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

net/publication/227304781

Discrete Time Transfer Matrix Method for Multibody System Dynamics

Article  in  Multibody System Dynamics · January 2005

DOI: 10.1007/s11044-005-5006-1

CITATIONS READS

81 819

5 authors, including:

Bin He
Nanjing Tech University
37 PUBLICATIONS   322 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

wind turbine blade dynamics View project

All content following this page was uploaded by Bin He on 27 February 2017.

The user has requested enhancement of the downloaded file.

Multibody System Dynamics (2005) 14: 317–344

C Springer 2005

Discrete Time Transfer Matrix Method

for Multibody System Dynamics

XIAOTING RUI, BIN HE, YUQI LU, WENGUANG LU and GUOPING WANG
Institute of Power Engineering, Nanjing University of Science and Technology, Nanjing 210094,
P.R. China

(Received: 19 April 2004; accepted in revised form: 18 March 2005)

Abstract. A new method for multibody system dynamics is proposed in this paper. This method,
named as discrete time transfer matrix method of multibody system (MS-DT-TMM), combines and
expands the advantages of the transfer matrix method (TMM), transfer matrix method of vibration of
multibody system (MS-TMM), discrete time transfer matrix method (DT-TMM) and the numerical
integration procedure. It does not need the global dynamics equations for the study of multibody
system dynamics. It has the modeling flexibility and a small size of matrices, and can be applied to a
wide range of problems including multi-rigid-body system dynamics and multi-flexible-body system
dynamics. This method is simple, straightforward, practical, and provides a powerful tool for the study
of multibody system dynamics. Formulations of the method as well as some numerical examples of
multi-rigid-body system dynamics and multi-flexible-body system dynamics to validate the method
are given.
Keywords: multibody system, dynamics, discrete time, transfer matrix method

1. Introduction
Lots of methods for multibody systems dynamics have been studied by many authors
on theory and computational method [1, 2]. In general, almost all methods of
the multibody system dynamics have the same characteristics as follows: (1) It is
necessary to develop global dynamics equations of the system. And if the system
structure is changed, generally, corresponding global dynamics equations must
be deduced again. (2) The orders of involved system matrices increase with the
increase of the number of the degrees of freedom of the system; hence the orders
of matrices involved in global dynamics equations are rather high for complex
multibody system.
The transfer matrix method (TMM) has been developed for a long time and
has been used widely in structure mechanics and rotor dynamics of linear time
invariant system. To linear system, Holzer (1921) initially applied TMM to solve
the problems of torsion vibrations of rods [3], Myklestad (1945) applied TMM to
determine the bending-torsion modes of beams [4], Thomson (1950) applied TMM
to more general vibration problems [5], Pestel (1963) listed transfer matrices for
elasto-mechanical elements up to 12th-order [6], Rubin (1964, 1967) provided a
general treatment for transfer matrices and their relation to other forms of frequency
318 X. RUI ET AL.

response matrices [7, 8]. Transfer matrices have been applied to a wide variety of
engineering programs by a number of researchers, including Targoff [9], Lin [10],
Mercer [11], Lin [12], Mead [13, 14], Henderson [15], McDanie [16, 17] and Murthy
[18–20], dealing with beams, beam-type periodic structures, skin-stringer panels,
rib-skin structures, curved muti-span structures, cylindrical shells, stiffened rings
etc. Dokanish (1972) developed finite element-transfer matrix method to solve the
problems of plate structure vibration analysis, by combining finite element method
and transfer matrix method [21]. Many researchers, such as, Ohga (1987), Xue
(1994) and Loewy (1985, 1999), studied and improved the finite element transfer
matrix for structure dynamics [22–25]. Horner (1978) proposed Riccati transfer
matrix method in order to circumvent the numerical stability of the boundary value
problem [26]. Up to the present, Riccati transform also is an important tool to
overcome the ill condition of the transfer matrix method.
Rui and Lu (1989, 2000) developed transfer matrix method of multibody system
(MS-TMM) for vibrations analysis of linear multibody system by developing new
transfer matrices and orthogonal property of multibody system [27–29]. Kumar
and Sankar (1986) developed discrete time transfer matrix method (DT-TMM) for
structure dynamics of time variant system by combining the transfer matrix method
with the numerical integration procedure [30]. DT-TMM gives an important clue
for dynamics of time variant system. MS-TMM provides an important thought that
multibody system dynamics may be solved using transfer matrix method.
In this paper, a new analytical method of multibody system dynamics, namely
discrete time transfer matrix method of multibody system (MS-DT-TMM) is de-
veloped. This method combines and expands the TMM, DT-TMM, MS-TMM and
the numerical integration procedure. When using this method, the global dynamics
equations of the system are not needed and the orders of involved system matrices
are decreased greatly. It can be applied to a wide range of problems of general
multibody system dynamics. This method is simple, straightforward, practical, and
provides a powerful tool for the multibody system dynamics. Using the new method,
dynamics of multi-rigid-body system and dynamics of multi-flexible-body systems,
including chain multibody system, branched multibody system, close-looped multi-
body system, network multibody system, have been discussed in detail. And cor-
responding numerical examples have been given. The simulating results obtained
by MS-DT-TMM and by ordinary dynamic method have a good agreement.
This paper is organized as follows. In Section 2, the sign convention and the steps
of the method are shown. In Section 3, the transfer matrices of typical elements
are developed, including rigid body moving in space, rigid body moving in plane,
elastic hinge, damper hinges, smooth ball-and-socket hinge, and smooth pin hinge.
In Section 4, the algorithm for dynamical analysis of multibody system is presented.
In Section 5, the numerical results of multibody dynamics got by MS-DT-TMM
and by ordinary methods are given, to validate the method. In Section 6, some
advantages of the method are shown. The conclusion and future works are presented
in Section 7.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 319

2. General Theorems and Steps of the Method

2.1. COORDINATE SYSTEM AND SIGN CONVENTION

We shall use the right-handed inertial Cartesian coordinate system ox yz as the

reference system, use x, y and z as the position coordinates of involved points,
use space-three-angles 1–2–3, which was defined in the reference [31], as the
orientation angles of involved bodies, that is, by rotating about x, y and z axis
directions successively in the same coordinate system, then we get θ1 , θ2 and θ3
[31].
Use the following sign convention: Positive position coordinates and orientation
angles coincide with positive directions of the coordinate system. Inboard forces
and outboard torques acting on the elements are positive (negative) if their vectors
are in the positive (negative) directions, outboard forces and inboard torques acting
on the elements are positive (negative) if their vectors are in the negative (positive)
directions. The first subscripts and the second subscripts denote the body indices
and the hinge indices of state vectors respectively.

2.2. DISCRETIZATION OF MULTIBODY SYSTEMS

According to the natural attribute of bodies, a complex multibody system may be

divided into a certain number of subsystems, which can be represented by various
elements including bodies (rigid bodies, elastic bodies, lumped masses etc.) and
hinges (joints, ball-and-socket, pins, linear springs, rotary springs, linear dampers
and rotary dampers, etc.).

2.3. DYNAMICS EQUATIONS OF ELEMENTS

It should be pointed out that the positions of the bodies and hinges are considered
equivalent in transfer equations and transfer matrices in this method. So, the dy-
namics equations of every body and hinge should be developed relatively to the
inertial reference system defined in Section 2.1 respectively.

2.4. LINEARIZATION OF DYNAMICS EQUATIONS OF ELEMENTS

According to numerical integration procedures, the motion parameters of multibody

system ξ̈ and ξ̇ at the time instant ti are expressed as the linear function of ξ in
form

ξ̈(ti ) = χ1 ξ(ti ) + χ2,ξ (1)

ξ̇(ti ) = χ3 ξ(ti ) + χ4,ξ (2)

where, the variable ξ may represent column matrix of the positions coordinates
x, y, z or the orientation angles θ1 , θ2 , θ3 respectively; ξ̈ and ξ̇ represent one order
320 X. RUI ET AL.

and two order derivative of ξ with respect to time, that is, corresponding column
matrices of acceleration and velocity, or corresponding column matrices of angular
acceleration and angular velocity for planar motion, at the same time instant ti .
The quantities χ1 , χ2,ξ , χ3 and χ4,ξ , will have different expressions for different
numerical integration procedures. There are many accurate and commonly avail-
able numerical integration procedures can be chosen, for example, if Newmark- β
method [32, 33] is used, then we obtained

1
χ1 = Ik,
βT 2

  
1 1
χ2,ξ =− ξ(ti−1 ) + ξ̇(ti−1 )T + − β ξ̈(ti−1 )T 2
(3)
βT 2 2
χ3 = γ χ1 T, χ4,ξ = γ χ2,ξ T + ξ̇(ti−1 ) + (1 − γ )ξ̈(ti−1 )T (4)

Where, time step T = ti − ti−1 , β and γ are the coefficients of Newmark-β

method. Bold capital symbol I k is the unit matrix, its subscripts k denotes the order
of the unit matrix and equals to three for a system moving in space or equals to two
for a system moving in plane.
Using the orientation angles θ1 , θ2 and θ3 got by rotating about x, y and z
successively defined in Section 2.1, the direction cosine matrix can be expressed
as follows
   
c3 −s3 0 c 2 0 s2 1 0 0
A =  s3 c3 0   0 1 0   0 c1 −s1 
0 0 1 −s2 0 c2 0 s1 c1
 
c2 c3 s1 s2 c3 − c1 s3 c1 s2 c3 + s1 s3
=  c2 s3 s1 s2 s3 + c1 c3 c1 s2 s3 − s1 c3  (5)
−s2 s1 c2 c1 c2

where

ci = cos θi , si = sin θi , i = 1, 2, 3, (6)

Using Taylor expansion theorem, at time instant ti direction cosine matrix A(ti ) can
be approximately expressed with respect to ti−1 by the truncated Taylor series of
order 3, that is

A(ti ) = A(ti−1 )T̃ 1 (ti−1 )θ1 (ti ) + A(ti−1 )T̃ 2 (ti−1 )θ2 (ti ) + A(ti−1 )T̃ 3 (ti−1 )θ3 (ti )
+ Φ(ti−1 ) + Bo(T 2 ) (7)
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 321

where, T̃ is the cross product matrix of T ;

Φ(ti−1 ) = A(ti−1 ) + A(ti−1 ) −T̃ 1 (ti−1 )θ1 (ti−1 ) − T̃ 2 (ti−1 )θ2 (ti−1 )

T 2 2 T 2 2
− T̃ 3 (ti−1 )θ3 (ti−1 ) + T̃ 1 (ti−1 )θ̇12 (ti−1 ) + T̃ 2 (ti−1 )θ̇22 (ti−1 )
2 2
T 2 2
+ T̃ 3 (ti−1 )θ̇32 (ti−1 ) + T̃ 2 (ti−1 )T̃ 1 (ti−1 )θ̇2 (ti−1 )θ̇1 (ti−1 )T 2
2
+T̃ 3 (ti−1 )T̃ 2 (ti−1 )θ̇3 (ti−1 )θ̇2 (ti−1 )T 2

+T̃ 3 (ti−1 )T̃ 1 (ti−1 )θ̇3 (ti−1 )θ̇1 (ti−1 )T 2

 
1 1 1
B =  1 1 1  , T 1 = [1, 0, 0]T , T 2 = [0, c1 , −s1 ]T ,
1 1 1
T 3 = [−s2 , s1 c2 , c1 c2 ]T (8)

Equation (7) will be used in the expressions of position coordinates of corre-

sponding points Equations (21) and (25). In the items of product of the trigonometric
functions and the elements of state vectors, the trigonometric functions at time ti
are expanded with respect to ti−1 using the truncated Taylor series of order 3, that is

sin θ (ti ) = sin[θ (ti−1 ) + θ ] = s̄ + o(T 2 )

cos θ (ti ) = cos[θ (ti−1 ) + θ ] = c̄ + o(T 2 )

where

 1
s̄ = sin θ (ti−1 ) 1 − [θ̇(ti−1 )T ] 2
2


1
+ cos θ (ti−1 ) θ̇ (ti−1 )T + θ̈(ti−1 )T 2
(9)
2

 1
c̄ = cos θ (ti−1 ) 1 − [θ̇ (ti−1 )T ]2
2


1
− sin θ(ti−1 ) θ̇ (ti−1 )T + θ̈(ti−1 )T 2
(10)
2

It can be proved that the multinomial in the dynamics equations can be approximated
by

[a(ti ) − a(ti−1 )][b(ti ) − b(ti−1 )] = ȧ(ti−1 )ḃ(ti−1 )T 2 + o(T 2 ) (11)

322 X. RUI ET AL.

The following equations can be obtained by Equations (2) and (11),

 2      2  
θ̇1 θ1 θ̈1 o(T 2 )
 2
 θ̇2  = κ1 χ3  θ2  + κ3  +  θ̈22  T +  o(T 2 ) 
2

θ̇32 θ3 t θ̈3 t
2
o(T 2 )
i i−1
          (12)
θ̇1 θ̇2 θ1 θ̈1 θ̈2 o(T 2 )
 θ̇2 θ̇3  = κ2 χ3  θ2  + κ3  +  θ̈2 θ̈3  T 2 +  o(T 2 ) 
θ̇1 θ̇3 θ3 t θ̈1 θ̈3 t o(T 2 )
i i−1

where
   
θ̇1 0 0 θ̇2 θ̇1 0
1
κ1 = 2  0 θ̇2 0, κ2 =  0 θ̇3 θ̇2  , κ3 = χ4,θ − θ̇(ti−1 )
0 0 θ̇3 t θ̇3 0 θ̇1 t 2
i−1 i−1

The truncation errors caused in Equations (7) and (9)–(11) are all o(T 2 ). If
necessary, it is not difficult at all to get higher computational accuracy by multi-
steps numerical methods, such as, Houbolt method, predictor-corrector methods
[32, 33] etc.
The motion quantities z(ti−1 ), ż(ti−1 ), z̈(ti−1 ) at the previous time instant are all
known at time instant ti . Thus, these quantities χ1 , χ2,ξ , χ3 , χ4,ξ , c̄ and s̄ etc. are
all definable for any subsystem for the time interval (ti − ti−1 ), and hence above
formulations are valid.

2.5. STATE VECTORS , TRANSFER EQUATIONS AND TRANSFER MATRICES

OF ELEMENTS

In order to describe conveniently, the chain multibody system is taken as an example

in the following. According to the dynamics equations of elements and structure of
multibody systems (such as closed-loop system, branched system, chain system,
network system, etc.), corresponding state vectors, transfer equations and trans-
fer matrices of the bodies and hinges can be developed. The state vectors of the
connection point among any rigid bodies and hinges moving in space are defined
as

z = [x, y, z, θ1 , θ2 , θ3 , m x , m y , m z , qx , q y , qz , 1]T (13)

where x, y, z, θ1 , θ2 and θ3 are the position coordinates of the connection point

with respect to the inertial reference system and the corresponding orientation
angles rotating in the directions of x, y, z, successively defined in Section 2.1;
m x , m y , m z , qx , q y , qz are the corresponding internal torques and internal forces in
the same reference system respectively.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 323

For a chain system moving in plane, the state vectors of the connection point
among any rigid bodies and hinges are defined as

z = [x, y, θ3 , m z , qx , q y , 1]T (14)

where, the meaning of the elements in the state vectors are similar to the meaning
described above.
Then the dynamics equations of the jth element that have been linearized using
numerical integration procedure can be assembled into a single transfer equation

z j, j+1 (ti ) = U j (ti )z j, j−1 (ti ) (15)

The meaning of the subscripts of the state vectors z follows the convention in
Section 2.1. The transfer equation describes the mutual relationship between the
state vectors at two ends of the jth element and has the similar form in contrast with
a general transfer equation used in MS-TMM, DT-TMM or TMM. Here; the matrix
U j (ti ) is the transfer matrix of the jth element. It is the functions of the motion
quantities (z(tk ), ż(tk ) and z̈(tk ), k = i − 1, i − 2, . . .) which are all known at
time instant ti . Its order always is (13 × 13) for dynamics of chain multi-rigid-body
system moving in space, or (7 × 7) for dynamics of chain multi-rigid-body system
moving in plane.

2.6. TRANSFER EQUATION AND TRANSFER MATRIX OF OVERALL SYSTEM

By using the same method used in MS-TMM, DT-TMM or TMM, the overall
system transfer equation and transfer matrix U, which relates the state vectors at
ends of the system, can be assembled and calculated. That is,

z n,n+1 = U z 1,0 (16)

U = U n U n−1 . . . U 2U 1 (17)

For a chain system, the order of the overall transfer matrix of the system is equal
to the order of the transfer matrix of the element, and it does not increase when
the degrees of freedom of system increase. Irrespective of the size of a multibody
system, the highest order of the overall transfer matrix U is the same with the
order of transfer matrix of single body, that is (13 × 13) for dynamics of chain
multi-rigid-body system moving in space, or (7 × 7) for dynamics of chain multi-
rigid-body system moving in plane. So, the matrices involved in the MS-DT-TMM
are always small, which greatly reduces the computational time and the memory
storage requirement.
324 X. RUI ET AL.

2.7. SOLUTIONS OF SYSTEM MOTION

Once the overall transfer matrix of the multibody system is known, the bound-
ary conditions of the system can then be applied and the unknown quantities in
the boundary state vectors can be computed. Now, knowing the boundary state
vectors completely, the state vectors and hence the motion quantities at each el-
ement at time ti can be computed by the repeated use of corresponding transfer
equations of element similar to Equation (15). The velocity, angular velocity, ac-
celeration and angular acceleration quantities at time ti are then obtained using
Equations (1) and (2) respectively. Then entire procedure can be repeated for time
ti+1 and so on. It can be seen clearly from Equations (16) and (17) that the global
dynamics equations of multibody system are not needed if using MS-DT-TMM
to solve the problems of multibody system dynamics. The overall transfer ma-
trix of multibody systems can be assembled easily just using the transfer matrices
of elements. So this method simplifies solving procedure of multibody system
dynamics.

3. Transfer Matrices of Typical Elements

3.1. RIGID BODY MOVING IN SPACE

As shown in Figure 1, points I , O and C denote the inboard end, outboard end
and mass center of the rigid body respectively; the subscript 2 denotes the body-
fixed coordinate system whose initial point O2 is on the inboard end I of the rigid
body, ox yz is the inertial coordinate system. So geometrical equations can been
obtained

r O = r I + Ar 2,O (18)

Figure 1. Rigid body moving in space.

DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 325

where, r K and r 2,K are the column matrices of the position coordinate of point K
with respect to the inertial coordinate system and the body-fixed coordinate system
respectively, K can be used to represent I , O and C; A is direction cosine matrix
that has been given in Equation (5).
Using Newton–Euler method and considering the symbol convention in
Section 2.1, the dynamics equations of the rigid body can be gained

m r̈ C = q I − q O + f C (19)
Ġ I = −m I + m O − I  I + mC + IC
Oq O − m ICa  fC (20)

where, m and r̈ C are the mass and the column matrix of mass center acceleration
of the rigid body, q I and q O are the column matrices of internal forces acted on
the point I and O respectively, f C and mC are the column matrices of external
force and the external torque acted on the mass center of the rigid body; G I is the
column matrix of moment of momentum with respect to point I , m I and m O are
the column matrices of internal torques acted on the points I and O, I O and IC
are the column matrices of position vectors from I to O and to C respectively, a I
is the column matrix of acceleration of point I .
To substitute Equation (7) into (18), we can obtain

r O = [I 3 , Ψ O , O 3×3 , O 3×3 , Φ(ti−1 ) r 2,O ]z I (21)

where

Ψ O = A(ti−1 )[T̃ 1 (ti−1 )r 2,O , T̃ 2 (ti−1 )r 2,O , T̃ 3 (ti−1 ) r 2,O ] (22)

O 3×3 is zero square matrix whose order equal to three; Φ(ti−1 ) is defined in
Equation (8).
Because the orientation angle of any point of the same rigid body is uniform
and can be expressed as

θ O = [O 3×3 , I 3 , O 3×3 , O 3×3 , O 3×1 ]z I (23)

where, θ is defined as

θ = [θ1 , θ2 , θ3 ]T

To substitute Equation (1) into (19), we can obtain

mχ1r C (ti ) + mχ2,r C = q I − q O + f C (24)

326 X. RUI ET AL.

To replace subscript O with C, Equation (21) becomes

r C = r I + ΨC θ I + Φ(ti−1 ) r 2,C (25)

To substitute Equation (25) into (24), we can obtain

q O = [−mχ1 , −mχ1 ΨC , O 3×3 , I 3 , U 4,5 ]z I (26)

where

U 4,5 = f C − mχ1 Φ(ti−1 )r 2,C − mχ2,r C (27)

The bold capital letters U i, j denote submatrix of the transfer matrix U , the first
subscript and the second subscript of U i, j denote the sequence numbers of row and
column of U i, j in the block matrix of the transfer matrix respectively.
Similar to the process from Equations (19) to (26), substituting Equations (1),
(2), (9), (10), (13) and (26) into Equation (20), we can obtain a matrix equation
 
m O = (r˜ I C − r˜ I O )mχ1 , κ4 − m r˜ I O χ1 ΨC , I 3 , r˜ I O , U 3,5 z I (28)

where

r I O = Ār 2,O , r I C = Ār 2,C ,

 
U 3,5 = κ5 − mC + r˜ I C mχ2,r I − f C + r˜ I O U 4,5
 
κ4 = ĀJ Hχ1 + (H 1 κ1 + H 2 κ2 )χ3 , H = T̄ 1 , T̄ 2 , T̄ 3 ,
 
H1 = Ā T̄˜ 1 J T̄ 1 , T̄˜ 2 J T̄ 2 , T̄˜ 3 J T̄ 3 ,
 T
κ5 = ĀJ Hχ2,θ + (H 1 κ1 + H 2 κ2 )κ3 + H 1 θ̈12 , θ̈22 , θ̈32 ti −1 T 2
 T
+ H 2 θ̈1 θ̈2 , θ̈2 θ̈3 , θ̈3 θ̈1 ti −1 T 2

H2 = Ā J T̄ 12 + T̄˜ 1 J T̄ 2 + T̄˜ 2 J T̄ 1 , J T̄ 23 + T̄˜ 2 J T̄ 3

+ T̄˜ J T̄ , J T̄ + T̄˜ J T̄ + T̄˜ J T̄ ,
3 2 31 3 1 1 3

T 12 = [0, −s1 , −c1 ] ,T

T 23 = [−c2 , −s1 s2 , −c1 s2 ]T , (29)
T 31 = [0, c1 c2 , −s1 c2 ]T

J is the inertia matrix, the symbol T̄ is the matrix got from matrix T by replaced
trigonometric functions s and c using s̄ and c̄ which defined in Equation (9) and
(10) respectively. For example, from A expressed in Equation (7), we can obtain
Ā, that is,
 
c̄2 c̄3 s̄1 s̄2 c̄3 − c̄1 s̄3 c̄1 s̄2 c̄3 + s̄1 s̄3
Ā =  c̄2 s̄3 s̄1 s̄2 s̄3 + c̄1 c̄3 c̄1 s̄2 s̄3 − s̄1 c̄3 
−s̄2 s̄1 c̄2 c̄1 c̄2
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 327

Equations (21), (23), (26) and (28) can be rewritten as the transfer equation

zO = U zI (30)

and the transfer matrix of the rigid body moving in space

 
I3 ΨO O 3×3 O 3×3 Φ(ti−1 )r 2,O
 O 3×3 I3 O 3×3 O 3×3 O 3×1 
 

U =  (r˜ I C − r˜ I O )mχ1 κ4 − m r˜ I O χ1 ΨC I 3 r˜ I O U 3,5  (31)

 −mχ1 −mχ1 ΨC O 3×3 I 3 U 4,5 
O 1×3 O 1×3 O 1×3 O 1×3 1

where, z O and z I are the state vector of the inboard end I and outboard end O on
the rigid body, the order of the transfer matrix of the rigid body moving in space is
(13 × 13).

3.2. RIGID BODY MOVING IN PLANE

With similar method used in Section 3.1, we can obtain the transfer matrix of a
rigid body moving in plane. And the order of the transfer matrix of a rigid body
moving in plane is (7 × 7). There are the elements of the transfer matrix

u h,h = 1 (h = 1, 2, . . . , 7), u 1,3 = −x2,O s − y2,O c,

u 1,7 = x2,O G 1 − y2,O G 2 , u 2,3 = x2,O c − y2,O s,
u 2,7 = x2,O G 2 + y2,O G 1 , u 4,1 = mχ1 (y I O − y I C ),
u 4,2 = mχ1 (x I C − x I O ), u 4,3 = u 6,3 x I O − u 5,3 y I O + JI χ1 ,
u 4,5 = −y I O , u 4,6 = x I O ,
u 4,7 = −m C + u 6,7 x I O − u 5,7 y I O + JI χ2,θ + [mχ2,yI − f yC ]x I C
(32)
+ [ f xC − mχ2,x I ]y I C ,
u 5,1 = −mχ1 , u 5,3 = mχ1 (x2,C s + y2,C c),
u 5,7 = f xC − mχ1 (x2,C G 1 − y2,C G 2 ) − mχ2,xC ,
u 6,2 = −mχ1 , u 6,3 = −mχ1 (x2,C c − y2,C s),
u 6,7 = f yC − mχ1 (x2,C G 2 + y2,C G 2 ) − mχ2,yC ,
G 1 = c + θs − c(θ̇ T )2 /2, G 2 = s − θc − s(θ̇T )2 /2,

Other elements are all zero.

where

x I C = x2,C c̄ − y2,C s̄, y I C = x2,C s̄ + y2,C c̄,

x I O = x2,O c̄ − y2,O s̄, y I O = x2,O s̄ + y2,O c̄
328 X. RUI ET AL.

the minuscule symbol u i, j denotes the element of the transfer matrix, the first
subscript and the second subscript of u i, j denote the sequence numbers of row and
column of u i, j in the transfer matrix respectively; angles and angular velocities
and trigonometric functions in Equation (32) all have the values of these parameter
at time ti−1 .

3.3. ELASTIC HINGE AND DAMPER HINGE

Here, the elastic hinge and damper hinge including linear spring hinge, rotary spring
hinge, linear damper hinge and rotary damper hinge that are parallel connected each
other with elastic hinge. And the mass of the spring and dampers are zero.
According to the equilibrium of forces and torques at point O or I , the following
equations can be obtained,

q O + k(r O − r I ) + c(ṙ O − ṙ I ) = 0 (33)

−m O + k (θ O − θ I ) + c (AO ω O − AI ω I ) = 0 (34)

where
   
kx 0 0 k x 0 0
k=0 ky 0 , k
=  0 k y 0  ,
0 0 kz 0 0 k z
    
cx 0 0 cx 0 0
c=0 cy 0 , c
=  0 cy 0 
0 0 cz 0 0 cz

k x , k y , k z and k x , k y , k z are the stiffness coefficients of linear springs and rotary

springs, cx , c y , cz and cx , cy , cz are damper coefficients of linear dampers and
rotary dampers, ω O and ω I are angular velocity matrices of points O and I
expressed in the body fixed coordinate system, AO and AI are direction cosine
matrix of the body fixed coordinate system of points O and I with respect to the
inertia reference system.
Considering that the mass of the hinge is zero, and the forces (torques) acted on
the two ends of the hinge are the same. Substituting Equations (2), (5), (9) and (10)
into Equation (33), then the transfer matrix of the hinge moving in space can be
obtained, that is, there are the submatrices and the elements of this transfer matrix

U h,h = I 3 (h = 1, 2, 3, 4), U 1,4 = −(k + cχ3 )−1 ,

 
U 1,5 = −U 1,4 c χ4,r I − χ4,r O ,
(35)
U 2,3 = (k
+ c
AO H O χ3 )−1 , U 2,2 = U 2,3 (k
+ c
AI H I χ3 ),
 
U 2,5 = U 2,3 c
AI H I χ4,θ I − AO H O χ4,θ O , u 13,13 = 1,

Other elements are all zero.

DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 329

3.4. SMOOTH BALL - AND - SOCKET HINGE

For a smooth ball-and-socket hinge, its mass and size have been neglected, the
position coordinates and internal forces are equal and internal torques equal to zero
in its two ends.

3.4.1. Smooth Ball-and-Socket Hinge Whose Outboard Hinge Is Also a Smooth

Ball-and-Socket Hinge
For a smooth ball-and-socket hinge j whose outboard hinge is also a smooth ball-
and-socket hinge, according to the internal torques are all zero in two ends of its
outboard body, using the transfer equation of the outboard body j + 1

z j+1, j+2 = U j+1 z j+1, j , (36)

the relationship between the orientation angles in two ends of the hinge can be
developed

θ O = [U 2,1 , O 3×3 , O 3×3 , U 2,4 , U 2,5 ]z I , (37)

where

−1 −1 −1
U 2,1 = −Ū 3,2Ū 3,1 , U 2,4 = −Ū 3,2Ū 3,4 , U 2,5 = −Ū 3,2Ū 3,5 (38)

Ū k, j is the submatrix of the transfer matrix Ū of its outboard body j + 1.

So the transfer equation and transfer matrix of the smooth ball-and-socket hinge
can be obtained as follows

zO = U zI (39)
 
I3 O 3×3 O 3×3 O 3×3 O 3×1
 −Ū −1 Ū −1
−Ū 3,2Ū 3,4 −Ū 3,2Ū 3,5 
−1
 3,2 3,1 O 3×3 O 3×3 
U =
 O 3×3 O 3×3 I3 O 3×3 O 3×1   (40)
 O 3×3 O 3×3 O 3×3 I3 O 3×1 
O 1×3 O 1×3 O 1×3 O 1×3 1

3.4.2. Smooth Ball-and-Socket Hinge Whose Outboard Hinge Is a Fictitious Hinge

For a smooth ball-and-socket hinge whose outboard hinge is a fictitious hinge (that
is, this end is free), the internal torques all equal to zero in two ends of its outboard
body too. So the transfer equation and transfer matrix of this kind of smooth ball-
and-socket hinge are the same as Equations (39) and (40).
330 X. RUI ET AL.

3.4.3. Smooth Ball-and-Socket Hinge Whose Outboard Hinge Is Neither a Smooth

Ball-and-Socket Hinge Nor a Fictitious Hinge
For a smooth ball-and-socket hinge j whose outboard hinges j + 1, j +
2, . . . , j + r − 1 are neither smooth ball-and-socket hinge nor fictitious hinge
until smooth ball-and-socket hinge or fictitious hinge j + r , the transfer matrix
can be obtained from Equation (40) by replacing Ū j+1 → Ū j, j+r , where
Ū j, j+r denote the submatrices of the transfer matrix of the subsystem described
the transfer relationship between the state vectors z j, j and z j j+r −1, j+r , that is
Ū j, j+r = U j+r −1U j+r −2 · · · U j+2U j+1 .

3.5. SMOOTH PIN HINGE

For a multibody system moving in plane, the transfer matrix of the smooth pin hinge
can be obtained using similar method to Section 3.4. In fact, the transfer matrix of
the smooth pin hinge can be obtained directly from the transfer matrix of smooth
ball-and-socket hinge as the special case of space motion.

4. Algorithm for Dynamics Analysis

Following the formulations given above, the motion quantities of a multibody sys-
tem at different time instants for different subsystems can now be obtained as
follows:

1. Decide the initial conditions and system properties of the multibody system.
2. Set i = 1.
3. Knowing the initial conditions r (ti−1 ), ṙ (ti−1 ), r̈ (ti−1 ), θ(ti−1 ), θ̇(ti−1 ), θ̈(ti−1 )
etc. and the system properties at time ti , calculate the quantities χ1 , χ2 , χ3 , χ4 ,
c̄ and s̄ etc. for each subsystem.
4. Formulate the transfer matrix for each subsystem and the overall transfer matrix
respectively.
5. Apply boundary conditions to the end state vectors of the system and calculate
the unknown quantities in the boundary state vectors as a function of the elements
of the overall transfer matrix.
6. Now, knowing all the elements in the boundary state vector, the motion quantities
at each subsystem at time instant ti can be computed by successive multiplication
of the transfer matrices, using Equation (15).
7. By using the computed values of the position coordinates r (ti ) and orienta-
tion angles θ(ti ) at ti , compute the values of ṙ (ti ), r̈ (ti ), θ̇(ti ), θ̈(ti ) etc. using
Equations (1) and (2).
8. Let i = i + 1, use the computed values of the last step as the initial conditions,
and return to step 3, until the time required for complete analysis.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 331

Figure 2. Motion of rigid double pendulum.

5. Applications and Numerical Examples

5.1. DYNAMICS OF MULTI - RIGID - BODY SYSTEM MOVING IN PLANE

Numerical example 1: Consider a rigid double pendulum moving in plane under

the effect of gravity. The initial conditions and salient parameters of the system are
listed in Figure 2(a).
The motion of the system is first computed by developing the global dynamics
equations of the system using Newton–Euler method and integrating the equations
using Runge–Kutta (R–K) method and Schiehlen method [34] respectively. Then
the same system is simulated by the proposed MS-DT-TMM procedure. The sim-
ulation results are shown in Figure 2(b). For these cases, the same integration time
step is taken. As can be seen from Figure 2(b), the motion of the system obtained by
MS-DT-TMM is almost identical to the motion got by other methods of multibody
dynamics.

5.2. DYNAMICS OF MULTI - FLEXIBLE - BODY SYSTEM MOVING IN PLANE

Numerical example 2: Compute the large planar motion of a flexible pendulum

under the effect of the gravity. The initial parameters of the system are listed in
Figure 3(a).
The dynamical equations of the beam system can be developed with an ordinary
method multi-flexible-body system. However, for making comparison between the
ordinary dynamics method and MS-DT-TMM, the flexible beam is divided into
twenty same parts represented by rigid bodies connected with rotary spring hinges
successively. The stiffness of the rotary spring hinges and the mass of the rigid
bodies can be decided by the method described in reference [35, 36]. The large
motion of the beam system is first computed by developing the global dynamics
equations of the system with Newton-Euler method and solved with R–K method.
Then the large motion is computed with MS-DT-TMM. The large angle motion
332 X. RUI ET AL.

Figure 3. Motion of the flexible pendulum.

of the beam end got by the two methods, as shown in Figure 2(b), have a good
agreement. These confirm the suitability of MS-DT-TMM for the dynamics of
multi-flexible-body system.
It can be seen from this example and Figure 3(b) that:

1. The order of the system matrix of corresponding global dynamics equations

is forty. But the order of corresponding overall transfer matrix of the system
is only seven. And the computational time required decreases greatly if using
MS-DT-TMM, see Figure 10.
2. With time step properly chosen, we have not found problems in the computa-
tional stability using MS-DT-TMM for the large system in the long time history.
3. The MS-DT-TMM is valid for the large system including twenty bodies.

Numerical example 3: Compute the response of the cantilever beam moving in

plane under the effect of external force A sin t in the free end. The initial conditions
and salient parameters of the system are listed in Figure 4(a).
The motion of the beam is studied using the similar model to that used in
Numerical example 2. Then the motion of the beam system is computed with MS-
DT-TMM and with analytical method respectively. The time history of the deflection
response of the cantilever beam end got by MS-DT-TMM is almost identical to that
got by analytical method, as shown in Figure 4(b).

5.3. DYNAMICS OF MULTI - RIGID - BODY SYSTEM MOVING IN SPACE

Numerical example 4: Compute the motion of a space pendulum moving in space

under the effect of the gravity. The initial parameters of the system are listed in
Figure 5(a).
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 333

Figure 4. Response of the cantilever beam.

Figure 5. Motion of the space pendulum.

Time history of the orientation angles of the space pendulum is computed by

MS-DT-TMM procedure and by using R–K method and integrating the dynamics
equations of the system respectively. The results of the system motion got by MS-
DT-TMM are almost identical to that got by Newton-Euler method, as shown in
Figure 5(b).
Numerical example 5: Compute the system motion of rigid three pendulums
moving in space under the effect of the gravity, as shown in Figure 6(a). In which,
three same rigid bodies connected by three smooth ball-and-socket hinges succes-
sively, and the parameters of every rigid body is the same with the parameters of
the rigid body of the example 5, the initial orientated angles of the three bodies and
the initial angular velocities of body 2 and body 4 are all zero, the initial angular
velocities of body 6 is [0, 0.1, 0]T rad/s.
334 X. RUI ET AL.

Figure 6. Motion of the space three pendulum.

For making comparison between the ordinary dynamics method and MS-DT-
TMM, the procedure to study the multibody system dynamics using two methods
is given in detail as follows.
If using ordinary methods of multibody dynamics, for example, Newton-Euler
method, the global dynamics equations of the system can be obtained as follows
    
E 22 E 24 E 26 ω̇ 2 F2
 E 42 E 44 E 46   ω̇ 4  =  F 4  (41)
E 62 E 64 E 66 ω̇ 6 F6

where

E 66 = J, E 44 = J − m r˜2,O r˜2,O , E 22 = J − 2m r˜2,O r˜2,O ,

E 24 = −r˜2,O A2T A4 m(r˜2,C + r˜2,O ), E 26 = −m r˜2,O A2T A6r˜2,C ,
E 42 = −m(r˜2,O + r˜2,C )A4 A2T r˜2,O , E 46 = −m r˜2,O A4T A6r˜2,C ,
E 62 = −m r˜2,C A6T A2r˜2,O , E 64 = −m r˜2C A6T A4r˜2,O ,
F 2 = −ω̃ 2 Jω 2 − mg[r˜2,C + 2r˜2,O ]A2T [0 1 0]T − r˜2,O A2T [2m A2 ω̃ 2 ω̃ 2r 2,O
+A4 ω̃ 4 ω̃ 4 m(r 2,O + r 2,C ) + A6 ω̃ 6 ω̃ 6 mr 2,C ],
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 335

F 4 = −ω̃ 4 J ω̃ 4 − mg[r˜2,C + r˜2,O ]A4T [0 1 0]T − r˜2,O A4T m[A2 ω̃ 2 ω̃ 2r 2,O

+A4 ω̃ 4 ω̃ 4r 2,O + A6 ω̃ 6 ω̃ 6r 2,C ] − m r˜2,C A4T A2 ω̃ 2 ω̃ 2r 2,O ,
F 6 = −ω̃ 6 Jω 6 − mgr˜2,C A6T [0 1 0]T
−m r˜2,C A6T [A2 ω̃ 2 ω̃ 2r 2,O + A4 ω̃ 4 ω̃ 4r 2,O ] (42)

g is the gravity acceleration. Because the structure parameters of three bodies are
the same, here we do not select the different denoted symbol. The order of the
system matrix of the system is eighteen in ordinary method.
Using MS-DT-TMM and considering the symbol convention in Section 2.1, we
can obtain the overall transfer equation of the system

Z 6,7 = U Z 2,1 (43)

transfer matrix of the system

U = U 6U 5U 4U 3U 2 (44)

transfer equations of the elements

Z 2,3 = U 2 Z 2,1 , Z 4,3 = U 3 Z 2,3 Z 4,5 = U 4 Z 4,3 Z 6,5 = U 5 Z 4,5

Z 6,7 = U 6 Z 6,5 (45)

and boundary conditions of the system

z 2,1 = [0, 0, 0, θ1 , θ2 , θ3 , 0, 0, 0, qx , q y , qz , 1]T

(46)
z 6,7 = [x, y, z, θ1 , θ2 , θ3 , 0, 0, 0, 0, 0, 0, 1]T

where, U 2 , U 4 , U 6 and U 3 , U 5 are the transfer matrices of corresponding rigid bod-

ies 2, 4, 6 and smooth ball-and-socket hinges 3, 5 described in Section 3 respectively.
The order of the transfer matrix of the system is thirteen in the MS-DT-TMM.
The system motion (orientated angles of three space pendulums) can be got by
direct integrating the dynamics equations of the system using R–K method and
by solving the transfer equations of the system using MS-DT-TMM procedure
respectively. The results obtained by two methods are almost identical, as shown
in Figure 6(b–d).

5.4. BRANCHED SYSTEMS , NETWORK SYSTEMS , CLOSED - LOOP SYSTEMS

TMM and MS-TMM are applicable not only for chain systems, but also for branched
systems, network systems, closed-loop systems etc. [6, 29]. In the MS-DT-TMM,
various complex multibody systems can be modeled in the same way as in the
336 X. RUI ET AL.

Figure 7. Closed-loop system.

TMM and MS-TMM. Showing the applicability of the proposed method for general
multibody systems, some explanatory examples about branched system, network
system, closed-loop system, have been discussed in this section. A detailed study
of these is yet to be undertaken.

5.4.1. Dynamics of Closed-Loop System

For a closed-loop system moving in space or in plane, as shown in Figure 7, the
boundary conditions are not so obvious. In the same way as in the TMM, we
“cut” and start at any arbitrary point of the closed system, where the state vector is
z l,1 . Then the closed-loop system can be considered as a chain system, proceeding
around the system until we return to the original point from which we started. Then
the overall transfer equations

Z l,1 = U Z l,1

or

U all Z l,1 = 0 (47)

and the transfer matrix of the system

U all = U − I (48)

are obtained, where

U = U l U l−1 . . . U n−1U n−2 . . . U 2U 1 (49)

The motion of the closed-loop system can be simulated by solving the transfer
equations of the system as same as chain system.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 337

Figure 8. Branched system.

5.4.2. Dynamics of Branched System

A branched system moving in space or in plane, can be divided into three branches
I, II and III respectively, each branch can be considered as a chain subsystem and
is connected at branched point j by hinge, as shown in Figure 8.
There are transfer equations for these branches

z j−1, j = U I z 2,1 , z n,n+1 = U II z j+1, j , z l+1, j = U III z n+3,n+2 (50)

where U I , U II and U III are the transfer matrix of the chain subsystem I, II and III.
If the system moves in a plane and the hinge on the branched point j is
smooth pin hinges, because the displacement is continuing and the forces are
equilibrium as well as the torques are zero in the branched point j, we can
obtain

z j−1, j (1) = z j+1, j (1) = z l+1,l (1),

z j−1, j (2) = z j+1, j (2) = z l+1,l (2),
z j−1, j (5) − z j+1, j (5) + z l+1,l (5) = 0, (51)
z j−1, j (6) − z j+1, j (6) + z l+1,l (6) = 0,
z j−1, j (4) = z j+1, j (4) = z l+1,l (4) = 0

where, symbol z(i) means the element of corresponding state vector, the num-
ber i in brackets is the sequence number of the element in corresponding state
vector.
338 X. RUI ET AL.

Substitute Equation (50) into (51), we can obtain

U b1U I z 2,1 = U b1U −1

II z n,n+1 − U b1 U III z n+3,n+2 ,
U b2U I z 2,1 = U b2U −1
II z n,n+1 = U b2 U III z n+3,n+2 , U b3 U I z 2,1 = 0,
U b3U −1
II z n,n+1 = 0, U b3U III z n+3,n+2 = 0 (52)
U b1 = [O 2×4 , I 2 , O 2×1 ] , U b2 = [I 2 ,O 2×5 ],
U b3 = [O 1×3 , 1, O 1×3 ] (53)

So the overall transfer equation can be obtained

 T T
U all z 2,1 T
z n,n+1 T
z n+3,n+2 =0 (54)

where
 
U b1U I −U b1U −1
II U b1U III
 U b2U I −U b2U −1 O 2×7 
 II 
 U b2U I O 2×7 −U b2U III 
U all =
 U b3U I
 (55)
 O 1×7 O 1×7  
 O 1×7 U b3U −1 O 1×7 
II
O 1×7 O 1×7 U b3U III

The motion of the branched system can be simulated by using boundary condi-
tions and solving the transfer equations of the system as same as chain system.
5.4.3. Dynamics of the Network System
A network system as shown in Figure 9, can be divided into two subsystems I, and II
respectively, each subsystem can be considered as a chain system and is connected
at network point j by hinge.
There are transfer equations for these subsystem I and II

z j−1, j = U I z 2,1 , z n, j = U II z j+1, j (56)

where, U I and U II are the transfer matrix of the chain subsystem I and II.

U I = U j−1U j−2 . . . U 3U 2 , U II = U n U n−1 . . . U j+1 (57)

If the system moves in a plane and the hinge on the network point j is smooth pin
hinges, because the displacement is continuing and the forces are equilibrium as
well as the torques are zero in the network point j, we can obtain

U b1U I z 2,1 = U b1 z j+1, j − U b1U II z j+1, j , U b3U I z 2,1 = 0,

(58)
U b2U I z 2,1 = U b2 z j+1, j = U b2U II z j+1, j , U b3U II z j+1, j = 0
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 339

Figure 9. Network system.

where, U b1 , U b2 , and U b3 are defined by Equation (53).

So the overall transfer equation can be obtained
 T T
U all z 2,1 , z Tj+1, j = 0 (59)

where
 
U b1U I U b1 (I 7 − U II )
 U b2U I −U b2 
 
U all =
 O 2×7 U b2 (I 7 − U II ) 
 (60)
 U b3U I O 1×7 
O 1×7 U b3U II

The motion of network system can be simulated by using boundary conditions

of the system and solving the transfer equations of the system as same as chain
system.

5.5. COMPUTATIONAL TIME , MEMORY STORAGE AND ACCURACY

Consider the motion of a chain multibody system including rigid bodies connected
with rotary spring hinges, the initial parameters of the system are as follows: the
number of the rigid bodies is n, the length and mass of each rigid body are (1/n) m
and (1/n) kg respectively, the stiffness of each rotary spring hinge is 20N · m/rad,
the initial angles and initial angular velocities of the bodies are all (−π/3) and zero.
The computational time required of the system dynamics using MS-DT-TMM
and ordinary method is shown in Figure 10. And the CPU of the used microcomputer
is an AMD-1.2 GHz. It can be seen clearly from Figure 10 that the computational
time increase very slowly when the number of the degrees of freedom increases
340 X. RUI ET AL.

Figure 10. Computational time vary with number of the degrees of freedom.

if using MS-DT-TMM, and the computational time increases very quickly when
the number of the degrees of freedom increases if using ordinary methods, the
MS-DT-TMM has a higher computational speed.
Of course, now, the building of the global dynamics equations of multibody
systems can also be relatively simple by using some effective methods; and by
using the sparse matrix techniques, the computer time is direct proportion to the
number of the bodies for chain system so that computational speed can also be
increased greatly [37].
There is no general method of exactly estimating the accumulated errors involved
in the computed solution, so far in the authors’ experience with MS-DT-TMM in
solving problems of varied multibody system sizes, with time step properly chosen,
the truncation error accumulation was not significant. On the other hand, the round
off error accumulation is a function of the word length of computing system as well
as the number of elements to be modeled and is very significant for MS-DT-TMM.
It can be seen from numerical examples shown in Section 5.2 that MS-DT-TMM
is also valid for large system. The simulation results of the motion of the large
system got by MS-DT-TMM and by ordinary dynamics methods always have a
good agreement in over thirty periods. In fact, MS-DT-TMM is also valid for
very large system if the Riccati transform is introduced. In authors’ experience in
computing the motion of the system including one hundred thousand rigid bodies
connected by elastic hinges, it is difficult to obtain correct results if only MS-DT-
TMM is used. But by using MS-DT-TMM and Riccati transform, we have not found
difficulty in the computational stability and accuracy for this very large system. The
order of the overall transfer matrix of the very large system is only seven, and the
order of the system matrix will be two hundred thousand if using ordinary method,
which greatly reduces the computational time and the memory storage requirement.
A detailed study of these errors on the accuracy of MS-DT-TMM solutions is yet
to be undertaken and will be discussed in detail in another paper.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 341

6. Highlights of the Method

1. The method avoids global dynamic equations of multibody system and simplifies
solving procedure of multibody system dynamics.
2. Irrespective of the size of a multibody system, the matrices involved in the MS-
DT-TMM are always small, which greatly increases the computational speed.
3. The method avoids the computing difficulties caused by too high matrix orders
of complex multibody system.
4. In contrast with the conventional TMM and MS-TMM, the matrices involved
in the MS-DT-TMM are always real, even when the damping is included. This
simplifies numerical computation algorithms of multibody system dynamics.
5. Unlike the TMM and MS-TMM, which are restricted to small vibration of system
alone, MS-DT-TMM is capable of analyzing linear time invariant, linear time
variant, nonlinear, multi-rigid-body and multi-flexible-body systems.
6. The method provides flexibility in modeling multibody systems with varying
configuration. That is, by creating a library of transfer matrices for commonly
occurring elements, such as rigid bodies moving in plane, rigid bodies moving
in space, elastic beams moving in plane, elastic beams moving in space, rotary
spring hinges, linear spring hinges, smooth ball-and-socket hinges, smooth pin
hinges, linear damper hinges and rotary damper hinges, etc., and by assembling
these at the required locations, various configurations can be modeled easily.
7. Any suitable numerical integration scheme [32, 33] can be included in this
method, thus providing researchers with flexibility in the computation of multi-
body systems dynamics.

7. Conclusions
A new method, named as discrete time transfer matrix method of multibody system,
is introduced for the study of multibody systems dynamics. The formulation of the
method as well as some numerical examples of multi-rigid-body system dynamics
and multi-flexible-body system dynamics to validate the method are given. Several
possible areas of applications are identified. The proposed method is based on the
TMM, DT-TMM, MS-TMM and the numerical solution procedures of differen-
tial equations. It combines the advantages of these methods. It has the modeling
flexibility, smaller core size requirement of TMM, DT-TMM, MS-TMM and the ex-
tended applicability of numerical integration procedures. It avoids global dynamics
equations of multibody system so that simplifies solution procedure of multibody
system dynamics. This method is simple, straightforward and practical for the sim-
ulation of multibody system dynamics. By developing transfer matrices of flexible
body elements, such as, elastic beam and shell elements etc. this method would be
a more power tool for dynamics of multi-flexible-body system with flexible body
elements; this is one subject that we are studying.
342 X. RUI ET AL.

It should be noted that the propagation of round off error in TMM, and the in-
fluence of time step selection on computational accuracy in numerical integration
procedures will also be important for the proposed method. The numerical difficul-
ties will arise when analyzing very large systems on small computing systems in
using transfer matrices. The accumulation of round off error, which is a function of
the word length of computing system as well as the number of elements to be mod-
eled, is very significant. Sometimes the multiplication of the transfer matrices of the
different elements of the system lead to very large numbers for the larger system,
when several very large numbers are multiplied and several very small numbers are
multiplied successively in nature sequence. Because the word length of comput-
ing system is limited, when computing the multiplication of the transfer matrices,
the solution process probably eventually collapses, if the absolute values of these
numbers are too large or too small. To avoid the problem of numerical difficulties
arising in the multiplication of the transfer matrices, there are some methods can
be chosen, such as the Delta-matrix method [6], Modified transfer-matrix method
[6, 38] etc. One of effective methods to avoid the problem of numerical difficulties
is to adjust the computing sequence of these numbers, so that large absolute value is
multiplied by small absolute value successively when computing the multiplication.
Riccati transfer matrix method [26] provides an effective approach for numerical
stability of the boundary value problem. Riccati transform can be used not only to
decrease the orders of matrices involved in the MS-DT-TMM, but also to overcome
the problem of numerical difficulties, see Section 5.5.

Acknowledgements
The research was supported by German Academic Exchange Service (DAAD). We
owe special thanks to Professor Werner Schiehlen who helped the first author’s
work in Germany and gave us some revelatory suggestions and valuable research
material. Thanks for things done by Doctor Laifeng Yun and Wei Zhang in this
paper.

References
1. Schiehlen, W., ‘Multibody system dynamics: Roots and perspectives’, Multibody System
Dynamics 1, 1997, 149–188.
2. Shabana, A.A., ‘Flexible multibody dynamics: Review of past and recent developments’,
Multibody System Dynamics 1, 1997, 189–222.
3. Holzer, H., Die Berechnung der Drehsenwingungen, Springer, Berlin, 1921.
4. Myklestad, N.O., ‘New method of calculating natural modes of coupled bending-torsion vibra-
tion of beams’, Transaction of The ASME 67, 1945, 61–67.
5. Thomson, W.T., ‘Matrix solution of vibration of nonuniform beams’, Journal of Applied
Mechanics 17, 1950, 337–339.
DISCRETE TIME TRANSFER MATRIX METHOD FOR MULTIBODY SYSTEM DYNAMICS 343

6. Pestel, E.C. and Leckie, F.A., Matrix Method in Elastomechanics, McGraw-Hil1, New York,
1963.
7. Rubin, S., ‘Transmission matrices for vibrators and their relation to admittance and impedance’,
Journal of Engineering Materials and Technology 86, 1964, 9–21.
8. Rubin, S., ‘Review of mechanical immittance and transmission matrix concepts’, Journal of the
Acoustical Society of America 41, 1967, 1171–1179.
9. Targoff, W.P., The associated matrices of bending and coupled bending-torsion vibrations,
Journal of Aeronautics Science 14, 1947, 579–582.
10. Lin, Y.K., Probabilistic Theory of Structure Dynamics, McGraw-Hill, New York, l967.
11. Mercer, C.A. and Seavey, C., ‘Prediction of natural frequencies and norma1 modes of skin-
stringer panel rows’, Journal of Sound and Vibration 6, 1967, 149–162.
12. Lin, Y.K. and McDaniel, T.J., ‘Dynamics of beam-type periodic structures’, Journal of
Engineering Materials and Technology 91, 1969, 1133–1141.
13. Mead, D.J. and Gupta, G.S., Propagation of Flexural Waves in Infinite, Damped Rib-Skin
Structures, United States Air Force Report, AFML-TR-70-l3, l970.
14. Mead, D.J., ‘Vibration response and wave propagation in periodic structures’, Journal of Engi-
neering Materials and Technology 93, 1971, 783–792.
15. Henderson, J.P. and McDaniel, T.J., ‘The analysis of curved multi-span structures’, Journal of
Sound and Vibration l8, 1971, 203–219.
16. McDaniel, T.J., ‘Dynamics of circular periodic structures’, Journal of Aircraft 8, 1971, l43–149.
17. McDaniel, T.J. and Logan, J.D., ‘Dynamics of cylindrical she1ls with variable curvature’, Journal
of Sound and Vibration 19, 1971, 39–48.
18. Murthy, V.R. and Nigam, N.C., ‘Dynamics characteristics of stiffened rings by transfer matrix
approach’, Journal of Sound and Vibration 39, 1975, 237–245.
19. Murthy, V.R. and McDaniel, T.J., ‘Solution bounds to structural systems’, AIAA Journal 14,
1976, 111–113.
20. McDaniel, T.J. and Murthy, V.R., ‘Solution bounds for varying geometry beams’, Journal of
Sound and Vibration 44, 1976, 431–448.
21. Dokanish, M.A., ‘A new approach for plate vibration: Combination of transfer matrix and finite
element technique’, Journal of Mechanical Design 94, 1972, 526–530.
22. Ohga, M. and Shigematus, T., ‘Transient analysis of plates by a combined finite element transfer
matrix method’, Computers & Structures 26, 1987, 543–549.
23. Xue, H., ‘A combined dynamic finite element riccati transfer matrix method for solving non-
linear eigenproblems of vibrations’, Computers & Structures 53, 1994, 1257–1261.
24. Loewy, R.G., Degen, E.E. and Shephard, M.S., ‘Combined finite element-transfer matrix method
based on a mixed formulation’, Computers & Structures 20, 1985, 173–180.
25. Loewy, R.G. and Bhntani, N., ‘Combined finite element-transfer matrix method’, Journal of
Sound and Vibration 226(5), 1999, 1048–1052.
26. Horner, G.C. and Pilkey, W.D., ‘The riccati transfer matrix method’, Journal of Mechanical
Design 1, 1978, 297–302.
27. Rui, X.T. and Lu, Y.Q., ‘Transfer matrix method of vibration of multibody system’, Chinese
Journal of Astronautics 16(3), 1995, 41–47.
28. Rui, X.T., Sui, W.H. and Shao, Y.Z., ‘Transfer matrix of rigid body and its application in
multibody dynamics’, Chinese Journal of Astronautics 14(4), 1993, 82–87.
29. Lu, Y.Q. and Rui, X.T., ‘Eigenvalue problem, orthogonal property and response of multibody
system’, in J.H. Zhang and X.N. Zhang (eds.), ICAPV 2000, Proceedings of International
Conference on Advanced Problems in Vibration Theory and Applications, Science Press,
Beijing, 2000.
30. Kumar, A.S. and Sankar, T.S., ‘A new transfer matrix method for response analysis of large
dynamic systems’, Computers & Structures 23, 1986, 545–552.
 344 X. RUI ET AL.

31. Kane, T.R., Likine, P.W. and Levinson, D.A., Spacecraft Dynamics, McGraw-Hill, New York,
1983.
32. Dokainish, M.A. and Subbaraj, K., ‘A study of direct time-integration methods in computational
structural dynamics-I. Explicit methods’, Computers & Structures 32(6), 1989, 1371–1386.
33. Subbaraj, K. and Dokainish, M.A., ‘A study of direct time-integration methods in computational
structural dynamics-II. Implicit methods’, Computers & Structures 32(6), 1989, 1387–1401.
34. Kübler, R. and Schiehlen, W., ‘Two methods of simulator coupling’, Mathematical and Computer
Modelling of Dynamical Systems 6, 2000, 93–113.
35. Banerjee, A.K. and Nagarajan, S., ‘Efficient simulation of large overall motion of Beams under-
going large deflection’, Multibody System Dynamics 1, 1997, 113–126.
36. Wang, Y. and Huston, R.L., ‘A lumped parameter method in the nonlinear analysis of flexible
multibody system’, Computers & Structures 50(3), 1994, 421–432.
37. Valasek, M., On the Efficient Implementation of Multibody System Formulations, Technical
Report institutsbericht, IB-17, University Stuttgart, 1990.
38. Li, Y., ‘Modified algorithm of transfer matrix method of eigenvalue problem’, Chinese Journal
of Aerospace Power 6(3), 1991, 249–250.

View publication stats