Professional Documents
Culture Documents
A Seck FVM PDF
A Seck FVM PDF
Abstract: Although derived from the principles of conservation of mass and momentum, the water-hammer equations integrating dynamic
friction are almost never expressed in conservative form. This is because the pressure and volume discharge are used as variables but these are
not conserved quantities, especially when the one-dimensional velocity profile is distorted from its assumed steady state shape due to the large
accelerations imposed on the fluid particles across the cross section. This paper presents the derivation of the water-hammer equations in
conservation form incorporating dynamic friction. With the dynamic friction taken into account, a source term appears in the basic partial
differential equations as presented by Guinot. The numerical algorithm implements the Godunov approach to one-dimensional hyperbolic
systems of conservation laws on a finite-volume stencil. Two case studies are used to illustrate the influence of the various formulations.
A comparative study between the analytical solution, the numerical solution with quasi-steady friction only, the numerical solution with
dynamic friction, and the measurements has been presented. The results indicate that the dynamic friction formulation reduces the peak water
hammer pressures when compared with a quasi-steady representation. DOI: 10.1061/(ASCE)HY.1943-7900.0001333. This work is made
available under the terms of the Creative Commons Attribution 4.0 International license, http://creativecommons.org/licenses/by/4.0/.
Author keywords: Hydraulic transients; Water hammer; Unsteady friction; Finite volume; Hyperbolic source term; Riemann problem;
Godunov scheme; Wave attenuation.
all types of transient flows and all roughness values. Shamloo and
∂U ∂F
Mousavifard (2015) use a two-dimensional finite-difference method þ ¼S
incorporating the k-ω turbulence model to study the dynamics of ∂t ∂x
turbulence during different periods of water hammer in a polymeric μ Qm 0
U¼ ; F¼ ; S¼ ð1Þ
pipe. However, comparison of their model with experimental tests Qm A0 p −fjVjV
reveal that the model underestimates the peak water-hammer
pressures, which could pose a significant risk for pipe safety. where t ¼ time; x = unit vector in the x-direction; U = flow variable
Although derived from the principles of conservation of mass vector; F = flux vector in the x-direction; S = source term vector;
and momentum, the water-hammer equations integrating dynamic μ = mass of fluid per unit length of pipe; Qm = mass discharge;
friction are almost never expressed in conservation form. This is A0 = cross-sectional area of pipe; p = pressure; V = fluid velocity;
because the presence of the dynamic friction terms preclude pre- and f = friction coefficient dependant on the pipe roughness and
sentation in the integral (conservation) form and also cause a source the fluid viscosity.
term to appear. The finite-volume method that was developed for For unsteady friction in transient flow with a constant pipe
hyperbolic equations in conservation form (e.g., the Euler equa- diameter, the friction factor f is split into the sum of the quasi-
tions in fluid dynamics) has to be adapted to tackle systems that steady fq and unsteady fu part i.e., f ¼ f q þ f u as in the model
are not in proper conservation form. Hyperbolic systems admit by Brunone et al. (1991) and modified by Bergant et al. (2001)
weak solutions in the form of discontinuities or shocks. The treat- pffiffiffiffiffiffi
ment of these discontinuities requires special treatment of the C D ∂V ∂V
flux function to avoid spurious oscillations. A large body of liter- f ¼ fq þ þ asignðVÞ ð2Þ
2jVjV ∂t ∂x
ature is available on finite-volume schemes developed to handle the
solution in the vicinity of these shocks. Among the most modern where a = celerity of the pressure waves; D = diameter of the pipe;
numerical schemes are total variation diminishing (TVD) (Zijlema and C = Vardy’s shear decay constant depending on the flow
and Wesseling 1998), essentially nonoscillatory (ENO) (Harten regime (Reynolds number R) expressed as
et al. 1987), and weighted essentially nonoscillatory (WENO)
schemes (Liu et al. 1994; Jiang and Shu 1995). One limitation 0.00476 if laminar flow
of the TVD schemes is unsatisfactory performance near extrema C ¼ 7.41
if turbulent flow ð3Þ
0.05 Þ
Rlogð14.3=R
(Osher and Chakravarthy 1984; Titarev and Toro 2003; Park et al.
2010). Essentially nonoscillatory and WENO schemes have been Eq. (2) may be rewritten in terms of the flow variables μ and
developed to overcome this limitation and provide a better scheme Qm as
that can handle both sharp interfaces and smooth gradients (Shu
1998). According to Gallerano and Cannata (2011) and Gallerano pffiffiffiffiffiffi
C D 1 ∂Qm Qm ∂μ 1 ∂Qm Qm ∂μ
et al. (2012), ENO and WENO shock capturing higher-order f ¼ fq þ − 2 þ aψ − 2
2jVjV μ ∂t μ ∂t μ ∂x μ ∂x
schemes for the solution of hyperbolic systems can be considered
as a further development of ideas that gave rise to the TVD ð4Þ
schemes. The original contribution of this paper is to present the
derivation of the water-hammer equations in conservation form where ψ is expressed as
incorporating dynamic friction. The hyperbolic system is then
solved using a finite-volume scheme. The Godunov scheme was þ1 if V ∂V
∂x > 0
ψ¼ ð5Þ
retained to obtain the numerical solutions to the present problem. −1 if V ∂V
∂x < 0
It is hoped that this paper will help practical engineers who are
at the detailed design phase to further safely optimize any incorpo-
For turbulent flow, the Colebrook-White equation may be iter-
rated surge controls by assisting in the development and numerical
atively solved for the quasi-steady part f q of the fiction factor
simulation of the water-hammer equations with dynamic friction.
1 k =D 2.51
pffiffiffiffiffi ¼ −2log10 S þ p ffiffiffiffiffi ð6Þ
fq 3.7 R fq
Methodology
In which kS = absolute roughness of the pipe.
This paper focuses on implementation of the Godunov approach to Streeter and Wylie (1993) provide the following expression for
one-dimensional hyperbolic systems of conservation laws that de- the celerity (sonic velocity) a in a circular pipe:
scribe the phenomenon of water-hammer. It is first order accurate in sffiffiffiffiffiffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
both space and time. In the present context, given the simple geom- EV 1
etry, the basic Godunov scheme is considered adequate by the au- a¼ ð7aÞ
ρ 1 þ DE V
thors. First, the computation of the fluxes is detailed followed by a eE
∂U ∂F
þ ¼0 ð12Þ
By taking into consideration the unsteady friction, Eq. (1) may ∂t ∂x
be written as
Downloaded from ascelibrary.org by Indian Institute of Technology Kharagpur on 09/22/19. Copyright ASCE. For personal use only; all rights reserved.
The second step solves for the nonconservative term of Eq. (10)
8 ∂μ
>
> h þ ∂Q
i ¼ p0ffiffiffiffi
m
∂U ∂U
< ∂t
∂x þR ¼0 ð13Þ
∂Qm
þ ∂ 2μ
pffiffiffiffi A p − pffiffiffiffi
C D
aψV ∂μ ð8Þ ∂t ∂x
∂t ∂x 2μþ C D 0 2μþ C D ∂x · · ·
>
> pffiffiffiffi
: þ Cpffiffiffiffi ðaψ − VÞ ∂Q 2μ
pffiffiffiffi
∂x ¼ − 2μþ C D f q jVjV
D
2μþ C D
m
The final step uses the Toro (2001) and Guinot (2003) treatment
of the nonhyperbolic source term of Eq. (10)
In pipe flows, the flow velocity V is negligible compared with a ∂U
(the celerity ¼S ð14Þ
pffiffiffiffiffiof
ffi the pressure waves). Thepffiffiffiffiffiauthors
ffi also note that ∂t
2μ=2μ þ C D ≈ 1 because the term C D is negligible com-
pared with 2μ. Eq. (8) may now be rewritten
8
< ∂μ ∂Qm Guinot’s Solutions at the Internal Cells and at the
pffiffiffiffi ∂t þ ∂x ¼ p
0
ffiffiffiffi Boundaries for the Conservation Terms
∂Qm ∂μ
ðaψÞ ∂Q
: ∂t þ ∂x ðA0 pÞ − 2μþ C D aψV ∂x þ 2μþCpffiffiffiffi
∂ C
p ffiffiffiffi
D D
C D ∂x
m
¼ −fq jVjV
The conservation component of Eqs. (10) and (12) is the same con-
ð9Þ servation component as Eq. (1) in Guinot (2003). For the internal
cells, Guinot (2003) constructs the Riemann problem Uðx; tn Þ and
In conservation form the governing Eq. (9) may be written, presents the solution Unþ1=2
iþ1=2 of this Riemann problem in Eq. (15)
respectively
computing the flux Fnþ1=2 n
iþ1=2 between the time intervals t and t
nþ1
∂U ∂U ∂U
þA þR ¼S where pref is a reference pressure at which the density ρref is
∂t ∂x ∂x known. μref is calculated as μref ¼ A0 ρref .
0 1 0
1 The flux solutions for a prescribed pressure pb at the left-hand
A¼ ¼ ð11Þ
k μ þ a2 0
A0 p
a2 0 (pb;L ) and the right-hand (pb;R ) boundary are respectively given by
nþ1=2 Qnm;1 þ ðμnþ1=2 − μn1 Þa
This is the general form of the water-hammer equations with F1=2 ¼ 1=2 ð17Þ
A0 pb;L
unsteady friction.
In Eq. (11), A ¼ ∂F=∂U is the Jacobian matrix of F with
respect to the matrix U. In the expression for the matrix A, Qnm;N þ ðμnN − μnþ1=2
Fnþ1=2 Nþ1=2 Þa
kA0 p=μ ≪ a2 . The general Eq. (10) is equivalent to the Guinot Nþ1=2 ¼ ð18Þ
A0 pb;R
Eq. (1) if unsteady friction is not considered. The inclusion of
the unsteady friction results in the appearance of the quasi-linear In Eqs. (17) and (18), pb represents a pressure to be prescribed
term R∂U=∂x in the hyperbolic system [Eq. (10)] thus destroying at the boundary. The mass of fluid per unit length of pipe μnþ1=2 (at
1=2
the integral form of the equations. the left hand) and μnþ1=2 (at the right hand) is obtained from the
Nþ1=2
The Godunov method is a conservative numerical scheme for prescribed pressure pb using
solving partial differential equations (PDE) in computational fluid
dynamics. This conservative finite-volume method solves an exact A0
μb ¼ μref þ ðpb − pref Þ ð19Þ
or approximate Riemann problem (depending on the friction for- a2
mulation) at each intercell boundary. It is first order accurate in both
space and time. The next section describes the solution process for The fluxes for a prescribed discharge Qb at the left-hand and the
these equations with the Godunov method. right-hand boundary are respectively given by
Qb μnþ1=2
Nþ1=2 =A0 Solution for the Hyperbolic Source Term at the
Fnþ1=2 ¼ ð21Þ Internal Cells
Nþ1=2
A0 pref þ ðμnþ1=2
Nþ1=2 − μref Þa2
The intermediate solution Uinþ1;x is taken as a starting point for
where μnþ1=2
1=2 in Eq. (20) and μnþ1=2
Nþ1=2 in Eq. (21) are obtained, the computation of the hyperbolic source term. This nonhyperbolic
respectively, from source term is solved according to the procedure detailed
Downloaded from ascelibrary.org by Indian Institute of Technology Kharagpur on 09/22/19. Copyright ASCE. For personal use only; all rights reserved.
subsequently.
Qnm;1 − aμn1 h i
μnþ1=2 ¼ ð22Þ 0 0
1=2 Qb
−a In Eq. (13), R ¼ kaψ . The two eigenvalues λð1Þ and
A0 −V 1
λð2Þ of the matrix R are therefore
Qnm;N þ aμnN
μnþ1=2 0; if ψ ¼ 1
Nþ1=2 ¼ ð23Þ λð1Þ ¼ ð25Þ
Qb
A0 þa −ka; if ψ ¼ −1
ka; if ψ ¼ 1
For all cells, the balance is performed omitting the source term λð2Þ ¼ ð26Þ
using 0; if ψ ¼ −1
0 ¼ 1 ; across dt ¼ ka
dμ dQm dx
" #
μinþ1;x
where Uiþ1=2 ¼ 1
μnþ1;x
0 ¼ 1 ; across dt ¼ −ka
dμ dQm
A20
dx A0 i
Qnþ1;x
for ψ ¼ −1 ð29Þ μnþ1;x μnþ1;x miþ1
1 ¼ V ; across dt ¼ 0
dμ dQm dx iþ1 iþ1
d Qρm ¼ 0; across dxdt ¼ 0 for ψ ¼ 1 ð30Þ By substituting Eqs. (33) and (34) into Eq. (35), the following
dt ¼ ka
dμ ¼ 0; across dx
equation is obtained:
)
¼0; across dt ¼ −ka
dx
dμ kaψΔt nþ1;x kaψΔt
for ψ ¼ −1 ð31Þ Unþ1;h ¼ 1þ Ui − Uiþ1=2 ð36Þ
d ρ ¼ 0; across dx
i
Δxi Δxi
dt ¼ 0
Qm
Fig. 2. Comparison of pressure profiles at pipe midpoint (Δt ¼ 0.0002 s, N ¼ 50) between the various formulations for the friction term
Because the scheme is explicit, the CFL condition for stability must if t → ∞;
be respected, i.e., the maximum permissible time step ΔtMax as pre-
jpinþ1 − pb;L j
sented by Guinot (2003) is as follows: η∞
1 ≈ η ∞
2 ≈ · · · ≈η ∞
N ≈ lim 1 − 100 ≈ 100%
n→∞ pb;L
Δxi ð40Þ
ΔtMax ¼ Min min ; ΔtMax;S ð38Þ
i¼1; : : : ;N a
Resolution Algorithm
Water-Hammer Wave Attenuation Ratio
The proposed algorithm is presented in the following (Fig. 1).
Downloaded from ascelibrary.org by Indian Institute of Technology Kharagpur on 09/22/19. Copyright ASCE. For personal use only; all rights reserved.
Fig. 3. Comparison of discharge profiles (Δt ¼ 0.0002 s, N ¼ 50) between the various formulations for the friction term: (a) pipe midpoint;
(b) downstream of the reservoir
puted. The end of the pipe is considered closed. Initial conditions the objectives of this paper, results were obtained for the analytical
specify all velocities to be zero. In the Guinot example, the friction resolution, the numerical solution with quasi-steady friction only
is assumed constant, consequently the source term does not appear. and the numerical solution with dynamic friction. A time step of
The physical data are given as follows (these constitute the initial Δt ¼ 0.0002 s was used. Comparison between the results indicates
conditions): that the inclusion of all components of the friction has an influence
• Pipe diameter: 100 mm; on the pressure profiles at pipe midpoint (Fig. 2) and the flow os-
• Pipe length: 50 m; cillation profiles at pipe midpoint and at the downstream end
• Absolute roughness of the pipe: 0 mm (smooth pipe); [Figs. 3(a and b)].
Fig. 4. Test no. 1: comparison of pressure profiles (Δt ¼ 0.002 s, N ¼ 100) between the various formulations for the friction term: (a) Section no. 2;
(b) Section no. 3
Fig. 5. Test no. 1: comparison of discharge profiles (Δt ¼ 0.002 s, N ¼ 100) between the various formulations for the friction term: (a) Section no. 1;
(b) Section no. 2
Fig. 6. Test no. 2: comparison of pressure profiles (Δt ¼ 0.002 s, N ¼ 100) between the various formulations for the friction term: (a) Section no. 2;
(b) Section no. 3
Results and Discussion At the middle of the pipe and at the downstream end, inclusion
of the dynamic friction causes the peak pressure to decay at a faster
The water-hammer equations with unsteady friction were solved
rate than that with the quasi-steady friction term only. It is expected
using the Godunov scheme detailed previously. In keeping with
that better agreement between measured and computed dynamic
the objectives of this paper, results were obtained for the analytical
friction profiles would be obtained if other key parameters are prop-
resolution, the numerical solution with quasi-steady friction only,
erly modeled and incorporated into the formulation. The resolution
and the numerical solution with dynamic friction using a time step
will be able to better capture discontinuities in the computed
of Δt ¼ 0.002 s. These results are compared with the experimental profiles.
measurements. Comparison of the results (Figs. 4–7) indicates that
the inclusion of all components of the friction has an influence on
the pressure profiles [Figs. 4(a and b) and 6(a and b)] and the flow Conclusions and Recommendations
oscillation profiles [Figs. 5(a and b) and 7(a and b)]. At the valve
[Figs. 4(b) and 6(b)] the peak pressures are shown as sloping The water-hammer equations in conservation form incorporating
upward in high pressure and downward in low pressure. The dynamic friction is developed and numerically solved using a
quasi-steady friction model and dynamic friction model are able finite-volume formulation. The computational algorithm based
to reproduce almost exactly the evolution of the typical shape of on the Godunov scheme for one-dimensional hyperbolic systems
the pressure oscillations. However, the quasi-steady friction model is presented in some detail. Introduction of the dynamic friction
underestimates the damping and dispersion predicted by the physi- results in the appearance of a source term in the hyperbolic system
cally more accurate dynamic friction model. The dynamic friction of governing partial differential equations. Two case studies have
tends to rapidly attenuate the magnitude of the overpressures. been presented to compare and contrast the separate impacts of
Fig. 7. Test no. 2: comparison of discharge profiles (Δt ¼ 0.002 s, N ¼ 100) between the various formulations for the friction term: (a) Section no. 1;
(b) Section no. 2
steady friction only and of dynamic friction on wave attenuation in interaction, viscoelastic behavior of the pipe wall, leakage, and
time. The findings indicate that inclusion of the dynamic friction blockage have to be properly modeled and incorporated into the
reduces the peak water-hammer pressures when compared with the formulation. This paper may be used as a basis for the study of
standard quasi-steady assumption for the friction. The inclusion of real mass variations in a surge tank. It should also prove useful
the dynamic friction results in a better agreement between calcu- as a tool for the optimal design of surge controls.
lated and measured values. In these two case studies, grid size and
time step were reduced to test for convergence; the numerical sol-
ution presented is the converged solution. Computational time for Acknowledgments
each of the cases studied took only 3 min on a PC (operating sys-
tem: Windows 7, 64 bits; CPU: Intel(R) Xeon(R) CPU E5620 @ The authors gratefully acknowledge the partial financial support of
2.40 GHz x 2 processors; mainboard: ASUS Z8NA-D6(C); Natural Sciences and Engineering Research Council of Canada
memory: 12280 MB triple channel DDR3 ECC @ 1333 MHz). (NSERC).
Thus several variants may be studied in a short period of time.
The efficient algorithm introduced in this paper may be adapted,
with some modifications, to incorporate higher-order schemes that Notation
may yield less computational effort. Because the principal objective
of this paper was to refine the physics of the phenomenon, exten- The following symbols are used in this paper:
sive numerical experimentation with different schemes was not A = Jacobian matrix of the flux F respecting to U;
undertaken. However, for more precision in engineering applica- A0 = cross-sectional area of pipe;
tions, other key parameters such as cavitation, fluid–structure a = celerity of the pressure waves;
kS = absolute roughness of the pipe; pipe flow by local velocity measurement.” J. Hydraul. Eng., 10.1061
N = number of cells in the computational domain; /(ASCE)HY.1943-7900.0000234, 716–726.
p = pressure; Brunone, B., Golia, U., and Greco, M. (1991). “Modelling of fast transients
by numerical methods.” Proc., Int. Conf. on Hydraulics Transients with
Q = volume discharge; Water Column Separation, IAHR, Valencia, Spain, 273–280.
Qm = mass discharge; Chaudhry, M. (1987). Applied hydraulic transients, Van Nostrana Reinhold
R = matrix for the hyperbolic source term; Co., New York.
R = Reynolds number; Chaudhry, M. H., Sabbah, M. A., and Fowler, J. E. (1985). “Analysis and
S = vector source term; stability of closed surge tanks.” J. Hydraul. Eng., 10.1061/(ASCE)
0733-9429(1985)111:7(1079), 1079–1096.
t = time;
Chaudhry, M. H., and Silvaaraya, W. F. (1992). “Stability diagrams for
U = vector variable; closed surge tanks.” Proc., Int. Conf. on Unsteady Flow and Fluid
V = fluid velocity; Transients, HR Wallingford Ltd. and IAHR, Durham, U.K., 221–228.
x = unit vector in the x-direction; Finnemore, E., and Franzini, J. (2002). Fluid mechanics with engineering,
Δt = computational time step; 10th Ed., McGraw-Hill, New York.
Δx = cell size in the x-direction; Gallerano, F., and Cannata, G. (2011). “Central WENO scheme for the in-
tegral form of contravariant shallow-water equations.” Int. J. Numer.
μ = mass of fluid per unit length of pipe; and Methods Fluids, 67(8), 939–959.
ρ = mass density. Gallerano, F., Cannata, G., and Tamburrino, M. (2012). “Upwind WENO
scheme for shallow water equations in contravariant formulation.”
Comput. Fluids, 62, 1–12.
Subscripts Guinot, V. (2003). Godunov-type schemes: An introduction for engineers,
b = value to be prescribed at a boundary; Elsevier, Amsterdam.
b, L = value to be prescribed at the left-hand boundary; Harten, A., Engquist, B., Osher, S., and Chakravarthy, S. R. (1987).
“Uniformly high order accurate essentially non-oscillatory schemes,
b, R = value to be prescribed at the right-hand boundary; and III.” Upwind and high-resolution schemes, Springer, New York,
i = cell number. 218–290.
Jaeger, C. (1933). “Théorie générale du coup de bélier.” Diss. Techn. Wiss.
ETH Zürich, Zürich, Switzerland.
Superscripts Jiang, G. S., and Shu, C. W. (1995). “Efficient implementation of
n = time level; weighted ENO schemes.” Technical Rep. ICASE 995-73, ICASE,
Hampton, VA.
n, h = numerical solution obtained at the time level n after the
Kim, S. H. (2010). “Design of surge tank for water supply systems using
solution of the hyperbolic source term of the equation;
the impulse response method with the GA algorithm.” J. Mech. Sci.
n, x = numerical solution obtained at the time level n after the Technol., 24(2), 629–636.
solution of the conservation part of the equation in the Lescovich, J. E. (1967). “The control of water hammer by automatic
x-direction; valves.” J. Am. Water Works Assoc., 59(5), 632–644.
n þ 1=2 = Average value between the time level n and n þ 1; Liu, X. D., Osher, S., and Chan, T. (1994). “Weighted essentially non-
∞ = time level if t gets larger; and oscillatory schemes.” J. Comput. Phys., 115(1), 200–212.
* = intermediate region in the solution of the Riemann Menabrea, L. F. (1858). Note sur les effets du choc de l’eau dans les
conduites, Mallet-Bachelier, Paris.
problem.
Michaud, J. (1878). “‘Coups de bélier dans les conduites, Étude des moy-
ens employés pour en atteneur les effects.” Bull. Soc. Vaudoise Ing.
Arch., 4(3), 4.
References Michaud, J. (1903). Intensité des coups de bélier dans les conduites d’eau,
H. Vallotton & Toso, Lausanne, Switzerland.
Allievi, L. (1903). “Teoria generale del moto perturbato dell’acqua nei tubi Moghaddam, M. A. (2004). “Analysis and design of a simple surge tank
in pressione (colpo d’ariete) [General theory of the variable motion of (research note).” Int. J. Eng. Trans. A: Basics, 17, 339–345.
water in pressure conduits.].” Annali della Società degli Ingegneri ed Osher, S., and Chakravarthy, S. (1984). “High resolution schemes and the
Architetti Italiani, 17(5), 285–325 (in Italian). entropy condition.” SIAM J. Numer. Anal., 21(5), 955–984.
Allievi, L. (1913). Teoria del colpo d’ariete, Tipografia della R. Accademia Park, J. S., Yoon, S. H., and Kim, C. (2010). “Multi-dimensional limiting
dei Lincei, Rome. process for hyperbolic conservation laws on unstructured grids.”
Allievi, L. (1932). Il colpo d’ariete e la regolazione delle turbine, Industrie J. Comput. Phys., 229(3), 788–812.
grafiche italiane Stucchi, Milano, Italy. Ramadan, A., and Mustafa, H. (2013). “Surge tank design considerations
Anderson, A. (1976). “Menabrea’s note on waterhammer: 1858.” for controlling water hammer effects at hydro-electric power plants.”
J. Hydraul. Div., 102(1), 29–39. Univ. Bull., 15(3), 147–160.
Bergant, A., Ross Simpson, A., and Vìtkovsk, J. (2001). “Developments in Roche, E. (1975). “Assainissement rural: Protection des conduites de
unsteady pipe flow friction modelling.” J. Hydraul. Res., 39(3), 249–257. refoulement.” TSM l’Eau, 365–378.
New York. Zijlema, M., and Wesseling, P. (1998). “Higher-order flux-limiting schemes
Streeter, V. L., and Wylie, E. B. (1993). Fluid transients in systems, for the finite volume computation of incompressible flow.” Int. J.
Prentice-Hall, Englewood Cliffs, NJ. Comput. Fluid Dyn., 9(2), 89–109.