Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

P. B.

Usoro
A Finite Element/Lagrange
Division Director.
Approach to Modeling Lightweight
R. Nadira
Research Engineer.
Flexible Manipulators
Scientific Systems, Inc., This paper presents a finite element/Lagrangian approach for the mathematical
Cambridge, Mass. 02140 modeling of lightweight flexible manipulators. Each link of the manipulator is
treated as an assemblage of a finite number of elements for each of which kinetic
S. S. Mahil and potential energies are derived. These elemental kinetic and potential energies are
Assistant Professor, then suitably combined to derive the dynamic model for the system. It is contended
Department of Electrical Engineering, that satisfactory modeling and analysis of the manipulator dynamics can lead to the
Purdue University Calumet, use of advanced control techniques to solve some of the problems associated with
Hammond, Ind. 46323 the flexure of otherwise attractive lightweight manipulator arms. Detailed model
development and simulation results for the case of a two-link manipulator system
are presented.

Introduction
Flexible manipulator systems exhibit many advantages over element and Lagrangian methods and the use of the General-
their traditional (rigid-arm) counterparts: they require less ized Inertia Matrix (GIM) concept [9-11], This modeling ef-
material, have less (arm) weight, consume less power, are fort represents a first step towards solving the flexible
more maneuverable, require smaller actuators, and are more manipulator control problem. The manipulator system is
transportable. However, they have not been much favored in modeled as being composed of links attached to each other
production industries due in part to the fact that manipulators with the first link attached to a fixed base. Each link is assum-
are required to have a reasonable accuracy in the response of ed to be symmetrical about its longitudinal axis in the absence
the arm's end-point to the joint control system input com- of deformation (e.g., cylindrical cross-section), and is con-
mands and this is severely deteriorated by structural deforma- sidered as an assemblage of distributed elements of equal
tion, especially in the case of flexible links where the deforma- lengths which are rigid in compression but not in bending.
tion is oscillatory. Traditionally, these vibrations have been Primary control inputs (torques) are applied at the joints and
eliminated by increasing the rigidity of the arms, but this solu- can be used to effect the movement of the end point of the link
tion is not available in the case of flexible manipulators; along feasible trajectories. Additional controls may be applied
therefore, if the advantages associated with lightweight are not at intermediate points along the link for the purpose of damp-
to be sacrificed, efficient controls have to be developed. ing the vibration modes and also to enhance the quality of the
The flexible manipulator control problem is complicated by overall control of the system.
the fact that the dynamics of the system is highly nonlinear Viscous damping at the joints is ignored. Also, only the out-
and complex (this is true even in the rigid arm case). The effec- puts of the actuators (torques in this case) are considered and
tive inertia of the system changes as a nonlinear function of so the dynamics as well as damping associated with the ac-
the robot's configuration. Also, when the manipulator moves tuator system are not modeled. Although it is recognized that
with high speed, the coriolis and centrifugal forces become in many existing robots the flexibility of the power train con-
significant and cannot be neglected. Adding flexural dynamics stitutes a major consideration, often much more than the flex-
complicates the problem much further, and serious problems ibility of the links, the derivation presented here does not treat
are encountered in both the modeling and control of the the power train and only the flexure of the manipulator link is
system [1, 2]. The dynamics of a flexible manipulator system represented.
can be described by a set of partial differential equations
which represent a distributed (continuous) system. This is ef- Model Development Approach
fectively an infinite dimensional system. Two commonly used
approximations to such a system include finite element models A flowchart of the overall modeling approach for the
and modal expansion models [3-8]. general case of an /w-link system is shown in Fig. 1. The
This paper presents a methodology for modeling lightweight overall approach involves treating each link of the
flexible manipulators. The methodology is based upon finite manipulator (say link 0 as an assemblage of «,• elements of
length /,-. For each of these elements (say ij where subscript j
denotes the element number) the kinetic energy (or, in fact,
Contributed by the Dynamic Systems and Control Division for publication in
kinetic co-energy) Ty and potential energy Vy are computed in
the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript terms of a suitably selected system of n generalized variables
received at ASME Headquarters, May 19, 1986. Q = (<7i. <?2> • • • In ] a n d their rate of change q. These energies

198/Vol. 108, SEPTEMBER 1986 Transactions of the ASME


Copyright © 1986 by ASME
Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms
Let * <• 0, i = 1

z
Divide link i into ns finite elements of length /f
T
Compute T{j for a generic element ij
Compute V,y for a generic element ij

I
Combine Ty, j = 1, 2, ..., n, to form J", = J T y

Combine Vy, /' = 1, 2 n, to form V, = %V,j

I
Compute X, = T, - V,
T Fig. 2 Schematic of a two-link flexible manipulator
Apply Boundary Conditions

Let X = ^ + forces. The Generalized Inertia Matrix M is guaranteed, for


physical systems, to be positive definite [9].
/ =i +1 The model in (5) is flexible to the extent that different levels
of model accuracy can be obtained by selecting an appropriate
\ Is f > number of links m? y~ number of elements to represent each link; obviously, the
\ Yes higher the number of elements, the higher the order of the
Apply Lagrange's Equation resulting model.
The specific application of the methodology to the case of a
. 1 £ = C„ / - 1, 2, .... m two-link manipulator is demonstrated below.
to obtain the dynamic equations for the system
Model Development for a Two-Link Manipulator
Consider a two-link manipulator as depicted in Fig. 2. The
manipulator comprises links 1 and 2. Consider link 1 to be
Fig. 1 Flow chart of the model development approach divided into elements ' 1 1 ' , '12', . . . , 'If, . . . ' I n , ' of equal
length /,, and link 2 to be divided into elements ' 2 1 ' , '22',
. . . , ' 2 / . . . '2n 2 ' of equal length / 2 , (refer to Fig. 2). Let us
are then combined to obtain the total kinetic energy X and define the following notation (where subscript / refers to link /,
potential energy V for the entire system. and subscript ij refers to they'th element of link i):
That is,

(1) OXY inertial system of coordinates


7Xq,q)= Ld Ld '•ij
;'=1 j=\ body-fixed system of coordinates attached to
and link i. (In this case O, is the same as O.) QXX is
the direction of the unflexed link 1. 02X2 is the
(2) direction of the unflexed link 2.
i/(q)= L £ F« X Y body-fixed system of coordinates attached to
®ij ij U
i=l y=l element 'if of link i with point 0, y fixed to the
where w is the number of links in the system. Knowledge of undeformed link / at the common junction bet-
the kinetic and potential energies is tantamount to specifying ween elements ' / ( y - 1)' and 'if. That is, O , ^ -
the Lagrangian £ of the system, given by is attached to 0,vT, and OyYy is parallel to
£(q,q)=r-K (3) 0,Y,.
By using results from variational calculus [12], the governing x
nJii = distances along body-fixed system OyX ii Y,u
„ iS
dynamic equations for the system are derived through the from common junction between elements
Lagrange's equations: 'i(j-l)' and 'if of link i
vector in inertial system from O to a given point
d T 3£ 1 3£ _ , , . P in the manipulator
(4)
dt L dqk J dqk vector from O to a point on element' if (in the
where Qk are the generalized nonconservative forces. body-fixed system OiXlYl)
Equation (4) along with the associated necessary boundary vector from 0 2 to a point on element 'If (in the
conditions provide the desired dynamic equations for the 0 2 ^ 2 ^ 2 system)
system, which are of the form: distance vector from O to the origin of the
system of coordinates 0 2 ^ 2 Y2
Mq-f = Q (5) number of elements of links 1 and 2,
nun2
where M=M(q) is a nonlinear function of the generalized respectively
variables q and is called the Generalized Inertia Matrix [9]; 7i>"2 angular displacements of joints 1 and 2,
f = f (q, q) is a vector of nonlinear functions of the generalized respectively
variables q and their rate of change q; and Q represents the li,h length of each element of links 1 and 2,
torques applied at the joints and any other additional applied respectively

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 1986, Vol. 108/199

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


L{,L2 = total lengths of links 1 and 2, respectively where, zJ = [0l u2j_x u2J u2j+l u2J+2\T.
ml ,m2 — uniform density (mass per meter) of links 1 and Thus, from (6) and (11), the kinetic energy Ty of element ' 1 /
2, respectively may be expressed as:
u2j-\ = flexural displacement at the common junction
of elements '1(/— 1)' and 'If of link 1
1
u2J = flexural slope at the common junction of tjMyZj (12)
elements 'l(J—l') and ' 1 / of link 1. This slope
is measured with respect to axis OXt. where
2B,+i == flexural
11CAUI&1 displacement
UlSpictCCIUClU at
i l l the
L11C tip
Lip of
U l the
L11C first
111 SI link
1111K „/ p Q -, J *j
= flexural slope at the tip of the first link Mxj(i,k) = \ m, dxxj, i,k=l,2, . . . ,5 (13)
J L
"2,-1 = = flexural displacement at the common junction junction ° °zjiJ °zjk
of elements '2(y- 1)' and ' 2 / of link 2
w2j = flexural slope at the common junction of and Zy,- is the /th element of zy-. It can be shown that Mxj is of
elements '2(j- 1)' and ' 2 / of link 2. This slope the form
is measured with respect to axis 02X2.
M, ,M2 = generalized inertia matrices of links 1 and 2, ~My(l,l) M, y (l,2) • • • MK (1,5) •
respectively M l y (l,2)
K\ ,K2 = stiffness matrices of links 1 and 2, respectively M„.= p (14)
E = Young's modulus V
/.,/, moment of inertia of links 1 and 2, respectively
bold face is used to denote a column vector M„(l,5) _
an overscore " is used to denote intermediate results
where
Kinetic Energy Computation. The kinetic energy for the
overall system is obtained by computing the kinetic energy for " 156 22/, 54 - 1 3 / , "
each element ij and then summing over all the elements.
»?,/, 22/, 4/1 13/, -3/1
Kinetic Energy for an Element 'lj' of Link 1: The kinetic (15)
energy TXJ for they'th element of link 1 may be computed as: 420 54 13/1 156 - 2 2 / ,
1 f'i VdrT 9r "1
^ = ly (6) -13/, -3/1 -22/, 4/l_
2 Jo 'L at dt J
It is convenient to express the vector r in terms of a vector r, in Also,
the body-fixed system of coordinates OlXl Yx. This is ac- mj\
complished by using the transformation matrix P0 which A/y(l,l) = Vf-y+D + tljPyiij (16)
transforms from coordinate system OxXxY{ to inertial system
OXY, as follows: with

r _rir rl _r cos(?
i -sine,] ^ y = [«2j-i «2y « y + i % + 2 l 7 ' , a n d

My(l,2) = - ^ - ( l Q / - 7 ) ,
and 0, is the joint angle between 01XI and OX (refer to Fig.
2). The vector r, is given by:
M lyJ (l,3) = — L l ( 5 y - 3 ) ,
(8) 60

The displacement yly can be described in terms of shape M,,.(l,4) = ^ - ( 1 0 y - 3 ) , a n d


functions (Hermitian Polynomials) [3-6] <f>k(xlj) as:

-M*iy.0= XJ<M*,y)"2,- 2 + k V) (9) My(l,5)=~^-(5y-2). (17)


where <j>(Xjj)
k is given by:
Total Kinetic Energy for Link 1. As link 1 is divided into
X2 X* X2 X3
*,(*)= 1 4>2 (*) == J C - 2 -+- «, elements, the total kinetic energy of link 1 is computed by
~p' / p summing over all elements ' If of link 1,
2 2 3
x x- X X
«3(jf) = 3 04 (*) = +- (10) Tx = L Txj = £ -i-i/Myi, = 4 - 4 ^ . 4 . (18)
P P' / p
So, from (7)-(10), where
qi=[0i.1?n r .
\—Y ^ [ W , M 2 " 3 " " 2 « , - l "2n, «2n,+l ^ n ^ F - (19)
Lac?, J M, is a Generalized Inertia Matrix and its elements can be
3r r dr 3r 3r 3r
~dT' MM'
L
9" 2 y-1 J
-86, du V-i du 2/+ 2 •]*
found in [13].
Kinetic Energy for an Element '2j' of Link 2. The kinetic
energy T2j for they'th element of link 2 may be computed as:

1 p'2 r3rr 3r 1
r—^-r
L
a"2y + 2 J
(11)
where
'2j- (20)

200/Vol. 108, SEPTEMBER 1986 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


ment ' 1 / of link 1 comprises two components, viz: VglJ due to
gravity and VelJ due to elasticity. That is,
r = rn, + r , •n + T] (21)
vv=vslj+veij

T\ is the transformation matrix from body-fixed system


° 2 ^ 2 Y2 to body-fixed system C^X, Y",, i.e.,
i r'i r3S22 ,^,,"i2 (29)
COS(02 + H 2 „ 1 + 2 ) -sin(0 2 + H2„1+2)T
(22)
n= sin(0 + u „ 2 2 i+2) cos(02 + « 2 „ 1+2 )J By substituting for y^ from (9) and performing the integra-
y2J in (21) may be computed as tion, the elemental potential energy becomes:
4

yy(x,t)= T,<t>iWw2j_2+i(t), (23)

where x2j is replaced by x for convenience and the shape func- Ky=/»,g[o nn
0-4-)
tions <j>j(x) are as given in (10). From (21) and (22) we con-
clude that r is independent of uh for / = 1, 2, . . . 2M,. *2y-l " 2 , + — u 2/+1 *2y + 2
12 12
T2j in equation (20) may be expressed as:
1
Ty- -YzjMyZj (24) (30)

where where \pxj is as defined in (17), and


z/ = [0i "2„,+i «2«,+2 02^/1. 12 6/, -12 6/,
fyT=lw2J-i w2J w2J+l w2j+2] (25)
and M2j is an (8 x 8) matrix whose elements are 6/i 4/? -6/, n
T
Eh -12 12 -61, (31)
P '2 r 5r l dr -6/,
M (i k) = l k=1 2 8 (26)
v - \o " ' L a d feT^'
•jk ' ' 6/, 2/? -6/, 4/?
where Zj, is the /th element of z,-. The elements of M2J are
straightforward to compute and can be found in [13]. Total Potential Energy for Link 1: Since link 1 comprises
nx elements, its total potential energy is
Total Kinetic Energy for Link 2. The kinetic energy T2 for
link 2 is computed by summing over all elements ' 2 / of link 2, 1
i.e., n\{\
"2 "2
T + -±-#•£,0, (32)
T M
^ = y'=it, y =J =EI 4~*/
2 "^' */ =24-<^<k
y y
(27)

where where $x is as defined in (19) and


Q2 = tfi «2«! + I M2„, +2 ^ 2 f D r . and
1?2 = [Wi W2 . . . W 2 „, + 1 W2n_+2]T (28) R 0 =[/, 01/, 0 1 . . . I/, 01-^—^-] (33)

Potential Energy Computation. The potential energy for The elements of Kx, the stiffness matrix, can be found in
the overall system is obtained by computing the potential [13].
energy for each element of the assemblage and summing over
Potential Energy for a Single Element '2j'. Considering
all the elements.
again OX as the reference, the potential energy V2J of they'th
Potential Energy for a Single Element 'lj' of Link 1: Con- element of link 2 is the sum of two components, one due to
sidering OX as the reference, the potential energy Vxj of ele- gravity and the other due to elasticity, i.e.,

Vy=\„
•i: ™2g[0 1]

L
n
. u
2n{ +1 _
+nn
(j-l)l2+x2J

yy
dx0: +-
i
\>m
r'2
dx2J

(y- i/2)/i
Li
2g[o im\u ]i2+n-n + -Y^2jKy^2j (34)
v 2j v
V-l 12 2 2y+l 2;+2
12

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 1986, Vol. 108/201

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


where \py is as in (25) and The Lagrangian and Final Dynamic Model
' 12 6/2 -12 6/2 From (37) and (39), the Lagrangian for Link 1 is

EI2 6/ 2 4/1 -6/2 2/1 •Ci^i-^-^-qfMiqi


(35)
-12 -6/2 12 -6/2

6/2 2/1 -6/2 4/1 -i»,g[o i ] r j 2 ' ' -tf*i*i- (41)


Potential Energy For Link 2. Summing over all elements
'2/ of link 2 the total potential energy of the link becomes Ro^i
The Lagrangian for link 2 is £ 2 = T2 - V2. Therefore, from
"2
(38) and (40),
«,/,
M
2«, +1
£i=-^-(llM2q2-m2g[0 l]T[0

—nlti
+ rt -$lKi$i (36) L, \n\ti
"2/2 + T\ -^K2\p2
U
R,<^2 . 2n{ + 1 _
_ Riife . (42)
where iA2 = [w1 w2 . . . . w2„2+2],and
The overall Lagrangian for a two-link flexible manipulator
R, = [/2 o| .. /, 0
/2 /I can then be written as:
12
£ = £l(dl,u3,u4, . . . u2n +2 )
The elements of K2, the stiffness matrix, can be found in [13]. + £2(6l,u2„1+l,u2„l+2,e2,wi,w4, . . . w2„2+2) (43)
Boundary Conditions. The joint of link 1 is constrained to By applying Lagrange's equations (4) to the Lagrangian in (43)
have zero displacement (translational) and angular displace- and performing algebraic manipulations the dynamic model
ment of 0, with respect to the inertial axis OX This constrains of the system becomes
variables ux and u2 to be zero, i.e., u1(f) = 0 and u2(/)=0.
Also, the second link is constrained to have an angular where
displacement of 82 relative to the first. This constrains
variables wx and w2 to be zero. qr= [0,^02^].
By enforcing these boundary conditions, the kinetic and
potential energies for the links become Q r = [r1Ffr2F2r],Ff=[F11F12
F
r = [^21 E22 . . . F'2,2n
: \ 2
T^—qfM,q, (37)
_a e
i,2A:-i.^i,2/t PPli d force and torque at junction of
elements 'Ik' and '1(A:+ 1)' of link 1 {k<n{)
T2=— q\M2i\2 (38)
^ t - 1 > ^2,2*: ~ aPPlied force and torque at junction of
elements '2A:' and '2{k+ 1)' of link 2 {k<n2)
1
n\t\ Tj — driving torque at joint i
V^m.glO i m T + —+fK^l. (39)
M-{In [ + 2n2 + 2) x (2n, + 2«2 + 2) symmetric
positive definite Generalized Inertia Matrix of
V2 = m2gl0 1]TJ the system
1 and
•nW
nM + T\ 2 d£ [d(£1+£2) d£, d£,
f=-
*2n, +1 -[•
R11A2
d(£,+£2) d(£,+£2) d£2 d£2
+ —tlK24,2 (40) (45)
du1 du2n. +2 dd2 ' tyT2
where
Modification of the Model to Include a Point Mass at the
«2„ 1+ 2]. ^=[01^?] Tip as Load. The extension of the model to the case where a
i/-2r=[w3)>v4 w 5 V
point mass (m) is added at the tip of the manipulator follows
2n2+2J
from our discussion above in a straightforward manner. One
<5r can show that the matrix differential model of the overall
r= [<9l W2«j +1 «2n, +2 e»2^JJ system (i.e., with m at the tip) is of the form:
M[, M2, A'1, and K2 are the generalized inertial and stiffness Mrq — fT = QT (46)
matrices for the links and are obtained by eliminating the ap-
propriate rows and columns from M : , M2, Kx, and K2, con- where
sistent with the above specified boundary conditions. MT=M+Mm, tT = t + tm, Q r = Q + Q„

202/Vol. 108, SEPTEMBER 1986 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


and the subscript 'V refers to the total system, and subscript are superimposed. Furthermore, the responses of the end ef-
'm' represents the influence of the additional mass. fector (tip-X2TIP and Y2TIP) are also undesirably
oscillatory.
Simulation Results
Test 2: Simulation of the Nonlinear Model With
In order to initially check the validity of the model, the Payload. This case is similar to Test 1 with the exception that
following two tests were performed. First, the response of the a mass (payload) of 1 kg is attached at the tip of the second
system to initial conditions corresponding to both links hang- link. The response of the system to the initial conditions
ing freely under gravity was simulated. The system response specified by Fig. 3 is essentially similar to that presented in
showed no change of state as expected. Second, the response Fig. 4 for the case without payload. A major difference lies in
of the system with very large elastic constant to an initial con- the fact that the added mass (payload) increases the inertia of
dition corresponding to 0, = - 95 deg and 02 = 0 deg (refer to the system and therefore the corresponding inertia forces.
Fig. 2), was simulated. The response of the system was in Consequently, the amplitudes of oscillation of the system are
agreement with the harmonic motion of an inelastic bar hang-
ing freely under gravity.
Several additional simulations of the system were perform-
ed to study the transient behavior of the flexible manipulator
system. They were selected to illustrate the response of the
nonlinear model with and without a payload at the tip. The
parameter values of the model used in these simulation studies
were L, Im, /, mH Wj = w 2 = 5 k g / m ,
and £•,=.&,= 2.0 x 10''N/m 2 .
In these simulations, each of the manipulator links was
discretized into two elements (i.e., /? 1 =2, n2 = 2). Also, the : Is
nominal equilibrium position was that where each link was ?
->
5:7
t—
hanging freely under gravity with no torques applied at the
joints. This correponds to dl = — 90 deg, d2 = 0 deg, the other
variables set equal to zero, and torques TX — T2 = 0 N»m. The
results presented show the system responses to an initial condi- '0.00 1.00 '2.00 3.00 .00
tion perturbation of 5 deg in 82, the relative angle between the TIME!SEC)
first and second links. In other words, we simulate the open
loop system response to the following initial conditions:
0, = - 90 deg, d2 = 5 deg, TX = T 2 = 0, and all other variables set
equal to zero. This initial condition is illustrated in Fig. 3.
Test 1: Simulation of the Nonlinear Model Without
Payload. The response of the system without payload to the
initial condition in Fig. 3 is shown in Fig. 4. The variables u},
u4, « 5 , w6, w3, vf4, w5, vf6 in Fig. 4 represent the flexural
responses of the system. The responses of these variables por-
tray the vibration modes of the system, and their influence on
the quality of the system response. Observe that two modes of 2.00 3.00 5.00
o.oo
oscillations are featured in each of the responses: a relatively TIME(SEC)
fast mode corresponding to high frequency vibration and a
relatively slower oscillatory mode upon which the vibrations

.00 1.00 2.00 3.00 4.00 5.00


* TIME!SEC)
en

l.A
a.

ih
J p
AA/W u
>*
R. M\ r^ H
v\r V^
v V V, —M—
M v
— i —
'0.00 1.00 2.00 3.00 4.00 5.00
TIME(SEC)
Fig. 3 Initial conditions for the simulations
Fig. 4 Simulation results of the model without payload

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 1986, Vol. 108/203

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


2g
x •
« §
33 B<

1.00 2.00 3.00 5.00 'o.oo 1.00 2.00 3.00 4.00 5.00
TIME(SEC) TIME(SEC)

c u
1 llIII,' , iliki ML .ilUiL .Jl Ui M
* -8
SH o
PkJfluUi mlm
•3." X n^f ' W f W W W »P
en
a-a =8
E?' i S''
in
D
O
— i — 1 1 — i — 1
'o.oo 2.00 3-00 1.00 2.00 3.00 4.00 5.00
TIME(SEC) TIME!SEC)

2.00 3.00 'o.oo 1.00 2.00 3.00 4 .00 5 .00


TIME1SEC) TIME!SEC)
Fig. 4 (cont.)

relatively higher for the case where a mass is attached at the tip into which each link is divided. Of course the higher the
than when no mass is attached. number of elements per link, the higher the order of the
resulting model. The final model derived for the flexible
Discussion
manipulator is extremely nonlinear and complex. It reflects
A mathematical model of a two-link flexible manipulator the variation in the effective inertia of the system as the
has been developed based upon finite element and Lagrange's manipulator configuration changes as well as the interaction
methods. The modeling methodology involves lumping each between the rigid body dynamics and vibration modes of the
link into a number of elements, determining the Lagrange links. Simulation results show that the response of the flexible
functions for the overall system and invoking variational prin- manipulator is highly undesirable and in order to get the
ciples to derive the governing dynamic equations. The model is dynamics of the system to be acceptable for most practical
flexible to the extent that different levels of model accuracy purposes, very effective controls are needed to control the
can be obtained by selecting appropriate number of elements vibration modes.

204/Vol. 108, SEPTEMBER 1986 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms


Acknowledgment 6 Zienkiewicz, D. C , The Finite Element Method in Engineering Science,
McGraw-Hill, London, 1971.
This paper is based upon work supported by the National 7 Melosh, R. J., "Finite Element Approximations in Transient Analysis,"
Science Foundation under Award Number ECS-8260419. Proc. NCKU/AAS Symp. on Eng. Sci. and Mech., Tainan, Taiwan, Dec. 1981,
pp. 701-723.
8 Johnson, T., "Progress in Modelling and Control of Flexible
Spacecraft," J. Franklin Institute, Feb. 1983.
9 Mahii, S. S., "On the Application of Lagrange's Methods to the Descrip-
tion of Dynamic Systems," IEEE Trans, on Syst. Man and Cybernetics, Vol.
References SMC-12, No. 6, Nov./Dec. 1982.
10 Shabana, A., and Wehage, R. A., "Variable Degree-of-Freedom Compo-
1 Whitney, D. E., Book, W. J., and Lynch, P. M., "Design and Control nent Mode Analysis of Inertia Variant Flexible Mechanical Systems," ASME
Configurations for Industrial and Space Manipulators," Proc. JACC, 1974. Journal of Mechanisms, Transmissions, and Automation in Design, Vol. 105,
2 Book, W. J., and Majett, M., "Controller Design for Flexible Distributed Sept. 1983, pp. 371-378.
Parameter Mechanical Arms via Combined State Space and Frequency Domain 11 Sunada, W. H., and Dubowsky, S., " O n the Dynamic Analysis and
Techniques," ASME JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CON- Behavior of Industrial Robotic Manipulators with Elastic Members," ASME
TROL, Vol. 105, No. 4, Dec. 1983. Journal of Mechanisms, Transmissions, and Automation in Design, Vol. 105,
3 Meirovitch, L., Computational Methods in Structural Dynamics, Sijthoff Mar. 1983, pp. 42-51.
and Noordhoff, Alphen aan den Rijn, Netherlands, 1980. 12 Crandall, S. H., et al., Dynamics of Mechanical and Electromechanical
4 Cook, R. D., Concepts and Applications of Finite Element Analysis, Systems, McGraw-Hill Book Company, New York, 1968.
Wiley, New York, 1981. 13 Usoro, P. B., Nadira, R., Mahil, S. S. and Mehra, R. K., "Advanced
5 Huebner, K. H., The Finite Element Method for Engineers, Wiley, New Control of Flexible Manipulators," Final Report, NSF Award No.
York, 1975. ECS-8260419, Apr. 1983.

Journal of Dynamic Systems, Measurement, and Control SEPTEMBER 1986, Vol. 108/205

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 09/09/2013 Terms of Use: http://asme.org/terms

You might also like