See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/303977687

The influences of the operating characteristics of an Electronic Expansion

Valve (EEV) on the operational stability of an EEV controlled direct expansion
air conditioning system

Article  in  International Journal of Refrigeration · June 2016

DOI: 10.1016/j.ijrefrig.2016.06.008

CITATIONS READS

13 895

2 authors:

Yudong Xia Shiming Deng

Hangzhou Dianzi University The Hong Kong Polytechnic University
22 PUBLICATIONS   143 CITATIONS    259 PUBLICATIONS   5,714 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Optimal Thermal Energy storage (TES) technology using phasechange material impregnated into open cell metal foams View project

Refrigerant View project

All content following this page was uploaded by Yudong Xia on 09 March 2020.

The user has requested enhancement of the downloaded file.

 international journal of refrigeration 69 (2016) 394–406

Available online at www.sciencedirect.com

ScienceDirect

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / i j r e f r i g

The influences of the operating characteristics of

an Electronic Expansion Valve (EEV) on the
operational stability of an EEV controlled direct
expansion air conditioning system

Yudong Xia, Shiming Deng *

Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China

A R T I C L E I N F O A B S T R A C T

Article history: This paper reports on a study on the influences of the operating characteristics of a
Received 10 March 2016 proportional–integral (PI) controlled Electronic Expansion Valve (EEV) on the operational sta-
Received in revised form 19 May bility of a direct expansion (DX) air conditioning (A/C) system. Using the classical control
2016 theory, EEV’s PI settings and time constant of EEV’s temperature sensor were analyzed. The
Accepted 5 June 2016 theoretical analysis results using the classical control theory were further verified experi-
Available online 11 June 2016 mentally using an experimental DX A/C system. The study results showed that a larger
proportional or integral gain would lead to a high chance for the EEV–evaporator control
Keywords: loop to become unstable, while slowing down the rate of degree of superheat (DS) signal
Operational stability transfer by increasing EEV’s time constants may help mitigate system’s operational insta-
Sensor dynamics bility. The results confirmed that the operating characteristics of an expansion valve in a
Electronic expansion valve refrigeration system could impact its operational stability.
Classical control theory © 2016 Elsevier Ltd and IIR. All rights reserved.
DX A/C system
PI settings

Les influences des caractéristiques de fonctionnement d’un

détendeur électronique (EEV) sur la stabilité opérationnelle
d’un système de conditionnement d’air à détente directe
contrôlé par un EEV
Mots clés : Stabilité opérationnelle ; Dynamique du capteur ; Détendeur électronique ; Théorie du contrôle classique ; Système de
Conditionnement d’air (A/C) à détente directe (DX) ; Paramètres PI

* Corresponding author. Department of Building Services Engineering, The Hong Kong Polytechnic University, Hong Kong SAR, China.
Tel.: +852 27665859; Fax: +852 27657198.
E-mail address:besmd@polyu.edu.hk (S. Deng).
http://dx.doi.org/10.1016/j.ijrefrig.2016.06.008
0140-7007/© 2016 Elsevier Ltd and IIR. All rights reserved.
 international journal of refrigeration 69 (2016) 394–406 395

refrigeration system could exhibit unstable operation as a

Nomenclature mixture–vapor transition point moving toward evaporator exit,
was proposed by Huelle (1967). Huelle (1972) later introduced
T temperature [°C]
mass flow rate [kg h−1] conceptually a so-called MSS line which is a monotone conic
Mre
Kp proportional gain [―] curve starting from the original point on a DS (X-axis) – cooling
Ki integral gain [―] capacity (Y-axis) chart, for predicting the conditions of hunting,
Ti integral time [s] and considered the MSS line for a refrigeration system as one
R thermal resistance [K kW−1] of the inherent characteristics of an evaporator itself. Chen et al.
Cp specific heat [kJ kg−1 K−1] (2002) confirmed the existence of an MSS of ~ 4 °C at a spe-
cific cooling capacity for a TEV controlled refrigeration system,
V volume [m3]
e relative error [―] and attributed the primary reason of hunting to the flow type
ue EEV control signal [―] and nonlinear variation of heat transfer coefficient in an evapo-
rs reference setting [―] rator. However, the second view tried to explain the cause of
DS degree of refrigerant superheat [oC] hunting based on the influence of the operating characteris-
t time [s] tics of an expansion valve on system stability, as it would take
some time for the DS signal to propagate through the expan-
Greek letters sion valve to adjust the refrigerant mass flow required, which
ρ density [kg m−3] was considered as the fundamental reason for hunting. A
θ time delay [s] number of related modeling studies suggested that either in-
τ time constant [s] creasing or decreasing the time constant of the TEV sensing
ϕ0 offset adjustment parameter for a bulb for measuring the DS at the exit of an evaporator would
particular EEV [―] help reduce the hunting (Broersen and Vanderjagt, 1980;
Ibrahim, 1998; Mithraratne et al., 2000). For TEV controlled
evaporators, decreasing valve gain and increasing static su-
Subscripts
perheat setting would also be beneficial to system operational
a air side
stability (Mithraratne and Wijeysundera, 2001, 2002; Mithraratne
m measured by the temperature sensor
et al., 2000).
e evaporator
Among these reported stability studies, only limited numbers
v valve
of studies on the influences of the operating characteristics of
re refrigerant side
an Electronic Expansion Valve (EEV) on system operational sta-
se sensor
bility have been carried out. In an EEV-controlled refrigeration
system, proportional and integral (PI) and proportional, inte-
gral and derivative (PID) control algorithms are extensively used
for its EEV to regulate the EEV’s opening for controlling refrig-
1. Introduction erant mass flow rate in response to DS at evaporator exit (Jolly
et al., 2000; Qi et al., 2010). In practice, a PI controller is ad-
Instability in a refrigeration system, conventionally known as equately capable to provide an acceptable control performance,
hunting, is the phenomena of the oscillation of certain system without the need to consider the problems associated with the
operational parameters such as the degree of refrigerant su- derivative actions, namely the need of properly filtering out
perheat (DS), refrigerant mass flow rate and evaporating the measurement noise (Visioli, 2006). In addition, in an EEV-
pressure. Hunting has been noticed not only in the refrigera- controlled refrigeration system, a temperature sensor is used
tion systems controlled by thermostatic expansion valves (TEVs) to measure the temperature of refrigerant at evaporator exit,
(Eames et al., 2014; Huang et al., 2014; Wedekind, 1971; so that the actual operating DS may be evaluated and sent to
Wedekind and Stoecker, 1968; Yasuda et al., 1983), but also those the EEV for control action. The temperature sensor acts in a
controlled by electronic expansion valves (EEVs) (Chen et al., similar manner to a TEV’s sensing bulb in a TEV-controlled re-
2008; Fallahsohi et al., 2010; Li et al., 2004; Qi et al., 2010). frigeration system. On the other hand, it was previously shown
Hunting leads to a lower operational safety and a higher energy that in a TEV-controlled system, the TEV’s valve gain and the
consumption (Liang et al., 2010), and therefore, should be time constant of the TEV’s sensing bulb (Mithraratne and
avoided as far as possible for the safe and energy efficient op- Wijeysundera, 2001, 2002; Mithraratne et al., 2000) did impact
eration of a refrigeration system. the system operational stability, because the former directly
There have been two different views on the causes of un- affected the value of opening and thus the refrigerant mass
stable system operation. The first concentrated on the inherent flow rate, and the latter the time for the DS signal to propa-
characteristics of an evaporator. Random fluctuation in a re- gate through the TEV to adjust the refrigerant mass flow rate.
frigerant mixture–vapor transition point was first described by This suggested that the operating characteristics of a PI con-
Zahn (1964). Wedekind and Stoecker (1968, Wedekind, 1971) trolled EEV in terms of PI settings and the time constant of its
further indicated that the random oscillation of a transition temperature sensor may also similarly impact on the opera-
point in an evaporator was due to the nature of two phase flow, tional stability of an EEV-controlled system. However, no
which appeared to be related to the same mechanism causing previous investigations on the influences of EEV’s PI settings
the density wave instabilities. The theory of minimal-stable- and the time constants of EEV’s temperature sensor on the op-
signal (MSS), defined as a critical minimal DS at which a erational stability in an EEV-controlled refrigeration system may
396 international journal of refrigeration 69 (2016) 394–406

be identified. Furthermore, it is observed that the Transfer Func-

Table 1 – The specifications of the DX evaporator and
tion method has been commonly used to describe the
EEV used in the experimental DX A/C system.
relationships between the input and output of a controlled
DX Evaporator
process, including evaporators used in refrigeration systems
(Aprea and Renno, 2001; Broersen and Vanderjagt, 1980; Chen Transverse tube pitch 21 mm
and Jiang, 1990; Stoecker, 1966). Longitude tube pitch 12.7 mm
Fin pitch 1.3 mm
Therefore, a study on the influences of EEV’s PI settings and
Fin thickness 0.115 mm
time constants of EEV’s temperature sensor in the PI-controlled Outside tube diameter 7 mm
EEV–evaporator control loop on the operational stability of a Inside tube diameter 6.4 mm
direct expansion (DX) air conditioning (A/C) system using the Length of the windward area 450 mm
classical control theory has been carried out and the study Height of the windward area 504 mm
results are reported in this paper. Firstly, the descriptions of Number of the windward transverse tube 24
Number of the tube row 6
the experimental DX A/C system and the PI-controlled EEV–
Number of refrigerant loop 8
evaporator control loop are presented. Secondly, the transfer
EEV
functions for each of the components in the PI controlled EEV– Pulse range 0~500
evaporator control loop are presented. Thirdly, using the transfer Nominal capacity 8.5 kW
functions developed, and following the Frequency Response Port diameter 1.8 mm
Method and Nyquist stability criterion in the classical control
theory, the influences of EEV’s PI settings and the time con-
stant of EEV’s temperature sensor on the operational stability shown schematically in Fig. 1, the experimental DX A/C system
of the DX A/C system were analyzed and the analysis results was composed of two parts, i.e., a DX refrigeration plant (re-
are presented. Finally, the analysis results were experimentally frigerant side) and an air distribution sub-system (air side). The
verified using the experimental DX A/C system and the veri- major components in the DX refrigeration plant included a
fication results are reported. variable speed rotor compressor, a PI controlled EEV, a high-
efficiency tube-louver-finned DX evaporator and an air-
cooled tube plate-finned condenser. The specifications of the
2. The experimental DX A/C system and its DX evaporator and EEV used in the experimental DX A/C system
EEV–evaporator control loop are shown in Table 1. The evaporator was placed inside the
supply air duct to work as a DX air cooling coil. The design air
For the purpose of this study, an experimental DX A/C system face velocity for the DX cooling coil was 1.98 m s−1, and the
having a conventional PI controlled EEV was established. As nominal output cooling capacity from the DX refrigeration plant

Fig. 1 – Schematic diagram of the experimental DX A/C system.

 international journal of refrigeration 69 (2016) 394–406 397

in the DS signal transfer in the EEV–evaporator control loop.

Like in a TEV controlled refrigeration system, such a delay may
also affect the operational stability in an EEV controlled re-
frigeration system.

3. Development of a transfer function for the

EEV–evaporator control loop

Fig. 3 shows the block diagram of the PI-controlled EEV–

evaporator control loop in the experimental DX A/C system
shown in Fig. 2. The measured DS at evaporator exit, DSm, is
Fig. 2 – Schematic diagram of the PI controlled
directly used as a feedback DS signal to be compared with the
EEV–evaporator control loop in the experimental DX A/C
reference DS setting, rs. Based on the error between DSm and
system.
rs, or e, the PI controller outputs a corresponding control signal,
ue, to regulate the EEV’s opening. Then, the refrigerant mass
7.5 kW. The refrigerant of the plant was R410a, with a total flow entering the evaporator, Mre, will be regulated continu-
charge of 5.8 kg. ously until e is within its preset range. Consequently, the open-
The air-distribution sub-system included an air-distribution loop transfer function, G(s), for characterizing the transient
ductwork with return air dampers, a variable speed centrifu- response for DSm to a change in EEV’s control signal, ue, can
gal supply fan, and a conditioned space. Inside the space, there be expressed as:
were sensible heat and moisture load generating units (LGUs)
for simulating the space cooling load. ΔDSm (s)
G (s) = = H1 (s) H2 (s) H3 (s) (1)
In the experimental DX A/C system, DS at its evaporator exit Δue (s)
as a representative operating parameter for indicating system
stability in the current study was PI-controlled. Fig. 2 shows The closed-loop transfer function for the PI controlled EEV–
the schematic diagram of the PI controlled EEV–evaporator evaporator control loop, Gc(s), can therefore be expressed as:
control loop which was made of the EEV, evaporator, EEV’s tem-
perature sensor and PI controller. As shown, the DS at C (s) G (s) C (s) H1 (s) H2 (s) H3 (s)
Gc (s) = = (2)
evaporator exit, as a feedback control signal to the EEV for 1 + C (s) G (s) 1 + C (s) H1 (s) H2 (s) H3 (s)
modulating its opening, can be evaluated by measuring the re-
frigerant temperature and evaporating pressure at evaporator In Eqs. (1) and (2), C(s) is the transfer function for EEV’s PI
exit using a temperature sensor and a pressure transducer. Con- controller, H1(s) that for the EEV, H2(s) that for the evaporator
sidering the heat transfer between the temperature sensor and H3(s) that for EEV’s temperature sensor. The details of each
which was usually attached to the outer surface of the refrig- of these transfer functions are separately presented as follows.
erant pipe at evaporator exit, and the vapor refrigerant inside
the pipe, it would take some time for EEV’s temperature sensor 3.1. Transfer function for each of the four components in
to detect the refrigerant temperature which fluctuates all the the EEV–evaporator control loop
time. On the other hand, for a pressure transducer which is
usually in direct contact with refrigerant, the response time 3.1.1. EEV’s PI controller
of the pressure signal transfer may be negligible because pres- For a conventional PI controller, its theoretical output signal,
sure wave travels at speed of sound in tubes. Therefore, the ue, can be written as:
presence of temperature sensor and its installation method in
an EEV-controlled refrigeration system for measuring the re- Kp t

frigerant temperature at evaporator exit would cause a delay

ue (t ) = K pe (t ) +
Ti ∫
0
e (τ ) dτ + φ0 (3)

rs + e ue Mre DS DSm
C(s) H1(s) H2(s) H3(s)
-
PI controller EEV Evaporator Temperature
sensor

C(s) - Transfer function for the PI controller

H1(s) - Transfer function for the EEV
H2(s) - Transfer function for the evaporator
H3(s) - Transfer function for the temperature sensor

Fig. 3 – Block diagram of the PI controlled EEV–evaporator control loop.

398 international journal of refrigeration 69 (2016) 394–406

or in its discretized form:

Temperature sensor
k Tse
ue (t k ) = K pe (t k ) + Ki ∑ e (t i ) Δt + φ0 (4) Ta
i =1
Refrigerant pipe

Kp
Ki = (5)
Ti Tre
Vapor refrigerant
For a digitally implemented PI controller in a refrigeration
system, an incremental algorithm can be written as follows:
Tre Tse Ta
Rre Ra
ue (t k ) = ue (t k −1 ) + K p (e (t k ) + e (t k −1 )) + Ki e (t k ) Δt (6)
Thermal circuit
In Eqs. (3)–(6), ϕ0 is an offset adjustment parameter for a par-
ticular EEV, Kp is the proportional gain, Ki is the integral gain, Fig. 4 – Schematic diagram of the installation of EEV’s
Ti is the integral time, Δt is the sampling time interval temperature sensor attached to the refrigerant pipe at
(Δt = tk−tk-1), and e(tk) is the error between feedback DS signal evaporator exit and its equivalent thermal circuit.
and the DS set point.
Applying the Laplace transform to Eq. (3), the transfer func-
tion for EEV’s PI controller, C(s), can be expressed as: the temperature sensor and the refrigerant at evaporator exit
yielded
Ki
C (s) = K p + (7)
s dTse (t ) ( Tre (t ) − Tse (t )) ( Ta (t ) − Tse (t ))
(ρCpV )se = + (10)
dt Rre Ra
3.1.2. The PI controlled EEV
When the variation range of EEV’s opening is not large, a linear where Tre is the temperature of vapor refrigerant at evapora-
valve characteristic can be assumed. Therefore, the mass flow tor exit, Ta is the ambient temperature, Tse is the temperature
rate passing through an EEV, Mre, can be considered linearly measured by the sensor. Rre is the total thermal resistance
proportional to its control signal, ue. The transfer function for between the vapor refrigerant and temperature sensor, and Ra
the EEV, H1(s), can hence be written as is the convective thermal resistance between the sensor and
its surroundings.
ΔMre (s) Normally, in a refrigeration system, its EEV’s temperature
H1 (s) = = Kv (8) sensor and the refrigerant pipe at evaporator exit are ther-
Δue (s)
mally insulated to reduce heat lose. Therefore, the natural
where Kv is the valve gain which is defined as the ratio of the convection heat transfer between the sensor and its surround-
change in refrigerant mass flow rate to the corresponding ings can be neglected. Thus, Eq. (10) can be simplified to
change in EEV’s control signal.
dTse (t ) 1 1
+ Tse (t ) = Tre (t ) (11)
dt τ se τ se
3.1.3. DX evaporator
According to the previous studies (Aprea and Renno, 2001), the
transient response for the DS at evaporator exit to a sudden τ se = Rre (ρCpV )se (12)
change in refrigerant mass flow rate supplied to the evapora-
tor can be approximated by a second-order plus dead-time where τse is the time constant of the temperature sensor, which
process. Thus the transfer function for the evaporator, H2(s), is affected by the thermal resistance between the sensor and
can be expressed as vapor refrigerant inside the pipe, as well as the heat capacity
of the temperature sensor itself.
ΔDS (s) ω n2 Using the Laplace Transform, Eq. (11) can be transformed to:
H2 (s) = = Ke 2 e −θ s (9)
ΔMre (s) s + 2ζω ns + ω n2
ΔTse (s) 1
= (13)
where Ke is the evaporator gain, indicating the variation of DS ΔTre (s) τ ses + 1
from one steady-state to another against the variation of the
refrigerant mass flow rate supplied. ζ is the damping ratio, ωn Therefore, the transfer function for EEV’s temperature sensor
is the natural frequency of the system. can be expressed as:

ΔDSm (s) ΔTse (s) 1

3.1.4. EEV’s temperature sensor H3 (s) = = = (14)
ΔDS (s) ΔTre (s) τ ses + 1
Fig. 4 shows the installation of EEV’s temperature sensor at-
tached to the refrigerant pipe at evaporator exit and an
equivalent thermal circuit for the heat transfer from refriger- 1
ΔDSm (s) = ΔDS (s) (15)
ant inside the pipe to the sensor. The heat transfer between τ ses + 1
 international journal of refrigeration 69 (2016) 394–406 399

Eq. (15) suggests that due to the dynamics of EEV’s tempera-

Table 2 – Specifications of EEV’s temperature sensor
ture sensor, the transient response for DSm, as a feedback control
used in the experimental DX A/C system.
signal to regulate EEV’s opening, is slowed down following a
Sensor diameter Sensor length Heat capacity of
change in the actual DS of refrigerant at evaporator exit.
(mm) (mm) the sensor (ρCpV)se
(J K−1)
3.2. Identification of transfer function parameters
4 30 1.42

Using the transfer functions for each of the components in the

PI controlled EEV–evaporator control loop, the open-loop trans-
fer function, G(s), for characterizing the transient response for When carrying out experiments, compressor and supply fan
the measured DS, DSm, to a change in EEV’s control signal, ue, speeds were fixed at 4680 rpm and 2880 rpm, respectively, and
can be expressed as: the inlet air state to the DX A/C system was maintained at 25 oC
and 50% relative humidity (RH) by modulating the LGUs in the
ΔDSm (s) K v Keω n2 conditioned space. The EEV’s opening was varied between 35%
G (s) = = H1 (s) H2 (s) H3 (s) = 2 e −θ s
Δue (s) (s + 2ζω ns + ω n2 ) (1 + τ ses) and 45% of its full opening with an increment of 2%. By track-
ing the dynamic responses for the DSm to a step change in EEV’s
(16)
control signal, the transfer function, G(s), can be approxi-
The parameters in Eq. (16) were identified experimentally mated to:
using the experimental DX A/C system.
Firstly, it was noted that in the experimental DX A/C system, −1.6
G (s) = e−31s (18)
for the purpose of reducing the influence of EEV’s tempera- (207s2 + 29.5s + 1) (1.6s + 1)
ture sensor on the rate of DS signal transfer as far as possible,
the temperature sensor used in experiments hence had a quick Eq. (18) was further simplified by using Padé Approxima-
response characteristic and was attached well to the refriger- tion (Golub and Van Loan, 2012) to:
ant pipe at evaporator exit. For the EEV’s temperature sensor
in the experimental DX A/C system, its specifications are shown 1.6s − 0.1032
G (s) ≈ (19)
in Table 2. Assuming a contact resistance of 0.4 ~ 1 × 10−4 m2 (270s3 + 42.85s2 + 2.903s + 0.06452) (1.6s + 1)
K W−1 (Lienhard, 2013) under normal sensor installation, the
time constant of the EEV’s temperature sensor was about For validating Eq. (19), a further experiment was carried out
0.95 ~ 2.3 s. Therefore, using an average time constant of 1.6 s when the EEV’s opening was firstly changed from 39.5% to 42.5%
for the EEV’s temperature sensor, Eq. (16) can be written as: at 640 s, and then from 42.5% to 40.5% at 1814 s. Fig. 5 shows
comparison between the experimental and the predicted re-
ΔDSm (s) K v Keω n2 sponses using Eq. (19). As seen, the predicted transient
G (s) = = H1 (s) H2 (s) H3 (s) = 2 e −θ s
Δue (s) ( s + 2ζω n s + ω n ) (1 + 1.6s)
2
responses for DSm agreed very well with the experimental
(17) responses.

12
10
Experimental response
8
Predicted response
DSm ( C)

6
o

4
2
0
0 500 1000 1500 2000 2500
Time (s)
EEV's opening (%)

44

42

40 640s 1814s

38
0 500 1000 1500 2000 2500
Time (s)

Fig. 5 – The comparison between the predicted and experimental responses for the DSm.
400 international journal of refrigeration 69 (2016) 394–406

Therefore, the open-loop transfer function, L(s), for the PI 1.6 (s − 0.0645)
L (s) = ( K ps + Ki ) (21)
controlled EEV–evaporator control loop in the experimental DX s (s + 0.0645) (18s + 1) (11.5s + 1) (1.6s + 1)
A/C system can be expressed as:
From Eq. (21), the poles for L(s) were 0, −0.0645, −0.0555,
L (s) = C (s) G (s) −0.0869 and −0.625, respectively. None of the poles lay on the
1.6s − 0.1032
= ⎛ Kp + i ⎞ ⎛ ⎞⎛
K 1 ⎞ right hand side of s-plane. Hence, the stability of the closed
(20)
⎝ 6452 ⎠ ⎝ 1.6s + 1 ⎠
s ⎠ ⎝ 270s3 + 42.85s2 + 2.903s + 0.6 EEV–evaporator control loop may be determined by examin-
ing whether the contour for the open-loop transfer function
L(s), ГL, encircled the point (−1, 0) on its Nyquist diagram.
4. Stability analysis for the EEV–evaporator To examine the influences of EEV’s PI settings on the sta-
control loop using the Frequency Response bility of the control loop, based on the commonly used PI
Method settings for a PI controlled EEV at normal operation condi-
tions, two groups of different EEV’s PI settings, each containing
four study cases as shown in Table 3, were set to obtain their
In a DX A/C system, its operational stability is mainly deter-
corresponding Nyquist diagrams shown in Fig. 6. As seen in
mined by the stability of its EEV–evaporator control loop. If the
Fig. 6(a) which is for Group I results, at a given integral gain,
EEV–evaporator control loop is stable, the operation of the DX
Ki = −0.025, when increasing the proportional gain, Kp, ГL changed
A/C system will be stable and vice versa. Therefore, in this paper,
to tend to encircle the point (−1, 0). When Kp was less than −1,
the analysis of the system operational stability of the experi-
ГL did not encircle the point (−1, 0). However, when the values
mental DX A/C system was via examining the stability of the
of Kp were greater than −1, ГL encircled the point (−1, 0), sug-
EEV–evaporator control loop.
gesting the instability of the EEV–evaporator control loop. On
With the availability of the open-loop transfer function, L(s),
the other hand, as seen in Fig. 6(b) which is for Group II results,
for the PI-controlled EEV–evaporator control loop, the Fre-
at a fixed proportional gain, Kp, increasing the integral gain,
quency Response Method in the classical control theory was
Ki, may also lead to instability of the EEV–evaporator control
used to assess the stability of the control loop by computing
loop. Therefore, the EEV’s PI settings could impact on the sta-
gain and phase margins, which can be used to determine if
bility of the EEV–evaporator control loop.
the control loop was stable or not.
4.1.2. The influences of the time constant of EEV’s
4.1. Nyquist stability criterion temperature sensor on the stability of the EEV–evaporator
control loop
In using the Frequency Response Method in the classical control For examining the effects of different time constants of EEV’s
theory, Nyquist stability criterion is a graphical technique to temperature sensor on stability, Eq. (20) can also be repro-
determine the stability of a dynamic controlled system and has duced by designating τse as a variable, as:
been widely used for designing and analyzing systems with
feedback. The importance of Nyquist stability lies in the fact 1.6 (s − 0.0645)
L′ (s) = ( K ps + Ki ) (22)
that it can be used to determine the relative degree of system s (s + 0.0645) (18s + 1) (11.5s + 1) (τ ses + 1)
stability by producing the phase and gain stability margins. With
the help of Nyquist stability criterion, the stability of the EEV– In Eq. (22), since none of the poles for L‘(s) lay on the right
evaporator control loop was investigated by analyzing the hand side of the s-plane, the stability of the closed EEV–
frequency response of its open-loop transfer function, L(s). evaporator control loop may also be determined by examining
whether the contour of L’(s), ГL’, encircled the point (−1, 0) on
4.1.1. The influences of EEV’s PI settings on the stability of its Nyquist diagram.
the EEV–evaporator control loop In the experimental DX A/C system, the average time con-
The open-loop transfer function, L(s), i.e., Eq. (20), for the PI- stant, τse, of its EEV’s temperature sensor was estimated at ~
controlled EEV–evaporator control loop can be reproduced by 1.6s, as mentioned in Section 3.2. However, if the contact
designating Kp and Ki as variables, as: between the temperature sensor and refrigerant pipe was

Table 3 – The stability of the EEV–evaporator control loop at two different groups of EEV’s PI settings.
Group Study case Proportional gain Integral gain Gain margin Phase margin Stability
(Kp) (Ki) (dB) (°)
I I - 1* −0.5 −0.025 0.636 4.93 Stable
I - 2* −1 −1.08 −12.3 Unstable
I-3 −1.5 −3.51 −40.9 Unstable
I-4 −2 −5.56 −61.6 Unstable
II II - 1 −0.5 −0.01 5.63 63.9 Stable
II - 2 −0.020 2.44 20.2 Stable
II - 3* −0.025 0.636 4.93 Stable
II - 4* −0.035 −2.61 −18.7 Unstable
* Study cases further experimentally verified in Section 5.1.
 international journal of refrigeration 69 (2016) 394–406 401

1 1
Ki= -0.010 Ki= -0.020
(-1,0) Ki= -0.025 Ki= -0.035

0 0

Kp = -0.5
Kp = -1.0
-1 Kp = -1.5 -1
(-1,0)
Kp = -2.0

Imaginary axis
Imginary axis
-2 -2

-3 -3

-4 -4

-5 -5
-2 -1 0 -2 -1 0
Real axis Real axis
(a) Group I (b) Group II

Fig. 6 – Nyquist diagrams for L(s) at two different groups of EEV’s PI settings.

insufficient, or EEV’s temperature sensor and refrigerant pipe that the EEV–evaporator control loop was unstable. When τse
were not adequately insulated thermally, the time constant of was further increased to over 50s, the ГL’ did not encircle the
the sensor would become relatively large. Therefore, three point (−1, 0) again, suggesting the stable operation of the control
groups of different time constants, τse, each containing five study loop. Therefore, for an initially stable EEV–evaporator control
cases for EEV’s temperature sensor ranging from 0s to 200s at loop, slowing down its rate of DS signal transfer by increas-
fixed PI settings, shown in Table 4, were set to examine their ing τse may first lead to instability, although further increasing
influences on the stability of the EEV–evaporator control loop τse can return the loop to be stable again.
using Eq. (22). Figs. 7(b) and 7(c) show the Groups IV and V results, where
The Nyquist diagrams for L’(s) at the three groups of dif- the stability of an initially unstable EEV–evaporator control loop
ferent time constants are shown in Fig. 7. From Fig. 7(a) which at different values of τse was examined. As seen in Figs. 7(b)
is for Group III results, it can be seen that at fixed Kp = −1 and and 7(c), when τse = 0s, the control loop at their PI settings, i.e.,
Ki = −0.015, when τse = 0s, ГL’ did not encircle the point (1, 0). Kp = −1.5, Ki = −0.025 and Kp = −2, Ki = −0.025, were unstable
However, when τse was at 10s, ГL’ encircled the point (−1, 0), so initially. When increasing τse, ГL’ would firstly move away from

Table 4 – The stability of the EEV–evaporator control loop at three different groups of time constant of EEV’s temperature
sensor.
Group Study case Proportional gain Integral gain Time constant, Gain margin Phase margin Stability
(Kp) (Ki) τse (s) (dB) (°)
III III - 1 −1 −0.015 0 0.56 7.29 Stable
III - 2 10 −0.44 −6.16 Unstable
III - 3 50 1.65 13.6 Stable
III - 4 100 4.02 20.8 Stable
III - 5 200 7.02 21.2 Stable
IV IV - 1 −1.5 −0.025 0 −3.14 −34.2 Unstable
IV - 2 10 −4.19 −51.5 Unstable
IV - 3* 50 −2.39 −19.3 Unstable
IV - 4* 100 −0.28 −1.5 Unstable
IV - 5* 200 2.33 7.44 Stable
V V-1 −2 −0.025 0 −5.2 −53.4 Unstable
V-2 10 −6.13 −72.5 Unstable
V-3 50 −3.66 −32.4 Unstable
V-4 100 −0.916 −6.03 Unstable
V-5 200 2.62 10.9 Stable
* Study cases further experimentally verified in Section 5.2.
402 international journal of refrigeration 69 (2016) 394–406

1 1 1
Kp= -1 Kp= -1.5 Kp= -2
Ki= -0.015 Ki= -0.025 Ki= -0.025

0 0 0
(-1,0) (-1,0) (-1,0)
=100s se
=200s
se

-1 =0s -1 -1
se
se
=200s
=100s

Imaginary axis
Imaginary axis

se

Imaginary axis
se
=0s =100s =0s
se se
-2 -2 -2
se
=10s

-3 -3 -3

se
=10s
=10s
-4 =50s -4 se -4
se =50s
se

se
=50s

-5 -5 -5
-2 -1 0 -4 -3 -2 -1 0 -4 -3 -2 -1 0
Real axis Real axis Real axis
(a) Group III (b) Group IV (c) Group V

Fig. 7 – Nyquist diagrams for L’(s) at three different groups of time constants of EEV’s temperature sensor.

the point (−1, 0), but if further increasing τse, it would move back presented in this Section. The experiments were conducted
to be close to the point (−1, 0), suggesting that the EEV– using the experimental DX A/C system described in Section
evaporator control loop tended to be stable as τse became larger. 2. When carrying out the experiments, for simulating the dif-
When τse was at 200 s, the contour in Figs. 7(b) and 7(c), ГL’, did ferent time constants of EEV’s temperature sensor, τse, a first-
not encircle the point (−1, 0), suggesting the stable operation order transfer function, i.e. Eq. (14), was incorporated into the
of the EEV–evaporator control loop. Therefore, it can be ob- EEV–evaporator control loop in the experimental DX A/C system,
served when the EEV–evaporator control loop was initially as shown in Fig. 8.
unstable, increasing the time constant of the EEV’s tempera- During the experiments, the compressor and supply fan
ture sensor to a certain value for slowing down the rate of DS speeds were fixed at 4680 rpm and 2880 rpm, respectively. On
signal transfer may help eliminate the instability of the control the other hand, there were an air heater and a water heater
loop. in the LGUs. The two heaters were program controlled to match
Therefore, it can be seen that the rate of the DS signal trans- the sensible and latent loads in the conditioned space, so that
fer, as reflected by the time constant of EEV’s temperature the air temperature and relative humidity entering the DX
sensor, did impact the operational stability of the EEV– evaporator were maintained at 25oC and 50%, respectively. The
evaporator control loop, thus the DX A/C system. Furthermore, DS setting was fixed at 8 oC.
the observations from the analysis were that in general a larger
time constant would lead to a stable operation of the DX A/C
system.

5. Experimental validation

In order to validate the theoretical analysis results using the

classical control theory presented in Section 4, experimental
work was carried out for those study cases marked with an Fig. 8 – Block diagram of the PI controlled EEV–evaporator
asterisk in Tables 2 and 3, and experimental results are control loop after incorporating H3(s).
 international journal of refrigeration 69 (2016) 394–406 403

5.1. The influences of EEV’s PI settings on the stability of controlled parameter, for the study cases I-1, I-2 and II-3, II-4,
the EEV–evaporator control loop respectively, where different EEV’s PI settings on the stability
of the EEV–evaporator control loop were examined. As seen,
Figs. 9 and 10 show the experimental results of the DSm, an in the experimental results for both parameters verified the sta-
directly controlled parameter, and EEV’s opening, a directly bility analysis results using the classical control theory

12 Study case I-1, Study case I-2

10
DSm ( C)
o

0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
(a)
44
EEV's opening (%)

42 Study case I-1, Study case I-2

40

38

36

34
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
(b)

Fig. 9 – Experimental validation results for the study cases I-1 and I-2.

12 Study case II-3, Study case II-4

10
DSm ( C)
o

0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
(a)
44

Study case II-3, Study case II-4

EEV's opening (%)

42

40

38

36

34
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (s)
(b)

Fig. 10 – Experimental validation results for the study cases II-3 and II-4.
404 international journal of refrigeration 69 (2016) 394–406

16
14 Study case IV-3, Study case IV-4, Study case IV-5
12
10
DSm ( C)
o
8
6
4
2
0 200 400 600 800 1000 1200 1400 1600 1800
Time (s)
(a)
46
44 Study case IV-3, Study case IV-4, Study case IV-5
EEV's opening (%)

42
40
38
36
34

0 200 400 600 800 1000 1200 1400 1600 1800

Time (s)
(b)

Fig. 11 – Experimental validation results for the study cases IV-3, IV-4 and IV-5.

presented in Section 4.1.1, that the EEV–evaporator control loop characteristics of an expansion valve. In a TEV-controlled re-
was stable for the study cases I-1 and II-3, but not stable for frigeration system, it was previously shown that both of the TEV’s
the study cases I-2 and II-4. control characteristics, such as its static superheat setting and
valve gain (Lenger et al., 1998; Mithraratne and Wijeysundera,
5.2. The influences of the time constants of EEV’s 2001, 2002; Mithraratne et al., 2000; Tassou and Al-Nizari, 1993)
temperature sensor on the stability of the EEV–evaporator and the characteristics of TEV’s sensing bulb, including the time
control loop constant of TEV’s sensing bulb, the thermal resistance between
the sensing bulb and refrigerant pipe and the bulb location
Figs. 11 shows the experimental results of DSm and EEV’s (Broersen and Vanderjagt, 1980; Chen and Jiang, 1990; Ibrahim,
opening for study cases IV-3, IV-4 and IV-5, respectively, where 1998, 2001; Lenger et al., 1998; Mithraratne and Wijeysundera,
the influences of different time constants on the stability 2001, 2002; Mithraratne et al., 2000; Mulay et al., 2005; Stoecker,
of the EEV–evaporator control loop were examined. As seen from 1966; Tassou and Al-Nizari, 1993), did influence the stability of
the Figures, the experimental results for both indirectly and refrigeration systems. On the other hand, since there were no
directly controlled parameters verified the stability analysis previous studies on the influences of EEV’s PI settings and time
results using the classical control theory presented in Section constant of the EEV’s temperature sensor on the stability of an
4.1.2, that the EEV–evaporator control loop was not stable for EEV controlled refrigeration system, the study results presented
the study cases IV-3 and IV-4, but stable for the study case IV-5. in this paper filled the gap and further confirmed that the char-
In addition, it is noted that the experimental and analysis acteristics of an EEV can also influence the stability of an EEV
results reported in this paper were based on only one set of controlled refrigeration system in the following two aspects:
the operating condition for the experimental DX A/C system.
While further follow-up studies based on other sets of oper- • The control characteristics of a PI-controlled EEV im-
ating conditions are required, the general variation trends pacted the stability of the EEV–evaporator control loop, thus
observed and the related analysis should remain valid, al- the refrigeration system. Generally, a larger proportional gain
though the absolute numerical value may be different. or integral gain would lead to a high chance for the EEV-
controlled refrigeration system to become unstable.
• The theoretical analysis using the classical control theory
and the experimental results shown in Sections 4.1.2 and
6. Discussions 5.2, respectively, for the influences of the dynamic charac-
teristics of EEV’s temperature sensor on the operational
As mentioned earlier in Introduction, there have been two views stability of the EEV–evaporator control loop, are the first of
on the causes for the hunting in a refrigeration system: the in- its kind reported in open literature. The results demon-
herent characteristics of an evaporator and the control strated that a larger time constant of EEV’s temperature
 international journal of refrigeration 69 (2016) 394–406 405

sensor can also lead to a high chance for the EEV–evaporator Chen, W., Chen, Z.J., Zhu, R.Q., Wu, Y.Z., 2002. Experimental
loop to become stable, at however the expense of reduc- investigation of a minimum stable superheat control system
ing the sensitivity of superheat control for the EEV during of an evaporator. Int. J. Refrigeration 25 (8), 1137–1142.
Chen, Y.M., Deng, S.M., Xu, X.G., Chan, M.Y., 2008. A study on the
normal operation.
operational stability of a refrigeration system having a
variable speed compressor. Int. J. Refrigeration 31 (8), 1368–
The analysis and experimental results reported clearly sug- 1374.
gested that a larger change, as reflected by the controller’s P Chen, Z.J., Jiang, W.Q., 1990. Stability analysis of refrigeration of
and I settings, and a faster change, as reflected by the rate of evaporator and thermo expansion valve control loop.
DS signal transfer, or the time constant of EEV’s temperature J. Shanghai Jiaotong Univ. 24, 58–66.
Eames, I.W., Milazzo, A., Maidment, G.G., 2014. Modelling
sensor, of the refrigerant mass flow supplied to an evapora-
thermostatic expansion valves. Int. J. Refrigeration 38, 189–
tor, would very likely lead to an instability operation of the EEV–
197.
evaporator control loop, thus the DX A/C system. These results Fallahsohi, H., Changenet, C., Place, S., Ligeret, C., Lin-Shi, X.,
further confirmed that the operating characteristics of an ex- 2010. Predictive functional control of an expansion valve for
pansion valve in a refrigeration system could impact its minimizing the superheat of an evaporator. Int. J.
operational stability. Refrigeration 33 (2), 409–418.
Furthermore, the experimental results also suggested an ef- Golub, G.H., Van Loan, C.F., 2012. Matrix Computations. JHU Press.
Huang, D., Zhao, R.J., Gong, Q.Q., Tang, X.Q., 2014. Effects of water
fective approach to mitigate instability problem encountered
mal-distribution on thermostatic-expansion-valve stability
in an EEV controlled refrigeration system by incorporating a
and its controlled two-circuit evaporator performance. Int. J.
first-order transfer function in its EEV–evaporator control loop Refrigeration 48, 38–47.
to slow down the rate of DS signal transfer, which can however Huelle, Z., 1972. MSS line – a new approach to hunting problem.
reduce the sensitivity of its DS control. ASHRAE J. 14 (10), 43–46.
Huelle, Z.R., 1967. Heat load influences upon evaporator
parameters. Proceedings of XII International Congress of
Refrigeration Meeting, Madrid, pp. 985–999.
7. Conclusion Ibrahim, G.A., 1998. Theoretical investigation into instability of a
refrigeration system with an evaporator controlled by a
thermostatic expansion valve. Can. J. Chem. Eng. 76 (4), 722–
A study on the influences of the operating characteristics of a
727.
PI controlled EEV on the operational stability of a DX A/C system Ibrahim, G.A., 2001. Effect of sudden changes in evaporator
has been carried out and study results are reported in this paper. external parameters on a refrigeration system with an
Using the classical control theory, the influences of EEV’s PI set- evaporator controlled by a thermostatic expansion valve. Int.
tings and time constants of EEV’s temperature sensor on the J. Refrigeration 24 (6), 566–576.
system operational stability were analyzed and further experi- Jolly, P.G., Tso, C.P., Chia, P.K., Wong, Y.W., 2000. Intelligent control
to reduce superheat hunting and optimize evaporator
mentally verified using an experimental DX A/C system.
performance in container refrigeration. HVAC&R Res 6 (3),
The theoretical analysis and experimental results showed that 243–255.
EEV’s PI settings and the time constant of EEV’s temperature Lenger, M.J., Jacobi, A.M., Hrojak, P.S., 1998. Superheat Stability of
sensor which affected the rate of DS signal transfer did impact an Evaporator and Thermostatic Expansion Valve. Air
the operational stability of the EEV–evaporator control loop, thus Conditioning and Refrigeration Center. College of
the DX A/C system. A larger proportional gain or integral gain Engineering. University of Illinois at Urbana-Champaign.
of the EEV’s PI controller, and faster DS signal transfer due to Li, X.Q., Chen, J.P., Chen, Z.J., Liu, W.H., Hu, W., Liu, X.B., 2004. A
new method for controlling refrigerant flow in automobile air
a smaller time constant of EEV’s temperature sensor would very
conditioning. Appl. Therm. Eng. 24 (7), 1073–1085.
likely lead to system instability. The study results further con- Liang, N., Shao, S.Q., Xu, H.B., Tian, C.Q., 2010. Instability of
firmed that the operating characteristics of an expansion valve refrigeration system – A review. Energy Conv. Manage 51 (11),
impacted a refrigeration system operational stability. 2169–2178.
Lienhard, J.H., 2013. A Heat Transfer Textbook, fourth ed. Courier
Corporation.
Mithraratne, P., Wijeysundera, N.E., 2001. An experimental and
Acknowledgments numerical study of the dynamic behaviour of a counter-flow
evaporator. Int. J. Refrigeration 24 (6), 554–565.
Mithraratne, P., Wijeysundera, N.E., 2002. An experimental
The financial supports for the Hong Kong Polytechnic Univer-
and numerical study of hunting in thermostatic-expansion-
sity (Project No. G-YBHH) are gratefully acknowledged.
valve-controlled evaporators. Int. J. Refrigeration 25 (7), 992–
998.
Mithraratne, P., Wijeysundera, N.E., Bong, T.Y., 2000. Dynamic
REFERENCES
simulation of a thermostatically controlled counter-flow
evaporator. Int. J. Refrigeration 23 (3), 174–189.
Mulay, V., Kulkarni, A., Agonafer, D., 2005. Effect of the location
Aprea, C., Renno, C., 2001. Experimental analysis of a transfer and the properties of thermostatic expansion valve sensor
function for an air cooled evaporator. Appl. Therm. Eng. 21 (4), bulb on the stability of a refrigeration system. J. Heat Transfer
481–493. 127 (1), 85–94.
Broersen, P.M.T., Vanderjagt, M.F.G., 1980. Hunting of evaporators Qi, Q., Deng, S.M., Xu, X.G., Chan, M.Y., 2010. Improving degree of
controlled by a Thermostatic Expansion Valve. J. Dyn. Sys., superheat control in a direct expansion (DX) air conditioning
Meas. Control 102 (2), 130–135. (A/C) system. Int. J. Refrigeration 33 (1), 125–134.
406 international journal of refrigeration 69 (2016) 394–406

Stoecker, W.F., 1966. Stability of an evaporator-expansion valve Wedekind, G.L., Stoecker, W.F., 1968. Theoretical model for
control loop. ASHRAE Trans. 72 (2), IV.3.1–IV.3.8. predicting the transient response of the mixture-vapor
Tassou, S.A., Al-Nizari, H.O., 1993. Effect of refrigerant flow transition point in horizontal evaporating flow. J. Heat
control on the thermodynamic performances of reciprocating Transfer 90 (1), 165–174.
chillers. Appl. Energy 45 (2), 101–116. Yasuda, H., Touber, S., Machielsen, C., 1983. Simulation-model of
Visioli, A., 2006. Practical PID Control. Springer Science & a vapor compression refrigeration system. ASHRAE J. 89 (2A),
Business Media. 408–425.
Wedekind, G., 1971. An experimental investigation into the Zahn, W.R., 1964. A visual study of two-phase flow while
oscillatory motion of the mixture-vapor transition point in evaporating in horizontal tubes. J. Heat Transfer 86 (3), 417–
horizontal evaporating flow. J. Heat Transfer 93 (1), 47–54. 429.

View publication stats