra and Deschamps (2015)
combined a differential compressor model with a thermal model and an electrical model. They compared their
results to measurements and to results from a standard differential model and found that implementing an
additional electrical model improves the accuracy of the results compared to the standard model. All of these
studies consider only one or two different fluids; the possibility to predict the compressor efficiencies for
different fluids, based on a model parameter fitting for just a single fluid has not been considered so far.
In order to find replacement fluids for existing plants, also from large fluid databases, a compressor model is
needed that predicts efficiencies for different fluids from different substance groups with an adequate
accuracy. The computational effort should be moderate and the number of fitting parameters should be small,
in order to reduce the number of needed measurements. In the present work, a differential model for a piston
compressor is derived, which solves the energy balance for gas and compressor shell and the mass balances.
The heat transfer between gas and shell, shell and environment is included, as well as a sub-model for the
valve flows. The model predicts volumetric and isentropic efficiencies and depends in total on four fitting
parameters and four geometrical values, which are normally known from the manufacturer. The focus is on
the compressor behaviour for different fluids; in this respect, we found, as will be explained, that the valve
flows lead to the major differences between different fluids and thus, a correlation for the flow resistances was
fitted based on measurements for different fluids. These correlations can be adapted to further compressors
5
Page 5 of 29
and reduces the number of fitting parameters to two. This enables adapting the model to a compressor based
on only one measured point (isentropic and volumetric efficiency). The model is validated with our heat pump
test rig (Venzik et al., 2017) containing a semi-hermetic reciprocating compressor, utilizing 63 experimental
operation points at a single rotation speed with several working fluids. The working fluids are the
hydrocarbons propene (R1270), isobutane (R600a), propane (R290) and mixtures containing
isobutane/propane, isobutane/propene and isobutane/pentane, the hydrofluorocarbons 1,1,1,2-
tetrafluoroethane (R134a) and 1,1-difluoroethane (R152a) as well as the natural fluid dimethyl ether DME.
The fluids were selected because some of them are used today and probably the others may be used in future.
Subsequently, the model is analysed with respect to different rotation speeds. In order to evaluate the
applicability of the model to further compressors, the model is finally tested for a second compressor driven
with R134a.

2 Modelling
The compressor model is based on a standard differential model for reciprocating compressors as used
previously by many authors (Belman-Flores et al., 2015; Dutra and Deschamps, 2015; Stouffs et al., 2001).
However, at some points crucial modifications were made, which on the one hand simplify the model, and on
the other hand include fluid-dependent physics, leading to suitable results for different fluids and different
operation conditions.

2.1 Physical Model

The compressor is modelled by a simple piston cylinder system with two valves as shown in Fig. 1.
Furthermore, a thermal mass representing the cylinder wall (compressor shell) is implemented. The rotation of
the crankshaft (crank length a) is converted to the reciprocating motion of the piston (diameter D) by the rod
(length L). The compressor is assumed to be free of leakage. The cylinder has a piston displacement volume
of V0 (stroke H, bottom dead center BDC to top dead center TDC) and the clearance volume Vcl is defined
relatively to V0:
Eq. 1

Here, ccl is a function of the compressor geometry and is one of the fitting parameters. The actual piston
position x depends on the crank angle θ and is given by:

6
Page 6 of 29
 Eq. 2

This equation is well known and basically used in the field of combustion engine modelling and can be
derived from the compressor geometry (Fig. 1). Regarding the cylinder wall, Adair et al. (1972) found that its
temperature varies less than ± 0.56 K during one cycle (dθ = 2π). As a result, the cylinder wall temperature
was set constant also for different fluids and different operation conditions by many authors (Belman-Flores et
al., 2015; Stouffs et al., 2001). However, the cylinder wall temperature is an important value in compressor
simulations; it depends on the heat transfers between the cylinder wall and the enclosed gas and also from or
to the surrounding environment. The heat transfer between gas and cylinder smooths the in-cylinder gas
temperature regime during the cycle and depends strongly on the inlet conditions and the pressure ratio
(Tuhovcak et al., 2016). Rather, the heat flow rate to the environment is commonly minor, due to small
convection coefficients, small surface areas and small temperature differences between the surface and the
environment (Brok et al., 1980). Nevertheless, regarding other operation conditions the heat losses become
more important and thus, they were also implemented here. The heat flow rate w
between cylinder wall and

gas is calculated using Newton’s law of coolin :

Eq. 3

The accessible heat transfer area A depends on the piston position x and the areas of the piston and the head,
assuming a regular cylinder. Re ardin the calculation of the convective heat transfer coefficient α (deviatin
from the usual nomenclature α was chosen to prevent conflicts with the symbol for enthalpy) Tuhovcak et al.
(2016) compared results from CFD simulations with well-known empirical correlations. They found that it is
best to take different correlations for the different geometrical parts (head, cylinder wall, piston) as well as for
the different working steps into account. However, the differences between the correlations investigated by
Tuhovcak et al. (2016) seem to be minor and due to the generally small influence of the heat transfer on the
efficiencies (Brok et al., 1980), the Woschni (1970) correlation is applied to calculate the convective heat
transfer coefficient α for all parts and working steps here. Originally, this equation was derived for internal
combustion engines, but it was found that, by omitting the combustion source term, the common peak of the
convection coefficient close to a crank angle of 180° (burning of fuel) is shifted to the end of the compression
step as it is expected for reciprocating compressors. Thus, modified in this way, it is also suitable to predict
convection coefficients within a compressor (Tuhovcak et al., 2016).
Eq. 4

In Eq. 4, p is the mean piston velocity and the value of c1 is 6.18 s·m-1 for suction and discharge and 2.28
s·m-1 for compression and expansion. Contrary to the units indicated in the nomenclature, the pressure has to

7
Page 7 of 29
be inserted in MPa in Eq.4. The heat flow rate between the thermal mass (wall) and the environment (at
Tenv = 298.15 K) is given by:

Eq. 5

The coefficient for the natural convection outside the cylinder is for the common parameter range not strongly
sensitive to the wall temperature and thus, it is set constant a typical value of αnat = 6 W·m-²·K-1. The outside
surface area can simply be estimated based on the compressor geometry. Finally, the temperature derivative of
the thermal mass, mainly consisting of the cylinder walls, piston and valves, is derived from the energy
balance.

Eq. 6

As already explained, the wall temperature is nearly constant for steady state operation of the compressor and
thus can be assumed to be time independent. In this case, the mean cylinder wall temperature does not depend
on the heat capacity mw·c. Thus, those parameters are chosen such that the iterations of the wall temperature
converge fast but without numerical oscillations. The transferred work to the fluid is split into two parts one
reversible piston work and some irreversible work. For all process variables, the European sign convention is
used, meaning that the value is positive when energy is transferred to the system. Assuming quasi-equilibrium
changes of state, the reversible piston work is given by:

Eq. 7

while p represents the time or crank angle dependent pressure of the fluid. Furthermore, some work to
overcome sliding friction is implemented, to take irreversibilities between the piston and the cylinder wall into
account. This part of the work has a mechanical background and in engine modelling (Merker et al., 2006),
the irreversibilities due to friction are indicated as dissipated work and calculated by means of a friction
pressure pfr; in the following we will follow this convention. The friction pressure pfr is one of the fitting
parameters and combines different irreversibilities, but the major part is the mechanical friction between the
piston rings and the wall.

Eq. 8

Generally, sliding friction mainly depends on the rotational speed, the geometry and the lubricant, and thus,
also on the oil temperature and pressure. The present work addresses a fluid retrofit for existing cycles without
changing their application. Therefore it is assumed that both, temperatures and pressures, will not vary grossly
for the different fluids. Also, presently existing heat pumps or refrigeration cycles are typically controlled by a
start/stop mechanism and not by changing the rotational speed. As a result, it is assumed that, regarding a
constant rotation speed, pfr will not be changed due to changing pressure ratios, inlet conditions and the

8
Page 8 of 29
refrigerant. If rotation speeds are different to the fitting rotation speed or a wider range in operation condition
shall be considered then the friction pressure has to be fitted again.
The time discretization is done by means of the crankshaft angle and the rotary frequency:

Eq. 9

In-cylinder gas temperatures and in-cylinder masses are easily calculated from the conservation of energy and
mass; the individual balances depend on the working step (suction, compression, discharge and expansion);
here changes in kinetic or potential energy are generally neglected.

Compression
The mass balance leads to the simple term for a closed system:

Eq. 10

Further, the energy balance is derived from the first law of thermodynamics:

Eq. 11

Discharge
The discharging process starts when the in-cylinder pressure exceeds the discharge pressure. Regarding the in-
cylinder gas volume and neglecting backflow the continuity equation may be written as follows:

Eq. 12

The valve model is deliberately held simple, it permits only two conditions: fully opened or closed. This is
mainly valid for fast opening/closing valves. If the in-cylinder gas pressure exceeds the discharge pressure, the
valve opens fully. Mass, acceleration and velocity of the valve are neglected. The mass flow rate through the
valve is deduced considering non-compressible flow; the following equation is well-known from the field of
fluid mechanics and is originally derived for pressure losses of fluids flowing through orifices and was often
included to the standard approach for compressors (Belman-Flores et al., 2015; Stouffs et al., 2001).

Eq. 13

ρ and p are in-cylinder gas density and pressure and Aeff,d represents an effective flow area, which is a
combination of the geometrical flow area (Fig. 1) and a flow resistance coefficient ζ (dimensionless).

Eq. 14

Various publications on the field of compressor models did not focus on the influence of different fluids
(Castaing-Lasvignottes and Gibout, 2010; Dutra and Deschamps, 2015; Navarro et al., 2007); thus, the flow

9
Page 9 of 29
resistance was often assumed to be constant. With respect to our investigation, it turns out that the flow
resistance is one of the important values needed to predict isentropic and volumetric efficiencies as a function
of the working fluid. As a result, general correlations for the effective flow areas for discharge as well as for
suction are empirically fitted to selected characteristic fluid parameters (see section 4.2). Those correlations
are also useable for further fluids and can be adapted to further compressors.
Finally, the first law of thermodynamics leads to:

Eq. 15

Expansion
The procedure and equations of the expansion are similar to the compression step. The continuity equation
leads to:

Eq. 16

The energy balance is given by (work due to friction is transferred to the system):

Eq. 17

Suction
The suction model follows the same concepts as the discharging model and leads to the following equations.
Continuity equation:

Eq. 18

Suction mass flow rate:

Eq. 19

Conservation of energy (work due to friction is transferred to the system):

Eq. 20

Finally, the model calculates the isentropic and volumetric efficiencies. The isentropic efficiency is calculated
from the mean discharged enthalpy (integration of the enthalpy flow rate along the discharge time duration),
the inlet state and the isentropic work:

Eq. 21

and the volumetric efficiency is given by:

10
Page 10 of 29
 Eq. 22

2.2 Programming and fitting

All programs are written in the programming language Python (Python.org) and the needed state variables like
internal energies, enthalpies, entropies or volumes (all specific) are basically taken from the NIST REFPROP
database (Lemmon et al., 2013). The REFPROP calls have the major impact on the computational time; in
order to enable larger investigations, regarding different fluids or compressor geometries, short computational
time is required. To address this issue, the required fluid properties are fitted within the relevant superheated
gas regime to polynomials for every program call with a new fluid. This largely reduces the calculation time,
especially if various operation points with the same fluid are calculated, without influencing the accuracy
markedly. After fitting the fluid properties, the program flow starts with the piston at BDC and the cylinder
volume is filled with gas at suction conditions; subsequently, all further iterations start with the final
conditions of the previous iteration. The criterion for reaching a steady state is based on the least squares
method regarding differences of the in-cylinder gas temperature T and mass m as well as of the temperature of
the thermal mass Tw between start and end point of one iteration. When the sum of squares of the change
between the actual values and the values of the previous iteration is lower than a prescribed value (here: 0.01)
it is assumed that steady state conditions are reached and the iteration process is stopped. Finally, the
isentropic and volumetric efficiencies are calculated.
The fitting procedure for estimating the different parameters is based on numerical optimization using the
NLP-algorithm (non-linear-programming) from the OpenOpt-network. The parameters to be fitted are
optimized with respect to the sum of error squares; the error is the difference between measured and
calculated values of the isentropic and volumetric efficiency for a given number of measured working points.

3 Experimental investigation
The investigated compressor in this study is part of a heat pump cycle, details were published in Venzik et al.
(2017). The experimental setup with the main components of the heat pump is schematically shown in Fig. 2;
it includes the compressor, an expansion valve, a condenser, an evaporator and further auxiliary valves for
controlling the mass flow rate of the heat source and sink. The semi-hermetic reciprocating compressor
(model: HG12P5.4, manufacturer: GEA Bock) is especially developed for hydrocarbons. The compressor has
two cylinders with a bore of 34 mm and a stroke of 34 mm, a maximal power consumption of 2.2 kW and 2
pole pairs (electric motor). The motor is controlled by a frequency inverter enabling different rotation speeds
between 1050 and 2100 min-1 (electrical frequency between 35 and 70 Hz). At the inlet and outlet of the
compressor, both, temperature and pressure sensors are installed and the power consumption of the

11
Page 11 of 29
compressor is monitored. The refrigerant mass flow rate is quantified accurately by a Coriolis flowmeter
(KROHNE MFS3081 K 1.5 E). Both heat exchangers, evaporator and condenser are designed as counter flow
double-pipe heat exchangers. A fine metering needle valve is used as expansion valve, and coupled with a
servo-motor which allows controlling the evaporation pressure via PC. The pressure levels or pressure ratios
are adjusted by varying the flow rate of the secondary fluid in the condenser and by controlling the orifice of
the expansion valve.
Regarding temperature and pressure measurements, resistance thermometers (PT 1000) of type AA and
calibrated relative pressure transducer are used. The software tool LabVIEW (LabVIEW Professional
Development System, 2013) is used for monitoring, recording and controlling the experiments. The
corresponding values are recorded every second within a measuring cycle. After the process reaches steady
state conditions, the data is recorded for about 20 minutes; the data analysis is based on averaged values.
Based on the given tolerances from the manufacturers of the used thermometers, pressure sensors and Coriolis
flowmeter, the statistical error was estimated based on a typical working point (propene, T e = 290.47 K,
pe = 500 kPa, pa = 2044 kPa) by using error propagation with respect to the isentropic and volumetric
efficiencies. The maximum relative errors are 1.37 % for the isentropic efficiency and 1.96 % for the
volumetric efficiency. The influence of the working condition on the statistical error is marginal, and thus, it
can be expected that the exemplarily calculated values are also similar to the ones of the other fluids and
operation conditions, which are investigated here.

For a systematic analysis of the physical behavior of the compressor, different combinations of inlet pressures
and pressure ratios are investigated for the pure fluids isobutane, propane, propene, R134a, R152a and
dimethyl ether. Furthermore, respectively three different blends of isobutane/propene and isobutane/propane
as well as one mixture of isobutane and pentane are investigated. The investigated inlet pressures and pressure
ratios strongly depend on the considered fluid but generally inlet pressures are between 150 and 650 kPa and
pressure ratios between 3.0 and 7.0 are investigated. The compressor speed was initially held constant at 1500
min-1 (electrical grid frequency: 50 Hz); the compressor inlet temperature is always kept constant at 290.15 K.
Finally, the isentropic and volumetric efficiencies are calculated based on the measured inlet and outlet state
and on the mass flow of the fluid.
Isentropic efficiency (assuming a reversible adiabatic compressor for comparison) is calculated from the
measured temperatures and pressures utilizing the REFPROP (Lemmon et al., 2013) database to obtain
enthalpies, according to:

12
Page 12 of 29
 Eq. 23

and the volumetric efficiency is calculated as:

Eq. 24

4 Results and discussion

With respect to the physical model, the needed input parameters are listed in Table 1. Most of them are
geometrical parameters and are specified for the used compressor; however, with exception of the piston
diameter D and the stroke length H, they are usually unknown to the user. The surface area is simply
estimated by the compressor geometry to be Aenv = 0.04 m². With respect to the geometrical ratio of rod/crank
it turns out that the model is not strongly sensitive to this value, and thus, the standard value of 3.5, as often
used for combustion engines (Heywood, 1988), is used. The remaining parameters have to be fitted for the
specific compressor, the rotation speed and the given working fluid. For the relative clearance volume and the
friction pressure, it is assumed that for a given rotation speed these parameters only depend on the specific
compressor, while the effective flow areas depend on both, the compressor and the fluid.

4.1 Results for propene

Initially, the four parameters ccl, pfr, Aeff,s and Aeff,d are fitted only for one fluid, while it is initially assumed
that the effective flow areas are regardless of the operation point and thus only depend on the compressor and
the fluid. Propene was selected for this procedure. The fitting is based on two measured working points with
the same inlet conditions but different outlet pressures (at Ts = 290.44 K, ps = 600 kPa, pd,1 = 1870 kPa,
pd,2 = 2360 kPa and rotational speed = 1500 min-1); the results are listed in Table 2. As an example, a relative
clearance volume of 6.07 % is calculated and the friction pressure is 49 kPa; the values are all within
physically meaningful parameter ranges. Fig. 3 shows the isentropic and volumetric efficiencies for propene
and an inlet pressure of 600 kPa as a function of the pressure ratio. Here, lines represent model results and
points indicate the experimental results; the open symbols indicate the measured points which were the basis
for the parameter fitting. The error bars representing the statistical errors were also included, showing that the
measurements are highly accurate and it can be expected that the influence of the experimental error on model
validation is marginal. Regarding the isentropic efficiency, it is well known from the literature (Hundy et al.,
2008) that ηis typically increases with higher pressure ratios, goes through a flat optimum and decreases again.
The optimum pressure ratio depends on the compressor geometry, the fluid and the inlet conditions. The
model predicts such a behavior, the line of the isentropic efficiency starts for a pressure ratio of 2.0 at 57 %,

13
Page 13 of 29
then increases slightly for higher ratios and finally ends with a value of 71 % and a gradient near to zero for a
pressure ratio of 5.0. The expected optimum is also predicted by the model, but not sketched in Fig. 3, due to
the small experimentally accessible pressure ratio range of the heat pump cycle.

Comparing the model results with the measured points it gets clear that the model is able to calculate the
values as well as the trend reliably. The same applies to the volumetric efficiency; the model shows analog to
literature (Hundy et al., 2008) a linear decrease of ηvol from 81 % to 65 % with increasing pressure ratio and
fits well to the measurements. In order to simulate or calculate thermodynamic cycles like heat pumps or
refrigeration cycles, it is, apart from the pressure ratio dependence, also important to predict volumetric and
isentropic efficiencies for different inlet pressures. The performance of the compressor generally varies with
the inlet density. Fig. 4 shows the efficiencies as a function of the inlet pressure for a pressure ratio of 3.5 and
propene as working fluid while the inlet temperature is constant at 290.15 K. In general, the trends of the
isentropic and volumetric efficiencies are reproduced well and the maximum mean errors are smaller than
1.1 % and 1.5 %, respectively. Therefore, the model reproduces isentropic and volumetric efficiencies also for
different inlet conditions at a constant rotation speed well. Another interesting aspect is the importance of the
work due to friction and the integral heat flow transferred to the wall w ) on the calculated isentropic and
volumetric efficiencies. Regarding one operation point with propene (Ts = 290.15 K; ps = 500 kPa;
pd = 1900 kPa) as an example, it turns out that the influence of the net transferred heat is minor related to the
transferred work Wfr. During one working cycle Qwg,net = -0.216 J is rejected from the gas and transferred to
the wall and finally, to the environment while the work due to friction amounts to Wfr = 3.0 J. Due to the
considered operation points, with moderate process temperatures close to the environment temperature, the
results for the isentropic and volumetric efficiencies are only marginally changed (<1 %) if the heat transfer is
neglected; this was already found by other authors (Brok et al., 1980). However, if operating conditions with
higher temperature differences to the environment are considered, the transferred heat becomes more
important. On the contrary, neglecting the work due to friction leads to an increased isentropic efficiency by
18 % while the volumetric efficiency is only slightly increased.
As already mentioned, the fluid properties are initially fitted by polynomials in order to reduce the
computational effort; this takes approximately one second (CPU: I7-4790, memory: 24 GB). The calculation
of one operation point takes then only 5 seconds and also enables extended studies in an appropriate time.

14
Page 14 of 29
4.2 Applying the model to different fluids

So far, it was shown that target values can accurately be calculated also for different operation conditions, as
long as the fluid is not changed from the one used for fitting (propene) the model parameters. The next step is
to apply the model to different fluids. An approach, to use the same effective flow areas also for other fluids
leads to inaccurate results and thus, fluid dependent expressions for the effective flow areas have to be found.
Assuming incompressible laminar flow, several correlations predicting the flow resistance as a function of the
Reynolds number and geometrical parameters can be found in numerous fluid mechanics textbooks (e.g.
Spurk and Aksel, 2008). However, it turned out that these correlations are not adequate to describe the fluid
dependence here. The valve flows are probably neither laminar nor incompressible but influenced by
deflection, separation and heat transfer. Especially for semi-hermetic compressors, the fluid is, prior to
entering the suction plenum, piped alongside the motor winding and thus superheated further. Suction and
discharge plenum are commonly located side by side and a heat transfer from the hot discharge gas to the
lower tempered suction gas occurs. Flowing into the cylinder, the cold gas is initially heated by the hot
cylinder wall, what commonly leads to a convective flow in the opposite direction. These findings led to a
search for empirical or semi-physical correlations for the effective flow areas as a function of some fluid
parameters and the operation conditions; the correlations should also be applicable to other piston
compressors. In order to address this issue, the effective flow areas were in a first step fitted numerically (c.f.
section 2.2) to 25 experimentally obtained operating points at different working conditions from five fluids
from different substance groups (propene, isobutane, propane, R134a, dimethyl ether). Here, inlet and outlet
pressures were varied while the inlet temperature (Ts = 290.15 K) and the rotation speed (rpm = 1500 min-1)
were held constant. Effective areas between 0.5·10-5 m² and 3.0·10-5 m² are found, and it turns out that the
suction effective flow area is only a weak function of the operation point (Fig. 5), but depends strongly on the
fluid, while, the discharge effective flow area depends on both (Fig. 6). The second step was finding
parameters, which are fluid and operation condition dependent, and correlate well to the optimized effective
flow areas. In the best case, a physical interpretation of the related parameters would be obtained. Several
different parameters were tested, which commonly are used to characterise flows. As an example, it turns out
that the Reynolds number, as one of the major flow describing parameter, is not adequate to correlate any of
the effective flow areas; the same is true for the Prandtl or the Mach number. Since Aeff,s strongly depends on
the fluid, but is nearly independent of the working conditions, as can be observed in Fig. 5 the points for each
fluid are quite coincident, it was assumed to be a function of fluid characteristic parameters, only. It is found
that Aeff,s can be described as a function of the specific gas constant Ri as depicted in Fig. 5. Due to its
connection to the specific heat capacities and the molecular weight, the ideal gas constant is a reasonable
parameter in the field of gas dynamics, and thus, it appears to be an appropriate parameter to characterise
15
Page 15 of 29
flows, especially if heat transfer occurs. Since the fluid state for the flows through the valves are strongly
different in density and velocity for the entering and the exiting fluid, it turns out that the discharge effective
flow areas do not correlate to the specific ideal gas constants, and this approach used for the discharge leads to
inaccurate results. Instead, a good matching with the mass flux through the valve (defined as ratio of mass
flow rate through the valve and effective flow area) was found; the results are shown in Fig. 6. It is obvious
that Aeff,d decreases with increasing mass fluxes; this indicates that the discharging flow is probably dominated
by flow losses induced by separation. As a result from the previous analysis, the following empirical
correlations of the effective flow areas depending on the specific gas constant (Aeff,s) and the mass flux (Aeff,d)
were fitted, while the units of the input as well as of the output values refer to the units indicated in the
nomenclature.
Eq. 25

Eq. 26

Since the mass flux is a function of the effective flow area, the correlation for the discharging valve is not
solvable analytically, and thus, the effective flow area of the discharging valve has to be iterated within the
program flow. The correlations are also depicted in Fig. 5 and Fig. 6. It is seen that these fits work well for the
investigated fluids; the coefficients of determination are R2=0.94 for suction and R2=0.92 for discharging.

Using the derived correlations, the model was tested for 63 measured points under different operation
conditions for pure fluids (40 records) and for mixtures (23 records). In order to evaluate the transferability to
other fluids, besides the fluids, used for fitting the effective areas, an additional fluid, 1,1-difluoroethane
(R152a) was also tested. The regarded mixtures are blends containing isobutane/propene and
isobutane/propane with different compositions and one isobutane/pentane mixture (mole fractions: 85/15);
pentane was also not considered in the fitting process. The results are shown in Fig. 7 and Fig. 8; Fig. 7
depicts the isentropic efficiency and Fig. 8 the volumetric efficiency while the respective measured values are
plotted on the x-axis and the calculated ones on the y-axis. Besides, a deviation area of ± 5 % is indicated. It is
obvious that the calculated results generally fit well to the measured values, the absolute mean errors are
3.0 % for the isentropic efficiency and 2.3 % for the volumetric efficiency while the maximum errors are 6 %
for both. Comparing the mean errors of mixtures and pure fluids no significant difference is observed. A
special focus is on the results for those fluids, which were not basis for the fitting (R152a open circles,
isobutane/pentane: open triangles). It is obvious from Fig. 7 and Fig. 8 that some of the larger deviations are
found for these fluids, but in general, the isentropic and volumetric efficiencies of these fluids are reproduced
with an adequate accuracy by the model. Therefore it is concluded that the correlations for the effective flow
areas also work well for fluids, not used in the fitting procedure; however, it should be kept in mind that the

16
Page 16 of 29
fits are basically empirical and validity outside the considered parameter ranges is probably not given; at least
it should be tested, prior to usage.

4.3 Applying the model to different rotation speeds

As already mentioned, the friction pressure depends on the rotation speed and has to be fitted again if a
different rotation speed shall be considered. Based on the previous results (effective flow areas and ccl) the
friction pressure was fitted again for the same compressor but now to a measured operation point with
propene (Ts = 290.4 K; ps = 552 kPa; pd = 1623 kPa) at a lower rotation speed of 1050 min-1 (electrical
frequency f = 35 Hz) resulting in a friction pressure of pfr = 62.3 kPa. Contrary to the expected physical
behaviour, the fitted friction pressure for a rotation speed of 1050 min-1 is larger as that for the higher rotation
speed (rpm = 1500 min-1: pfr = 49 kPa). It was found that the effective flow areas are overestimated by Eq. 25
and Eq. 26 if the rotation speeds are lower (1050 min-1) than the one used for fitting. This in turn also results
in overestimated efficiencies. Due to the mathematical covariance, the friction pressure and the flow area can
compensate each other within the fitting procedure. Here, for the determination of the correlations for the
effective flow areas, the focus was on describing different fluids at the same rotation speed correctly, however
future work should analyse the valve flow in more detail and the correlations (Eq. 25 and Eq. 26) will have to
be further expanded, with respect to the rotation speed dependence.
Based on the new friction pressure isentropic and volumetric efficiencies were calculated for 18 different
measured operation points with different fluids and a rotation speed of rpm = 1050 min-1; Fig. 9 and Fig. 10
show the results.

It gets clear that the calculated isentropic and volumetric efficiencies fit well to the measured values, the
absolute mean errors are 3.6 % for the isentropic efficiency and 2.1 % for the volumetric efficiency while the
maximum errors are smaller than 7 % for both. Thus, it is obvious that the entire model can also be applied
well to different rotation speeds, if the friction pressure is fitted to one measured operation point with the
respective rotation speed.

4.4 Applying the model to different compressors

It was stated that one of the main targets of this work is to develop a compressor model that can be used after
fitting only one measured working point with one fluid to a compressor, which is then able to predict
isentropic and volumetric efficiencies accurately for other fluids and different operating points. The advantage
of the correlations for the effective flow areas is that the number of needed fitting parameters are reduced to
two: the relative clearance volume ccl and the friction pressure pfr. Thus, the model can be fitted to a different
compressor, as soon as the isentropic and volumetric efficiencies at one working point are available. However,
17
Page 17 of 29
the correlations for the effective flow areas have to be adapted if a compressor with different geometry is
considered. From Eq. 14 it is seen that the effective flow area is a combination of the geometrical flow area of
the valve and a dimensionless flow resistance; if the geometrical valve flow area is different to that of the
compressor used here, the correlations are no longer valid in this form. However, it can commonly be
assumed that the valve geometry (flow area) is proportional to the piston surface area while the flow
resistance is constant.

Eq. 27

The model is tested in the following, by applying Eq. 27 to a different compressor. The compressor geometry
and three measured points (ηis and ηvol) for different operation conditions with the fluid R134a are taken from
example 1 of the paper of Castaing-Lasvignottes and Gibout (2010). The relative clearance volume and the
friction pressure were fitted to only one of the three measured points. Due to the unknown compressor
geometry, the surface area could not be estimated directly; however, it is assumed that the surface area is
proportional to the number of cylinders ncyl, the piston diameter D and the stroke length H:

Eq. 28

Based on the similar specifications of the compressor and due to its minor influence, it was assumed that the
rod-to-crank length ratio is not changed. The fitting results, as well as the geometry data of the compressor,
are shown in Table 3.

Fig. 11 shows the calculated values and the measured values by Castaing-Lasvignottes and Gibout (2010) of
the isentropic and volumetric efficiencies. The operation point 1 which was the basis for the model fitting is
accurately reproduced, and also, the relative mean errors for the further points are smaller than 3.4 % for the
isentropic and 4.4 % for the volumetric efficiencies, and are smaller than the deviations of the model of
Castaing-Lasvignottes and Gibout (2010) (isentropic efficiency: 7.4 %, volumetric efficiency: 8.8 %). As a
result it looks, as if the model could successfully be applied to different compressors without decreasing the
accuracy compared to the compressor, on which the modelling is based. Due to the limited data available in
the literature, no further test cases were investigated, but further validations with other piston compressors are
recommended, whenever the data are available.

18
Page 18 of 29
5 Conclusions
The presented differential compressor model calculates volumetric and isentropic efficiencies as a function of
the inlet condition and the outlet pressure and the fluid, aiming to help retrofitting compressors from a fluid
used so far to a new, less harmful fluid. It is based on the energy and mass balances, heat transfer from the in-
cylinder gas to the compressor shell and between shell and environment and on the valve flows. The model
depends basically on four fitting parameters: friction pressure, relative clearance volume and effective flow
areas for the inlet and outlet valves and four other parameters, which are normally known from the
manufacturer. The key observation in the present work is the role played by the valve flows: it is of crucial
importance when the compressor performance for different fluids shall be reproduced. Based on several
measurements with different fluids under various operation conditions two empirical equations, correlating
the effective flow areas were derived. Applying these equations, the transferable parameters are reduced to the
friction pressure and the relative clearance volume and thus, the model can be fitted to an existing compressor
based on only one measured point with one fluid. The results for different fluids under various operation
conditions showed a good matching with the measured values at a constant rotation speed; the relative mean
errors for the isentropic and the volumetric efficiencies are smaller than 3.0 % for the compressor that was
basis for the fitting procedure and lower than 4.4 % for a different compressor from literature. If the model
shall be applied to different rotation speeds, the friction pressure has to be fitted again based on one measured
point. It was found that the correlations for the effective flow areas are inaccurate for rotation speeds, differing
from that used for the fitting; however, this can be compensated by the fitted friction pressure, which in this
case should only be regarded as a fitting parameter. But in order to apply such a model to different rotation
speeds without refitting the friction pressure, the proposed correlations for the effective flow areas have to be
further developed especially with respect to different rotation speeds. In summary, the model, as proposed
here, seems to be a good basis for fluid replacement studies, although further validations with different piston
compressors and further fluids would be desirable.
.
References
Adair, R. P., Qvale, E. B., Pearson, J. T., 1972. Instantaneous Heat Transfer to the Cylinder Wall in
Reciprocating Compressors. Proceedings of the International Compressor Conference at Purdue, 521–526.
ARI standard 540: Standard for positive displacement refrigeration compressors and compressor units, 1999.
Belman-Flores, J.M., Ledesma, S., Barroso-Maldonado, J.M., Navarro-Esbrí, J., 2015. A comparison between
the modeling of a reciprocating compressor using artificial neural network and physical model.
International Journal of Refrigeration 59, 144–156. 10.1016/j.ijrefrig.2015.07.017.
Brok, S. W., Touber, S., Van der Meer, J. S., 1980. Modeling of Cylinder Heat Transfer - Large Effort, Little
Effect? Proceedings of the International Compressor Conference at Purdue.
19
Page 19 of 29
Castaing-Lasvignottes, J., Gibout, S., 2010. Dynamic simulation of reciprocating refrigeration compressors
and experimental validation. International Journal of Refrigeration 33 (2), 381–389.
10.1016/j.ijrefrig.2009.10.007.
Cavallini, A., Doretti, L., Longo, G. A., Rossetto, L., Bella, B., Zannerio, A., 1996. Thermal Analysis of a
Hermetic Reciprocating Compressor. Proceedings of the 1996 Purdure International Compressor
Engineering Conference, 535–540.
Dutra, T., Deschamps, C.J., 2015. A simulation approach for hermetic reciprocating compressors including
electrical motor modeling. International Journal of Refrigeration 59, 168–181.
10.1016/j.ijrefrig.2015.07.023.
Farzaneh-Gord, M., Niazmand, A., Deymi-Dashtebayaz, M., Rahbari, H.R., 2015. Thermodynamic analysis
of natural gas reciprocating compressors based on real and ideal gas models. International Journal of
Refrigeration 56, 186–197. 10.1016/j.ijrefrig.2014.11.008.
Heywood, J.B., 1988. Internal combustion engine fundamentals. McGraw-Hill, New York, 930 pp.
Hundy, G.F., Trott, A.R., Welch, T.C., 2008. Refrigeration and air-conditioning, 4th ed. Elsevier Butterworth-
Heinemann, Amsterdam, 381 pp.
LabVIEW Professional Development System, 2013. National Instruments.
Lampe, M., Stavrou, M., Bücker, H.M., Gross, J., Bardow, A., 2014. Simultaneous Optimization of Working
Fluid and Process for Organic Rankine Cycles Using PC-SAFT. Ind. Eng. Chem. Res. 53 (21), 8821–
8830. 10.1021/ie5006542.
Lemmon, E.W., Huber, M.L., McLinden, M.O., 2013. NIST Standard Reference Database 23: Standard
Reference Data Program. National Institute of Standards and Technology, Gaithersburg.
Merker, G.P., Otto, F., Schwarz, C., Stiesch, G., 2006. Simulating Combustion: Simulation of combustion and
pollutant formation for engine-development. Springer-Verlag Berlin Heidelberg, Berlin, Heidelberg.
Navarro, E., Granryd, E., Urchueguía, J.F., Corberán, J.M., 2007. A phenomenological model for analyzing
reciprocating compressors. International Journal of Refrigeration 30 (7), 1254–1265.
10.1016/j.ijrefrig.2007.02.006.
Navarro-Peris, E., Corberán, J.M., Falco, L., Martínez-Galván, I.O., 2013. New non-dimensional performance
parameters for the characterization of refrigeration compressors. International Journal of Refrigeration 36
(7), 1951–1964. 10.1016/j.ijrefrig.2013.07.007.
OpenOpt. http://openopt.blogspot.de/. Accessed 31 July 2017.
Pérez-Segarra, C.D., Rigola, J., Sòria, M., Oliva, A., 2005. Detailed thermodynamic characterization of
hermetic reciprocating compressors. International Journal of Refrigeration 28 (4), 579–593.
10.1016/j.ijrefrig.2004.09.014.
Python.org. https://www.python.org/. Accessed 31 July 2017.
20
Page 20 of 29
Roskosch, D., Atakan, B., 2015. Reverse engineering of fluid selection for thermodynamic cycles with cubic
equations of state, using a compression heat pump as example. Energy 81, 202–212.
10.1016/j.energy.2014.12.025.
Schilling, J., Lampe, M., Gross, J., Bardow, A., 2017. 1-stage CoMT-CAMD: An approach for integrated
design of ORC process and working fluid using PC-SAFT. Chemical Engineering Science 159, 217–230.
10.1016/j.ces.2016.04.048.
Spurk, H., Aksel, N., 2008. Fluid Mechanics, 2nd ed. Springer-Verlag, Berlin, Heidelberg.
Stouffs, P., Tazerout, M., Wauters, P., 2001. Thermodynamic analysis of reciprocating compressors.
International Journal of Thermal Sciences 40 (1), 52–66. 10.1016/S1290-0729(00)01187-X.
Tuhovcak, J., Hejcik, J., Jicha, M., 2016. Heat Transfer Analysis in the Cylinder of Reciprocating
Compressor. Proceedings of the International Compressor Conference at Purdue.
Venzik, V., Roskosch, D., Atakan, B., 2017. Propene/isobutane mixtures in heat pumps: An experimental
investigation. International Journal of Refrigeration 76, 84–96. 10.1016/j.ijrefrig.2017.01.027.
Woschni, G., 1970. Die Berechnung der Wandverluste und der thermischen Belastung der Bauteile von
Dieselmotoren. MTZ (12), 491–499.
.

21
Page 21 of 29
 Ts , ps Td , pd
Aeff,s
Aeff,d
TDC
thermal
mass dQwg (α , A)
T, p
Tw, m, c D

H
x
BDC

Tenv rod L
crank a

Fig. 1: Schematic of the compressor model

22
Page 22 of 29
 p

T
FI
evaporator

expansion valve

M compressor
Motor
liquid tank
reservoir

p CF
condenser

suction line T temperatur

discharge gas line p pressure
discharge liquid line CF Coriolis flow meter
FI frequency inverter

Fig. 2: Schematic diagram of the experimental setup

23
Page 23 of 29
Fig. 3: Isentropic and volumetric efficiencies as a function of the pressure ratio for propene and a suction
pressure of 600 kPa.

Fig. 4: Isentropic and volumetric efficiencies as a function of the suction pressure for propene and a pressure ratio of 3.5.

24
Page 24 of 29
 Fig. 5: Effective flow area for suction as a function of the specific gas constant for different fluids.

Fig. 6: Effective flow area for discharging as a function of the mass flux through the valve for different fluids.

25
Page 25 of 29
Fig. 7: Calculated against measured isentropic efficiencies for various fluids, rpm = 1500 min-1.

Fig. 8: Calculated against measured volumetric efficiencies for various fluids, rpm = 1500 min-1.

26
Page 26 of 29
Fig. 9: Calculated against measured isentropic efficiencies for various fluids, rotational speed: 1050 min-1.

Fig. 10: Calculated against measured volumetric efficiencies for various fluids, rotational speed: 1050 min-1.

27
Page 27 of 29
Fig. 11: Comparison between calculated and measured isentropic and volumetric efficiencies.

28
Page 28 of 29
Table 1: Model input parameters and values
parameter value
piston diameter* D 34 mm
stroke length* H 34 mm
surface area* (estimated) Aenv 0.04 m²
rod/crank L/a 3.5
effective flow area (suction) Aeff,s
effective flow area (discharge) Aeff,d
relative clearance volume ccl
friction pressure pfr
* manufacturer information

Table 2: Fitting results for pure propene at rpm = 1500 min-1

parameter value
effective flow area (suction) Aeff,s 1.27·10-5 m²
effective flow area (discharge) Aeff,d 1.61·10-5 m²
relative clearance volume ccl 6.07 %
friction pressure pfr 49 kPa

Table 3: Model values for the compressor of Castaing-Lasvignottes and Gibout (2010)
parameter value
piston diameter D 39.995 mm
stroke length H 47.625 mm
surface area (estimated) Aenv 0.066 m²
rod/crank L/a 3.5
relative clearance volume ccl 8.1 %
friction pressure pfr 19 kPa

29
Page 29 of 29