using steady flow force for unstable valve design ...lixxx099/papers/yuanli03 ver 1.pdfusing steady...
TRANSCRIPT
1
Using Steady Flow Force for Unstable Valve Design:Modeling and Experiments
Qinghui Yuan and Perry Y. Li
Abstract— In single stage valves, the main spoolsare stroked directly by solenoid actuators. Theyare cheaper and more reliable than multistagevalves. Their use, however, is restricted to lowbandwidth and low flow rate applications due tothe limitation of the solenoid actuators. Our re-search focuses on alleviating the need for large andexpensive solenoids in single stage valves by ad-vantageously using fluid flow forces. For example,in a previous paper, we proposed to improve spoolagility by inducing unstable transient flow forces bythe use of negative damping lengths. In the presentpaper, how steady flow forces can be manipulatedto improve spool agility is examined through fun-damental momentum analysis, CFD analysis andexperimental studies. Particularly, it is found thattwo previously ignored components - viscosity ef-fect and non-orifice momentum flux, have strong in-fluence on steady flow forces. For positive dampinglengths, viscosity increases the steady flow force,whereas for negative damping lengths, viscosity hasthe tendency of reducing steady flow forces. Also,by slightly modifying the non-orifice port geometry,the non-orifice flux can also be manipulated so as toreduce steady flow force. Therefore, both transientand steady flow forces, can also be used to improvethe agility of single stage electrohydraulic valves.
Keywords: Unstable valve, damping length, steadyflow force, viscosity, non-orifice flux, computationalfluid dynamics (CFD)
This research is supported by the National Science Founda-tion ENG/CMS-0088964. Submitted to the ASME Journal ofDynamic Systems, Measurement and Control, July 2003. Por-tions of this work appeared or will appear at the 2002 BathWorkshop on Power Transmission and Motion Control, 2002American Control Conference, and 2003 ASME-IMECE.
Q-H Yuan and P. Y. Li are with the Department of MechanicalEngineering, University of Minnesota, Minneapolis, MN 55455.E-mails: {qhyuan, pli}@me.umn.edu. Please send all corre-spondence to Professor Perry Y. Li.
Nomenclature
α viscosity related geometry coefficient∆1 land area difference compensation∆2 land tapers compensationτsleeve wall stresses on the sleeveτrod wall stresses on the rodθ Vena contracta jet angleµ dynamic viscosityρ densityAl annular area of the left landAo orifice areaAr annular area of the right landCc Vena contracta contraction coefficientcin meter-in chamber non-orifice outlet flux
coefficientcout meter-out chamber non-orifice inlet flux
coefficientFefflux net momentum effluxFland pressure force acting on the landsFrod viscous force acting on the rodFsleeve viscous sleeve force acting on fluidFspool fluid flow force acting on spoolFsteady steady flow forceL damping lengthL1 meter-in chamber damping lengthL2 meter-out chamber damping lengthLd dead space distancen unit vector outward normal to surfacenx positive x unit vectorkL steady force sensitivity to damping
lengthPl pressure on the left landPr pressure on the right landQ flow rateP fluid pressureRi radius of the rodRo radius of the sleevev fluid velocity vectorvx longitudinal fluid velocity
2
I. Introduction
In a single stage electrohydraulic flow control valve,
the main spool is stoked directly by solenoid actua-
tors. In high flow rate and high bandwidth applica-
tions, the solenoids must overcome significant forces.
The force and power limitations of the solenoid ac-
tuators make single stage valves un-suitable for high
performance applications. In these situations, multi
stage valves are generally used in which the spools are
driven by one or more pilot stage hydraulic valves.
However, multi stage valves tend to be more suscep-
tible to contamination, and more expensive in terms
of manufacturing and maintenance, than single stage
valves.
Our research aims at improving the flow and band-
width capabilities of single stage valves by alleviating
the need for large and expensive solenoid actuators.
Our approach is to utilize the inherent fluid flow forces
to increase the agility of the spool, so that less power
or less force is needed from the solenoids while achiev-
ing fast spool responses. Flow forces can be classified
as either steady or transient. Steady flow forces are
those exerted by the fluid on the spool during steady
flow conditions, while transient forces are the addi-
tional forces due to the time variation of the flow con-
dition. Our initial approach [1] proposes to configure
the valve to achieve transient flow force that has an
unstable (negative damping) effect on the spool, and
then to stabilize the system by closed loop feedback.
Analysis and computer simulations show that valves
configured to have unstable transient flow forces can
have faster step responses under solenoid saturation
than their stable counterparts. They also take less
positive power but more negative power (braking)
Ps Sleeve Pt
Fe
Q Q
Rod
Xv
L1 L2
Actuator
Land
Solenoid
Fig. 1. Typical configuration of a four way direction flow
control valve. The two “Q” ports are connected to
the load (hydraulic actuator), and Ps is connected to
the supply pressure, and Pt is connected to the return.
In a single stage valve the spool is stroked directly by
solenoid actuators. Damping length L is defined to be
L := L2 − L1.
power to track a sinusoidal flow rate. The saving in
positive power becomes more significant at higher fre-
quencies. A basic premise of this study is that the
sign of damping length determines whether the tran-
sient flow force is stabilizing or destabilizing (see Fig.
1 for a definition). The study in [1] distinguishes from
other studies of transient flow force induced instabil-
ity in the 1960’s [2], [3] in that our goal is to utilize the
instability advantageously rather than to eliminate it
as in the previous studies.
In this paper, experimental results are first pre-
sented to confirm that spool agility is indeed improved
when valves are configured to have small or nega-
tive damping lengths. Then, it is argued that the
improved agility may not due solely to the unstable
transient flow forces as was originally hypothesized.
The flow force models in [1] as well as in [2], [3] fo-
cus only on the momentum flux at the variable orifice
and neglect the effects of viscosity and the non-orifice
3
momentum flux. They predict that the steady flow
forces should be unaffected by damping lengths. In
this paper, we present models for flow forces that take
into account fluid viscosity and non-orifice momentum
flux. The new models are developed based on funda-
mental momentum theory and are verified using com-
putational fluid dynamics (CFD) analysis and experi-
ments. It turns out that both viscosity and non-orifice
momentum flux have significant effects on steady flow
forces, and hence spool agility. In particular, viscos-
ity reduces the steady flow force for negative damping
lengths, and increases it for positive damping lengths.
For sufficiently negative damping lengths, the steady
flow force can even become unstable. Thus, the exper-
imentally observed improvement in spool agility for
negative damping lengths, can be attributed, to the
reduction in steady flow forces in addition to the un-
stable transient flow force. Furthermore, we show that
the non-orifice momentum can be manipulated by mi-
nor modification of the port geometry, so as to reduce
the steady flow force. Consideration of the viscosity
effect and the non-orifice momentum flux presents ad-
ditional design opportunities to improve spool agility
via the reduction of steady flow forces.
Steady flow force compensation techniques have
been studied in the 1960’s (see [3], [2] for a summary).
These focus on altering the flow pattern at the vari-
able orifice (which involves shaping the lands or the
sleeves, or using multiple small orifices) or on induc-
ing a differential force on the spool ends at high flows.
The steady flow force reduction methods that make
use of viscosity or the non-orifice momentum flux may
be simpler to implement.
The rest of paper is organized as follows. In section
II, we present some experimental results that confirm
the influence of damping lengths on spool agility. In
section III, we present models for the fluid flow in-
duced steady flow forces that take into account vis-
cosity and non-orifice momentum flux. Section IV
presents CFD analysis of the fluid forces and compare
the accuracies of the various approximations. The ex-
perimental study on the viscosity effect and the non-
orifice flux effect is presented in section V. Section VI
contains some concluding remarks.
II. Effect of damping Lengths on Spool
Agility: Experimental results
The experimental setup (Fig. 2) consists of a cus-
tom built valve that allows for different (both pos-
itive and negative) damping lengths L. The valve
sleeve has eleven (11) ports, fitted with quick couplers
for connection to the inlets and outlets. The ports
that are not connected are blocked. By connecting
the supply and return to different ports, the dimen-
sions L1 and L2 in Fig. 1 and the damping length
L := L2 − L1 can be changed. Although both meter-
in and meter-out chambers (see Section III) can be
achieved simultaneously, our experiments investigate
each single chamber separately so that either L1 = 0
or L2 = 0. The spool consists of a threaded hardened
precision anodized aluminum shaft on which several
bronze lands can be arbitrarily positioned to be con-
sistent with the chosen inlet / outlet locations. The
spool is actuated by a pair of solenoids and the spool’s
displacement is measured by a linear potentiometer.
The solenoids are driven by a MOSFET circuit so that
the gate voltage Vg applied to the MOSFET would be
roughly proportional to the solenoid current. The con-
4
(a) Photograph of the valve
A B C D E F G H I J K
6.35 mm
1 2
. 7 m
m
216 mm 117 mm
(b) Dimension of the valve
Fig. 2. Custom built valve with adjustable damping
length. By connecting the inlet / outlet and actua-
tor to the various ports, different damping lengths can
be achieved. The configuration shown has only one
valve chamber.
trol of the solenoid and the data acquisition system
are coordinated by a PC within the Matlab (Math-
works, MA USA) environment via the Matlab’s Real-
time Toolbox (Humusoft, CZ Rep.).
The spool displacement trajectories in response to
step inputs of Vg = 2.3V for the various damping
lengths are shown in Figure 3. As the damping length
decreases, it takes less time for the spool to travel the
full stroke of 7 mm (transit time). The transit times
for Vg = 2.3V and Vg = 2.4V for the different damping
lengths are shown in Fig. 4. In both cases, the transit
times decrease as damping length decreases. These
experimental results show clearly that damping length
has a significant effect on spool agility. Moreover, the
damping length has a more significant effect for Vg =
2.3V than for Vg = 2.4V as indicated by the slopes in
20 30 40 50 60 70 80−2
0
2
4
6
8
10x 10
−3
time(ms)
x v(m)
−0.216
−0.118
0.118
0.216
Fig. 3. Spool displacement trajectories for various damp-
ing lengths (from left to right: damping lengths are
L = 0.216, 0.118,−0.118,−0.216m.) when a step in-
put of Vg = 2.3V is applied.
Fig.4. This latter result was qualitatively predicted by
the numerical analysis in [1] which shows that spool
instability has a greater effect on spool agility when
the solenoid limitation is more severe. In fact when
Vg = 2.2V , the solenoid limitation is so severe that
the spool is halted by friction and is unable to travel
the entire stroke length (figure 5). However, a longer
stroke is attained when the damping length is smaller.
For L = −0.216m, the spool is able to nearly complete
the entire stroke.
III. Modeling of steady flow forces
The qualitative increase in spool agility observed
in the experiments in Section II was originally ex-
plained by the traditional premise [1], [3], [2] that
transient flow forces tend to be destabilizing for neg-
ative damping lengths. In this section, we show that
the improved agility can also be attributed in part
to reduction in steady flow force due to viscosity and
non-orifice flux effects.
A typical four way directional control valve such as
5
−0.25 −0.2 −0.15 −0.1 −0.05 0 0.05 0.1 0.15 0.2 0.2528
30
32
34
36
38
40
42
44
46
48
damping length(m)
resp
onse
tim
e(m
s)V
g=2.3V
Vg=2.4V
Fig. 4. Spool transit times versus damping length for step
inputs of 2.3V and 2.4V.
0 20 40 60 80 100 120−1
0
1
2
3
4
5
6
7
8x 10
−3
time(ms)
x v(m)
L= 0.216m
L= 0.118m
L= −0.118m
L= −0.216m
Fig. 5. Displacement trajectory of the spool displacement
various damping lengths for step input of Vg = 2.2V
in Figure 1 consists of a meter-in chamber (left) and
a meter-out chamber (right). The fluid flow induced
force on the spool for the meter-in and the meter-out
chambers are analyzed separately. In single chamber
experiments, negative damping lengths correspond to
meter-in chambers (L = −L1) and positive damping
lengths correspond to meter-out chambers (L = L2).
A. Meter-in chamber
Consider first the meter-in chamber (Fig. 6). The
force Fspool (positive to the right) that the spool ex-
L1
x
Xv
P r P l
−τ sleeve
θ
−τ rod
τ rod
Ld
P l P r
Ri
Ro
(x=0)
Fig. 6. Meter-in valve chamber. The lower gray block is
the control volume (C.V.) which includes all the fluid
in the chamber, and is surrounded by the sleeve, rod,
land ends and the inlet/outlet surfaces.
periences from the fluid can be calculated in various
ways. Most fundamentally, it is given by:
Fspool = Fland + Frod
=
∫
Ar
Pr dA−∫
Al
Pl dA+
∫
rodτrod dA
(1)
where Fland is the pressure force acting on the lands,
Frod is the viscous force acting on the spool rod, Pr
and Pl are the pressures on the right and left lands,
Ar and Al are the annulus area of the right and left
lands, and τrod is the shear stress exerted on the rod
by the fluid.
Fspool can also be computed from the reaction of
the force that the spool acts on the fluid. The conser-
vation of longitudinal (left to right) momentum equa-
tion for the control volume (C.V.) that encompasses
the fluid in the meter-in chamber (marked gray in Fig.
6
6) is given by:
− Fspool +
∫
sleeve−τsleeve dA
︸ ︷︷ ︸Fsleeve
=d
dt
∫
C.V.ρvxdV +
∫
C.S.ρvx v · ndA
︸ ︷︷ ︸Fefflux
(2)
where the LHS is the sum of the forces that the spool
and the sleeve act on the fluid, and the RHS is the
sum of the rate of change of the longitudinal momen-
tum and the net momentum flux out of C.V.. Here, ρ
is the fluid density, v and vx are the fluid velocity and
its longitudinal component, and n is the outward unit
normal vector associated with the differential surface
dA. The areas of and the shear stresses at the inlet
and outlet surfaces are assumed to be negligible com-
pared to the sleeve area and the sleeve shear force.
From (2), Fspool can be decomposed into a steady
flow force component and a transient flow force com-
ponent:
Fspool = −Fefflux + Fsleeve︸ ︷︷ ︸steady flow force
− d
dt
∫
C.V.ρvx dV
︸ ︷︷ ︸transient flow force
. (3)
Previous models for steady flow force [2], [3] do not
consider the viscosity effect, i.e. Frod = Fsleeve = 0.
Consider now Fefflux and Fsleeve separately. The net
momentum efflux in (2) is given by:
Fefflux,1 =
∫
outletρvxv · ndA+
∫
inletρvxv · ndA (4)
since, if there is no leak, the only surfaces that have
non-zero flux through them are the inlet and outlet.
The efflux at the non-orifice outlet is often neglected
in previous studies [1], [2], [3] on the assumption that
the fluid velocity is nearly normal to the spool axis
at the outlet surface (vx = 0). However, CFD analy-
sis in Section IV reveals that the outlet efflux can be
significant.
Generally, knowledge of the flow profile at the inlet
and outlet is needed to evaluate Eq. (4). However,
the flow at the variable orifice inlet is often modeled
by a Vena contracta in which the flow is uniform [3].
If we further assume that the flow profile at the non-
orifice outlet is also proportional to the flowrate Q,
then, Eq. (4) can be approximated by:
Fefflux,2 =
(−cin +
ρ cos θ
CcAo(xv)
)Q2 (5)
where Ao(xv) is the orifice area when the spool dis-
placement is xv, θ is the Vena contracta jet angle,
Cc < 1 is the contraction coefficient so that CcAo(xv)
is the jet area, and cin is a coefficient that summarizes
the normalized flow profile at the non-orifice outlet.
The sign convention in (5) has been chosen so that
cin > 0 in typical port geometries.
Consider now the viscous sleeve force Fsleeve. If the
fluid is Newtonian, then fundamentally,
Fsleeve,1 = −∫
sleeveτsleeve dA =
∫
sleeveµ∂vx∂n
dA (6)
where µ is the fluid dynamic viscosity and ∂vx∂n is the
longitudinal velocity gradient with respect to the out-
ward normal at the sleeve surface. To evaluate Eq.
(6) generally requires knowledge of the flow profile at
the non-orifice outlet.
Using the assumptions that 1) the friction force in
the dead space between the outlet and the left hand
land in Fig. 6 is negligible; 2) the flow is laminar and
fully developed along the full length L1 between the
inlet and outlet; Eq.(6) can be approximated by
Fsleeve,2 = αµL1Q (7)
for some coefficient α > 0 that depends only on valve
7
geometry. Generally, α increases as the cross sectional
area of the flow passage decreases.
For the cylindrical valve geometry in Fig. 6 (see the
Appendix),
α =4(2R2
o ln(Ro/Ri) +R2i −R2
o)
(R4o −R4
i ) ln(Ro/Ri)− (R2i −R2
o)2. (8)
The assumption that the flow in the chamber is lam-
inar is valid when the Reynolds number Re < 2100
(e.g. when the flow rate is sufficiently low). For
our experimental system in Fig. 2, Re ≈ 1000 when
Q = 38 lpm (10 gpm). For higher flow rate, the flow
may become turbulent and the shear stress must in-
clude also the Reynolds stresses caused by the random
velocity components. The analysis can be extended to
this case using friction / flow relation obtained from
tabulated empirical friction factors. For example, in
the fully turbulent regime, the friction factor is pro-
portional to 1/R0.25e [4] and Eq. (7) can be replaced
by
Fsleeve,3 = βµ0.25Q1.75L1 (9)
where β is a function of the geometry and the fluid
density. In this paper, we focus on laminar flows.
In summary, the steady flow force for the meter-in
chamber is given by
Fsteady = −Fefflux + Fsleeve (10)
where Fefflux is given either fundamentally by (4) or
approximately by (5); and Fsleeve is given either fun-
damentally from (6) or approximately by (7).
Since the flow rate Q increases monotonically with
the orifice area Ao(xv), and Ao(xv) is roughly linear
w.r.t. xv, Eqs. (5) shows that the steady flow force
component due to the variable orifice will be roughly
linear in xv. This is the well known stable linear spring
L2
x
Xv θ
P r P l
−τ sleeve −τ rod
τ rod
Ri
Ro
P l P r (x=0)
Ld
Fig. 7. Meter-out valve chamber.
effect on the spool [2], [3]. On the other hand, both
the non-orifice flux component (−cinQ2) in Eq. (5)
and Fsleeve in Eq.(7) correspond to unstable quadratic
and unstable linear spring forces with negative spring
constants. Therefore, for the meter-in chamber, both
the previously neglected viscosity effect and the non-
orifice flux effect tend to reduce the steady state flow
force.
B. Meter-out valve chamber
A similar analysis can be performed for the meter-
out valve chamber (Fig. 7). The steady flow force is
also given by Eq. (10):
Fsteady = −Fefflux + Fsleeve (10)
with Fefflux given by one of the following:
Fefflux,1 =
∫
outletρvxv · ndA+
∫
inletρvxv · ndA (11)
Fefflux,2 =
(−cout +
ρ cos θ
CdAo(xv)
)Q2 (12)
where cout is a coefficient that summarizes the flow
pattern at the non-orifice inlet for meter-out cham-
ber. The sleeve force is given by one of the following
8
expressions:
Fsleeve,1 = −∫
sleeveτsleeve dA =
∫
sleeveµ∂vx∂n
dA
(13)
Fsleeve,2 = −αµL2Q (14)
where L2 is the distance between the entry port and
the meter-out orifice in Fig. 7. For the valve geometry
in Fig. 7, the same α given in Eq. (8) for the meter-in
case can be used.
For the meter-out chamber, the steady flow force
due to the momentum flux at the variable orifice can
again be represented by a stable linear spring. The
non-orifice momentum flux (−coutQ2) also acts like
an unstable quadratic spring with a negative spring
constant and serves to reduce the steady flow force.
Unlike the meter-in chamber, the sleeve force compo-
nent of the steady flow force in Eq.(14) now acts like a
stable spring with a positive spring coefficient. There-
fore, the neglected viscosity effect tends to increase
the steady flow force in the case of a meter-out cham-
ber. Moreover, since typically, 0 < cout < cin, the
reduction in steady flow force due to the non-orifice
flux in a meter-out chamber (L = L2) will be less than
that in a meter-in chamber (L = −L1). Both of these
effects can contribute to the observed improvement in
spool agility in Section II.
C. Valve with both meter-in and meter-out valve
chambers
In a 4-way symmetric valve, the spool is acted on
by the flow forces in both the meter-in and the meter-
out chambers (Fig. 1). Hence, the net force that
acts on the spool is the sum of the forces and can be
approximated by:
Fsteady =
[−(cin + cout) +
2ρ cos θ
CcAo(xv)
]Q2
− αµ (L2 − L1)︸ ︷︷ ︸L
Q (15)
where it is assumed that for a given spool displace-
ment, the Vena contracta coefficient is the same for
both meter-in and meter-out chambers in a symmetric
valve. In commercial valves, L := L2−L1 is designed
to be positive to ensure positive damping effect. In
[1] we proposed that by choosing L < 0 to induce
a negative damping effect through the transient flow
forces so as to improve the agility and responsiveness
of the spool. The new results in this paper are that,
the steady flow forces can also be used to improve the
agility of the single stage valve via the viscosity effect
and the non-orifice flux.
IV. CFD analysis of flow forces
In this section, we present CFD analysis to ver-
ify and evaluate the various flow force models pre-
sented in Section III. The 3D computational mod-
els for a given xv are shown in Fig. 8. Notice that
the valve is not axis symmetric because of the inlet
and outlet ports. The mesh and the boundary condi-
tion for each geometry are generated by the GAMBIT
pre-processor. Each computation volume uses about
1,000,000 nodes and 500,000 elements.
The incompressible Navier-stokes equations with-
out body forces are given by [5]:
Continuity:
∇ · v = 0 (16)
Momentum:
ρ∂v
∂t+ ρv · ∇v = −∇P + µ∇2v (17)
9
(a) L1 = 0.216m or L2 = 0.216m
(a) L1 = 0.118m or L2 = 0.118m
Fig. 8. Two fluid models for a given xv. The orifice is
at the right hand side port. To model the meter-in
chamber (left chamber in Fig. 1), the right hand side
port is the entry port, and the left hand port is the
outlet port. To model the meter-out chamber (right
chamber in Fig. 1), the left hand side port is the entry
port, and the right hand side port is the outlet port.
where ρ is the fluid density, v is the fluid velocity
vector, P is the pressure and µ is the dynamic vis-
cosity. The SIMPLE pressure correction approach
[5] is applied to decouple the continuity and mo-
mentum equations. No slip conditions are imposed
on all land faces, rod and sleeve walls. Fluid den-
sity of ρ = 871kg/m3, and dynamic viscosity µ =
0.0375kg/m/s are used. These correspond to a typ-
ical hydraulic fluid (Mobil DTE 25) at 40◦C. The
rod and inner sleeve radii of the valve are Ri =
Fig. 9. Flow patterns and estimation of jet angles
for the meter-in case: (from left to right) xv =
−0.635,−1.27,−1.905 and −2.54 mm.
3.175mm and Ro = 6.35mm. Sixteen models corre-
sponding to combinations of four spool displacements
xv = 0.635mm,1.27mm, 1.905mm and 2.54mm, and
four single chamber damping lengths L = −L1 =
−0.216m, −0.118m (meter-in chamber) and L = L2 =
0.118m, 0.216m (meter-out chamber) are investigated.
Constant inlet pressure Ps = 689475.7Pa (100psi) and
outlet pressure PT = 101300Pa (1 atm) are imposed.
Solutions are obtained using FLUENT 6 (Fluent Inc.,
NH) on the IBM SP supercomputer at the University
of Minnesota. The various components of the steady
flow force and momentum fluxes can be directly mea-
sured and are used to evaluate the various models in
Section III.
A. Variable Orifice Flux
First, we investigate the variable orifice flux. The
meter-in and meter-out flow patterns are shown in
Figs. 9-10. The key parameter group cos(θ)/Cc in
Eq.(5) is potentially a function of the spool displace-
ment. The jet angle θ for both meter-in and meter-out
chambers at different spool displacements may be ob-
tained from the flow patterns shown in figures 9 and
10. For sufficiently large xv, θ ≈ 69◦ as predicted in
[3]. A contraction coefficient of Cc = 0.6 has been
cited in the literature [3]. This gives an estimate of
cos(θ)/Cc ≈ 0.60.
10
Fig. 10. Flow patterns and estimation of jet an-
gles for meter-out case (from left to right) xv =
0.635,1.27,1.905 and 2.54 mm.
0 1 2 3 4 5 6 70
0.5
1
1.5
2
2.5
3
3.5
ρ Q2/ Ao (N)
Var
iabl
e or
ifice
flux
(N)
L=−0.216mL=−0.118mL=0.118mL=0.216m
Fig. 11. CFD computed variable orifice momentum versus
ρQ2/Ao(xv). The solid line is the estimate based on
the least squares fit with cos(θ)/Cc = 0.454.
Fig. 11 shows the relationship between the
measured variable orifice momentum flux and
ρQ2/Ao(xv), where Ao(xv) is the orifice area of the
circular ports used in this study. The linearity of the
data indicates that a constant least squares estimated
average cosθ/Cc = 0.454 suffices for all spool displace-
ments and damping lengths. This value suggests that
an equivalent contraction coefficient of Cc ≈ 0.79,
which is larger than suggested in [3].
B. Non-orifice Flux
Next, we consider the non-orifice flux component of
the steady flow force. To illustrate, the flow patterns
at the non-orifice inlet / outlet for a 2D (i.e. 3D ax-
−0.2 −0.1 0 0.1 0.2 0.30
0.5
1
1.5
2
2.5
3
Damping length (m)
Var
iabl
e O
rific
e flu
x (N
)
xv=0.635mmxv=1.27mmxv=1.905mmxv=2.54mm
Fig. 12. Variable orifice fluxes as the function of L, for
various xv. The red solid lines are direct CFD mea-
surements, while the blue dotted lines are estimated
based on cos(θ)Cc
= 0.454.
(a) meter-out (inlet) (b) meter-in (outlet)
Fig. 13. Non-orifice flow pattern with the hose designed
to be normal to the spool axis.
ial symmetric) valve model are shown in Figure 13.
Notice that despite the hoses are normal to the spool
axis, the flow patterns at the non-orifice inlet or out-
let are not normal to it, so that non-zero longitudinal
momentum fluxes are expected.
Figure 14 shows the relationship between the mea-
sured non-orifice flux and Q2 for various flowrates
and damping lengths. The linearity of the meter-
in and meter-out data confirms that the momentum
fluxes in Eqs. (5) and (12) can indeed be mod-
eled using constants cin = 4.26 × 106Ns2m−6 and
cout = 1.79× 106Ns2m−6 for all flows, spool displace-
11
0 0.2 0.4 0.6 0.8 1
x 10−7
−0.45
−0.4
−0.35
−0.3
−0.25
−0.2
−0.15
−0.1
−0.05
0
Q2 (m6 s−2)
Non
orif
ice
flux
(N)
meter−in (L=−0.216m)meter−in (L=−0.118m)meter−out (L=0.118m)meter−out (L=0.216m)
Fig. 14. Non-orifice flux as a function of Q2 for the sixteen
3D CFD models. The regression lines correspond to
cin = 4.26×106Ns2m−6 and cout = 1.79×106Ns2m−6
ments, and damping lengths. Notice that the non-
orifice outlet flux coefficient cin in the meter-in cham-
ber is more than twice that of the non-orifice inlet
flux coefficient cout in the meter-out chamber. This
is consistent with the observation in Fig. 13 that the
flow is less obtuse in the meter-in outlet than in the
meter-out inlet.
To further confirm that cin and cout are not de-
pendent on damping length and spool displacement,
extensive studies using 2D (equivalent to 3D axis-
symmetric) CFD models with multiple flowrates at
each spool displacements and damping lengths were
performed. Figure 15 shows that cin and cout can be
treated as constants in these 2D cases as well.
C. Viscosity effect
Fig. 16 shows Fsleeve, computed from the funda-
mental equations (6) and (13) as a function of µQL for
the 16 test geometries. The linear nature of the curve
indicates that the laminar and fully developed flow ap-
0 0.5 1 1.5 2 2.5 30.5
1
1.5
2
2.5
3x 10
5
xv (mm)
cin
an
d c
out (N
s2/m
6)
cin
c
out
−250 −200 −150 −100 −50 0 50 100 150 200 2500.5
1
1.5
2
2.5
3x 10
5
damping length (mm)c
out (N
s2/m
6)
cin
c
out
Fig. 15. Estimated cin and cout for various spool displace-
ments and damping lengths L in a 2D CFD model.
Top: Various xv and L = 0.15m. Bottom: Various L
and xv = 0.645mm.
proximations in (7) and (14) are accurate. The least
squares estimate for the coefficient α is 1 × 106m−2
and the one computed analytically from Eq. (8) is
0.8× 106m−2. Fig. 16 shows that these estimates are
quite close.
D. Total Steady Flow Force
Consider the following expressions for the steady
flow force that use different approximations:
Fsteady,0 = Fland + Frod (18)
Fsteady,1 = −Fefflux,1 + Fsleeve,1 (19)
Fsteady,2 = −Fefflux,2 + Fsleeve,2 (20)
Fsteady,inviscid = −Fefflux,2 (21)
12
−3 −2 −1 0 1 2 3
x 10−6
−4
−3
−2
−1
0
1
2
3
4
µQL(Kg m3 /s2)
Sle
eve
forc
e (N
)
Fsleeve,1
Fsleeve,2
Fig. 16. Measured sleeve force Fsleeve as the function
of µQL. Sixteen diamond points correspond to the
sleeve forces computed from sixteen CFD models, and
the dotted line is the least squares fit. The solid line is
calculated from Eq. (7) or Eq. (14) with α computed
from Eq. (8).
where Fsteady,0 is the most fundamental computed di-
rectly based on Eq.(1); Fsteady,1 is the approach based
on the fundamental momentum equation in Eqs. (4),
(6) and (13); Fsteady,2 is based on the approximate mo-
mentum equations Eqs. (5), (7), (12), (14) using the
estimated parameters; Fsteady,inviscid is the estimate of
the steady flow forces that neglects viscosity. Notice
in Fig. 17 shows that Fsteady,0, Fsteady,1, and Fsteady,2
are almost exactly the same. Both the fundamental
and approximate momentum methods to calculate the
steady flow forces accurate. Furthermore, Fig. 17 and
18 shows that the viscosity effect plays a significant
role in determining the steady flow forces. It reduces
the steady flow force for negative damping lengths,
while increasing it for positive damping lengths. The
extent to which the steady flow force varies increases
with the damping length.
(a)L = −L1 = −0.216m
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10−3
−6
−5
−4
−3
−2
−1
0
xv (m)
Stea
dy flo
w fo
rce
(N)
Fsteady,0
Fsteady,1
Fsteady,2
Fsteady,Inviscid
(b)L = −L1 = −0.118m
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10−3
−6
−5
−4
−3
−2
−1
0
xv (m)
Stea
dy flo
w fo
rce
(N)
Fsteady,0
Fsteady,1
Fsteady,2
Fsteady,Inviscid
(c)L = L2 = 0.118m
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10−3
−6
−5
−4
−3
−2
−1
0
xv (m)
Stea
dy flo
w fo
rce
(N)
Fsteady,0
Fsteady,1
Fsteady,2
Fsteady,Inviscid
(d)L = L2 = 0.216m
0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4
x 10−3
−6
−5
−4
−3
−2
−1
0
xv (m)
Stea
dy flo
w fo
rce
(N)
Fsteady,0
Fsteady,1
Fsteady,2
Fsteady,Inviscid
Fig. 17. Steady flow forces computed from various meth-
ods as a function of the orifice displacement. Four
different damping lengths are considered.
13
−0.2 −0.1 0 0.1 0.2 0.3−6
−5
−4
−3
−2
−1
0
1
Damping length (m)
Ste
ady
forc
e (N
)
xv=0.635mmxv=1.27mmxv=1.905mmxv=2.54mm
Fig. 18. Steady flow forces as a function of the single
chamber damping length.
Fig. 18 shows the dependence of the steady flow
force for each meter-in / meter-out chamber on damp-
ing lengths. Notice the discontinuity of the flow
force at L = 0. This is related to the difference
between the meter-in and meter-out non-orifice flux,
(cin − cout)Q2 so that the discontinuity decreases for
smaller xv (hence Q). Fig. 19 shows that total steady
flow force for the two-chamber 4 way valve according
to Eq. (15). Compared to Fig. 18, no discontinuity
is present because both the non-orifice inlet and out-
let fluxes are simultaneously present in all damping
lengths. Fig. 19 also shows that both the viscos-
ity and non-orifice flux are more significant at high
flow rate. The non-orifice effect is illustrated by the
parallel shift and is given by (cin + cout)Q2. CFD
analysis reveals that the non-orifice flux as a propor-
tion of the variable orifice flux will be even larger for
axis-symmetric (i.e. 2D) valves, than in the 3D axial-
assymmetric case in Fig. 8.
−0.2 −0.1 0 0.1 0.2 0.3−10
−9
−8
−7
−6
−5
−4
−3
−2
−1
0
A
A
A
A
B
B
B
B
Damping length (m)
Ste
ady
forc
e (N
)
xv=0.635mmxv=1.27mmxv=1.905mmxv=2.54mm
Fig. 19. Steady flow forces and their estimates in a two
chamber 4-way valve as a function of the damping
length for fours sets of spool displacements. “A” -
take into account non-orifice flux, “B” - do not take
into account non-orifice flux. The horizontal bars at
L = 0 indicate the estimates if viscosity effect is ig-
nored.
(a) Meter-out (inlet) (b) Meter-in (outlet)
Fig. 20. Non-orifice flow pattern for a 2D CFD model with
the hose is designed to be +30◦ clockwise rotated from
the normal direction to the spool axis.
(a) meter-out (inlet) (b) meter-in (inlet)
Fig. 21. Non-orifice flow pattern for a 2D CFD model with
with the hose is designed to be −30◦ clockwise rotated
from the normal direction to the spool axis.
14
Hose angle meter-in meter-out sum
30◦ 1.17 0.09 1.26
0◦ 1.00 0.45 1.45
−30◦ 1.20 0.80 2.00
TABLE I
Meter-in and meter-out non-orifice flux for
various non-orifice hoses angles, normalized by
that of the meter-in, 0circ hose angle case.
E. Modifying the non-orifice flux
Since the non-orifice fluxes tend to reduce steady
flow forces, it would be advantageous if it can be ma-
nipulated by design. One straightforward idea is to
change the angle of the non-orifice hose to the spool
axis. Figs. 13, 20 and 21 illustrate the differences in
flow patterns for three hose angles for a 2D (or 3D
axis-symmetric) CFD model. Table I shows the non-
orifice flux values for various hose angles, normalized
by the value for the 0◦ hose angle, meter-in chamber
case. It can be seen that the non-orifice flux is suscep-
tible to changes in geometric configuration of the non-
orifice hose. In particular, by rotating the inlet and
outlet hoses in the counter-clockwise direction, the
non-orifice fluxes in both the meter-in and meter-out
chambers increase. If the hose is rotated 30◦ counter-
clockwise, the sum of the non-orifice fluxes of the two
chambers increases by 37%, thus having the effect of
further reducing the steady flow force.
V. Experimental study
The experiment aims at verifying the steady flow
force models in section III and that the viscosity
and the non-orifice flux have significant effects on the
steady flow forces, and hence on the spool agility.
Inlet/outlet hoses
Blocked quick coupler
opened quick coupler
Orifice
Force sensor
A B C D E F G H I J K
Fig. 22. The diagram of experimental setup for measuring
the steady flow forces
The experimental set up (Fig. 22) uses the custom
built valve in Fig. 2. The geometry of the valve is
similar to the one used in the CFD studies in Section
IV. In order to measure the steady flow force, the
shaft of the valve is connected in series with a force
sensor (67M25A, JR3 Inc.) that can measure both
compression and tension along the spool axis. A dig-
ital flow meter (DFS-3, Digiflow Systems) connected
downstream to the valve and an adjustable bypass
needle valve are used to adjust the desired flow rate
through the valve.
In the experiment, the orifice area is set to be
6×10−6m2. The flow forces were measured in 24 con-
ditions: combination of four positive (L = L2), four
negative (L = −L1) single chamber damping lengths:
L = ±236mm, ±175mm, ±118mm, ±62mm); and
three flow rates Q = 3.7 × 10−5m3/s (2.2LPM),
6.7× 10−5m3/s (4LPM), and 1× 10−4m3/s (6LPM).
Fig. 23 shows the measured steady flow forces and
the corresponding estimates based on Eqs. (15) using
parameters estimated in Section IV. The measured
steady flow force values are qualitatively similar to the
model in Fig. 18 in that i) the steady flow force has
an affine dependence on the damping length; ii) the
15
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
Ste
ad
y f
orc
e(N
)
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
Ste
ad
y f
orc
e(N
)
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
Ste
ad
y f
orc
e(N
)
Damping length(mm)
Fig. 23. Measurement and estimation of steady flow forces
as a function of the damping length, for various flow
rates. The solid line is a regression curves assuming
that for each Q, the sensitivity to damping length kL
is the same for all damping lengths. Top: Q = 2.2
LPM, Middle: Q = 4.0 LPM, Bottom: Q = 6.0 LPM.
sensitivity to damping length kL := ∂Fsteady/∂L in-
creases with flow rate; iii) the steady flow force is dis-
continuous at L = 0 with the discontinuity increases
with flow rate. Notice in particular, that as predicted,
the steady flow force is reduced for negative damping
lengths, and for sufficiently negative damping lengths,
the steady flow force becomes marginally stable or
even unstable (reverses sign).
Notice that for each Q, the steady flow force can
be fitted by two straight lines (L > 0 and L < 0)
using the the same slope kL := ∂Fsteady/∂L. The re-
0 0.5 1 1.5 2 2.5 3 3.5 4
x 10−6
0
0.5
1
1.5
2
2.5
3
3.5
4
µ Q(kg m2 s−2)
k L(N m
−1 )
Fig. 24. Relationship between flow force sensitivity to
damping length kL and flow rate Q. The slope should
be αµ.
lationship between kL and Q are shown in Fig. 24 and
indicates a straight line with a slope which should be
α/µ where µ is the dynamic viscosity. This gives a
value of α = 0.978 × 106 which is close to the val-
ues obtained by CFD in Section IV (1.0× 106) or via
Eq.(8) (0.8× 106).
Despite excellent match in the viscosity effect, there
are some significant quantitative discrepancies be-
tween the experimental results in Fig. 23 and the
model. First, the discontinuity at L = 0 between
the meter-in and meter-out chamber is significantly
larger than that predicted by the expected non-orifice
flux differences. Second, the experimental data tend
to be more negative than the prediction, especially
when L > 0. Extensive effort was spent to investi-
gate the source of these disparities. Two limitations
of our experimental setup can be used for their expla-
nations. Firstly, since the valve lands are machined
by hand, it turns out that the areas of the two spool
lands are different. The area difference was estimated
by measuring the spool force when the closed cham-
16
ber formed using the two lands is pressurized. The
land at the variable orifice was found to be smaller
by ∆A = 0.186mm2. This leads to an extra orifice-
closing force ∆1 ≈ ∆A · P (negative force in the fig-
ure) as the chamber is pressurized. Using the pressure
measured using the digital pressure transducers dur-
ing the experiments, the steady force offset ∆1 are
shown in table II.
Secondly, the shape of the lands need to be consid-
ered. To avoid hydraulic locking and to reduce friction
between the lands and the sleeve, the lands were in-
tentionally machined to have a 2.5◦ taper to form a
hydrostatic bearing. The effect of the taper can be
seen in Fig. 25 where the control volume is different
from that in Figs. 6 and 7. Modifying Eqs. (1)-(2)
accordingly, the spool force of the tapered land valve
is:
Fsteady,t = −Fefflux + Fsleeve + ∆1
+
∫
s−lP (nx · n)dA−
∫
o−pP (nx · n)dA
︸ ︷︷ ︸∆2
(22)
where n is the outward normal vector of the surface,
and nx represents the unit vector in the positive x
direction. The offset force due to the taper, ∆2, con-
sists of the longitudinal pressure force applied on the
orifice plane (o-p) and on the side surface (s-l) respec-
tively. It can be seen that ∆2 > 0 for the meter-in
case because the first term of ∆2 is greater than the
second term, while ∆2 < 0 for the meter-out case
because of the opposite effect. The numerical values
of ∆2 was obtained via CFD analysis of a 2D (i.e.
axis-symmetric 3D) model at various pressure differ-
ences across the orifice. The ∆2 values for the 3D
axis-asymmetric case were then estimated from the
Q (LPM) 2.2 4.0 6.0
Meter-in ∆1 (N) -0.16 -0.20 -0.25
(L < 0) ∆2 (N) 0.17 0.27 0.40
Meter-out ∆1 (N) -0.14 -0.16 -0.20
(L > 0) ∆2 (N) -0.058 -0.089 -0.141
TABLE II
Estimation of ∆1 and ∆2 (N)
x
Side sufrace of the right land
Orifice plane
Fig. 25. Modified control volume of the valve with tapered
lands. The variable orifice land area is also smaller
than the area of the left hand side land. The control
volume is the gray colored block.
2D case with the same pressure difference, corrected
for the port area difference (3D case value is approxi-
mately 50% of the equivalent 2D case value). The spe-
cific values for the experimental conditions are listed
in Table II. The predicted flow forces after incorpo-
rating the offsets ∆1 + ∆2 match the experimental
results very closely (Fig. 26). Even better match can
be obtained by slightly adjusting ∆2. The fact that
∆1 and ∆2 are of different signs for L < 0 explains the
observation that there is less discrepancies for L < 0
than for L > 0 in Fig. 23. In a commercial valve,
where the lands can be precisely machined, the land
area difference is expected to be much smaller, and no
taper or a much smaller taper will generally be used.
Therefore, in practice, the ∆1 +∆2 correction will not
be needed.
17
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
−250 −200 −150 −100 −50 0 50 100 150 200 250−2
−1
0
Damping length L (mm)
Ste
ad
y flo
w fo
rce
s(N
)
Fig. 26. Experimentally measured steady flow force com-
pared with models that take into account limitation of
the experimental setup. Top: Q = 2.2 LPM, Middle:
Q = 4.0 LPM, Bottom: Q = 6.0 LPM.
VI. Conclusion
In this paper, we demonstrate that the fluid viscos-
ity and the non-orifice flux are very important in the
estimation of the steady flow forces. On one hand, in
the negative damping length region, the steady flow
forces, conventionally regarded to be always stabiliz-
ing, can be reduced to be marginally stable or unsta-
ble, thus improving the agility of the spool. Paradoxi-
cally, using a valve with negative damping lengths, the
spool’s agility will improve as the fluid becomes more
viscous. On the other hand, the non-orifice flux can
be tuned by changing the angle of the non-orifice hose
to the spool axis, so that the net momentum efflux,
which is traditionally regarded to be always stabiliz-
ing, could possibly be reduced to be marginally stable
or unstable. The viscosity and the non-orifice effects
predicted by the proposed models and CFD analysis
are verified in experiment.
The viscosity and non-orifice flux effects generate
other valve design parameters that can be used to
improve spool agility in single stage electrohydraulic
valves. Since the steady flow force is a significant force
that the solenoid actuator has to overcome, the reduc-
tion of the steady flow force via the viscosity effect will
be useful for developing high performance single stage
valves.
References
[1] Kailash Krishnaswamy and Perry Y. Li, “On using unsta-
ble electrolydraulic valves for control,” in 2000 ACC Pro-
ceedings, May 2000, Also to appear in ASME Journal of
Dynamic Systems, Measurement and Control.
[2] J. F. Blackburn, G. Reethof, and L. L. Shearer, Fluid Power
Control, MIT Press, 1960.
[3] Hebert E Merritt, Hydraulic Control System, John Wiley
and Sons, 1967.
[4] B. R. Munson, D. F. Young, and T. H. Okiishi, Fundamen-
tals of Fluid Mechanics, John Wiley and Sons, New York,
1998.
[5] John C. Tannehill, Dale A. Anderson, and Richard H.
Pletcher, Computational fluid mechanics and heat transfer,
Taylar and Francis, Philadephia, PA, 1997.
Appendix
For the valve with geometry in Fig. 6, the sleeve
force is fundamentally given by:
Fsleeve,1 = 2πRoµ
∫ 0
−(Ld+L1)
∂vx∂r
∣∣∣∣r=Ro
dx (23)
where ∂vx∂r
∣∣r=Ro
is the longitudinal velocity gradient
in the outward radial (r) direction evaluated at the
18
sleeve wall (r = Ro). If the friction in the dead space
is negligible,
Fsleeve,1 ≈ 2πRoµ
∫ 0
−L1
∂vx∂r
∣∣∣∣r=Ro
dx. (24)
For laminar and fully developed flow, it can be shown
that [4]:1
r
d
dr(rdvxdr
) =1
µ
dP
dx(25)
where P (x) is the pressure. Applying the no-slip con-
ditions that vx = 0 at the rod wall (r = Ri) and the
sleeve wall (r = Ro), the velocity distribution can be
determined to be:
vx =1
4µ
dP
dx
[r2 −R2
o +R2i −R2
o
ln(Ro/Ri)ln
r
Ro
](26)
Since
−∫ Ro
Ri
2πvxdr = Q,
∂vx∂r can be expressed as a function of Q so that α in
(7) is given by:
α =4(2R2
o ln(Ro/Ri) +R2i −R2
o)
(R4o −R4
i ) ln(Ro/Ri)− (R2i −R2
o)2
(8)