G Model

JOCS-249; No. of Pages 7 ARTICLE IN PRESS

Journal of Computational Science xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Computational Science

journal homepage: www.elsevier.com/locate/jocs

Mathematical modelling and parameter optimization of pulsating

heat pipes
Xin-She Yang a,∗ , Mehmet Karamanoglu a , Tao Luan b , Slawomir Koziel c
a
School of Science and Technology, Middlesex University, London NW4 4BT, UK
b
School of Energy and Power Engineering, Shandong University, Jinan, China
c
School of Science and Engineering, Reykjavik University, IS-103 Reykjavik, Iceland

a r t i c l e i n f o a b s t r a c t

Article history: Proper heat transfer management is important to key electronic components in microelectronic appli-
Received 4 October 2012 cations. Pulsating heat pipes (PHP) can be an efficient solution to such heat transfer problems. However,
Received in revised form mathematical modelling of a PHP system is still very challenging, due to the complexity and multiphysics
25 November 2013
nature of the system. In this work, we present a simplified, two-phase heat transfer model, and our anal-
Accepted 13 December 2013
Available online xxx
ysis shows that it can make good predictions about startup characteristics. Furthermore, by considering
parameter estimation as a nonlinear constrained optimization problem, we have used the firefly algo-
rithm to find parameter estimates efficiently. We have also demonstrated that it is possible to obtain
Keywords:
Asymptotic good estimates of key parameters using very limited experimental data.
Firefly algorithm © 2013 Elsevier B.V. All rights reserved.
Heat transfer
Mathematical modelling
Optimization
Oscillations
Pulsating heat pipe
Two-phase flow

1. Introduction Both loop heat pipes (LHP) and pulsating heat pipes (PHP)
may provide a promising solution to such challenging prob-
A pulsating heat pipe is essentially a small pipe filled with both lems, and thus have attracted renewed attention in recent years
liquid and vapour and the internal diameter of the heat pipe is at [1,4,7,12,13,24,25,42]. In many microelectronic applications, con-
the capillary scale [1,2]. The liquids in the pipe can form segments ventional solutions to heat management problems often use fans,
or plugs, between vapour segments. When encountering heat, part heat exchangers and even water cooling. For examples, the fan-
of the liquid many evaporate and absorb some heat, thus causing a driven air circulation system in desktop computers and many
differential pressure and driving the movement of the plugs. When laptops have many drawbacks such as bulky sizes and poten-
vapour bubbles meet a cold region, some of the vapour may con- tial failure of mechanical, moving parts. In contrast, loop heat
densate, and thus releasing some heat. The loop can be open or pipes have no mechanical driving system, and heat circulation
closed, depending on the type of applications and design. This con- is carried out through the pipes, and thus such LHP systems
tinuous loop and process will form an efficient cooling system if can be very robust and long-lasting. In addition, miniaturization
designed and managed properly for a given task. Therefore, such and high-performance heat pipes systems are being developed
systems have been applied to many applications in heat exchanger, [11,15,18,26,39,43]. Simulation tools and multiphase models have
space applications and electronics, and they can potentially have been investigated [16,17]. All these studies suggested that LHP sys-
even wider applications [43]. On the other hand, the emergence tems can have some advantages over traditional cooling systems.
of nanotechnology and the steady increase of the density of the A PHP system may often look seemingly simple; however,
large-scale integrated circuits have attracted strong interests in its working mechanisms are relatively complex, as such systems
modelling heat transfer at very small scales, and the heat manage- involve multiphysics processes such as thermo-hydrodynamics,
ment of microdevices has become increasingly important for next two-phase flow, capillary actions, phase changes and others.
generation electronics and miniaturization. Therefore, many challenging issues still remain unsatisfactorily
modelled. There are quite a few attempts in the literature to model
a PHP system with various degrees of approximation and success.
∗ Corresponding author. Tel.: +44 2084112351. In this paper, we will use a mathematical model based on one
E-mail address: [email protected] (X.-S. Yang). of the best models [24,25], and will carry out some mathematical

1877-7503/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jocs.2013.12.003

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
2 X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx

analysis and highlight the key issues in the state of the art models. 3. The brightness of a firefly is determined by the landscape of the
In this paper, we intend to achieve two goals: to present a simpli- objective function.
fied mathematical model which can reproduce most characteristics
of known physics, and to provide a framework for estimating key
As a firefly’s attractiveness is proportional to the light inten-
parameters from a limited number of measurements. The rest of the
sity seen by adjacent fireflies, we can now define the variation of
paper is organized as follows: we first discuss briefly design opti-
attractiveness ˇ with the distance r by
mization and metaheuristic algorithms such as firefly algorithm,
we then outline the main multiphysics processes in the mathemat- 2
ical models. We then solve the simplified model numerically and ˇ = ˇ0 e−r , (1)
compared with experimental data drawn from the literature. After
such theoretical analysis, we then form the parameter estimation where ˇ0 is the attractiveness at r = 0.
as an optimization problem and solve it using the efficient firefly The movement of a firefly i attracted to another more attractive
algorithm for inversely estimating key parameters in a PHP system. (brighter) firefly j is determined by
Finally, we highlight the key issues and discuss possible directions
−r 2
for further research. xt+1
i
= xti + ˇ0 e ij (xtj − xti ) + ˛ ti , (2)

2. Design optimization of heat pipes where the second term is due to the attraction. The third term is
randomization with ˛ being the randomization parameter, and ti is
The proper design of pulsating heat pipes is important, so that a vector of random numbers drawn from a Gaussian distribution or
the heat in the system of concern can be transferred most effi- uniform distribution at time t. If ˇ0 = 0, it becomes a simple random
ciently. This also helps to produce designs that use the least amount walk. Furthermore, the randomization ti can easily be extended to
of materials and thus cost much less, while lasting longer without other distributions such as Lévy flights.
the deterioration in performance. Such design tasks are very chal- The Lévy flight essentially provides a random walk whose ran-
lenging, practical designs tend to be empirical and improvements dom step size s is drawn from a Lévy distribution
tend to be incremental. In order to produce better design options,
we have to use efficient design tools for solving such complex Lévy∼s− (1 <  ≤ 3), (3)
design optimization problems. Therefore, metaheuristic algorithms
are often needed to deal with such problems. which has an infinite variance with an infinite mean. Here the steps
essentially form a random walk process with a power-law step-
2.1. Metaheuristics length distribution with a heavy tail. Some of the new solutions
should be generated by Lévy walk around the best solution obtained
Metaheuristic algorithms such as the firefly algorithm and bat so far; this will speed up the local search. Lévy flights are more
algorithm are often nature-inspired [28,31], and they are now efficient than standard random walks [29].
among the most widely used algorithms for optimization. They Firefly algorithm has attracted much attention [3,10,23,35].
have many advantages over conventional algorithms, such as sim- A discrete version of FA can efficiently solve non-deterministic
plicity, flexibility, quick convergence and capability of dealing with polynomial-time hard, or NP-hard, scheduling problems [23], while
a diverse range of optimization problems. There are a few recent a detailed analysis has demonstrated the efficiency of FA for a
reviews which are solely dedicated to metaheuristic algorithms wide range of test problems, including multiobjective load dis-
[28,29,34]. Metaheuristic algorithms are very diverse, including patch problems [3,32]. Highly nonlinear and non-convex global
genetic algorithms, simulated annealing, differential evolution, optimization problems can be solved using firefly algorithm effi-
ant and bee algorithms, particle swarm optimization, harmony ciently [9,35]. The literature of firefly algorithms have expanded
search, firefly algorithm, cuckoo search, flower algorithm and oth- significantly, and Fister et al. provided a comprehensive literature
ers [29,30,33,36]. review [8].
In the context of heat pipe designs, we have seen a lot of interests
in the literature [5,7]. However, for a given response, to identify the
3. Mathematical model for a PHP system
right parameters can be considered as an inverse problem as well
as an optimization problem. Only when we understand the right
3.1. Governing equations
working ranges of key parameters, we can start to design better
heat-transfer systems. To our knowledge, this is the first attempt to
Mathematical modelling of two-phase pulsating flow inside a
use metaheuristic algorithms to identify key parameters for given
pulsating heat pipe involves many processes, including interfa-
responses. We will use the firefly algorithm to achieve this goal.
cial mass transfer, capillary force, wall shear stress due to viscous
action, contact angles, phase changes such as evaporation and con-
2.2. Firefly algorithm densation, surface tension, gravity and adiabatic process. All these
will involve some constitutive laws and they will be coupled with
Firefly algorithm (FA) was first developed by Xin-She Yang in fundamental laws of the conservation of mass, momentum and
2008 [28,30] and is based on the flashing patterns and behavior of energy, and thus resulting in a nonlinear system of highly coupled
fireflies. In essence, FA uses the following three idealized rules: partial differential equations. Consequently, such a complex model
can lead to complex behaviour, including nonlinear oscillations and
1. Fireflies are unisex so that one firefly will be attracted to other even chaotic characteristics [14,19,20,24,27,41].
fireflies regardless of their sex. A mathematical model can have different levels of complexity,
2. The attractiveness is proportional to the brightness and they and often a simple model can provide significant insight into the
both decrease as their distance increases. Thus for any two flash- working mechanism of the system and its behaviour if the model
ing fireflies, the less brighter one will move towards the brighter is constructed correctly with realistic conditions. In most cases, full
one. If there is no brighter one than a particular firefly, it will mathematical analysis is not possible, we can only focus on some
move randomly in the form of a random walk. aspects of the model and gain some insight into the system.

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx 3

A key relationship concerning the inner diameter of a typical In the above equations, we also assume the vapour acts as an
pulsating heat pipe is the range of capillary length, and many design ideal gas, and we have
options suggest to use a diameter near the critical diameter [5]
 mv RTv
pv = . (13)
 V
dc = 2 , (4)
g(l − v ) The shear stress sw between the wall and the liquid plug is given
where  is the surface tension (N/m), while g is the acceleration by an empirical relationship
due to gravity, which can be taken as 9.8 m/s2 . l and v are the 1
densities of liquid and vapour, respectively. In the rest of the paper, sw = C  v2 , (14)
2 f l p
we will focus on one model which may be claimed as the state-of-
the-art, as it is based on the latest models [24,25], with a simplified where Cf ≈ 16/Re when Re = l vp di /l ≤ 1180; otherwise,
model for a system of liquid plugs and vapour bubbles as described
Cf ≈ 0.078Re−1/4 . (15)
in [37].
Here, we also used that vp = dxp /dt.
3.1.1. Temperature evolution The mathematical formulation can include many assumptions
The temperature Tv in a vapour bubble is governed by the energy and simplifications. When writing down the above equations, we
balance equation have tried to incorporate most the constitutive relationships such
dTv dV contact angles, capillary pressures and viscous force into the equa-
mv cvv = −hlfv (Tv − )L(di − 2ı) − rm hv L(di − 2ı) − pv , tions directly so that we can have as fewer equations as possible.
dt dt
(5) For details, readers can refer to [5,24,25]. This way, we can focus on
a few key equations such as the motion of a plug and temperature
where variations, which makes it more convenient for later mathematical
20 1 pv p analysis.
rm = √ (  − √l ). (6)
(2 − 0 ) 2R Tv 
3.2. Typical parameters
Here L is the mean plug length, di is the inner diameter of the pipe,
and ı is the thickness of the thin liquid film. h corresponds to the
The properties of the fluids and gas can be measured directly,
heat transfer coefficient and/or enthalpy in different terms. Here
and typical values can be summarized here, which can be relevant
 0 is a coefficient and R is the gas constant.
to this study. Most of these values have been drawn from earlier
In addition, the volume in the above equation is given by
studies [6,24,38].
di2 Typically, we have L = 0.18 m, di = 3.3 × 10−3 m, ı = 2.5 × 10−5 m.
V= (L + Lv ), (7) The initial temperature is Tv ≈  = 20 ◦ C, while the initial wall
4
temperature is Tw = 40 ◦ C. The initial pressure can be taken
where the length of the vapour bubble is typically Lv = 0.02 m, and
as pv ≈ 5.5816 × 104 Pa. hlfw = 1000 W/m2 K, hv = 10 W/m2 K.
 ≈ 1.
cvl = 1900 J/kg ◦ C, R = 8.31 and cvv = 1800J/kg ◦ C.
The temperature  in the liquid film is governed by
In addition, we have the initial values: mv0 = pv0 V0 /(Rv (Tv0 +
d 273.15) where Tv0 = 20 ◦ C, pv0 = 105 Pa, Rv = 461 J/kg K, and mf 0 ≈
mf cvl = −hlfw ( − Tw )Ldi + hlfv (Tv − )L(di − 2ı)
dt mv0 /10.
+ rm hv L(di − 2ı), (8)
4. Nondimensionalization, analysis and simulation
where Tw is the initial wall temperature.
After some straightforward mathematical simplifications, the
3.1.2. Mass transfer
full mathematical model can be written as the following equations:
For a vapour bubble, rate of change in mass is governed by the
conservation of mass mv cvv
dTv
= −hlfv (Tv − )L(di − 2ı) − rm hv L(di − 2ı)
dt
dmv −pv

[(di − 2ı)rm − di Lv rv ],
= −(di − 2ı)rm . (9) v
dt
d
As mv = v V , we have mf cvl = −hlfw ( − Tw )Ldi + hlfv (Tv − )L(di − 2ı) + rm hv L(di − 2ı),
dt
(16)
dV dmv
v = (di − 2ı)rm − rv di Lv , (10) = −(di − 2ı)rm ,
dt dt

where v ≈ 1 kg/m3 and the transfer rate rv = 0 if Tw > Tv . d2 xp  2

mp = (di − 2ı) (pvk − pvk ) − di Lp sw + mp g.
dt 2 4

3.1.3. Motion of a plug

The position xp of the liquid plug is governed by the momentum
Though analytically intractable, we can still solve this full model
equation
numerically using any suitable numerical methods such as finite
d2 xp  2 difference methods, and see how the system behaves under var-
mp = (d − 2ı) (pv1 − pv2 ) − di Lp sw + mp g, (11)
dt 2 4 i ious conditions with various values of parameters. Preliminary
where studies exist in the literature [6,24], and we have tested our sys-
tem that it can indeed reproduce these results using the typical
di2 parameters given in Section 3.2 [37]. For example, the mean tem-
mp = l (L0 − xp ), (12)
4 perature in a plug is consistent with the results in [41], while the
with L0 ≈ 25di and l = 1000 kg/m3 . Here the term pv1 − pv2 is the oscillatory behaviour is similar to the results obtained by others
differential pressure between the two sides of the plug. [5,12,25,40,41].

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
4 X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx

4.1. Non-dimensionalized model

For the current purpose of parameter estimation, the system can

be considered stationary under appropriate conditions, and thus
many terms can be taken as constants. Then, the full mathematical
model (16) can be non-dimensionalized and written as
dT v
u = a(Tv − ) − ˛1 − ˛2 ,
dt
d
= b(Tv − ) −  + ˛3 ,
dt
du (17)
=− ,
dt
  2 
d2 xp 1 dxp
= ˇ− ,
dt
2 (ˇ1 − ˇ2 xp ) dt

where
hlfv L(di − 2ı) rm hv L(di − 2ı)
a=− , ˛1 = , (18) Fig. 1. Comparison of the full numerical results with approximations and experi-
cvv cvv mental data [25].

pv  hlfv L(di − 2ı)

˛2 = [(d − 2ı)rm − di Lv rv ], b=− , (19) which provides a similar startup behaviour for small times as that
v cvv mf cvl
in [41].
hlfw Ldi rm hv L(di − 2ı) The approximation solutions are compared with the full numer-
= , ˛3 = , (20)
mf cvl mf cvl ical solutions and experimental data points, and they are all shown
2
in Fig. 1. Here the experimental data points were based on [5,24,25].
(di − 2ı)rm (di − 2ı) (pv1 − pv2 ) It is worth pointing that the approximation can indeed provide
= , ˇ= + g, (21)
mv0 4 some basic characteristics of the fundamental behaviour of the
l L0 di2 l di2 startup, and the full numerical results can also give a good indi-
di Lp Cv l
= , ˇ1 = , ˇ2 = . (22) cation of the final steady-state value.
2 4 4
This simplified model also confirms that some earlier studies
In the above formulation, u is the dimensionless form of mv , that using mainly xp as the dependent variable can indeed give good
is u = mv /mv0 . insight into the basic characteristics of the complex system. For
example, Wong et al. used an equivalent viscous damping system
4.2. Simplified model without considering the actual heat transfer system [27], and their
main governing equation can be written as
Under the assumptions of stationary conditions and xp /L0  1
(for example at the startup stage), the last equation is essentially d2 x dx
+a + b(k + t)x = 0, (28)
decoupled from the other equations. In this case, we have an dt 2 dt
approximate but simplified model where a, b, k are constants. This system has typical features of a vis-
d cous damping system. On the other hand, Yuan et al. used a system
= b(Tv − ) −  + ˛3 , with primarily a second-order nonlinear ODE for xp
dt
  2  (23)  2
d2 xp 1 dxp d2 xp 2C dxp 2g (pv1 − pv2 )
= ˇ− , + + xp = , (29)
2 ˇ1 dt dt 2 di dt L Ll
dt
where C is the friction coefficient. This system can have even richer
Now the second equation can be rewritten as
characteristics of oscillatory behaviours [40].
 2 In comparison with earlier simpler models, our simplified model
d2 xp dxp ˇ 
=A−B , A= , B= . (24) can produce even richer dynamic features of the system without
dt 2 dt ˇ1 ˇ1
using much complex mathematical analysis, thus provides a basis
For the natural condition xp = 0 at t = 0, the above equation has a for further realistic analysis and simulations.
short-time solution
1 √ 5. Inverse parameter identification and optimization
xp = ln[cosh( AB t)], (25)
B
The aim of an inverse problem is typically to estimate impor-
which is only valid when xp /L0 is small and/or t is small. Indeed, it
tant parameters of structures and materials, given observed data
gives some main characteristics of the startup behaviour as shown
which are often incomplete. The target is to minimize the differ-
as the dashed curve in Fig. 1.
ences between observations and predictions, which is in fact an
Furthermore, the first equation for the liquid plug temperature
optimization problem. To improve the quality of the estimates, we
is also decoupled from the PDE system. We have
have to combine a wide range of known information, including any
d prior knowledge of the structures, available data. To incorporate
= Q1 − Q2 , Q1 = bTv + ˛3 , Q2 = b + . (26)
dt all useful information and carry out the minimization, we have to
For an initial condition  = 0 at t = 0, we have the following solution deal with a multi-objective optimization problem. In the simplest
case, we have to deal with a nonlinear least-squares problem with
 = Q1 (1 − e−Q2 t ), (27) complex constraints [21,22].

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx 5

The constraints for inversion can include the realistic ranges of Table 1
Least-squares parameter estimation.
physical parameters, geometry of the structures, and others. The
constraints are typically nonlinear, and can be implicit or even Parameters Estimates True values
black-box functions. Under various complex constraints, we have L 0.15–0. 24 0.18 m
to deal with a nonlinear, constrained, global optimization problem. di 0.001–0.005 0.0033 m
In principle, we can then solve the formulated constrained prob- Tv 10–27 20 ◦ C
lem by any efficient optimization techniques [37,29]. However, Tw 20–49 40 ◦ C
pv 40–129 100 kPa
as the number of the degrees of freedom in inversion is typically
large, data are incomplete, and non-unique solutions or multiple
solutions may exist; therefore, metaheuristic algorithms are par- More sophisticated metaheuristic algorithms have the potential to
ticularly suitable. provide better solution strategies.
Though above inverse problems may be solved if they are
5.1. Parameter estimation as an inverse problem well-posed, however, for most inverse problems, there are many
challenging issues. First, data are often incomplete, which leads to
Inverse problems in heat transfer and other disciplines tend non-unique solutions; consequently, the solution techniques are
to find the best parameters of interest so as to minimize the dif- often problem-specific, such as Tikhonov regularization. Secondly,
ferences between the predicted results and observations. In the inverse problems are highly nonlinear and multimodal, and thus
simplest case, a linear inverse problem can be written as very difficult to solve. In addition, problems are often large-scale
with millions of degrees of freedom, and thus requires efficient
u = Kq + w, (30)
algorithms and techniques. Finally, many problems are NP-hard,
where u is the observed data with some noise w, q is the parameter and there is no efficient algorithms of polynomial time exists. In
to be estimated, and K is a linear operator or mapping, often writ- many of these cases, metaheuristic algorithms such as genetic algo-
ten as a matrix. An optimal solution q* should minimize the norm rithms and firefly algorithm could be the only alternative. In fact,
||u − Kq||. In case when K is rectangular, we have the best estimate metaheuristic algorithms are increasingly popular and powerful
−1 [28,37]. In the rest of this paper, we will use the firefly algorithm dis-
q∗ = (KT K) KT u. (31) cussed earlier to carry out the parameter estimations for pulsating
But in most cases, K is not invertible, thus we have to use heat pipes.
other techniques such as regularization. In reality, many inverse
problems are nonlinear and data are incomplete. Mathematically 5.2. Least-squares estimation
speaking, for a domain ˝ such as a structure with unknown but
true parameters q* (a vector of multiple parameters), the aim is Now we try to solve the following problem for parameter esti-
find the solution vector q so that the predicted values yi (i = 1, 2, mation: Suppose we measure the response (i.e., the location xp of a
. . ., n), based on a mathematical model y = ( x, p) for all x ∈ ˝, are plug) of a pulsating heat pipe, as the results presented in Fig. 1, can
close to the observed values d = (d1 , d2 , . . ., dn ) as possible. we estimate some key parameters using these results through our
The above inverse problem can be equivalently written as a simplified mathematical model? One way to deal with this prob-
generalized least-squares problem [21,37] lem is to consider it as a nonlinear, least-squares best-fit problem
as described earlier in Eq. (33).
min f = ||d − (x, q)||2 , (32) Let us focus on the key parameters L, Tv , Tw , di , and pv . When
we carry out the estimation using the least squares methods, we
or
 2
can only get crude estimates as shown in Table 1. Here we have
min [di − (xi , q)] , (33) used 25 data points to establish estimates for 5 key parameters.
i The wide variations of these parameters, though near the true val-
ues, suggest that there are insufficient conditions for the inverse
Obviously, this minimization problem is often subjected to a set
problem to have unique solutions. This can be attributed to the fac-
of complex constraints. For example, physical parameters must be
tors that the measured data points were sparse and incomplete,
within certain limits. Other physical and geometrical limits can also
no constraints were imposed on the ranges of the parameters, and
be written as nonlinear constraints. In general, this is equivalent to
additional constraints might be needed to make the inverse prob-
the following general nonlinear constrained optimization problem
lem well-posed.
min f (x, q, d) (34)
5.3. Constrained optimization by the firefly algorithm
subject to

hj (x, q) = 0, (j = 1, ..., J), gk (x, q) ≤ 0, (k = 1, ..., K). (35) In order to get better and unique solutions to the problem, to
increase of the number of measurements is not the best choice, as
where J and K are the numbers of equality and inequality con- experiments can be expensive and time-consuming. Even with the
straints, respectively. Here all the functions f, hj , and gk can be best and most accurate results, we may still be unable to get unique
nonlinear functions. This is a nonlinear, global optimization for q. In solutions. We have to make this problem well-posed by imposing
the case when the observations data are incomplete, thus the sys- enough conditions.
tem is under-determined, some regularization techniques such as First, we have to impose more stringent bounds/limits. Thus,
Tikhonov regularization are needed. Therefore, the main task now we apply L ∈ [0.15, 0.22] m, di ∈ [0.002, 0.004] m, Tv ∈ [15, 25] ◦ C,
is to find an optimal solution to approximate the true parameter Tw ∈ [35, 45] ◦ C, and pv ∈ [80, 120] kPa. Then, we have to minimize
set q*. the parameter variations, in addition to the residual sum of squares.
From the solution point of view, such optimization can be in Thus, we have
principle solved using any efficient optimization algorithm. How-
ever, as the number of free parameters tends to be very large,  2

5

and as the problem is nonlinear and possible multimodal, conven- Minimize [di − (xi , q)] + k2 , (36)
tional algorithms such as hill-climbing usually do not work well. i k=1

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
6 X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx

Table 2 References
Parameter estimation by constrained optimization.
[1] A. Akachi, Structure of microheat pipe, US Patent number 5,219,020 (1993).
Parameters Estimates (mean ± standard deviation) True values [2] H. Akachi, F. Polášek, Pulsating heat pipes, in: 5th Int. Heat Pipe Symposium,
L 0.17 ± 0.1 0.18 m Melbourne, Australia, November, 1996.
di 0.0032 ± 0.0004 0.0033 m [3] T. Apostolopoulos, A. Vlachos, Application of the firefly algorithm for solving
the economic emissions load dispatch problem, International Journal of Com-
Tv 20.4 ± 0.9 20 ◦ C
binatorics 2011 (2011), Article ID 523806 (online), Available at http://www.
Tw 39.2 ± 1.1 40 ◦ C
hindawi.com/journals/ijct/2011/523806.html (14.01.12).
pv 98 ± 8.8 100 kPa
[4] L. Cheng, T. Luan, W. Du, M. Xu, Heat transfer enhancement by flow-induced
vibration in heat exchangers, International Journal of Heat and Mass Transfer
52 (2009) 1053–1057.
[5] R.T. Dobson, T.M. Harms, Lumped parameter analysis of closed and open oscil-
where k2 is the variance of parameter k. Here we have 5 parameters
latory heat pipes, in: 11th Int. Heat Pipe Conf., Tokyo, 12–16 September,
to be estimated. Now we have a constrained optimization problem 1999.
with imposed simple bounds. [6] R.T. Dobson, Theoretical and experimental modelling of an open oscillatory
By using the firefly algorithm with n = 20,  = 1, ˇ = 1 and 5000 heat pipe including gravity, International Journal of Thermal Sciences 43 (2)
(2004) 113–119.
iterations, we can solve the above constrained optimization. We [7] A. Faghri, Heat Pipe Science and Technology, Taylor & Francis, New York, 1995.
have run the results 40 times so as to obtained meaningful statistics, [8] I. Fister, I. Fister Jr., X.S. Yang, J. Brest, Swarm and Evolutionary Computation. A
using the same set of 25 observed data points, and the results are Comprehensive Review of Firefly Algorithms, 2013 http://www.sciencedirect.
com/science/article/pii/S2210650213000461
summarized in Table 2. [9] A.H. Gandomi, X.S. Yang, A.H. Alavi, Mixed variable structural optimiza-
We can see clearly from the above results that unique parameter tion using firefly algorithm, Computers and Structures 89 (23–24) (2011)
estimates can be achieved if sufficient constraints can be imposed 2325–2336.
[10] A.H. Gandomi, X.S. Yang, S. Talatahari, A.H. Alavi, Firefly algorithm with chaos,
realistically, and the good quality estimates can be obtained even Communications in Nonlinear Science and Numerical Simulation 18 (1) (2013)
for a number of sparse measurements. This implies that it is fea- 89–98.
sible to estimate key parameters of a complex PHP system from [11] M. Groll, M. Schneider, V. Sartre, M.C. Zaghdoudi, M. Lallemand, Thermal control
of electronic equipment by heat pipes, Renue Générale de Thermique 37 (5)
experimental data using the efficient optimization algorithm and (1998) 323–352.
the correct mathematical models. [12] S. Khandekar, M. Schneider, M. Groll, Mathematical modelling of pulsating heat
pipes: state of the art and future challenges, in: Proc. 5th ISHMT-ASME Joint
Int. Conf. Heat & Mass Transfer, India, 2002, pp. 856–862.
6. Discussions and conclusions [13] T. Luan, L. Cheng, H.Z. Cao, Y. Qu, Effects of heat sources on heat transfer of
axially grooved heat pipe, Journal of Chemical Industry Engineering 4 (April)
(2007).
Though a pulsating heat pipe system can be hugely complex
[14] S. Maezawa, F. Sato, K. Gi, Chaotic dynamics of looped oscillating heat pipes,
as it involves multiphysics processes, we have shown that it is in: 6th Int. Heat Pipe Symposium, Chiang Mai, November, 2000.
possible to formulate a simplified mathematical model under one- [15] R.J. McGlen, R. Jachunck, S. Lin, Integrated thermal management techniques for
high power electronic devices, Applied Thermal Engineering 24 (8–9) (2004)
dimensional configuration, and such a model can still be capable
1143–1156.
to reproduce the fundamental characteristics of time-dependent [16] P. Naphon, S. Wongwises, S. Wiriyasart, On the thermal cooling of central
startup and motion of liquid plugs in the heat pipe. By using scaling processing unit of the PCs with vapor chamber, International Communications
variations and consequently non-dimensionalization, we have ana- in Heat and Mass Transfer 39 (8) (2012) 1165–1168.
[17] M. Rahmat, P. Hubert, Two-phase simulations of micro heat pipes, Computers
lysed the key parameters and processes that control the main heat & Fluids 39 (3) (2010) 451–460.
transfer process. Asymptotic analysis has enabled us to simplify [18] D. Reay, P. Kew, R. McGlen, Heat Pipes, Sixth ed., Elsevier, Amsterdam,
the model further concerning small-time and long-time behavior, 2014.
[19] W. Qu, H.B. Ma, Theoretical analysis of startup of a pulsating heat pipe, Inter-
so that we can identify the factors that affect the startup charac- national Journal of Heat and Mass Transfer 50 (2007) 2309–2316.
teristics. Consequently, we can use the simplified model to predict [20] P. Sakulchangsatjatai, P. Terdtoon, T. Wongratanaphisan, P. Kamonpet, M.
and then compare the locations of a plug with experimental data. Murakami, Operation modelling of closed-end and closed-loop oscillating heat
pipes at normal operating condition, Applied Thermal Engineering 24 (7)
The good comparison suggests that a simplified model can work (2004) 995–1008.
very well for a complex PHP system. [21] M. Sambridge, Geophysical inversion with a neighbourhood algorithm. I.
Even though we have very good results, however, there are still Search a parameter space, Geophysical Journal International 138 (1999)
479–494.
some significant differences between the full numerical model and
[22] M. Sambridge, K. Mosegaard, Monte Carlo methods in geophysical inverse prob-
experimental data, which can be attributed to unrealistic param- lems, Reviews of Geophysics 40 (2002), 3-1-29.
eter values, incomplete data, oversimplified approximations and [23] M.K. Sayadi, R. Ramezanian, N. Ghaffari-Nasab, A discrete firefly meta-heuristic
with local search for makespan minimization in permutation flow shop
unaccounted experimental settings. Further work can focus on the
scheduling problems, International Journal of Industrial Engineering Compu-
extension of the current model to 2D or 3D configuration with tations 1 (2010) 1–10.
realistic geometry and boundary conditions. [24] G. Swanepoel, A.B. Taylor, R.T. Dobson, Theoretical modelling of pulsating heat
By treating parameter estimation as an inverse problem and pipes, in: 6th International Heat Pipe Symposium, Chiang Mai, Thailand, 5–9
November, 2000.
subsequently a nonlinear constrained optimization problem, we [25] G. Swanepoel, Thermal Management of Hybrid Electrical Vehicles
have used the efficient metaheuristic algorithms such as firefly Using Heat Pipes (MSc thesis), University of Stellenbosch, South Africa,
algorithm to obtain estimates of some key parameters in a 1D PHP 2001.
[26] L.L. Vasiliev, Micro and miniature heat pipes—electronic component coolers,
system. We have demonstrated that a least-squares approach is Applied Thermal Engineering 28 (4) (2008) 266–273.
not sufficient to obtain accurate results because the inverse prob- [27] T.N. Wong, B.Y. Tong, S.M. Lim, K.T. Ooi, Theoretical modelling of pulsating
lem was not well-posed, with insufficient constraints. By imposing heat pipe, in: 11th Int. Heat Pipe Conf., Tokyo, 12–16 September, 1999, pp.
159–163.
extra proper constraints on parameter bounds and also minimizing [28] X.S. Yang, Nature-Inspired Metaheuristic Algorithms, Luniver Press, Bristol,
the possible variations as well as minimizing the best-fit errors, we 2008.
have obtained far more accurate estimates for the same five key [29] X.S. Yang, Engineering Optimization: An Introduction with Metaheuristic
Applications, John Wiley and Sons, USA, 2010.
parameters, and these parameter estimates are comparable with
[30] X.S. Yang, Firefly algorithms for multimodal optimization, stochastic algo-
their true values. rithms: foundations and applications (SAGA2009), Lecture Notes in Computer
There is no doubt that further improvement concerning the Sciences 5792 (2010) 169–178.
[31] X.S. Yang, A.H. Gandomi, Bat algorithm: a novel approach for global engineering
mathematical model and parameter estimation will provide further
optimization, Engineering Computations 29 (5) (2013) 464–483.
insight into the actual behaviour of a PHP system, and subsequently [32] X.S. Yang, Multiobjective firefly algorithm for continuous optimization, Engi-
allows us to design better and more energy-efficient PHP systems. neering with Computers 29 (2) (2013) 175–184.

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003
G Model
JOCS-249; No. of Pages 7 ARTICLE IN PRESS
X.-S. Yang et al. / Journal of Computational Science xxx (2014) xxx–xxx 7

[33] X.S. Yang, S. Deb, Two-stage eagle strategy with differential evolution, Interna- Mehmet Karamanoglu is Professor in Design Engineer-
tional Journal of Bio-Inspired Computation 4 (1) (2012) 1–5. ing at Middlesex University. He graduated with a BEng
[34] X.S. Yang, S. Deb, Engineering optimisation by cuckoo search, International degree in Mechanical Engineering and followed onto to
Journal of Mathematical Modelling and Numerical Optimisation 1 (2010) complete his PhD in numerical methods, supported by
330–343. British Aerospace. He is currently heading the department
[35] X.S. Yang, S.S. Hossein, A.H. Gandomi, Firefly algorithm for solving non-convex of Design Engineering and Mathematics. His expertise
economic dispatch problems with valve loading effect, Applied Soft Computing includes manufacturing automation, CAD, design engi-
12 (3) (2012) 1180–1186. neering, modelling and robotics. His research interests
[36] X.S. Yang, M. Karamanoglu, X.S. He, Multi-objective flower algorithm for opti- include numerical analysis, process simulation, design
mization, Procedia Computer Science 18 (1) (2013) 861–868. strategies and robotic systems.
[37] X.S. Yang, T. Luan, Modelling of a pulsating heat pipe and startup asymptotics,
Procedia Computer Science 9 (2012) 784–791.
[38] C.L. Yaws, Chemical Properties Handbook, McGraw-Hill, New York, 1999.
[39] J. Yeom, M.A. Shannon, Micro-Coolers, in: Y. Gianchandani, O. Tabata, H. Zappe
(Eds.), in: Comprehensive Microsystems, vol. 3, Elsevier, London, 2008, pp.
499–550. Tao Luan is Professor in Energy and Power Engineering at
[40] D. Yuan, W. Qu, T. Ma, Flow and heat transfer of liquid plug and neighboring Shandong University. He graduated from Shandong Uni-
vapour slugs in a pulsating heat pipe, International Journal of Heat and Mass versity, China, in 1983, and obtained his PhD from the
Transfer 53 (2010) 1260–1268. University of Leeds, UK, in 1997. He has served in both
[41] Y. Zhang, A. Faghri, Heat transfer in a pulsating heat pipe with open end, Inter- industry and research institutes for many years in China,
national Journal of Heat and Mass Transfer 45 (2002) 755–764. Europe and Canada. His main research areas are in the field
[42] V.V. Zhirnov, R.K. Cavin, J.A. Hutchby, G.I. Bourianoff, Limits to binary of heat transfer, combustion, energy utilisation, environ-
logic switch scaling—a Gedanken model, Proceedings of the IEEE 91 (2003) mental protection, and heat exchanger/condenser design.
1934–1939.
[43] Z.J. Zuo, M.T. North, L. Ray, Combined Pulsating and Capillary Heat Pile Mecha-
nism for Cooling of High Heat Flux Electronics, 2001 http://wwww.thermacore.
com

Xin-She Yang received his DPhil in Applied Mathematics Slawomir Koziel received the MSc and PhD degrees in
from the University of Oxford. He worked at Cambridge electronic engineering from Gdansk University of Tech-
University and then at National Physical Laboratory as a nology, Poland, in 1995 and 2000, respectively. He also
Senior Research Scientist. He is Reader in Modelling and received the MSc degrees in theoretical physics and in
Simulation at Middlesex University, UK and an Adjunct mathematics, in 2000 and 2002, respectively, as well as
Professor at Reykjavik University, Iceland. He is the IEEE the PhD in mathematics in 2003, from the University
CIS Task Force Chair on Business Intelligence and an of Gdansk, Poland. He is currently a Professor with the
Honourary Fellow of Australia Institute of High Energy School of Science and Engineering, Reykjavik University,
Materials. Iceland. His research interests include CAD and mod-
elling of microwave circuits, simulation-driven design,
surrogate-based optimization, evolutionary computation
and numerical analysis.

Please cite this article in press as: X.-S. Yang, et al., Mathematical modelling and parameter optimization of pulsating heat pipes, J.
Comput. Sci. (2014), http://dx.doi.org/10.1016/j.jocs.2013.12.003