Sie sind auf Seite 1von 11

Energy Conversion and Management 78 (2014) 776786

Contents lists available at ScienceDirect

Energy Conversion and Management


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

Numerical simulation and experimental validation of the cavitating ow


through a ball check valve
Jos R. Valds a,, Jos M. Rodrguez a, Ral Monge a, Jos C. Pea a, Thomas Ptz b
a
Instituto Tecnolgico de Aragn, Mara de Luna 8, 50018 Zaragoza, Spain
b
TRW Automotive, Carl Spaeter Strasse, 56070 Koblenz, Germany

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

Article history: The main objective of this work is to perform and validate a series of CFD simulations of the cavitating
Received 20 September 2013 ow through a ball check valve. Experimental tests are performed in order to obtain the mass ow rate
Accepted 25 November 2013 through the valve under different operating conditions, inducing or preventing the appearance of cavita-
Available online 24 December 2013
tion by conveniently adjusting the pressure level on the valve outlet port. The measurements are com-
pared with the results of numerical (CFD) simulations of the uid ow through the valve, with and
Keywords: without the inclusion of a cavitation model. The characteristic ow coefcient of the valve and the
Cavitation
hydraulic forces on the ball are analysed.
Valves
Fluid ow
2013 Elsevier Ltd. All rights reserved.
Numerical simulation
CFD
Testing
Validation

1. Introduction ow inside the valve, as well as the ow coefcient for its subse-
quent application in design and validation phases; for example,
The aim of this work is to run and validate numerical CFD sim- in the development of lumped parameter models of complete
ulations of the liquid ow through a ball check valve with and hydraulic systems for control, performance or stability analysis
without cavitation. The ultimate aim is to validate the accuracy [1]. In many cases, because the valve must work with outlet pres-
of the cavitation model used in the CFD simulations that have been sure levels that are close to the atmospheric pressure, the appear-
run in order to reproduce the incompressible, turbulent ow ance of cavitation is not avoidable. However, it is necessary to
through a ball check valve used in the Hydraulic Control Unit of know it and take into account its inuence on the mass ow rate,
an automotive ABS system. Depending on the operating conditions ow forces and ow coefcients. The inception of cavitation yields
of the valve, cavitation phenomena may or may not appear, yield- additional energy losses and density variations that can signi-
ing signicant differences in the mass ow rate through the valve cantly decrease the mass ow rate, rendering the uid discharge
and thus in the ow coefcient of said valve. process less efcient. The ow forces acting on the moving parts
Cavitation is a phase transformation that occurs in liquid ows are also modied due to the limitation of the minimum static pres-
when the local pressure drops below the saturation pressure, with sure that can be reached. In hydraulic control valves, having differ-
gas cavities that appear, develop and nally collapse when the ent forces for similar pressure differences is a fact that must be
pressure is high enough. In hydraulic components, cavitation accurately known in order to correctly design the control software.
may lead to problems such as vibrations, pressure pulsations, noise In check valves, the net forces acting on the ball must be analysed
and erosion on solid surfaces. Therefore, cavitation and its poten- in order to ensure that the valve opens when necessary.
tial effects must be considered in the design process of hydraulic Existing cavitation models fall into two classes: the Volume Of
systems. Fluid method based on the interface tracking and the homoge-
Valves are generally used to control the ow rate in hydraulic neous equilibrium ow method [2]. The VOF method deals with
systems. Ball check valves are designed to open under certain con- bubble dynamics by solving the vapourliquid interface, but most
ditions and allow an alternative uid discharge path. In the design of the practical cavitating ows are approached using the
of such valves, it is important to know the characteristics of the homogeneous ow theory, due to its reduced complexity and
computational cost. In this theory, the uid is considered as a
Corresponding author. Tel.: +34 976011160. vapourliquid mixture without explicit phase interfaces containing
E-mail address: jrvaldes@ita.es (J.R. Valds). a large number of spherical bubbles where the liquidvapour mass

0196-8904/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.enconman.2013.11.038
J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786 777

Nomenclature

A cross-sectional area Greek symbols


d contraction small diameter a vapour volume fraction
D contraction large diameter D increment
K loss coefcient / valve orice diameter
n bubble density n ow coefcient
P static pressure l viscosity
Pin inlet pressure q density
PT total pressure
Q volume ow Subscripts
Rb bubble radius l liquid phase
V velocity v vapour phase
b bubble

transfer is governed by a vapour transport equation. This equation changes and the Standard ke approach to model turbulent uctu-
contains two source terms that account for the mass transfer and ations. An experimental validation of the mass ow rate as a func-
that are modelled based on the RayleighPlesset equation [3], tion of the pressure difference showed a good correlation using the
which describes the growth of a single vapour bubble in a liquid. two-phase ow, but not so with a one-phase ow. Mimouni et al.
Many cavitation models based on the homogeneous ow theory [12] performed an analysis of the numerical simulation of cavita-
have been put forward in recent years, such as Singhal et al. [4], tion phenomena with the NEPTUNE CFD code, which is based on
Schnerr and Sauer [5], Kunz et al. [6] or Zwart et al. [7]. the resolution of the mass, momentum and energy balance for both
The full cavitation model developed by Singhal [4] accounts liquid and vapour phases, and the occurrence of cavitation by
for all rst order effects (phase change, bubble dynamics, turbulent nucleation at the wall or by pre-existing cavitation nuclei. The
pressure uctuations and non-condensable gases dissolved in the model was validated by comparison with experimental measure-
uid). The expressions for the phase-change rates, which depend ments of the void fraction in a critical water ow in a nozzle,
upon the local ow conditions as well as the uid properties, are obtaining a good agreement under the assumption that most of
derived from a reduced form of the RayleighPlesset equation for the cavitation nuclei come from the vapour micro bubbles gener-
bubble dynamics. The rate expressions employ two empirical con- ated at the wall. Computations of cavitation development down-
stants, calibrated with experimental data covering a very wide stream an orice showed also good agreement with experimental
range of ow conditions. visualizations.
The SchnerrSauer model [5] follows a similar approach to Li et al. [13] presented a study using a modied k-x model to
derive the expression for the net mass transfer from liquid to predict the unsteady cavitating ows around 2D and 3D hydrofoils
vapour, but uses a different expression to connect the vapour and modelling cavitation with the SchnerrSauer cavitation model.
volume fraction to the number of bubbles per volume of liquid. The results were qualitatively in agreement with experimental
The ZwartGerberBelamri model [7] uses a mass transfer expres- observations of formation and transport of cavitating vortices,
sion which is similar to that of Singhals, except that it is related to but under-predicted the lift coefcients. Zhao et al. [2] performed
the vapour phase density, but not to the liquid phase and mixture a numerical simulation and validation of the cavitating ow on a
densities. It also employs two empirically calibrated coefcients 2D NACA0015 hydrofoil under high pressure and temperature.
for evaporation and condensation and a correction to the vapour The Singhal cavitation model was adopted combined with an
volume fraction to account for the incorrect assumption that the improved RNG ke turbulence model.
cavitation bubbles do not interact with each other, which is only These analyses and validations indicate that CFD simulations
true during the earliest stages of cavitation when the bubble grows using homogeneous cavitation models are a valid tool to investi-
from the nucleation site. gate cavitating ows. All three main ke approaches for modelling
Experimental measurements and CFD simulations of the cavi- turbulence (standard, RNG and Realizable) seem to provide good
tating ow in different hydraulic systems are widely reported in approximation to the measured ows. However none of the previ-
the literature. Chern et al. [8] performed an analysis, based on ously mentioned works apply the SchnerrSauer cavitation model
experimental observations, of the ow through a ball valve and for valve ow simulation, neither do other more recent works such
the appearance of cavitation. Jia et al. [9] performed an analysis as those of Shang [14], who has applied and validated the Schnerr
of the cavitating ow in a conical spray nozzle by means of numer- Sauer model in external ows around blunt bodies of submarine
ical CFD simulations using Singhals cavitation model and the Real- shape; Aung and Li [15] have applied the Singhal model to an elec-
izable ke turbulence model. The mass ow results are compared tro-hydraulic servo valve; Mohan et al. [16] have coupled the Schn-
with those of experimental measurements, showing a fairly good errSauer model to a spray model for simulating a fuel spray; Li
correlation, with some signicant differences for high pressure dif- et al. [17] have applied the SchnerrSauer model to the cavitating
ferences, due to the use of a 2D model. Salvador et al. [10] per- ow around a hydrofoil and Zhang and Chen [18] have applied the
formed a study of the internal ow in diesel injector nozzles ZwartGerberBelamri model to investigate the cavitating ow
modelling the cavitating ow as a homogenous mixture of liquid within a slanted axial-ow pump.
and vapour and an RNG ke model for modelling turbulence. Before The present work aims to validate the SchnerrSauer model for
carrying out an analysis of the inuence of the needle lift by means the ow through a ball check valve. Furthermore, no papers have
of CFD, a validation of the mass ow rate and momentum ux is been located that provide numerical and experimental results for
performed at full needle lift conditions, showing a good agreement. a hydraulic valve with and without cavitation, under the same
Casoli et al. [11] performed a CFD analysis of a homogenizing valve pressure jump conditions. This paper will show numerical and
using a two-phase ow, the Singhal model to describe phase experimental results of the mass ow through a ball valve, under
778 J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786

a given set of pressure jumps, with and without cavitation, analys- An unstructured tetrahedral mesh of 700,000 cells has been
ing the differences and demonstrating that the SchnerrSauer built. Although this type of mesh could lead to less accurate results
cavitation model is able to capture the difference between both. than a structured hexahedral mesh, the complexity of the geome-
The way to turn off cavitation in the experimental tests is by try in the ball region makes it difcult to use hexahedral cells.
raising the pressure level of the valve outlet port to a sufciently Building a tetrahedral mesh allows us to reduce the mesh setup
high level. time while facilitating the use of a large number of small cells in
Lumped parameter models of check, poppet and spool valves the critical region that exists between the ball and the housing.
are widely used for the dynamic simulation of hydraulic systems Furthermore, the mesh can be easily rened by adapting the cells
[1]. The mass ow through the valve is computed from the pres- by pressure or velocity gradients. Thus, mesh renement studies
sure upstream and downstream the valve, the area of the critical are conducted by rening the mesh in regions of high pressure gra-
restrictions and the characteristic ow coefcient of the valve. dients, in order to ensure that the mass ow rate result is indepen-
The ow coefcient depends on the Reynolds number and the geo- dent of the mesh. Fig. 2 shows the original mesh and the rened
metric features of the ow path, except for high Reynolds number mesh.
ow, in which the coefcient is a constant which depends only on The valve is designed so that it opens completely even for small
the geometry of the ow passage [1922]. Sometimes a constant DPs. There is no spring, so the ball is thrown against the opposite
ow coefcient is used for modelling purposes, with the corre- wall as soon as the inlet pressure is higher than the outlet pressure.
sponding loss of accuracy [19], [23]. This work will also show that Therefore, stationary simulations are run for different DPs assum-
not only the Reynolds number, but also the cavitation conditions ing that the ball is in the position shown in Fig. 2. As will be shown
have a signicant inuence on the ow coefcient, and must be in Section 4, the ow values correlate well with the experimental
taken into account to increase the accuracy of lumped parametric ones, which is an indication that the ball position is correct. The
models. net force on the ball, derived from the CFD simulations, points
downwards for all the simulated DPs. For other possible positions
of the ball, this downward force would be even higher as the high
pressure region on top of the ball would be larger than the one
2. Description of the valve. CFD model
shown in Fig. 7, due to the restriction between the ball and the
housing. To be completely sure, CFD simulations for different ball
The ball check valve under study is situated within a control
positions (or a full transient dynamic simulation of the ball motion
valve located in the Hydraulic Unit of an ABS system of an automo-
with deforming mesh) should be run, but the computational effort
bile (see Fig. 1). The aim of this valve is to allow or cut the ow be-
would be high and the assumption that the ball is at the farthest
tween the brake cylinder and the brake calliper. In normal
position seems very likely.
operation, the ball check valve is closed, and all the uid ows
Pressure differences between 0.5 and 100 bar have been simu-
through the valve main restriction. When the uid ows from
lated. The resulting Reynolds numbers at the valve orice are
the brakes, the check valve opens in order to facilitate the pressure
between 400 and 7000 (see Table 1). Some of the cases lie within
release in the brake callipers. The diameter of the ball is 1.4 mm,
the turbulent regime, while others belong to the laminar or transi-
which gives an idea of the reduced dimensions of the valve.
tion regimes. The turbulence model used in this analysis is based
CFD simulations using the commercial software ANSYS/FLUENT
on the RANS approach where the instantaneous velocity is divided
are performed in order to obtain the mass ow rate through the
into a mean velocity and a uctuation. The so called Reynolds
valve as well as the pressure, velocity and phase contours. The con-
stresses that appear when this division is introduced in the
tinuity, momentum and turbulence generation and destruction
NavierStokes equations are modelled using the Bousinessq
equations are solved by means of the Finite Volume commercial
hypothesis [24].
code FLUENT [24], assuming the hypotheses of steadystate,
While the Standard ke model [25] is valid only for high-
incompressible and isothermal ow. The uid is a typical brake
Reynolds numbers, the RNG [26] and Realizable [27] ke models
liquid, which can be considered Newtonian, and has the following
provide an analytical formula for effective viscosity that accounts
properties at room temperature: density 1055 kg/m3 and viscosity
for low Reynolds number effects. The ball valve cases have been
0.017 kg/m s. The saturation pressure of the uid, considering an
solved using the Realizable ke model, and for Re < 1500, they have
air content of 3%, is 0.22 bar of absolute pressure.

Fig. 1. Location of the ball check valve within the control valve: closed (left) and open (right).
J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786 779

Fig. 2. Original mesh (left) and rened mesh (right).

Table 1 able to resolve the laminar sublayer, the Enhanced Wall treatment
Mass ow rate values obtained with different turbulent models at low Reynolds [24] solves the whole turbulent boundary layer, including the
numbers.
viscous sublayer. However, the use of a very ne near-wall mesh
DP (bar) Re Mass ow (gr/s) might impose a large computational requirement. The Enhanced
Realizable ke RNG ke Transition SST Laminar ow Wall treatment allows to have a near-wall formulation that can
0.5 412 3.66 3.66 3.68 3.70
be used with coarse meshes as well as ne meshes, without incur-
1 623 5.54 5.52 5.54 5.60 ring in excessive error for intermediate meshes where the rst
5 1502 13.36 13.34 13.54 13.67 node is neither in the fully turbulent region nor at y+ = 1.
10 2170 19.30 19.29 19.37 19.60 The resulting y+ values on the walls for the most critical case
(DP 100 bar) are below 15, as can be seen in Fig. 3. The use of
the Enhanced Wall treatment should ensure accurate results. How-
also been solved considering laminar ow. The differences are ever, an y+-based mesh renement has been performed in order to
lower than 2% (see Table 1). Solving these cases with the RNG analyse the variation in the ow results. Rening the near-wall
ke model or with the transition SST model does not yield signi- mesh so that the y+ is below 11 yields a ow reduction of 0.6%,
cant differences either. and a further renement that lowers it below 7 (see Fig. 4) yields
Turbulent variables have been set by specifying the hydraulic an additional reduction of 1%. Therefore, the original mesh with
diameter of the inlet port and a typical value of the turbulent y+ < 15 is deemed adequate.
intensity of 5%. Cavitation is modelled using the Scherr and Sauer cavitation
The simulations have been carried out using the Enhanced Wall model [5], [24]. Both the Singhal and the SchnerrSauer models
treatment that combines the two-layer near wall model with en- follow a quite similar approach and both are implemented in
hanced wall functions. If the near-wall mesh is ne enough to be ANSYS FLUENT. The main advantage of the Singhal model is that

Fig. 3. y+ Values on the ball wall (left) and housing walls (right).
780 J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786

Fig. 4. y+ Values on the ball (left) and housing walls (right) after two renements of the near-wall mesh.

it can be used to account for the effect of non-condensable gases, Different static pressure levels are imposed at the inlet and out-
with a mass fraction that is assumed to be a known constant. let ports of the valve, in order to evaluate the mass ow rate across
The SchnerrSauer model is the default model, does not have a range of pressure jumps between 0.5 and 100 bar. According to
empirically calibrated constants and its numerical implementation the incompressible ow theory, the ow rate is determined by
is more robust, stable and easy to converge than the Singhal model the net pressure difference. However, if cavitation is taken into ac-
[24]. count, the ow rate depends not only on the net pressure differ-
As in the Singhal model, in the SchnerrSauer model the liquid ence, but also on the intensity of the cavitation effects, which is
vapour mass transfer is governed by the vapour transport directly related with the outlet pressure level.
equation: The SIMPLE scheme [28] is used for the pressurevelocity cou-
pling, and second order upwind spatial discretization is used.
@ ! ! q q da
aqv r aqv V v v l 1
@t q dt
3. Experimental setup
and the simplied bubble dynamics equation derived from the Ray-
leighPlesset equation [3] A hydraulic test rig (see Fig. 5) has been designed and developed
 1=2 in order to measure pressure and mass ow rate in valves and
dRb 2 Pb  P other hydraulic components. The gure on the left shows a picture
2
dt 3 ql of the complete test bench, with the reservoir, the pump and the
electronic equipment. A specic tooling has been designed in
where v is the vapour phase, a is the vapour volume fraction, qv is which different types of valves can be installed for testing, con-
the vapour density and Vv is the vapour phase velocity. The term on necting the inlet and outlet ports to the rest of the hydraulic circuit,
the right hand side of the vapour transport equation is the mass avoiding the apparition of leakages. A detail of the testing tool can
source term, where q is the mixture density and ql is the liquid be observed in Fig. 5 (right).
phase density. In the bubble dynamics equation, Rb is the bubble ra- A scheme of the hydraulic circuit can be seen in Fig. 6. The uid
dius, Pb is the bubble surface pressure (approximated by the satura- is pumped from a reservoir into the circuit where the valve is in-
tion vapour pressure) and P is the local far-eld pressure stalled. For controlling the ow level, a proportional control valve
(approximated by the cell centre pressure). is installed at the outow of the pump. The test bench can generate
The particularity of the SchnerrSauer model is the expression input pressure levels of up to 250 bar at the sample inlet, and ow
used to connect the vapour volume fraction to the number of bub- rates of up to 15 l/min.
bles per volume of liquid For characterizing the valve, it is necessary to measure the pres-
sure at both sides of the valve and the ow passing through it. This
n 43 pR3b
a 3 is done by means of two pressure sensors at the inow and outow
1 n 43 pR3b ports, with operation ranges up to 250 bar and 100 bar, respec-
tively. The ow meter is placed before the inow pressure sensor,
where n is the number of bubbles per volume. Working with the
and has a measuring range of 0.021.8 l/min.
previous expressions, the mass source term can be written as
In order to achieve different levels of outlet pressure during the
 1=2 experimental tests, orices with different diameters are installed
qv ql 2 jPv  Pj 3
a1  a   4 at the outow of the valve, downstream the outow pressure sensor.
q 3 ql a 3 1 1=3
Fig. 7 shows a detail of this sort of orices, which have been manufac-
1a 4p n
tured by means of laser technology from stainless steel plates
In this model the only parameter is the number of bubbles per vol- 0.5 mm thick and are placed between two hydraulic tube adapters.
ume of liquid. Typical values of the bubble density are between 1012
and 1014 bubbles/m3 [5]. The default value in FLUENT is 1013 bub-
bles/m3. Increasing or decreasing this value by an order of magni- 4. Results
tude changes the mass ow rate in less than 0.5%. A similar
conclusion is reached for the steady cavitating ow around a 2D Fig. 8 shows the comparison between the measured mass ow
hydrofoil [2]. rate and the values predicted by the CFD simulation when the
J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786 781

Fig. 5. Hydraulic test rig (left) and detail of the valve tooling and the pressure sensors (right).

Pressure Pressure
Pressure Gauge Flow Proportional Sensor 1 Sensor 2
control valve
Flow Sensor

Pressure
relief valve
Testing Tool

Motor Pump
Manual flow control
valve
Filter
Bundy Pipes

Tank

Fig. 6. Diagram of the experimental setup.

Inlet pressure
sensor

Outlet pressure sensor

Fig. 7. Orice used for raising the outlet pressure level.


782 J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786

to compute the difference in the whole pressure range, a power


function that ts the CFD results with a correlation coefcient R2
of 99.9% is used to estimate the numerical results for all the pres-
sure values that have been experimentally registered (see Fig. 9).
Fig. 10 shows the static pressure contours and Fig. 11 shows the
liquid phase contours for two of the cases that have been run with
atmospheric outlet pressure and cavitation. The intensity of the
cavitation effects can be observed. They appear in the narrow
restriction between the ball and the housing where the uid is
accelerated as well as in those regions where intense energy dissi-
pation losses take place. These regions are the ow separation
zones in the sharp bends upstream the ball and the wake down-
stream the ball, and can be clearly observed in Fig. 12, which shows
the total pressure in the domain.
Cavitation can be eliminated by raising the value of the outlet
Fig. 8. Mass ow rate results (atmospheric outlet pressure). pressure to a sufciently high level. This is achieved by inserting
a 0.7 mm diameter orice downstream the outlet port. The result-
ing outlet pressure levels are shown in Fig. 13.
These values of the outlet pressure are high enough to avoid the
appearance of cavitation effects. From the CFD simulations per-
formed without activating the cavitation model, by registering
the minimum pressure value in the domain, we can deduce the
value of the outlet pressure above which there should be no cavi-
tation effects (see Fig. 14 and Table 2). For example, the right hand
picture of Fig. 14 shows a lowest static pressure of 40 bar. As the
ow is incompressible, this value is not an absolute pressure value
(which would have no physical meaning), but rather the pressure
relative to the zero reference pressure imposed at the valve outlet.
The minimum absolute pressure with physical meaning is the
saturation pressure of 0.22 bar (0.8 bar below the atmospheric
pressure). This means that if the absolute outlet pressure is higher
than 41 bar, the static pressure will be higher than 1 bar every-
where and there will be no cavitation. For lower outlet pressures,
Fig. 9. Power function used to t the CFD results (atmospheric outlet pressure). cavitation effects may appear and therefore a cavitation model
should be included in order to limit the minimum pressure to
the saturation pressure value.
outlet port is at atmospheric pressure. In this case, CFD simulations Fig. 15 shows the comparison between the measured mass ow
are run with the cavitation model activated. rate and the values predicted by the CFD simulation when the
Fig. 8 shows a very good correlation, with differences between pressure at the outlet port has been raised by inserting a 0.7 mm
the experimental results and the CFD values well below 5%, in fact orice. The CFD simulations have been performed in this case
below 2% in most cases (the solid line on the bottom of the graph without activating the cavitation model.
shows the% difference between the numerical and the experimen- The agreement is also excellent, with differences between the
tal values). The differences are shown in absolute values. In order experimental results and the CFD values below 5%, in fact below

Fig. 10. Static pressure (bar) atmospheric outlet: DP 20 bar (left) and DP 50 bar (right).
J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786 783

Fig. 11. Liquid phase volume fraction atmospheric outlet: DP 20 bar (left) and DP 50 bar (right).

Fig. 12. Total pressure (bar) atmospheric outlet: DP 5 bar (left) and DP 50 bar (right).

2% in many cases, except for some of the values in the lower DP


range.
For a given pressure jump, the appearance of cavitation phe-
nomena can signicantly change the value of the mass ow rate.
Fig. 16 puts together the experimental results, with and without
orice, and the CFD results, with and without cavitation model.
The excellent agreement between tests and simulations can be ob-
served, as well as the signicant difference in mass ow rate be-
tween both situations. Without cavitation, the mass ow rate is
up to 20% higher (see Fig. 16).
According to the theory of internal incompressible ow [19],
[23], the major energy losses in valves are not the linear viscous
losses typical of straight pipes, but the secondary losses that occur
in contractions, expansions, bends, etc. The ow conguration in
these elements is usually quite complex, and the energy losses
are commonly measured experimentally or numerically (CFD mod-
Fig. 13. Outlet pressure caused by the insertion of an orice. els) and correlated with the ow parameters. The loss coefcient K
784 J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786

Fig. 14. Static pressure (bar) without cavitation model: DP 5 bar (left) and DP 50 bar (right).

Table 2
Minimum outlet pressure needed for avoiding cavitation and actual pressure
measured in the tests with orice at the valve outlet.

DP (bar) Minimum outlet pressure (bar) Test outlet pressure (bar)


100 70 Not performed
50 40 66
20 15 23
10 6 8.6
5 3 3

Fig. 16. Mass ow rate results (with and without cavitation).

high Reynolds number ow, the loss coefcient is a constant which


depends only on the geometry of the ow passage.
In some simple cases there exist expressions that agree well
with the experiments [23]. For example, for sharp contractions:
Fig. 15. Mass ow rate results (without cavitation). !
2
d
K 0:42  1  6
D2
is usually given as the ratio of the total pressure drop through the where d is the diameter of the small duct and D is the diameter of
device to the kinetic energy: the large duct. For sharp expansions:
!2
 2 2
1 Q d
DP K  q  5 K 1 7
2 A D2

These expressions are valid for fully turbulent regime. For laminar
where q is the uid density, Q is the volume ow rate, A is the and transition regimes, the viscous forces start to gain importance
cross-sectional area of the restriction and K is the ow resistance and K varies with Re. As can be seen in [20], the loss coefcient com-
coefcient associated to the element, which is dependent on the puted at the contraction of a hydraulic valve tends asymptotically
Reynolds number and the geometric features of the ow path. For towards the value computed with Eq. (6).
J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786 785

Another way to put Eq. (5) is


s
2  DP
Q An 8
q
where n is the so-called the ow coefcient and represents the ow
reduction with respect to the maximum ow that can be obtained
for a given pressure jump across a given restriction of area A.
In dynamic simulation models of hydraulic systems, the ow
through a restriction is calculated from the pressure upstream
and downstream the restriction and therefore the ow coefcient
is one of the main parameters of the ow calculation block, along
with the restriction area and the uid density.
Sometimes the dependence of K or n on Re is ignored and a con-
stant ow coefcient is used to calculate the ow rate in valves, Fig. 18. Flow coefcient vs. cavitation number.
orices and other hydraulic elements for low Reynolds number
ows and even for slightly compressible ows [19]. However, to
accurately calculate the ow for any regime, the ow coefcient
should be expressed as a function of Re. In the ball check valve,
the Re number has been calculated as

4qQ
Re 9
lp/
where q is the uid density, Q is the ow rate, l is the viscosity and
/ is the inlet diameter of the valve (see Fig. 14).
Not only the dependence of the ow coefcient on the Reynolds
number, but also the consideration of the inuence of cavitation is
important for the accurate modelling of the valve. The ow coef-
cient of the valve, calculated from Eq. (8) and shown in Fig. 17, is sig-
nicantly different with and without cavitation. If the outlet
pressure is high enough to prevent cavitation, the ow coefcient
Fig. 19. Forces on the ball (square markers: atmospheric outlet, diamonds: high
follows the normal trend that can be found in the literature
outlet pressure; empty markers: viscous force; crossed markers: pressure force; full
[20,21,29]: linear increase with Re for purely viscous ow, and markers: net force).
asymptotical increase in the transient regime towards the constant
value typical of fully turbulent regime. As the outlet pressure level
decreases, the trend is quite different: at low Reynolds number, drops towards the constant value typical of high Reynolds numbers
the behaviour is similar due to the low intensity of cavitation, but and full cavitation conditions.
drops sharply once cavitation effects become intense, yielding addi- Thus, in order to develop parametric models of valves as accu-
tional energy losses and thus a reduced ow rate for a similar DP. rately as possible, different functions n(Re, Cv) should be consid-
The inuence of cavitation on the ow coefcient can be estab- ered, in order to cover the whole range of situations.
lished as a function of the degree of cavitation through the cavita- Finally, the CFD simulation also allows us to analyse the net
tion number, dened as [16]: forces acting on the moving parts of the valve. This knowledge is
essential for the design of the control software in control valves
Pin  Pv
Cv 10 and for the analysis of the opening and closing conditions in check
DP valves. In this case, the net forces acting on the ball are very differ-
where Pin is the pressure at the valve inlet, Pv is the uid saturation ent under cavitating and non-cavitating conditions, due to the dif-
pressure and DP is the pressure difference between the inlet and ferent static pressure distribution around the ball and the viscous
outlet ports. As can be observed in Fig. 18, there is a critical cavita- forces associated with the different ow rates. Fig. 19 shows the
tion number below which cavitation starts and the ow coefcient net forces for the two extreme situations, that is, atmospheric out-
let and high outlet pressure. Both the viscous and pressure forces
act downwards, trying to open the valve. Under full cavitation,
the minimum static pressure is limited and the pressure forces in-
crease rapidly with DP, with a small contribution from the viscous
forces, which also increase with DP. Under non-cavitation condi-
tions, the lowest static pressure is not limited, and therefore the
pressure forces are much lower, although the viscous forces are
higher due to the larger ow rate. The result is that, with no cavi-
tation, the net forces increase much more slowly. This different
behaviour of the net forces must be taken into account when
designing the check valve.

5. Conclusions

A series of numerical CFD simulations of the ow through the


Fig. 17. Flow coefcient of the ball check valve vs. Reynolds number. ball check valve of an ABS control valve have been performed, with
786 J.R. Valds et al. / Energy Conversion and Management 78 (2014) 776786

and without the inclusion of a cavitation model, under similar con- [9] Jia M, Xie M, Liu H, Lam W-H, Wang T. Numerical simulation of cavitation in
the conical-spray nozzle for diesel premixed charge compression ignition
ditions of pressure difference between the inlet and outlet ports. In
engines. Fuel 2011;90(8):265261.
parallel, using a test rig specically designed and built for testing [10] Salvador FJ, Martnez-Lpez J, Caballer M, De Alfonso C. Study of the inuence
hydraulic valves, a batch of tests have been run, measuring the of the needle lift on the internal ow and cavitation phenomenon in diesel
mass ow rate through the valve as a function of the pressure injector nozzles by CFD using RANS methods. Energy Conver Manage
2013;66:24656.
jump. Cavitation has been induced by setting the outlet valve pres- [11] Casoli P, Vacca A, Berta GL. A numerical procedure for predicting the
sure at atmospheric level or eliminated by raising the outlet pres- performance of high pressure homogenizing valves. Simul Modell Practice
sure up to a high enough value to prevent the appearance of gas Theory 2010;18(2):12538.
[12] Mimouni S, Boucker M, Laviville J, Guel A, Bestion D. Modelling and
phase. The correlation between the measurements and the CFD computation of cavitation and boiling bubbly ows with the NEPTUNE_CFD
predictions is excellent in both cases, thus validating the accuracy code. Nucl Eng Des 2008;238(3):68092.
of the simulations and the cavitation model for valve ow predic- [13] Li D, Grekula M, Lindell P. Towards numerical prediction of unsteady sheet
cavitation on hydrofoils. J Hydrodyn Ser B 2010;22(5):7416 [Supplement 1].
tion. The signicant difference between the cavitating and non- [14] Shang Z. Numerical investigations of supercavitation around blunt bodies of
cavitating conditions shows the importance of taking into account submarine shape. Appl Math Modell 2013;37(2021):883645.
the cavitation effects in the prediction of the ow rate. The ow [15] Aung NZ, Li S. A numerical study of cavitation phenomenon in a apper-nozzle
pilot stage of an electrohydraulic servo-valve with an innovative apper
coefcient function has been extracted as a function of the shape. Energy Conver Manage 2014;77(January):319.
Reynolds number and the cavitation number, in order to [16] Mohan B, Yang W, kiang Chou S. Development of an accurate cavitation
characterize the valve as accurately as possible in lumped coupled spray model for diesel engine simulation. Energy Conver Manage
2014;77(January):26977.
parameter simulation blocks. The hydraulic forces acting on the
[17] Li Z, Pourquie M, Van Terwisga TJC. A numerical study of steady and unsteady
moving parts must also be analysed in detail, as their behaviour cavitation on a 2d hydrofoil. J Hydrodyn Ser B 2010;22(5):7707 [Supplement
is quite different under cavitating and non-cavitating ow 1].
conditions. [18] Zhang R, Chen H. Numerical analysis of cavitation within slanted axial-ow
pump. J Hydrodyn Ser B 2013;25(5):66372.
[19] Manning N. Hydraulic control systems. John Wiley & Sons; 2005.
[20] Valds JR, Miana MJ, Nez JL, Ptz T. Reduced order model for estimation of
References uid ow and ow forces in hydraulic proportional valves. Energy Conver
Manage 2008;49(6):151729.
[1] Ors J, Rodrguez JM, Ptz T, Schwanke W. Real-time simulation of hydraulic [21] Payri F, Bermdez V, Payri R, Salvador FJ. The inuence of cavitation on the
control unit for brake systems. In: Presented at the 12th EAEC european internal ow and the spray characteristics in diesel injection nozzles. Fuel
automotive congress; 2009. 2004;83(45):41931.
[2] Zhao W, Zhang L, Shao X. Numerical simulation of cavitation ow under high [22] Posa A, Oresta P, Lippolis A. Analysis of a directional hydraulic valve by a Direct
pressure and temperature. J Hydrodyn Ser B 2011;23(3):28994. Numerical Simulation using an immersed-boundary method. Energy Conver
[3] Brennen CE. Cavitation and bubble dynamics. Oxford University Press; 1995. Manage 2013;65(January):497506.
[4] Singhal AK, Athavale M, Li HY. Mathematical basis and validation of the full [23] White FM. Fluid mechanics. McGraw-Hill; 1979.
cavitation model. J Fluids Eng 2002;124(3):61724. [24] ANSYS FLUENT Theory Guide; 2012 [release 14.5.12].
[5] Schnerr GH, Sauer J. Physical and numerical modeling of unsteady cavitation [25] Launder BE, Spalding DB. The numerical computation of turbulent ows.
dynamics. In: Presented at the 4th int. conference of multiphase ow, New Comput Method Appl Mech Eng 1974;3(2):26989.
Orleans; 2001. [26] Yakhot V, Orszag SA. Renormalization group analysis of turbulence. J Sci
[6] Kunz RF, Boger DA, Stinebring DR, Chyczewski TS, Lindau JW, Gibeling HJ, et al. Comput 1986;1(1):151.
A preconditioned NavierStokes method for two-phase ows with application [27] Shih TH, Liou WW, Shabbir A, Yang Z, Zhu J. A new k-e eddy-viscosity model
to cavitation prediction. Comput Fluids 2000;29(8):84975. for high reynolds numer turbulent ows-model development and validation.
[7] Zwart PJ, Gerber AG, Belamri T. A two-phase ow model for predicting Comput Fluids 1995;23(3):22738.
cavitation dynamics. In: Presented at the 5th int. conference on multiphase [28] Patankar S, Spalding D. A calculation procedure for heat, mass and momentum
ow, Yokohama, Japan; 2004. transfer in three-dimensional parabolic ows. Int J Heat Mass Transfer
[8] Chern M-J, Wang C-C, Ma C-H. Performance test and ow visualization of ball 1972;15(10):1787806.
valve. Exper Therm Fluid Sci 2007;31(6):50512. [29] McCloy D, Martin HR. The control of uid power. Wiley; 1973.

Das könnte Ihnen auch gefallen