See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/277415747

Dynamic Modelling of a Flexible Manipulator System Incorporating Payload:

Theory and Experiments

Article in Journal of Low Frequency Noise Vibration and Active Control · December 2000
DOI: 10.1260/0263092001493056

CITATIONS READS

20 133

3 authors:

Mohammad Osman Tokhi Zaharuddin Mohamed

London South Bank University Universiti Teknologi Malaysia
579 PUBLICATIONS 5,841 CITATIONS 171 PUBLICATIONS 3,371 CITATIONS

SEE PROFILE SEE PROFILE

Mohammad Hasan Shaheed

Queen Mary, University of London
102 PUBLICATIONS 1,785 CITATIONS

SEE PROFILE

All content following this page was uploaded by Zaharuddin Mohamed on 31 May 2018.

The user has requested enhancement of the downloaded file.

 Dynamic Modelling of a Flexible Manipulator
System Incorporating Payload:
Theory and Experiments
M. O. Tokhi, Z. Mohamed and M. H. Shaheed

Department of Automatic Control and Systems Engineering

The University of Sheffield, Mapping St. Sheffield, S1 3JD, U.K.

Received 4 November 2000

ABSTRACT
This paper presents theoretical and experimental investigations into the
dynamic characterisation of a flexible manipulator system. A constrained
planar single-link flexible manipulator is considered. A dynamic model of the
system, incorporating structural damping, hub inertia and payload, is
developed using finite element methods. Effects of damping and payload on the
response of the flexible manipulator are discussed. Performance of the
algorithm in describing the dynamic behaviour of the system is assessed in
comparison to an experimental test-rig. Experimental results are presented for
validation of the developed finite element model in the time and frequency
domains.
Keywords: Dynamic modelling, flexible manipulator, finite element method.

1. INTRODUCTION
Most existing robotic manipulators are designed and built in a manner that
maximises stiffness, in an attempt to minimise system vibration and achieve
good positional accuracy. High stiffness is achieved by using heavy material.
This, in turn, limits the speed of operation of the robot manipulation, increases
the size of actuator, boosts energy consumption and increases the overall cost.
Moreover, the payload to robot weight ratio, under such situations, is low.
Conversely, flexible robot manipulators exhibit many advantages over rigid
robots: they require less material, are lighter in weight, consume less power,
require smaller actuators, are more manoeuvrable and transportable, have less
overall cost and higher payload to robot weight ratio (Azad, 1995 ).
However, control of flexible manipulators to maintain accurate positioning
is an extremely important problem. Due to the flexible nature of the system, the
dynamics are highly non-linear and complex. Problems arise due to lack of
sensing, vibration due to system flexibility, imprecise positional accuracy and
the difficulty in obtaining an accurate model for the system (Piedboeuf et al.,
1993; Yurkovich, 1992 ). Therefore, flexible manipulators have not been
favoured in production industries, due to un-attained end-point positional
accuracy requirements in response to input commands. In this respect, a control
mechanism that accounts for both the rigid body and flexural motions of the
system is required. The complexity of this problem increases dramatically when
a flexible manipulator carries a payload. Practically, a robot is required to
perform a single or sequential task such as to pick up a payload, move to a
specified location or along a pre-planned trajectory and place the payload.

Journal of Low Frequency Noise, 209

Vibration and Active Control
Vol. 19 No. 4 2000
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

However, the dynamic behaviour of the manipulator is significantly affected by

payload variations. If the advantages associated with lightness are not to be
sacrificed, accurate models and efficient controllers have to be developed.
Various approaches have previously been developed for modelling of
flexible manipulators (Azad, 1995 ). These can be divided into two categories.
The first approach looks at obtaining approximate modes by solving the partial
differential equation (PDE) characterising the dynamic behaviour of a flexible
manipulator system. Previous investigations utilising this approach for a single-
link flexible manipulator have shown that the model eigenvalues agree well
with experimentally determined frequencies of the vibratory model (Book,
1984; Cannon and Schmitz, 1984; Hasting and Book, 1987 ). However, with this
approach, the model does not always represent the fine details of the system
(Hughes, 1987 ).
The second approach uses numerical analysis techniques based on finite
difference (FD) and finite element (FE) methods to solve the PDE. Previous
simulation studies using FD methods have shown that the method is simple in
mathematical terms and is more appropriate in applications involving uniform
structures, such as flexible manipulator systems. Further studies have shown
the relative simplicity of the method (Kourmoulis, 1990 ). This approach has
previously been utilised in the dynamic characterisation of single-link flexible
manipulator systems incorporating damping, hub inertia and payload (Tokhi
and Azad, 1995; Tokhi et. al., 1995 ). Experiments have also been conducted,
where acceptable agreement between simulation and experimental results has
been achieved.
The FE method has been successfully used in solving many material and
structural problems. The method involves discretising the actual system into a
number of elements with associated elastic and inertia properties of the system.
This gives approximate static and dynamic characterisation of the actual system
(Rao, 1989 ). The performance of this technique in modelling flexible
manipulators has also been investigated (Aoustin et al., 1994; Tokhi and
Mohamed, 1999; Tokhi et al., 1997; Usoro et al., 1986 ). These investigations
have shown that the method can be used to obtain a good representation of the
system. It has been reported that in using FE methods, a single element is
sufficient to describe the dynamic behaviour of a flexible manipulator
reasonably well. Using a single element, the first two modes of vibration are
well described (Aoustin et al., 1994 ). Moreover, the FE method exhibits several
advantages over the FD method (Tokhi et al., 1997 ). However, in modelling of
the manipulator using FE methods, the effects of structural damping and
payload have not been adequately addressed. Inclusion of these parameters in
the mathematical model is essential for a closer representation of the actual
system. Moreover, most of the investigations utilising the FE method have not
been supported by experimental evaluations. The effect of payload on the
manipulator is important for modelling and control purposes, as successful
implementation of a flexible manipulator control is contingent upon achieving
acceptable uniform performance in the presence of payload variations. The
damping in the real system is expected to make the residual motion converge to
zero as the energy is dissipated, and not to change the resonance modes of the
system (Poerwanto, 1998 ). These provide important performance indications
that can be utilised in the development of an acceptable FE model of the system.
Moreover, experimental investigations form an important step in assessing the
performance of the FE model.
This paper presents theoretical and experimental investigations into the
dynamic characterisation of a single-link flexible manipulator system. A

210
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

dynamic model of the system incorporating structural damping, hub inertia and
payload, is developed using the FE method. The simulation algorithm thus
developed is implemented in Matlab. The performance and accuracy of the
modelling approach is assessed in comparison to a laboratory scale
experimental flexible manipulator. To study the effect of number of elements,
the results are evaluated with a varying number of elements in the algorithm.
Theoretically, more accurate results will be obtained with increasing number of
elements, but at the expense of higher execution times. Effects of damping and
payload on the performance of the manipulator are investigated. Experiments
are performed using a flexible manipulator experimental rig and the results are
presented for validation of the developed modelling approach. Simulation and
experimental results are compared and analysed in both the time and frequency
domains to assess the accuracy of the model in representing the actual system.
The work presented forms the basis of design and development of suitable
control strategies for flexible manipulator systems. The rest of the paper is
structured as follows: Section 2 provides a brief description of the flexible
manipulator system considered in this study. Section 3 describes the modelling
process of the system using the FE method. Simulation and experimental results
are presented in Section 4 and the paper concluded in Section 5.

2. THE FLEXIBLE MANIPULATOR SYSTEM

This section describes the flexible manipulator system used in this study.
Several assumptions utilised in the process of obtaining the dynamic equations
of motion of the system using FE methods are briefly discussed. Moreover, the
main components of the flexible manipulator experimental-rig are described.
A description of the single-link flexible manipulator system considered in
this work, is shown in Figure 1, where XOY and POQ represent the stationary
and moving co-ordinates respectively, t represents the applied torque at the
hub. E, I, r , A, Ih and Mp represent the Young modulus, second moment of
inertia, mass density per unit volume, cross sectional area, hub inertia and
payload of the manipulator respectively. Since the manipulator is long and
slender, transverse shear and rotary inertia effects are neglected. This allows the
use of Bernoulli-Euler beam theory to model the elastic behaviour of the
manipulator. The manipulator is assumed to be stiff in vertical bending and
torsion, thus allowing it to vibrate dominantly in the horizontal direction. In this
work, the effects of gravity are neglected as the motion of the manipulator is
confined to the XOY plane. Moreover, the manipulator is considered to have
constant cross section and uniform material properties throughout.

Figure 1. Description of the flexible manipulator system

211
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

The experimental-rig used in this work consists of three main parts: a

flexible arm, measuring devices and a processor. Figure 2 shows a schematic
diagram of the experimental-rig. The flexible arm is constructed using a piece
of thin aluminium alloy with length, L = 0.9 m, width = 19.008 mm, thickness
= 3.2004 mm, E = 71x10 9 N/m2, I = 5.1924 m4, r = 2710 kg/m3 and Ih =
5.8598x10 –4 kgm 2. The test-rig is equipped with a U9M4AT type printed circuit
motor at the hub driving the flexible manipulator. The motor is chosen as the
drive actuator due to its low inertia, low inductance and physical structure (PMI
Motion Technologies, 1988 ). In this work, a linear drive amplifier LA5600
manufactured by Electro-Craft Corporation is used as a motor driver
(Electrocraft Corporation, 1985). The motor drive amplifier produces a current
proportional to the input voltage.
The measuring devices used in this work are the shaft encoder, tachometer
and an accelerometer along the arm. The shaft encoder, with a resolution of
2048 pulses, is used for measurement of hub-angle of the manipulator. A
precision interface circuit consisting of a TCHT2000 incremental encoder
interface chip and MP7636A double buffered 16 bit multiplexing digital to
analogue (D/A) converter is used to convert the shaft encoder output to an
analogue signal. The tachometer is used for measurement of the hub-velocity.
A miniature integrated circuit piezoelectric accelerometer 303A03 is located at
the end-point of the flexible arm and used for measurement of the end-point
acceleration. The accelerometer has a built-in FET source follower which
lowers the output impedance level. The low impedance output allows the use of
long cables without an appreciable signal loss or distortion.
The processor used for this experimental rig is an IBM-PC compatible based
on 486DX2 50 MHz processor. Data acquisition and control are accomplished
through the utilisation of RTI - 815 I/O board. This board can provide a direct
interface between the processor, actuator and sensors. The experimental set-up
requires one analogue output to the motor driver amplifier and four analogue
inputs from the hub-angle, hub-velocity, end-point acceleration and motor
current sensor. The interface board is used with a conversion speed of 25 m sec
for A/D conversion and settling time of 20 m sec for D/A conversion, which are
adequate for the system under consideration.

Figure 2. Schematic diagram of the experimental rig

212
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

3. ALGORITHM DEVELOPMENT
This section focuses on the development of the FE simulation algorithm
characterising the dynamic behaviour of the flexible manipulator system.
Firstly, the FE method is briefly discussed. Then formulations to obtain the
mass, stiffness and damping matrices and the dynamic equations of motion of
the manipulator utilising the Lagrange equation are presented. The procedure is
further extended to incorporate hub inertia, structural damping and payload into
the dynamic model. The equations of motion are then expressed in a state-space
form, so as to be solved using control system approaches.

3.1 The Finite Element Method

Since its introduction in the 1950s, the FE method has been continually
developed and improved (Rao, 1989 ). The method involves decomposing a
structure into several simple pieces or elements. The elements are assumed to
be interconnected at certain points, known as nodes. For each element, an
equation describing the behaviour of the element is obtained through an
approximation technique. The elemental equations are then assembled to form
the system equation. It is found that by reducing the element size of the
structure, that is, increasing the number of elements, the overall solution of the
system equation can be made to converge to the exact solution.
The main steps in an FE analysis include (1) discretisation of the structure
into elements; (2) selection of an approximating function to interpolate the
result; (3) derivation of the basic element equation; (4) calculation of the system
equation; (5) incorporation of the boundary conditions and (6) solving the
system equation with the inclusion of the boundary conditions. In this manner,
the flexible manipulator is treated as an assemblage of n elements and the
development of the algorithm can be divided into three main parts: the FE
analysis, state-space representation and obtaining and analysing the system
response.

3.2 Simulation Algorithm

For an angular displacement u (t) and an elastic deflection w(x,t), the total
displacement y(x,t) of a point along the manipulator at a distance x from the hub
can be described as a function of both the rigid body motion u (t) and elastic
deflection w(x,t) measured from the line OX as

y(x,t) = xu (t) + w(x,t) (1)

Using the standard FE method to solve dynamic problems, leads to the well-
known equation

w(x,t) = Na (x)Qa(t) (2)

where Na (x) and Qa(t) represent the shape function and nodal displacement
respectively. For the flexible manipulator under consideration, w(x,t) in
equation (2) represents the residual motion of the system. The manipulator is
approximated by partitioning it into n elements. As a consequence of using the
Bernoulli-Euler beam theory, the FE method requires each node to possess two
degrees of freedom, a transverse deflection and rotation. These necessitate the
use of Hermite cubic basis functions as the element shape function (Ross,
1996 ). Hence, for the elemental length l, the shape function can be obtained as

213
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

Na (x) =[f 1(x) f 2(x) f 3(x) f 4(x)]

where
3x2 2x3 2x2 x3
f 1 (x ) =1– + 3 ;f 2(x) = x– + 2 ;
l2 l l l

3x2 2x3 x2 x3
f 3 (x ) = – 3 ;f 4 (x ) =– + 2 .
l2 l l l
For element n the nodal displacement vector is given as

Qa(t) = [wn–1(t) u n–1(t ) wn(t) u n(t)]

where wn–1(t) and wn(t) are the elastic deflections of the element and u n–1(t) and
u n(t) are the corresponding rotations. Substituting for w(x,t) from equation (2)
into equation (1) and simplifying yields

y(x,t) = N(x)Q(t) (3)

where

N(x) = [x Na(x)] and Q(t) = [u (t) Qa(t)]T

The new shape function N(x) and nodal displacement vector Q(t) in equation (3)
incorporate local and global variables. Among these, the angle u (t) and the
distance x are global variables while Na(x) and Qa(t) are local variables.
n–1
Defining s = x –
Si=1
li as a local variable of the nth element, where li is the

length of the ith element, the kinetic energy of an element n can be expressed
as

(4)

and the potential energy of the element can be obtained as

(5)

d2 N
where B =
ds2
Defining Mn and Kn as
l
Mn = º r A(N TN)ds = element mass matrix (6)
0

214
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

l
Kn = º EI(BTB)ds = element stiffness matrix (7)
0

and solving equations (6) and (7) for the n elements, the element mass and
stiffness matrices can be obtained as

3 4
m11 m12 m13 m14 m15
m21 156 22l 54 –13l
r Al m31 22l 4l2 13ll –3l2
Mn =
420 m41 54 13l 156 –22l
m51 –13l –3l2 –22l 4l2

3 4
0 0 0 0 0
0 12 6l –12 6l
EI
Kn = 0 6l 4l2 –6l 2l2
l3
0 –12 –6l 12 –6l
0 6l 2l2 –6l 4l2

where

m11 = 140l2 (3n2 – 3n + 1)

m12 = m21 = 21l (10n – 7)
m13 = m31 = 7l2 (5n – 3)
m14 = m41 = 21l (10n – 3)
m15 = m51 = –7l2 (5n – 2)

Assembling the element mass and stiffness matrices, the total kinetic and
potential energies form equations (4) and (5) an be written as
? T ?
T= 1Q M Q
2

T
P= 1Q KQ
2

where Q(t) = [u w0 u 0 . . . w n u n] T ,

M and K are global mass and stiffness matrices of the manipulator respectively.
The dynamic equations of motion of the flexible manipulator can be derived
utilising the Lagrange equation;

d
dt h j h j
¶L
¶Q?
–
¶L
¶Q
=F

where L = T – P is the Lagrangian and F is a vector of external forces and

215
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

moments. Considering the damping, the desired dynamic equations of motion

of the system can accordingly be obtained as

? ? (t) + DQ? (t) + KQ(t) = F(t)

MQ (8)

where D is a global damping matrix, normally determined through

experimentation.
For the flexible manipulator under consideration, the global mass matrix can
be represented as

3 4
Mu u Mu w
M=
Mu w Mww

where Mww is associated with the elastic degrees of freedom (residual motion),
Mu w represents the coupling between these elastic degrees of freedom and the
hub angle u and Mu u is associated with the inertia of the system about the motor
axis. Similarly, the global stiffness matrix can be written as

3 4
0 0
K=
0 Kww

where Kww is associated with the elastic degrees of freedom (residual motion).
It can be shown that the elastic degrees of freedom do not couple with the hub
angle through the stiffness matrix.
The global damping matrix D in equation (8) can be represented as

3 00 4
0
D=
Dww

where Dww denotes the sub-matrix associated with the structural damping. The
matrix is obtained by assuming that the manipulator exhibits the characteristics
of Rayleigh damping. This proportional damping model has been assumed
because it allows experimentally determined damping ratios of individual
modes to be used directly in forming the global matrix. It also allows
assignment of individual damping ratios to individual modes, such that the total
manipulator damping is constituted with the sum of the dampings associated
with the modes (Chapnik et al., 1991 ). Using this assumption, the damping can
be obtained as

Dww = a Mww + b Kww (9)

where

with j 1, j 2, f1 and f2 representing the damping ratios and natural frequencies of

modes 1 and 2 respectively.

216
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

3.3 Incorporation of Hub Inertia and Payload

By incorporating the hub inertia and payload of the flexible manipulator, the
kinetic energy of the system can be obtained as (Azad, 1995 )

Hence

or

(10)

where w n is the elastic deflection at the end-point of the manipulator.

In equation (10) the second term on the right hand side shows the
contribution of hub inertia and payload to the rotary inertia of the system about
the motor axis and contributes to the 1 x 1 sub-matrix Mu u . The third term
shows the effect of payload on the end-point residual motion of the manipulator
and contributes to Mww whilst the last term contributes to the coupling matrix
Mu w. Utilising equation (10) and the FE formulation (Meirovitch, 1975 ), a new
global mass matrix can be obtained.

3.4 State-Space Representation

The M, D and K matrices in equation (8) are of size m 3 m and F(t) is of size m
3 1, where m = 2n + 1. For the manipulator, considered as a pinned-free arm
with the applied torque t at the hub, the flexural and rotational displacement,
velocity and acceleration are all zero at the hub at t = 0 and the external force
is F(t) = [t 0 … 0]T. Moreover, in this work, it is assumed that Q(0) = 0.
The matrix differential equation in equation (8) can be represented in a state-
space form as

v? = Av + Bu
y = Cv

where

0m is an m 3 m null matrix, Im is an m 3 m identity matrix, 0m 3 1 is an m 3 1

null vector,

217
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

u = [t 0 … 0 ]T ,

Solving the state-space matrices gives the vector of states v, that is, the
angular, nodal flexural and rotational displacements and velocities.

4. RESULTS
In this section, simulation and experimental results of the dynamic behaviour of
the flexible manipulator system are presented in the time and frequency
domains. The system is considered with and without payload. A bang-bang
signal of amplitude 0.3 Nm, shown in Figure 3, is used as an input torque,
applied at the hub of the manipulator. A bang-bang torque has a positive
(acceleration) and negative (deceleration) period allowing the manipulator to,
initially, accelerate and then decelerate and eventually stop at a target location.
System responses are monitored for a duration of 3 sec, and the results are
recorded with a sampling time of 4 msec.

4.1 Simulation Results

The developed FE model was implemented within the Matlab environment on a
Pentium II 333MHz processor. To demonstrate the effect of damping on the
system response, investigations with and without damping were carried out.
Previous investigations have shown that the damping ratio typically ranges from
0.007 to 0.01 (Hasting and Book, 1987). In this work, the damping ratios were
assumed as 0.007 and 0.01 for vibration modes 1 and 2 respectively. Using 12
Hz and 35 Hz, obtained from experiments, as the first two resonance frequencies
(Azad, 1995), a and b in equation (9) can be obtained as 0.6105 and 7.8294x10 –5

Figure 3. The bang-bang input torque

218
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

respectively. To investigate the accuracy of the FE simulation algorithm in

characterising the behaviour of the flexible manipulator, the algorithm was
implemented on the basis of varying the number of elements from 1 to 20. Three
system responses namely the hub-angle, hub-velocity and end-point acceleration
with the corresponding spectral densities (SDs) are obtained and evaluated.

4.1.1 System without Payload

Figures 4 and 5 show the response of the system with the corresponding SDs
without the presence of damping using 1 and 10 elements respectively. It is noted
that a steady-state hub-angle level of 38° was achieved within 0.9 sec using one or
more elements. These proved that a satisfactory dynamic behaviour of a flexible
manipulator, up to the second mode, could be achieved with one or more elements.
Further nodes of the system are obtained with increasing the number of elements

Figure 4. Simulated response of the flexible manipulator without damping

and payload. Number of elements = 1

219
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

as shown in Figure 5. As expected, without the damping effect, the system

response exhibits persistent oscillation. Using one element, the poles of the system
transfer function from torque input to end-point displacement output were
obtained as 0 , 0 , ± j90.82 and ± j300.07. The corresponding zeros were obtained
as ±58.59 and ±229.14 respectively. Since there are two system zeros on the right-
hand s-plane, the system response characterises a non-minimum phase behaviour.
This agrees with previously developed models (Cannon and Schmitz, 1984).
The residual motion of the system is found to be characterised by the first two
modes of vibration. Resonance frequencies of the system were obtained by
transforming the time domain representation of the system into the frequency
domain using FFT analysis. As shown in Figures 4 and 5, with one element, the
resonance frequencies of the system were obtained as 14.49 Hz and 47.70 Hz
whereas with ten elements these were 11.99 Hz, 35.22 Hz and 65.2 Hz. It is noted
that by increasing the number of elements, the system resonance frequencies
converge to more accurate values, but at the expense of higher execution times.

Figure 5. Simulated response of the flexible manipulator without damping

and payload. Number of elements = 10

220
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

This is mainly due to an increase in the size of the mass, damping, stiffness and
state-space matrices. The inter-relation between the number of elements, execution
time and resonance frequencies of the system is summarised in Table I.

TABLE I
Relation between the number of elements, execution times and
resonance frequencies of the flexible manipulator

Number of Execution
Resonance frequencies (Hz)
Elements time (sec)
Mode 1 Mode 2 Mode 3
1 0.38 14.49 47.7 –
2 0.44 11.99 35.71 77.17
3 0.55 11.99 35.46 65.68
5 0.67 11.99 35.46 65.43
10 0.98 11.99 35.22 65.2
20 2.72 11.99 35.22 65.2

Figure 6. Simulated response of the flexible manipulator with damping and

without payload. Number of elements = 10

221
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

Figure 6 shows the dynamic behaviour of the system in the presence of

damping with 10 elements. It is noted that the damping has resulted in
considerable attenuation in the system residual response amplitude. Analysing
the system time response, it is noted that the hub-velocity and end-point
acceleration of the system converged to zero within 1.8 sec. It is also evidenced
from the SDs of the system response, that the damping has not affected the
resonance frequencies of vibration of the system. However, with damping, the
level of vibration reduces as expected.

4.1.2 System with Payload

To demonstrate the effect of payload on the dynamic behaviour of the system,

Figure 7. Simulated response of the flexible manipulator with payload of

20grams. Number of elements = 10

222
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

various payloads of up to 60 percent of the manipulator weight were simulated.

Figures 7 and 8 show the system response with payloads of 20grams and
60grams respectively. These results were obtained with damping using 10
elements. It is noted that the hub-angle decreases with increasing payloads. For
payloads of 20grams and 60grams, the steady-state hub-angle levels were
obtained as 27.5° and 17.5° respectively. With increasing payload, it is also
noted that the magnitudes of end-point acceleration and hub-velocity of the
manipulator decrease. It is also evidenced from the SD of the system response,
that the resonance modes of vibration of the system shift to lower frequencies
with increasing payloads. This implies that the manipulator oscillates at lower
frequency rates than those without payload. Table II summarises the relation
between payload and the resonance frequencies of the system.

Figure 8. Simulated response of the flexible manipulator with payload of

60grams. Number of elements = 10

223
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

TABLE II
Relation between payload and resonance frequencies of the flexible
manipulator. Number of elements = 10

Payloads Resonance frequencies (Hz)

(grams)
Mode 1 (Hz) Mode 2 (Hz) Mode 3 (Hz)
10 11.65 33.22 61.19
20 11.49 32.22 59.44
30 10.99 31.72 58.69
40 10.74 31.22 58.19
50 10.49 30.97 57.69
60 10.24 30.97 57.44

4.2 Experimental Results

Experiments using the experimental rig were conducted for validation of the
developed FE model. In the experiments, the hub-angle, hub-velocity and end-

Figure 9. Response of the flexible manipulator experimental-rig without

payload

224
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

point acceleration were measured and the corresponding SDs were obtained.
These were then compared with the simulation results. Figure 9 shows the hub-
angle, hub-velocity, end-point acceleration, with their SDs, of the flexible
manipulator without payload. It is noted that for the hub-angle, the steady-state
level of 38° was achieved within 1.8 sec. The first three modes of vibration
were obtained as 11.72 Hz, 35.15 Hz and 65.60 Hz.
To investigate the effect of payload on the performance of the manipulator,
experiments were performed using various payloads ranging from 10grams to
60grams. Figures 10 and 11 show the hub-angle, hub-velocity, end-point
acceleration, with their SDs with payloads of 20grams and 60grams
respectively. The system response shows that levels of hub-angle, hub-velocity
and end-point acceleration decrease with increasing payloads. For 20grams and
60grams, the steady-state level of hub-angle was obtained as 31° and 24°
respectively. It is also noted that the settling time of the manipulator response
was affected by variations in the payload. This shows that the manipulator
response has higher settling times with increasing payloads. Analysing the SDs,
it is noted that the resonance frequencies of the system decrease with increasing
payloads. The relation between payload and resonance frequencies of the
system is summarised in Table III.

TABLE III
Relation between payload and resonance frequencies of the flexible
manipulator experimental rig

Payloads Resonance frequencies (Hz)

(grams)
Mode 1 (Hz) Mode 2 (Hz) Mode 3 (Hz)
10 10.97 33.40 61.85
20 10.23 32.95 59.35
30 9.97 31.42 59.35
40 9.48 30.92 58.85
50 9.23 30.42 58.10
60 8.97 30.17 56.61

4.3 Model Validation

Validation of a dynamic model for use in simulation and control is an important
step before the model can be employed with confidence. Typically, model
validation can be considered in two parts: frequency-domain validation, which
involves the resonance frequencies of the system, and time-domain validation,
which focuses on the time response of various system states to an input
command. Matching of natural frequencies is a good indication of accurately
modelled mass and stiffness properties. Time-domain results show the effects
of assumptions concerning the non-linear terms in the equations of motion.
Validation of the developed FE model was carried out by comparing the
simulation and experimental results in time and frequency domains.
Comparisons of Figures 6 and 9, for the manipulator without payload, show that
a close agreement between experimental and simulation results in the time
responses and resonance frequencies was obtained. For the hub-angle, a steady-
state level of 38° was achieved within 1.8 sec in both cases. Similar
characteristics are also noted in the transient response of the system.
Furthermore, reasonably close agreement between the simulation and
experimental results is noted with the hub-velocity and end-point acceleration
responses. It is also noted in Table I that the first three modes of vibration of the

225
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

system converged to 11.99 Hz, 35.22 Hz and 65.20 Hz with 10 elements or

more. The experimental results, however, gave 11.72 Hz, 35.15 Hz and 65.60
Hz. The corresponding errors between the simulation and experimental results
for modes 1, 2 and 3 are accordingly 2.3 %, 0.2 % and 3.9 %, which are
considered negligibly small. It can thus be concluded that FE methods can
successfully be used for modelling of a flexible manipulator. Moreover, these
validate the assumptions used in this work.
For the manipulator with payload, comparing Figures 7, 8, 10 and 11, it can
be observed that simulation and experimental results were slightly different in
the steady-state level and settling time of the hub-angle of the system. It is noted
that the differences increase with increasing payloads. The differences are
within 2° – 6°. These could mainly be due to the gravity effect, which is ignored
in the simulation, whereas a payload that might be affected by gravity was used
in the experiments. Theoretically, gravity contributes to the potential energy

Figure 10. Response of the flexible manipulator experimental-rig with

payload of 20grams

226
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

and thus the mass matrix of the system. Moreover, payload rotary inertia is
ignored in the simulation. However, with increasing payloads, both cases show
a decrease in the levels of hub-angle, hub-velocity and end-point acceleration.
In the frequency-domain, comparisons of experimental and simulation results
in Tables II and III, for the first three modes of vibration, show that reasonably
close results were achieved for all payloads. Although slight differences were
noted in the time-response, the simulation algorithm provided a close
characteristic behaviour of the experimental rig with increasing payloads. Such
characteristic behaviour is important and essential for development of suitable
control strategies for flexible manipulator systems.

5. CONCLUSION
Theoretical and experimental investigations into the dynamic modelling and
characterisation of a single-link flexible manipulator system have been

Figure 11. Response of the flexible manipulator experimental-rig with

payload of 60grams

227
DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

presented. A dynamic model of the manipulator incorporating damping, hub

inertia and payload has been developed using FE methods. The performance
and accuracy of the algorithm has been studied in comparison to an
experimental rig. The effects of damping and payload on the system behaviour
have been addressed. Experiments have been performed using the experimental
rig and used for validation of the FE model. A very close agreement between
simulation and experimental results has been achieved. Thus, confidence in the
accuracy of the model for utilisation in subsequent investigations, at
development of control strategies for flexible manipulator systems, has been
established.

6. REFERENCES
Aoustin, Y. Chevallereau, C. Glumineau, A. and Moog, C. H. (1994 ).
“Experimental results for the end-effector control of a single flexible
robotic arm”, IEEE Transactions on Control Systems Technology, Vol. 2,
pp. 371 –381.

Azad, A. K. M. (1995). Analysis and design of control mechanisms for flexible

manipulator systems, PhD. Thesis, Department of Automatic Control and
Systems Engineering, The University of Sheffield, UK.

Book, W. J. (1984 ). “Recursive lagrangian dynamics of flexible manipulator

arms”, International Journal of Robotics Research, Vol. 3, pp. 87–101.

Cannon, R. H. and Schmitz, E. (1984 ). “Initial experiment on the end-point

control of a flexible one-link robot”, International Journal of Robotics
Research, Vol. 3, pp. 62–75.

Chapnik, B. V., Heppler, G. R. and Aplevich, J. D. (1991 ). “Modeling impact

on a one-link flexible robotic arm”, IEEE Transactions on Robotics and
Automation, Vol. 7, pp. 479 –488.

Electrocraft Corporation. (1985 ). DC motors speed control servo systems,

Electrocraft Corporation / Robbins & Mayers, Minnesota.

Hasting, G. G. and Book, W. J. (1987 ). “A linear dynamic model for flexible

robot manipulators”, IEEE Control Systems Magazine, Vol. 7, pp. 61–64.

Hughes, P. C. (1987 ). “Space structure vibration modes: How many exist?

Which are important”, IEEE Control Systems Magazine, Vol. 7, pp.
22–28.

Kourmoulis, P. K. (1990 ). Parallel processing in the simulation and control of

flexible beam structures, PhD. Thesis, Department of Automatic Control
and Systems Engineering, The University of Sheffield, UK.

Meirovitch, L. (1975 ). Elements of vibration analysis, McGraw-Hill Inc., New

York.

Piedboeuf, J. -C. Farooq, M. Bayoumi, M. M. Labinaz, G. and Argoun, M. B.

(1993 ). “Modelling and control of flexible manipulators – Revisited”,
Proceedings of 36th Midwest Symposium on Circuits and Systems,
Detroit, pp. 1480 –1483.

228
 DYNAMIC MODELLING OF A FLEXIBLE MANIPULATOR SYSTEM

PMI Motion Technologies. (1988 ). General application of printed motors,

PMI Motion Technologies, New York.

Poerwanto, H. (1998 ). Dynamic simulation and control of flexible manipulator

systems, PhD. Thesis, Department of Automatic Control and Systems
Engineering, The University of Sheffield, UK.

Rao, S. S. (1989). The finite element method in engineering, Pergamon Press,

Oxford.

Ross, C. T. F. (1996 ). Finite element techniques in structural mechanics,

Albion Publishing Limited, West Sussex.

Tokhi, M. O. and Azad, A. K. M. (1995 ). “Real time finite difference

simulation of a single-link flexible manipulator incorporating hub inertia
and payload”, Proceedings of IMechE-I: Journal of Systems and Control
Engineering, Vol. 209, pp. 21–33.

Tokhi, M. O. and Mohamed, Z. (1999 ). “Finite element approach to dynamic

modelling of a flexible robot manipulator: Performance evaluation and
computational requirements”, Communications in Numerical Methods in
Engineering, Vol. 15, pp. 669 –676.

Tokhi, M. O. Mohamed Z. and Azad, A. K. M. (1997). “Finite difference and

finite element approaches to dynamic modelling of a flexible
manipulator” , Proceedings of IMechE-I: Journal of Systems and Control
Engineering, Vol. 211, pp. 145 –156.

Tokhi, M. O. Poerwanto, H. and Azad, A. K. M. (1995 ). “Dynamic

simulation of flexible manipulators incorporating hub inertia, payload
and damping”, Machine Vibration, Vol. 4, pp. 106 –124.

Osoro, P. B. Nadira, R. and Mahil, S. S. (1986 ). “A finite element/lagrange

approach to modelling lightweight flexible manipulators”, Transactions
of ASME: Journal of Dynamic Systems, Measurement and Control, Vol.
108, pp. 198 –205.

Yurkovich, S. (1992 ). “Flexibility effects on performance and control”, Robot

Control, IEEE Press, Part 8, pp. 321 –323.

229

View publication stats