International Journal of Refrigeration 28 (2005) 481–488

www.elsevier.com/locate/ijrefrig

Performance investigation of a variable speed vapor compression

system for fault detection and diagnosis
Minsung Kim, Min Soo Kim1*
School of Mechanical and Aerospace Engineering, Seoul National University, Seoul 151-744, South Korea
Received 26 August 2004; received in revised form 26 November 2004; accepted 26 November 2004
Available online 15 January 2005

Abstract
An experimental study has been performed to investigate the effect of four artificial faults on the performance of a variable
speed vapor compression system. Experimental setup to test several artificial faults was made by modifying the conventional
vapor compression test rig. Four major faults of compressor fault, condenser fault, evaporator fault, and refrigerant leakage,
were implemented by observing the variation of cooling capacity. Two different rule-based modules for constant and variable
speed operations were organized for an easy diagnosis of system faults. These two modules were applied differently as the
cooling capacity satisfies the necessary air conditioning load. As a result, COP degradation due to the fault in a variable speed
system is severer than that in a constant speed system.
q 2004 Elsevier Ltd and IIR. All rights reserved.

Keywords: Compression system; Compressor; Variation; Speed; Correlation; Fault; Performance

Etude sur la performance d’un système de détection et de

diagnostic d’anomalies appliqué à un système à compression de
vapeur à vitesse variable
Mots clés : Système à compression ; Compresseur ; Variation ; Vitesse ; Corrélation ; Panne ; Performance

1. Introduction component of the system. For this reason, fault detection

and diagnosis (FDD) systems are installed to prevent the
Since HVAC&R systems are getting complicated and main system from the deadly failure or performance
extended, there is a great need of advanced technology to degradation caused by various faults. FDD system was
observe, monitor, and check up real-time behaviors of each originally developed for the purpose of safety issues. [1] For
example, in nuclear power plants or aircrafts, FDD systems
are equipped for guaranteed operation even with high
* Corresponding author. Tel.: C82 2 880 8362; fax: C82 2 883 installation cost. On the other hand, a number of recent
0179. industrial applications pursue the reduction of management
E-mail address: minskim@snu.ac.kr (M.S. Kim). cost related with equipment downtime, service costs, or
1
Associate member of IIR. utility costs. Since FDD systems for HVAC&R industry are
0140-7007/$35.00 q 2004 Elsevier Ltd and IIR. All rights reserved.
doi:10.1016/j.ijrefrig.2004.11.008
482 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488

Nomenclatures
COP coefficient of performance TCO.D compressor discharge temperature [8C]
3 fault alarm threshold DTSC liquid line subcooling [8C]
m_ r mass flow rate of refrigerant [g/s] DTW temperature difference of water inlet and outlet
N compressor speed [rpm] [8C]
PC condenser outlet pressure [kPa] DTB temperature difference of brine inlet and outlet
PE evaporator inlet pressure [kPa] [8C]
TBO brine outlet temperature [8C] Subscript
TC condensation temperature [8C] ref reference
TE evaporator inlet temperature [8C]

mainly considered to reduce the management cost, it should diagnostic module assuming individual features as a series
be approached from the view of its usefulness and cost of independent probabilistic accidents. Chen and Braun
effectiveness. (2001) [6] developed simple rule-based FDD methods for a
Grimmelius (1995) [2] developed an on-line failure packaged air conditioner with a thermal expansion valve
diagnosis system for a refrigeration system used in a vessel (TXV). They suggested a fault diagnostic scheme by
or a plant. A symptom matrix based on the combination of comparing the sensitivity of residuals based on the fault
casual analysis was used for the FDD system. Stylianou characteristic chart. Castro (2001) [7] applied a nearest
and Nikanpour (1996) [3] diagnosed faults of a reciprocat- neighbor method and a nearest prototype method for fault
ing chiller. The authors suggested three fault detection detection of a chiller. The author calculated Euclidean
modules for startup, stop, and steady-state operations based distances of the current state based on the selected two
on a thermodynamic model and expert knowledge of the largest residuals, and estimated the possibility of fault from
chiller. For a successive research, Stylianou (1997) [4] the distance information.
presented the fault diagnostic methodology using a rule- Even though various FDD studies have been carried out
based statistical method. Rossi and Braun (1997) [5] for a vapor compression system, mostly they focused on the
developed a statistical FDD method for a roof-top air systems with fixed speed compressors. Considering several
conditioner. They suggested a fault detection module HVAC systems adopt variable speed compressors, the FDD
applying Bayesian classifier, and developed a fault researches on variable speed systems are very important. In

Fig. 1. Schematic diagram of the experimental setup.

 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488 483

this research, a laboratory-scale experimental apparatus of based on the parameters measured at the reference operating
vapor compression system with a variable speed drive was condition. Brine flow rates of 60, 75, 90 g/s represent low,
set up for several fault tests. The apparatus equipped with medium, high cooling loads, respectively.
two controllers for the variation of a compressor speed and Four artificial faults are tested in this study. Compressor
the opening of an electronic expansion. The performance fault was simulated by hot gas bypass using a fine-tuning
variations and other characteristics are investigated under valve which connects compressor outlet to inlet. The
several artificial faults scenarios. compressor fault is modulated by the opening of the bypass
valve, where the level was based on the reduction ratio of

2. Experimental setup and test conditions

Fig. 1 illustrates the schematic diagram of a variable

speed system for FDD experiments in this study. The test rig
consisted of a compressor, a condenser, an evaporator,
and an electronic expansion valve. The compressor was an
open-type vehicle reciprocating compressor with a design of
two inline cylinders. Displacement of the compressor is
3.92 m3/h at 1450 rpm. The condenser and the evaporator
were two concentric heat exchangers of the same size and
length. R22 was used as a working fluid, and water was used
as a secondary fluid for the condenser. Brine of 40%
ethylene-glycol aqueous solution was used for the evaporator.
The compressor was driven by motor-inverter assembly, and
its speed was controlled to maintain the brine outlet
temperature of the evaporator within 16.7G0.1 8C. Elec-
tronic expansion valve opening was controlled to give the
uniform evaporator exit superheat of 5G0.3 8C. Pro-
portional-integral-derivative (PID) control based on the
second Ziegler–Nichols tuning method was applied.
Table 1 shows the experimental conditions for the
reference test and the fault test. Compressor speed is limited
below 1600 rpm, which is properly set as a maximum speed
as in the case of a real machine. At a reference charge of
R22, the system is adjusted to satisfy 5 8C of liquid line
subcooling and 16.7 8C of brine outlet temperature of the
evaporator. The severity of faults was labeled in percentage

Table 1
Experimental conditions for the reference test and the fault test

Parameters Units Reference test Fault test

Brine inlet 8C 26.7 26.7
temperature
Brine outlet 8C 16.7 16.7
temperature
Brine flow g/s 75 60/75/90
rate
Water inlet 8C 35.0 30.0/35.0/40.
temperature 0
Water flow g/s 100 100
rate
Suction line 8C 5.0 5.0
superheat
Compressor rpm 1185 w1600
speed Fig. 2. Performance variation of the compressor fault with respect to
Charge of kg 0.84 – cooling load. (a) Compressor speed. (b) Brine temperature at
R22 evaporator exit. (c) Refrigerant flow rate. (d) Coefficient of
performance.
484 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488

refrigerant flow rate obtained from the reference test. 1600 rpm, brine outlet temperature did not reach the given
Refrigerant leakage is set by reducing the total amount of set value. Trends of compressor speed and brine outlet
refrigerant charge. Fault level of refrigerant leak was the temperature were very similar in all other fault tests.
ratio of charged mass reduction from the reference charge. Distinct characteristics of compressor fault are observed
Fault concept in the condenser and evaporator is in Fig. 2(c). The refrigerant flow rate remains the same
basically the same. Malfunction of a fan or deterioration regardless of the fault level except at the ‘out of capacity’
of surface feature may cause the degradation of heat transfer situation. Similarly, most of other parameters are found to
performance. Heat exchanger fault in this study was keep their own values. Under compressor fault, the
designed to reduce the capacity of the heat exchanger by compressor is accelerated to compensate the lost capacity.
changing the heat transfer area. In order to reduce heat Since the fault is caused only by hot gas bypass, almost all
transfer area, flow passage of the secondary fluid was other parameters are not affected if the refrigerant flow rate
changed by bypass ports installed at the each end of heat in the main loop maintains the original value. Fig. 2(d)
exchanger subsection. Fault designs for both condenser and shows the degradation of coefficient of performance (COP)
evaporator are alike. Reduction ratio of heat exchanger area with respect to the increase of fault level and cooling load.
is set to fault level. Since the higher compressor speed results in the lower
efficiency, COP at the constant speed of 1600 rpm,
representing ‘out of capacity’ conditions, degrades less
3. Test results of fault performance than the COP at variable speed conditions.

3.1. Compressor fault 3.2. Refrigerant leakage

Fig. 2 compares the performance variation under three Fig. 3 shows trends of liquid line subcooling and COP
different cooling loads. Data points specified as ‘out of change with respect to load requirements. The vertically
capacity’ imply that cooling capacity is insufficient for the dotted line represents optimal charge, and the left side of the
required cooling load. Fig. 2(a) shows that the compressor is optimal charge line shows overcharged operations. Over-
accelerated as the increase of fault level and cooling load. charge caused by technical inexperience should be regarded
From Fig. 2(b), brine outlet temperature is well controlled as a fault. But sometimes overcharge is hardly considered as
near 16.7 8C, but becomes uncontrollable at the high cooling a fault for some large systems with a liquid receiver installed
load when the fault level is over 4%. That is, even though in the liquid line. Such systems might be overcharged
compressor speed approached at the maximum speed of intentionally to retard performance degradation caused by
unpredicted leakage of refrigerant. In this case, overcharge

Fig. 3. Performance variation of the refrigerant charge with respect Fig. 4. Performance variation of the condenser fault with respect to
to cooling load. (a) Liquid line subcooling. (b) Coefficient of cooling load. (a) Condensation pressure. (b) Coefficient of
performance. performance.
 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488 485

itself may not be regarded as a fault. By this reason, this cooling load. Since low cooling load operation requires
study did not regard an overcharge as a fault and focused on reduced heat transfer, the reduction of heat exchanger area
leakage only. has a little impact on COP for the low load condition.
From Fig. 3(a), liquid line subcooling goes down
gradually as the leakage portion increases. The liquid line
3.4. Evaporator fault
subcooling approaches zero in the high leakage condition,
which means the refrigerant at the condenser outlet turns
Fig. 5(a) shows the trends of evaporator pressure which
into a saturated state. When the system load increases,
temperature difference at the heat exchangers should decreases at a higher fault level. Little change was detected
increase. Thus the subcooling is getting greater due to the
increase of pressure difference between condenser and
evaporator. From Fig. 3(b), COP is reduced as the fault level
increases. COP change does not appear distinctively under
overcharge or small leakage conditions. However, if
refrigerant is leaked out of the system beyond a certain
extent of approximately 15%, COP degrades very much.

3.3. Condenser fault

Fig. 4(a) shows the increase of condenser pressure along

with the fault level. When heat transfer area is reduced, the
capacity of a heat exchanger is decreased. As a result, the
total capacity of condensation reduces at a constant speed
condition. But at a variable speed condition, the system tries
to catch up the deficient capacity through compressor speed
change then the entire heat capacity is kept as the same
level. Therefore, the temperature difference between
refrigerant and water is raised and the condenser pressure
increases. Fig. 4(b) shows COP degradation for several fault
levels. COP degrades less for low cooling load than for high

Fig. 5. Performance variation of the evaporator fault with respect to Fig. 6. Comparison of COP changing rates with respect to fault type.
cooling load. (a) Evaporation pressure. (b) Coefficient of (a) Compressor fault. (b) Refrigerant leakage. (c) Condenser fault.
performance. (d) Evaporator fault.
486 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488

in absolute pressure of the evaporator inlet with respect to operated at its off-design point and the performance
the cooling load. It is because temperatures at the brine inlet becomes worse. Additionally, when the compressor oper-
and outlet are controlled uniformly and the only refrigerant ates at a high speed, the ratio of capacity increase cannot
flow rate is closely relevant to the cooling capacity. Fig. 5(b) follow the increase of compressor work. Thus the faults are
shows the degradation of COP as the increase of the more influential to a variable speed system. From Fig. 6(a)–
evaporator fault level. The overall trend is quite similar with (d), COP degradation is commonly redeemed when the
the condenser fault, but the rate of COP degradation is quite compressor operates at the maximum speed. However, COP
less. reduces more when the compressor operates in variable
Fig. 6 shows COP change for the applicable faults in this speed conditions. If the compressor speed is not restricted, it
study. There are remarkable difference in performance with will speed up over the maximum to retrieve from the ‘out of
respect to the load requirements, and fault levels. The capacity’ condition. In this case, the compressor shows
compressor fault in Fig. 6(a) shows COP reduces similarly lower performance than at the maximum speed because of
regardless of cooling load conditions. Since the compressor lower efficiency at higher speed. Moreover, the increased
fault mostly results in a loss of refrigerant flow rate, the capacity will pull down the mean brine temperature. Thus,
system parameters are hardly affected by the fault level as the points of ‘out of capacity’ in Fig. 7 will be moved down.
long as the flow rate reduction due to faults is compensated. It means COP degradation by fault is severe in a variable
However, refrigerant leakage or heat exchanger fault result speed system. Therefore, the development of FDD method
in the change of operating point, therefore, their influences is much more important for a variable speed system than for
a constant speed system.
are strong as the changes of cooling load condition. From
the test results of the refrigerant leakage or the heat
exchanger fault under a low load condition, COP degrades
weakly. 4. Establishment of FDD system
Normally, COP decreases as the fault level or cooling
load increases. When the fault level increases, the system is Fig. 7 shows a prototype of FDD system for a variable

Fig. 7. Prototype of a real-time FDD algorithm for a variable and constant speed system.
 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488 487

speed refrigeration unit. First, the current state of compres- two sorts of classifiers were used to check whether the
sor speed or brine outlet temperature should be monitored to cooling capacity satisfies the cooling load requirement.
see if the compressor speed is constant or variable. If the When the cooling capacity satisfies the cooling load (i.e.
compressor speed is constant at its maximum, the compressor speed is below maximum or brine outlet
subsequent diagnosis follows a series of FDD system for a temperature is at the set temperature), fault diagnosis is
constant speed system. On the contrary, when the compres- applied to the classifier of a variable speed system. If the
sor speed is below the maximum speed, the subsequent capacity does not meet the cooling load (i.e. the compressor
procedure follows the variable speed system. is operated in its maximum speed), fault diagnostic should
To determine whether the system is faulty or not, an ideal follow the classifier for a constant speed system. Therefore,
way is to evaluate the possibility of fault from the variation in this experiment, two sort of rule-based diagnostic
of COP. However, it is not cost-effective to measure COP classifiers are established separately for a variable speed
directly using expensive measuring devices. Instead of system and a constant speed system. These rules were
measuring COP, other parameter like a compressor speed developed and tested through experiments over a range of
can be used as a fault detection parameter for a variable operating conditions in this study.
speed system. The compressor speed is easily estimated by Table 2(a) and (b) shows the rule-based fault classifi-
monitoring the frequency input modulated by the inverter. cation charts for a constant speed system and a variable
By the way, when the compressor operates at its maximum speed system, respectively. Refrigerant leakage fault is
speed, compressor speed is inappropriate for a fault divided into two groups by the leak amount of 15% of the
detection parameter since no speed change was carried initial charge. When the leak is over 15%, the system is
out. In this case, a brine outlet temperature is good as an always operated in the maximum speed and the variable
alternative parameter. Then the selected fault detection speed classifier could not be applicable. From the two tables,
parameter is compared with its reference value. In case each parameter in a constant speed system shows clear
that the difference of the two values is larger than the increase or decrease compared in a variable speed system.
threshold 3, the fault alarm is activated to notify the Even though there is a fault in the system, compressor and
occurrence of fault. expansion valve are well controlled to maintain the required
When the system is expected to be faulty, diagnostic performance in a variable speed system. This is the reason
module performs fault classification using the residuals why the rise and fall of the parameter values are not
between expected and measured parameters. In this study, distinctively conceived for a variable speed system.

Table 2
Changes in parameters of a vapor compression system

Fault type TE TC TSC TCO.D DTW DTB N

Constant speed operation
Compressor [ Y Y [ Y Y –
fault
Refrigerant
leakage
!15% – – Y – Y Y –
O15% Y Y –a – Y Y –
Condenser Y [ – [ Y Y –
fault
Evaporator Y Y [ – Y Y –
fault
Variable speed operation
Compressor – – – [ – – [
fault
Refrigerant
leakage
!15% – – Y [ – – [
O15% Not applicableb
Condenser – [ Y [ – – [
fault
Evaporator Y – [ [ – – [
fault
a
The exit of liquid line reached saturated liquid state.
b
Compressor speed remained 1600 rpm (the maximum speed properly set in this study) when the refrigerant leaked over 15%.
488 M. Kim, M.S. Kim / International Journal of Refrigeration 28 (2005) 481–488

5. Conclusions performed. Additionally, a variety of researches should be

accompanied to operate the system appropriately in spite of
Performance of a variable speed vapor compression fault occurrence.
system was experimentally investigated for specified faults.
The test rig was devised by modifying a conventional water-
to-refrigerant system with an open-type reciprocating
compressor. The system parameters are less sensitive to References
the faults at a variable speed condition because the
compressor speed is controlled to compensate the reduced
[1] J.E. Braun, Automated fault detections and diagnostics for
capacity, which suppresses the changeability of the system. vapor compression cooling equipment, Int J HVAC&R Res 5
The compressor fault test did not provide clear difference in (2) (1999) 85–86.
the system parameters for a variable speed system. The [2] H.T. Grimmelius, J.K. Woud, G. Been, On-line failure
variation of refrigerant charge or heat exchanger has a great diagnosis for compression refrigeration plant, Int J Refrig 18
influence on the overall operation. However, the perform- (1) (1995) 31–41.
ance variations due to those faults are largely influenced by [3] M. Stylianou, D. Nikanpir, Performance monitoring, fault
initial charge and heat exchanger sizing. detection, and diagnosis of reciprocating chillers, ASHRAE
Algorithm of the fault detection and diagnosis for a Trans 102 (1) (1996) 615–627.
variable speed vapor compression system was applied [4] M. Stylianou, Application of classification functions to chiller
separately as the system satisfies the required load. Rule- fault detection and diagnosis, ASHRAE Trans 103 (1) (1997)
based fault classifiers were established for a constant speed 645–656.
[5] T.M. Rossi, J.E. Braun, A statistical, rule-based fault detection
system and a variable speed system. The FDD system
and diagnostic method for vapor compression air conditioners,
developed in this study provided a prototype FDD sequence
Int J HVAC&R Res 3 (1) (1997) 19–37.
of a various speed system. From this experimental study, the [6] B. Chen, J.E. Braun, Simple rule-based methods for fault
system with an open-type reciprocating compressor was detection and diagnostics applied to packaged air conditioners,
found to show different trends of performance with respect ASHRAE Trans 107 (1) (2001) 847–857.
to the operating condition. COP degradation by fault is [7] N.S. Castro, G. Remington, Performance evaluation of a
much severer in a variable speed condition than in a constant reciprocating chiller using experimental data and model
speed condition. Therefore, researches on FDD systems for predictions for fault detection and diagnosis, ASHRAE Trans
a variable speed system should be more carefully 108 (1) (2002) 889–903.