Dynamics of Large Constrained

F. M. L. Amirouche
Department of Mechanical Engineering,
University of Illinois at Chicago,
Flexible Structures
Chicago, III. 60680 This paper presents an automated procedure useful in the study of large constrained
flexible structures, undergoing large specified motions. The structure is looked upon
as a "partially open tree" system, containing closed loops in some of the branches.
R. L Huston The governing equations are developed using Kane's equations as formulated by
Mechanical and Industrial Engineering, Huston et al. The accommodation of the constraint equations is based on the use of
University of Cincinnati, orthogonal complement arrays. The flexibility and oscillations of the bodies is
Cincinnati, Ohio 45221 modelled using finite segment modelling, structure analysis, and scaling techniques.
The procedures developed are expected to be useful in applications including
robotics, space structures, and biosystems.

Introduction
In recent years there has been considerable interest in pre- vibration characteristics as well as dynamic response of struc-
dicting the dynamic response of large mechanical structures. tures with closed loops and constraint motions.
The development of the governing dynamical equations for a This paper itself is divided into four parts: the first provides
structure with constrained motions or with closed loops has preliminary development of the equations of motion for a par-
been of particular interest. The methods used to treat this par- tial open tree system; the second part discusses the flexibility
ticular problem have been motivated by applications in analysis; the third part illustrates the reduction of the equa-
robotics, mechanisms, space structures and biosystems. tions of motion using orthogonal complement arrays; the final
The procedures employed for studying a system subject to part presents the application and contains concluding
these constraints can be computationally expensive particular- remarks.
ly with the introduction of parameters such as lagrange
undetermined multipliers [1, 2, 3]. Huston et al. [4, 24] have I Governing Equations of Motion
shown not only how Kane's equations can be used for
holonomic and simple non-holonomic systems without in- In studying the dynamics of large structures, such as the one
troducing undetermined multipliers, but the equations of mo- shown in Fig. 1, it is shown that using Kane's equations [6] the
tion could easily be obtained through the use of orthogonal equations of motion may be expressed in the form:
complement arrays. a,Pxp=fi d,P=l, . . . ,6N) (1)
There are a few methods available to automatically develop
orthogonal complement arrays. These include the singular where the alp form elements of the generalized mass matrix,
value decomposition method as exposited by Likins and Singh the Xp form elements of the generalized coordinate array, and
[11], and Mani [12], Another approach is the use of the zero the // represent the generalized force array. It can also be
eigenvalue theorem exposited by Walton and Steeves [5] as shown that the generalized mass matrix takes the following
presented by Kamman and Huston [4]. Still other methods form:
have been developed [2, 15]. These use generalized coordinate a
lp — mk Vklm Vkpm + hmn^klm^kpn (2)
partitioning to isolate the independent coordinates. However where mk denotes the mass of typical body Bk, Iklm are com-
these methods usually require more effort than needed. In ponents of the central inertial dyadic of Bk, and Vklm are the
many cases the analyst must make an appropriate selection of components of the partial velocity and partial angular velocity
the coordinates to avoid singularities. vectors (see references [6, 16, 8]). The notation is adopted
In this paper the use of orthogonal complement arrays is from [6]. For example, there is a sum on repeated indices from
used in developing the governing equations of motion of large 1 to N on k, from 1 to 6N on / andp, and from 1 to 3 on m and
flexible structure undergoing large motions. The flexibility ef- n, where A' is the number of bodies. The generalized force
fects of the bodies is modelled using finite segment modelling elements are
with data from analytical solutions, or experimental modal
analysis. This approach is expected to be useful for studying fl=~ (Fl ~ mk Vklm VkumXu

+ hmn aklm ^kun^u + enmhhmr01kun <"> ksru kln^u^x ) 0)

where the Ft are elements of the "generalized active force" ar-
ray [6] and enmh are components of the permutation function.
Contributed by the Dynanic Systems and Control Division for publication in
the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL. Manuscript
For small oscillations (see reference [22]), equation (3) reduces
received by the Dynamic Systems and Control Division, July 25, 1986. to the simple form:

78/Vol. 110, MARCH 1988 Transactions of the ASME

Copyright © 1988 by ASME
Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 ft=-Fi (4)
where the generalized active forces components may be ex-
pressed in the form:

F, =Ff (X,) -Ff (X,) + £y]}(X h X„t) (5)

The Ff(Xi) represent the contribution to the generalized

system by the elastic properties of the system. The Ff(Xt)
represent the contribution by the internal frictions and damp-
ing forces. Finally,/); (X„ Xh t) represent the contribution to
the generalized forces by externally applied forces such as
gravity and contact forces. The attractive feature of equation
(5) is that axial forces bending moments and torsional forces
occur singly—that is one per equation.

II Equations of Motion of Flexible Bodies Using

Finite Segment Modelling
Suppose a typical body Bk of the structure can be
represented by Nk segments. Let Sk (i = 1 , 2 , . . .A^) denote
the rth segment of the body k. Let the coordinates be divided
into two sets Xr and r\kP, where Xr (r = 1, 2, . . .6k) represent
the orientation and displacement of the body reference frame
fixed at 0k, and ij„ (p = 1, 2, . . . 6 ^ ) represent the relative
orientation and displacement of the segments in body k with
respect to 0k. The coordinates -r\p are the elastic coordinates. Fig. 1 A partially closed tree-like structure
They are divided into two sets of triplets. The first set contains
3Nk. It denotes the orientation. In this s e t p varies from 1 to
3Nk. In the second set p varies from 3Nk+l to 6Nk. This set
the angular velocity of Sk with respect to Bk system could be
represents the relative translation, for example.
expressed as
r\\
< • *

0k y,Ar„
CO ilm
(9)
V2k orientation of Sk
expressing nkm with respect to the unit vectors fixed in R,
Vk3 relative to Bk system equation (9) takes the following form
k
r,k^
CO = < m ^" n"-0 m (10)

orientation of S£ where
k o4,=«?r_'SOK„
i)5 j relative to S
k
Note that if Sf is fixed in the Bk system then -q = T]-, = n] = 0 k k In equation (9) u ,' is the partial angular velocity array of the
as vw,•*+l ~~ 7,™H _ „k__
r — 0. If the bodies undergo segment of body k in the local body frame, whereas tok. is the
k +2 ~ l3Nk+i
small deformation and if we assume that the segments are partial angular velocity array of the rth segment of body k in
originally aligned, then we can express the transformation the reference frame R. Substituting equation (10) and equa-
matrix between any two Sk and Sk segments as tion (11) into the right-hand side of equation (8) leads to
R-Sf
1 -Vii 13/-1 = [uklmXi + aikJk]n0 (12)
•ni l -•^31-2 (6) Note how we are able to isolate two arrays with the generalized
coordinates derivatives X: and T),-, respectively. The angular
-ni-i •n k
-2 1 . acceleration of segment Sf is given by differentiation of equa-
Using procedures developed by Huston et al. [6], the absolute tion (12)
angular velocity of any body could be expressed as
(X =lukimX, + akimXl + uklJk+o}klmrik]n0m (13)
« * = wklmXln0m (7)
The rate of change of the partial angular velocities wWm and
(/=1,2, .6N) co*- is function of the rate change of the transformation
matrices. Expression for these matrices have been formulated
( m = 1,2,3)
in previous research [6]. A position to the center mass of body
where wkim, known as the "partial angular velocity array," is a k in a given structure could be defined as
block matrix composed of transformation matrices. Ex-
tending this approach to include the segments of the bodies
and using the addition theorem for the angular velocities [16], P*=E*i+X)<li + r* (14)
an expression for the angular velocity of the rth segment is
found to be where J,- denotes the translation between bodies, q,- is the body
R-sf R-k 0k-Sk vector, and rk and the position from the body reference frame
00 = (JO + CO (8) of Bk to its mass center. The summation on q, and ij, is simply

Journal of Dynamic Systems, Measurement, and Control MARCH 1988, Vol. 110/79

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 tional we can drop the coupling terms associated with b' as
well as the one for d'. This assumption is based on the fact
that their contribution is minimal if a large number of
segments is used to model the flexibility effects. Adding the
contribution of joint stiffness between segments to equation
(20) we obtain
d<f.rjf+klprip + cilpnl=-F(XlXlXit) (21)
Equation (21) represents the governing equations of motion of
Fig. 2 A typical finite segment modelling of body ft large structures including the flexibility effects. klp represent
the generalized stiffness matrix of the elastic bodies. An il-
the path from a point 0 fixed in R (inertial frame) to a point 0^. lustration on how to estimate the stiffness matrix could be
fixed at the center of the unit vectors nki (i = 1, 2, 3). derived based on procedures developed by Amirouche in
By simple modification of equation (14) we can write similar reference [7].
equation for the position vector to the center mass of a seg-
ment Sf in body k. Let vf (/ = 1, 2 Nk) locate the ith II Estimation of the Stiffness Matrix
mass center relative to the adjacent lower unumbered segment
Consider a typical body Bk, modelled by n segments. Let
as shown in Fig. 2. Let the position of the fth segment in body
these segments be connected by stiffness and damping
k be given by
matrices. The stiffness matrix Km of body Bk could be written
as a partitioned matrix with the form:
*f= £«;+!>,+ £ ' , (15) ' KB 0 "
j=\ J=2 9=1

expressing all the position vectors in equation (12) with respect

to unit vectors n0m fixed in R, we obtain
0 K'B _
k k
p r
f=Ti ijmSQJpmn0p + D qjmS0(J- l)pmn0p Where K B is the element stiffness matrix for rotation (that is
torsion along nx and bending along n,, and nz, respectively)
/ and where K'B represents the stiffness due to translation (along
nx, ny, and nz). If the body BK is undergoing oscillatory mo-
+ L rksS0kQmsS0Kpmn0p (16)
tion, then the relative orientation between segments can be
made arbitrarily small by increasing the number of segments.
where S0kQ denotes the transformation matrix between the Then the stiffness matrices KrB and K'B can be obtained by us-
segment axis and body k reference axis; SOA'is the transforma- ing standard procedures of structural analysis as shown in [9].
tion matrix between body k and the reference frame R, £-m, The global stiffness matrix for a tree-like structure could
qjm, and rks are the scalar components associated with their then be expressed in the partitioned form
respective vectors given by equation (15). Differentiation of
equation (16) yields the mass center velocity of segment i with Kr 0
respect to R. After some computation and using the defini-
tions given by Huston et al. in reference [6], the mass center
velocity of segment Sf could be expressed as
0 K, _
k
y-~k={ VklmXt + V rJf]n0m (17) where Kr is a diagonal matrix whose elements are given by
where Vklm is the "partial velocity array" associated with the Kr=KrBiS0I
rate of change of the generalized coordinates Xh and Vk. is
the partial velocity array associated with the elastic coor- (;=1 NandJ=i) (24)
dinates rif. The absolute mass center acceleration is obtained SOI is an orthogonal transformation matrix between reference
by differentiation of equation (14): frame fixed in body B{ and in the inertia frame R. The element
stiffness KrBj is given by
* ; , * = [ VUmX, + VklmX, + VkfJ,+ VkfJ,]n0m (18)
Krm = k ( G / 4 / ( y , E,P„/Ut £ / „ / , ) (25)
The equations of motion of large structures undergoing small J is a coefficient function of the number of segments used in
deformations can now be readily obtained using Kane's equa- the modelling [9]. Using a similar analysis K, is given by
tions together with the kinematics previously defined. That is,
K,=KBiS0I (26)
fr+f} = 0 (19)
K'Bj may be expressed as a matrix whose elements are
where/,- are elements of the generalized active force array, and
fj are elements of the generalized inertia force. Using the K<Bi = i*(ElAl/„, 3E,7V«. 3£,A,/„) (27)
definitions of both terms in equation (19) provided in In equations (23) and (24) G, is the shear modulus, Et is the
references [6, 7], we can find an explicit form of the equations elastic modulus, /,• is the segment length, P is the central
as second moments of area relative to the X^ axes, and Aj is the
cross-sectional area.
di[-m+bi[pi,[Xp + Cifpf,l + dl[n.r,!yiJ =
-fr-F(X,X,Xt) Ill Constraint Equations
the function F(X, X, X, t) provides the equations of the rigid Restrictions placed on the movement of the structures as on
body motion. They are in a form similar to equation (1). The its geometry produce constraints equations which need to be
generalized forces/,- could be expressed as in equation (5). The solved simultaneously with the dynamical equations. For ex-
coefficients a', b', C', d' are arrays function of Vklm, Vk- , ample, if there are closed loops in the structure there occur a
o)kf , mk, Ikm„, and emst. To make the analysis less computa- holonomic constraint equations of the form

80/Vol. 110, MARCH 1988 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 0,(^1.^2. • • • ,XN,t)=0 Y, = t„zr (38)
(1=1,2, . . . ,m) (28) (/=1,2, . . . ,6N
differentiating equation (12) yields a linear relation in the Xt r=l,2, • • • J>)
and may be expressed in the matrix form: or
*«-*/ = 0 Y=fz (39)
i = l , 2 , . . . ,m (29)
where tlr are elements of the matrix associated with f. The
where the \p,•_, are functions of Xl and t. The constraint equa- reduced equations of motion can be obtained by premultiply-
tions given by equation (29) must be satisfied when solving the ing equation (1) by tlr. To avoid mixings between the rigid and
governing dynamical equations of motion. flexible motion coordinate description it is helpful to treat the
right hand side of equation (21) separately. Then the con-
Reduction of the Equations of Motion. In reference [4] straint can be applied to the rigid body motion alone.
Kamman and Huston outlined a procedue based upon a zero However, if there are constraints to be imposed on the flex-
eigenvalue theorem which reduces the governing dynamical ibility effects the appropriate orthogonal complement array is
equations from N equations to N—m equations. A brief then used to reduce the left hand side of equation (18). In any
outline of the method is as follows: Let equation (29) to be case, the form of the governing equations as presented are
written as suitable for computer implementation. The finite segment
modelling provides explicit expressions needed to describe the
^ =0 (30) dynamics of flexible bodies.
where y is a column vector equal to X and denotes the
generalized speeds. \[/ is a matrix of order m x 6N. Next, a
matrix S is formed as
S=W (31) IV Applications
where \pT is the transpose of *//. Since S is symmetric, there ex- A computer program based upon the ideas presented herein
ists an orthogonal matrix has been written. A numerical example is developed which in-
corporates the constraint equations and test the flexibility ef-
a r , such that fects in the structure. Specifically consider a windshield wiper
TrST=A (32) mechanism as shown in Fig. 3. This example is presented by
Haug [10]. The model consists of 6 bodies. The crank arm
where A is an 6N x 6N diagonal matrix with real eigenvalues (body 2) drives the link (body 3) and connecting link (body 5)
elements A,-. From [4], it is shown that those eigenvalues are are modeled as rigid bodies. In the right wiper arm (body 4),
non-negative, with at least 6N—m of them being zero. Next, the link connecting bar and wiper arm are modelled as rigid
let T be rearranged so that the zero eigenvalues will appear at and flexible respectively. The left wiper arm (body 6), the link
the end of the diagonal matrix A. Define a matrix D such that connecting bar and wiper are modeled as rigid and flexible,
respectively.
D = i>T (33)
The chassis of the vehicle is body 1. The joint modelling is
Equation (33) then shows that DTD is equal to A. Therefore D shown in Fig. 3. The friction force on the arm tip is modeled
could be partitioned into two matrices as a linearly decreasing force with increasing tip velocity, as
shown in Fig. 4. A constant crank arm speed of 60 rpm is
"D given and the movement of the tip of the wiper for the case of
rigid and flexible are presented in Fig. 5 and Fig. 6.

6 V Conclusion

the order of D is n x p where 6N—m < p < 6N. From the or- A method used in the study of the dynamics of large con-
thogonality of T, and noting that 7T7" is the identity matrix (n strained flexible structures undergoing large motions is
x n) the constraint equations may be expressed in the form presented. The method is based upon the use of Kane's equa-
tions as developed by Huston et al. The constraint equations
xlsy = tTTry^DZ =0 (35) are used to describe the closed loops formed on the structure.
An orthogonal complement array derived from the eigen-
where Z is T^y. By inverting T, y may be expressed as
values of the zero eigenvalue of the equation By = Xy is used
y=TZ Zisa67Vxl vector (36) to modify the governing equations of motion. The new set of
the reduced equations of motion along with the constraint
T may be partitioned as equations form the general equations of motion describing the
structural dynamics. The uncoupling of the generalized forces
" 0 0"
make the equations tailored for the incorporation of the flex-
ibility effects. The global stiffness matrix used to account for
the elastic deformation of the bodies is based upon the local
0 f deformations associated with equivalent beam deflections.
From structural analysis those stiffness matrices are for-
where the first zero matrix 6 is of order of 6N—p and T is an mulated by expressing them with respect to the reference
6N x p matrix whose columns are those associated with the frame R. In the application; results are presented for both
zero eigenvalues of S. Substituting into equation (36) we can rigid and flexible constrained wiper mechanism.
accommodate the constraint equation (16). Walton and The advantages to the method presented stems from the use
Steeves [5] provide a description of the zero-eigenvalue of Kane's equations, finite segment modelling and orthogonal
theorem. The solution to equation (27) can be expressed in complement arrays. The equations of motion are presented in
terms of p independent elements of z as a form suitable to control algorithms as in robotics where the

Journal of Dynamic Systems, Measurement, and Control MARCH 1988, Vol. 110/81

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 differential equations terms of the elastic coordinates are used 6 Huston, R. L., and Passarello, C. E., "On Multi-Rigid Body System
Dynamics," Computer and Structures, Vol. 10, 1979, pp. 439-446.
to adjust the body position. 7 Amirouche, F. M. L., "Flexibility Effects—Estimation of the Stiffness
Matrix in the Dynamics of Large Structure," ASME Journal of Vibration,
Stress, Acoustics and Reliability in Design, June, 1987.
References 8 Huston, R. L., and Passarello, C. E., "On the Dynamics of Chain
Systems," ASME Paper No. F4-WA/AM. 11, 1974.
1 Boland, P., Samin, J. C , and Wiltons, P. Y., "Stability Analysis of In- 9 Syriac, Joseph, "Flexibility Effects in Large Structures Dynamics," M.S.
terconnected Deformable Bodies with Closed-Loop Configuration," AIAA Jr., thesis, Department of Mechanical Engineering, University of Illinois at
Vol. 13, 1975, pp. 864-867. Chicago, Fall 1985.
2 Wehage, R. A., and Haug, E. J., "Generalized Coordinate Partitioning 10 Yoo, W. S., and Huag, E. J., "Dynamics of Flexible Mechanical Systems
for Dimension Reduction in Analysis of Constrained Dynamic Systems," Using Vibration and Static Correction Modes," ASME Design Engineering
ASME Journal of Mechanical Design, Vol. 104, Jan. 1982, pp. 247-255. Technical Conference, Cincinnati, Ohio, Sept. 10-13, 1985.
3 Sunada, W., and Dubowsky, S., "The Application of Finite Element 11 Singh, C. L., and Likins, P, W., "Singular Value Decomposition for Con-
Methods to the Dynamic Analysis of Flexible Spatial Coplanar Linkage strained Dynamical Systems," ASME Journal of Applied Mechanics, Vol. 52,
Systems," ASME Journal of Mechanical Design, Vol. 103, No. 3, July 1981, 1985, pp. 843-948.
pp. 643-651. 12 Mani.N. K., "Use of Singular Value Decomposition for Analysis and Op-
4 Kamman, J. W., and Huston, R. L., "Constrained Multibody System timization of Mechanical Systems Dynamics," Ph.D. thesis, The University of
Dynamics—An Automated Approach," J. Computers and Structures, Vol. 18, Iowa, 1984.
No. 6, 1984, pp. 999-1003. 13 Hollerback, J. M., " A Recursive Lagrangian Formulation of Manipulator
5 Walton, W. C , Jr., and Steeves, E. C , "A New Matrix Theorem and Its Dynamics and a Comparative Study of Dynamics Formulation Complexity,"
Applications for Establishing Independent Coordinates for Complex Dynamical IEEE Transactions, Systems, Mass and Cybernetics, Vol. SMC 10, 1980, pp.
Systems with Constraints," NASA Tech. Rep. TRR-326, 1969. 730-736.

FRICTION FORCE IN)

TIP VELOCITY
IM/SI

RJ7

BODY I - CHASSIS RVI - REVOLUTE JOINT 183.52,7.84,2.81

BODY 2 - CRANK LINK SJ2 - SPHERICAL JOINT (40.4,3.8.1.51

BODY 3 - DRIVE LINK UJ3 - UNIVERSAL JOINT I-7.5,2.3.4.71

BODY 4 - RIGHT WIPER ARM RJ4 - REVOLUTE JOINT I -2.85,5.B,7.41

BODY S - CONNECTING LINK SJ5 - SPHERICAL JOINT 11.5.10.5,8.41

BODY 6 - LEFT WIPER ARM UJ8 - UNIVERSAL JOINT 188.4,14.2.2.81

RJ7 - REVOLUTE JOINT 183.52,7.B4,2.91

Fig. 3 Windshield wiper mechanism Fig. 4 Friction force versus tip velocity

0.0 0.2 0.4 0.6 0.8 1.0

Time (sec)
Fig. 5

82/Vol. 110, MARCH 1988 Transactions of the ASME

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 2.5

u
u 0 -
^2'
E
L
01 5 -
r
CD

1.0 -
a

0.5

41
>
0.0
0.0 0.2 0.4 0.6 0.8 1 .0

Time (sec)
Fig. 6

14 Hooker, W. W., "Equations of Motion for Interconnected Rigid and 20 Chace, M. A., "Analysis of the Time Dependence of Multifreedom
Elastic Bodies," Celestial Mechanics, Vol. 11, 1975, pp. 357-359. Mechanical Systems in Relative Coordinates," ASME Journal of Engineering
15 Hemami, H., and Weiner, F. C , "Modelling of Nonholonomic Dynamic for Industry, Vol. 89, 1967, pp. 119-125.
Systems with Applications," ASME Journal of Applied Mechanics, Vol. 48,
21 Kane, T. R., "Dynamics of Nonholonomic Systems," ASME Journal of
1981, p. 177.
Applied Mechanics, Vol. 28, 1961, pp. 574-578.
16 Kane, T. R., and Levinson, D. A., "Formulation of Equations of Motion
for Complex Spacecraft," Journal of Guidance and Control, Vol. 3, 1980, pp. 22 Amirouche, M. L., "Dynamic Analysis of Tree-Like Structures Undergo-
99-112. ing Large Motions—A Finite Segment Approach," J. of Engineering Analysis
17 Kane, T. R., Likins, P. W., and Levinson, D. A., Spacecraft Dynamics, and Computational Mechanics, Vol. 3, No. 2, pp. 111-117.
McGraw-Hill, New York, 1982. 23 Orlandea, N., Chace, M. A., and Calahan, D. A., " A Sparsity Oriented
18 Huston, R. L., and Passarello, C. E., "On Lagrange's Form of Approach to Dynamic Analysis and Design of Mechanical Systems; Part I / I I , "
d'Alembert's Principle," Matrix and Tensor Quarterly, Vol. 23, 1973, pp. ASME Journal of Engineering for Industry, Vol. 99, 1977, p. 773.
109-112.
19 Jerkovsky, W., "The Transformation Operator Approach to 24 Huston, R. L., and Kamman, J. W., " A Discussion on Constraint Equa-
Multisystems Dynamics, Part I: The General Approach," Matrix and Tensor tions in Multibody Dynamics," Mechanics Research Communications, Vol. 9,
Quarterly, Vol. 27, 1976, pp. 48-59. 1982, p. 251.

Journal of Dynamic Systems, Measurement, and Control MARCH 1988, Vol. 110/83

Downloaded From: http://dynamicsystems.asmedigitalcollection.asme.org/ on 01/13/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use