58) Dynamic Characterization
58) Dynamic Characterization
58) Dynamic Characterization
Research Paper
h i g h l i g h t s
Phase plot, FFT, correlation dimension, LLE are used to conform flow dynamics.
Phase plots shows limit cycle in the oscillatory regime.
Phase plot shows Lorenz like strange attractor in the flow reversal regime.
For oscillatory region correlation dimension is 1 and LLE is negative.
For flow reversal region correlation dimension is fractional and LLE is positive.
a r t i c l e i n f o a b s t r a c t
Article history: Natural circulation loop (NCL) has many applications in the energy sector as a passive heat transfer sys-
Received 5 October 2016 tem. The passive nature of its operation increases the reliability of natural circulation loops and makes
Revised 20 April 2017 them crucial for the safety of the system. However, depending on the operating conditions, the dynamics
Accepted 18 September 2017
of the system can vary significantly. For this reason proper analysis of the fluid flow dynamics in the NCL
Available online 19 September 2017
is very important. In this paper instability associated with the loop fluid flow in the NCL is investigated
from a dynamical systems perspective with the help of phase plot, Fast Fourier Transform (FFT), correla-
Keywords:
tion dimension and largest Lyapunov exponent. From the numerical analyses it is found that with the
Natural circulation loop
Nonlinear dynamics
increase in heater power, dynamics of the fluid flow in the loop changes significantly. Loop characteristics
Correlation dimension remain steady up to a certain heater power level but with the increase in heater power level first it shows
Lyapunov exponent periodic oscillations and then it shows chaotic behavior. The existence of periodic and chaotic regimes is
established using tools of dynamic systems analysis.
Ó 2017 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.applthermaleng.2017.09.092
1359-4311/Ó 2017 Elsevier Ltd. All rights reserved.
R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1127
Nomenclature
A area of cross section (m2) ho heat transfer coefficient between wall & ambient
Ain internal cross section area of loop (m2) (W/m2K)
A0 external cross section area of loop (m2) hex heat transfer coefficient between wall & coolant
Cp specific heat of fluid at constant pressure (J/kgK) (W/m2K)
Cpw specific heat of wall at constant pressure (J/kgK) kf thermal conductivity of fluid (W/mK)
Cpex specific heat of coolant at constant pressure (J/kgK) kw thermal conductivity of wall (W/mK)
din internal loop diameter (m) kex thermal conductivity of heat exchanger (W/mK)
dz change in length (m) Lloop total loop length (m)
f friction factor LH loop height (m)
Gr Grashof number NG length to diameter ratio (Lloop/D)
Grm modified Grashof number ((D3q2f bavgLHQ_ )/(Am3f Cp) Red Reynolds number
Gz Graetz number Q_ volumetric heat generation (W/m3)
Nu Nusselt number vf fluid velocity (m/s)
Pr Prandtl number Vw wall volume (m3)
t time (s) Vex heat exchanger fluid volume (m3)
Tf temperature of fluid (°C) bav thermal volumetric expansion coefficient (K1)
Tw loop wall temperature (°C) mf dynamic viscosity of fluid (kg/ms)
Tex coolant temperature (°C) t kinematic viscosity of fluid (m2/s)
Tref reference temperature (°C) qf fluid density (kg/m3)
Ta ambient temperature (°C) qw density of wall (kg/m3)
Ra Rayleigh number qex coolant density (kg/m3)
hi heat transfer coefficient between fluid & wall (W/m2K)
accurate characterization of the instabilities associated with the using computer code SUCLIN was carried out by Sharma et al. [22].
flow dynamics of NCLs. Proper identification and characterization Vijayan et al. [23] carried out different experiments and numerical
of these instabilities are prerequisites for control of instabilities simulation by computer code ATHLET to observe that instability
and stable operation of these systems. As the system is highly non- grows with the increase in diameter and the flow dynamics changes
linear in nature, common characterization tools like Fast Fourier with the increase in power level. Vijayan et al. [24] studied nature of
Transform (FFT) cannot fully characterize the systems. On the flow dynamics and stable regime both experimentally and numer-
other hand, nonlinear dynamic systems show certain general char- ically. Experimentally three different flow oscillations were found
acteristic features irrespective of the physical nature of the system. but numerically only two different flow oscillations were observed.
Thus dynamical systems approach has proved to be useful for They also observed that the flow dynamics shows a conditionally
dynamic characterization of complex systems. This approach has stable region near the stability threshold.
been successfully applied to various engineering systems like com- In spite of the large volume of work on stability and dynamics of
bustion systems [11,12], condition monitoring of gear boxes [13] NCL, very few works have been reported, which systematically
and vibration fault diagnosis of roller bearings [14]. characterize the system dynamics using quantitative tools of non-
Single phase natural circulation loops have been used in many linear dynamics. Zhang et al. [25] analysed the time series data
engineering systems owing to their lower susceptibility to instabil- using tools of nonlinear dynamics but they considered only peri-
ities compared to their two-phase counterparts. Consequently, a odic regime and not the flow reversal regime.
significant volume of literature exists on dynamics and control of In the present study, numerical simulations have been carried
single phase natural circulation loops [15–24]. Nayak et al. [7] out for different power levels with the help of Simulink model
showed that the instability associated with the operation of single which was developed earlier by our group [26,27]. The dynamic
phase NCL significantly reduced with the use of Al2O3 nano-fluids. analysis of loop fluid flow behavior has been carried out to quantify
By two dimensional numerical analysis, dynamical behavior of fluid the transition of dynamic states from steady state to chaos through
flow for a rectangular NCL due to the change in Rayleigh number periodic oscillation. For this we use phase plot, Fast Fourier Trans-
was performed by Desrayaud et al. [15]. The flow dynamics for form (FFT), correlation dimension and largest Lyapunov exponents
the annular thermosyphon was investigated experimentally and as non-linear dynamic analysis tools [11,28–35]. A major motiva-
numerically by Desrayaud et al. [16]. They found that flow dynam- tion for such analysis is to distinguish possibly chaotic systems,
ics changes from steady flow to Lorenz-like chaotic flow through which are deterministic in nature from those dominated by noise,
periodic motion. Fichera et al. [17–19] carried out experimental which are stochastic in nature. Visual appearance of time series
and numerical study and find that flow dynamics of the NCL is data and FFT cannot make this distinction. Confirmation of deter-
changing with the change in heater power. But there is only quali- ministic nature has important implications for control of the loop
tative observation no quantitative studies have been made. Mathe- dynamics as different control strategies are required for chaotic
matical simulation was done to identify the stable, unstable and and stochastic systems.
neutrally stable points by Nyquist stability criterion for fluid flow Several tools have been suggested for studying stability of
by Nayak et al. [20]. Ridouane et al. [21] carried out numerical sim- stochastic and chaotic nonlinear systems. One of the early works
ulation of the fluid flow in the unstable convection regime for a nat- in studying stability of stochastic discreet systems is by Morozan
ural convection loop. The flow reversal phenomenon was observed [36]. While Yang and Miminis [37] derived some sufficient
to grow with the increase in Rayleigh number. Steady state and lin- conditions for local asymptotic stability and instability for discrete
ear stability analysis for super critical water natural circulation loop nonlinear deterministic and stochastic systems. Lin and Cai [38]
1128 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138
Numerical model and the geometry (Fig. 1) which is used in From the numerical investigation the system dynamics can be
present work is given in detailed in our previous paper [27]. The divided into following three categories; for heater power less than
equations for conservation of mass, momentum and energy for 625 W the system exhibits steady state behavior. But when the
the loop fluid and the energy equation for coolant are similar to heater power is increased further, an oscillatory flow is observed
our earlier works [26,27] and is mentioned briefly below. for heater power till 742 W. Heater power above 742 W, flow
@ qf @ reversal in the loop is noted. For identifying flow dynamics we
þ ðqvÞf ¼ 0 ð1Þ are analysing the results after the initial transient part is over.
@t @z
For heater power of 600 W the variation of loop mass-flux with
I q
dðqvÞf f in ðqvÞ2f time is shown in Fig. 2. The flow in a NCL is governed by the tem-
Lloop ¼ Lloop þ gbav ðTf Tref Þdz ð2Þ
dt 2qf din 0
perature difference between its two legs – the riser (having higher
temperature) and the down-comer (having lower temperature).
@ @ @ @Tf 4hi ðTf Tw Þ The difference in temperature causes a density difference, and this
ðqf Cp Tf Þ þ ððqvÞf Cp Tf Þ ¼ kf ð3Þ
@t @z @z @z din in turn causes fluid flow due to a balance of buoyancy and friction.
From Fig. 2 it can be observed that the mass flux shows a stable
@ @
ðq Cpex Tex Þ þ ððqvÞex Cpex Tex Þ steady behavior.
@t ex @z For heater power of 650 W and 740 W the variation of mass-
@ @Tex hex A0 ðTw Tex Þ flux with time is shown in Fig. 3. From Fig. 3 it can be observed that
¼ kex þ ð4Þ
@z @z Vex mass-flux shows an oscillatory nature. In the oscillatory region,
oscillation amplitude increases with the increase in heater power,
The heat conduction equation for tube wall is
which is concluded from the comparative study of heater power
@ @ @Tw hi Ain ðTf Tw Þ 650 W and 740 W as shown in Fig. 3. For heater power of 650 W,
ðqw Cpw Tw Þ ¼ kw þ the oscillation amplitude of mass flux (difference between maxi-
@t @z @z Vw
mum and minimum mass flux) is about 2.5 kg/m2 s while for hea-
h0 A0 ðTw Ta Þ Q_
þ ð5Þ ter power 740 W, the oscillation amplitude increases to about 10
Vw Vw kg/m2 s.
R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1129
Fig. 6. Spatial distribution of temperature at different instant of time for heater power 740 W.
heater 0.3 to 0.7
cooler 2.3 to 2.9 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1131
Fig. 7. Spatial distribution of temperature at different instant of time for heater power 800 W.
Fig. 8. Comparison of loop fluid temperature and wall temperature for different heater powers.
other frequencies are the harmonics, which indicate periodic Figs. 15 and 16 shows the FFT of temperature difference time
behavior. series data for heater power 750 W and 800 W, respectively. From
Fig. 14 shows variation of the dominant frequency with heater the figures it can be observed that no fundamental frequency can
power for the oscillatory region. It can be observed that fundamen- be identified. The broadband nature of the FFT clearly shows
tal frequency increases slightly with the increase in heater power. absence of periodicity in the dynamics of the system.
1132 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138
Fig. 10. Reynolds number versus Grashof number phase plot at 740 W.
From the phase plot and FFT analysis, it is difficult to infer tions. Dynamical systems analysis allows one to characterize the
whether the underlying dynamics is chaotic or dominated by dynamics of a multidimensional system from the time series of a
noise. This distinction, however, is very crucial for the selection single scalar variable by representing the data in time delayed
of appropriate strategy to control the system dynamics [11,28–35]. phase space using the embedding theorem [45]. From the time
series data of a scalar like temperature difference a vector
4.1. Quantitative dynamic characterization ½xðt0 Þ; xðt 0 þ sd Þ; xðt0 þ 2sd Þ; . . . ; xðt 0 þ ðm 1Þsd Þ can be recon-
structed where sd and m denote the delay time and embedding
For proper characterization of the system dynamics, particu- dimension respectively. These are crucial parameters in the
larly for distinction between deterministic and stochastic behavior, reconstruction of phase space. The delay time is obtained using
quantitative characterization tools based on dynamical systems autocorrelation function which is shown in Eq. (6) [46]. The time
analysis can be convenient [11,28–35]. One of these is correlation when the autocorrelation function shown below first crosses the
dimension [12]. In addition, largest Lyapunov Exponent (LLE) zero is taken as the delay time here.
[44] can be used to confirm the existence of chaos. Correlation PN
1
i¼1 ½xði þ sÞ x½xðiÞ x
dimension and LLE are determined using the time series data for C L ðsÞ ¼ N PN ð6Þ
1 2
i¼1 ½xðiÞ x
temperature difference which we get from the numerical simula- N
R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1133
Fig. 11. Reynolds number versus Grashof number phase plot at 750 W (Negative Reynolds number signifies flow in reversed direction).
Fig. 12. Reynolds number versus Grashof number phase plot at 800 W (Negative Reynolds number signifies flow in reversed direction).
1X N HðxÞ ¼ 1; x > 0
x ¼ xðiÞ
N i¼1 ¼ 0; x < 0
e is the radius of the hyper sphere in a m-dimensional space
Here x(m) denotes the instantaneous temperature at t = t0 + where x(i) and x(j) are two neighbouring vectors. The correlation
(i s), where t0 and s are the initial time and sampling time integral Cm (e) scales with the hyper sphere radius, e as Cm (e)
respectively for the temperature time series and i is an integer eDc(m) where the correlation dimension is the DC (m) for the dimen-
and N is the total number of temperature data available. sion m which is chosen. From the linear part of log Cm (e) versus log
For determination of the embedding dimension we are using (e) curve the correlation dimension is evaluated as shown in Eq. (8)
the Grassberger and Procacia algorithm [45]. According to this and it is a function of the embedding dimension, m.
algorithm, first we need to calculate correlation integral as shown
in Eq. (7). log C m ðeÞ
DC ðmÞ ¼ lim ð8Þ
e!0 logðeÞ
1 X N
C m ðeÞ ¼ 2
Hðe kxm
i xj kÞ
m
ð7Þ For a random system DC increases linearly with the increment
N i;j¼1 of m and for a deterministic system DC increases with the incre-
i–j
ment of m, for some time and then becomes practically constant.
Here kxm
i xj k represents the Euclidean distance between the
m
This final value of DC is called correlation dimension and the value
two vectors x(i) and x(j). H() represents the Heaviside function of m where final DC value appear first is called embedding
given by: dimension.
1134 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138
If the system is deterministic then to confirm the existence of 650 W and 800 W respectively. From the figures it can be observed
chaos LLE can be used [44,47–51]. To find out the LLE we need to that as the hyper sphere radius (e) increases correlation integral
find out nearest neighbor for each point of the reconstructed data also increases but thereafter it saturates at a higher radius. It can
series. be observed that with increasing radius numbers of neighbouring
vectors are increased. However, after reaching the critical radius,
dj ð0Þ ¼ min kxm
i xj k
m
ð9Þ
which incorporates all the neighbouring vectors, if the radius fur-
ther increases, the results do not change and leads to saturation
1 1 X Mi
dj ðiÞ in correlation integral. Also it can be observed that saturation
k1 ðiÞ ¼ ln ð10Þ
iDt ðM iÞ j¼1 dj ð0Þ comes faster for a lower embedding dimension because of identifi-
cation of false neighbors due to inadequate embedding dimension.
where Dt is the sampling time of the time series data and dj(i) is the Also for lower heater power level saturation in correlation integral
distance between the jth pair of nearest neighbors after i discrete- comes faster i.e., at a lower value of hyper sphere radius.
time interval and k1 ðiÞ is the Lyapunov exponent at ith time instant. From the linear part of the correlation integral we can find out
Figs. 17 and 18 show the variation of correlation integral with the correlation dimension for different embedding dimension by
hyper sphere radius for different dimension for heater powers of using Eq. (8). Variation of correlation dimension with embedding
R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1135
dimension for different heater power is shown in Fig. 19. From the region, correlation dimension saturates at a fractional value more
figure it can be observed that with the increase in embedding than 2, which indicates a possibly chaotic nature.
dimension the correlation dimension first increases but after a cer- A more complete picture of the variation of correlation
tain value of embedding dimension, the correlation dimension dimension with heater power emerges can get from Fig. 20. For
becomes saturated at all the power levels. This establishes the heater power less than 625 W, the system dynamics shows
deterministic nature of the system whereas for random noise, the steady state. Here correlation dimension shows zero value but
correlation dimension increases unbounded with increase in when the system dynamics becomes oscillatory (for power
embedding dimension [12]. 625–742 W) correlation dimension becomes unity which con-
For a periodic data series correlation dimension is one, and a firms that the oscillation is periodic. For further increase in hea-
fractional correlation dimension indicates a strange attractor, ter power, the fluid flow dynamics shows the flow reversal
which could be chaotic [28]. From Fig. 19 it can be observed that characteristics and the correlation dimension becomes fractional
for heater powers 650 W and 740 W correlation dimensions satu- with values more than two which signify strange attractor. With
rates at one which properly signifies that the oscillation we get increase in heater power, the correlation dimension becomes
at this power levels are purely periodic. On the other hand, for increasingly fractional, which is implying a higher level of com-
the powers 750 W and 800 W, which are in the flow reversal plexity of the attractor.
1136 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138
Fig. 17. Variation of correlation integral with hyper-sphere radius for different embedding dimensions at 650 W.
Fig. 18. Variation of correlation integral with hyper-sphere radius for different embedding dimensions at 800 W.
Although fractional correlation dimension indicates strange Flow reversal is an undesirable phenomenon for practical appli-
attractors, but such attractors may not be necessarily chaotic cations involving NCLs and in particular for power plants. Hence
[28]. To confirm the chaotic nature, study of the LLE becomes help- such states need to be avoided. As these reversals take place at
ful. LLE for different heater powers are presented in Table 1. From irregular intervals and frequency of their occurrence increases with
Table 1 we can observe that for the oscillatory region LLE is nega- increase in heater power, these dynamics need to be suppressed
tive, but for the flow reversal region LLE is positive. LLE is a mea- using active feedback control strategies. Confirmation of chaotic
sure of divergence of two initially close states as the system behavior has important implications for such control actions.
evolves in time along a characteristic direction in the phase space.
A positive LLE indicates that the two initially close points in phase
5. Conclusion
space diverge with time. As chaotic systems are characterized by
high rate of divergence of nearby points in phase space, Lyapunov
Dynamics of a single phase NCL has been investigated through
exponent along one of the characteristic directions must be posi-
numerical simulations. From the numerical investigation it is
tive. Thus a positive value of the LLE is an indication of the chaotic
found that fluid flow dynamics are divided into three categories
nature of the system while negative values for all Lyapunov expo-
steady state, stable oscillatory flow and flow reversal. When the
nents indicate that two initially close points in phase space remain
heater power is in the oscillatory region, oscillation amplitude
close throughout i.e., the system is not chaotic.
for mass flux or temperature difference increases with the increase
R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138 1137
Acknowledgement
References
[1] R.K. Sinha, A. Kakodkar, Design and development of the AHWR – the Indian
thorium fuelled innovative nuclear reactor, Nucl. Eng. Des. 236 (2006) 683–
700.
[2] P.K. Vijayan, H. Austregesilo, Scaling laws for single-phase natural circulation
loops, Nucl. Eng. Des. 152 (1994) 331–347.
[3] M. Misale et al., Analysis of single-phase natural circulation experiments by
system codes, Int. J. Therm. Sci. 38 (1999) 977–983.
Fig. 20. Distribution of correlation dimension with power variation. [4] D.N. Basu, S. Bhattacharyya, P.K. Das, Effect of heat loss to ambient on steady-
state behaviour of a single-phase natural circulation loop, Appl. Therm. Eng. 27
(2007) 1432–1444.
[5] D.N. Basu, S. Bhattacharyya, P.K. Das, Effect of geometric parameters on steady-
Table 1
state performance of single-phase natural circulation loop with heat loss to
Largest Lyapunov exponent for different heater power. ambient, Int. J. Therm. Sci. 47 (2008) 1359–1373.
[6] P.K. Vijayan et al., Effect of loop diameter on the steady state and stability
Heater Power (W) Largest Lyapunov Exponent (LLE)
behaviour of single-phase and two-phase natural circulation loops, Sci.
650 4.710 Technol. Nucl. Install. 2008 (2008) 672704-1–672704-17.
740 3.482 [7] A.K. Nayak, M.R. Gartia, P.K. Vijayan, Thermal–hydraulic characteristics of a
750 0.1293 single-phase natural circulation loop with water and Al2O3 nanofluids, Nucl.
800 0.3327 Eng. Des. 239 (2009) 526–540.
[8] J.H. Choi, J. Cleveland, N. Aksan, Improvement in understanding of natural
circulation phenomena in water cooled nuclear power plant, Nucl. Eng. Des.
241 (2011) 4504–4514.
[9] M. Misale et al., Influence of thermal boundary conditions on the dynamic
behaviour of a rectangular single-phase natural circulation loop, Int. J. Heat
in heater power and when the heater power is in the flow reversal Fluid Flow 32 (2011) 413–423.
region, the occurrence of flow reversal increases with the increas- [10] B.T. Swapnalee, P.K. Vijayan, A generalized flow equation for single-phase
ing heater power. natural circulation loops obeying multiple friction laws, Int. J. Heat Mass
Transf. 54 (2011) 2618–2629.
From the phase plot, FFT, correlation dimension and LLE analysis [11] H. Gotoda, T. Ueda, Transition from periodic to non-periodic motion of a
it is confirmed that the fluid flow oscillation obtained in the oscilla- bunsen-type premixed flame tip with burner rotation, Proc. Combust. Inst. 29
tory region is purely periodic and fluid flow dynamics in the flow (2002) 1503–1509.
[12] S. Datta et al., An investigation of nonlinear dynamics of a thermal pulse
reversal region is chaotic. In the oscillatory region the phase plots combustor, Combust. Theor. Model. 13 (2009) 17–38.
show limit cycle and FFT has only one fundamental frequency [13] J.D. Jiang, J. Chen, L.S. Qu, The application of correlation dimension in gear box
and other frequencies are the harmonics. The correlation dimension condition monitoring, J. Sound Vib. 223 (1999) 529–541.
[14] D. Logan, J. Mathew, Using the correlation dimension for vibration fault
obtained here is 1 with a negative value for the LLE. In the flow
diagnosis of rolling element bearings—I basic concepts, Mech. Syst. Signal
reversal region phase plots show Lorenz like strange attractor and Process. 10 (1996) 241–250.
no fundamental frequency can be identified in the FFT, in place of [15] G. Desrayaud, A. Fichera, G. Lauriat, Two-dimensional numerical analysis of a
that a broad band of frequencies are present. The correlation rectangular closed-loop thermosiphon, Appl. Therm. Eng. 50 (2013) 187–196.
[16] G. Desrayaud, A. Fichera, M. Marcoux, Numerical investigation of natural
dimension saturates at a fractional value more than 2 and LLE is circulation in a 2D-annular closed-loop thermosyphon, Int. J. Heat Fluid Flow
positive which confirms that the system dynamics is chaotic. 27 (2006) 154–166.
1138 R. Saha et al. / Applied Thermal Engineering 128 (2018) 1126–1138
[17] A. Fichera, M. Froghieri, A. Pagano, Comparison of the dynamical behaviour of [34] H.R. Karimi et al., Haar wavelet-based approach for optimal control of second-
rectangular natural circulation loops, J. Process Mech. Eng. Part E 4 (2001) order linear systems in time domain, J. Dyn. Control Syst. 11 (2005) 237–252.
273–284. [35] H. Zhang et al., Robust H1 sliding mode control with pole placement for a fluid
[18] A. Fichera, A. Pagano, Neural network-based prediction of the oscillating power electrohydraulic actuator (EHA) system, Int. J. Adv. Manuf. Technol. 73
behaviour of a closed loop thermosyphon, Int. J. Heat Mass Transf. 45 (2002) (2014) 1095–1104.
3875–3884. [36] T. Morozan, Stability of stochastic discrete systems, J. Math. Anal. Appl. 23
[19] A. Fichera, A. Pagano, Modelling and control of rectangular natural circulation (1968) 1–9.
loops, Int. J. Heat Mass Transf. 46 (2003) 2425–2444. [37] X. Yang, G. Miminis, Stability of discrete deterministic and stochastic nonlinear
[20] A.K. Nayak et al., Mathematical modelling of the stability characteristics of a systems, J. Math. Anal. Appl. 168 (1992) 225–237.
natural circulation loop, Math. Comput. Modell. 22 (1995) 77–87. [38] Y.K. Lin, G.Q. Cai, Stochastic stability of non-linear systems, Int. J. Non-Linear
[21] E.H. Ridouane, C.M. Danforth, D.L. Hitt, A 2-D numerical study of chaotic flow Mech. 29 (1994) 539–553.
in a natural convection loop, Int. J. Heat Mass Transf. 53 (2010) 76–84. [39] L. Its’haki-Allerhand, U. Shaked, H1 control of a class of nonlinear systems and
[22] M. Sharma et al., Steady state and linear stability analysis of a supercritical its application to control of electronic circuit with nonlinear elements, in: IEEE
water natural circulation loop, Nucl. Eng. Des. 240 (2010) 588–597. 25th Convention of Electrical and Electronics Engineers in Israel, Israel, 2008.
[23] P.K. Vijayan, H. Austregesilo, V. Teschendorff, Simulation of the unstable [40] Z. Li, Robust control of PM spherical stepper motor based on neural networks,
oscillatory behavior of single-phase natural circulation with repetitive flow IEEE Trans. Industr. Electron. 56 (2009) 2945–2954.
reversals in a rectangular loop using the computer code ATHLET, Nucl. Eng. [41] Y. Xia et al., Robust sliding-mode control for uncertain time-delay systems
Des. 155 (1995) 623–641. based on delta operator, IEEE Trans. Industr. Electron. 56 (2009) 3646–3655.
[24] P.K. Vijayan et al., Experimental and numerical investigations on the nature of [42] P.K. Vijayan et al., A generalized flow correlation for single-phase natural
the unstable oscillatory flow in a single-phase natural circulation loop, in: XVII circulation loops, in: XVII National and VI ISHMT/ASME Heat and Mass
National and VI ISHMT/ASME Heat and Mass Transfer Conference, IGCAR, Transfer Conference, IGCAR, Kalpakkam, 2004.
Kalpakkam, 2004. [43] Y.Y. Jiang, M. Shoji, Spatial and temporal stabilities of flow in a natural
[25] W. Zhang et al., Non-linear time series analysis on flow instability circulation circulation loop: Influences of thermal boundary condition, J. Heat Transf. 125
under rolling motion condition, Ann. Nucl. Energy 65 (2014) 1–9. (2003) 612–623.
[26] R. Saha, Experimental and Numerical Study of a Single Phase Square Natural [44] M.T. Rosenstein, J.J. Collins, C.J.D. Luca, A practical method for calculating
Circulation Loop, Jadavpur University, Kolkata, 2013. largest Lyapunov exponents from small data sets, Physica D 65 (1993) 117–
[27] R. Saha et al., Experimental and numerical investigation of a single-phase 134.
square natural circulation loop, J. Heat Transf. 137 (2015) 121010-1–121010- [45] P. Grassberger, I. Procaccia, Measuring the strangeness of strange attractors,
8. Physica D 9 (1983) 189–208.
[28] R.C. Hilborn, Chaos and Nonlinear Dynamics, second ed., Oxford University [46] H.D.I. Abarbanel et al., The analysis of observed chaotic data in physical
Press, New York, 2000. systems, Rev. Mod. Phys. 65 (1993) 1331–1392.
[29] G.P. Williams, Chaos Theory Tamed, Taylor & Francis, London, 1997. [47] A. Wolf et al., Determining Lyapunov exponents from a time series, Physica D
[30] Z. Wu et al., Stability of stochastic nonlinear systems with state-dependent 16 (1985) 285–317.
switching, IEEE Trans. Autom. Control 58 (2013) 1904–1918. [48] X. Zeng, R. Eykholt, R.A. Pielke, Estimating the Lyapunov exponent spectrum
[31] X. Zhao et al., Novel stability criteria for T-S fuzzy systems, IEEE Trans. Fuzzy from short time series of low precision, Phys. Rev. Lett. 66 (1991) 3229–3232.
Syst. 22 (2014) 313–323. [49] B.J. Kim, G.H. Choe, High precision numerical estimation of the largest
[32] X. Zhao et al., Robust control of continuous-time systems with state- Lyapunov exponent, Commun. Nonlinear Sci. Numer. Simulat. 15 (2010)
dependent uncertainties and its application to electronic circuits, IEEE Trans. 1378–1384.
Industr. Electron. 61 (2014) 4161–4170. [50] H.F. Liu et al., Noise robust estimates of the largest Lyapunov exponent, Phys.
[33] H.R. Karimi, A computational method for optimal control problem of time- Lett. A 341 (2005) 119–127.
varying state-delayed systems by Haar wavelets, Int. J. Comput. Math. 83 [51] H. Kantz, A robust method to estimate the maximal Lyapunov exponent of a
(2006) 235–246. time series, Phys. Lett. A 185 (1994) 77–87.