Numerical and Experimental Investigation On A Conical Poppet Relief Valve With Flow Force Compensation
Numerical and Experimental Investigation On A Conical Poppet Relief Valve With Flow Force Compensation
To cite this article: Roberto Finesso & Massimo Rundo (2017): Numerical and experimental
investigation on a conical poppet relief valve with flow force compensation, International Journal of
Fluid Power, DOI: 10.1080/14399776.2017.1296740
a dynamic mesh and of a user-defined function (UDF). A detail of the valve cartridge in regulating conditions
The UDF is used to update the position of the poppet at is reported in Figure 2. Since the diameters of the poppet
each time instant, on the basis of the evaluation of the seat and of the damper are the same, when the valve is
forces acting on the poppet itself, including the contribu- closed, only the pressure acting of the right frontal sur-
tions of the spring and of the fluid; the latter contribution face of the damper generates the opening force Fact to
is evaluated taking into account the pressure field infor- oppose the spring force Fspr. In steady-state conditions
mation that derives from the CFD solution. The CFD such pressure coincides with the inlet pressure at the
mesh is then deformed at the new time instant, on the port P.
basis of the new position of the poppet that is evaluated In regulating conditions two additional forces arise:
by the UDF, and the flow and pressure field solutions the flow force Fjet due to the reduction of the pressure
are then solved by the CFD code. This method allows along the conical poppet just upstream the metering
a direct simulation of the flow-pressure characteristic edge and the force Fdef generated by the flow deflector.
of the valve to be obtained, for a given spring preload. Three different configurations of the flow deflec-
It is in fact sufficient to set the preload of the spring in tor were tested experimentally. In particular, in addi-
the UDF, as well as the initial poppet opening position, tion to the baseline geometry referred to as ‘Geometry
and to set the inlet flow rate; the dynamic simulation is A’, a spacer of 0.5 mm (Geometry B) and of 1.0 mm
then run, and the inlet valve pressure can be evaluated at (Geometry C) were added in order to investigate the
the end of the simulation when the poppet achieves the effect of the deflector depth on the flow-pressure char-
steady-state position. The flow-pressure characteristic acteristic. The geometrical parameters of the valve are
can thus be derived point by point, by setting different listed in Table 1.
inlet flow rates and different spring preloads. The CFD The hydraulic scheme of the test rig is reported in
simulations were carried out using the cavitation model Figure 3. The flow rate was generated by a 40 cc/rev
embedded in ANSYS Fluent, in order to avoid the occur- variable displacement pump equipped with absolute
rence of negative absolute pressure values (Gao et al. pressure limiter, which was set to 100 bar. The pump
2002, Bernad and Resiga 2012). An alternative approach was driven by a constant speed 15 kW electric motor.
to evaluate the flow-pressure characteristic of the valve A two-port flow control valve RQ2 was used to impose
is based on lumped-parameter modelling environments. the flow rate through the valve under test.
These tools are based on analytical models to evaluate
the forces acting on the poppet, but the force generated
by the flow deflector is not included. A correct estima-
tion of such contribution would require a large number
of steady-state 3D-CFD simulations to be carried out
(for different flow rate and poppet opening conditions),
in order to build a look-up table of the force which can
be implemented in the lumped-parameter modelling
tool. This procedure is however time consuming. For
this reason a customized 0D model has been developed
and implemented in the LMS Imagine.Lab AmesimTM
Figure 1. 3D section view of the valve.
environment in order to evaluate the forces acting on
the flow deflector. It will be shown that this model is
characterized by a very simple tuning, as it requires only
a few 3D-CFD simulations to be carried out.
equation k – ε model to resolve the Reynolds stress 3.2.2. Dynamic mesh and UDF
and close the equation set. The scalable wall-function A dynamic mesh was used to deal with the motion of
approach was adopted instead of the standard wall-func- the poppet; in particular, the surface of the poppet was
tion one (ANSYS FluentTM theory guide, 2011). defined as the moving zone with rigid body motion solu-
It is worthwhile recalling that the standard wall func- tion, while all the remaining inner surfaces of the valve
tion approach is based on the Lauder and Spalding results were set as stationary zones. The smoothing, layering
(Lauder et al. 1974), according to which the boundary and remeshing functions embedded in ANSYS Fluent
layer is divided into two sub-layers. The inner one is were used to prevent excessive mesh distortion and the
the viscous sub-layer (y* < 11.225), for which laminar occurrence of negative volumes of the cells during the
stress-strain relation occurs, so that y* = U*. The quan- poppet displacement. The remeshing was activated if
tities y* and U* are the dimensionless wall distance and the maximum cell skewness exceeded a threshold equal
dimensionless velocity respectively, defined as follows: to 0.9. A UDF was written in C code and compiled in
ANSYS Fluent, in order to calculate the axial rigid body
1∕2
𝜌C𝜇1∕4 kp yp motion of the poppet. This function was used to cal-
∗
y = (1)
𝜇 culate the net axial force acting on the poppet at each
time instant, taking into account the spring and the
fluid forces. In the UDF it is necessary to set the spring
1∕2
stiffness, the initial spring preload, as well as the initial
Up C𝜇1∕4 kp poppet opening position and the surfaces of the poppet
U =∗
(2)
𝜏w ∕𝜌 on which the fluid force is calculated.
where kp is the turbulence kinetic energy at the near-wall 3.2.3. Fluid type and cavitation sub-model
node P, C𝜇is an empirical constant equal to 0.09, yp is Preliminary simulations were carried out using a sin-
the distance from point P to the wall, μ is the dynamic gle-phase incompressible fluid; however, it was verified
viscosity of the fluid, ρ is the density of the fluid, Up is that negative values of the absolute pressure (up to about
the mean velocity at node P and τw the wall shear stress. −20 bar) occurred in some zones of the fluid domain (in
In the log-law sub-layer (y* > 11.225), logarithmic particular, near to sharp edges with high fluid velocity
dependence between dimensionless velocity and wall gradients, as well as in recirculation vortexes). Negative
unit occurs: absolute pressure values are unphysical, and may result
in a wrong estimation of the forces exerted by the fluid
1 ( ∗)
U∗ = ln Ey (3) on the poppet. Although it was verified that the influ-
𝜅
ence of such negative pressure regions is little, for a more
where κ is the von Karman constant (equal to 0.4187) rigorous approach a two-phase mixture was chosen as
and E is an empirical constant, equal to 9.793. The log- working fluid, which is constituted by incompressible
arithmic law for mean velocity is known to be valid for liquid phase (DTE25 oil at 40 °C) and vapour phase,
30 < y* < 300. setting a constant vaporization pressure. The Schnerr
The standard wall function approach requires that the and Sauer cavitation sub-model (Schnerr et al. 2001)
first cell outside the walls is in the log-layer region (y* > embedded in ANSYS Fluent was used to estimate the
11.225). The scalable wall-function approach prevents expression of the net mass transfer from liquid to vapour.
from the deterioration of the results in case of highly Obviously in the real valve the gaseous phase will be also
refined grid (y* < 11.225), thus avoiding an erroneous constituted by separated air, however such aspect goes
modelling of the viscous sub-layer and buffer regions. beyond the scope of the present paper.
In particular, a limiter in the y* calculation is introduced
in Equation (3), so that: 3.2.4. Boundary conditions, discretization schemes
and solution methods
ỹ ∗ = MAX y ∗ , ylimit
∗
(4) For all simulations, a steady-state flow solution was pre-
( )
liminarily calculated and set as the initial condition for
where y*limit = 11.225. the time-dependent solution. The inlet mass flow rate
These wall functions produce consistent results for at port P and the outlet pressure at port T were set as
grids of arbitrary refinement. For grids which are coarser boundary conditions. The vapour mass fraction at the
than y* > 11.225, the scalable wall-function approach will inlet and outlet of the valve was set to zero.
provide identical results to the standard wall function The pressure-based coupled algorithm was used to
one. In the present investigation, it was found that the solve the momentum and pressure-based continuity
use of the standard wall function approach, instead of equations together, so as to speed-up convergence.
the scalable wall function one, leads to errors in the pre- The gradients were evaluated using the Green–Gauss
diction of the valve characteristic which can be of the node based method, which is usually recommended for
order of 6% with respect to the experimental data, if the tetrahedral meshes. The PRESTO! discretization scheme
near-wall mesh cell size is not appropriate. was used for the pressure equation, the QUICK one for
INTERNATIONAL JOURNAL OF FLUID POWER 5
50
3.3.2. Comparison between the predicted and
experimental valve characteristics
Flow rate (L/min)
40
The dynamic 3D-CFD model has been validated by
carrying out a comparison between the experimental 30
60
Geometry B Predicted Experimental
50
Flow rate (L/min)
40
30
20
10
0
0 15 30 45 60 75 90
Pressure difference P-T (bar)
50
Flow rate (L/min)
40
30
20
10
0
0 15 30 45 60 75 90
Low spring setting
Pressure difference P-T (bar)
area in the metering edge. This would lead to a benefit Figure 18. Velocity fields, at 50 L/min, different deflector
in terms of the valve characteristic (i.e. to a reduction positions, high spring setting.
in the inlet pressure at equal flow rate), because a lower
spring compression is necessary to achieve the same
cross-sectional area. However, it can be seen in Table 2
that an increase in the poppet cone angle also leads to a
decrease in the flow force acting on the flow deflector,
which has instead a negative effect on the valve charac-
teristic. This can also be observed in Figure 17, where it
can be seen that lower pressure distributions occur on
the flow deflector at increasing cone angles, as a con-
sequence of a different angle of the flow impacting the
deflector and of the fluid outflowing from it.
As a result, the net effect of a variation in the poppet
cone angle on the valve characteristic depends on which
of the two effects is predominant. For the analysed con-
ditions, both a reduction and an increase in the poppet
cone angle of 5 deg lead to a detriment in the valve char-
acteristic (i.e. to an increase in the inlet pressure at fixed Baseline Distance +1.5 mm +3.0 mm
flow rate), so that the baseline geometrical configuration
Figure 19. Pressure fields, at 50 L/min, different deflector
can be considered as the optimal one.
positions, high spring setting.
70
Pred.(with Fdef & recirc.) Pred.(no Fdef)
Pred.(with Fdef no recirc.) Experimental
60
15 bar 45 bar P4 P3 75 bar
50
30
Figure 22. LMS Amesim model of the valve.
20
10
70
Pred.(with Fdef) Pred.(no Fdef)
Experimental 0
60 0 15 30 45 60 75 90 105
15 bar 45 bar P2 75 bar P1 Pressure difference P-T (bar)
50
Flow rate (L/min)
30
20
However the CFD simulations show that a fraction of the
10
flow rate is recirculated and the coefficient k31 is about
0 0.25. The angle θ3 is not easy to be determined, so a
0 15 30 45 60 75 90 105 120
value of 45° can be a good compromise. In this case P4
Pressure difference P-T (bar)
becomes the new final point and the force is 146 N with
Figure 23. Predicted vs. experimental curves, Geometry C.
an error of 8%. It should, however, be noted that such
value represents the maximum error, since for lower flow
rates or lower pressures the simulated curves are closer
Component Design library is reported (Amesim User’s to the experimental data.
Guide, 2015) and the customized component imple-
menting the expression (11) is highlighted.
5. Conclusion
A flow source is connected to the inlet port, while
the spring chamber is connected directly with the tank. In this paper two modelling approaches for assessing the
steady-state characteristic of a flow force compensated
pressure relief valve are presented. A 3D CFD model
4.2. Model performance
with interaction between the pressure field and the pop-
The comparison between the experimental points and pet position has been developed in ANSYS Fluent. It has
the simulated characteristics is reported in Figure 23 been demonstrated that such model can be used as a
for three different pressure settings and Geometry C. fully predictive tool for the evaluation of the valve behav-
In such a case, based on the CFD simulations, the recir- iour, since it does not require any tuning on experimen-
culation is negligible and, accordingly, the term k31 has tal tests. It was verified that, in order to obtain physically
been set to 0. The simulations have been performed with consistent results, a cavitation model has to be adopted,
θ2 = 120° and k21 = 0.4; both values come from the CFD and particular attention must be paid to the choice of
simulation at 50 L/min and maximum pressure setting the correct wall function approach.
(see Figure 14). In the same figure the curves simulated Moreover a very good agreement between the pre-
without the force Fdef are also superimposed (no Fdef ), dicted and measured curves was found. The model was
allowing to appreciate the significant contribution of the also proved to be reliable in evaluating the effect of a
force induced by the deflector. In fact at the maximum geometrical modification of the flow deflector.
pressure setting and maximum flow rate, the deflector A significant outcome from the study is that the flow
allows to reduce the real regulated pressure from 103 bar force compensation is mainly affected by the depth of
(point P1) to 78.5 bar (point P2). In this operating point the deflector, while the cone angle and the deflector dis-
the force Fdef calculated with the Equation (7) is equal tance have a negligible effect. The lumped parameter
to 116 N. model requires a tuning on CFD data, however reason-
With the Geometry A (Figure 24), in the same oper- ably good results can be obtained with just one CFD
ating point the experimental final regulated pressure is steady-state simulation performed in the condition of
about 69 bar and the force Fdef can be estimated to be about maximum flow rate and maximum pressure setting.
159 N. If the recirculation is not taken into account (no Since the presented CFD model takes into account also
recirc.), the simulated deflector force is 132 N (point P3) the dynamics of the valve, it can reasonably be used to
with an error of 17%, using θ2 = 170°, based on Figure 11. investigate transient phenomena.
INTERNATIONAL JOURNAL OF FLUID POWER 11
Siemens PLM Software, 2015. LMS Imagine.Lab Amesim – Vaughan, N.D., Johnston, D.N., and Edge, K.A., 1992.
hydraulic component design library 14, user’s guide Numerical simulation of fluid flow in poppet valves.
Vaughan, N.D., Johnston, D.N., and Edge, K.A., 1991. Proceedings of the institution of mechanical engineers part
Experimental investigation of flow and force characteristics c journal of mechanical engineering science, 206 (2), 119–
of hydraulic poppet and disc valves. Proceedings of the 127. doi:10.1243/PIME_PROC_1992_206_105_02.
institution of mechanical engineers, part a: journal of
power and energy, 205 (3), 161–171, doi:10.1243/PIME_
PROC_1991_205_025_02.