Journal of Sound and Vibration (1999) 222(2), 171±189

Article No. jsvi.1998.2080, available online at http://www.idealibrary.com on

VIBRATION CONTROL OF A FOUR-BAR

MECHANISM WITH A FLEXIBLE COUPLER
LINK
M. KARKOUB AND A. S. YIGIT
Department of Mechanical and Industrial Engineering, Kuwait University,
P.O. Box 5969, Safat, 13060, Kuwait

(Received 16 July 1998, and in ®nal form 22 September 1998)

Dynamic modelling and controller design for a ¯exible four-bar mechanism

is studied. The fully coupled non-linear equations of motion are obtained
through a constrained Lagrangian approach. Resulting di€erential-algebraic
equations are solved numerically to obtain the system response. A linearized
dynamic model is developed which facilitates the design of various controllers.
The fully coupled nature of the governing equations facilitates control of elastic
motion through the input link alone. A simple PD controller is designed based
on a linearized model. These gains are subsequently tuned using the actual
model to achieve the desired response. The resultant controller is shown to be
ecient in suppressing the vibrations of the ¯exible link as well as controlling
the rigid body motion.
# 1999 Academic Press

1. INTRODUCTION

Four-bar mechanisms are widely used in the industry and can be found in
machines such as photocopiers and card feeders [1]. The quest for high speed
light weight machinery requires a redesign of the current mechanisms.
Unfortunately, reducing the weight of four-bar mechanisms and/or increasing
their speed may lead to the onset of elastic oscillations which causes performance
degradation such as misfeeding in the case of the card feeder mechanism [1].
Moreover, the dynamics of these mechanisms become more involved and make
the control process a challenging task. Suppressing the vibrations of the coupler
mid-point of a four-bar mechanism using torque actuators mounted on the input
link is the subject of this work.
Traditionally, dynamic analysis and control of mechanisms have been based
on the assumption that the links behave as rigid bodies. In mechanical systems
operating at high speeds, some oscillatory elastic motion is inevitable. This
motion becomes a major concern when performance requirements are such that
high precision is important. Because of the trend toward increasing operating
speeds and reducing weights in modern machinery, it may be inaccurate to treat
certain links in such systems as rigid bodies. The effect of ¯exibility on the

0022±460X/99/170171 ‡ 19 $30.00/0 # 1999 Academic Press

172 M. KARKOUB AND A. S. YIGIT

dynamic behavior of mechanical systems has been the subject of numerous

investigations (see e.g., reference [2]). Mathematical modelling of ¯exible
mechanisms is complicated. The rigid body motion involves changing geometries
resulting in varying system parameters, and these in¯uence the elastic
deformations. The elastic deformations themselves in turn in¯uence the rigid
body motion [3]. Though the importance of including the effect of elastic motion
in dynamic modelling has been recognized for quite some time, most models
were based on a prescribed rigid body motion (see e.g., references [4] and [5]).
The assumption here is that the elastic deformations do not have signi®cant
effect on rigid body motion. In other words, there is only a one-way coupling
between elastic deformations and rigid body motion. For lightweight and high-
speed mechanisms where there are no large ¯ywheel type inertia elements to
maintain the rigid body motion independent of the elastic deformations, this
modelling approach results in gross errors [6, 7]. Furthermore, in a control
system formulation, this type of modelling results in an unobservable system
with respect to the rigid body motion. Therefore, in these cases, a fully coupled
formulation is essential. There are two distinct modelling approaches to fully
coupled formulations. The ®rst one is a ®nite element formulation where both
rigid and elastic degrees of freedoms are considered as generalized co-ordinates
[3]. The second approach uses the Lagrange multiplier technique to incorporate
joint constraints into the equations of motion, and results in a mixed set of
differential and algebraic equations (see e.g. reference [8]).
Although dynamic analysis of ¯exible mechanisms has been the subject of
numerous investigations, the control of such systems has not received much
attention. Sung and Chen [9] proposed an optimal control scheme to suppress
the vibrations of a four-bar linkage with a ¯exible output link. They used
piezoceramic sensors and actuators which are located on the ¯exible link. The
effect of elastic deformations as well as control action on the rigid body motion
was not considered since the authors used a model with only one-way coupling.
Beale and Lee [10] investigated the feasibility of applying fuzzy control for a
¯exible slider±crank mechanism. A piezoelectric actuator placed on the ¯exible
link was required to implement this controller. Liao et al. [11] also proposed the
use of piezoelectric ®lms and designed a robust controller based on a linearized
state±space model of the mechanism. The effect of parameter variations and the
instability caused by the control action is studied.
Most of the work available in the literature which deals with vibration control
of ¯exible mechanisms employ an actuator which acts directly on the ¯exible
link. The effect of the control forces and moments on the overall motion is
neglected. In addition, the implementation of such controllers may require
sophisticated and expensive design. An alternative method would be to control
the vibrations through the motion of the input link. Since the elastic
deformations are both observable and controllable through the rigid body
motions with a fully coupled model, such a control strategy should be feasible.
The current study deals with control of a four-bar mechanism with a very
¯exible coupler link. An actuator is assumed to be placed on the input link
which applies a control torque. A simple PD control is designed which requires
 FOUR-BAR MECHANISM 173
measurement of the position and angular velocity of the input link only. This is
different from the control technique proposed by Sung and Chen [9] where the
actuator and the sensor are placed directly on the ¯exible link. Moreover, the
control scheme proposed here controls both the ¯exible as well as the rigid body
motion unlike the work done in reference [9] where the control of the rigid body
motion is assumed to be done independently. It is evident from this study that
the rigid body motion and the elastic deformations are strongly coupled;
therefore, they need to be taken care of simultaneously. In this study, a
constrained modelling approach is used. The elastic motion is treated by an
assumed modes method instead of a ®nite element formulation. The resulting
differential-algebraic equations are numerically solved for both open and closed-
loop simulations. Simulation results demonstrate that the proposed controllers
are effective in suppressing the vibrations as well as in accurate positioning of
the mechanism.

2. DYNAMIC MODEL
The kinetic energy of the four-bar mechanism shown in Figure 1 is given by:
…
1 _ 2 1 _ 4 1 l3 2
T ˆ Ic y2 ‡ Io y4 ‡ r_r dx, …1†
2 2 2 0
where y2 and y4 are the input and output angles, respectively, Ic and Io are the
mass moments of inertia of the input and output links with respect to the joint

z1

B
e2

w
z2 l3
x

A l4

l2

2 4
O C e1

Figure 1. Sketch of the four-bar mechanism.

174 M. KARKOUB AND A. S. YIGIT

axes, respectively, r is mass per unit length of the coupler link, and r is the
position vector of a point on the coupler link given as
r ˆ r2 ‡ r3 ‡ w, …2†
where
r2 ˆ l2 cos…y2 †e1 ‡ l2 sin…y2 †e2 , …3†

r3 ˆ x cos…y3 †e1 ‡ x sin…y3 †e2 , …4†

w ˆ wz2 …5†
where w represents the elastic deformation of the coupler link at an arbitrary
position x. The unit vectors z1 and z2 can be written as
z1 ˆ cos…y3 †e1 ‡ sin…y3 †e2 , …6†

z2 ˆ ÿ sin…y3 †e1 ‡ cos…y3 †e2 : …7†

The strain energy is given by

… l3  2 2
1 @ w
Uˆ EI dx, …8†
2 0 @x2
where EI is the ¯exural rigidity of the coupler link.
The variables y2, y3, and y4 are related by the following geometrical
constraints:
 
ÿl1 ‡ l2 cos…y2 † ‡ l3 cos…y3 † ÿ l4 cos…y4 †
Fˆ ˆ 0: …9†
l2 sin…y2 † ‡ l3 sin…y3 † ÿ l4 sin…y4 †

It should be noted that in general the constraint equations depend on the elastic
deformations. However, because the longitudinal deformations are neglected and
the transverse deformations are assumed small, the distance AB (see Figure 1) is
assumed to be constant. Therefore, the constraint equations here do not depend
on the elastic deformations.
The assumed modes method along with a constrained Lagrangian approach is
used to obtain the discretized equations of motion. Let the de¯ection of the
coupler link be written as
X
N
w…x, t† ˆ ci …x†qi …t†, …10†
iˆ0

where qi(t) are the modal co-ordinates and ci(x) are the mode shapes of a
pinned±pinned beam given as
 
ip
ci …x† ˆ sin x : …11†
l3
 FOUR-BAR MECHANISM 175
The constrained Lagrange's equations are
   T
d @L @L @F
F
ÿ ‡ L ˆ Fi , …12†
dt @ Z_ i @Zi @Zi
where
L ˆ T ÿ U,
and Zi are the generalized co-ordinates which includes rigid body co-ordinates as
well as elastic modal co-ordinates. F is the vector containing the constraint
equations and L is the vector of Lagrange multipliers, l1 and l2; and Fi are the
generalized forces. The force vector (@F F/@Zi)TL is substituted in to replace the
removed joint forces.
The resulting differential equations are given as:
1 1
Ic 
y2 ‡ rl22 l3 
y2 ‡ rl2 l23 
y3 cos…y2 ÿ y3 † ÿ rl2 l23 …y_ 2 ÿ y_ 3 †y_ 3 sin…y2 ÿ y3 †
2 2
X
N X
N
ÿ l2 sin…y2 ÿ y3 †…y_ 2 ÿ y_ 3 † q_ i Qi ‡ l2 cos…y2 ÿ y3 † qi Qi
iˆ1 iˆ1

X
N X
N
‡ l2 
y3 sin…y2 ÿ y3 † qi Qi ‡ l2 y_ 3 cos…y2 ÿ y3 †…y_ 2 ÿ y_ 3 † q i Qi
iˆ1 iˆ1

X
N X
N
‡ l2 y_ 3 sin…y2 ÿ y3 † q_ i Qi ‡ l2 y_ 2 sin…y2 ÿ y3 † q_ i Qi
iˆ1 iˆ1

1 XN
‡ rl2 l23 y_ 2 y_ 3 sin…y2 ÿ y3 † ÿ l2 y_ 2 y_ 3 q_ i Qi ÿ l2 sin…y2 †l1 ‡ l2 cos…y2 †l2 ˆ t, …13†
2 iˆ1

1 3 XN XN X N XN X N
rl y3 ‡ 
qi Si ‡ y3 _
mij qi qj ‡ y3 mij qi qj
3 3 iˆ1 iˆ1 jˆ1 iˆ1 jˆ1

1 XN
ÿ rl2 l23 y_ 2 …y_ 2 ÿ y_ 3 † sin…y2 ÿ y3 † ‡ l2 y2 sin…y2 ÿ y3 † qi Qi
2 iˆ1

X
N X
N
‡ l2 y_ 2 …y_ 2 ÿ y_ 3 † cos…y2 ÿ y3 † qi Qi ‡ l2 y_ 2 sin…y2 ÿ y3 † q_ i Qi
iˆ1 iˆ1

X
N
1 1
ÿ l2 y_ 2 sin…y2 ÿ y3 † q_ i Qi ÿ rl2 l23 y_ 2 y_ 3 sin…y2 ÿ y3 † ‡ rl2 l23 y2 cos…y2 ÿ y3 †
iˆ1
2 2

X
N
ÿ l2 y_ 2 y_ 3 cos…y2 ÿ y3 † Qi ÿ l3 sin…y3 †l1 ‡ l3 cos…y3 †l2 ˆ 0, …14†
iˆ1
176 M. KARKOUB AND A. S. YIGIT

Io 
y4 ‡ l4 sin…y4 †l1 ÿ l4 cos…y4 †l2 ˆ 0, …15†

X
N

y 3 Si ‡ mij qj ‡ l2 y2 cos…y2 ÿ y3 †Qi
jˆ1

X
N
ÿ l3 y_ 2 …y_ 2 ÿ y_ 3 † sin…y2 ÿ y3 †Qi ÿ y_ 23 mij qj
jˆ1

X
N
ÿ l2 y_ 2 y_ 3 sin…y2 ÿ y3 †Qi ‡ kij qj ˆ 0 i ˆ 1, 2, . . . N, …16†
jˆ1

where
… l3 … l3
Qi ˆ rci dx, Si ˆ rxci dx,
0 0
… l3 … l3
00 00
mij ˆ rci cj dx, kij ˆ EIci cj dx,
0 0

and t is the torque applied at the input joint.

3. OPEN LOOP SIMULATIONS

A set of (N ‡ 3) differential equations are obtained by an N-mode
approximation from the Lagrange's equations. Therefore, the four-bar
mechanism system is governed by a total of (N ‡ 3) differential and two
algebraic constraint equations. As it is well known solving differential-algebraic
equations numerically is a challenging task. For simulations presented here an
IMSL subroutine (DASPG) which is based on Petzold±Gear method is used [12].
The number of modes to be taken to guarantee convergence depends on the type
of excitations. Since the primary objective of this study is to investigate the
feasibility of controlling the vibrations through the input link, a smooth
sinusoidal torque input is used to drive the system. In this case only the ®rst
mode is dominant. Therefore, a two-mode approximation is considered to be
adequate. For comparison purposes, another set of simulations was carried out
by assuming that the coupler link behaves as rigid (i.e., neglecting the elastic
deformations). The values of the parameters used in the simulations are given in
Table 1. The torque input is given by

t0 sin…2pt=Tm † tETm ,
t…t† ˆ …17†
0 teTm ,

where t0 is the peak torque and Tm is the duration of the torque.

Figures 2±5 compare open-loop responses of rigid and ¯exible mechanism
with a peak torque magnitude of t0 = 0
003 Nm and Tm = 1 s. As mentioned in
 FOUR-BAR MECHANISM 177
TABLE 1
Parameters used in the simulations
Constant Description Value

l1 Length of the ground link 0
4064 m

l2 Length of the input link 0
0635 m
l3 Length of the coupler link 0
3048 m
l4 Length of the output link 0
3048 m
Ic Moment of inertia of the input link 2
002 6 10ÿ3 kgm2
Io Moment of inertia of the output link 7
46 6 10ÿ6 kgm2
r Mass per unit length 0
2237 kg/m
E Modulus of elasticity 2
06 6 1011 Pa
I Area moment of inertia of the coupler link 5
34 6 10ÿ12 m4
m2 Mass of the input link 0
0142 kg
m3 Mass of the coupler link 0
0682 kg
m4 Mass of the output link 0
0682 kg

section 1, because of the fully coupled nature of the equations, the rigid body co-
ordinates (e.g., input and output link angular displacements) are affected by the
elastic deformation of the coupler link. However, this effect is negligible and can
only be seen in the joint angular velocities (see Figures 4 and 5).
Another set of simulations are carried out with a peak torque magnitude of
t0 = 0
01 Nm and Tm = 1 s. The open-loop responses for the ¯exible and rigid
models of the four-bar mechanism are shown in Figures 6 through 9. The effect
of ¯exibility is now more pronounced, since the larger torque causes larger

3.2

3.0

2.8

2.6
Crank angle (rad)

2.4

2.2

2.0

1.8

1.6

1.4
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 2. Crank angle open-loop time response, t0 = 0
003 Nm; Ð, ¯exible; . . . , rigid.
178 M. KARKOUB AND A. S. YIGIT

2.50

2.45

2.40
Output angle (rad)

2.35

2.30

2.25

2.20

2.15
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 3. Output angle open-loop time response, t0 = 0
003 Nm; key as Figure 2.

3.0

2.5
Crank angular velocity (rad/s)

2.0

1.5

1.0

0.5

0.0

–0.5
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 4. Crank angular velocity open-loop time response, t0 = 0
003 Nm; key as Figure 2.
 FOUR-BAR MECHANISM 179
0.6

0.5

Output angular velocity (rad/s)

0.4

0.3

0.2

0.1

0.0

–0.1
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 5. Output angular velocity open-loop time response, t0 = 0
003 Nm; key as Figure 2.

10

0
Crank angle (rad)

-5

–10

–15

–20

–25
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 6. Crank angle open-loop time response, t0 = 0
01 Nm; key as Figure 2.
180 M. KARKOUB AND A. S. YIGIT

2.6

2.5

2.4
Output angle (rad)

2.3

2.2

2.1

2.0
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 7. Output angle open-loop time response, t0 = 0
01 Nm; Ð, ¯exible; key as Figure 2.

20

15
Crank angular velocity (rad/s)

10

–5
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 8. Crank angular velocity open-loop time response, t0 = 0
01 Nm; key as Figure 2.
 FOUR-BAR MECHANISM 181

2.5

2.0

Output angular velocity (rad/s)

1.5

1.0

0.5

0.0

–0.5

–1.0

–1.5

–2.0
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 9. Output angular velocity open-loop time response, t0 = 0
01 Nm; key as Figure 2.

elastic deformations. The effect of ¯exibility is signi®cant not only on the joint
angular velocities, but also on the joint angles. Clearly, in this case, a rigid
model or a ¯exible model without full coupling will yield gross errors with
respect to rigid body motions since the effect of elastic deformation will not be
seen in rigid body co-ordinates.
It is apparent from Figures 2 through 9 that the ¯exibility of the coupler link
is felt in the input as well as the output links. This makes it possible to suppress
the vibrations of the coupler link through the input torque. However, it is clear
that models based on decoupled rigid and ¯exible motions cannot be used for
the proposed type of control.

4. CONTROL DESIGN AND CLOSED LOOP SIMULATIONS

The governing equations are highly non-linear differential equations. Most
control designs require a linear unconstrained model. In this case a
straightforward linearization is not possible since it may be dif®cult to ®nd an
operating point. However, the equations can be linearized around an open-loop
rigid body trajectory. Since the main goal of the controller is to suppress
vibrations, a rigid body trajectory can be used as the nominal trajectory and the
deviations from this can be assumed small.
For small deviations from the open-loop trajectory, the constraint equations
can also be linearized to obtain an expression for the coupler and output angles
as a function of the input angle. A set of unconstrained linear system of
differential equations describing the mechanism for a small motion range can be
182 M. KARKOUB AND A. S. YIGIT

obtained as follows: Let:

y2 ˆ y2e ‡ dy2 , y3 ˆ y3e ‡ dy3 , y4 ˆ y4e ‡ dy4 ,
where y2e, y3e, y4e represent the open-loop trajectory. Using the constraint vector
F given in equation (9), one can obtain the deviations from the open-loop
trajectory dy3 and dy4 as a function of dy2 as:
l4 ÿ l3 cos…y3e ÿ y4e † ‡ l1 cos…y3e † ÿ l2 cos…y2e ÿ y3e †
dy3 ˆ ÿ
l2 sin…y3e ÿ y4e †
l4 sin…y2e ÿ y3e †
ÿ dy2 : …18†
l2 sin…y3e ÿ y4e †

ÿl3 ‡ l1 cos…y3e † ÿ l2 cos…y2e ‡ y3e †

dy4 ˆ ÿ
l3 sin…y3e ÿ y4e†
l4 cos…y2e ÿ y3e † ‡ l2 sin…y2e ÿ y3e †
ÿ dy2 : …19†
l3 sin…y3e ÿ y4e †
Replacing these linearized constraint equations in the linearized equations of
motion, the resultant unconstrained equations can be written as follows:
Md  Z_ ‡ KdZ
 Z ‡ Cd  ˆ dQ, …20†
where M,  and K
 C,  are the associated condensed matrices, and

dZT ˆ ‰dy2 q1 q2 ... qn ŠT , …21†

dQT ˆ ‰tc 0 ... 0ŠT , …22†

where tc is the control torque applied to the input link.
In order to investigate whether it is possible to suppress vibrations of the
¯exible link by a control torque applied to the input link, a simple control
strategy is tried. The control torque is given by
tc ˆ ÿKp dy2 ÿ Kd dy_ 2 , …23†
where Kp and Kd are the controller gains and dy2 is given by:
dy2 ˆ y2 …t† ÿ y2d …t†, …24†
where y2d(t) is the desired input angle pro®le which can be taken as the desired
open-loop rigid body trajectory. The chosen control scheme employs a motor for
actuation, an encoder for the sensor, all placed at the input link joint. This type
of control arrangement makes real time implementation easy.
For the ®rst set of simulations, the desired trajectory is assumed to be a
constant (regulator type of problem) resulting in a constant set of operation
angles (i.e., zero open-loop input torque). The system is disturbed by an initial
de¯ection of the coupler. The control objective is to suppress the vibrations
while keeping the mechanism's input, coupler, and output angles unchanged.
 FOUR-BAR MECHANISM 183
The initial controller gains are obtained using the linear model and the pole
placement technique (the poles are chosen to be s = ÿ21 2 38
5i, ÿ7
5 2 3i).
The gains associated with the chosen poles are found to be: Kp = 0
008 and
Kd = 0
002; however, these gains are tuned using the nonlinear constrained
model as Kp = 0
08 and Kd = 0
004.
Figures 10 through 12 show the performance of a PD controller. For these
simulations, the system is excited by an initial mid-point de¯ection of 4 mm.
Both open and closed-loop responses are shown. For the open-loop simulations,
there is no torque input applied. The initial de¯ection in the coupler link results
in some initial velocities which cause the joint angles to grow. In other words,
part of the initial elastic energy is transferred to the rigid body motion. This
clearly shows the interaction between the elastic deformations and the rigid body
motion. In the closed-loop case, the energy transferred from the elastic
deformation to the joint motion is absorbed by the control action. Consequently,
the vibrations are suppressed and the rigid body motion is stopped. In short, the
controller is able to suppress the vibrations and control the link angular
motions. The required control torque is shown in Figure 13.
Next, another set of simulations are performed by requiring the input angle to
follow a desired trajectory for a rest to rest maneuver while suppressing the
vibrations of the coupler link. The desired trajectory is selected as
8  
< y ‡ t ÿ sin 2pt , t < Tm ,
2e Tm
y2d …t† ˆ …25†
:
y2e ‡ Tm , t > Tm ,

1.90

1.85

1.80
Input angle (rad)

1.75

1.70

1.65

1.60

1.55

1.50
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Time (s)

Figure 10. Time response with PD control for the regulator problem, the input angle; y2; Ð,
closed-loop; . . . , open-loop.
184 M. KARKOUB AND A. S. YIGIT

2.25

2.24

2.23

Output angle (rad)

2.22

2.21

2.20

2.19

2.18

0.0 0.5 1.0 1.5 2.0 2.5 3.0

Time (s)

Figure 11. Time response with PD control for the regulator problem, the output angle, y4; key
as Figure 10.

10 –3
6

4
Mid-point coupler deflection, w (m)

–2

–4

–6
0 1 2 3 4 5 6
Time (s)

Figure 12. Time response with PD control for the regulator problem, coupler mid-point de¯ec-
tion, w; key as Figure 10.
 FOUR-BAR MECHANISM 185

10 –3
2.5

2.0

1.5

PD controller output torque 1.0

0.5

0.0

–0.5

–1.0

–1.5

–2.0
0 1 2 3 4 5 6
Time (s)

Figure 13. PD controller output torque.

where Tm is the duration of the maneuver. In order to evaluate the performance

of the controller, an open-loop input torque pro®le which would result in a
similar input and output angle is used to obtain the open-loop response. Note
that no inverse dynamics procedure is used to prescribe the open-loop torque.
The purpose here is to investigate the performance of the closed-loop and to
provide quantitative comparisons with respect to the required control action.
Figures 14±16 show both open and closed-loop responses. The search for the
controller gains is done using the linear and non-linear models as in the previous
controller design. The PD controller gains are found to be Kp = 0
05 and
Kd = 1
0. With the designed controller in place, the desired trajectory is followed
with very good accuracy. The vibrations are suppressed within 2 s, and the ®nal
position is reached without any signi®cant vibration. It is clear that the proposed
controller results in a very good closed-loop behavior without requiring excessive
control effort as can be seen by comparing open and closed-loop torques (see
Figure 17).
Proportional plus derivative (PD) control essentially imitates a passive
controller and thus it is unconditionally stable. In fact, it can be thought of as a
variable coef®cient torsional spring and damper. Since it does not require
measurements of the elastic deformations, there is no problem of spillover [13].
However, the performance of a PD controller is limited by the ®rst modal
frequency of the elastic link, i.e., it can not be made arbitrarily fast [14]. This is
because there is no feedback of elastic deformations, and therefore the closed-
loop frequencies related to ¯exible motion can not be altered. Furthermore, the
performance will not be robust for a wide range of input angles because of the
186 M. KARKOUB AND A. S. YIGIT

3.2

3.0

2.8

2.6
Crack angle (rad)

2.4

2.2

2.0

1.8

1.6

1.4
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 14. Time response with PD control for start-stop input command, the crank angle, y2;
key as Figure 10.

2.50

2.45

2.40
Output angle (rad)

2.35

2.30

2.25

2.20

2.15
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 15. Time response with PD control for start±stop input command, the output angle, y4;
key as Figure 10.
 FOUR-BAR MECHANISM 187

10 –4
8

Coupler mid-point deflection (m)

4

–2

–4

–6

–8
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 16. Time response with PD control for start-stop input command, the coupler mid-
point de¯ection, w; key as Figure 10.

10 –3
3

1
Torque (N–m)

–1

–2

–3
0 1 2 3 4 5 6 7 8 9 10
Time (s)

Figure 17. Open-loop and closed-loop torques for start-stop input command; key as Figure
10.
188 M. KARKOUB AND A. S. YIGIT

effect of non-linearities and varying coef®cients. Despite these limitations,

however, the controller can be easily implemented due to its simplicity. The
performance can be improved by gain scheduling where several sets of gains are
used for a large angle maneuver.

5. CONCLUSIONS
Modelling and control of a ¯exible four-bar mechanism has been investigated.
Governing non-linear equations are obtained through a constrained Lagrangian
approach. The equations are fully coupled such that the mutual dependence
between rigid body and elastic motions is preserved. Resulting differential-
algebraic equations are solved numerically to simulate the system open-loop and
closed-loop behavior. A linearized dynamic model is developed which facilitates
design of a simple PD controller which does not require measurement of the
elastic deformations. The gains obtained are used as initial estimates for
controlling the actual mechanism system described by a set of coupled non-linear
differential-algebraic equations. The resultant controller has been shown to be
ef®cient in suppressing the vibrations of the ¯exible link as well as controlling
the rigid body motion. The work is ongoing to improve the performance of the
proposed controllers by gain scheduling.

REFERENCES
1. G. N. SANDOR and A. G. ERDMAN 1984 Advanced Mechanism Design: Analysis and
Synthesis. Englewood Cli€s, NJ: Prentice-Hall.
2. A. A. SHABANA 1989 Dynamics of Multibody Systems, New York: Wiley-
Interscience.
3. S. NAGARAJAN and D. A. TURIC 1990 Journal of Dynamic Systems, Measurement
and Control 112, 203±224. Lagrangian formulation of the equations of motion for
elastic mechanisms with mutual dependence between rigid body and elastic motions,
part 1: element level equations, part 2: system equations.
4. W. SUNADA and S. DUBOWSKY 1981 Journal of Mechanical Design 103, 643±651.
The application of ®nite element methods to the dynamic analysis of spatial and co-
planar linkage systems.
5. A. L. SCHWAB and J. P. MEIJARD 1997 Proceedings of the 1997 ASME Design
Engineering Technical Conferences, Sacramento, CA, 1±7. Small vibrations superim-
posed on non-linear rigid body motion.
6. A. YIGIT, R. A. SCOTT and A. G. ULSOY 1988 Journal of Sound and Vibration 121,
201±210. Flexural motion of a radially rotating beam attached to a rigid body.
7. H. EL-ABSY and A. A. SHABANA 1996 Journal of Sound and Vibration 198, 617±637.
Coupling between rigid body and deformation modes.
8. Y. A. KHULIEF and A. A. SHABANA 1986 ASME Journal of Mechanisms,
Transmissions, and Automation in Design 108, 38±45. Dynamic analysis of con-
strained system of rigid and ¯exible bodies with intermittent motion.
9. C. K. SUNG and Y. C. CHEN 1991 ASME Journal of Vibration and Acoustics 113,
14±21. Vibration control of the elastodynamic response of high-speed ¯exible link-
age mechanisms.
10. D. G. BEALE and S. W. LEE 1995 ASME Design Engineering Technical Conferences
DE-Vol. 84±1, 203±209. The applicability of fuzzy control for ¯exible mechanisms.
 FOUR-BAR MECHANISM 189
11. W. H. LIAO, J. H. CHOU and I. R. HORNG 1997 Smart Materials and Structures 6,
457±463. Robust vibration control of ¯exible linkage mechanisms using piezoelectric
®lms.
12. C. W. Gear and L. R. PETZOLD 1984 Journal of Numerical Analysis 21, 716±728.
ODE methods of the solution of di€erential/algebraic systems.
13. M. J. BALAS 1978 IEEE Transactions on Automatic Control AC-23, 673±679.
Feedback control of ¯exible systems.
14. A. S. YIGIT 1994 ASME Journal of Dynamic Systems, Measurement, and Control
116, 208±215. On the stability of PD control for a two-link rigid-¯exible manipula-
tor.