Sunteți pe pagina 1din 12

Energy Conversion and Management 155 (2018) 218–229

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage:

1-D transient numerical modeling of counter-current two-phase stratified MARK

flow inside a medium temperature solar linear collector

R. López, A. Lecuona , R. Ventas, J. Nogueira
ITEA Research Group, Thermal Engineering and Fluids Department, Universidad Carlos III de Madrid, Madrid, Spain


Keywords: Two-phase gravity-driven and gravity-stratified flow regime inside a pipe, which is present in many engineering
1-D two-phase flows applications, is an attractive option for solar cooling/heating/power production using Rankine cycles, absorp-
Runge-Kutta methods tion cycles or any other thermodynamic application by means of vapor as working fluid. This paper offers a
Finite volume method numerical model of this flow configuration that copes with transient phenomena, like unsteadiness of solar
Quasi-homogeneous model
radiation, among others. The mathematical model consists of 1-D balance equations for mass and momentum for
Solar collectors
Direct steam production
both fluids and energy for both fluids and the wall of the pipe that absorbs the solar radiation. The model is
Gravity driven and stratified counter-current characterized by the fact that the area (or height) of the liquid layer is treated as a dependent variable forming
flow part of the solution. The numerical method consists in a finite volume staggered grid discretization of the
governing equations. Mass flow and liquid area are calculated with a semi-implicit pressure based method and
the transient terms are treated with the explicit first stage singly implicit Runge-Kutta (ESDIRK) method. The
calculation of the mass transfer rate from liquid to vapor is calculated iteratively by a guess-and-correct mass
transfer algorithm, specially developed for stratified flows. The results show the applicability and benefits of this
model for the not so well known counter-current stratified two-phase with evaporation/boiling. Additionally, the
performance of the mass transfer algorithm is discussed showing that it is monotonic decreasing and linearly

1. Introduction mixture, depending on the application and operating conditions re-

Solar energy is the most abundant energy resource, its advantages Nowadays with the applications mentioned above, the liquid oc-
over other energy sources are clear, no polluting emissions, renewable, cupies entirely the straight pipe, which is normally horizontal; however
widely available and no fuel cost. All that makes this kind of source a in some other applications such as petroleum transport, steam genera-
very interesting choice in the years to come. Solar energy systems are tion equipment, chemical processes, distillation and nuclear reactors
encountered in many different applications, such as power generation, the simultaneous flow of liquid and vapor is used, e. g. Newton and
heating, hot sanitary water preparation and refrigeration applications Behnia [2] and the pipe can be inclined. If the vapor velocity is high
only to mention a few. Many are the available technologies for solar enough an annular stratification can be established owing to the pre-
thermal energy production. Among these technologies, thermal con- valence of shear stresses over gravity force, Rohsenow, Hammet and
centrated solar technology stands out for its high flexibility, high Cho [3], Wallis [4] and Friedel [5] among others, with the so-called
thermal efficiency, and low cost, Jebasingh and Joselin Herbert [1]. liquid holdup. This flow layout has been much studied because it
Either parabolic trough or Fresnel type linear solar collectors are the happens in vertical pipes and also in horizontal and inclined setups.
most common. They work on the same principle and their layout is Usually, the pipe is completely filled with liquid at its inlet side and it is
similar for our purposes. The working principle is simple, a straight subjected to intense heating, releasing so much vapor that the vapor
absorber pipe is located along the focal line of either a single or a group velocity increases also very much downstream, especially at low pres-
of large area mirrors that concentrate solar radiation from below on its sures, driving the liquid flow by shear stresses. In those layouts, a
external surface, tracking the sun, Fig. 1. With the resulting high ra- complete transformation of liquid into vapor is pursued. This is not the
diation flux, a fluid stream that flows inside the straight pipe is heated. present case, where the velocities involved are low, making gravity
This fluid can be a pure substance, like water or thermal oil, or can be a predominant. This type of regime is a gravity stratified flow, which

Corresponding author.
E-mail address: (A. Lecuona).

0196-8904/ © 2017 Elsevier Ltd. All rights reserved.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Nomenclature ε fluid zone integer function

φ liquid cross angle
A cross section area θ angle coordinate
a butcher table of coefficient matrix ω pointing vector
B discretised linear system coefficient vector τ shear stress
b source term vector λ thermal conductivity
c butcher tabl of quadrature point location α convective heat transfer coefficient
D pipe diameter μ dynamic viscosity
G mass flux ν kinematic viscosity
h enthalpy Δ finite difference
hlv phase change enthalpy
K mass transfer coefficient Subscripts
ṁ mass flow rate
N number of finite volumes h hydraulic
n̂ unitary normal vector l liquid
Nu Nusselt number (= αDh / λ ) v vapor
P perimeter i inner
Pr Prandtl number (= Cpμ/ λ ) ip interphase
p pressure k zone index
r radius o output, outer
Re Reynolds number (= GDh / ν ) wall wall
S surface area 0 at z = 0
T temperature w,e,Ee,W ,P ,E cardinal coordinate system location index
t time
u velocity Superscripts
x quality (=(hl−hlsat )/ hlvsat )
z axial coordinate per unit length symbol, correction
′′ per unit area symbol
Greeks ∗ correction
sat saturation condition
β pipe tilting angle T transposed
ρ density (n) time level
δ flow vein cross height (s ) stage level

Fig. 1. Schematic representation of the

counter-current gravity stratified flow

excludes vertical and quasi-vertical pipes. It happens in horizontal and With this layout, the vapor is produced directly without intervening
inclined pipes, making the gravity force responsible for its stratification any intermediate heat carrying fluid, for that using pressurized water,
so that the heaviest phase occupies the bottom side, Rohsenow, steam or thermal oil, as customary. Cost, weight, and bulk are reduced
Hammet, and Cho [3], Wallis [4] and Taitel and Duckler [6]. As in our with direct steam production as well as system complexity. For many
case, for low enough velocities, a smooth interphase is maintained se- industrial processes, the temperatures involved are in the order of
parating two continuous phases, not even reaching the wavy regime. 120–250 °C (medium temperature solar collection, Coccia et al. [8]),
The moderate vapor production and the large void fraction makes that depending upon the pressure levels. In any case, these temperatures are
this regime does not switch to another one. The flow can be co-current lower than those required by steam generation for power production,
or counter-current depending upon the relative motion of both phases, around 400–500 °C, thus the pipe thermoelastic bending problem is of a
here each fluid stream defines a phase. If both phases flow in the same much lesser importance, e. g. [9,10]. With this layout, the flow is es-
direction, it is called co-current, otherwise is counter-current, Kandlikar tablished when a pure substance liquid (or multicomponent mixture for
[7], which is the layout of interest in this paper. absorption cycle applications), slides continuously, under the effect of

R. López et al. Energy Conversion and Management 155 (2018) 218–229

gravity, inside and inclined straight pipe of high slenderness, thus being dimensions much larger than surface tension curvatures. The descrip-
gravity driven. The inclination can be very different, up to around tion of the main phenomena occurring with this innovative solar col-
β = −50°, see Fig. 1, depending on the latitude of the location and on lecting layout with a mathematical model is of high interest as it allows
the inclination of the roof where the solar collector is installed. quantifying the effect of the design and operating variables on this
As a result of the heat produced by the sun rays absorbed at the rather unexplored architecture. It allows studying the sensitivity to the
external surface of the pipe, the liquid is heated, generating vapor, main parameters and also its extension to include further mechanisms,
which is collected on the upper side of the pipe and then is exhausted such as extended surfaces and flow retardants.
on the upper end. No dry out is considered in this paper. The remaining The pipe is of a constant circular cross section, as customary in
liquid is exhausted in the lower end of the pipe. This layout is shown linear solar collectors, making both flows a circular segment. This
schematically in Fig. 1. This double fluid exhaustion system can be used shape, although relatively well studied for turbulent flow friction and
for several applications; for example, the vapor produced can be con- heat transfer, has not been much described for evaporation and boiling.
densed and used as a refrigerant, as a product, or simply as steam for This work establishes a model in a progressive way by integrating into it
heat or power generation. In parallel, the exhausted liquid can be used the different refinements necessary to reach a correct and re-
as the hot fluid for any other thermal process required or even it can be presentative physical behavior.
recirculated. Actually, this boiler configuration adds the function of The heat conduction through the pipe wall determines the pipe
liquid/vapor separation in favor of compactness. This layout is original bending in the longitudinal z axis, Khanna, Singh and Kedare [17],
and no relevant open literature has been found on the topic, excepting what is important both for solar focusing and for structural integrity.
from the authors [11]. Considering the pipe wall as an extra phase, which is thermally coupled
As with any other technological development, the ultimate goal is to with both fluid ones, allows determining its cross-section temperature
build prototypes and later a precommercial series; therefore numerical distribution under a non-conventional flow regime. Its study will shed
modeling serves as an outstanding tool to perform optimisations before light on the possible structural problems. Moreover, the present study
materialisation, reducing drastically the overall cost of the development adds a more precise calculation than previously performed Lecuona,
and of the end product (design to minimize cost of maintenance and an Rosner, and Ventas [18].
easy operation, being the costs reduction evaluated in terms of CAPEX Within this scenario, the paper is organized as follows: In Sections 3
and OPEX). In that sense 1-D numerical models offer high flexibility, the basic geometry characteristics of stratified flow are presented. In
capturing all the relevant physics of the problem in a reasonable Section 4, the derivation of the mathematical model is explained jointly
computational time. Many researchers have developed 1-D numerical with the empirical information needed to close the system. In Section 5,
models for parabolic mirror trough solar collectors for at least two in order to obtain approximate numerical solutions, the discrete form of
decades up to date, taking advantage of the high slenderness of the pipe the equations are formulated, followed by the thermal and mass
and the accumulated experience on these models. Good examples of transfer coupling algorithms, specially developed for this work. In
overall energy balance formulations can be found in Kalogirou et al. Section 6 the numerical experiments performed in this paper are stated
[12], more realistic models based on differential balances of mass, aiming at validating the model. The first numerical experiment consists
momentum, and energy in transient-state can be found in Sivaram et al. in a steady-state fully adiabatic case, thus without mass transfer; the
[13], Nallusamy [14], Garcia- Valladares et al. [15] and Vázquez Pa- second test case consists in a transient thermal evolution of a start-up
dilla et al. [16], among others. Unfortunately, all the models mentioned operation problem. In Section 7, the results are shown and a compre-
above were developed for direct steam generators in which the pipe is hensive discussion is constructed around them with especial attention
completely filled with the entering fluid thus with an initial null void on the correct physics, on benefits of the present model and on the mass
fraction. No other model was found in the open literature capable of transfer algorithm performance. Finally, in Section 8, the conclusions of
simulating both gravity driven and stratified internal flow in a counter- this work are pointed out.
current layout with mass transfer between both phases. In the opinion
of the authors, the development of such a 1-D transient model seems
useful and relevant for future developments, Lecuona et al. [11]. 3. Gravity stratified counter-current flow

Under the assumptions described in the introduction, Section 1, one

2. Scope of the work can assume that in the pipe cross section a straight interphase is located
between liquid and vapor, but the height of the liquid vein δl can
The main interest of this research is the counter-current gravity change downstream. According to Fig. 2, some geometric relations can
stratified flow layout, as described in the introduction, Section 1, with a be found:
smooth interphase owing to the low velocities involved and the

Fig. 2. (a) Control volume, (b) cross section.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Al =
(2φ−sin2φ) Di2
(1) εk = {−11ififkk==l v (9)
1 Since the model is 1-D, heat fluxes are approximated by means of
Atotal = πDi2 = Al + Av
4 (2) the Newton cooling law, using average temperatures:

D ″ ,k = α wall,k (Twall,k−Tk )
Qwall (10)
φ = π −cos−1 ⎡ ⎛δl− i ⎞ (Di /2)⎤
⎣ ⎝ 2⎠ ⎦ (3)
″ k = α ip (Tip−Tk )
Qip, (11)
Alv = Disinφ (4)
Here α is the convective heat transfer coefficient, T is for absolute
temperature and the superscript (¯) refers to its surface mean value for
4. Mathematical formulation and discretised equations the wall:
∫φk,ki,f rTwall (φ)dθ
In this section, the mathematical formulation is presented. For this Twall,k = φk ,f
purpose, the entire domain is separated as three independent zones ri ∫φ dθ
k ,i (12)
(phases) which interact with each other, Fig. 1. These three zones are:
(i) The liquid zone, located at the bottom side of the pipe. Here the On the other hand, the interphase heat flux can be calculated as
heavy phase is located, which can be sub-cooled or saturated liquid. (ii) ″ = α ip,l (Tl−Tip) or Q ïp″ = α ip,v (Tip−Tv ) , depending upon which zone is
Q ïp
The vapor zone, located on the upper side of the pipe; this zone is where considered. In order to guarantee: temperature continuity between the
all the vapor generated in the liquid zone is collected and finally flows liquid zone and the vapor zone, and heat flux continuity in the inter-
away from the upper end of the pipe. (iii) The solid wall, it receives the phase Tip is defined as:
net heating of the radiation heat flux on its external surface, transmit- α ip,l Tl + α ip,v Tv
ting this heat to both flows through its internal surface. Tip =
α ip,l + α ip,v (13)
The assumptions for the flows are: both fluids are Newtonian, the
flow is 1-D in the pipe axis direction z , the flows are completely stra-
tified, gravity driven and separated by a continuous smooth interphase; 4.1. Empirical information
each substance is perfectly mixed in their respective zone, and mass
transfer from the liquid to the vapor zone is made under saturated The model formed by Eqs. (5)–(8) requires some empirical in-
conditions. The fluid velocities and the respective velocity between formation. For the momentum equations, the shear stresses are calcu-
them are small enough ∼ 1 ms−1 to guarantee that no significant ripples lated using the Darcy formula.
or waves are present in the interphase [6,19]. The pipe inner diameter 1
Di is in the order of some centimeters, so that surface tension is af- τw,k = f ρ Uk2
2 k k (14)
fecting only a tiny region near the wall. The vertical hydrostatic pres-
sure can be neglected in front of the main pressure imposed by the 1
τip = f ρ Uip2
vapor-liquid equilibrium, thus both streams share the same pressure. 2 ip ip (15)
For the solid wall, the assumptions are that the axial heat conduction is Here f is the Darcy friction factor, which takes into account the wall
negligible in front of circumferential one, all properties are constant roughness. Uip is the average (bulk) velocity difference between both
and finally, the heat flux by convection on the outer surface is taken phases in contact.
into account by a net solar radiation flux. Many correlations to calculate the friction factors f have been de-
The mathematical formulation is based on balances for mass and veloped, unfortunately all of them are for stratified co-current flows,
momentum for both flows and energy balance for them and for the solid therefore if we assume that the total amount of vapor generated in the
wall. The equations are applied to an initially differential element of liquid zone is low compared to the liquid mass flow, we can infer with a
length dz , according to Fig. 2. For such balances, 1st order Taylor series high level of certainty that the behavior of counter-current stratified
expansions are used, neglecting the higher order terms. The mass and flow under this condition is similar than the co-current stratified
momentum for both fluids (k = l or v ) and energy balance for both smooth pattern. In that sense, according to Andreussi and Persen [19],
fluids and the solid wall can be written in integral form as: the shear stress at the wall interphases can be calculated as it were a
single-phase flow, with the hydraulic diameter (Dh,k = 4Ak / Pk ) as the
∫z ∂∂t (ρk Ak )dz + ∫z ṁ k dz + εk ṁ ip′ = 0 (Mass) (5) characteristic length. For the liquid/vapor interphase, the shear stress is
highly important when relative velocity between both phases is high,
∫z ∂∂mṫ k dz + ∫z ⎛ ρ 1Ak ⎞ ṁ k2 dz = − ∫z pAk dz− ∫Sk τw,k dSk−εk ∫Sk τip dSip
⎜ ⎟
being responsible for the interfacial wave generation mechanism and
⎝ k ⎠ liquid holdup. Anyhow, the value for fip is again calculated by the
− ∫z ρk gAk sinβ dz (Momentum) criterion of Andreussi & Persen [19]. This criterion is based on their
observations, later confirmed independently by Hanratty and Andritsos
[20], in which they concluded that for stratified smooth pattern the
∫z ∂∂t (ρk Ak hk )dz + ∫z ṁ k hk dz + εk ṁ ip′ hv sat = Qwall
′′ ′′
,k + εk Qip,k (Energy−flows)
interfacial friction factor is approximately equal to the one of the gas
(7) flow ( fip = fv ). Some correction can be added because of the effusion
velocity by vapor transfer, but it seems to be quite low so that it has
∂Tw λ wall ∂Twall
(ρCp) w ∫V ∂t
″ −Qwall
dV = Qrad ″ ,l−Qwall
″ ,k + ∫Sθ r ∂θ
dA (Energy−wall) been neglected.
For the local heat transfer coefficients (both fluids/wall and liquid/
(8) solid interphase), the Gnielinski correlation [21] is used for the calcu-
The balance equations for both flows are similar, the only difference lation of the fully developed turbulent flow Nusselt number, Eq. (16). It
is in the direction in which mass, momentum, and energy are trans- is modified by the entrance correction of Al-Arabi [22], Eq. (17). Both
ferred from one zone to the other, hence the fluid zone integer function correlation and correction correlations can also be found in Sigalés
εk is introduced. This function defines the normal vector direction at the [23]. The solar heat flux reaches a maximum in the order of 10 kW/m2,
interphase, being oriented in the direction of the liquid to the vapor a low value for fired boilers. Consequently, in the present study, we
zone. This function is defined as: neglect the enhancement by bubbles formation at the wall, Eq. (28);

R. López et al. Energy Conversion and Management 155 (2018) 218–229

this way being conservative. Momentum ⎫

Be,k = max[−ṁ w,k ;0] KP,k + max[ṁ e,k ;0] KE ,k + Sp ⎪
αDh 8
(Re−1,000) Pr B w,k = −max[ṁ w,k ;0] KP,k BEe,k = −max[−ṁ e,k ;0] KE ,k ⎪
Nu∞ = = ⎪
λ 1+ 12.7 f /8 (Pr 2/3−1) (16) 1
∼(j) ⎪
(s ) f
bk = (PP AP −PE AE )k + a ∑ asj F k + Sc + Sc ⎪
j=1 ⎪
Nu (z ) D 0.8 Swall,k

= 1 + 0.56 ⎛ h ⎞ ⇒ Nu (z ) Δz ∗
Sp = a Δt + fk 4ρ A 2 sign(ṁ e,k ) ṁ e,k ∗
Nu∞ ⎝ z ⎠ ss k k

Δz ⎪
D 0.8
(Re−1,000) Pr Sc = a Δt ṁ e(,nk)
= ⎡1 + 0.56 ⎛ h ⎞ ⎤ 8 ⎪
⎢ ⎝ z ⎠ ⎥ 1 + 12.7 f /8 (Pr 2/3−1) (17)
S w ,k
Scf = fk 8ρ A 2 sign(ṁ e∗,k )(ṁ e∗,k )2−ρk gAk sin(β )Δz ⎪
⎣ ⎦ ⎪
k k
∼(j) ⎪
(j ) (j ) (j ) (j ) (j )
F k = −Bi,k ui,k − ∑ Bnb,k unb,k + bi,k ⎪
nb ⎪
5. Discretised equations nb ≠ i ⎪ (18)

The model developed in the last section consists of non-linear PDEs. Energy(Flows) ⎫
As these equations do not have analytical solutions, the use of numer- BP,k = max[−ṁ w,k ;0] + max[ṁ e,k ;0] + Sp ⎪
ical methods is mandatory. In this work, the system of equations is ⎪
BW ,k = −max[ṁ w,k ;0]BE ,k = −max[−ṁ e,k ;0] ⎪
solved numerically using the finite volume method (FVM). Refs. s−1 ⎪
1 ∼ (j ) ⎪
[24–26] offer more detailed information over FVM. Four coupled grids bk = a ∑ asj F k −εk ṁ ip h vsat + Sc0 + ScQ
ss ⎪
are used to transform the continuum model into a discrete algebraic j=1

system. The first two grids (flows) consist of a staggered grid arrange- Δz ⎪
Sp = a Δt ρk(s) Ak(s)
ment; here scalar variables Ø (pressure, enthalpy, area, vapor quality, ss ⎬
Δz ⎪
etc.) are collocated at the center of the volume and the mass flow rate at Sc0 = a Δt ρk(s) Ak(s) hk(s)
ss ⎪
the boundaries, Fig. 3a. The integration of the momentum equations is (n) (n) (s ) (s ) (n) (n) (s ) (s ) ⎪
ScQ = α wall ,kAP ,k (T wall,k−TP ,k ) + εk α ip A ip (T ip −Tk ,P ) ⎪
made on the contrary, Fig. 3b. The third grid consists of a collocated
∼(j) ⎪
grid for the wet surface average wall temperature Fig. 3a–c. This tem- F k = −Bi(,jk) ui(,kj)− ∑ Bnb (j ) (j ) (j )
,k unb,k + bi,k ⎪
perature is collocated at the center of the volume and in the same axis nb ⎪
nb ≠ i ⎪ (19)
of the scalar variables in the fluid main grid. Finally, an auxiliary grid is ⎭
also used for the collocation of the wall temperatures in order to cal-
Energy(Wall) ⎫
culate angular wall heat conduction. This temperature is also at the
BP,k = (ρCp) w a
λw Δr ⎪
center of each volume Fig. 3d. ss Δt r Δθ ⎪
The dependent variable vector is W = ( ṁ l ṁ v hl h v Twall p Al )T . λ Δr λ Δr ⎪
BW ,k = − rwΔθ BE ,k = − rwΔθ ⎪
The first five variables are determined with the momentum and energy
s−1 ⎪
equations of each fluid and the wall respectively. The other two vari- 1 ∼(j) ⎪
bk = Sc0 + ScQ + a ∑ asj F k
ables, pressure and liquid cross section area, are calculated with a ss ⎪
j=1 ⎪
pressure correction method. Δ V
Sc0 = a Δt (ρCP ) w Ti(n) ⎬
For W = ( ṁ l ṁ v hl h v Twall )T the balance equations are integrated ss ⎪
(space and time) over the staggered flow grid for ṁ k , the staggered grid ScQ = πDO ΔzQrad ″ −αl(,wn) Al(,Pn) (Tl,z−Tl,P )−α v(,nw) Av(,nP) (Tv,z−Tv,P ) ⎪

main grid for hk and the auxiliary angular wall temperature grid for ∼(j) ⎪
F P = −B(Pj) uP(j)− ∑ Bnb (j ) (j )
unb + bnb(j )
Twall . To perform this integration a 1st order upwind scheme is used for ⎪
nb ⎪
the non-linear terms due to convection; also the non-linear friction nb ≠ P ⎪
terms are approximated with a 1st order Taylor series expansion. For all ⎪
ΔV = r Δr ΔθΔz ⎭ (20)
the equations the transient terms are evaluated using the Explicit First
Stage Diagonally implicit Runge-Kutta method (ESDIRK), [27,28]. With The two equations left are for mass conservation of both fluids,
these simplifications the model written in the discrete form which integrated over the staggered main grid, in semi-implicit form,
Be / P We / P + B w / W Ww / W + BEe / E WEe / E = b is: results as follows:

Fig. 3. Meshes: (a) main fluid staggered grid (scalar variable Ø), (b) fluid staggered grid, (c) auxiliary wet surface average grid, (d) auxiliary angular wall temperature grid.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

∂ 5.1. Flow and thermal algorithms

ṁ e,k −ṁ w,k = − (ρ Ak )Δz−εk ṁ P,ip
∂t k (21)
The 1-D flow transient discrete model consists of momentum and
Here the mass accumulation term can be approximated under the
energy Eqs. (18), (19) and (20) plus the pressure/liquid area correction
ESDIRK fashion at every (s) stage as:
Eqs. (23) and (24). The system of discrete equations is solved with the
(s ) TDMA algorithm taking advantage of the form of the coefficient ma-
∂ (ρk Ak )(s)−(ρk Ak )n
(ρ Ak ) ≈ trices. The numerical methodology consists in solving the corre-
∂t k P cs Δt (22)
sponding system of equations for each zone in a sequential manner in
This equation acts more of a restriction to the mass flow field than the following order: liquid, vapor, and wall. Because all variables are
as a dynamic equation for pressure and liquid area, therefore a pres- highly coupled the entire system of balance equations must be solved
sure/liquid cross area evolution equation must be constructed. Focusing iteratively on each (s) ESDIRK stage inside the interval t + Δt . To ensure
on both, liquid and vapor zones momentum equations, the driving force convergence a tolerance value must be settled, and this convergence
is given by pAk , so the same procedure to build this evolution equation criterion is applied to the mean average wall and flows interphase
applies for pressure ( p ) in the vapor zone and for the liquid area ( Al ) in temperatures. The details of the algorithm are shown in Fig. 4.
the liquid zone. This is an old issue in incompressible flows giving birth
to the so-called semi-implicit-pressure-linked-equation (SIMPLE) algo- 5.2. Mass transfer correction algorithm
rithm, Patankar [29]. The pressure/liquid area correction method starts
with a guessed mass flow field (ṁ k∗), and both mass flow fields are At this point, the discrete conservation and pressure/liquid area
corrected on each iteration as ṁ = ṁ k∗ + ṁ k′. The mass flow guess field correction equations are solved with the mass transfer rate at the in-
is calculated with a guessed pressure/liquid area field p = p∗ + p′ and terphase (ṁ ip ) as a parameter. Mass transfer through the interphase is
Al = Al∗ + Al′, With these corrections, equations combined with the controlled by many factors such as: shape of the interphase, interphase
discrete Eq. (18) and mass conservation Eq. (21) the pressure/liquid surface tension, molecular diffusion, turbulent mixing, plus convection
area correction on each (s ) level can be built. and concentration gradient in multicomponent mixtures. All this in-
(s ) formation is here modeled through the so-called mass transfer coeffi-

(dv,e + dv,w ) P′P −dv,e P′E −dv,w P′W = −(ṁ v∗,e−ṁ v∗,w )− (ρ Av ) Δz −ṁ P,ip cient (K ); hence, if we assume that all the vapor is generated by the
∂t v P addition of latent heat and all this vapor is transferred instantly to the
(23) vapor phase (K = 1), the mass transfer rate can be modeled as
(s )
ṁ ip = xṁ . Particularized in discrete form in our integration stencil, the
∂ mass transfer rate finally becomes.
(dl,e + dl,w ) A′l,P −dl,e A′l,E −dl,w A′l,W = −(ṁ l∗,e−ṁ l∗,w )− (ρ Al∗,P ) Δz + ṁ P,ip
∂t l P
0 if xP,l ⩽ 0 (sub−cooled state)
(24) ṁ P,ip = ⎧
In the above equations, the coefficients dk,e are obtained by algebraic ⎩ ̇ w,l xP,l if xP,l > 0 (liquid + vapor mixture)
⎨ m (25)
manipulation of the discrete momentum equation. More details can be Application of Eq. (25) to calculate ṁ P,ip requires to know ṁ w,l , hl
found in Patankar [24,29]. and p on each (s) stage, however, mass transfer rate in strongly coupled

Fig. 4. Flowchart of thermal fluid for (s)

stage ESDIRK algorithm.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

with mass flow, enthalpy and pressure, therefore Eqs. (18), (19), (20), are derived using the prescribed pressure and vapor flow at both ends,
(24) and (25) must be solved sequentially in an iterative fashion. Un- in [31] some more details can be found. The boundary conditions for
fortunately Eq. (25) is not a dynamic equation for ṁ P,ip , so the way to the Poisson equation in the liquid there they are not straightforward so
fulfill the correct mass transfer rate is by a correction numerical pro- that a more detailed insight must be reserved for. One of the assump-
cedure, using Eq. (26) tions is that the flow is gravity driven, this means that the inlet flow is
accelerated by gravity until an equilibrium terminal velocity is reached.
ṁ ip = ṁ ip∗∗ + Δṁ ip (26) This transition from the inlet is supposed to occurs in a very small re-
Here ṁ ip∗∗ stands for the guess mass flow through the interphase field gion, thus at the inlet, the proper BC is of the Dirichlet type in mass
and Δṁ ip stands for its correction, which is calculated after every flow. For the outlet, to ensure that the flow is fully developed a Neu-
iteration level with Eq. (25) logically using the value of quality calcu- mann BC for the driving force ∂ (pAl )/ ∂z = ∅ is needed. At first sight,
lated with enthalpies of the same iteration level.The idea behind this these BC’s are enough, but both BC’s do not produce a unique solution
correction procedure is that by successive iterations the mass transfer of the liquid area Poisson equation. The reason of the non-uniqueness is
rate is adjusted, such that the generated vapor is instantly transferred because the Dirichlet liquid mass flow condition is numerically
and the liquid stream is always in saturation conditions. The mass equivalent to a Neumann liquid area condition, thus the liquid area is
transfer correction algorithm for the transient regime under the ESDIRK determined excepting adding a constant; in other words, a family of
method is shown in Fig. 5. solutions for Al is found on every iteration level. To avoid this problem,
the boundary condition at the inlet is transformed into a Dirichlet type
6. Numerical experiments condition. This change of the natural condition into a coherent math-
ematical boundary condition makes that the liquid area at the inlet
Verification and validation both conform the essential procedure to must be iterated in a steady-state preconditioning algorithm. According
determine the reliability and accuracy of CFD codes and models [30]. to the previous discussion, the correct and mathematical coherent
The way to proceed is simple, with the aim of test cases as a benchmark, boundary conditions for vapor pressure and liquid area Poisson equa-
comparative simulations are performed. Unfortunately, there is a lack tions are:
of experimental data or analytical solution to be used as a benchmark
for the counter-current flows with mass transfer between phases, a Liquid ⎫
Al (z = 0,t ) = Al,0 ⎪
layout that is of interest in this paper. For this reason, two numerical
−ṁ ip,z = L ṁ l,z = L ⎬
experiments will be carried out. First of all, a simple steady-state pro- (∂pAl / ∂z )|(z = L,t ) = ϕ = Δz (ρ A )|
l l (z = L,t ) ⎭
⎪ (29)
blem with no heat and mass transfer is solved in order to show the
reliability of the numerical model. Secondly, a simple conceptual
transient test case is proposed as benchmark hoping that in the near
future experiments can be carried out and serve as a basis for a required
fully comparison and from the results not only a further validation re-
sults, but the fine-tuning of the empirical data used.
Both experiments share the same geometrical and initial data: The
pipe of the collector is 6 m long and is made of commercial steel,
λ wall = 15 W m−2 K−1, ρwall = 7900 kg m−3 , and Cpwall = 500 J kg−1 K−1
with an inner diameter Di = 30 mm and 5 mm wall thickness. The inlet
temperatures of the liquid and vapor streams are respectively 118.21 °C
(sub-cooled water), 125 °C and 30 °C for the wall, so that a transient
effect will be evident when starting. Pressure for the liquid and vapor
are equal to 200 kPa. The initial mass flow rates for the liquid and vapor
are 0.01 kg s−1 and 0 kg s−1 respectively inlet and for this case ∀z. The
inclination angle of the pipe is set to β = −30°.
For the second test case besides the data mentioned above other
inputs are required. Solar net irradiance is a function of the angle of the
pipe, Eq. (27). This function is for these cases the same for any z po-
sition at any time.

500 W m−2 for 330° ⩽ θ < 210°

″ =⎧
Qrad −2

⎩ 800 W m for 210° ⩽ θ < 330° (27)

This solar irradiance is concentrated with a parabolic device, such

that the total net heat flux at the outer diameter is, including losses to

500 W m−2 for 330° ⩽ θ < 210°

″ =⎧
Qrad −2

⎩ 7,680 W m for 210° ⩽ θ < 330° (28)

6.1. Boundary, initial conditions, and numerical parameters

To well pose the problem, initial and boundary conditions must be

settled. According to our test cases, the initial conditions are:
pl (t = 0,z ) = pv (t = 0,z ) , Tl (t = 0,z ) = 118,21 °C, Tv (t = 0,z ) = 125 °C
and Twall (t = 0,z ,θ) = 30 °C. Additionally, the solution method requires
solving an intermediate pressure/liquid area Poisson equation. For the
Fig. 5. Flows interphase mass transfer algorithm.
pressure Poisson equation in the vapor zone, the boundary conditions

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Vapor ⎫ 6.3.1. Grid sensitivity analysis

ṁ v (z = L,t ) = 0 kg s−1 In order to guarantee the consistency of the numerical model pro-
⎬ posed in this paper, with the initial and boundary conditions explained
pv (z = 0,t ) = 200 kPa ⎭ (30)
in Section 6.1, a grid sensitivity analysis has been carried out before the
Here ϕ stands for the equilibrium force between the pressure-area simulation results are presented in the next sub- section, now in-
driving force and the momentum transfer from the liquid to the vapor corporating mass transfer. As three different grids are used, the analysis
zone. It is important to notice that ϕ = 0 implies that no mass is being is carried out in two ways. First, the number of volumes in the wall is
transferred ṁ ip = 0 . This term is small compared to other terms of the fixed (Nw = 90 ) varying the volumes of both flows (N ) and finally, the
momentum equation and can be neglected for practical purposes. same procedure is repeated in an opposite manner, fixing the number of
In addition, some numerical parameters are also required. In this flows volumes and varying the wall ones. The selected tracking vari-
work, the temporal integration is performed with the 3rd order of ac- ables are the vapor enthalpy and mass flow, both measured at the vapor
curacy ESDIRK method developed by Alexander [32]. This temporal outlet (z = L ). Fig. 7 shows the temporal evolution of these variables. It
numerical scheme is implicit, making the numerical method to be un- can be observed that the grid refinement makes that the solution tends
conditionally stable, not requiring any restriction on the time step. asymptotically to a stable solution while N → ∞ or Δz → 0 . This be-
There is no rapid variation in the boundary conditions, so no rapid havior evidences the consistency of the numerical schemes; in other
transient phenomena will occur, hence the time step was set to words, the numerical fluxes that cross the volumes boundaries tend to
Δt = 0.25 s . The total simulation time is 120 s, a time long enough to the real numerical flux as Δ z→ 0 . This result is not surprising due to
reach the steady-state condition. The relaxation factor for pressure the fact that the employed 1st order UPWIND scheme is consistent. On
chosen is constant and set to 0.5. On the other hand, the relaxation the other hand, in the case of vapor mass flow of Fig. 7a, it can also be
factor for liquid area correction equation is inversely proportional to observed that the solutions for N = 50 and N = 100 volumes are very
the number of volumes employed; in our numerical experiments, it was close with an error respect to the solution with N = 150 of −0.067%
varied from 0.05 for N = 150 to 0.25 for N = 20 . The selection of the and of −0.051% respectively. A similar result is observed for the en-
liquid area relaxation factor is an important issue in this numerical thalpy evolution, Fig. 7b. From an asymptotic point of view, both so-
model. The reason behind this unusual small value for the area cor- lutions are acceptable. When the number of volumes is increased the
rection equation is because the liquid area must be always a positive computational effort increases. This behavior is explained not only
quantity and the use of the area correction equation makes that the because the dimension of the linear systems to solve is larger, but also
intermediate solution can oscillate around the solution point at any because of the relaxation factors for the liquid area correction Eq. (24)
iteration level, bringing the risk of negative liquid areas. must be lower for a larger number of volumes, see Section 6.1 for the
definition of the relaxation factors.
For the next sensitivity experiment carried out, the number of flow
6.2. Test case #1: Steady-state test
volumes is settled to N = 50 and the influence of the number of wall
volumes Nw is analyzed. Fig. 8 shows the results. They indicate that for
Stratified gravity driven flows are characterized by an initial hy-
a low number of volumes Nw < 20 a spurious high vapor enthalpy ap-
drodynamic development. Once the fluid enters the pipe, gravity acts as
pears for large times; this occurs because for a low number of volumes
the main driven force accelerating or decelerating the flow to a terminal
the mean wall temperatures, calculated with Eq. (13), are not well
velocity leading to a certain liquid area. In the physical modeling used
approximated, so that in this case the vapor mean wall temperature is
in this paper this transition zone is neglected, thus for this particular
high in excess, transferring more heat to the vapor zone. With the grid
case the liquid area is constant along the pipe. The purpose of this first
refinement, the solution tends asymptotically to a stable solution, but
case is to verify that the liquid area calculated with the numerical
again the computational cost increases, in this case, when Nw > 48. It
model reproduces the stratified gravity driven behavior. An analytical
also can be observed that respectively for 50 s< t < 100 s and t > 100 s
expression can be derived with the aim of the momentum and mass
a decrease and later an increase in vapor enthalpy is found; this effect is
conservation equations for the liquid zone, Eqs. (5) and (6) for k = l .
because the initial wall temperature is lower than the initial vapor
According to the assumptions it can be supposed that ∂ (pAl )/ ∂z = 0 ∀ z .
temperature (Section 6), therefore heat is transferred from the vapor
Correspondingly, the momentum equation can be reduced to an ana-
zone to the wall for t < 100 s and later on the heat flux is reversed when
lytic expression where Ul is the terminal velocity:
the mean vapor wall temperature equals that of the vapor zone.
1 According to this analysis, N = 50 (Δz = 0.03 m ) for the liquid and
fl Sw,l (Al ) fl Sw,l (Al ) ṁ l2 ⎤3
Al (z ) = Ul2 = ⎡
⎢ 2 ⎥
vapor zones and Nw = 48 volumes (Δθ = 0.1308 rad ) for the wall at each
2g sinβ ⎣ 2ρl g sinβ ⎦ (31) z location are enough to ensure an asymptotically consistent solution.

With this simple equation one can calculate the liquid area in terms
of the mass flow. Eq. (31) is implicit, so it must be solved iteratively.
Several simulations were carried out for different angles of inclination
of the pipe varying the mass flow. Fig. 6 shows the comparison of the
numerical solution with the theoretical solution for the steady state test
case #1. The numerical solution is for the mass flow and liquid area at
the inlet and the theoretical solution is found solving Eq. (31). In can be
observed that both results coincide for different angles of inclination.
With this test case, it can be ensured that the numerical model jointly
with the boundary conditions reflects satisfactorily the stratified gravity
driven steady-state hydrodynamics.

6.3. Test case #2: Transient start-up operation

As stated in the beginning of this section, a transient test case has

been performed. Before showing the numerical results it is necessary to
Fig. 6. Theoretical and numerical solution comparison for the steady-state test case #1.
check the consistency in a grid sensibility analysis.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Fig. 7. Grid sensitivity for Nw . fixed and varying N for two tracking variables, (a) vapor mass flow rate, (b) vapor enthalpy.

6.3.2. Transient numerical solution generated and instantly is transferred to the upper vapor zone. Even
Once the number of volumes is settled, the spatial distribution is that initial conditions for temperature were constant for both liquid and
analyzed in this section. Fig. 9 shows the results for liquid temperature. wall, the way in which they evolve thermally downstream is not uni-
For t < 90 s there is the washout of the cold entering liquid with a form, the reason of this is that the convective coefficient αl is not
thermal wave traveling downstream. This phenomenon is caused by the constant along the pipe, as Fig. 10b shows, mainly due to the entrance
initial wall temperature that was set at a lower than the initial liquid correction in the Nusselt number, Eq. (17). This non-uniformity affects
temperature, hence heat is transferred from the liquid to the wall and the vapor generation and can be clearly observed in Fig. 10a as a bulge
therefore the wall is heated not only by sun radiation but by the liquid shaped curve for ṁ ip , especially for 50 s< t < 120 s .
heat as well, until its large heat capacity allows a temperature gain. One of the novelties incorporated in the model is the simultaneous
This is coherent with the results shown in Fig. 9, which shows that for calculation of the wall temperature dynamics. Fig. 11 shows the wall
t = 40 s , the pipe wall is still cold well downstream. This kind of be- peripheral temperature distribution at four different locations for
havior is an important aspect to take into account if an analysis of t = 40 s . It can be observed how the lower part of the pipe is hotter than
startup or shutdown of the installation has to be carried out. Later on, the upper part, mainly near the inlet of the pipe in spite of the increased
for example at t > 50 s , in some sections of the pipe near the liquid inlet, heat transfer near the liquid inlet, also it shows the heating of the upper
the wall temperature is recovered, switching the direction of heat flux. part by thermal diffusion, affecting this way the efficiency of the solar
Once the heat flux is towards the liquid, it begins to increase its tem- system and eventually superheating the vapor. The full description of
perature until the saturation condition is reached at the inlet pressure of wall temperature distribution allows studying problems with non-
200 kPa. It is important to highlight that as part of the numerical al- symmetric solar angular concentration that happens when tracking the
gorithm the pressure inside the pipe is maintained constant until vapor sun along the day. Relevant information can be extracted for extended
is generated, this is because pressure is determined by the coupling of analysis like the determination of pipe thermal fatigue and for the
the momentum and mass equations in the vapor zone; hence this cou- evaluation of pipe longitudinal bending.
pling is not performed until there is vapor flow on top; this is to avoid
having zeros in the main diagonals of the matrices involved in the
SIMPLE algorithm for the vapor zone. The exact subcooled pressure 7. Performance of the mass transfer algorithm
downstream gradient would be slightly positive owing to the counter-
current vapor flow. The subcooled region corresponds to the almost The mass transfer algorithm developed for this work is a key part of
linear temperature increase wit z near the liquid inlet, more clearly the numerical procedure. It allows logically determining the mass
shown in Fig. 9b, as only sensible heating is produced. transfer rate and also to fulfill the here assumed saturation conditions of
Once the saturation condition is locally reached, vapor begins to be the liquid stream. This is the reason why some attention must be paid to
its performance. Fig. 12 shows the mass transfer correction (Δṁ ip ) for

Fig. 8. Grid sensitivity for N fixed varying Nw , (a) vapor mass flow rate, (b) vapor enthalpy.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Fig. 9. (a) Liquid temperature transient evolution, (b) close-up of (a).

Fig. 10. Transient evolution, (a) mass transfer through the flows interphase ṁ ip , (b) wall/liquid convective coefficient.

Fig. 11. Angular wall temperature distribution, t = 60 s .

different locations at t = 54 s , the instant at which vapor generation has same for all z points and moreover, Δṁ ip approaches zero mono-
already started, coherently with Figs. 9 and 10. Inside the interval tonically decreasing. Hence, it can be argued that the algorithm is al-
0.5 m< z < 1.8 m is where saturation is reached and the mass transfer ways convergent; furthermore, if one calculates the convergence se-
algorithm operates; outside this interval, the fluid is still sub-cooled and quence as Δṁ ipk + 1/Δṁ ipk , where k is the iteration level, we found that this
no mass correction is needed because x < 0 . It can be observed from sequence converges to 0.921, 0.904 and 0.909 for z = 0.541 m ,
Fig. 12a how the mass transfer correction is larger in the first iteration z = 0.78 m and z = 1.5 m respectively. This means that the mass
levels. After the iteration level advances, the mass correction mono- transfer corrections are linearly convergent with a mean order of con-
tonically decreases toward zero. This means that the energy transfer vergence of 0.901. These results evidence the potential of using this
through the interphase source term εk ṁ ip h vsat is being recalculated for algorithm for the determination of transient mass transfer rates in
accuracy. In other words, the algorithm is transferring the correct en- stratified evaporation/boiling flows.
ergy flux so that the liquid stream remains saturated. It could be ob-
served that at all locations the mass transfer correction exhibits the 8. Conclusions
same convergence performance. Fig. 12b shows a sample, again for
t = 54 s . It can also be observed that the convergence rate is almost the A 1-D numerical model has been developed for the transient

R. López et al. Energy Conversion and Management 155 (2018) 218–229

Fig. 12. Mass transfer algorithm performance, (a) mass transfer correction performance, (b) convergence historical.

simulation of heat and mass transfer in counter-current gravity driven [3] Rohsenow WM, Harnett JP, Cho YI. Handbook of heat transfer. 3rd ed. New York:
McGraw-Hill; 1998. ISBN 0-07-053555-8.
and gravity stratified liquid and vapor flows inside the receiver pipe of [4] Wallis GB. One-dimensional two-phase flow. New York: McGraw-Hill; 1969.
a concentration linear solar collector, aiming at direct vapor production [5] Friedel L. Dimensionless relationship for the friction pressure drop in pipes during
and separation. It has been performed using well-known methodologies two-phase flow of water and of R 12. Verfahrenstechnik 1979;13(4):241–6.
[6] Taitel Y, Duckler AE. A model for predicting flow regime transitions in horizontal
and implementing further advances over current thermal 1-D modeling. and nearly horizontal gas-liquid flow. AIChE J 1976;22:47–55.
The test cases used for checking the algorithms show all the phe- 10.1002/aic.690220105.
nomena involved in the correct way and are coherent, demonstrating [7] Kandlikar S, Shoji M, Vijay D. Handbook of phase change: boiling and condensation.
Taylor & Francis; 1999.
good performance.
[8] Coccia G, Di Nicola G, Giovanni H, Hidalgo A. Parabolic trough collector prototypes
The method of solution allows in the entire domain the calculation for low-temperature process heat. Springer; 2016. ISBN 978-3-319-27084-5.
of pressure, liquid cross area (or height), mass flow rates and tem- [9] Alguacil M, Prieto C, Rodriguez A, Lohr J. Generation in parabolic trough collectors.
Energy Proc 2014;49:21–9.
perature for both streams (liquid and vapor). Additionally, this method
[10] Serrano-Aguilera J, Valenzuela L, Parras L. Thermal 3D model for direct solar steam
determines axial and traverse distributions of the pipe wall temperature generation under superheated conditions. Appl Energy 2014;132:370–82. http://
as well as mass transfer between both phases. With this model, the
transient performance of solar linear receivers can be modeled with [11] Lecuona A, Ventas R, Vereda C, Legrand M. Linear tube solar receiver as stratified
flow vapor generator/separator for absorption machines using NH3/LiNO3. In: de
only a reduced set of input data. This tool can be of help to perform 11th EUROSUN 2016 ISES conference proceedings, Palma (Mallorca), Spain; 2016.
even more realistic analysis for the design and optimization of this kind [12] Kalogirou SA. A detailed thermal model of a parabolic trough collector receiver.
of devices. Energy 2012;48:298–306.
[13] Sivaram PM, Nallusamy N, Suresh M. Experimental and numerical investigation on
The counter-current layout makes the treatment of boundary con- solar parabolic trough collector integrated with thermal energy storage unit. Int J
ditions specific, shedding light on similar problems. Energy Res 2016;40:1564–75.
One relevant feature of this modeling methodology comes from a [14] Nallusamy N, Sivaram PM, Suresh M. Numerical modelling of solar parabolic
trough receiver employed for thermal energy storage system. J Clean Technol
numerical point of view: the mass transfer algorithm presents an out- 2017;5(2).
standing performance; the correction to be applied is monotonically [15] Garcia- Valladares O, Velázquez N. Numerical simulation of parabolic trough solar
decreasing when iterating and it is lower bounded. This makes the al- collector: improvement using counter flow concentric circular heat exchangers. Int
Heat Mass Transfer 2009(52):597–609.
gorithm linearly convergent with a calculated mean order of con- ijheatmasstransfer.2008.08.004.
vergence of 0.901 for the problem proposed. [16] Vazquez Padilla R, Demirkaya G, Yogi GD, Elias S, Rahman MM. Heat transfer
The model can be improved by adding a correction on the radiant analysis of parabolic trough solar receiver. Appl Energy 2011;88:5097–110. http://
flux absorbed by modeling the heat losses to ambient, depending on the
[17] Khanna S, Singh S, Kedare SB. Effect of angle of incidence of sun rays on the
optical and thermal layout external to the pipe. Further refinement is bending of absorber tube of solar parabolic trough concentrator. Energy Proc
necessary to include the low vapor head losses in the liquid subcooled 2014;48:123–9.
zone. More precise heat transfer coefficients that eventually include [18] Lecuona A, Rosner M, Ventas R. Transversal temperature profiles of two-phase
stratified flow in the receiver tube of a solar linear concentrator. Simplified analysis.
subcooled and saturated bubble formation could improve the accuracy In: 11th ISES EuroSun conference EUROSUN 2016, Palma de Mallorca, Spain,
of the model, especially for higher optical concentrations of the sun October 13–16.
irradiance. Multicomponent working fluids and no saturation condition [19] Andreussi P. Stratified gas-liquid flow in downwardly pipes. Int J Multiph Flow
can be implemented in the model. A purposely performed experimental [20] Hanratty TJ, Andritsos N. Effect of pipe diameter on stratified flow in horizontal
data set would allow to tune the method and add definite validation. pipes; 1984.
[21] Gnielinski V. New equations for heat and mass transfer in turbulent pipe and
channel flow. Int Chem Eng 1976;16(2):359–68.
Acknowledgements [22] Al-Arabi M. Turbulent heat transfer in the entrance region of a tube. Heat Transfer
Eng 1982;3.
This work has been partially funded by the Ministry of Economy and [23] Sigalés B. Transferencia de Calor Técnica, Barcelona. Spain: Editorial Reverté, S.A;
Competitiveness of Spain through the project ENE2013-45015-R, [24] Patankar S. Numerical heat transfer and fluid flow. McGrawHill; 1980.
whose help is much appreciated. [25] Ferziger JH, Peric M. Computational methods for fluid dynamics. 3rd ed. Springer;
[26] Versteeg HK, Malasekera W. An introduction to computational fluid dynamics: the
finite volume method. Longman Group; 1995.
[27] Butcher JC. Implicit RUNGE-KUTTA processes. Math Comp 1964;18:50–64. doi:
[1] Jebasingh VK, Joselin Herbert GM. A review of solar parabolic trough collector. stable/2003405.
Renew Sustain Energy Rev 2016(54):1085–91. [28] Ascher UM, Ruuth S, Spiteri R. Implicit-explicit Runge-Kutta methods for time-
2015.10.043. dependent Partial Differential Equations. Appl Numer Math 1997;25:151–67.
[2] Newton CH, Behnia M. A numerical model of stratified wavy gas–liquid pipe flow.
Chem Eng Sci 2001(56):6851–61. [29] Ashok K. Application of Runge-Kutta method for the solution of non-linear partial
353. differential equations. Appl Math Model 1977;1(4):199–204.

R. López et al. Energy Conversion and Management 155 (2018) 218–229

1016/0307-904X(77)90006-3. Further reading

[30] Patankar SV, Spalding DB. A calculation procedure for heat, mass and momentum
transfer in three-dimensional parabolic flows. Int J Heat Mass Transfer [33] Alexander R. Design and implementation of DRIK integrators for stiff systems. Appl
1972;15:1787–806. Numer Math 2003;46:1–17.
[31] Oberkampf WL, Trucano TG. Verification and validation in computational fluid [34] Dekker K, Verwer JG. Stability of Runge-Kutta Methods for stiff nonlinear differ-
dynamics. Prog Aerospace Sci 2002;38:209–72. ential equations. Math Comput 1986;47(176):753–5.
0421(02)00005-2. [35] Churchill SW. Friction factors equations spans all fluid regimes. Chem Eng
[32] Lopez R, Lecuona A, Nogueira J, Vereda C. Numerical solution of one-dimensional 1977;84:91–102.
transient, two-phase flows with temporal fully implicit high order schemes: sub- [36] Johanessen T. A theoretical solution of the lockhart and martinelli flow model for
cooled boiling in pipes. Nucl Eng Des 2017. calculating two-phase pressure drop and holdup. Int J Multiph Flow
nucengdes.2016.12.027. 1972;15:1443–9.