Professional Documents
Culture Documents
Abstract
In this paper, we developed a detailed mathematical model of dual action pneumatic
actuators controlled with proportional spool valves. Eects of nonlinear ow through
the valve, air compressibility in cylinder chambers, leakage between chambers, end of
stroke inactive volume, and time delay and attenuation in the pneumatic lines were
carefully considered. System identication, numerical simulation and model validation
experiments were conducted for two types of air cylinders and dierent connecting
tubes length, showing very good agreement. This mathematical model will be used in
the development of high performance nonlinear force controllers, with applications in
teleoperation, haptic interfaces, and robotics.
Introduction
This article has appeared in the September 2000 issue of ASME Journal of Dynamic Systems Measurement and Control, Vol. 122, No.3, pp. 416-425
mechanisms. Yet, applications such as teleoperation master arms with gravitational compensation, require long duration, static high force output. In these cases, direct drive electrical
actuators necessitate special cooling systems to dissipate the excessive heat.
We believe that pneumatic cylinders can oer a better alternative to electrical or hydraulic
actuators for certain types of applications. Pneumatic actuators provide the previously
enumerated qualities at low cost. They are also suitable for clean environments and safer
and easier to work with. However, position and force control of these actuators in applications
that require high bandwidth is somehow dicult, because of compressibility of air and highly
nonlinear ow through pneumatic system components. In addition, due to design and space
considerations, in many applications the command valve is positioned at relatively large
distance from the pneumatic cylinder. Thus, the eects of time delay and attenuation due
to the connecting tubes becomes signicant. Due to these diculties, early use of pneumatic
actuators were limited to simple applications that required only positioning at the two ends
of the stroke. Subsequently, more complete mathematical models for the thermodynamic
and ow equations in the charging-discharging processes were developed (Shearer, 1956).
As a result more complex position controllers, based on the linearization around the mid
stroke position, were developed (Burrows, 1966; Liu and Bobrow, 1988). These simplied
models provided only modest performance improvements. During the last decade, nonlinear
control techniques were implemented using digital computers. Bobrow and Jabbari, 1991,
and McDonell and Bobrow, 1993 used adaptive control for force actuation and trajectory
tracking, applied to an air powered robot. Improved results were presented for low frequencies
(approx. 1 Hz). Sliding mode position controllers were also tested (Arun et al., 1994; Tang
and Walker, 1995), again with improved results at low frequencies. As a general characteristic
the mathematical models used in these controllers assumed no piston seals friction, linear
ow through the valve, and neglected the valve dynamics. Ben-Dov and Salcudean, 1995
developed a forced-controlled pneumatic actuator that provided a force with an amplitude
of 2 N at 16 Hz. Their model included the valve dynamics and the nonlinear characteristics
of the compressible ow through the valve. A comparison between linear and nonlinear
controllers applied to a rotary pneumatic actuator is presented by Richard and Scavarda,
1996. The mathematical model accounted for the leakage between actuators chambers and
the nonlinear variation of the valve eective area with the applied voltage. Their model
depends heavily on curve tting using experimental values, making dicult the application
of the model to even slightly dierent systems.
The goal of this article is to provide an accurate model of a pneumatic actuator system controlled using a proportional spool valve. The model takes into consideration the
friction in piston seals, the dierence in active areas of the piston due to the rod, inactive
volume at the ends of the piston stroke, leakage between chambers, valve dynamics and ow
nonlinearities through the valve orice, and time delay and ow amplitude attenuation in
valve-cylinder connecting tubes. Since the ultimate purpose of the modeling eort is for
control applications, the proposed equations should be suitable for on-line implementation.
We designed special experiments in order to identify the unknown characteristics of the pneumatic system, such as: valve discharge coecient, valve spool viscous friction coecient, and
2
piston friction forces. The model was nally tested using two experiments that allowed the
measurement of the actuator force output and piston displacement. The experimental results
were compared with the results obtained by numerical simulation.
System Dynamics
A typical pneumatic system includes a force element (the pneumatic cylinder), a command
device (valve), connecting tubes, and position, pressure and force sensors. The external load
consists of the mass of external mechanical elements connected to the piston and perhaps a
force produced by environmental interaction. A schematic representation of the pneumatic
actuator system is shown in Fig. (1), with variables of interest specied for each component.
;;;;;;;;;;;
;;;;;;;;;;;
Cylinder
;;;;;;;;;;;
;;;;;;;;;;;
P ,V , A
P ,V , A
M A
;;;;;;;;;;;
;;;;;;;;;;; F
;;;;;;;;;;;
;;;;;;;;;;;
P Pressure sensors
P
-
+ Position sensor
p,
Load
r
ML
FL
xs +
At, Lt
;;;
;;; Valve
;;;;;;;
;;;
;;;;;;;
;;;
;;;;;;;
ic
Pa
Pa
Exhaust
Exhaust
Ps
Supply
2.1
Piston-Load Dynamics
The equation of motion for the piston-rod-load assembly can be expressed as,
x + x + Ff + FL = P1 A1 P2 A2 Pa Ar
(ML + Mp )
(1)
where ML is the external load mass, Mp is the piston and rod assembly mass, x is the
piston position, is the viscous friction coecient, Ff is the Coulomb friction force, FL is
the external force, P1 and P2 are the absolute pressures in actuators chambers, Pa is the
absolute ambient pressure, A1 and A2 are the piston eective areas, and Ar is the rod cross
sectional area. The right hand side of Eq. (1) represents the actuator active force, produced
by the dierent pressures acting on the opposite sides of the piston. In order to control the
actuator force output, one has to nely tune the pressure levels in the cylinder chambers
using the command element (the pneumatic valve). This requires detailed models for the
dynamics of pressure in both chambers of the actuator, valve dynamics, and connecting
tubes.
2.2
In this section we seek to develop a mathematical relationship linking the pressure change
with the mass ow rate and piston translational speed is necessary for each pneumatic
cylinder chamber. In previous works (see Liu and Bobrow, 1988; Ben-Dov and Salcudean,
1995; Richard and Scavarda, 1996) the authors derived this equation using the assumption
that the charging and discharging processes are both adiabatic. Al-Ibrahim and Otis, 1992,
found experimentally that the temperature inside the chambers lays between the theoretical
adiabatic and isothermal curves. The experimental values of the temperature were close
to the adiabatic curve only for the charging process. For the discharging of the chamber
the isothermal assumption was closer to the measured values. In this article we derive the
pressure dynamics equation in a way that accounts for the dierent thermal characteristics
of the charging and discharging processes of the cylinder chambers.
The most general model for a volume of gas consists of three equations (see Hullender
and Woods, 1985): an equation of state (ideal gas law), the conservation of mass (continuity)
equation, and the energy equation. Assuming that: (i) the gas is perfect, (ii) the pressures
and temperature within the chamber are homogeneous, and (iii) kinetic and potential energy
terms are negligible, these equations can be written for each chamber. Considering the
control volume V , with density , mass m, pressure P , and temperature T , the ideal gas law
can be written as,
P = RT
(2)
where, R is the ideal gas constant. Applying the continuity equation the mass ow rate can
be expressed as,
d
(3)
m
= (V )
dt
(4)
where, m
in and m
out are the mass ows entering and leaving the chamber.
The energy equation can be written as follows:
= U
in Tin m
out T ) W
qin qout + kCv (m
(5)
where qin and qout are the heat transfer terms, k is the specic heat ratio, Cv is the specic
is the rate of
heat at constant volume, Tin is the temperature of the incoming gas ow, W
change in the work, and U is the change of internal energy. The total change in internal
energy is,
1
1 d
d
(P V ) =
(V P + P V )
(6)
U = (Cv mT ) =
dt
k 1 dt
k1
= P V and
in which we used the ideal gas relation, Cv = R/(k 1). Now, substituting W
Eq. (6), into Eq. (5),
qin qout +
k P
k
1
out T )
(m
in Tin m
P V =
V P
k 1 T
k1
k1
(7)
Assuming that the incoming ow is already at the temperature of the gas in the chamber
considered for analysis, the energy equation becomes,
1
V
k1
P
out ) V =
(qin qout ) + (m
in m
k
kP
(8)
Further simplication can be made by analyzing the heat transfer terms in Eq. (8). If
the process is considered to be adiabatic (qin qout = 0), the time derivative of the chamber
pressure is,
P
P
(9)
P = k
out ) k V
(m
in m
V
V
or, substituting from Eq. (2),
RT
P
P = k
(m
in m
out ) k V
V
V
(10)
(11)
and Eq. (8) can be written as,
P
qin qout = P V (m
out )
in m
(12)
(13)
A comparison of Eqs. (10) and (13) shows that the only dierence is the specic heat
ratio term k. Thus, both equations can be written as,
RT
P
P =
in out m
out ) V
(in m
V
V
(14)
with , in , and out taking values between 1 and k, depending on the actual heat transfer
during the process. In equation (14) one does not have to know the exact heat transfer
characteristics, but merely estimate the coecients , in , and out . The fact that the
uncertainty of the estimation is bounded by k 1 is also very important from the control
design perspective. For the charging process, a value of in close to k is recommended, while
for the discharging of the chamber out should be choose close to 1. The thermal characteristic
of compression/expansion process, due to the piston movement is better described using
= 1.2 (see Al-Ibrahim and Otis, 1992).
Choosing the origin of piston displacement at the middle of the stroke, the volume of
each chamber can be expressed as,
1
Vi = V0i + Ai ( L x)
2
(15)
where i = 1, 2 is the cylinder chambers index, V0i is the inactive volume at the end of stroke
and admission ports, Ai is the piston eective area, L is the piston stroke, and x is the piston
position. The dierence between the piston eective areas for each chamber A1 and A2 is
due to the piston rod. Substituting Eq. (15) into (14), the time derivative for the pressure
in the pneumatic cylinder chambers becomes:
P i =
RT
P Ai
in out m
out )
(in m
x
1
V0i + Ai ( 2 L x)
V0i + Ai ( 12 L x)
(16)
In this new form the pressure dynamics equation accounts for the dierent heat transfer
characteristics of the charging and discharging processes, air compression or expansion due
to piston movement, the dierence in eective area on the opposite sides of the piston, and
the inactive volume at the end of stroke and the admission ports. The rst term in the
equation represents the eect on pressure of the air ow in or out of the chamber, and
the second term accounts for the eect of piston motion. There are two sources for the
ow entering a cylinder chamber: a) the pressure tank, through the pneumatic valve and
connecting tube, b) the neighboring chamber if it has a higher pressure and piston seals are
leaking. The air can ow out to the atmosphere through the valve or piston rod seals, or
to the second chamber if it has a smaller pressure. The leakage between chambers can be
neglected for regular pneumatic cylinders with rubber type seals, but can be signicant for
low friction cylinders that have graphite or Teon seals. The expressions for the input and
output ows will be derived in the following section.
6
2.3
The tubes that connect the valve with the actuator have two eects on the system response.
First, the pressure drop along the tube will induce a decrease in the steady state air ow
through the valve. Second, the ow prole at the outlet will be delayed with respect to the
one at the inlet by the time increment necessary for the acoustic wave to travel the entire
length of the tube. This will aect the transient response of the ow in the cylinder chambers.
The problem of the pressure losses and time delay in long pneumatic lines was analyzed by
many authors: Schuder and Binder, 1959; Hougen et al., 1963; Andersen, 1967; Whitmore
et al., 1990, Elmadbouly and Abdulsadek, 1994. Most of these investigators assumed fully
developed laminar ow through the tube. They provide innite series solutions for the
pressure dynamics, or approximate the response to harmonic pressure inputs using a second
order linear system.
Because we model for online control applications, in our derivation we seek a simpler
expression for the mass ow through the tube. The expression should not require intensive
computation like the series solution and should not increase the order of the system as
the second order linear approximation does. We also extend the analysis to include wholly
turbulent ow.
s
At, Lt
mt (0, t)= h(t)
mt (Lt, t)
(17)
(18)
where P is the pressure along the tube, u is the velocity, is the air density, c is the sound
speed, s is the tube axis coordinate, and Rt is the tube resistance. Introducing the mass
ow through the tube as m
t = At u, where At is the tube cross sectional area,
P
s
m
t
s
These equations are similar to those
term in Eq. (19) that accounts for
1 m
Rt
t
m
t
(19)
At t
At
At P
= 2
(20)
c t
in Elmadbouly and Abdulsadek, 1994, with the second
resistance of the tube. Dierentiating Eq. (19) with
=
respect with t and Eq. (20) with respect to s, we obtain the equation for the mass ow
through the tube as,
2
t
t
t Rt m
2m
2 m
c
+
=0
(21)
2
2
t
s
t
This equation represents a generalization of a wave equation, with an additional dissipative
term. The proposed mass ow equation can be solved by using the following form (see
Chester, 1971):
(22)
m
t (s, t) = (t)v(s, t)
where v(s, t) and (t) are new unknown functions. Substituting Eq. (22) into (21) yields,
2v
2v
Rt
v
Rt
2 c2 2 + + 2
+
+ v = 0
t
s
(23)
In order to simplify the equation for v, we determine (t) such that, after substitution in
Eq. (23), the remaining equation in v contains no rst derivative term (Chester, 1971),
2 +
Rt
=0
Rt
(24)
(t) = e 2 t
(25)
2
Rt2
2v
2 v
c 2 + 2v = 0
t2
s
4
(26)
which is a dispersive hyperbolic equation. Chester, 1971, showed that Eq. (26) had a solution
in the form of a progressive wave that propagates along the tube with any constant velocity
dierent than c. The fact that the solution waves do not propagate with the same velocity
is called dispersion, and it is caused by the term (Rt /2)2 v. The tubes under consideration
are not very long, so we assumed that the dispersion is small, and can be neglected. This
assumption yields,
2
2v
2 v
=0
(27)
t2
s2
This is the classic one-dimensional wave equation, and can be solved for specic boundary
and initial conditions. We assume that there is no ow through the tube at t = 0, the ow
at the inlet (s = 0) is an arbitrary time dependent function h(t), and there is no reection
from the end connected at the pneumatic cylinder. The corresponding initial and boundary
conditions become,
v(s, 0)
= 0
v
(s, 0) = 0
(28)
v(0, t)
= h(t)
8
0
s
v(s, t) =
h t
if t < s/c
if t > s/c
(29)
The input wave will reach the end of the tube in a time period = Lt /c. Replacing t by
Lt /c in Eq. (25), and substituting from the equation of state, the attenuation component
in Eq. (22) becomes,
Rt RT Lt
= e 2P c
(30)
where P is the end pressure. The mass ow at the outlet of the tube (s = Lt ) is,
m
t (Lt , t) =
Rt RT Lt
2P
c
Lt
h t
c
if t < Lt /c
if t > Lt /c
(31)
Equation (31) describes in a simple form the mass ow at the tube outlet, for any inlet
ow. It shows that the ow at the outlet of the tube is attenuated in amplitude and delayed
by Lt /c, which represents the time required by the input wave to travel through the entire
length of the tube. This solution can not account for the dependence of the amplitude
attenuation on the frequency of the input ow. Its application have to be restricted to
relatively small frequencies. The experimental results presented by Hougen et al., 1963,
showed very small amplitude dependence on frequencies up to 50 Hz, for tubes up to 15 m in
length. Considering the fact that Eq. (31) is applied to the valve-cylinder connecting tubes,
this limitations are considered more than satisfactory for usual applications.
The tube resistance Rt can be obtained from the expression for the pressure drop along
the tube (see Munson et al, 1990),
p = f
Lt u2
= Rt uLt
D 2
(32)
where f is the friction factor, and D the inner diameter of the tube. For fully developed
laminar ow f = 64/Re, where Re is the Reynolds number. The tube resistance becomes,
Rt =
32
D2
(33)
where is the dynamic viscosity of air. This result is also derived in Schuder and Binder,
1959. Extending the analysis for wholly turbulent ow in smooth tubes (the plastic tubes
used in this work can be considered smooth), the friction factor can be computed using the
Blasius formula (Munson et al, 1990),
f=
0.316
Re1/4
(34)
-3
mt [x10 Kg/s]
-2
Input flow
Solution from Eq. (31)
Solution from Eq. (21)
-4
-6
0
20
40
60
t [x10
-3
80
100
sec ]
Re3/4
D2
(35)
In order to use Eq. (31) one has to evaluate the Reynolds number from the input ow values,
and compute Rt using Eq. (33) or (35) accordingly.
We compared the results given by Eq. (31) with those obtained by numerical integration
of Eq. (21), for a plastic tube with 3.2 mm internal diameter and 3 m length. The input
ow was considered to be sinusoidal, with 6 103 Kg/s amplitude and 30 Hz frequency
(see Fig. (3)). One can observe from the gure that the simplied equation can eectively
predict time delay and the amplitude attenuation.
2.4
Valve Model
The pneumatic valve is a critical component of the actuator system. It is the command
element, and should be able to provide fast and precisely controlled air ows in and out of
the actuator chambers. There are many available designs for pneumatic valves, which dier
by geometry of the active orice, type of ow regulating element, number of paths and ports,
type of actuating, etc. We restricted our study to proportional spool valves, actuated by
voice coil. This design presents several advantages: quasi-linear ow characteristic, small
time constant, small internal leakage, ability to adjust both chambers pressure using one
10
control signal, very small hysteresis, and small internal friction. In addition, there are many
commercially available quality valves. We used a PositioneX, four-way, proportional valve
produced by Numatics Inc. The valve features a lapped spool-sleeve assembly, with very low
friction. The spool is balanced with respect to pressure and positioned at the equilibrium
(closed) position using two coil springs. This design permits fast and precise adjustments
of the valve orice area, providing accurate ow control. If the spool is displaced in the
positive direction, one chamber will be connected to the pressure tank through the supply
out 0). The other chamber will
path, and the compressed air will ow inward (m
in > 0, m
be connected to the atmosphere through the exhaust path, and the air will ow outward
out > 0).
(m
in 0, m
Now we present a model that is developed for the PositioneX valve. This analysis,
however, can be easily tailored to model any commercially available pneumatic proportional
spool valve. Modeling of the valve involves two aspects: the dynamics of the valve spool,
xs
Ms
Fc
ks
+
cs
ks
Ff
(36)
where xs is the spool displacement, xso is the spring compression at the equilibrium position,
Ms is the spool and coil assembly mass, cs is the viscous friction coecient, Ff is the Coulomb
friction force, ks is the spool springs constant, and Fc is the force produced by the coil.
Simplifying the spring force expressions yields:
Ms xs + cs x s + Ff + 2ks xs = Fc
(37)
(38)
m
v=
Pu
Cf Av C1
if
1/k
Pu Pd
Pd (k1)/k
1
Cf Av C2
Pu
Pu
Pd
Pcr
Pu
(39)
Pd
if
> Pcr
Pu
where m
v is the mass ow through valve orice, Cf is a nondimensional, discharge coecient,
Pu is the upstream pressure, Pd is the downstream pressure and,
C1 =
2
R k+1
k+1
k1
C2 =
2k
;
R(k 1)
2
Pcr =
k+1
k
k1
(40)
are constants for a given uid. For air (k = 1.4) we have C1 = 0.040418, C2 = 0.156174,
and Pcr = 0.528. The meaning of the upstream and downstream pressure in Eq. (39) is
dierent for the charging and discharging process of the cylinder chambers. For charging,
the pressure in the supply tank should be considered the upstream pressure and the pressure
in the cylinder chamber is the downstream one. For discharging process, the pressure in
the chamber is the upstream, and the ambient pressure is the downstream pressure. Same
expression can be used for the leakage ow between the chambers, with the valve area
replaced by an experimentally determined leakage area, and the upstream and downstream
pressures as the ones from the chambers.
The area of the valve is given by the spool position relative to the radial holes in the
valve sleeve, as it is shown in Fig (5). The area of the segment of the circle delimited by the
edge of the spool can be expressed as,
Ae = 2
xe
0
Rh2
( Rh
)2
d = 2
xe
0
(2Rh ) d
(41)
where Ae is the eective area for one radial sleeve hole, xe is the eective displacement of
the valve spool, and Rh is the hole radius. Integrating Eq. (41) and considering all active
holes for an air path in the sleeve (nh ), the eective area of one path in the valve is,
Av = nh
2Rh2
arctan
xe
2Rh xe
(Rh xe ) xe (2Rh xe )
(42)
The spool width (2 pw ) is slightly larger than the radius of the hole, to ensure that the
air paths are closed even in the presence of small valve misalignments. Thus, the eective
displacement of the valve spool xe will be dierent from its absolute displacement xs ,
xe = xs (pw Rh )
12
(43)
xe
2pw
Rh
xs
Avin
0
if xs pw Rh
Rh p w + xs
2
2
n 2R2 arctan
(pw xs ) Rh (pw xs )
h
h
=
R
+
p
x
h
w
s
if pw Rh < xs < pw + Rh
nh Rh
(44)
if xs pw + Rh
and,
Avex
nh Rh2
if xs pw Rh
p
+
|x
|
h
w
s
n 2Rh2 arctan
(pw |xs |) Rh2 (pw |xs |)2
= h
(45)
Rh + pw |xs |
if pw Rh < xs < Rh pw
0
if x R p
s
Valve areas for input and exhaust paths versus the spool displacement are presented in
Fig. (6). The direction of ows in cylinder chambers have opposite signs, when one chamber
is charged the other one is discharged, thus the role of the curves should be switched for the
second chamber.
Substituting Eqs. (44) and (45) in Eq. (39) we obtain the expressions for the input and
output valve ows.
13
30
25
Av [mm ]
20
15
10
0
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
xs [mm ]
System Identication
The mathematical model of the pneumatic system derived in the previous sections includes
a number of geometric and functional characteristics or parameters. Accurate values of
this parameters are required in order to use the model in practical applications. Some
parameters, such as cylinder bore diameter, piston stroke, or the length of the connecting
tubes are provided by the manufacturer or they can be easily measured. Other parameters,
such as the critical pressure ratio for chocked ow, can be calculated using generally accepted
formulas and values for the physical constants involved. The parameters which can not be
directly measured or calculated, have to be estimated (identied) using specially designed
experiments. The valve discharge coecient, spool viscous friction coecient, static and
dynamic friction force between the piston and the cylinder bore, and the piston viscous
friction coecient are parameters of the model that have to be identied experimentally.
An important factor in the pressure dynamics equation is the mass ow, which can be
controlled using the pneumatic valve. Thus, we initiate the system identication process
by considering the command element (valve). The value for the coil force coecient is
provided in the valve user manual (Kf c = 2.78N/A), and Ms and ks can be easily measured
by dismantling the spool. With these values known, the steady state value for the spool
displacement can be computed for any constant coil current as,
xs =
Kf c
ic
2 ks
(46)
Measuring the ow for several coil currents and tting the theoretical mass ow expression,
Eq. (39), on the experimental values as it is shown in Fig. (7), we determined the discharge
14
coecient Cf = 0.25. The last unknown characteristic of the valve is the viscous friction co3.0
-3
mf [x10 Kg/s]
2.5
2.0
1.5
1.0
Experimental values
Theoretical flow
0.5
0.0
0.0
0.1
0.2
0.3
0.4
0.5
ic [A]
15
1.0
Experimantal values
Theoretical flow
0.6
-3
mf [x10 Kg/s]
0.8
0.4
0.2
0.0
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
t [sec]
Ff =
Fsf
if x = 0
if x
= 0
Fdf sign(x)
(47)
where Fsf and Fdf are the static and dynamic friction forces, and
1 if x
sign(x) =
1
0 if x =
0
1 if x
1
(48)
In order to determine the friction force we performed a simple experiment. With the
piston at rest and air ports disconnected, we applied an increasing force at the rod end. The
piston remained xed until the external force reaches the value of the static friction, and
then started moving, with a smaller friction force, which included the dynamic Coulomb
force and viscous friction, and piston inertia force. Figure (10) shows the total resistance
force for the Numatics actuator, and the corresponding position and velocity of the piston.
The static friction is identied from the curve peak as Fsf = 3.8N . If we neglect the inertia
of the piston,in the time interval 0.15 - 0.40 sec, the resistant force can be expressed as,
+ x
Frf = Fdf sign(x)
(49)
Plotting the resistance force versus piston velocity, and tting Eq. (49) to the experimental
values (see Fig. (11), we separated the dynamic friction force Fdf from the viscous friction.
16
1.0
0.6
-3
mf [x10 Kg/s]
0.8
0.4
Experimantal values
Theoretical flow
0.2
0.0
150
200
250
300
350
400
450
500
550
P [x10 Pa]
The complete mathematical model for the pneumatic actuator system consists of the valve
dynamics equation, two equations for the chamber pressure time derivatives, and the pistonload equation of motion. The chamber pressure rate of change described by Eq. (16), has to
be customized for the two cylinder chambers using the corresponding input and exhaust ows,
and the corresponding upstream and downstream pressures for the charging or discharging
process. The valve ows through the input and output paths can be written as,
Ps
r (Ps , P )
m
vin = Cf Avin m
T
Pi
m
vex = Cf Avex m
r (Pi , Pa )
T
17
(50)
(51)
3.5
3.0
Ff [N]
2.5
2.0
1.5
1.0
0.5
0.0
10
0.25
Displacement
Velocity
0.20
-10
0.15
-20
0.10
-30
0.05
-40
0.00
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
t [sec]
18
v [m/s]
-3
X [x10 m]
2.0
Experimental values
Theoretical friction force
Ff [N]
1.5
1.0
0.5
0.0
0.00
0.05
0.10
0.15
0.20
v [m/s]
Figure 11: Dynamic and viscous friction force versus piston velocity, for Numatics 0750D0203A.
19
3.5
3.0
Ff [N]
2.5
2.0
1.5
1.0
0.5
0.0
20
0.5
Displacement
Velocity
0.4
0.3
-10
0.2
-20
0.1
-30
0.00
0.0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
t [sec]
20
v [m/s]
-3
X [x10 m]
10
2.0
Experimental values
Theoretical friction force
Ff [N]
1.5
1.0
0.5
0.0
0.0
0.1
0.2
0.3
0.4
v [m/s]
Figure 13: Cylinder friction force versus piston velocity, for Airpel E16 D 3.0 N.
where Pi ,
C1
m
r=
C2
if
Pd
Pu
1/k
Pd
1
Pu
(k1)/k
Pd
Pcr
Pu
Pd
if
> Pcr
Pu
(52)
is the reduced ow function. Using Eq. (31) the ows through the connecting tubes become,
m
tin = in m
vin (t )
vex (t )
m
tex = ex m
(53)
(54)
where in and ex are the ow attenuations from Eq. (30), and is the tube time delay.
Substituting these equations in Eq. (16) we obtain the nal equations for chambers pressure
as,
C
T
R
f
v1 Ps m
1 ) ex ex Av1ex P1 m
1 , Pa )
A
P1 =
(P
,
P
(
P
in
in
r
s
r
in
V01 + A1 ( 12 L + x)
P1 A1
x
(55)
V01 + A1 ( 12 L + x)
Cf R T
P2 =
A
A
P
P
m
(P
,
P
)
(
P
,
P
)
in in v2in s r
s
2
ex ex v2ex 2 r
2
a
V02 + A2 ( 12 L x)
21
P2 A2
x
V02 + A2 ( 12 L x)
(56)
where the variables with overbars represent values delayed by time . These equations,
together with Eqs. (1) and (38), completely describe the pneumatic system. They include
the eect of the external load, piston friction, the dierence in the eective areas on the two
sides of the piston on the actuator force output, the dierent heat transfer characteristics at
charging or discharging process and compression-expansion of air due to piston motion, the
ow attenuation and delay due to the tubes, nonlinear compressible ow through the valve,
and the valve dynamics. They can be numerically solved for any valve command ic , and can
be used as a mathematical model for control design.
150
500
100
400
50
F [N]
600
Experimental P1
Experimental P2
Theoretical P1, P2
300
0
Experimental Fa
Theoretical Fa
-50
200
-100
100
0
20
40
60
80
-3
20
40
60
80
-3
t [x10 sec]
t [x10 sec]
(a)
(b)
600
150
500
100
400
50
F [N]
Figure 15: Numerical and experimental actuator pressures (a) and force (b) for 0.5 A step
coil current and 0.5 m tube length.
Experimental P1
Experimental P2
Theoretical P1, P2
300
0
Experimental Fa
Theoretical Fa
-50
200
-100
100
0
20
40
60
80
-3
20
40
60
80
-3
t [x10 sec]
t [x10 sec]
(a)
(b)
Figure 16: Numerical and experimental actuator pressures (a) and force (b) for 0.5 A step
coil current and 2 m tube length.
23
the valve dynamics. The additional 0.005 sec time delay for the 2 m tube case matches the
value computed theoretically. The attenuation of the mass ow amplitude induced by the
tube is reected by the signicantly larger (more than double) amount of time required to
reach the maximum value of the output force, when the long tubes are used. In the second
20
x [mm]
10
-10
-20
0.0
0.1
0.2
0.3
0.4
0.5
t [sec]
Figure 17: Experimental and numerical piston position for 1.5 V and 5 Hz sinusoidal valve
current for Numatics 0750D02-03A cylinder.
experiment a sinusoidal command current was applied to the valve coil and the piston motion
was recorded. Figures (17) and (18) show the experimental and numerically simulated results
for the Numatics 0750D02-03A and Airpel E16 D 3.0 N pneumatic cylinders. In both cases
the agreement between the outcome of the experiment and the theoretical model is excellent.
Conclusions
In this article we developed a detailed mathematical model for a dual action pneumatic
actuator controlled by a proportional, coil actuated valve. The proposed model is not only
accurate, but also suciently simple such that it can be used online in control applications. Specically, one can use the model to develop high performance force controller for
applications in robotics and automation. We developed a pneumatic cylinder model that
includes the piston seals friction, the dierence in eective areas on the opposite sides of
the piston, the inactive volumes at the ends of the stroke and the connecting ports, and the
dierence in the heat transfer characteristics for the charging and discharging processes of
24
40
30
x [mm]
20
10
-10
-20
0.0
0.1
0.2
0.3
0.4
0.5
t [sec]
Figure 18: Experimental and numerical piston position for 1.5 V and 5 Hz sinusoidal valve
current for Airpel E16 D 3.0 N cylinder.
the cylinder chambers. We introduced a new equation to account for the inuence of the
tubes that connect the pneumatic cylinder with the valve. Explicit formulas were derived
for the tube resistance for laminar and turbulent ow. Valve dynamics, the nonlinearity of
the valve eective area with respect to the coil current, and the nonlinear turbulent ow
through the valve orice were also considered. The results obtained by numerical simulation
were compared with the experimental data, and excellent agreement was found.
References
[1] Al-Ibrahim, A.M., and Otis, D.R., 1992, Transient Air Temperature and Pressure Measurements During the Charging and Discharging Processes of an Actuating Pneumatic
Cylinder, Proceedings of the 45th National Conference on Fluid Power, 1992.
[2] Andersen, B., 1967, The Analysis and Design of Pneumatic Systems, New York, John
Willey & Sons, Inc.
[3] Arun, P.K., Mishra, J.K., and Radke, M.G., 1994, Reduced Order Sliding Mode Control
for Pneumatic Actuator, IEEE Transactions on control Systems Technology, Vol. 2, No.
3, pp. 271-276.
25
[4] Ben-Dov, D., and Salcudean, S.E., 1995, A Force-Controlled Pneumatic Actuator, IEEE
Transactions on Robotics and Automation, Vol.11, No. 6, pp. 906-911.
[5] Bobrow, J.E., and Jabbari, F., 1991, Adaptive Pneumatic Force Actuation and Position
Control, Journal of Dynamic Systems, Measurement, and Control, Vol. 113, pp. 267-272.
[6] Burrows, C.R., and Webb, C.R., 1966, Use of the root Loci in Design of Pneumatic
Servo-Motors, Control, Aug., pp. 423-427.
[7] Chester, C.R., 1971, Techniques in Partial Dierential Equations, McGrew-Hill, Inc.,
New York.
[8] Elmadbouly, E.E., and Abdulsadek, N.M., 1994, Modeling, Simulation and Sensitivity
Analysis of a Straight Pneumatic Pipeline, Energy Conservation and Management, Vol.
35, No. 1, pp. 61-77.
[9] Hougen J.O., Martin, O. R. and Walsh, R. A., 1963, Dynamics of Pneumatic Transmission Lines, Energy Conservation and Management, Vol. 35, No. 1, pp. 61-77.
[10] Hullender,D.A., and Woods, R.L., 1985, Modeling of Fluid Control Components, Proceedings of the First Conference on Fluid Control and Measurement, FLUCOME 85,
Tokio, London: Pergamon Press.
[11] Liu, S., and Bobrow, J.E., 1988, An Analysis of a Pneumatic Servo System and Its Application to a Computer-Controlled Robot, Journal of Dynamic Systems, Measurement,
and Control, Vol. 110, pp. 228-235.
[12] McDonell, B.W., Bobrow, J.E., 1993, Adaptive Traking Control of an Air Powered
Robot Actuator, Journal of Dynamic Systems, Measurement, and Control, Vol. 115, pp.
427-433.
[13] Munson, B.R., Young, D.F., and Okiishi, T.H., 1990, Fundamentals of Fluid Mechanics,
John Willey & Sons, New York.
[14] Richard E., Scavarda, S., 1996, Comparison Between Linear and Nonlinear Control
of an Electropneumatic Servodrive, Journal of Dynamic Systems, Measurement, and
Control, Vol. 118, pp. 245-118.
[15] Schuder C. B., Binder, R. C., 1959, The Response of Pneumatic Transmission Lines to
Step Inputs, Journal of Basic Engineering, Vol. 81, pp. 578-584.
[16] Shearer, J.E., 1956, Study of Pneumatic Process in the Continuous Control of Motion
with Compressed Air-I, II, Transactions of ASME, Feb., pp. 233-249.
[17] Tang, J., and Walker, G., 1995, Variable Structure Control of a Pneumatic Actuator,
Transactions of the ASME, Vol. 117, pp. 88-92.
26
[18] Whitmore, S.A., Lindsey, W.T., Curry, R.E., and Gilyard, G.B., 1990, Experimental
Characterization of the Eects of Pneumatic Tubing on Unsteady Pressure Measurements, pp. 1-26, NASA Technical Memorandum 4171.
27