Accepted Manuscript

Modeling slack flow in hydraulic pipelines by means of a consistent

thermodynamic theory

Alexandre Hastenreiter Assumpção, Maria Laura Martins-Costa, Felipe

Bastos de Freitas Rachid, Rogério Martins Saldanha da Gama

PII: S0020-7462(17)30147-6
DOI: http://dx.doi.org/10.1016/j.ijnonlinmec.2017.05.015
Reference: NLM 2855

To appear in: International Journal of Non-Linear Mechanics

Received date : 24 February 2017

Revised date : 28 May 2017
Accepted date : 30 May 2017

Please cite this article as: A.H. Assumpção, M.L. Martins-Costa, F.B. Freitas Rachid, R.M.
Saldanha da Gama, Modeling slack flow in hydraulic pipelines by means of a consistent
thermodynamic theory, International Journal of Non-Linear Mechanics (2017),
http://dx.doi.org/10.1016/j.ijnonlinmec.2017.05.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
Modeling slack flow in hydraulic pipelines by means of a
consistent thermodynamic theory
Alexandre Hastenreiter Assumpção1
Maria Laura Martins-Costa2
Felipe Bastos de Freitas Rachid3
Mechanical Engineering Graduate Program (PGMEC), Department of Mechanical
Engineering (TEM), Universidade Federal Fluminense, Rua Passo da Pátria, 156,
Niterói, RJ, 24210-240 - Brazil

Rogério Martins Saldanha da Gama

Mechanical Engineering Graduate Program (FEN), Universidade do Estado do Rio de
Janeiro - Rua São Francisco Xavier, 524, Rio de Janeiro, RJ, 20550-013 - Brazil

Abstract

This work presents a consistent thermodynamic model to describe cavitat-

ing flows in hydraulic pipelines. Although the model is capable to describe
the vaporous cavitation phenomenon in unsteady as well as steady regimes,
the application presented herein is restricted to slack flow condition, which
takes place under steady state. The flow is assumed to be homogeneous
and isothermal and the fluid is regarded as a pseudo-mixture, comprising
the liquid, its vapor and an inert gas. The constituents are assumed to be
compressible and to coexist at every material point and time instant. The
balance equations of mass for each constituent are considered in the model,

1
Graduate Student.
2
Laboratory of Theoretical and Applied Mechanics (LMTA).
3
Corresponding Author. E-mail: rachid@vm.uff.br or rachidfelipe@gmail.com.
Tel./Fax: (+55 21) 2629-5468.

Preprint submitted to International Journal of Non-linear Mechanics May 28, 2017

along with one balance equation of momentum for the mixture as a whole, all
of them within the one-dimensional context. The existence of a small amount
of inert gas along with the vapor ensures the necessary thermodynamic con-
dition to describe the opening and closing of the cavity. It also allows the
description of the vaporous cavitation phenomenon in any region of the fluid
flow domain by means of a same set of equations. The phase change trans-
formation is properly accounted for as an irreversible process. The obtained
results by means of simple numerical simulations in steady state show that
model is capable to coherently describe the continuous opening and closure
of the cavities as long as the flow does not become sonic.
Keywords: Vaporous cavitation, inert gas, free gas, irreversible phase
change, thermodynamics of irreversible processes.

1. Introduction

Vaporous cavitation is the formation of vapor in a liquid as a result of

a pressure drop below the vapor pressure of the liquid at a constant tem-
perature approximately. It is present in several relevant and quite distinct
applications in engineering and so is a broad filed of research (Arndt , 1981;
Utturkar et al. , 2005). Due to its undesired consequences, vaporous cavita-
tion plays an important role in the design and operation of liquid pipelines,
both in unsteady and steady regimes, and therefore has been the subject
of intense theoretical and experimental studies in the past decades. For a
broad and updated review one should see Bergant et al. (2006) and refer-
ences therein.
Pipelines used for liquid transmission have lengths several orders of mag-

2
nitude greater than their diameters, therefore fluid flow analyses are com-
monly described by means of long-wave length theories (Lighthill , 1978;
Wylie and Streeter , 1993) giving rise to one-dimensional models. Because
vaporous cavitation is a localized phenomenon which takes place at discrete
and small regions of the fluid flow domain, it is customarily approached as
a single pressure homogeneous model, in which the liquid and vapor con-
stituents share the same velocity, instead of using one-dimensional two-fluid
flow models (Drew and Passman, 1999). It is within this context that va-
porous cavitation in pipelines has been extensively studied in unsteady as
well as in steady states (Wylie, 1984; Nicholas , 1995; Kessal and Amaouche
, 2001; Shu , 2003; Liu et al. , 2004; Berg et al. , 2005; Xie et al. , 2006;
Assumpção and Freitas Rachid, 2010; Sumam et al. , 2010; Ferrari , 2010;
Sadafi et al. , 2012; Jiang et al. , 2012; Yuanyuan et al. , 2012; Soares et al.
, 2015) .
In unsteady regimes, vaporous cavitation is induced by the passage of
low pressure waves generated by fluid transients. Typical sources of tran-
sients include, but are not limited to, opening and closing of valves as well as
start-up and shut-down of pumps, which are actions inherent to either line
operation or power failure. When the vapor cavities increase and agglomer-
ate to an extent capable to induce the liquid column separation, devastating
effects may arise due to the very high pressures which result of the rejoining
of the liquid columns, such as piping rupture and damage to system appurte-
nances. Such a kind of problem may be severely intensified when the piping
structure interacts with the fluid motion given rise to the important fluid-
structure interaction analyses (Tijsseling et al. , 1996; Tijsseling and Vardy

3
, 2005).
In steady state, vaporous cavitation can cause a two-phase flow pattern
of stratified nature along some stretches of the line, where the liquid flows
through the bottom of the pipe in free fall whereas the vapor recirculates
on its top (Nicholas , 1995; Stanley , 2005). A sketch of such a pattern is
shown in the cross-section view A-A in Fig. 1 where ”L” and ”G” denote the
liquid and the gaseous phases, respectively. Such a flow pattern is known in
the technical literature as slack line flow and is most likely to take place in
pipelines that lie in hilly terrains with steep slopes. To get a better picture of
this flow phenomenon, it is worthwhile to report to the (modified) piezometric
head H ∗ and elevation z profiles of a typical pipeline as a function of its
developed length s, as shown in Fig. 1. The modified piezometric head is
defined as,
p − psv
H ∗ := + z, (1)
ρl g
where p and psv = p̂sv (T ) stand, respectively, for the local pressure and the
saturated vapor pressure at the local temperature T , ρl is the liquid mass
density and g represents the local gravity acceleration. The piezometric head
defined by Eq. (1) is called modified with regard to the classical definition,
which is computed without subtracting from the pressure the saturated vapor
pressure of the liquid.
When the pipeline operates in steady state and the pressure in the flowing
liquid is everywhere greater than its vapor pressure, the head remains above
the pipeline elevation profile along its total extension L and the line is said
to operate under tight line flow condition. The hydraulic grade line (HGL)
corresponding to this condition is drawn in Fig. 1 with a dashed red line. In

4
contrast, if the pressure in the fluid falls below the vapor pressure, then vapor
pockets are formed in one or more line stretches as shown in the segments
B-C and D-E of the pipeline (see Fig. 1), giving rise to slack line flow. In such
segments, the head gets closer to the elevation and, consequently, the HGL
assumes the form depicted with the red solid line. Within these segments the
wavefront speed in the mixture is drastically reduced and the mixture velocity
accelerates and decelerates to compensate for the mass density variation as
a result of the phase change process.
In contrast to the unsteady state, there is no catastrophic event arising
from vaporous cavitation under steady state as long as the slack line regions
do not collapse abruptly. However, it is an unwanted event from the oper-
ational viewpoint of oil pipelines for which the existence of leak detection
systems is mandatory. Since slack line flow regions may spread over signif-
icant extensions of the pipeline, it may substantially alter the linefill and
the transit-time associated with transient events due to the reduction of the
wavefront speed (Stanley , 2005; Nicholas , 1995). On the other hand, if
the linefill and the wavefront speed are not accurately evaluated, the leak
detection may be severely compromised. The time required to detect a leak
may significantly increase. Small leaks may not be detected anymore and
the number of false alarms may intensify, as compared with the operation
under tight line flow condition. Thus, it becomes particularly important to
accurately predict when and where the cavity pockets will be formed as well
as their extensions, so as to ensure an appropriate performance of the leak
detection system (Schraml et al. , 2011; Yuanyuan et al. , 2012).
In spite of its relevance, this subject has received little attention in the

5
open literature in the past years. As a result, there are only few works
concerning this issue, all of them focusing on the implications of the slack
line flow on the leak detection or on the way the line should be operated
to circumvent such a problem rather than on the development of models for
properly describing cavitating flows under steady conditions (Stanley , 2005;
Schraml et al. , 2011; Yuanyuan et al. , 2012).
On the other hand, because of the catastrophic damage caused by fluid
transients, there is a great number of models about unsteady cavitating flows,
such as those proposed by Wylie (1984); Kessal and Amaouche (2001); Shu
(2003); Liu et al. (2004); Berg et al. (2005); Xie et al. (2006); Ferrari
(2010); Sumam et al. (2010); Sadafi et al. (2012); Jiang et al. (2012); Soares
et al. (2015). Curiously, a model that is capable to reasonably describe the
vaporous cavitation in unsteady regime is not necessarily able to do it in
steady state. Examples of such models are the Discrete and the Generalized
Discrete Vapor Cavitation Models (DVCM and GDCVM) (Bergant et al.
, 2006; Sadafi et al. , 2012; Soares et al. , 2015). Moreover, although
the physics behind the vapor cavitation stems on thermodynamic features,
only a few number of the existing models take it into account (Kessal and
Amaouche , 2001; Shu , 2003; Liu et al. , 2004; Berg et al. , 2005; Ferrari
, 2010; Jiang et al. , 2012). However, these models admit as a simplifying
assumption that the phase change process is reversible, what is equivalent to
assume that the vapor pressure remains constant and equal to the saturated
vapor pressure during the phase change transformations. Such assumption
may underestimate the overall dissipation in the medium and, consequently,
erroneously predict the attenuation and timing of the pressure waves.

6
 Aiming to bridge the gap between the vaporous cavitation modeling under
steady and unsteady states as well as the one associated with the thermo-
dynamics background, this work presents a one-dimensional homogeneous
model based on the thermodynamics of irreversible processes to describe the
cavitation phenomenon, whatever the flow regime is. Instead of admitting
only the existence of the liquid and its vapor in the mixture, a free inert
gas is taken into account as an additional constituent. Since it is assumed
that the inert gas coexists along with the vapor throughout the flow domain,
the fluid is now regarded as a two-constituent two-phase mixture. Therefore,
besides the momentum conservation equation for the mixture as a whole,
mass conservation equations for the liquid, the vapor and the gas are consid-
ered. To properly describe the wall friction in both slack flow and tight flow
conditions a suitable modification is introduced in the friction factor. The
temperature is supposed to be the same for the three constituents and the
cavitation process is assumed to be an isothermal irreversible transformation.
The macroscopic dissipative effect of the liquid-vapor transformation, which
drives the cavity growth and collapse, is properly accounted for in such a
way that the second law of thermodynamics is always satisfied.
To access the capability of the proposed model in properly describing
the opening and closure of a cavity in slack line flows, numerical simulations
are carried out in steady state under distinct topographic conditions. To
do so, an air-water mixture flowing in a simple pipeline arrangement made
of two segments of straight pipes which go sequentially uphill and downhill
is considered. By changing the downhill inclination of the second segment,
distinct elevation profiles are generated rendering different conditions for the

7
formation of a vapor cavity in the second pipe segment. The obtained results
demonstrate that the model is able to properly report important physical fea-
tures, such as: the description of the phase change as an irreversible process,
its implication on the wavefront speed in the medium and its capability in
assessing the rate of energy dissipation associated with the vaporous cavita-
tion.

2. Basic assumptions and balance equations

Because the pipeline length L is several order of magnitudes greater than

its diameter D, fluid flow in pipelines is assumed to be one-dimensional, be-
ing the fields, in the Eulerian description, expressed in terms of the spatial
coordinate s measured along the pipe centerline and the current time instant
t. Within this context and in contrast to two-phase fluid flows, in which
the phases can assume very different geometrical configurations along the
pipeline (Ishii and Hibiki , 2006), the cavitation in liquid pipelines is a local-
ized phenomenon which takes place at discrete and relative small stretches
of the line. Therefore, it is reasonable to assume that there exists no sig-
nificant relative motion between the liquid and gaseous phases, so that the
constituents are assumed to share the same velocity. In addition, if we as-
sume that both phases also share the same temperature, to properly describe
the fluid flow it suffices to consider the balance equations of mass for each
constituent, along with the balance equation of momentum and the entropy
inequality for the mixture as a whole.
Under suitable regularity assumptions, and also by restricting the anal-
ysis to isothermal transformations and undeformable pipes, the following

8
equations in the Eulerian description, for any inertial frame of reference, are
sufficient to fully describe the one-dimensional fluid flow through a pipe that
makes an angle θ with the horizontal, with −90o ≤ θ = θ̂(s) ≤ 90o , (Germain
and Muller , 1995):
∂ ∂
(αρg ) + (αρg v) = 0, (2)
∂t ∂s
∂ ∂
(αρv ) + (αρv v) = Γ, (3)
∂t ∂s
∂ ∂
((1 − α)ρl ) + ((1 − α)ρl v) = −Γ, (4)
∂t ∂s
∂ ∂ 4
ρv 2 + p = − τ − ρgsinθ,


(ρv) + (5)
∂t ∂s D
∂v 4
d := −Ψ̇ − (Ψ + p) + τ v ≥ 0, (6)
∂s D
for all (s, t) ∈ (0, L) × (0, +∞). In the above equations, v stands for the
mixture relative velocity field with respect to the pipe , p denotes the mixture
thermodynamic pressure, τ is the wall shear stress acting on the fluid and
g represents the local acceleration of the gravity. As it has been implicitly
assumed in Eqs. (2-6), the fluid is regarded a pseudo-mixture of a two-phase,
two-constituent homogeneous mixture with average properties. The gaseous
phase is formed by the vapor of a liquid and a free inert gas, whereas the
liquid phase is made up of liquid solely. The liquid and gaseous phases are
assumed to coexist at every material point and time instant. To take it into
account, it is considered by an internal variable α, α ∈ (0, 1), usually called
gaseous volume fraction, which is defined as being the ratio of the volume
occupied by the vapor along with the free inert gas and the total volume of
the mixture. As a result, the mass density of the mixture is expressed as:

ρ = (1 − α) ρl + α (ρv + ρg ) , (7)

9
in which ρl , ρv and ρg stand for the mass densities of the liquid, vapor and
free inert gas, respectively.
As usual, the superimposed dot stands for the time material derivative
and Ψ denotes the Helmholtz free energy per unit volume. The rate of
mass exchange per unit volume between the gaseous and liquid phases as a
result of the vaporous cavitation is represented by Γ. If vapor is transformed
into liquid, then Γ < 0. On the other hand, if liquid becomes vapor, then
Γ > 0. Finally, if there is no mass transfer at all, then Γ = 0 and the
mass conservation principles associated with the liquid and the vapor become
independent from each other.
Equations (2), (3) and (4) express the mass conservation principle for
the free inert gas, the vapor and for the liquid. Equation (5) expresses the
momentum conservation principle for the mixture as a whole and Eq. (6)
is a local version of the second law of the thermodynamics (SLT), usually
known as the Clausis-Duhen inequality, which states that the rate of energy
dissipation per unit volume, d, must be non-negative. It distinguishes the
possible processes (d ≥ 0) from the impossible ones (d < 0). The possible
processes are classed as reversible as long as d = 0 and irreversible whenever
d > 0.
To complete the problem description, constitutive relationships must be
provided. They encompass expressions for τ , p and Γ in such a way that
Eq. (6) be satisfied no matter the external actions, the initial and the bound-
ary conditions.

10
3. Constitutive theory

The constitutive relations describing the macroscopic mechanical behav-

ior of the liquid-vapor-gas mixture are derived in the framework of the ther-
modynamics of irreversible processes. In this theory, once the local state of
the material has been characterized by means of an appropriate choice of
the set of state variables, two thermodynamic potentials - the Helmholtz free
energy and a pseudo-potential of dissipation - are sufficient to derive a com-
plete set of constitutive equations. For this particular problem, we choose
as state variables the local mass densities of the liquid ρl , the inert gas ρg
and the vapor ρv , the gaseous volume fraction α and the absolute tempera-
ture T . As we shall see, the restriction associated with α is treated in this
work as a physical property in the constitutive equations. This approach
has already been used by Frémond and Nicolas (1990) in the modeling of
the sorption-desorption problem in saturated porous media and by Freitas
Rachid (2003); Silva and Freitas Rachid (2013); Freitas Rachid (2014) and
Freitas Rachid (2006) in the modeling of cavitation in flows of compressible
fluids and of surface tension effects in homogeneous liquid-gas flows.

3.1. Helmholtz free energy and state laws

The free energy per unit volume Ψ is supposed to be a function of the

state variables ρl , ρg , ρv , α and T . Since the fluid is regarded as a mixture,
its behavior is supposed to comprise a combination of the liquid, the inert gas
and the vapor thermo-mechanical properties, taking α as a weighting factor.
Thus, we choose

Ψ (ρl , ρg , ρv , α, T ) := Ψ0 (ρl , ρg , ρv , α, T ) + I(α), (8)

11
in which

Ψ0 (ρl , ρg , ρv , α, T ) := (1 − α)ρl Ψl (ρl , T ) + αρg Ψg (ρg , T ) + αρv Ψv (ρv , T ), (9)

(
0 , if α ∈ (0, 1)
I(α) := (10)
+∞ , otherwise.
In Eqs. (9) and (10), Ψ0 is a smooth function which describes the thermo-
mechanical properties of the fluid whereas I(α) represents the indicator func-
tion of the convex set (0, 1) (Moreau et al. , 1988). The term I(α) is the
non-smooth parcel of the free energy and is included to take the internal
constraint α ∈ (0, 1) into account as a constitutive assumption. In other
words, it prevents α from getting out of its admissible interval since it would
be required an infinite amount of energy to do it.
The terms Ψl , Ψg and Ψv represent the free energies per unit mass of
the liquid, the gas and the vapor, respectively. As suggested by its func-
tional dependence, these free energies are supposed to represent the thermo-
mechanical behavior of the liquid, the gas and the vapor as if they were single
constituents. According to the theory presented in Freitas Rachid (2014),
the following forms for these free energies are assumed:
   o  
T 2 ρl ρl
Ψl (ρl , T ) := −ClT log + al + log , (11)
T ∗ ρl ρ∗l
   
T 2 ρg
Ψg (ρg , T ) := −Cg T log + ag log , (12)
T ∗ ρ∗g
   
T 2 ρv
Ψv (ρv , T ) := −Cv T log + av log , (13)
T ∗ ρ∗v

in which ρ∗l , ρ∗g and ρ∗v along with T ∗ represent reference states for the liquid,
the inert gas and the vapor, respectively.

12
 In the above expressions, Cl , Cg and Cv represent the specific heats at
constant volume of the liquid, the inert gas and the vapor, respectively, which
are assumed to be constants. The material parameters ρol , a2l , a2g and a2v are
temperature dependent, being the last three the square of the isothermal
wavefront propagation velocities in the pure liquid, the inert gas and the
vapor, respectively.
The state laws for the fluid, relating the reversible components of the ther-
modynamic forces to the state variables, are obtained from the free energy
potential and defined as follows:

%l ∂Ψ0 pl
B := = (1 − α)(Ψl + ) = (1 − α)gl , (14)
∂ρl ρl
∂Ψ 0
pg
B %g := = α(Ψg + ) = αgg , (15)
∂%g ρg
∂Ψ 0
pv
B %v := = α(Ψv + ) = αgv , (16)
∂%v ρv
∂Ψ 0
B α := + h, with h ∈ ∂α I(α), (17)
∂α

in which;

∂Ψ0
= ρg Ψg + ρv Ψv − ρl Ψl , (18)
∂α
∂Ψl
pl := ρ2l = a2l (ρl − ρol ) , (19)
∂ρl
∂Ψg
pg := ρ2g = a2g ρg , (20)
∂ρg
∂Ψv
pv := ρ2v = a2v ρv . (21)
∂ρv

In the above equations, gl , gg and gv represent the Gibbs specific free

energies of the liquid, the gas and the vapor whereas pl , pg and pv stand for
the liquid, the gas and the vapor partial pressures, respectively. The term

13
h in Eq. (17) is an element of the sub-differential set ∂α I(α) (also called
generalized derivative) with respect to α of the convex function I(α). The
sub-differential of the indicator function I(α) at α is given by the set (Ekeland
and Teman, 1976; Moreau et al. , 1988);

∂α I(α) := {h ∈ IR|I(α? ) − I(α) ≥ h(α? − α); ∀α? ∈ (0, 1)}. (22)

A straightforward calculation shows that ∂α I(0 < α < 1) = {0} and ∂α I(α) =
∅ if α ∈
/ (0, 1). The restriction α ∈ (0, 1) is effectively taken into account
through the constitutive law Eq. (17), since this relation implies that the
sub-differential ∂α I(α) is not empty.
Keeping in mind that time derivatives must be left derivatives in order to
cope with the principle of determinism, it comes out that the material time
derivative of Ψ which appears in Eq. (6) can be written for α ∈ (0, 1) as:
∂Ψ0 ∂Ψ0 ∂Ψ0 ∂Ψ0
Ψ̇ = ρ̇l + ρ̇g + ρ̇v + α̇. (23)
∂ρl ∂ρg ∂ρv ∂α
When the above result is used in Eq. (6) along with the state laws
Eqs. (14-17), Eq. (6) can be rewritten as:
∂v 4
d := − (Ψ + p) − B ρl ρ̇l − B ρg ρ̇g − B ρv ρ̇v − B α α̇ + τ v ≥ 0. (24)
∂s D
To obtain a complete set of constitutive equations, it suffices to specify a
pseudo-potential of dissipation from which complementary laws are derived
in such a way that the local version of the SLT given by Eq. (6) or (24) is
always verified.

3.2. Pseudo-potential of dissipation and complementary laws

To introduce the irreversible behavior of the pseudo-mixture, and also
to ensure that the SLT is always satisfied, we assume that there exists a

14
pseudo-potential of dissipation Φ, which is an objective, convex and dif-
ferentiable function of B Γ and v. Moreover, the pseudo-potential Φ =
Φ(B Γ , v; ρl, ρg , ρv , α, T ) is assumed to have the following properties:

Φ(B Γ , v; ρl, ρg , ρv , α, T ) ≥ 0 and Φ(0, 0; ρl , ρg , ρv , α, T ) = 0, (25)

whatever the values of B Γ and v.

The additional information associated with the dissipative behavior of
the fluid is obtained from Φ through the following complementary laws:

∂Φ
Γ := , (26)
∂B Γ
∂Φ
τ := . (27)
∂v

In addition, if the rate of energy dissipation d is supposed to have the fol-

lowing form for α ∈ (0, 1),

4
d := B ΓΓ + τ v, (28)
D

then we get from the convexity property of Φ that, for any actual evolution
(Berger, 1977):

4
d := B Γ Γ + τ v ≥ Φ(B Γ , v; ρl , ρg , ρv , α, T ) − Φ(0, 0; ρl , ρg , ρv , α, T ). (29)
D

In view of Eq. (25), it is easy to see that d ≥ 0 for any actual evolution of
the fluid and so the SLT given by the inequality Eq. (6) is always satisfied.
From the mechanical viewpoint, Eq. (28) establishes that the dissipated
energy rate is due to the phase change process and to the fluid friction. The
constitutive assumptions made so far are sufficient to partially characterize
the mechanical behavior of the fluid. Since the mass balance equations given

15
by Eqs. (2-4) define a subspace of the linear space spanned by ∂v/∂s, ρ̇l , ρ̇g , ρ̇v
and α̇, then in order that Eq. (24) be equal to Eq. (28) for any actual evolution
it is possible to prove, by consistently exploring the entropy inequality by
using the extension of the Coleman-Noll procedure (Coleman and Noll , 1963;
Eringen and Ingram , 1965; Green and Naghdi , 1965), as it can be seen in
detail in (Silva and Freitas Rachid , 2013; Freitas Rachid , 2014), that for α ∈
(0, 1), one must have the following relationships among the thermodynamic
forces:

p = pl = pg + pv , (30)

B Γ = gl − gv . (31)

Equation (30) establishes that the mixture pressure is equal to the par-
tial pressure of the liquid, which in turn is equal to the sum of the partial
pressures of the vapor and the inert gas. As expected, Eq. (31) indicates that
the thermodynamic force associated with the rate of mass transfer between
the liquid and the vapor is equal to the difference of the Gibbs free energies
in these two phases.
In the sequel, we assume the following simple and appropriate choice for
Φ:
κ
2 ρv 2 |v|
Φ(B Γ , v; ρl , ρg , ρv , α, T ) := B Γ + fm . (32)
2 12
in which κ = κ̂(T ) and fm = fˆm (ρl , α, T ) are positive material constants.
Equation (32) establishes that there are only two dissipative mechanisms in
the fluid flow. One is associated with the liquid-vapor phase change process
and the other is related to the wall shear stress. Since the expression in
Eq. (32) satisfies the conditions set forth before, then the SLT given by

16
Eq. (24) is unconditionally satisfied. Taking Eq. (32) into account, along
with Eq. (31), Eqs. (26) and (27) give rise to the following complementary
laws:

Γ = κ (gl − gv ) , (33)
ρv|v|
τ = fm . (34)
8
Equations (19-21), (30) and (33-34) form a complete set of constitutive
equations for the balance equations given by (2-5), in such a way that the
Clausis-Duhen inequality given by Eq. (6) is unconditionally satisfied. To
complete the model keeping in mind the slack line flow phenomenon, it suf-
fices to specify the forms of the Gibbs free energy difference and the modified
friction factor.

3.3. Gibbs free energy difference and the modified friction factor
As stated by Eq. (33), the Gibbs free energy difference between the liquid
and the vapor is the thermodynamic driving force associated to the liquid-
vapor phase change. In the absence of free gas, it can be shown that Eq. (30)
reduces to the form p = pl = pv while Eq. (33) remains unaltered. Since
the thermo-mechanical equilibrium of the liquid-vapor mixture requires that
gl − gv = 0 (what is equivalent to ensure that there is no phase change
according to Eq. (33)), these conditions ensure the existence of a saturated
vapor pressure psv = p̂sv (T ) such that p = pl = pv = psv (Freitas Rachid
, 2003). In other words, for a pure liquid-vapor mixture there will exist
only one saturated vapor pressure psv for a fixed temperature for which the
mixture is at equilibrium. When these conditions are taken into account,
it can be proved that the Gibbs free energy difference between the liquid

17
and the vapor may be expressed as (Freitas Rachid , 2003; Silva and Freitas
Rachid , 2013):
 p + a2 ρo  psv  1 1 
l l l 4 o
gl −gv = a2l log 2
+av log +al ρl − . (35)
psv + a2l ρol pv psv + a2l ρol pl + a2l ρol
When a free gas is present in the mixture, Eq. (35) still expresses the
Gibbs free energy difference between the liquid and the vapor, although
p = pl = pv is no longer valid. Since now Eq. (30) must hold, it can be
verified based on Eq. (35) that the condition gl − gv = 0, which is a req-
uisite for the thermo-mechanical equilibrium of the mixture, is satisfied for
different values of pl and pv , according to the partial pressure of the gas pg
in the mixture. In other words, for a fixed temperature there will exist more
than one equilibrium state of the liquid-vapor-gas mixture, depending on the
partial pressure of the gas. As a matter of fact, by setting gl − gv = 0 in
Eq. (35) along with p = pl = pv + pg from Eq. (30), we can express the vapor
pressure at equilibrium pv for different values of the mixture pressure p. To
better quantify the influence of the inert gas on the equilibrium mixture of
a liquid-vapor-gas mixture, a normalized graph of pv /psv as a function of
p/psv = (pv + pg )/psv is shown in Fig. 2 for a water-vapor-air mixture at
T = 293 K, at which psv = 2.34 kPa. In constructing the graph of Fig. 2 the
values assigned to the parameters psv , ρol , al, av and ag for the water, vapor
and air are those used in the simulations reported in Section 5.
As it can be seen in Fig. 2, the net effect of the presence of the inert gas
in the mixture at equilibrium, which is felt by varying the mixture pressure,
is to slightly alter the equilibrium vapor pressure. For mixture pressures
up to 103 psv the increase in the equilibrium vapor pressure does not exceed
4% of psv . For significant high pressures of the order of 104 psv the equilib-

18
rium vapor pressure will be 40% greater than psv . The modification of the
equilibrium vapor pressure in a mixture of liquid, vapor and an inert gas is
a well-known result of the phase equilibrium thermodynamics that consis-
tently provides the appropriate conditions for the onset and the completion
of the phase change phenomenon. As we will see in Section 5, by assuming
that the fluid is a liquid-vapor-gas mixture, with very small values of gaseous
volume fractions, the vaporous cavitation in the fluid flow can be described
in a consistent thermodynamic framework as an irreversible process, without
the need to consider different set of equations for the extremal situations of
pure liquid α = 0 and pure vapor α = 1. With only one set of equations the
cavitating and non-cavitating regions of the fluid flow domain are automat-
ically identified as the gaseous volume fraction approaches to zero and one,
respectively.
Equation (34) expresses one way of evaluating the wall shear stress. In
this equation fm stands for the modified friction factor,
f
fm :=  4/3 , (36)
1 − sin (2ω)
2ω
in which ω is the angle which defines the liquid content in the cross-section
ˆ
of the pipe (see Fig. 1) and f = f(Re, ) is the Darcy-Weisbach steady
friction factor in tight flow conditions. This last term is a function of the
steady Reynolds number Re (with Re := (ρvo D)/µ, being vo the steady state
fluid relative velocity and µ the liquid dynamic viscosity) and of the relative
roughness of the pipe wall .
The modified friction factor has been proposed by Nicholas (1995) in or-
der to properly handle the adequate friction factor in both tight and slack flow

19
conditions by using a single expression. In contrast to the Darcy-Weisbach
friction factor, the modified friction factor is capable to reproduce the ap-
propriate friction factor in segments of the pipeline that are partially full,
where the slack line flow regions are approximated as open channel flow. It
is worthwhile noting that the proposed expression for the wall shear stress
does not take frequency-dependent effects into account which may be very
important in some transient regimes, specially in tight flow stretches of the
pipeline (Ferrari , 2010; Jiang et al. , 2012). However, as it will be clear later,
this paper will focus attention on the steady regimes only, what justifies the
use of Eq. (34).
Finally, it should be noticed that the gaseous volume fraction α, with
α ∈ (0, 1), can be expressed as a function of angle ω, with ω ∈ (0, π), i. e.
α = α̂(ω), according to the following expression:

ω sin(2ω)
α = α̂(ω) = 1 − + . (37)
π 2π

Since α = α̂(ω) is a monotonic function, its inverse exists and so it is

possible to write ω = α̂−1 (α), allowing the aforementioned representation
fm = fˆm (ρl , α, T ). To better understand the way the slack line flow condi-
tion affects the modified friction factor, the graph presented in Fig. 3 displays
the ratio fm /f as a function of the gas volume fraction α, for a same Reynolds
number and pipe roughness. As it can be seen, as long as the liquid fills more
than half cross-section of the pipe the modified friction factor practically re-
mains unaltered. However, as the pipe cross-section empties even further,
the modified friction factor increases substantially, by promoting stringent
head losses in slack line flow regions as it has been advocated in Section 1.

20
4. Governing equations in steady-state: an initial-value problem

Since the main purpose of this work is to analyze the capability of the
present model in describing coherently the slack line flow in hydraulic pipelines,
we hereafter restrict the analysis to steady state. Thus, the governing equa-
tions given by Eqs. (2) to (5) can be reduced, with the aid of Eqs. (19-21)
and (30), to the following set of ordinary differential equations in terms of
the spatial coordinate s for the dependent variables u := (α, p, pg , v)T :

du
A(u) = b(u, s) for s ∈ (0, L), (38)
ds

in which the matrix A and the vector b assume the following forms:
 αv 
ρg v 0 αρ g
 a2g 
 αv αv 
 ρv
 v − αρ v

a 2 a 2 
A(u) =  v v , (39)
 −ρ v (1 − α)v
 
l 2
0 (1 − α)ρ 
l 

 a l 
0 1 0 ρv
 
0
 
Γ
 
 
b(u, s) =  . (40)
−Γ
 
 
4τ
 
− − ρg sin θ
D

In the above equations, θ = θ̂(s) and a2g and a2v are constants defined in
Eqs. (20) and (21), which represent the isothermal wavefront propagation
velocities in the inert gas and vapor, respectively. The variables ρl , ρv , ρg
and ρ are to be interpreted as functions of the dependent variables vector
u := (α, p, pg , v)T , i. e., ρl = ρ̂l (p), ρv = ρ̂v (p, pg ), ρg = ρ̂g (p, pg ) and

21
ρ = ρ̂(α, p, pg ), what can be done by invoking Eqs. (7), (19-21) and (30).
Similarly, in view of Eqs. (30) and (35), the variable Γ, defined by Eq. (33),
can be written as Γ = Γ̂(p, pg ) and the variable τ , defined by Eq. (34), can
be expressed as τ = τ̂ (α, p, pg , v) by taking Eqs. (7), (19-21), (30), (36) and
(37) into account.
To explore the capability of the model in properly describing the slack
line flow, the system of ordinary differential equations given by Eq. (38) will
be solved in the context of an initial-value problem, with the initial condition
being prescribed at the pipe entrance;

u(s = 0) := uo = [αo po pgo vo ]T , (41)

in which αo ∈ (0, 1), po , pgo > 0 and vo > 0.

The existence and uniqueness of the solution of the initial-value problem
defined by Eqs. (38) and (41) is conditioned to some particular conditions
which can be better identified when Eq. (38) is rewritten as:
du
= f(u, s) := A−1 b with, (42)
ds    
αvΓ 4τ
 p (w + x + y) − z D + ρg sin θ 
   
 4τ 
 −vΓy − + ρg sin θ 
1  D  
f(u, s) =    , (43)
(1 − M 2 )  pg
 −vΓ (w + y) − pg 4τ 
+ ρg sin θ 

 p  p D 


 Γy v 4τ 
+ 2 + ρg sin θ
ρ ρa D
in which
c2v (1 − M 2 )
w := , (44)
α M2
p c2v
x := − , (45)
ρl v 2 v 2

22
 ρc2v ρ
y := − , (46)
p ρ
 l
ρa2

α
z := 2 1 − , (47)
ρa p
v
M := , (48)
a

ρl a2l ρv a2v + ρg a2g
a :=

. (49)
ρ ρv a2v + ρg a2g (1 − α) + αρl a2l

In Eqs. (48) and (49), M and a stand for the Mach number and the wavefront
speed with which disturbances propagate in the medium, respectively.
According to the theorems of existence and uniqueness for initial-value
problems for ordinary differential equations (Perko , 2001; Teschl , 2012),
continuity and differentiability of f(u, s) with respect to u are the funda-
mental sufficient conditions for ensuring local existence and uniqueness of
solution, respectively. Since α ∈ (0, 1), p > 0 and pg , pv > 0, if we assume
that Mo = vo /ao < 1, in which vo = v(s = 0) and ao = a(s = 0), it can be
easily verified (see Eqs. (43-49)) that the conditions set forth for f are verified
as long as M 6= 1 and v 6= 0. Since the main objective of this paper is to
evaluate the capability of the model in properly describing the slack line flow
phenomenon rather than to explore all the possibilities regarding the solu-
tions of the initial-value problem defined by Eq. (42) and (41), the numerical
simulations will be carried out next until the onset of sonic flow when M → 1
somewhere in the spatial domain; the unique condition susceptible to take
place in the context of the initial-value problem.

23
5. Numerical results and discussion

The objective of this section is to explore the capability of the proposed

model in properly describing the opening and closure of a cavity as a function
of distance due to the vaporous cavitation in steady state under distinct
topographic conditions. To do so, the initial-value problem defined in Section
4 is solved numerically for a specific pipe arrangement in which distinct
elevation profiles are generated. The line has a total length of L = 150 m
and is composed by two segments of straight pipes. The first segment is
50 m long and goes upward vertically, i.e. θ1 = 90o . By adhering to the
common practice in hydraulics, instead of working with the angle θ1 = 90o
with respect to the horizontal, we use its counterpart β1 = 45o (defined as
β1 := arctan(sin θ1)) which has a straightforward geometrical meaning when
the pipe elevation z is plotted against the spatial position s measured along
the pipe centerline, as presented in Fig. (4(a)). The second segment is 100 m
long and goes downhill according to different declinations, which in terms of
the angle β2, defined as β2 := −arctan(sin θ2), are β2 = 0o , 0.5o , 1o , 10o and
45o (see Fig. 4(a)), rendering five distinct elevations profiles. The extremal
conditions portrayed by the angles β2 = 0o and β2 = 45o represent the
situations in which the second pipe segment is horizontal and goes downward
vertically via a U-bend, respectively. The pipes have a constant internal
diameter D = 0.2032 m (8 in) through which a water-air mixture flows at
T = 293 K (20o C). At this temperature, the values of the parameters psv , ρol ,
al, av and ag are equal to 2.34 kPa, 998.1419 kg/m3 , 1477 m/s, 368 m/s and
290 m/s, respectively.
The initial conditions at the pipe inlet are characterized by the vector

24
given by Eq. (41), whose components are: α0 = 10−8 , po = 494 kPa, pgo =
491.64 kPa and v0 = 1.0 m/s. For these conditions Re =2.0×105 , which
results in f = 0.021 by assuming that the relative roughness of the pipe is
 =0.001. It is worth noting that the gaseous volume content in the mixture
is very small and so the mixture flowing at the pipe inlet is essentially made
of liquid water. The initial partial pressure of the inert gas pgo is computed
in terms of po , by considering Eq. (30) evaluated at s = 0 (po = plo =
pgo + pvo ) and Eq. (35) subjected to the condition gl − gv = 0. As a result,
it comes out that pvo = 2.36 kPa, which is a partial vapor pressure slightly
greater than the saturated vapor pressure (psv =2.34 kPa at T = 293 K)
as it has been commented in Section 3.3. By imposing this condition it is
tacitly assumed that the liquid and vapor are at thermodynamic equilibrium
within the mixture at the pipeline entrance. Finally, to assign an irreversible
character to the phase change transformation we have run the simulations
by adopting κ = 10−8 kg2/m3 Js (Freitas Rachid , 2003). As it has been
shown by Freitas Rachid (2003), κ of this order represents an intermediate
value between a lower and an upper bounds for which the phase change
transformations in water take place as reversible processes.
Although there are many numerical methods available to obtain approx-
imated solutions to the initial-value problem defined by Eqs. (42) and (41),
we have used the well-known Gear’s method herein (Gear , 1971a). It is an
implicit, multi-step, adaptive method that uses backward differentiation for-
mulas up to fifth order (Gear , 1971a,b; Byrne and Hindmarsh , 1987). This
choice has been based on the stiff nature of the system of equations given
by Eq. (42), for which the Gear’s method has been known to exhibit a very

25
good performance. The numerical simulations shown ahead were carried out
with initial mesh sizes of the order of 0.001 m and relative errors of 10−9 .
In what follows, the ability of the model in properly describing the slack
line flow is assessed by analyzing the numerical predictions of the proposed
model when the angle β2 assumes the values β2 = 0o , 0.5o , 1o , 10o and 45o .
The modified piezometric head (see Eq. (1))and the pipe elevation as a func-
tion of spatial position measured along the pipe centerline are presented in
Fig. 4(a) and Fig. 4(b) for these five configurations, respectively. As expected,
from the pipe inlet until the top of the line (s = 50 m), where a vapor cavity
opens, there is no difference on any flow property among these five configu-
rations. From that point on, however, different behaviors are observed in the
fluid flow due to the occurrence of the vaporous cavitation phenomenon. Ex-
cept to the case for β2 = 0o , in which the simulation has stopped because the
fluid flow has become sonic (the Mach number has reached the unity value,
= 1.73m/s or a/al ∼
with v = a ∼ = 1.17 × 10−3 ), all the other cases have quite
similar behaviors whether we look at the modified piezometric head against
the spatial position. However, as we shall see, different behaviors will be
observed for the cases β2 = 0.5o , 1o , 10o and 45o in which the vapor cavity
closes at different positions at the downhill segment of the pipe. Figure 5
presents the pipe elevation and the modified piezometric head against the
spatial position along the pipe for β2 = 0.5o , 1o , 10o and 45o . In contrast
to Figs. 5(c) and 5(d), it can be seen in Figs. 5(a) and 5(b) (for β2 = 0.5o
and β2 = 1o , respectively) that the modified piezometric head not only gets
closer to the pipe but also remains that way downstream. As we shall see
later, this particular feature is responsible for promoting quite distinct cavity

26
extensions.
Figures 6(a) and 6(b) show the fluid velocity and the Mach number,
respectively, as a function of the spatial position along the pipe for β2 = 0o ,
0.5o , 1o , 10o and 45o . To leave the effects of the opening and the closure of
the cavity in evidence, from now on we will only focus attention on the pipe
stretch s ∈ [40, 100] m where such features occur. The variations observed
in the fluid velocity and in the Mach number take place from the opening
to the closure of the vapor cavity, for each one of the cases portrayed by
the angle β2. This information can be corroborated by Fig. 9(b) (presented
a little later) which exhibits the gaseous volume fraction as a function of
the spatial position along the pipe. The opening of the vaporous cavity
promotes an increase in the gaseous volume fraction which in turn makes
the fluid to accelerate in order to satisfy the continuity equation. On the
other hand, the closure of the vapor cavity induces the reverse process, by
decreasing the gaseous volume fraction and decelerating the fluid velocity.
For β2 = 0o , the normalized wavefront speed a/al (see Eq. (49)) increases
faster than the fluid velocity what makes the Mach number to reach the unit
approximately at s = 57.8 m. At this stage the gaseous volume fraction is
greater than 40% (see Fig. 9(b)) and the cavity, which extends to almost
8 m, is no longer capable to close. The predicted extensions of the vapor
cavities are approximately 32 m, 8 m, 1 m and 0.1 m for β2 = 0.5o , 1o , 10o
and 45o , respectively, demonstrating that they vary significantly according
to the pipe topography. The fluid velocity and the Mach number apparently
return to their original values after the vapor cavity closes.
The mixture pressure and the gas partial pressure along the spatial posi-

27
tion of the pipe stretch s ∈ [40, 100]m are illustrated in Figs. 7(a) and 7(b),
respectively. As it can be noticed, the mixture pressure as well as the gas
partial pressure fall linearly due to the friction and the elevation profile in
the uphill segment of the pipe until the vaporous cavitation phenomenon
begins around the top of the line. From that point on, the pipe profile of the
downhill segment induces pressures of different magnitudes inside the pipe
which are responsible for different extensions of the vapor cavity. As it can
be observed in Figs. 7(a) and 7(b), the pressure gradient is the same for the
mixture pressure and the gas partial pressure at both the uphill and down-
hill segments of the line outside of the cavity. The more abrupt the angle
β2 is, the greater the pressure recovery at the downhill segment becomes. In
addition, the lower the pressure remains along the downhill side, the greater
the cavity extension becomes.
The way the mixture pressure varies inside the vapor cavity extension and
outside of it can be better understood by appealing to the zoom in windows
in Figs. 7(a) and 7(b). Outside the cavity, Γ = 0 (or equivalently, gl − gv = 0)
what implies that pv ∼
= constant and is of the order of psv . Thus, by virtue
of Eq. (30) dp/ds ∼
= dpg /ds. However, along the cavity extension Γ 6= 0 (or
equivalently, gl − gv 6= 0) with p → pv (which is still of the order of psv )
by implying by virtue of Eq. (30) that pg → 0 (see the zoom in window in
Figs. 7(b)) and so dp/ds ∼
= dpv /ds. Thus, the existence of a vapor cavity
alters the way the mixture pressure gradient varies throughout its extension.
To better characterize the phase change transformations that take place
along the cavity extension the vapor partial pressure and the rate of mass
transfer are plotted as a function of the spatial position along the pipe stretch

28
s ∈ [40, 100] m for β2 = 0o , 0.5o , 1o , 10o and 45o in Figs. 8(a) and 8(b),
respectively. As the cavity opens, the vapor partial pressure falls below
the equilibrium vapor pressure, which is slightly greater than the saturated
vapor pressure (see Fig. 2), and the liquid begins to be transformed into
vapor (the mass transfer rate is positive as observed in Fig. 8(b)) and the
cavity begins to expand. Except for β2 = 0o in which the vapor production
is so intense that the vapor partial pressure decreases continuously until
the flow becomes sonic, in the other geometrical configurations the vapor
partial pressure attains a minimum and increases until it becomes greater
than the equilibrium vapor pressure. At that condition the reverse process
initiates. Vapor starts to be transformed into liquid (the mass transfer rate
becomes negative as observed in Fig. 8(b)) and the cavity begins to shrink.
The positions along the pipe it effectively takes place and the spatial rate at
which the cavity grows and shrinks depends on the angle β2. The less this
angle is, the farther that point becomes with respect to the top, the less the
spatial rate the cavity expands and contracts and the larger the vapor cavity
becomes. Note that for β2 = 45o the cavity opens and soon after closes,
rendering an extension of the order of a few centimeters.
These processes follow the same pattern and are marked by an abrupt
change in the vapor partial pressure as well as in the mass transfer rate, as
it can be observed in Figs. 8(a) and 8(b). As it has been shown in Section
3.3, the thermodynamic driving force responsible for the opening and closure
of the vapor cavity is the Gibbs free energy difference which, for isothermal
transformations, depends solely on the mixture and vapor pressures. As long
as the mixture pressure remains well above the equilibrium vapor pressure,

29
the Gibbs free energy difference equals zero and no phase change transfor-
mation takes place. As the vapor partial pressure becomes less or greater
than the equilibrium vapor pressure, the phase change transformation takes
place by producing vapor or liquid, respectively.
Up to the onset of the cavity, while the partial pressure of the air varies
significantly with the position, the vapor partial pressure remains practically
unaltered as shown in Figs. 7(b) and 8(a). However, from that point on, the
partial pressure of vapor oscillates around the saturated vapor pressure by
promoting significant variations of the Gibbs free energy difference between
the liquid and the vapor (see Fig. 8(a)); the driving force responsible for the
phase change transformation.
The effect of the vapor cavity formation on the wavefront speed can be vi-
sualized in Fig. 9(a) where the normalized wavefront speed is plotted against
the spatial position along the pipe stretch s ∈ [40, 100] m for the five different
inclinations of the downhill segment, β2 = 0o , 0.5o , 1o , 10o and 45o . As the
vapor cavity remains open, the wavefront speed is drastically reduced as a
result of the increase of the compressibility of the mixture. It is due to the
combined effects of the low pressure and high gas volume fraction that pre-
dominates while the vapor cavity persists, as it can be noticed in Fig. 9(b),
which shows the gas volume fraction as a function of the spatial position
along the pipe stretch s ∈ [40, 100] m for β2 = 0o , 0.5o , 1o , 10o and 45o .
Except to the case in which the cavity opens but does not closes afterwards
(β2 = 0o ), it can be seen that the large-scale spatial variation of the gaseous
volume fraction allows the identification of the beginning and the end of the
vapor cavity, and therefore of its extension as it has been mentioned before.

30
 As the vapor cavity opens around the top of the line, a stringent increase
of the gaseous volume fraction from 10−8 to values up to 0.4 (see Fig. 9(b))
is observed due to the transformation of liquid into vapor. This statement
can be corroborated through the graphic of the rate of mass transfer as a
function of the position as illustrated in Fig. 8(b). From the pipe entrance
to approximately the top of the pipe where the vapor cavity opens there
is no vapor generation. However, from that point on until the sign of the
mass transfer change from positive to negative, vapor is generated causing
an appreciable increase of the gaseous volume fraction which in turn reduces
drastically the wavefront speed for all the angles β2 . When approximately
one-half of the pipe’s cross-section is filled by the vapor, the flow becomes
sonic as it can be observed in Fig. 9(b) for β2 = 0o .
The increase in the gaseous volume fraction of about 105 within a very
narrow neighborhood around the top of the line can be seen in Fig. 10(a)
which presents a zoom in around the spatial position s = 50 m. As a con-
sequence, the wavefront speed in the mixture jumps down sharply at the
beginning of this process, as it can be observed in Fig. 9(a) which exhibits
the mixture-to-liquid wavefront speed ratio as a function of the spatial posi-
tion along the pipe. As the vapor cavity closes, the wavefront speed abruptly
increases and progressively tends to recover its original value, by exhibiting
a typical space-dependent effect. Besides the pressure, such an effect is gov-
erned by the gaseous volume fraction which does not recover its initial value
due to the irreversible character of the phase change transformation. This
feature is better visualized in the graph of Fig. 10(b) which presents a zoom
in around the spatial position s ≈ 82 m where the vapor cavity closes for the

31
downhill inclination β2 = 10o . As it can be seen in Fig. 10(b), the value of
the gaseous volume fraction for β2 = 10o tends to approximately 1.0×10−5
after the vapor cavity closes instead of its original value αo = 10−8 . It can
also be seen that this same phenomenon takes place, in a somewhat moder-
ate fashion, for the angle β2 = 1.0o , for which the gaseous volume fraction
reaches a residual value of the order of 0.125×10−5 . In this case, however,
the vapor cavity closes around s ≈ 58, as it can be observed in Figs. 8(b)
and 9(b). Since the amount of mass of liquid generated during the collapse
phase of the cavity is slightly less than the mass of vapor produced during its
expansion phase, the difference has remained as vapor along with the inert
gas in the gaseous phase after the cavity closure. This is why the residual
value of the gaseous volume fraction, soon after the cavity closure, is slightly
greater than its initial value before its opening, when it is subjected to al-
most the same mixture pressure. These residual values are responsible for
lowering significantly the wavefront speed after the cavity has been closed
for the cases in which β2 = 1.0o and 10o (see Fig. 9(a)).
These abrupt discontinuities in the wavefront speed of the mixture, as
a result of the phase change process, is one of the major difficulties when
numerical techniques are applied to obtain approximated solutions in both
steady and unsteady regimes. In the case of steady state, this example clearly
demonstrates that Gear’s method is a good choice since it has captured this
sharp discontinuity with success.
Finally, it is presented in Fig. 11(a) the graph of the rate of energy dissi-
pation associated with the phase change process (first term of the right-hand
side of Eq. (28)) as a function of the spatial position along the pipe stretch

32
s ∈ [40, 100] m for the five different inclinations of the downhill segment
β2 = 0o , 0.5o , 1o , 10o and 45o . The regions of the pipe where the phase
change transformation takes place are marked by a non-null rate of energy
dissipation. It happens during the stages of expansion as well as collapse of
the vapor cavity when vapor and liquid are formed, respectively. One should
note that before the opening and after the closing of the vapor cavity there
is no phase change and consequently no dissipation. Since the phase change
transformation is modeled as an irreversible process - as it should be from
the correct physical viewpoint - the rate of energy dissipation associated with
the phase change is always non-negative. Even though the magnitude of the
rate of energy dissipation involved in the phase change process is quite small,
it seems to be capable to introduce a relevant irreversible feature, such as
the permanent reduction of the wavefront speed.
To have a better idea of the magnitude of the rate of energy dissipation
due to the mass transfer process, it is presented in Fig. 11(b) the rate of
energy dissipation associated with friction (second term of the right-hand
side of Eq. (28)) as a function of the spatial position for different declinations
β2 = 0o , 0.5o , 1o , 10o and 45o of the downhill part of the line. By comparing
Figs. 11(a) and 11(b) we can see, by excepting the case for which β2 =
0o , that the maximum rate of energy dissipation associated with the phase
change process is up to approximately 20% of that related with the fluid
friction.

33
6. Conclusions

This work has presented a theoretical thermodynamically consistent model

to describe one-dimensional cavitating flows in hydraulic pipelines. Although
the model is capable to deal with the cavitation phenomenon in unsteady
flows, the application presented herein is focused on slack line flow, which
takes place in steady state. The vaporous cavitation is modeled as a dissi-
pative phenomenon within the context of the thermodynamics of irreversible
processes in such a way that the second law of the thermodynamics is un-
conditionally satisfied. To ensure that the same set of equations is used
throughout the whole fluid domain, with sub-domains experiencing cavita-
tion and others not, a small amount of an inert gas is assumed to coexist
along with the vapor everywhere in the fluid domain. The presence of an
inert gas ensures the necessary thermodynamic conditions to coherently de-
scribe not only the opening and the closing of the vapor cavity but also the
phase transformations of liquid into vapor and vice-versa.
In order to access the capability of the model to properly describe the slack
line flow, a numerical example is presented for an air-water mixture flowing in
a simple pipeline arrangement made of two segments of straight pipes which
go sequentially uphill and downhill. By changing the downhill inclination
of the second segment, distinct elevation profiles are generated rendering
different conditions for the formation of a vapor cavity in the second pipe
segment. Making use of a suitable and robust classical numerical method for
solving the initial-value problem for the set of ordinary differential equations
that govern the problem, very different cavity extensions have been obtained
ranging from 0.1 to 32 m with rates of energy dissipation up to 12 W/m3 .

34
The obtained results demonstrate that the model is able to properly report
important physical features, such as the description of the phase change as
an irreversible process, its implication on the wavefront speed in the medium
and its capability in predicting the rate of energy dissipation. Although
experimental results are still required to definitively validate these findings,
they seem to qualify the model as a promising tool for practical applications.

Acknowledgments
The authors kindly acknowledge the partial financial support that has been
provided along the past years by the Brazilian Council of Scientific and
Technological Development (CNPq) and by the Carlos Chagas Foundation
(FAPERJ) of the Rio de Janeiro State.

References

Arndt, R. E. A., 1981. Cavitation in fuid machinery and hydraulic structures.

Annual Review of Fluid Mechanics, 13, 273-328.

Assumpção, A. H. and Freitas Rachid, F.B., 2010. Mechanical Model for Cav-
itating Flow in Hydraulic Pipelines. 13rd Brazilian Congress of Thermal
Sciences and Engineering, December 05-10, Uberlândia, MG, Brazil.

Berg, A., Iben, U., Meister, A. and Schmidt, J. 2005. Modeling and simula-
tion of cavitation in hydraulic pipelines based on the thermodynamic and
caloric properties of liquid and steam. Shock Waves, 14, (1), 111–121.

Bergant, A., Simpson, A. R., and Tijsseling, A. R. 2006. Water hammer with
column separation: A historical review. Journal of Fluids and Structures,
22, 135–171.

35
Berger, M. S. 1977. Nonlinearity and Functional Analysis. Academic Press,
London.

Byrne, G. D. and Hindmarsh, A. C., 1987. Stiff ODE Solvers: A review of

current and coming attractions. Journal of Computational Physics, 70,
(1), 1–62.

Coleman, B. D. and Noll, M., 1963. The thermodynamics of elastic materials

with heat conduction and viscosity. Archive for Rational Mechanics and
Analysis, 13, 167–178.

Drew, D. A. and Passman, S. L., 1999. Theory of Multicomponent Fluids.

Springer-Verlag, New York.

Ekeland, I. and Teman, R. 1976. Convex Analysis and Variational Problems.

North-Holland, Amsterdam.

Eringen, A. C. and Ingram, J. D., 1965. A continuum theory of chemically

reacting media I. International Journal of Engineering Science, 3, 197–212.

Ferrari, A. 2010. Modelling approaches to acoustic cavitation in transmission

pipelines. International Journal of Heat and Mass Transfer, 53, 4193–4203.

Freitas Rachid, F. B. 2003. A thermodynamically consistent model for cav-

itating flows of compressible fluids. International Journal of Non-Linear
Mechanics, 38, 1007–1018.

Freitas Rachid, F. B., 2006. Continuum thermo-mechanical model for homo-

geneous liquid-gas flows with internal surface tension effects. Mechanics
Research Communications, 33, 337–351.

36
Freitas Rachid, F. B., 2014. Modeling gaseous and vaporous cavitation in
liquid flows within the context of the thermodynamics of irreversible pro-
cesses. International Journal of Non-Linear Mechanics, 65, 245–252.

Frémond, M., and Nicolas, P., 1999. Macroscopic thermodynamics of porous

media. Continuum Mechanics and Thermodynamics, 2, 119–139.

Gear, C. W., 1971a. Algorithm 407: DIFSUB for solution of ordinary differ-
ential equations. Communications of the ACM, 14, (3), 185-190.

Gear, C. W., 1971b. Numerical Initial Value Problems in Ordinary Differ-

ential Equations. Prentice-Hall, Englewood Cliffs, NJ.

Germain, P. and Muller, P., 1995. Introduction à la Mécanique des Milieux

Continus. Masson, Paris.

Green, A. E. and Naghdi, P. M., 1965. A dynamical theory of interacting

continua. International Journal of Engineering Science, 3, 231–241.

Ishii, M. and Hibiki, T., 2006. Thermo-Fluid Dynamics of Two-Phase Flow.

Springer, New York.

Jiang, D., Li, S., Edge, K. A. and Zeng, W., 2012. Modeling and simulation
of low pressure oil-hydraulic pipeline transients. Computers & Fluids, 67,
79–86.

Kessal, M. and Amaouche, M., 2001 Numerical simulation of transient va-

porous and gaseous cavitation in pipelines. International Journal of Nu-
merical Methods for Heat and Fluid Flow, 11, (2), 121–137.

37
Lighthill, J., 1978. Waves in Fluids. Cambridge University Press, London.

Liu, T. G., Khoo, B. C. and Xie,W. F., 2004. Isentropic one-fluid modelling of
unsteady cavitating flow. Journal of Computational Physics, 201, 80–108.

Moreau, J. J., Panagiotopoulos, P. D., and Strang, G. 1988. Topics in non-

smooth mechanics. Birkhäuser.

Nicholas, R. E., 1995. Simulation of slack flow: a tutorial. In Proceedings of

the Pipeline Simulation Interest Group Conference. October 19-20, Albu-
querque, New Mexico, 1–10.

Perko, L. 2001. Differential Equations and Dynamical Systems. 3rd Ed.,

Springer-Verlag, New York.

Sadafi, M., Riasi, A. and Nourbakhsh, S. ., 2012 Cavitating flow during water
hammer using a generalized interface vaporous cavitation model. Journal
of Fluids and Structures, 34, 190–201.

Schraml, K. C., Fujikawa, M. Y., Ivanouskas, E. A., Braz, A. L., Andrade,

A. C. and Severino, J. P., 2011. Improving operational safety in oil
pipelines. In Proceedings of the 6th Pipeline Technology Conference. April
04-05, Hannover Messe, Germany, 1–8.

Shu, J. J. 2003. Modelling vaporous cavitation on fluid transients. Interna-

tional Journal of Pressure Vessels and Piping, 80, 187–195.

Silva, A. B. and Freitas Rachid, F. B., 2013. Modeling of release and absorp-
tion of gas in liquid-gas flows within a consistent thermodynamic frame-
work. International Journal of Engineering Science, 66-67, 21–43.

38
Soares, A. K., Martins, N. and Covas, D. I. C., 2015. Investigation of tran-
sient vaporous cavitation: experimental and numerical analyses. Procedia
Engineering, 119, 235–242.

Stanley, G., 2005. Sensitivity study for leak detection in slack line conditions.
In Proceedings of the Rio Pipeline 2005 Conference & Exposition. October
17-19, Rio de Janeiro, Brazil, Paper IBP1399-05, 1–16.

Sumam, K. S., Thampi, S. G. and Sajikumar, N., 2010. An alternate ap-

proach for modelling of transient vaporous cavitation. International Jour-
nal for Numerical Methods in Fluids, 63, 564–583.

Teschl, G. 2012. Ordinary differential equations and dynamical systems.

American Mathematical Society, Rhode Island.

Tijsseling, A. S., Vardy, A. E. and Fan, D., 1996. Fluid-structure interac-

tion and cavitation in a single-elbow pipe system. Journal of Fluids and
Structures, 10, 395–420.

Tijsseling, A. S. and Vardy, A. E., 2005. Fluid-structure interaction and tran-

sient cavitation tests in a T-piece pipe. Journal of Fluids and Structures,
20, 753–762.

Utturkar, Y., Wua, J., Wang, G. and Shy, W., 2005. Recent progress in
modeling of cryogenic cavitation for liquid rocket propulsion. Progress in
Aerospace Sciences, 41, 558–608.

Wylie, E. B., 1984. Simulation of vaporous and gaseous cavitation. ASME

Journal of Fluids Engineering, 106, 307–311.

39
Wylie, E. B. and Streeter, V. L. 1993. Fluid transients in systems. Prentice
Hall, New York.

Xie, W. F., Liu, T. G., Khoo, B. C., 2006. Application of a one-fluid model
for large scale homogeneous unsteady cavitation: The modified Schmidt
model. Computers &Fluids, 35, 1177–1192.

Yuanyuan, Y., Jing, G., Xiaoping, L., Tong, Z. and Lijun, Z., 2012. Study
improves control of slack line flow. Oil & Gas Journal, 110, 12, 1–15.

40
 List of Figure Captions

Figure 1: Elevation (black line) and modified piezometric head (red line) in
a pipeline operating under tight (dashed red line) and slack line flow (solid red
line) conditions under steady state. Cross-section view A–A of the pipeline
under slack line flow is shown in detail.

Figure 2: Normalized vapor pressure against normalized mixture pressure

under thermo-mechanical equilibrium.

Figure 3: Normalized modified friction factor against gaseous volume frac-

tion.

Fig. 4(a) Elevation against spatial position along the pipe.

Fig. 4(b) Modified piezometric head against spatial position along the
pipe.
Figure 4: Elevation and modified piezometric head against spatial position
along the pipe for different inclinations β2 = 0o , 0.5o , 1.0o , 10o and 45o of the
downhill part of the line.

Fig. 5(a) Downhill inclination β2 = 0.5o .

Fig. 5(b) Downhill inclination β2 = 1.0o .
Fig. 5(c) Downhill inclination β2 = 10o .
Fig. 5(d) Downhill inclination β2 = 45o .
Figure 5: Elevation and modified piezometric head against spatial position
along the pipe for different inclinations β2 = 0.5o , 1.0o , 10o and 45o of the
downhill part of the line.

Fig. 6(a) Velocity against spatial position along the pipe.

41
 Fig. 6(b) Mach number against spatial position along the pipe.
Figure 6: Velocity and Mach number against spatial position along the pipe
segment s ∈ [40, 100] m for different inclinations β2 = 0o , 0.5o , 1.0o , 10o and 45o
of the downhill part of the line.

Fig. 7(a) Mixture pressure against spatial position along the pipe.
Fig. 7(b) Gas partial pressure against spatial position along the pipe.
Figure 7: Mixture pressure and gas partial pressure against spatial posi-
tion along the pipe segment s ∈ [40, 100] m for different inclinations β2 =
0o , 0.5o , 1.0o , 10o and 45o of the downhill part of the line.

Fig. 8(a) Vapor partial pressure against spatial position along the pipe.
Fig. 8(b) Mass transfer rate against spatial position along the pipe.
Figure 8: Vapor partial pressure and mass transfer rate against spatial
position along the pipe segment s ∈ [40, 100] m for different inclinations
β2 = 0o , 0.5o , 1.0o , 10o and 45o of the downhill part of the line.

Fig. 9(a) Normalized wavefront speed against spatial position along the
pipe.
Fig. 9(b) Gaseous volume fraction against spatial position along the pipe.
Figure 9: Normalized wavefront speed and gaseous volume fraction against
spatial position along the pipe segment s ∈ [40, 100] m for different inclina-
tions β2 = 0o , 0.5o , 1.0o , 10o and 45o of the downhill part of the line.

Fig. 10(a) Zoom in of Fig. 9(b) at the cavity opening.

Fig. 10(b) Zoom in of Fig. 9(b) at the cavity closure.
Figure 10: Zoom in of Fig. 9(b) at the opening and at the closure of the
cavity against spatial position along the pipe segment s ∈ [40, 100] m for

42
different inclinations β2 = 0o , 0.5o , 1.0o , 10o and 45o of the downhill part of
the line.

Fig. 11(a) Rate of energy dissipation due to mass transfer.

Fig. 11(b) Rate of energy dissipation due to friction.
Figure 11: Rate of energy dissipation due to mass transfer and friction
against spatial position along the pipe segment s ∈ [40, 100] m for different
inclinations β2 = 0o , 0.5o , 1.0o , 10o and 45o of the downhill part of the line.

43
Figure 1: Elevation (black line) and modified piezometric head (red line) in a pipeline
operating under tight (dashed red line) and slack line flow (solid red line) conditions
under steady state. Cross-section view A–A of the pipeline under slack line flow is shown
in detail.

44
Figure 2: Normalized vapor pressure against normalized mixture pressure under thermo-
mechanical equilibrium.

Figure 3: Normalized modified friction factor against gaseous volume fraction.

45
 (a) Elevation against spatial position along the pipe

(b) Modified piezometric head against spatial position along the pipe

Figure 4: Elevation and modified piezometric head against spatial position along the pipe
for different inclinations β2 = 0o , 0.5o, 1.0o, 10o and 45o of the downhill part of the line.

46
 (a) Downhill inclination β2 = 0.5o (b) Downhill inclination β2 = 1.0o

(c) Downhill inclination β2 = 10o (d) Downhill inclination β2 = 45o

Figure 5: Elevation and modified piezometric head against spatial position along the pipe
for different inclinations β2 = 0.5o , 1.0o, 10o and 45o of the downhill part of the line.

47
 (a) Velocity against spatial position along the pipe

(b) Mach number against spatial position along the pipe

Figure 6: Velocity and Mach number against spatial position along the pipe segment
s ∈ [40, 100] m for different inclinations β2 = 0o , 0.5o, 1.0o, 10o and 45o of the downhill
part of the line.

48
 (a) Mixture pressure against spatial position along the pipe

(b) Gas partial pressure against spatial position along the pipe

Figure 7: Mixture pressure and gas partial pressure against spatial position along the pipe
segment s ∈ [40, 100] m for different inclinations β2 = 0o , 0.5o, 1.0o, 10o and 45o of the
downhill part of the line.

49
 (a) Vapor partial pressure against spatial position along the pipe

(b) Mass transfer rate against spatial position along the pipe

Figure 8: Vapor partial pressure and mass transfer rate against spatial position along the
pipe segment s ∈ [40, 100] m for different inclinations β2 = 0o , 0.5o, 1.0o, 10o and 45o of
the downhill part of the line.

50
 (a) Normalized wavefront speed against spatial position along the pipe

(b) Gaseous volume fraction against spatial position along the pipe

Figure 9: Normalized wavefront speed and gaseous volume fraction against spatial
position along the pipe segment s ∈ [40, 100] m for different inclinations β2 =
0o , 0.5o, 1.0o, 10o and 45o of the downhill part of the line.

51
 (a) Zoom in of Fig. 9(b) at the cavity opening

(b) Zoom in of Fig. 9(b) at the cavity closure

Figure 10: Zoom in of Fig. 9(b) at the opening and at the closure of the cavity against
spatial position along the pipe for different inclinations β2 = 0o , 0.5o , 1.0o, 10o and 45o of
the downhill part of the line.

52
 (a) Rate of energy dissipation due to mass transfer

(b) Rate of energy dissipation due to friction

Figure 11: Rate of energy dissipation due to mass transfer and friction against spa-
tial position along the pipe segment s ∈ [40, 100] m for different inclinations β2 =
0o , 0.5o, 1.0o, 10o and 45o of the downhill part of the line.

53
 Highlights

• A new model is proposed to describe vaporous cavitation in hydraulic pipelines.

• It is derived in the light of the thermodynamics of irreversible processes.

• It allows assessment of the rate of energy dissipation associated with the

phenomenon.

• Inert gas is assumed to coexist with vapor, ensuring the thermodynamic condition
to describe the opening and closing of the cavities.

• Numerical simulations illustrate the model ability in reporting slack line flows with
quite different cavity extensions.