You are on page 1of 14

Dynamic Modelling and Linearized Design of Electrohydraulic Systems

Important components in an electrohydraulic power system are actuators, valves and pumps. The
working and dynamic modelling for linear actuation systems realized by cylinders are explained next.
Similar analysis could be carried out for rotary actuation systems with hydraulic motor.
2.1 Linear Hydraulic Actuator or Cylinder
Figure 2.1 shows a double-acting single-rod cylinder of cap-end area Aa1 and the initial and
instantaneous volumes of V01 and (V01  Aa1 y) respectively at the inner dead centre and at the
displacement y of the piston away from the dead centre. The piston velocity y  v is supported by the
incoming discharge Q1 and pressure P1 at this chamber together with discharge Q 2 leaving the rod-end
side chamber at pressure P2 and flow area Aa 2 . Due to the piston motion, the rod-end side chamber
volume contracts to (V02  Aa 2 y) from the initial volume of V02 . The pressure differential (P1P2) is
usually referred as the load pressure. In case of an aircraft servo-actuation system, the load pressure
supports the pressure difference on the two sides of each wing. The increase or decrease of pressure in any
cylinder chamber can be attributed to oil bulk modulus  coupled with partly the chamber flow and partly
the piston velocity. Of course, the incoming flow in the cap-end chamber can be visualized as inducing a
volume compression rate by Q1 in absence of piston motion, whereas the piston motion alone induces an
expansion rate Aa1 y leading to net volumetric expansion of ( Aa1 y  Q1 )dt and the corresponding
volumetric strain of ( Aa1 y Q1 )dt /(V01  Aa1 y) . Following similar arguments, the rod-end volume can be
inferred to be associated with the net volumetric strain of (Q2  Aa 2 y )dt /(V02  Aa2 y) . Under the cases of
either piston extension or retraction, the increase in the respective chamber pressure and the volumetric
strain can be related through bulk modulus as
  dP1 /{( Aa1 y  Q1 )dt /(V01  Aa1 y)} dP2 /{(Q2  Aa2 y )dt /(V02  Aa2 y)} . (2.1a)
The above equation provides the expressions for the chamber pressure dynamics as

P1   (Q1  Aa1 y ) /(V01  Aa1 y) , (2.1b)


and P   ( A y  Q ) /(V  A
2 a2 2 02 a 2 y) . (2.1c)
The consequent equation of piston motion against external load FL and friction force F f can be written as

y {(P1Aa1  P2 Aa 2 )  F f  FL}/ ma . (2.1d)

A well known model shown in Figure 2.2 for variation of friction F f with piston velocity is due to
Tustin. This model conceives the walls of the piston and the cylinder as beds of bristles in contact. Prior to
the onset of motion, the bristles offer static friction, or stiction, with the maximum magnitude of F0 . Their
resistance eventually wanes off due to their deformation that sets in a lubricating oil film between the
deforming bristle beds. During the partial lubrication regime, the thickness of the oil film keeps on
increasing with increase in the velocity of the piston with respect to the cylinder. Consequently, the
lubricating effect gets strengthened causing the overall friction to reduce. Beyond a certain velocity called
Stribeck velocity v s , further growth of the film becomes weak and the fully lubricated viscous friction
regime sets in. The later regime is characterized by increase of friction in proportion to the increase in the
velocity involving a coefficient  v and the extension of this linear characteristic to the force axis signifies
what is referred as Coulomb friction Fc . Assuming symmetric behaviour for both the positive and negative
piston velocity, the overall friction model is expressed as

F f {Fc  ( F0  Fc ) exp( v 2 / v s2 )}sgn(v)  v v . (2.1e)


V01 V02
y  v F
P1 L
P2
Q1 y Ff Q2

Figure 2.1: Modeling and Representation of a Hydraulic Cylinder


Friction Force, Ff

F0 Fully lubricated regime


Partially lubricated Fc
friction regime
Piston extension, v0
Piston velocity, v

Piston retraction, v<0

-Fc
-F0

Figure 2.2: Typical Friction Characteristics of a Hydraulic Cylinder


In many real situations, the nonlinear zone of the variation of friction with velocity is dissimilar for
increasing and decreasing velocities. Such a variation gives rise to a hysteresis effect. On many occasions,
the positive and negative velocities lead to different stiction and Coulomb friction. Often, the friction also
involves a pre-sliding slip zone of constant friction with increase in velocity up to a small threshold value.

2.2 Modeling of Spool Valve and Servovalve


The spool valve depicted in Figure 2.3 is comprised of a two-land spool inside a bush. Flow
passages are formed in the body of the bush and these passages extend up to the spool at the port cuts,
marked as P, T and A respectively for the pump, tank and actuator or cylinder ports. With the driving
solenoid unexcited, ideally the left and right lands of the spool should cover both of the metered ports at P
and T. Such valve ports are termed as critically lapped. In case the land width is smaller or greater, the
valve port is termed as under-lapped or over-lapped respectively. Any servovalve can be considered as
close centred with the lapping conditions emerging from manufacturing tolerance. The overlapping in a
proportional valve is much wider that necessarily makes the valve close centred. At the neutral, all the
metered ports are axially covered for both critically lapped and overlapped lapped valves. An underlapped
valve has an opening of, say u, in the axial direction at the neutral position. A small leakage through the
valve occurs from the pump to the tank due to a small spool-bush radial clearance c r . This is usually within
2 to 3% of the maximum flow through the valve. In a simple analysis, the leakage flow can be neglected.
The direction of the spool movement depends on the external force Fe . When the spool moves to
the right by x, the axial opening in the main flow path from Ports P to A increases by that much for both
underlapped and critically lapped valves. When the spool moves to the left, the main flow path through the
valve is set from Ports A to T. If the valve overlap is o, which could be taken as negative underlap u, the
axial opening is created beyond the displacement o. With respect to the width of a cut of a metered port, the
port opening (x+u) is usually small, so that the flow can be considered to be through small orifice. At an
open short orifice corresponding to nonzero command voltage to the valve solenoid, the flow takes place in
the form of an oblique jet, say with velocity V j at an angle  with respect to the spool axis. With the spool
displacement x taken to be proportional to the command signal V with proportionality constant, say kv, the
orifice area in the main flow path of a spool valve can be obtained as

A p  w p (u  k vV ) 2  c r2  w p k vV , (2.2a)

where with reference to Figure 2.3(d), the circumferential width wp of the metered port can be expressed as
wp ncds sin{c /(2nc )}(ds /2)c , (2.2b)

for spool diameter ds and the total angular opening  c at each port made in nc cuts.
P A

x
Fe
T P
T A (b) Symbolic representation of a
(a) Spool valve with rectangular ports servo-class spool valve
P X (d s / 2)sin{ c /(2nc )}
Flow
P

cr
Bush x+u P

Vj
Land Stem Land  c /(2nc )
x

PT PA
T A X (d) Angular disposition of
(c) Flow through a port of a spool valve ports in a spool valve at XX
Figure 2.3: Modelling and representation of a spool valve
The pressure-velocity and pressure-discharge relations corresponding to the short orifice are
respectively written as

V j  Cvp 2( P1  P2 ) /  , (2.2c)

and Q  Cdp ( x  u)w p 2( P1  P2 ) /   CvV P1  P2 , (2.2d)


where ( P1  P2 ) is the pressures drop across the orifice, through which oil of density  flows, whereas Cdp ,
C vp and C v are respectively the discharge coefficient and velocity coefficient of the orifice and the flow-
gain coefficient of the valve. With reference to Figure 2.3, P1 and P2 are respectively
(a) PP and PA for positive x or V and (b) PA and PT for negative x or V .
The velocity and discharge coefficients of an orifice depend on the flow geometry. For rectangular port cuts
in a spool valve, it is customary to take Cvp= 0.98 and Cdp= 0.60 for the main flow.
Figure 2.4 shows a three-land spool valve with unmetered ports A and B. Such a valve is usually
connected to two chambers of a double-acting-double-rod cylinder with equal flow area Aa and discharges
Q in both the ports. In the metered ports P and T, the pressure drops always remain equal yielding

PP  PA  PB  PT . (2.3a)
The no-load condition for such a cylinder and the corresponding port pressures are given respectively by
FL  0 Pa1  Pa 2 and PA  PB  ( PP  PT ) / 2 , where PA  Pa1 and PB  Pa 2 . (2.3b)
Under the no-load condition (2.3b), it can be obtained from (2.2d) and (2.3a) that
Q  Cdp ( x  u)w p ( PP  PT ) /   CvV ( PP  PT ) . (2.3c)

The valve opens to allow flow through it, when the external force Fe causes spool displacement.
For small opening in the vicinity of a square edge of a spool land, the flow takes place in the form of a jet.
The jet angle  depends on the ratio of the length of the opening to the length of spool-bush radial clearance
as well as the neighbouring geometry of both spool land and porting in the bush. A non-radial jet exerts a
force on the spool along the spool axis. This is by virtue of the axial component of momentum of the jet.
This is termed as the steady state flow force.
PP
PT
T P T P2
P1 FL
Fe(+V) x Fe(V)
Aa1=Aa Aa2=Aa
PB=P2 B A PA=P1 y, y
Q Q

Figure 2.4: A three-land spool valve

At unmetered port like A and B in Figure 2.4 for the actuator lines, the flow remains radial. Hence,
no force is contributed by these flows at the steady state. For an underlapped symmetric spool valve shown
in Figure 2.4, the forces at inlet and exit of the small opening cancel each other at the null position. At non-
null positions, the leakages reduce and the effective force from the actuation discharges becomes
significant. Following conservation of linear momentum and Equations 2.2a to 2.2d and 2.3c, the
expressions for steady-state flow-forces on the spool in the direction opposite to the spool displacement
corresponding to Figures 2.3 and 2.4 are given respectively as

FS  Q{0  (V j cos  )}  2C dp Cvp w p [(u  x) 2  c r2 ]1/ 2 ( PP  PA ) cos  , (2.4a)


and FS  Q{V j cos   (V j cos  )}  2C dp Cvp w p [(u  x) 2  c r2 ]1/ 2 {(PB  PT )  ( Pp  PA )}cos 
 2C dp Cvp w p [(u  x) 2  c r2 ]1/ 2 ( PP  PT ) cos . (2.4b)

The variation of flow force with spool displacement is highly nonlinear for linear port cuts. This is
even stronger for small spool displacement around which the jet angle undergoes a rapid variation. A port is
referred as linear, if the change of area at the port cut varies linearly with axial displacement of the spool. In
order to counteract the flow force non-linearity and improve the near-null valve characteristics, either the
spool-port cut, near to the null, is often made nonlinear or the valve is made to operate in closed loop. For
small radial clearance and spool displacement that is large with respect to the underlap, the stiffness
corresponding to the steady-state flow force in Equation (2.4b) can be obtained as
FS / x  K sf  2C dp Cvp w p ( PP  PT ) cos  . (2.4c)

The opening of the spool is decided by the magnitude of control force on the spool. Thus, a
variable control force would result in valve opening over a range of values. Hence it is called an infinite-
position valve. For a given actuator load, the flow rate through the valve depends on the openings at the
metered ports controlled by the spool actuation force. Hence, this spool valve acts as a flow control valve.
In case the piston in the cylinder shown in Figure 2.4 is subjected to an external load FL , then the
modification to Equation (2.4b) can be easily obtained as

Fs  2CdpCvpw p [(u  x) 2  cr2 ]1/ 2 ( PP  PT  FL / Aa ) cos  . (2.4d)

A spool valve connected with a double-acting single-rod cylinder is shown in Figure 2.5. This
valve has four lands contrasting the three-land construction shown in Figure 2.4 that is meant for a double-
acting double-rod cylinder. While the bush enveloping the sleeve contains the flow communicating ports
marked as P, T, A and B respectively, the spool inside the sleeve has four lands 1 to 4. Each of Ports P, T, A
and B is shown in Figure 2.5(b) to have a circular hole at the valve base. These holes are meant for
connecting the respective external flow line from the pump, to the tank and with the cap and rod-end
chambers of the hydraulic cylinder. In Figure 2.4, the supply and return ports are metered, whereas the
cylinder ports are unmetered. In Figure 2.5(a), the supply and return ports are unmetered and the cylinder
ports are metered. From each circular hole of the port, there is a narrow hole providing flow connection
with the chambers separated by the lands within the sleeve of the valve. There are a number of window-like
cuts in the sleeve inside the bush. Each set of windows is connected by a common recess in the bush wall
and each recess is connected with the circular side of the port by a narrow hole in the bush wall. There are
two recesses for Port T. Each of the other ports has only one recess.
Vf V +V
Feedforward VCC
+ =
Controller + V
yd e Feedback Vb
+ =
 Controller
yLVDT

Ff
FL=ksy
y
V

y, y
P1,-Q1 P1,Q1
P2,-Q2 P2, Q2
B A

s1 4 xv 2 1 -xv 3 s2

+V T P T -V
PT,Q1 PT,Q2
Pp,Q1
Pp,Q2

:flow with s1 actuated :flow with s2 actuated : electrical command


(a)
Bush

T B Sleeve
T A T
P

(b)
Figure 2.5: (a) Position control by a double-acting single-rod cylinder with feedback-feedforward
controller and proportional valve with (v) details of bush-sleeve pair of the valve

Since the flow at the cap-end of the cylinder is larger, the total circumferential length w pA
provided by all the cuts of Port A is larger than the total circumferential length w pB provided by all the
cuts of the other metered port. The ratio of the lengths in Ports A and B should be set nearly in proportion to
the flow ratio at the cap and rod-end of the cylinders under incompressible flow assumption. A look at the
window-like cuts in the sleeve shown in Figure 2.5(b) explains the consequential dissimilar cut lengths.
Irrespective of extension or retraction of the actuating piston, a valve design satisfies (2.3a) for the port cuts
matching with the flow at the metered ports. For the system shown in Figure 2.5(a) corresponding to piston
displacement y and a constant piston velocity y , the force balance condition is expressed as
FL  F f  P1 Aa1  P2 Aa 2 . (2.5a)

Now, (2.3) and (2.5a) can be combined to obtain


P1 {(PP  PT ) Aa 2  FL  F f }/( Aa1  Aa 2 ) , (2.5b)
P2 {(PP  PT ) Aa1  FL  F f }/( Aa1  Aa 2 ) , (2.5c)
and PP  P1  P2  PT  ( PP Aa1  PT Aa2  FL  F f ) /( Aa1  Aa 2 ) . (2.5d)
It can be inferred from the figure that for a zero-lap or critically lapped valve, the direction of
discharge in and out of the cylinder is reversed as the sign of the voltage in the active solenoid of the valve
is reversed. For negligible effects of radial clearance on the valve flow and cylinder friction on the piston
motion, then following the same procedure that led to Equation (2.4d), the steady state flow force in case of
the arrangement depicted in Figure 2.5(a) can be obtained as
Fs  2CdpCvpx(wpA  wpB )( PP Aa1  PT Aa 2  ks y)cos /( Aa1  Aa 2 ) , (2.5e)
where k s is the stiffness of the spring. Such a loading arises, say on an actuation mechanism setting the
aeroplane wing at small angle of attack.
Flow rate through the spool valve, and hence the velocity of flow within the spool valve of annular
area Av change dynamically with spool movement. With this change in flow velocity is associated an
acceleration of the entrapped mass within the spool valve. For the spool valve depicted in Figure 2.3, the
acceleration to the mass over the distance xpa between Ports P and A in the spool chamber is in the opposite
direction to the spool motion. In case the spool valve is connected to a single-acting cylinder of flow area
Ac negotiating an external load FL and the radial clearance between the spool and its bush is negligibly
small, the force on the spool acting in the direction opposite to the spool motion can be expressed as
d (Q/ Av )
Ft  Av x pa Cdp x pa w p 2  ( PP PA ) x . (2.6)
dt
Therefore, the damping coefficient corresponding to the transient flow force is given by
Ctf Cdp x pa wp 2 ( PP PA ) . (2.7a)

It is easy to show these coefficients to bear opposite signs for the two pairs of metered and
unmetered ports for a spool valve meant for a double-acting cylinder. For zero tank pressure together with
(2.5d), it can be shown that
Ctf  C dp (w pA x pa  w pB xbt )  ( PP Aa1  FL  F f ) /( Aa1  Aa 2 ) . (2.7b)

Thus, the coefficient for this spool valve turns out to be zero, if Ports A and B for the configuration shown
in Figure 2.4 are placed midway between the metered ports on the either side. Zero coefficient of damping
could be achieved for a double-acting single-rod cylinder under the condition
wpA x pa wpB xbt wpA xat wpB x pb 0 . (2.7c)

Since the actuation force governs this displacement, any value to it is possible within a range
subjected to the range of available force variation. Thus, the valve is called an infinite-position valve.
Corresponding to the displacement and actuator chamber pressures, flow through the spool valve sets in.
Solenoid or force motors are used for driving the spool of such a valve.
2.3 Dynamic Modeling of Solenoid
Schematic arrangements of a bi-solenoid motor and a permanent-magnet motor are shown in
Figures 2.6 (a) and (b) respectively. Each type of motor has an armature A and two springs Sp1 and Sp2 at
both the ends of the armature core. On the other side of the springs lie a solenoid pair S1 and S2 in Figure
2.6(a) and a magnet pair M1 and M2 in Figure 2.6(b). A central solenoid S is placed around the armature in
Figure 2.6(b). From both ends of the armature core, connecting spokes protrude out that are meant for
connecting the element to be driven by the motor. Of course, physical arrangements of the components in
the actual devices are more complex. In case the force requirement to drive the spool is lower, permanent
magnets are not used. Magnets introduce substantial nonlinear effects through magnetic hysteresis and
nonlinear variation of magneto-motive force or MMF M m with magnetic flux m .
Considering the symbols for magnetic flux density as B , reluctance as R , MMF as M and field
intensity as H with subscripts m, a and c for magnet, air and solenoid coil respectively, i for current
through coil of turns N , g a and Aa for axial length and area of air gap respectively, lm for length of
magnet and  a for air permeability, the relationship among the terms can be summarized as

Ba  φa / Aa , Ra  g a /(a Aa ) , H a  Ba / μa  φa Ra / ga , M m  H mlm , M a  g a H a  a Ra and M c Ni . (2.8)


φc
S1 S2
Sp1 Sp2 Sp1 S Sp2
M1 M2
x x A N S
A S N

φl φr

I+i (a) I-i


(b)
Figure 2.6: Schematic arrangement of a moving armature A in a (a) solenoid pair and (b) force motor

By integrating M a corresponding to increase of flux from 0 to a as the coil current is increased


from 0 to i by holding the air gap constant, the stored energy in the air of gap area can be obtained as
a a
E 0 M a d a  0  a Ra d a   a2 g a /(2 a Aa ) . (2.9)

By using (2.8b), the force of attraction between the members across the gap can be obtained as

F dE/dga a2 /(2a Aa )Ba2 Aa /(2a )M a2a Aa /(2 ga2 ) . (2.10)

Since the net force F in case of Figure 2.6(a) is due to the difference of forces Fl and Fr
produced respectively by S1 and S2, it can be determined by using (2.8) and (2.10) for spool displacement
x g 0 and neutral current I  i in both S1 and S2 along with differential currents i and i through S1
and S2 respectively as

F Fl Fr N 2a Aa [{( I i)2 /{2( g0 x)2}{(I i)2 /{2( g0 x)2}]
{2 N 2 Iμa Aa /( g02 )}[i  Ix / g0 ]
 k f i  km x; k f  2 N 2 Iμa Aa /( g02 ) & km  ks I / g0 , (2.11)

where k f and k m are referred respectively as force constant and magnetic elastance. The latter coefficient
behaves in a manner exactly opposite to a centering spring force. In other words, the de-centering force due
to the magnetic action of the solenoids increases with displacement away from the neutral. Clearly, this is a
de-stabilizing force, which should be compensated by using a spring between each solenoid and the spool.
Of course, the spring stiffness should dominate over the magnetic elastance for stability.
For the armature of mass ma in a bi-solenoid motor supported by springs of effective stiffness ks and an
external force Fe, the dynamics for a control current i can be expressed in view of (2.11) as
ma x k s x  k f i  k m x  Fe . (2.12a)

Following the subsidence of the dynamics, the steady displacement would be obtained as
x st  (k f i  Fe ) /(k s  k m ) . (2.12b)

The nonlinear variation of the motor force for using only one solenoid at a time would result in possible
multiple steady solutions of the displacement emerging from the forcing model obtained from (2.10) and
the last equation of (2.8) yielding the dynamics of displacement x closing the gap g 0 between the active
solenoid and the spool separated by a spring in Figure 2.6(a). Writing the dynamics as

ma x k s x  k sol i 2 /( g 0  x) 2  Fe , (2.12c)


with k sol ( N  a Aa /2) ,
2
(2.12d)
the equation for the steady state displacement after the subsidence of the transient corresponding to a
constant current ic emerges as

k s x st  k sol ic2 /( g 0  x st ) 2  Fe . (2.12e)


Under the circumstances, a small perturbation at any instant would have caused the system to jump
from one steady state to another in absence of any compensation provided by a careful controller design. In
case the external force arises from coupling of a spool valve with a bi-solenoid motor, the nonlinear
variation of jet angle with valve opening would have given steady displacement equation with possible
multiple solutions as
xst Fe ( xst )k f i/(ks km ) . (2.12d)

Thus, the nonlinear variation of flow forces nullifies the linear behaviour of the driving force from a bi-
solenoid arrangement. Hence, a spool valve with two solenoids is usually actuated by only one solenoid at a
time so that the coil current in the solenoid does not need any large base current I. This minimizes the
resistive heating of the coils.
Corresponding to an excitation voltage V across the solenoid of resistance R and inductance L ,
the dynamics of the coil current is given by
Ldi / dt  Ri V , (2.13a)
This dynamics is indeed much faster than the dynamics of the spool motion (2.12a), which in turn is much
faster than the motion dynamics of the cylinder piston. In the time scale of the piston motion, the transient
terms in (2.12a) and (2.13a) are negligible. In addition, the spool displacement during most part of the
command excitation is larger than both the radial clearance and the underlap in the valve. Under these
assumptions, the discharge through the valve shown in Figure 2.3 is given by (2.2d)

where Cv  2 /  K f /[( K sf  k m )C dp w p }] / R . (2.13b)

A proportional valve is characterized by a deadband, whereas a servovalve has negligible


deadband. Only beyond this deadband in actuation voltage ( V0 , V0 ) or current ( i0 , i0 ), there is
perceptible discharge through the valve. In a proportional valve, the deadband is about 10 to 15% of the
maximum possible actuation command. Even with only one solenoid excited at a time in any of these two
classes of electrohydraulic valves, spool displacement proportional to the excitation voltage is usually
achieved by an embedded controller that employs LVDT feedback of the spool displacement. Hence, (2.2d)
remains applicable for servovalves as well.
A dynamic system represented by an ordinary differential equation in y up to second order is
always stable, in case its coefficients are positive constants. In such cases, the solution y ~ e at for a first-
order equation and y ~ e a1t , e a2t for a second-order equation involves negative real part in the arguments of
the exponential terms. In case an argument is complex, it is always paired with its conjugate values. The
solutions corresponding to the imaginary pair could also be expressed alternatively by sine and cosine
functions that are bounded and periodic in time, whereas the common real part involves an exponential
term. In all these cases, the decay of the exponential part with time ensures stability. Hence the transient
system dynamics for a constant current excitation is stable in (2.12a), as long as (ks km ) is positive.
Therefore, the coil current should be controlled to fulfil the condition x / g0  1 . It may also be noted that a
negative transient flow force coefficient as obtained in (2.7a) gives rise to an inherently unstable system.
The composition of combined feedforward-feedback controller is now presented for a nonlinear system
along with discussion of the stability aspect.

2.4 Feedforward Control Modeling for a Servosystem

The input to the system shown in Figure 2.5 is the displacement demand yd . An LVDT provides
the measurement y LVDT for the variation of the corresponding response y . A possible control approach is
to employ a feedforward controller to compensate nonlinearities of the system. This requires the demand
yd to be fed to a feedforward controller for extracting the velocity demand v d (t ) and the acceleration
demand a d (t ) , both emerging as known function of time, from the differentiated expressions

vd  y d . (2.14a)
and ad  yd . (2.14b)
In turn, these functions are used to estimate the feedforward voltage V f based on a system model that
should have a simple form for the sake of online implementation. A relatively smaller feedback voltage Vb
should be calculated in tandem, providing the total command as
V V f Vb , (2.15a)
where a PID feedback controller could be used to calculate


Vb  k p e  k i edt  k d e , (2.15b)
involving the proportional, integral and derivative gains designated respectively as k p , k i and k d and a
small control error
e  yd  y  yd  yLVDT . (2.15c)
Neglecting the effect of oil compressibility and taking the valve dynamics to be much faster than
the dynamics of piston motion, the nonlinear relations between piston velocity and the pressures across the
metered ports of the valve can be expressed with reference to Figure 2.5 and Equation (2.13a) as

y  (CvAV / Aa1 )( PP  P1 )1/ 2  (CvBV / Aa 2 )( P2  PT )1/ 2 for vd  0 , (2.16a)


and y  (CvAV / Aa1 )( P1  PT ) 1/ 2
 (CvBV / Aa 2 )( PP  P2 ) 1/ 2
for vd  0 , (2.16b)

where C vA and CvB are the valve gains respectively for Ports A and B of the valve.

Using Equations (2.14a), (2.15a) and (2.15c) along with the smallness of (e / vd ) and ( Vb / V f ) relative to
1, Equation (2.16a) can be approximated as

P1  PP {vd Aa1 /(CvAV f )}2 (1 2e / vd )(1 2Vb /V f ) for vd  0 , (2.16c)
and P2  PI {vd Aa 2 /(CvBV f )}2 (1 2e/ vd )(1 2Vb /V f ) for vd  0 , (2.16d)

whereas Equation (2.16b) yields

P1  PT {v d Aa1 /(CvAV f )} 2 (1 2e / v d )(1 2Vb /V f ) for vd  0 , (2.16e)

and P2  PP {v d Aa 2 /(CvBV f )} 2 (1 2e / v d )(1 2Vb / V f ) for vd  0 . (2.16f)

Instead of the more complex Tustin friction model expressed by Equation (2.1e), a popular
approximate form given by
~
Ff  F0 sgn(v)  vv  Ff , (2.17)
~
is often employed in controller formulation, where F f is the error due to modelling approximation.
Employing Equations (2.14a), (2.14b), (2.15c) and (2.17) together with a consideration of external loading
shown in Figure 2.7 by a spring of stiffness k s , Equation (2.1d) can be rewritten as
~
ad e{(P1 Aa1P2 Aa 2 )F0 sgn(vd ) v (vd e)F f k s ( yd e)}/ma , (2.18)

Consistent with an objective of formulating a feedforward-feedback controller, the above equation can be
split in two corresponding parts. The feedforward parts are obtained as

a d [( PP Aa1  PT Aa 2 )  (v d / V f ) 2 ( Aa31 / CvA


2
 Aa32 / CvB
2
)  F0  v v d  k s y d ] / ma for vd  0 , (2.19a)

and a d [( PT Aa1  PP Aa 2 )  (v d /V f ) 2


( Aa31 / CvA
2
 Aa32 / CvB
2
)  F0  v v d  k s y d ] / ma for vd  0 , (2.19b)

On the other hand, the feedback parts are given by

e ( v / ma )e  (k s / ma )e  (2 / ma ){ Aa31 | v d | /(CvAV f ) 2  Aa32 | v d | /(CvBV f ) 2 }e 


~
{2v d2 /(maV f3 )}{( Aa31 / CvA 2
)  ( Aa32 / CvB2
)}Vb  F f / ma , for vd  0 , (2.20a)
and e ( v / ma )e  (k s / ma )e  (2 / ma ){ Aa31 | v d | /(CvAV f ) 2  Aa32 | v d | /(CvBV f ) 2 }e 
~
{2vd2 /(maV f3 )}{( Aa31 / CvA 2
)  ( Aa32 / CvB
2
)}Vb  F f / ma , for vd  0 , (2.20b)

Equation (2.19a) can be rewritten by employing (2.16c) and (2.16d) as

V f  v d [( Aa31 / CvA
2
 Aa32 / CvB
2
) /{PP Aa1  PT Aa 2  (ma a d  v v d  k s y d  F0 ]1/ 2 for vd  0 , (2.21a)

and Equation (2.19a) together with (2.16e) and (2.16f) yield

V f  v d [( Aa31 / CvA
2
 Aa32 / CvB
2
) /{PP Aa 2  PT Aa1  (ma a d  v v d  k s y d  F0 ]1/ 2 for vd  0 . (2.21b)

It is a straight forward appreciation that if a positive coil voltage in the corresponding solenoid is expected
to induce spool displacement corresponding to piston extension, a negative coil voltage would induce the
reversed piston motion. Accordingly, the last two equations have been cast to have the sign of the
feedforward voltage in the excited solenoid identical with the sign of the velocity demand.
By considering the sign of feedforward voltage in (2.21a) and (2.21b), now (2.20a) and (2.20b) can
be recast in terms of positive coefficients a1 , a0 and b as

e a1e  a0 e  bVb  , (2.22a)

where a1 { v  2 Aa31 | v d | /(CvAV f ) 2  2 Aa32 | v d | /(CvBV f ) 2 }/ ma , (2.22b)

a0 ks /ma , (2.22c)

b {2v d2 /(ma |V f3 |)}{(Aa31 / CvA


2
)  ( Aa32 / CvB
2
)} , (2.22d)
~
and  F f /ma . (2.22e)

In case of a nonlinear system, the feedforward controller, despite involving approximations, ensures system
operation in the vicinity of the demand. The role of the additional feedback controller is to take care of the
modelling approximations and the parameter uncertainties so as to provide error with respect to the demand
as low as possible. The small feedback part guarantees the total command to lie in the close vicinity of the
feedforward part. This eliminates the possibility of hunting by the total command among various possible
steady states of the nonlinear system.

2.5 Role of Feedback Gains on Controller Performance


Corresponding to the feedback controller form (2.15b), the model (2.22a) can be rewritten as


e(a1bkd )e(a0 bk p )ebki edt  , (2.23)

The above differential equation models the error corresponding to a third order system model. Its
coefficients a1 , a0 and b depend on the modelling of the system at hand. For the system shown in Figure
2.5, these are given by (2.22b) to (2.22d) for the approximations discussed in the last section. The forms of
such equations for the system model (2.12c) are presented next.
It is considered that a solenoid drives a mass ma separated by a gap by a spring of stiffness k s . In
absence of any command signal, the gap is taken as g 0 . For a constant external force Fe , the models for
the displacement demand xd and the response x rewritten as ( xd e) are obtained respectively as

ma xd ks xd {ksol i 2f /( g0 xd )2}Fe , (2.24a)

and ma ( xd e)ks ( xd e){ksol i 2f /( g0 xd )2}(12ib /i f ){1e/( g0 xd )}2 Fe . (2.24b)

While (2.24a) yields the expression for feedforward current as

i f {(ma xd ks xd Fe )/ksol }1/2 ( g0 xd ) , (2.25a)


subtraction of (2.24a) from (2.24b) provides the differential equation (2.22a) for the variation of error e ,
where the coefficients are now given by
a10 , (2.25b)
a0  [ks {2ksol i 2f /( g0  xd ) }  bk p ] / ma ,
3
with ib  bk p (2.25c)
b2ksol i f /{ma ( g0 xd )2} , (2.25d)

and  is the neglected higher order terms used to obtain the simplified form (2.24b).
Some important observations emerging from (2.23) are
(a) (a1bk d ) and (a0 bk p )0 (2.26a)
for the system to be stable, which in the context of (2.25b) and (2.25c) imply
(i) the role of the derivative gain providing artificial damping leading to zero error with possible
overshoots and undershoots during the initial transient phase

(ii) the minimum value of the spring stiffness given by ks {2ksol i 2f /( g0 xd ,max)3} guarantees system
stability irrespective of the choice of the proportional gain
(b) having a nonzero integral gain for the system to acquire zero steady error at the subsidence of the
transient say for tts by virtue of having
ts
bki
0
edt  , (2.26b)
for compensating all the modeling approximations
(c) nonzero steady error es corresponding to zero integral gain given by

es  /(a0 bk p ) , (2.26c)


implying lower es for higher k p , as long as  is bounded

(d) a stable oscillation of the armature in absence of derivative and integral gains for the system (2.25a).

2.6 Control Objectives and System Requirements


In case of manual control, a valve is moved and locked at the desired position at specific time
instants. Under the locked conditions, flow forces do not influence the opening at the metered ports. In such
a case, the metered ports can be made wide open, thereby reducing the pressure losses at the metered ports
insignificant. These forces, however, influence the dynamics of passive or active valves. In an automated
system, the manual control of a valve is replaced by any command driven motor from the types described in
Figure 2.6. In applications with lower system cost, a fixed-displacement pump with relief valve bypass is
used to feed such a valve. While a constant flow bypass keeps the supply pressure of the pump constant, the
loss of power in the relief valve makes the system inefficient. In case of more sophisticated systems, where
higher cost of the system is acceptable, a pressure-compensated pump is used. Efficiency of such a system
is of course higher.
The core components of a pump shown in Figure 2.7 pertains to a fixed-displacement in-line axial
piston pump with nine pistons a to i. A shaft (1) is splined with a barrel (2) providing torque to rotate it
together with the pistons housed inside the respective cylindrical part (3) of the bores. One end of each of
these pistons has a hemi-spherical head that is held within a groove of a stationary inclined plate called the
swash plate (4). There is a valve plate (5) on the other end the barrel. This plate has two long kidney-shaped
grooves (6) and (7), acting as the delivery and suction manifolds respectively, since the said arrangement
makes the rotating pistons to reciprocate as well. Corresponding to the swash angle  and the diameter Dp
of the pitch circle, along which the axes of the pumping pistons are equally spaced, the stroking length of
each piston between the IDC and ODC is given by (0.5Dptan). Each barrel bore towards the valve plate
end is given a kidney shape (8) of matching width, but of much smaller angular extent. This ensures smooth
passage of flow between each bore and the manifold, over which it passes. The movement of a piston away
from the valve plate creates a partial vacuum at the suction manifold (7). This vacuum enables suction of oil
from a tank through the suction port (9) and manifold (7) into the piston bore. During the piston movement
towards the valve plate, the oil is forced through the delivery manifold (6) and then through the delivery
port (10) and the external circuit.
A piston remains closed to a dead centre, as it rotates over a bridge separating the manifolds.
During this period, the pressure inside the piston bore undergoes fast change that in turn leads to a torque
transient to be supported by bearings. An odd number of pistons ensures only one piston passing over any
one of the bridge at a time. With respect to an even number of pistons, the amplitudes of the consequent
torque imbalance and discharge oscillation reduce for a near-by odd number of pistons along with a
corresponding increase in their frequency. With increase in number of pistons, the amplitudes reduce
further. The diameter, number and stroke length of the pistons determine the geometric displacement of the
pump. Ideal characteristic of a fixed displacement, or FD, pump is of constant mean discharge irrespective
of the delivery pressure. However, the discharge reduces with increase in delivery pressure due to internal
leakages and oil compressibility. A variable-displacement, or VD, pump requires additional arrangement
for changing the swash angle . This could be through a simple spring-loaded set-screw arrangement, where
the knob of the set screw could be turned externally to change the angle.
A pressure compensator shown in Figure 2.8 could be an alternative self-adjusting arrangement. It
uses the delivery pressure to change the angle by an arrangement of a valve and a pair of single-acting
cylinders. These cylinders and the valve along with a VD pump are symbolically shown in Figure 2.8. The
role of the compensator is to reduce the displacement volume sharply from the high value at the cut-in
pressure to zero at cut-off pressure. A pump operating between the cut-in and cut-off conditions responds
quickly to variable flow demand corresponding to the displacement demand of the actuator piston. Among
the external connections protruding out from the box encompassing the symbolic representation of a pump
with pressure compensator, the usual ones are the shaft, the suction port and the delivery port. Two more
connections are associated with the valve. One is a drain port for returning flow to the tank and the other is
a set-screw knob. The inclined arrow in the VD pump symbol can be envisaged as the swash plate held at
an angle with respect to the normal to the axis of the driving shaft. Besides the rotating pumping pistons, a
pressure compensator involves two non-rotating pistons, each inside a single-acting cylinder. The
communication of the delivery pressure and a spring together causes extension of the piston in one such
cylinder. At the normal setting, the valve allows the flow from the other cylinder, called the stroking
cylinder, to return to the tank through the drain port together with other leakages inside the pump. The third
port of the valve is internally connected with the delivery line, which also joins with a pilot line.
In a hydraulic circuit having a pump without a pressure compensator, the part of the fixed pumped
flow not required in the cylinders returns through the pressure relief valve resulting in loss of power. This
valve also keeps the pump pressure constant in the circuit as long as flow takes place through this. Clearly,
a pressure compensator provides an energy-efficient but costly solution. A relief valve limits the maximum
pressure in the system in case of a valve failure. However in a circuit with fixed displacement pump, the
relief valve also ensures a fixed pressure at the pump delivery. In such a case, no coupled nonlinear effect
arises that are due to the pressure-compensated stroking dynamics of the more advanced pump design. The
availability of high pressure in the inlet manifold and low pressure at the outlet manifold makes a device
working as a hydraulic motor, in case the construction is similar to the pump shown in Figure 2.7.

5 2 3
3
a
6 6
a 7 b, i
ODC 4
b i
1
c h 9 c, h
Dp
10
10 g 
d
d, g
e f
IDC e
8 7 8 f

(a) (b)
Figure 2.7: Schematic of an axial-piston pump – (a) valve plate and barrel, (b) valve plate and
barrel in section with pistons and swash plate
Figure 2.8: Symbolic representation of a pressure-compensated variable-discharge pump with the
compensator cylinders and valve shown symbolically

Questions

1. Obtain the equation of piston motion and the equations of pressure dynamics in the chamber of a single-
acting cylinder and in the chambers across the piston of a double-acting cylinder with (a) double rods and
(b) single rod.
2. Discuss Tustin model of friction on the piston in a hydraulic cylinder varies. What are meant by stiction
and Coulomb friction?
3. Comment on the valve laps for a spool type servovalve.
4. What is the origin of steady-state flow force in a spool valve? Show this force to be centring in nature
with reference to a spool valve with (a) three lands and (b) four lands.
5. What is meant by jet angle at a metering port of a spool valve? Obtain an expression for steady-state flow
force on a four-land spool valve and explain the effect of external load on the cylinder on this force.
6. By obtaining an expression of transient flow force on a spool valve with two lands metering the pump
and tank ports with an in-between cylinder port, comment on its effect on the stability on spool
displacement.
7. How is zero flow force damping ensured on spool valve with (a) three lands metering the pump and tank
ports with two unmetered actuator ports and (b) four lands metering the two actuator ports with the pump
and tank ports unmetered?
8. The cylindrical core of an armature is supported by two mechanical springs, each placed on a flat end of
the core on one side and a solenoid core on the other side. Obtain the expressions for force constant and
magnetic elastance for concurrent bi-solenoid excitation of an armature undergoing linear displacement.
Comment on the role of the mechanical springs in stabilizing the system.
9. With the help of a schematic, describe the operation of a bi-directional linear force motor.
10. Why is only one solenoid excited at a time, despite having two solenoids in a proportional valve with
three-land spool?
11. Consider an armature with one of the flat end of its cylindrical core maintained at a gap from the similar
core of a solenoid by a mechanical spring. The other end of the armature is connected to an external mass
meant to slide on a solid horizontal surface corresponding to the excitation current in the solenoid coil.
Obtain the equation of motion of the mass and comment on the stability, the final ‘steady’ state and the gain
setting requirements corresponding to
(a) a constant current input in the solenoid and the opposing friction as constant
(b) a feedforward controller and a proportional feedback
(c) a feedforward controller and a PID feedback
(d) a feedforward controller and a PID feedback with the dynamic friction reducing with increase in
armature velocity.
12. Discuss the roles of the feedforward and the feedback in the control of a nonlinear system, when applied
together.
13. Consider an electrohydraulic circuit with a fixed displacement pump, relief valve, critically lapped
servovalve and a single-rod (or double-rod) double-acting cylinder driving a constant external load. For a
feedforward-PID controller, obtain an estimate of the feedforward current input to the solenoid for the
extension (or retraction) stroke of the piston and the equation for the error dynamics for the system. Discuss
the roles of the feedback gains in the control.
14. What are the relative merits and demerits of a fixed-displacement pump and a pressure-compensated
variable-displacement pump in an electrohydraulic motion control system?
15. Why does an axial-piston pump have odd number of pistons? What is the effect of increasing the
number of pistons for pumping?
16. In terms of a schematic diagram, describe the operation of a pressure-compensated variable-
displacement pump.

You might also like