Académique Documents
Professionnel Documents
Culture Documents
R
e
e
c
c
u
u
r
r
D
D
y
y
n
n
/
/
S
S
o
o
l
l
v
v
e
e
r
r
T
T
h
h
e
e
o
o
r
r
e
e
t
t
i
i
c
c
a
a
l
l
M
M
a
a
n
n
u
u
a
a
l
l
Copyright 2005 FunctionBay, Inc. All rights reserved
This document may not be copied, disclosed, or modified without the prior
written consent of FunctionBay, Inc.
Information described in this document is furnished for general information
only, is subject to change without notice, and should not be constructed as a
warranty or inaccuracies that may appear in this manual.
The software described in this document is provided under written license only,
contains valuable trade secrets and proprietary information, and is protected by
the copyright laws of the United States and Other countries. UNAUTHORIZED
USE OF SOFTWARE OR ITS DOCUMENTATION CAN RESULT IN CIVIL DAMAGES AND
CRIMINAL PROSECUTION.
Edition Note
This theoretical manual documents the theoretical background of the
RecurDyn / Solver.
Trademarks of FunctionBay, Inc.
RecurDyn is a registered trademark of FunctionBay, Inc.
RecurDyn/Professional, RecurDyn/SOLVER, RecurDyn/SOLID,
RecurDyn/FLEX, RecurDyn/NodalFlex, RecurDyn/LINEAR,
RecurDyn/CONTROL, RecurDyn/TRACK_HM,
RecurDyn/TRACK_LM, RecurDyn/CHAIN, RecurDyn/MTT2D,
RecurDyn/MTT3D, RecurDyn/BELT, RecurDyn/HAT,
RecurDyn/Hydraulic, RecurDyn/Gear, RecurDyn/Hydraulic,
RecurDyn/Tire are trademarks of FunctionBay, Inc.
Revision History
First printed, April 2001
1
st
Revision, January 2002
2
nd
Revision, July 2002
3
rd
Revision, August 2002
4
th
Revision, September 2003
5
th
Revision, September 2005
RecurDyn
=
33 32 31
23 22 21
13 12 11
a a a
a a a
a a a
| (2-1)
where , , and h are unit vectors along the f g x , y , and axes,
respectively. The
z
z y x frame is the body reference frame and the
frame is the inertial reference frame. Z Y X
Z
X
Y
rp
r
p
x
y
z
s
o
Fig. 2.1 Coordinate systems and a rigid body
Velocities and virtual displacements of point in the X O Z Y frame are
defined as
(2-2a)
(
w
r&
(
r
(2-2b)
Their corresponding quantities in the z y x frame are defined as
RecurDyn
=
w A
r A
w
r
Y
T
T
& &
(2-3b)
(
T
T
A
r A r
1.2.2 RELATIVE KINEMATICS FOR A PAIR OF CONTIGUOUS BODIES
A pair of contiguous bodies is shown in Fig. 2.2. Body 1) (i is assumed
to be an inboard body of body and the position of point is i
i
O
1) i(i 1)i (i 1)i (i 1) (i i
+ + = s d s r r (2-4)
The angular velocity of body in its local reference frame, using Eq. 2-3a and
defining , is
i
i
T
1) (i 1)i (i
A A A
=
1)i (i 1)i (i
T
1)i (i 1) - (i
T
1)i (i i
+ = q H A w A w & (2-5)
where is determined by the axis of rotation. H
zi
X
Z
Y
r
i-1
r
i
s(i-1)i
si(i-1)
yi-1
x
i-1
zi-1
zi-1
x
i-1
x
i
y
i
x
i
z
i
y
i
yi-1
d(i-1)i
o
i o
i-1
Fig. 2.2 Kinematic relationship between two adjacent rigid bodies
1-5
Differentiation of Eq. 2-4, using Eq. 2-3a, yields
1)i (i
'
1)i (i 1) (i
'
i
'
1) i(i i
'
1) (i
'
1)i (i 1) (i
'
1) (i
'
1)i (i 1) (i
'
1) (i 1) (i
'
i i
1)i (i
) (
~
~
~
+
+
=
q d A
s A d A
s A r A r A
q
&
& &
(2-6)
where symbols with tildes denote skew symmetric matrices comprised of their
vector elements that implement the vector product operation (Ref. 1) and
denotes the relative coordinate vector. Substituting of Eq. 2-5 and
multiplying both sides of Eq. 2-6 by yields
1)i (i
q
'
i
T
i
A
1)i (i 1)i (i
1)i (i
T
1) i(i 1)i (i 1)i (i
1)i (i
T
1) (i
1)i (i
T
1) - i(i 1)i (i 1)i (i 1)i (i
1)i (i
T
1) (i
1)i (i
T
i
)
~
) ( (
)
~
~
~
(
1)i (i
+ +
+
=
q H A s A d A
A s A d s A
r A r
q
&
& &
(2-7)
where
i i i
~
= A A
&
is used. Combining Eqs. 2-5 and 2-7 yields the recursive
velocity equation for a pair of contiguous bodies.
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Y B Y
&
(2-8)
where
(
+
(
I 0
A s A d s I
A 0
0 A
B
)
~
~
~
( 1)i (i
T
1) i(i 1)i (i 1)i (i 1)i (i
1)i (i
T
1)i (i
T
1)i1 (i
(
(
+
(
1)i (i
1)i (i
1)i (i
T
1) i(i 1)i (i 1)i (i
1)i (i
T
1)i (i
T
1)i2 (i
~
) (
1)i (i
H
H A s A d
A 0
0 A
B
q
(2-9)
It is important to note that matrices and are functions of only
relative coordinates of the joint between bodies
1)i1 (i
B
1)i2 (i
B
(i 1) and . As a
consequence, further differentiation of the matrices and in Eq.
i
1)i2 1)i1 (i
B
(i
B
RecurDyn
+ = q B Z B Z
1.3. GENERALIZATION OF THE VELOCITY RECURSIVE FORMULA
1
n-1
n
0
Fig. 3.1 A serial chain mechanism
Before proceeding to generalize the recursive velocity formula, the
computational equivalence between the recursive method and the velocity
transformation method is demonstrated using the mechanical system shown in
Fig. 3.1. The Cartesian velocity is obtained by replacing by in
Eq. 2-8.
m
i m
1)m (m 1)m2 (m 1) (m (m m
+ = q B B Y & (3-1)
Substitutions of Eq. 2.8 for , , . . . , and yield Y
2) - (m
Y
0
Y
1-7
1)m - (m 1)m2 - (m
1 - m
1 j
1)j - (j 1)j2 - (j
j - m
1 k
1)1 - m m)(-k k (
m
1 k
0 1)1 - m m)(-k k ( m
q B
q B B
Y B
&
&
+
)
`
|
|
.
|
\
|
+
=
= =
+ +
=
+ +
Y
(3-2)
Thus, the Cartesian velocity for all bodies is obtained as Y
q B Y & = (3-3)
where is the collection of coefficients of and B
1)i (i
q&
| |
T
1 nc
T T
2
T
1
T
0
=
n
Y , , Y , Y , Y Y K (3-4)
| |
T
1 nr
T
) 1 (
T
12
T
01
T
0
=
n n
q , , q , q , Y q & K & & & (3-5)
where and nc nr denote the numbers of the Cartesian and relative coordinates,
respectively.
The Cartesian velocity , with a given , can be evaluated either
by using Eq. 3-3 or by using Eq. 2-8 with recursive numerical substitution of .
Since both formulas give an identical result, and recursive numerical substitution
is proven to be more efficient in Ref. 4, matrix multiplication with a given
will be evaluated by using Eq. 2-8.
nc
R Y
nr
R q&
i
Y
q B&
q&
Since in Eq. 3-3 is an arbitrary vector in , Eqs. 2-8 and 3-3, which are
computationally equivalent, are actually valid for any vector such that
q&
nr
R
nr
R x
x B X & = (3-6)
and
1)i - (i 1)i2 - (i 1) - (i 1)i1 - (i i
x B X B X + = (3-7)
where is the resulting vector of multiplication of and . As a
nc
R X B x
RecurDyn
* T
Q q =
+
*
) ! i(i
Q
( {
+
+
1 i 1)2
Q
(
T
q =
=
=
1 - n
0 i
=
+
1 - n
0 i
T
i(i
q
T
=0
1.4. GENERALIZATION OF THE FORCE RECURSIVE FORMULA
It is often necessary to transform a vector in into a new vector
in . Such a transformation can be found in generalized force
computation in the joint space with a known force in the Cartesian space. The
virtual work done by a Cartesian force is
nc
R
G B g
T
=
nr
R
= (4-1)
where must be kinematically admissible for all joints in Fig. 3.1.
Substitution of
Z
q B Z = into Eq. 4-1 yields
T
Q B W (4-2)
where . Equation 4-2 can be written in a summation form as Q B Q
T *
+
T
1) i(i
q W (4-3)
On the other hand, the symbolic substitution of the recursive virtual
displacement relationship Eq. 2-10 into Eq. 4-1, along the chain in Fig. 3.1
starting from the body n toward inboard bodies, yields
)}
)
+ +
=
1 i
T
i(i 1)
S B W (4-4)
where
2 i 2 i
2)1 1)(i (i
1 i
0
+ +
+ +
+
+ S Q B S
S
(4-5)
1-9
Equating the right sides of Eqs. 4-3 and 4-4, the following recursive formula for
is obtained:
*
Q
( ) 0 ...., 1, - n i ,
1 i 1 i
)2 1 i(i
T
1) i(i
*
= +
+ +
+ + S Q B Q (4-6)
where is defined in Eq. 4-5.
1 i+
S
Since is an arbitrary vector in , Eqs. 4-5 and 4-6 are valid for any
vector in . As a result, the matrix multiplication of is
evaluated to achieve computational efficiency by
Q
G
nc
R
nc
R G B
T
( )
( ) 0 ...., 1, i
1 i 1 i
)1 1 i(i
T
i
n
1 i 1 i
T
1)2 i(i 1) i(i
= +
=
+ =
+ +
+
+ + + +
S Q B S
0 S
S G B g
(4-7)
where is the result of . g G B
T
1.5. GRAPH REPRESENTATIONS OF MECHANICAL SYSTEMS
In the previous section, a serial chain mechanism is considered to derive
recursive formulas for and B where is a vector in and G
in . In general, a mechanical system may have various topological
structures. To cope with the various topological structures, an automatic
preprocessing is required for a general purposed program, which employs a
relative coordinate formulation. The preprocessing identifies the topological
structure of a constrained mechanical system to achieve computational efficiency.
A graph theory was used to represent bodies and joints for mechanical systems
in Refs. 1 and 4. A node and an edge in a graph represented a body and a joint,
respectively. The preprocessing based on the graph theory yielded the path and
distance matrices that are provided to automatically decide execution sequences
Bx G
T
x
nr
R
nc
R
RecurDyn
T
= +
&
( (6-1)
where Z must be kinematically admissible for all joints except cut joints [1].
In the equation, and , respectively, denote the cut joint constraint and the
corresponding Lagrange multiplier. The mass matrix and the force vector
are defined as
M
Q
(
nbd 2 1
, , , diag M M M M L = ) (6-2)
(
=
i
i
i
J 0
0 I m
M (6-3)
| |
T
nbd
T
3
T
2
T
1
, , , , Q Q Q Q Q L = (6-4)
~
(
=
J n
r m f
Q
~
i
&
(6-5)
where denotes the number of bodies, denotes the identity matrix,
denotes the moment of inertia, denotes the external force, and denotes
the external torque. Substituting the virtual displacement relationship into Eq.
6-1 yields
nbd I J
f n
{ } 0 ) Q Y M B q
T
= +
&
(
T
(6-6)
Since q is arbitrary, the following equations of motion are obtained:
0 ) Q Y M B F
T
= + =
&
( (6-7)
The equations of motion, the constraint equations, q v = & , and
constitute the following differential algebraic equations[8]:
a v = &
RecurDyn
) (
) (
) (
&
&
& &
&
t , , ,
t , ,
t ,
t , , ,
(6-8)
Application of 'tangent space method' in Ref. 7 to Eq. 6-8 yields the following
nonlinear system that must be solved at each time step:
0
a v U
v q U
a v q
v q
q
) a , v q ( F
p H =
(
(
(
(
(
(
(
(
+ +
+ +
) (
) (
) (
=
) (
) (
t , , ,
t , ,
t ,
t , , ,
) (
2 n 0 n
T
0
1 n 0 n
T
0
n n n
n n n
n n
n n n n n
n
n
& &
&
(6-9)
where | |
T T
n
T
n
T
n
T
n n
, a , v , q p =
0
U
, , , and are determined by the
coefficients of the BDF, and is an
0
nr
2
U
T
0
Applying the Newton's method to solve the nonlinear system in Eq. 6-9 yields
H p H
p
= (6-10)
1,2,3,... i ,
i
n
1 i
n
= + =
+
p p p (6-11)
where
1-13
(
(
(
(
(
(
(
(
=
0 U U 0
0 0 U U
0
0 0
0 0 0
F F F F
H
T
0
a v q
v q
q
q q q q
p
T
0 0
T
0 0
T
0
q G B
T
To derive the formulas systematically, bodies in a graph are divided into four
disjoint sets (associated with a generalized coordinate ) as follows:
k
q
( )
k
q q I having joint the of body outboard body outborad {
k
=
} coordinate d generalize its
as
( ) ( ) { }
k k
of bodies outboard all q I q II =
( )
( )
)
`
=
body inboard and base the including
, of body inboard the and body base e between th bodies all
k
k
q I
q III
RecurDyn
&
q v
F
q
q q
& &
F
q
F
1.8.1 COMPUTATION OF THE RESIDUAL F
The generalized force , and the Cartesian acceleration need to be
computed to obtain shown in Eq. 6-7. The term is obtained by applying
the recursive formula in Eq. 7.1. The recursive formula with
in Eq 4.7 can be applied to evaluate in since
is a vector in
Q
Z
T
Y
&
F
)
nc
R .
Y
&
G B
T
nr
R (
T
Q Y M G
Z
+ =
&
G
F
1-19
1.8.2 COMPUTATION OF THE JACOBIAN
q
F
In Eq. 6-7, differentiation of matrix with respect to vector results in a
three dimensional array. To avoid the complexity, Eq. 6-7 is differentiated with
respect to a typical generalized coordinate . Thus,
B q
k
q
nr ....., 2, 1, k , ) (
) (
k
k k
T T
T T
= + +
+ =
q Z
Z q q
Q Y M B
Q Y M B F
&
&
(8-1)
Since the term can be easily expressed in terms of the Cartesian
coordinates, is obtained by applying the chain rule, as
) (
T
Q
Z
k
)
T
q Z
Q (
k
T T
) ( ) (
k
B Q Q
Z Z q Z
= (8-2)
where B
q
Z
=
(
k
Y
q
+
&
B Q
Z Z
)
T
T
)
Z
G B
T
Q nc
T
( B
G
T
Z
T
)
Z
Q
G B
T
RecurDyn
4
3
2
5
6
1
R4
R2
R3
R5
S1 S2
R1
T1
7
Y
X
Z
0
.
1
6
0.5
0.2
0
.
1
0
9
45
o
Fig. 9.1 A governor mechanism
1-21
Table 9-1 Inertia properties of the governor mechanism and spring and damping
constants
Mass
x
I
y
I
z
I
xy
I
yz
I
zx
I
Body 1 (Ground) not necessary
Body 2 200.0 25.0 50.0 25.0 0.0 0.0 0.0
Body 3 1.0 0.1 0.1 0.1 0.1 0.1 0.1
Body 4 1.0 0.1 0.1 0.1 0.1 0.1 0.1
Body 5 1.0 0.15 0.125 0.15 0.0 0.0 0.0
Body 6 0.1 0.1 0.1 0.1 0.0 0.0 0.0
Body 7 0.1 0.1 0.1 0.1 0.0 0.0 0.0
Spring constant 1000
Damping constant 30
Table 9-2 Integration output information
Program TOL Average step size No. fevals CPU time (sec)
Other
5
10 3
2
10 1 . 1
748 41
Proposed
5
10 3
2
10 2 . 1
441 7
PROPOSED
OTHER
Fig. 9.2 Z acceleration of Body 4
RecurDyn
3
10 4
1359 330
Proposed
4
10
3
10 6 . 6
1167 69
OTHER
PROPOSED
Fig. 9.4 Vertical acceleration of the chassis
RecurDyn
k
k
) (
) (
1 i
1)i (i
q q
x) B ( X
& &
=
1)i (i 1)i2 (i
1 i 1)i1 (i
1 i 1)i1 (i i
k
k
k k
) (
) (
) ( ) (
+
+
=
x B
X B
X B X
q
q
q q
&
&
& &
k
k k
) (
) ( ) (
1 i 1)i1 (i
1 i 1)i1 (i i
q
q q
X B
X B X
+
=
&
& &
v v
x B X ) (
& &
=
1)i (i 1)i2 (i
1 i 1)i1 (i i
k
k k
) (
) ( ) (
+
=
x B
X B X
v
v v
&
& &
k
k k
) (
) ( ) (
1 i 1)i2 (i
1 i 1)i1 (i i
v
v v
X B
X B X
+
=
&
& &
Recursive
formulas
) ( i
k
q III ) ( i
k
q IV
q q
Bx X ) ( = 0 X
q
=
k
) (
i
0 X
q
=
k
) (
i
q q
G B g ) (
T
=
k k
k k
) ( ) (
) ( ) (
i
T
1)i1 (i 1 i
i
T
1)i2 (i 1)i (i
q q
q q
S B S
S B g
=
=
0 S
0 g
q
q
=
=
k
k
) (
) (
1 i
1)i (i
q q
x B X ) (
& &
= 0 X
q
=
k
) (
i
&
0 X
q
=
k
) (
i
&
v v
x B X ) (
& &
= 0 X
v
=
k
) (
i
&
0 X
v
=
k
) (
i
&
Recursive
formulas
) ( i
k
q I or i or or ) (
k
q II ) ( i
k
q III ) ( i
k
q IV
Bx X = 1)i (i 1)i2 (i 1 i 1)i1 (i i
+ = x B X B X
G B g
T
=
) (
0
) (
1 i 1 i
T
1 ) 1 i ( i i
n
1 i 1 i
T
2 ) 1 i ( i ) 1 i ( i
+ + +
+ + + +
+ =
=
+ =
S G B S
S
S G B g
x B X
& &
= i ) 1 i ( 2 i ) 1 i ( 1 i 1 i ) 1 i ( 1 i 1 i ) 1 i ( i
+ + = x B X B X B X
& & & &
1-27
RecurDyn
F(x)
H(x)
T
2
T
1
q
q
2
T
2
1
T
1
=
(
(
(
(
(
(
(
(
(
(
(
|
|
.
|
\
|
|
|
.
|
\
|
=
(
(
(
(
(
(
(
(
(
(
(
|
|
.
|
\
|
|
|
.
|
\
|
=
2
0
1
0
0
0
b
h
b
h
b
h
b
h
&
& &
(2.3)
where
=
=
k
1 i
1 n i
0
1
b
b
1
v ,
=
=
k
1 i
1 n i
0
2
b
b
1
q
[ x =
i
U
T
q
) 2 , 1 i (
) m n ( n
=
R
(
(
(
T
q
U
i
i
U
n
R .
The number of equations and the number of unknowns in Eq. 2.3 are the same,
so Eq. 2.3 can be solved. Newton's numerical method can be applied to obtain
the solution . x
i i i
H x H
x
= (2.4.a)
i i 1 i
x x x + =
+
(2.4.b)
LU-decomposition of the matrix not only increases the computation time
i
x
H
RecurDyn
F U
0 F U x R v = + +
1
1
1
) (
T
q a
T
i
h h h (3.2)
where is a mass matrix and is generally nonsingular. The can be
singular if a parametric formulation is employed. If is singular, Eqs. 3.1
must be solved simultaneously to obtain
a
F
a
F
q , v , a and . The vector
is a new unknown variable. The a is thus obtained from Eq. 3.2 in
terms of as
1
T 1
a 1
) (
h
1
F x R v a
q
= (3.3)
2-5
Substituting Eq. 3.3 into Eq. 3.1.a yields
) ( ) ( ) (
h
1
x R F x F v
F
F q F
a 1
T
q
a
v q
+ = + |
.
|
\
|
+ + (3.4)
Equation 3.1.f can be rewritten in an equivalent inflated form by choosing
such that
2
U
0
F
F U
T
q
a
v
= |
.
|
\
|
+
1
T
2
h
as follows:
0
F
F x R q v
T
q
a
v 2
= |
.
|
\
|
+ + +
2
1
) (
h
h h h (3.5)
where
h
+
a
v
F
F is assumed to be a nonsingular matrix and is a new
unknown variable. The solution process for the case of a singular matrix will be
explained later in this section. Equation 3.5 can be solved for in terms of
as follows:
m
2
R
v
q
2
1
2
h
) (
h
1
F
F x R q v
T
q
a
v
|
.
|
\
|
= (3.6)
Substituting Eq. 3.6 into Eq. 3.4 and multiplying both sides of Eq. 3.4 by
yields
2
h
3
2
h R q K
T
q
*
= + (3.7)
where
a v q
*
F F F K + + h h
2
(3.8.a)
2 1
(3.8.b)
) ( ) h ( h ) ( h ) ( h
2 1
2 2
3
x R F F x R F x F R
a v a
+ + + (3.8.c)
RecurDyn
=
(
(
(
) ( h 0
3
2
x
R
K
q
T
q
*
(3.9)
Equation 3.9 is then solved for and . Note that is scaled by to
make the coefficient matrix of Eq. 3.9 ill-conditioned, even as approaches to
zero.
q
2
h
h
Multiplying both sides of Eq. 3.5 by
h
+
a
v
F
F yields
) (
h h
1
h
2 2
x R q
F
F v
F
F
a
v
T
q
a
v
|
.
|
\
|
= + |
.
|
\
|
+ (3.10)
Equations 3.10 and 3.1.c are combined to obtain
(
(
(
|
.
|
\
|
+
=
(
(
(
+
q x
x R q
F
F
F
F
q
a
v
q
T
q
a
v
) (
) (
h h
1
0
h
2
2
(3.11)
where has been obtained from Eq. 3.9. Equation 3.11 is solved for the
and . Multiplying both sides of Eq. 3.3 by yields
q
2
v
a
F
|
.
|
\
|
= + v x R F a F
a a
h
1
) (
1 1
T
q
(3.12)
Equations 3.12 and 3.1.b are combined to obtain
2-7
(
(
(
|
.
|
\
|
=
(
(
(
(
q v x
v x R F
F
q v
a
q
T
q a
& & & & & &
) (
h
1
) (
0
1
1
(3.13)
Equation 3.13 is solved for a and . Once , and are obtained, the
is evaluated from Eq. 3.8.b, as follows:
1
1
2 1
+ + = (3.14)
Since is a mass matrix and is a tangent damping matrix, is
generally not ill conditioned. If an ill-conditioned case is encountered, Eqs. 3.1
must be solved simultaneously to obtain
a
F
v
F
v a
F F h +
q , v , a and . However, the
and
a
F
v
F h
a
F + are rarely singular, so q , v , a and are obtained by
using Eqs. 3.9, 3.11, 3.13, and 3.14 for most of practical problems.
2.4. NUMERICAL ALGORITHM
The DASSAL subroutine [4] is employed to integrate the system variables.
Computational flow for the proposed DAE solution method is given in Fig.
1.(Page 2-7)
2.5. NUMERICAL EXAMPLES
2.5.1 QUICK-RETURN MECHANISM
The quick-return mechanism as shown in Fig. 5.1 is mounted on a body
translating with respect to the ground. The system consists of 6 bodies, 2
translational joints, and 5 revolute joints. The system has two degrees of freedom
if the redundant constraints are eliminated.
Dynamic analyses were performed for 1 sec with error tolerances of 10
-4
and
10
-6
by using the program developed in this paper and the other commercial
program
RecurDyn
F
a
or
is singular ?
Y
Fig. 1 Flowchart for the proposed DAE solution method
2-9
Front view Side view
Body1
Body2
Body3
Body5
Body6
Body4
Fig. 5.1 A quick-return mechanism
, which employs the implicit numerical integration with the BDF. The results are
shown in Fig. 5.2. and the integration information is shown in Table 5.1.
OTHER
PROPOSED
Fig. 5.2 Results of the quick-return mechanism
RecurDyn
i
f
p
Orientation of the virtual body is obtained as follows: 1 + i
1 , 1 + +
=
i i i
f
i i
A A A A (3)
where is the relative orientation matrix induced by the rotational
deformation and is the orientation matrix between the reference frames of
the flexible body and virtual body
i
f
A
1 , + i i
A
i 1 + i
i
f
in an undeformed state. If the Bryant
angle (1-2-3) [13] is employed, the is expressed as follows: A
(
(
(
+
+
=
i
y
i
x
i
z
i
y
i
x
i
z
i
x
i
z
i
y
i
x
i
z
i
x
i
y
i
x
i
z
i
y
i
x
i
z
i
x
i
z
i
y
i
x
i
z
i
x
i
y
i
z
i
y
i
z
i
y
i
f
cos cos sin sin cos sin sin cos sin cos sin sin
cos sin sin sin sin cos cos cos sin sin sin cos
sin sin cos cos cos
A (4)
If is infinitesimal, the matrix can be approximated as |
T
i
z
i
y
i
x
i
= |
i
f
A
(
(
(
1
1
1
i
x
i
y
i
x
i
z
i
y
i
z
i
f
A (5)
The rotational deformation vector can be represented by linear
combination of rotational mode shapes of body as
i
i
i
f
i i
p
= (6)
where is a modal matrix whose columns are composed of rotational mode
i
3-7
shapes and is the vector of modal coordinate.
C
i
q
|
=
0
0
0
i
f
p
Finally, kinematic constraints between two body frames of the flexible and
virtual bodies can be obtained from Eqs. (1) and (3) as follows:
( ) 0 u u A R r C = + =
+ i
f
i i i i i
R 0
1
(7)
0
h A A f h A A f
h A A g h A A g
g A A f g A A f
=
(
(
(
(
=
+ +
+ +
+ +
1 , 1
1 , 1
1 , 1
i i i
f
T i
T
i T
i i i
f
T i
T
i T
i i i
f
T i
T
i T
(8)
where
| |
(
(
(
=
1 0 0
0 1 0
0 0 1
h g f (9)
Orthogonality conditions would have been used in deriving the orientational
constraints. However, the in Eq. (8) is employed in this research for simple
implementation. Eqs. (7) and (8) yields algebraic constraint equations that
describe the flexible joint between flexible body and virtual body .
Taking variation of Eqs. (7) and (8) yields
i
C
i 1 + i
( )
( )
( )
0 q
C
C
q C
q
q
q
=
(
(
i
i
R i
flex
i
(10)
where
| |
T
T
i
T
i
T
i
f
T
i
T
i i 1 1 + +
= r p R (11)
and the constraint Jacobian matrix ( )
flex
i
q
C is obtained as
( ) |
( )
(
(
(
(
=
+
+
+
1
1
1
i
h
T i f
h
T i
h
T
i
h
T i f
h
T i
h
T
i
g
T i f
g
T i
g
T
i
i
R
i i i
R
B f 0 B f B f
B g 0 B g B g
B f 0 B f B f
C
0 I A B I C
q
q
(12)
RecurDyn
=
=
=
=
+ +
+
+
k A A B
k A B
A k A A B
u A B
(13)
and the vectors ) (
i
skew u , , , and are the
skew symmetric matrices of vectors,
) (
1
k A
+ i
skew ) (
1 ,
k A
+ i i
skew ) (k skew
i
u , , , and , respectively.
In order to obtain the acceleration level constraint, one can differentiate Eqs. (7)
and (8) twice with respect to time to yield
k A
1 + i
A
1 , + i i
k k
( ) ( ) ( ) ( )
( ) ( )
(
(
(
(
(
+ + +
+ + +
+ + +
=
= =
4 3 2 1
4 3 2 1
4 3 2 1
2
2
2
) ( 2 ) (
h
T
h
T
h
T
h
T
h
T
h
T
h
T
h
T
g
T
g
T
g
T
g
T
i
f
i
R
i i i
flex
i
c
i i
flex
i i
flex
i
skew skew
H f H f H f H f
H g H g H g H g
H f H f H f H f
p A u A
Q q q C q C
q
q q
&
& & & &
(14)
where the is the angular velocity with respect to the body reference frame
and the generalized velocity vector is q&
| |
T
T
i
T
i
T
i
f
T
i
T
i i 1 1 + +
= r p R q & &
&
& (15)
and
( ) ( )
h g k
skew
skew skew
skew skew skew
skew skew
i
f
i i i i
f
i
f
i
k
i i i
T
i
k
i i
T
i i
k
i
T
i i i
k
, ,
) (
) ( ) (
) ( ) ( ) (
) ( ) (
1 , 4
1 1 1 3
1 1 2
1 1
=
=
=
=
=
+
+ + +
+ +
+
p k A A p H
k A A H
k A A H
k A A H
& &
(16)
3-9
3.3. EQUATIONS OF MOTION
Even though the proposed method is applicable to a general system consisting
of many flexible bodies, a slider crank mechanism with one flexible body in Fig.
4(a) is used to clearly show the impact of the proposed method on the equations
of motion. An equivalent virtual system, modeled by using the rigid virtual
bodies proposed in this investigation, is shown in Fig. 4(b). The augmented
equations of motion for the system is obtained by using the general form of
equations of motion as [11]
(
+ +
=
(
(
(
c
s v e
T
Q
Q Q Q
q
0 C
C M
q
q
& &
(16)
where is the mass matrix of the system. The vector consists of
translational acceleration for rigid and flexible bodies, angular acceleration, and
modal acceleration for the flexible body.
M q& &
FLEXIBLE BODY FLEXIBLE BODY
RIGID BODY RIGID BODY
C
R
A
N
K
C
R
A
N
K
SLIDER SLIDER
C
O
U
P
L
E
R
C
O
U
P
L
E
R
1
2
3
Y
Z
X
P1 P1
(a) Two rigid bodies and one flexible body
RecurDyn
s
Q
v
Q
e
Q
c
Q
c
Q
3.3.1 COEFFICIENT MATRIX OF CONVENTIONAL AUGMENTED FORMULATION
The mass matrix for the system in Fig. 4(b) is
(
(
(
=
3
2
1
r
r
f
M 0
M
0 M
M (17)
where the mass matrices for virtual bodies, and are the mass matrix for
flexible body and for a rigid body, as
f
M
r
M
3-11
) 3 , 2 ( ,
6 6
) 6 ( ) 6 (
1
=
(
=
(
(
(
+ +
k
symmetric
k
k
rr k
r
nf nf
ff f fr
r
rr
f
m 0
0 m
m m m
m m
m
M
M
(18)
where is the number of modal coordinates. The constraint Jacobian matrix
of the slider crank mechanism with flexible crank is
nf
C
) (
q
C
( )
( )
( )
( ) (
(
(
(
(
=
int
30
int
23
,
12
,
01
) (
jo
jo
c flex
c flex
c
q
q
q
q
q
C
C
C
C
C (19)
where, ( )
c flex,
q
C is the constraint Jacobian matrix of the flexible joint obtained
by the conventional method[11].
3.3.2 COEFFICIENT MATRIX OF PROPOSED AUGMENTED FORMULATION
The mass matrix for the system in Figure 4b is
(
(
(
(
(
(
=
5
4
3
2
1
r
r
v
f
v
M
M 0
M
0 M
M
M (20)
where the mass matrix for virtual body, , the mass matrix for flexible body,
v
M
RecurDyn
(
int
int
int
,
,
int
p
p
| | 3 , 1 ,
6 6
= =
k
v
0 M
) 6 ( ) 6 (
2
nf nf
ff f fr
r
rr
f
+ +
(
(
(
=
m m
m m
m
M
(21)
) 5 , 4 ( =
(
= k
k
k
rr k
r
m 0
0 m
M
The proposed constraint Jacobian matrix ( )
p
q
C of the slider crank
mechanism with flexible crank is
( )
( )
( )
( )
( )
( )
( )
=
50
45
34
23
12
01
jo
jo
jo
flex
flex
jo
p
q
q
q
q
q
q
q
C
C
C
C
C
C
C (22)
where ( )
p flex,
q
C is the constraint Jacobian matrix of the flexible body joint
obtained by the proposed method. As shown in Eq. (22), the constraint Jacobian
matrix can be clearly divided into flexible and rigid body joint modules by
introducing rigid virtual bodies.
3.3.3 NON-SINGULARITY OF AUGMENTED MASS MATRIX
If the constraint Jacobian matrix C has a full row rank, the coefficient
matrix of Eq. (16) is non-singular, which can be proved by showing that the
following equations have only trivial solutions under the same assumption.
q
3-13
( ) 0 y C y
q
= +
3 1
N
T
N
M (23)
( ) 0 y C
q
=
3
V
T
(24)
( ) ( ) 0 y C y C
q q
= +
2 1
V N
(25)
where ,
N
M ( )
N
q
C , and ( )
V
q
C are the mass matrix of non-virtual body, the
Jacobian of Eqs. (23) and (24) after pre-multiplying by Eq. (23) and by
Eq. (24) yields
T
1
y
T
2
y
( ) ( ) ( ) 0 y M y y C y C y y M y
q q
= = + +
1 1 2 1 3 1 1 N
T
V N
T
N
T
(26)
where Eq. (25) is used. As a result, . Eqs. (23) and (24) reduces to 0 y =
1
0 y C
q
=
3
T
(27)
Since the has full row rank, must be zero. Substituting into Eq.
(25) yields
q
C
3
y 0 y =
1
( ) 0 y C
q
=
2
V
(28)
Since rank of ( )
V
q
C
y =
3
is the same as the size of , must be zero. Since
, , are only solutions of Eqs. (23), (24), and (25), their
coefficient matrix is non-singular.
2
y
2
y
0 y =
1
0 y =
2
0
3.4. COMPUTER IMPLEMENTATION AND DISCUSSIONS
In previous sections the flexible equations of motion and kinematic constraints
using virtual body techniques are presented. In this section, the computer
implementation methods for the equations developed in section 2 and 3 are
illustrated.
3.4.1 NUMERICAL ALGORITHM
RecurDyn
r
(1)
and
(
(2)
Their corresponding quantities in the z y x frame are respectively defined
as
(
(
=
(
=
A
r A
r
Y
T
T
& &
(3)
RecurDyn
=
(
=
A
r A
r
Z
T
T
(4)
where is the orientation matrix of the A z y x frame with respect to the
Z Y X frame. Two flexible bodies connected by a joint and their reference
frames are shown in Fig. 1.
Suppose there exists a joint between the
1 1 1 i i i
z y x and
frames, and a force applied at the origin of the
1 1 1 j j j
z y x
2 2 2 j j j
z y x frame. Kinematic
admissibility conditions among the reference frames can be divided into two
categories. One is the admissibility conditions between the two joint frames and
the other is the admissibility conditions among the frames within a flexible body.
These two types of conditions have been mixed in formulating the kinematic
joint constraints and generalized forces in the previous works. As a result, every
joint and force modules in a flexible multibody code, such as ADAMS [16] and
DAMS [17], have been developed separately for rigid and flexible bodies. This
would take long time for computer implementation and prone to coding errors.
Especially, flexible body programming requires much more effort than rigid
body programming does due to complexity associated with flexibility
generalized coordinates and the strain energy.
Figure 2 Two adjacent flexible bodies and three virtual bodies
4-5
In order to minimize the programming effort, a concept of the virtual body is
introduced in this section. At every joint and force reference frames, a virtual
rigid body, whose mass and moment of inertia are zero, is introduced. The
virtual body and the original flexible body are then connected by a virtual joint.
As an example, three virtual rigid bodies are introduced for two adjacent
deformable bodies as shown in Fig. 2. Note that the flexible bodies have no joint
or applied force except the virtual joints which are represented by the kinematic
admissibility conditions among the flexible body frame and the virtual body
frames. Therefore, the joint and force modules are developed only for rigid
bodies and one flexible body joint of the virtual joints to be added in the joint
module. The recursive kinematic relationships representing the admissibility
conditions of the flexible body joint are formulated in the following subsections.
4.2.2 RELATIVE KINEMATICS FOR A FLEXIBLE BODY JOINT
Figure 3 Flexible body joint between a flexible body and a virtual body
A virtual body is always connected to the original flexible body by a flexible
body joint. Origin of the virtual body reference frame in Fig. 4 can be expressed
as follows:
RecurDyn
1 i
f
i i ) 1 (
q
i i 1
The angular velocity in the local reference frame is obtained as follows
f
i i i
T
i i i
T
i i i ) 1 ( 1 ) 1 ( 1 ) 1 (
+ = q A A &
(7)
where is used. Differentiating Equation (5) and multiplying by
yields
i
T
i
T
i i
A A A
) 1 ( ) 1 (
=
T
i
A
f
i i i
T
i i i i i
T
i i i
T
i i i ) 1 ( 1 ) 1 ( 1 ) 1 ( ) 1 ( 1 ) 1 (
~
+ = q A s A r A r & & &
(8)
where
i i i i i i ) 1 ( ) 1 ( 0 ) 1 (
~
+ = u s s , symbol with tilde denotes skew symmetric matrix
which consists of their vector elements, and wide tilde are used.
Combining Equations (7) and (8) yields the following recursive velocity
equation for a flexible body joint.
i i i
A A =
~
&
i
f
i i
f
i i i
f
i i i ) 1 ( 2 ) 1 ( 1 1 ) 1 (
+ = q B Y B Y & (9)
where
4-7
(
(
=
(
(
1 ) 1 (
1 ) 1 (
2 ) 1 (
) 1 (
) 1 ( ) 1 ( ) 1 (
1 ) 1 (
~
i
T
i i
R
i
T
i i
f
i i
T
i i
i i
T
i i
T
i i f
i i
A
A
A 0
s A A
B
B
(10)
It is important to note that matrices and are function of only
modal coordinates of the flexible body i-1. As a result, further differentiation of
the matrices and B in Equation (9) with respect to other than bf
yields null. This property will play a key role in simplifying recursive
formulas in sections 4 and 5.
f
i i 1 ) 1 (
B
f
i i 2 ) 1 (
B
f
i i 1 ) 1 (
B
f
i i 2 ) 1 (
f
i i ) 1 (
q
Equation (9) defines the kinematic relationships between an inboard flexible
body and an outboard rigid body. The kinematic relationships between an
inboard rigid body and an outboard flexible body can be derived similarly.
Similarly, the recursive virtual displacement relationship between a flexible body
and a virtual body is obtained as follows
f
i i
f
i i i
f
i i i ) 1 ( 2 ) 1 ( 1 1 ) 1 (
+ = q B Y B Y & (11)
where
(
(
=
(
(
1
1 1 ) 1 (
2 ) 1 (
) 1 (
) 1 ( ) 1 ( ) 1 (
1 ) 1 (
~
~
i
R
i i i i
r
i i
T
i i
T
i i i i
T
i i r
i i
s
B
A 0
A s A
B
(12)
4.2.3 RELATIVE KINEMATICS FOR A RIGID BODY JOINT
The recursive velocity relationship for a rigid body joint connecting two rigid
bodies can be derived by following the similar steps as in Equations (5)-(9) as
r
i i
r
i i i
r
i i i ) 1 ( 2 ) 1 ( 1 1 ) 1 (
+ = q B Y B Y & (13)
RecurDyn
+
=
(
(
+
=
i i
T
i i
i i
T
i i i i i i i i
T
i i r
i i
T
i i
T
i i i i i i i i i i
T
i i
T
i i r
i i
i i
) 1 ( ) 1 (
) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 (
2 ) 1 (
) 1 (
) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 ( ) 1 (
1 ) 1 (
)
~
)
~
((
)
~
~
~
(
) 1 (
H A
H A s A d A
B
A 0
A s A d s A A
B
q
(14)
where is determined by the axis of rotation. Node that the matrices are
function of only .
i i ) 1 (
H B
r
i i ) 1 (
q
4.2.4 GRAPH REPRESENTATIONS OF MECHANICAL SYSTEMS
The graph theory was used to automatically preprocess mechanical systems
having various topological structures in References [1, 4]. A node and an edge in
a graph have represented a body and a joint, respectively. The preprocessing,
based on the graph theory, yields the path and distance matrices that are provided
to automatically decide computational sequences. Two computational sequences
are required in a general purpose program. One is the forward path sequence
starting from the base body and moving towards the terminal bodies. The other is
the backward path sequence starting from the terminal bodies and moving
towards the base body.
Figure 4 Flexible slider crank mechanism
4-9
Figure 5 Graph representation and computational sequence
In order to derive systematically the recursive formulas, bodies in a graph are
divided into four disjoint sets (associated with a generalized coordinate ) as
follows :
k
q
) (
k
q I ={adjacent outboard body of the joint having as its generalized co-
ordinate}
k
q
) (
k
q II ={all outboard bodies of , excluding all bodies in } ) (
k
q I ) (
k
q I
) (
k
q III ={all bodies between the base body and the inboard body of ,
including the base and inboard bodies and excluding all bodies in }
) (
k
q I
) (
k
q I
) (
k
q IV ={ the complementary set of } ) ( ) ( ) (
k k k
q q q III II I
RecurDyn
=
(
=
f
r
zf zr
f
(15)
where and q are the relative and modal co-ordinates vectors for a system,
respectively. The dimension of , and are, respectively, assumed to
be ,
r
q
n
f
Y
r
q&
f
q&
nc r , and . The velocity nf
nf nc
R
+
Y with a given can be
evaluated either by using Equation (15) obtained from symbolic substitutions or
by using (9), (11) and (13) with recursive numeric substitution of 's. Since
both formulas give an identical result and recursive numeric substitution is
proven to be more efficient [4], matrix multiplication with a given will
be actually evaluated by using Equations (9), (11) and (13). Since q in
nf nr
R
+
q&
i
Y
q B& q&
&
4-11
Equation (15) is an arbitrary vector in
nf nr
R
+
, Equations (9), (11), (13) and (15)
which are computationally equivalent, are actually valid for any vector
such that
Bx
x
X
=
(
(
BX BX+
q
) (Bx
i ) 1 (
+ B
i
i i
(
1 1
X
nf nr
R
+
x
X
(16)
and
X = (17)
where
nf nc
R
+
X is the resulting vector of multiplication of and and B
matrices depend on a joint type. As a result, transformation of
B x
nf nc
R
+
Y into
nf nc
R
+
Bx is actually calculated by recursively applying Equation (17) to
achieve computational efficiency in this research.
4.3.2 RECURSIVE FORMULA FOR
q
X =
Equation (17) is partially differentiated with respect to for
to obtain the recursive formula for as follows.
k
q
) ( ,..., 1 nf nr k + =
q
) (Bx
i i q i i q i i i q i i q i
k k k k
) 1 ( 2 ) 1 ( 1 1 1 1 ) 1 (
) ( ) ( ) ( ) (
+ = X B X X B X (18)
Since matrices depend only on the generalized coordinates for joint
, their partial derivatives with respect to generalized coordinates other than
become null. In other words, the partial derivatives become null if
does not belong to set . If body is an element of set , Equation
(18) becomes
B
i i ) 1 (
i i ) 1 (
q
k
q
) (
k
q I ) (
k
q II
(19)
k k
q i q i
) ) (
) 1 (
= B X
RecurDyn
=
f
c
fT
T T
Q
Q
q Z Q Z W (23)
where Z must be kinematically admissible for all joints in a system and
c
Q
RecurDyn
\
|
+ +
|
|
.
|
\
|
+ =
+
+
fjts i i q l
i
c
i
fT
i i
f
i i
fT
i i
rjts i i q l
i
c
i
rT
i i
rT
i i
i i
i i
) 1 ( ) (
1 1 2 ) 1 ( ) 1 ( ) 1 (
) 1 ( ) (
1 1 2 ) 1 ( ) 1 (
) 1 (
) 1 (
I
I
S Q B Q q
S Q B q W
(26)
where
+ +
+ +
+ + +
+
+
+
) (
2 2 1
1
) 2 )( 1 (
) 2 )( 1 (
) (
body terminal a is 1 if
i i
i i
q l
i
c
i
T
i
i
i
I
S Q B S
0 S
(27)
The recursive formula for bf and is obtained by equating Equations
(25) and (26) as follows:
* f
Q
* r
Q
+ + + + +
+ + =
) (
1 1 2 ) 1 ( ) 1 (
*
) 1 (
) 1 ( i i
q l
i
c
i
T
i i i i i i
I
) S (Q B Q Q (28)
4-15
where for a rigid body joint and for a flexible body joint connecting
an inboard flexible body and an outboard virtual body, and for a
flexible body joint connecting an inboard virtual body and an outboard flexible
body, and is defined in Equation (27).
0 =
+ ) 1 (i i
1 + i
S
Q
f
i i i i ) 1 ( ) 1 ( + +
= Q Q
Since Q in Equation (23) is an arbitrary vector in
nf nc
R
+
, Equations (23) and
(28)are valid for any vector in G
nf nc
R
+
. As a result, the matrix multiplication
of is actually evaluated to achieve computational efficiency in this
research by
G B
T
+ + + + +
+ + =
) (
1 1 2 ) 1 ( ) 1 ( ) 1 (
) 1 (i i
q I l
i
c
i
T
i i i i i i
) S (G B G g (29)
where is the result of and is defined as in Equation
(28) and
g G B
T
) 1 ( + i i
G
) 1 ( + i i
Q
body terminal a is 1 if
1
+
+
i
i
0 S (30)
+ +
+ + + + +
+
) (
2 2 1 ) 2 )( 1 ( 1
) 2 )( 1 (
) (
i i
q l
i
c
i
T
i i i
I
S Q B S (31)
Recursive formula in Equation (29) must be applied for all joints in the
backward path sequence to obtain where is a constant vector in G B g
T
= G
nf nc
R
+
.
4.4.2 RECURSIVE FORMULA FOR
k k
q
T
q
) ( G B g =
The Recursive formula for is obtained by replacing by in
Equation (29) and
k
q
T
) ( G B i 1 i
1 + i by 1 i in Equation (31) and taking partial derivative
with respect to yield
k
q
+ +
+ + =
) (
2 ) 1 (
) (
2 ) 1 ( ) 1 ( ) 1 (
) 1 (
) 1 (
) ( ) ( ) (
i i
k
i i
k k k
q l
q i
c
i
T
i i
q l
i
c
i q
T
i i q i i q i i
I
I
) S (G B
) S (G B G g
(32)
RecurDyn
+ + + =
) (
1 ) 1 (
) (
1 ) 1 ( ) 1 (
) 1 ( ) 1 (
) ( ) (
i i
k
i i
k k
q l
q i
c
i
T
i i
q l
i
c
i q
T
i i q i
I I
) S (G B ) S (G B S (33)
Since is a constant vector, . If
, matrices are not functions of . Therefore,
their partial derivatives with respect to become null. As a result, Equations
(32) and (33) can be simplified as follows.
nf nc
R
+
G
) (
k
q IV III
0 G =
k
q
k
q ) ( ) (
k k
q q i II B
k
q
=
) (
2 ) 1 ( ) 1 (
) 1 (
) ( ) (
i i
k k
q l
q i
T
i i q i i
I
S B g (34)
=
) (
) 1 ( 1
) 1 (
) ( ) (
i i
k k
q I l
q i
T
i i q i
S B S (35)
Since for the terminal bodies, for .
Thus, for , Equation (34) becomes
0 S =
k
q i
) (
) (
k
q i II
0 S =
k
q i
) ( ) ( ) (
k k
q q i IV II
) (
k
q IV
0 g =
k
q i i
) (
) 1 (
(36)
There are two recursive formulas in the case of body . If body
and body
) (
k
q i I
) (
k
q i I 1 + i belongs to set , and . Thus, Equation
(32) and (33) become
) (
k
q II 0 S =
k
q i
) (
+ =
) (
2 ) 1 ( ) 1 (
) 1 (
) ( ) (
i i
k k
q l
i
c
i q
T
i i q i i
I
) S (G B g (37)
+ =
) (
1 ) 1 ( ) 1 (
) 1 (
) ( ) (
i i
k k
q l
i
c
i q
T
i i q i
I
) S (G B S (38)
where must be saved when is computed. This recursive formula can
be applied to compute . As an example,
i
S G B
T
q
T
) ( G B ( )
34 34
) (
q
T
q
G B g =
34
q q
k
=
for the system
in Fig. 4 is obtained, as shown in Fig. 7 for the case of . Note that the
components of ( )
34
q
g are either zero or simple to compute.
4-17
Figure 7 Computation sequence of .
ii ii
q q
) ( ) ( BG g =
4.5. THE GOVERNING EQUATIONS OF SOLUTION
4.5.1 IMPLICIT INTEGRATION OF THE EQUATIONS OF MOTION
The dynamic equations of motion for a constrained mechanical system in the
joint space have been obtained in Reference [1] by the velocity transformation
RecurDyn
+ +
+ +
=
n n n n
n n n
n n
n n n n n
n n
T
n n
T
n
t
t
t
t
& &
&
,
2 0 0
1 0 0
(42)
where , , , and are determined by the coefficients
of the BDF. The must be chosen such that the augmented square matrix
is nonsingular. Applying Newton's method to solve the nonlinear system
in Equation (42) yields
|
T
T
n
T
n
T
n
T
n
T
n
a v q p , , , =
0
U
|
0
(
(
U
T
0
H p ) H(p =
n
(43)
,... 3 , 2 , 1 ,
) ( ) 1 (
= + =
+
i
i
n
i
n
p p p (44)
where
4-19
(
(
(
(
(
(
(
(
=
0
0 0
0 0 0
F F F F
0 U U 0
0 0 U U
) H(p
a v q
v q
q
a v q
& & & & & &
& &
T T
T T
n
0 0 0
0 0 0
(45)
Since and are highly nonlinear functions of q, and
q
F
q
a v,
, some
cautions must be taken in deriving the non-zero expressions in matrix so
that they can be efficiently evaluated.
p
H
4.5.2 APPLICATION OF THE GENERALIZED RECURSIVE FORMULAS
A set of the generalized recursive formulas has been developed in the sections
3 and 4. This section shows how these formulas can be utilized to efficiently
compute the in H in Equation (45). Inspection of reveals that partial
derivatives of F , , , , and are needed to be computed.
Only the is presented in this section and the rest can be derived similarly.
q
F
p
v
F
p
H
q a
F
q
&
q
& &
q
F
In Equation (41), differentiation of matrix with respect to vector q
results in a three dimensional matrix. To avoid the notational complexity for the
three dimensional matrix, Equation (41) is differentiated with respect to each
generalized coordinate one by one. Thus,
B
k
q
nf nr k
k k
k k
q
T
q
T
T T
q q
+ = + +
+ =
,..., 3 , 2 , 1 ), ( ) ( ) Q Y (M B
) Q Y (M B F
Z
Z
&
&
(46)
Since the term ) Q
Z
T
( can be easily expressed in terms of the Cartesian
coordinates,
k
q
T
) Q
Z
( is obtained by applying the chain rule as follows.
RecurDyn
i
1 + i Z Y X
k k k
z y x ) , ( j i k = is the nodal reference frame
attached to a node , and is a position vector of the node k
k
r k .
is the reference frame attached to a node and the first
subscript denotes the inboard node number of the second subscript . The
orientation of coincides with that of in
the undeformed state. The absolute nodal displacements measured in the
frame have been solved for in the conventional finite element analysis
methods(see Refs. 1-4). In contrast to conventional methods, the relative nodal
displacements measured in its inboard nodal reference frame are solved in this
paper.
i i i i i ) 1 ( ) 1 ) 1 (
z x
1 i
i 1 (
x
X
i (
y
Z Y
i
) 1 ( ) 1
z y
i
i
( i i ) i 1 (
z
i i ) 1 (
y
i )
) 1 (
x
i
i
x
i
z
i
y
1
x
i
1
z
i
1
y
i
i-1
i
X
Z
Y
i
r
1 i
r
i i ) 1 (
x
i i ) 1 (
z
i i ) 1 (
y
i+1
RecurDyn
Y X Z frame can be
expressed in terms of these of node 1 i and the relative nodal displacements as
follows:
( )
'
) 1 (
'
0 ) 1 ( ) 1 ( ) 1 ( i i i i i i i
+ + = u s A r r (1)
i i i i i i i i ) 1 (
'
) 1 ( ) 1 ( 1
) (
= C D A A (2)
where
[
T
i i i i i i i i
'
3 ) 1 (
'
2 ) 1 (
'
1 ) 1 (
'
) 1 (
= ] (3)
In Eqs.(1) and (2),
k
A ) , 1 ( i i k = denotes the transformation matrix for nodal
reference frame , denotes the constant transformation matrix from
to , denotes the location vector of node
measured in in the undeformed state, and denotes the
deformation vector of node relative to the nodal frame
k
i ) 1
)
i i ) 1 (
i i i ) 1 ( ) 1
z
) 1 (
z
i
i
C
i (
y
) 1 (
y
i
i i i
z y x
i (
x
1 ( i
'
0 ) 1 ( i i
s i
x
'
) 1 ( i i
u
1 i
i i ) 1 (
x
. is the
transformation matrix due to a rotational displacement of
relative to the nodal frame
i i ) 1 (
D
i i ) 1 (
y
i ) i 1 (
z
I 0
u s I
A 0
0 A
B
)
~ ~
(
'
) 1 (
'
0 ) 1 (
) 1 (
) 1 (
1 ) 1 (
i i i i
T
i i
T
i i
i i
(12)
i i
T
i i
T
i i
i i
) 1 ( ) 1 (
) 1 (
2 ) 1 (
H
I
A 0
0 A
B (13)
It is important to note that matrices and are only functions of the
relative displacement between nodes
1 ) 1 ( i i
B
2 ) 1 ( i i
B
i i ) 1 (
q 1 i and i .
The virtual displacement relationship between the absolute and relative nodal
coordinates for the whole system can be obtained by repetitive application of Eq.
(9) along a chain in a graph. As an example, the virtual displacement relationship
RecurDyn
=
342 232 341 122 231 341 012 121 231 341
232 122 231 012 121 231
122 012 121
012
B B B B B B B B B B
0 B B B B B B
0 0 B B B
0 0 0 B
B (17)
5.3. GOVERNING EQUATIONS OF EQUILIBRIUM
5.3.1 STRAIN ENERGY
The strain energy in a finite element having multiple nodes is affected only by
the relative displacements of nodes relative to the inboard nodal frame of the
element and is free from its rigid body motion. As a result, the variational form
of the strain energy for a system can be obtained in a summation form as
=
= =
n
k
T
k k k k
T
k k
W
1
) 1 ( ) 1 ( ) 1 (
Kq q q K q (18)
where q must be kinematically admissible for all constraints. Since the
stiffness matrix is generated in the nodal reference frame, the strain energy due
to a rigid body motion of a node does not appear in Eq. (18). The element
stiffness matrix is contributed from linear and nonlinear terms as (see
Ref. 3)
k k ) 1 (
K
nL
k k
L
k k k k ) 1 ( ) 1 ( ) 1 (
+ = K K K (19)
where
=
k k
l
k k
L
k k
T
k k
L
k k
dx
) 1 (
0
*
) 1 ( ) 1 (
*
) 1 ( ) 1 (
K (20)
5-7
+
=
) 1 (
0
*
) 1 ( ) 1 ( ) 1 (
*
) 1 ( ) 1 (
) (
i i
l
k k k k
nL
k k
T
k k
nL
k k
dx q K (21)
In Eqs. (19) - (21), denotes a linear stiffness matrix, denotes a
nonlinear stiffness matrix, and l denotes the undeformed length of the
element between the nodes and . Note that the significance of
depends on the magnitude of . becomes negligible when the
magnitude of is small, which is true when the element size is small. It is
very difficult analytically to prove the significance of . As a consequence,
the significance of has been demonstrated through a numerical example
in 5.
L
k k ) 1 (
K
nL
k k ) 1 (
K
nL
k k ) 1 (
K
k k ) 1 (
k k ) 1 (
q
1 k k
nL
k k ) 1 (
K
nL
k k ) 1 (
K
k k ) 1 (
q
nL
k k ) 1 (
K
5.3.2 EXTERNAL FORCE
The virtual work done by both nodal forces described in the absolute
nodal coordinate system and described in the relative nodal coordinate
system is obtained as follows:
Q
R
R q Q Z
T T
W + = (22)
where Z must be admissible for the kinematic relationship between Z and
q . Substitution of q B Z = into Eq. (22) yields
( )
*
Q q R Q B q
T T T
W = + = (23)
where
R Q B Q + =
T *
(24)
RecurDyn
=
= =
5
1
*
5 ) 1 ( ) 1 ( 0 5
k
k k k k
A C D A A (26)
where and are given by the boundary conditions at node 5. Since Eq.
(26) comprises of nine dependent equations, only three are independent. The
three independent constraint equations can be extracted by imposing
perpendicularity between the axes of reference frames. As a result, the six
independent constraint equations are given as
*
5
r
*
5
A
=
*
52 51
*
52 53
*
51 53
*
5 5
a a
a a
a a
r r
T
T
T
(27)
where
[ ]
53 52 51 5
a a a A = (28)
[ ]
*
53
*
52
*
51
*
5
a a a A = (29)
In Eqs. (28) and (29), and
i 5
a
*
5i
a ) 3 , 2 , 1 ( = i denote the -th column vector
of and , respectively.
i
5
A
*
5
A
RecurDyn
F q
0
F
q
q q
T
(33)
where
( )
q
q q
Q K F
*
+ =
T
(34)
By solving Eq. (33), the improved solution of for the next iteration can be
obtained as follows:
q
q q q + = (35)
5-11
By using Eqs. (33) and (35), the iteration continues until the solution variance
remains within a specified allowable error tolerance. Before solving Eq. (33), it is
necessary to calculate . However, the calculation of F is numerically
difficult and tedious.
q
F
q
In order to save computing time in solving Eq. (33), some numerical
approximation techniques may be applied. As an example, the coefficient matrix
of Eq. (33) may remain near constant if the variation of is small, which is the
case when the lengths of finite elements are small. In such case, the coefficient
matrix of Eq. (33) can be hold during Newton-Raphson iterations, which
significantly reduces the computation time. However, the approximation
technique may not converge for a system whose is large. To overcome this
numerical difficulty, a combined incremental and iterative method (see Ref. 16)
can be used.
q
q
5.4. NUMERICAL ALGORITHM
Kinematics of the relative nodal displacements and the equations of
equilibrium are presented in the section 3. This section explains how the
equations are implemented to obtain the relative and absolute nodal
displacements of a structure. The numerical algorithm for closed loop systems is
as follows:
1) Perform the graph theoretic preprocessing to determine computational path
sequences.
2) Form a stiffness matrix . K
3) Compute , , and for in the backward path sequence.
q
*
Q
k
q
4) Solve the Eq. (37) to obtain and q .
5) If and F q remains within the specified allowable error tolerance, then
go to step 6. Otherwise, improve the solution using Eq. (35). Go to step 3.
6) Compute the Cartesian deformations in the forward path sequence by using
Eqs. (1) and (2).
RecurDyn
L
Figure 6 A cantilever beam subjected to end moment
-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13
-2
0
2
4
6
8
10
12
14
16
Undeformed
Proposed
ANSYS: nonlinear
ANSYS: linear
Y
[
m
]
X [m]
Figure 7 Deformed shape of the beam
5-13
In the figure, E , , L , , and denotes Young's modulus, Poisson ratio,
the length of the beam, the cross sectional area of the beam, and the second area
moment of the cross section, respectively. M =6.54510
6
[Nm] is applied at
the end node. Fig. 7 shows the deformed shapes of the beam by the proposed
method by the proposed method and a commercial program ANSYS. In the
figure, Proposed, ANSYS: nonlinear, and ANSYS: linear denote numerical
results by the proposed method, a commercial program ANSYS using nonlinear
analysis, and ANSYS using linear analysis, respectively. It shows that the
numerical results obtained by the proposed method and ANSYS(nonlinear
analysis) are almost identical, but the numerical results by ANSYS(linear
analysis) shows large difference with the remaining two numerical results.
A I
0 2 4 6 8 10
-10
-8
-6
-4
-2
0
u
e
: Proposed
u
e
: ANSYS
u
e
[
m
]
The number of elements
Figure 8 Convergence of axial deformation at the end node vs. the number of elements
Fig. 8 shows the convergence of the axial deformation at the end node.
When fewer elements are used for static analysis, the numerical results of
ANSYS are more accurate than those of the proposed method, but obtained
e
u
e
u
RecurDyn
L
P
Figure 9 A closed loop system subjected to concentrated force and Moment
Figure 10 Comparison of deformed shapes of the closed system
5-15
Fig. 9 shows a closed loop system subjected to a concentrated force and
moment
F
M at a point P . When =[310 F
4
-110
4
]
T
[N] and M =0.0
[Nm] are applied at the point P the deformed shapes of the system are shown
in Fig. 10. It shows that the numerical results obtained by the proposed method
with 20 elements and a commercial program ANSYS are almost identical.
-1 0 1 2 3 4 5 6 7 8
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
(F=[3.0E+4, -3.0E+4]
T
[N], M=3.0E+4[Nm])
Y
[
m
]
X[m]
Undeformed shape
Deformed shape: 20 elements
Figure 11 Undeformed and deformed shapes of the closed loop system
0 1 2 3 4 5 6 7 8
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
(F=[3.0E+4, -3.0E+4]
T
[N], M=3.0E+4[Nm])
Y
[
m
]
X[m]
0.33F, 0.33M
0.67F, 0.67M
1.00F, 1.00M
Figure 12 Deformed shapes of the closed loop system at each load step
RecurDyn
+
=
) (
2 2
L L l
l
P P
i s
i
i
(2)
where , , and are the initial pressure and distances when the road arm
is in its initial configuration,
i
P
i
l
i
L
2
is a constant which is equal to 1.4, and is
the distance shown in Fig. 2.
2
L
6-5
Figure 2. Schematic diagram of spring-damper suspension units: hydro pneumatic
suspension unit and torsion bar systems.
The distance l can be adjusted by charging or discharging oil into the oil
chamber. The torsion bars are mounted on the middle stations for this vehicle
model. A simple torsional spring model is used in this investigation to represent
the stiffness of the torsional bars. The stiffness coefficient of the torsion bar
spring is approximately Nm/rad. Figure 2 shows the schematic diagram
of the HSU and the torsion bar systems. Figure 3 shows the spring characteristics
which are employed in this investigation.
s
4
10 5
Figure 3. Spring characteristics of suspension unit
Each track subsystem is modeled as a series of bodies connected by rubber
bushings around the link pins which are inserted into a shoe plate with some
radial pressure in order to reduce the non-linear effect of the rubber. When the
RecurDyn
0
0 C
K
K
Q
Q
&
&
R
R R R
j
j
R
C 0
0
where and are the 3
R R
, , C K K
C 3 diagonal matrices that contain the
stiffness and damping coefficients of the bushing, and is translational force
vector and is the vector of translational deformations of the frame
j
R
Q
R
j relative
to the frame . Similarly, is the rotational force vector and is the
vector of relative rotational deformations of the frame
i
j
Q
j relative to the frame .
The force and torque applied to the frame i are assumed to be equal in
i
RecurDyn
j j j
Z Y X frame with respect to
the
i i i
Z Y X frame along the Z axis, and are the stiffness and
damping coefficients.
Rz
K
Rz
C
The first two components of the bushing restoring torque as the result of the
relative rotation of link with respect to link i j are given by
i , j
x x
i , j
x x x
C K T
&
= (11)
i , j
y y
i , j
y y y
C K T
&
= (12)
where and are relative rotational deformations of the
i , j
x
i , j
y
j j j
Z Y X frame
about x-axis and y-axis with respect to the
i i i
Z Y X frame, respectively, ,
, , and are stiffness and damping coefficients. The restoring bushing
torque about the
x
K
y
K
x
C
y
C
j
Z axis due to the rotation of link j with respect to link i
ib , j
z z
ib , j
z z z
C K T
&
= (13)
RecurDyn
=
ij
z
ij
y
ij
x
ij
d
d
d
(14)
It follows that
( )
ij
x
ij
y
ib , i
z
d / d
-1
tan = (15)
where is the transformation matrix that defines the orientation of link
with respect to the global frame. Note that since the inertia of the connector
element is neglected, the resultant force acting on this element must be equal to
zero. Using the spring forces defined in this section, the generalized bushing
forces acting on the track links can be systematically defined.
i
A i
6.5. MEASUREMENT OF TRACK COMPLIANCE CHARACTERISTICS
In order to determine the stiffness and damping coefficients of the contact
force models used in this investigation, an experimental study is conducted to
examine the road wheel and track link contact as well as the interaction between
the track links. Since the experimental results are to be used in the dynamic
simulation of the multibody tracked vehicle, the dynamics of the contact is also
considered in the measurement process.
While a viscous damping force is proportional to the velocity, in many cases,
analytic expressions for the damping forces are not directly available. It is,
however, possible to obtain an equivalent viscous damping coefficient by
equating energy expressions before and after a contact. In this investigation, the
effective stiffness and damping coefficient are obtained by employing the
hysteresis loop method [10]. The effective stiffness and damping coefficient of
single degree of freedom system are given as follows [10]:
6-13
+ = cos
x
F
m
0
0 2
eff eff
K (16)
sin
0
0
x
F
C
eff
= (17)
where , , and are the effective mass, the magnitude of
applied force, the magnitude of displacement, the natural frequency, and the
phase angle of displacement, respectively.
eff
m ,
0
F
0
x ,
In these experimental studies, forces are applied to the center of the road
wheel which is in contact with a track link fixed to a rigid frame. A LVDT
sensor is attached between the center of the wheel and a track link fixed base to
measure the relative displacement. For a static test, the actuator force is
increased gradually up to 10 ton with 2mm/min velocity. For a dynamic test, the
actuator force is excited harmonically up to 35 Hz. Frequencies higher than 35
Hz are not considered in the measurement because of noise and system
resonance. The relationship between the effective stiffness, damping coefficient,
and frequencies is given by Park et al [11]. It can be shown that the effective
stiffness increases up to a frequency of 10 Hz and does not significantly change
after this frequency. On the other hand, the effective damping coefficient
decreases as the frequency increases.
A LVDT sensor is attached between two adjacent track links to measure the
relative displacement. For a static test, the actuator force is increased gradually
up to 10 ton with 2mm/min velocity. Figure 7 shows the resulting load-
displacement relationship. For a dynamic test, a harmonic actuator force with a
frequency up to 50 Hz is used. Figure 8 shows the hysteresis loop when the load
frequency is 10 Hz with 5 ton pre-static applied force. It can be observed that the
effective stiffness increases up to 12 Hz and does not significantly change after
this frequency. The effective damping coefficient, on the other hand, decreases
as the frequency increases.
RecurDyn
n n
HX X =
+1
} 1 ,......, 2 , 1 , 0 { N n (25)
where
T
n
2
n n n
] q , q , q [ & & & t t = X (26)
and
6-21
=
0 0
2 / 1 1 ) 2 / 3 (
54 / 29 1 ) 27 / 28 ( 1
2
2
2
H (27)
in which t = . The characteristic equation for
is obtained as follows: H
0 ) 27 / 1 ( } ) 27 / 2 ( 1 { } ) 27 / 28 ( 2 { ) det(
2 2 2 3
= + + = I H
(28)
where is the I 3 3 identity matrix and
denotes the eigenvalue. The
stability characteristics of the method are determined by the condition that the
roots of the characteristic equation remain in or on the unit circle of the complex
plane as follows:
1 , { }
3 2 1
, , max = (29)
where is called the spectral radius. Stability analysis can be assessed by
using the transformation of Eq. 7.9 to map the interior of the unit circle into the
left half-plane and by applying the Routh-Hurwitz criteria to the transformed
characteristic equation. The stability condition for the algorithm is obtained by
applying the Routh-Hurwitz criteria as follows:
0 ) 27 / 31 ( 4
2
(30)
0 ) 27 / 23 ( 4
2
(31)
which are reduced to
) / 8665 . 1 ( t (32)
Equation (32) provides a guideline in choosing a step size that satisfies the
stability condition.
RecurDyn
q
n
The step size determination algorithm is shown in Figure 12. Note that the
stability condition of instead of in Equation
(32) is used for conservative numerical integration. The integration step size
employed by the variable step integration algorithm used in this investigation,
when the vehicle maximum acceleration, steady state velocity at 50 Km/h and
stiff deceleration of braking, is shown in Figure 13. This figure shows that the
integration step size is relatively depended on the vehicle speed. The increment
of vehicle speed will enlarge impulsive contact forces and oscillation of track
links, and integration step size should be decreased, accordingly.
app
/ 5 . 0 < t
app
/ 8665 . 1 < t
6-23
Figure 12. Variable stepsize algorithm
Figure 13. Stepsize of variable step integration algorithm
RecurDyn
=
=
=
=
3 3 3
3 3 2
3 3 1
3 3
] [
] [
] [
] [
k k k
k k
k k
k k
z y x A
z x A
y z A
y x A
I
I
I
and (5)
T
] 1 1 1 [ = I
Then the unit normal vector of the plane is defined as n
T
a a
a a
] 1 [
1
1
2 1
2
2
2
1
+ +
= n
(6)
Virtual proving ground : Until recent development of computer simulation model
[6, 7, 9], the development process of tracked vehicle have been depended on
inefficient technologies of repeated procedures ; construction prototype vehicle
based on basic calculations and simple computer simulation, test on proving
ground, then modification. This expensive design procedure can be diminished
by recent developments of computer simulation.
In this investigation, only paved ground models are developed for the virtual
test of dynamic analysis of three dimensional tracked vehicle. The developed
computer models of grounds are stored into the created ground library.
As shown in Figs. 6 and 7 the variety of virtual proving grounds, symmetric
and unsymmetric bump courses, trench and ditch courses, longitudinal and
laternal inclined courses, and standard cross-country courses of RRC9 and
Profile IV, are constructed by using triangular patch elements. When a vehicle
runs over these virtually created proving grounds, the nonlinear behaviors of
track chains resulting from the interacting with the test grounds are obtained in
this numerical investigation.
7-11
(a) Single bump course (d) Obstacle course
(b) Trench course (e) Grade ability slope
(c) Ditch course (f) Side slope
Figure 6. Various paved virtual proving ground using triangular patch
RecurDyn
>
separate d
contact d
ij
k
ij
k
0
0
(8)
If this conditions is satisfied, the position vector shown in Fig. 3 is used
to compute the node location whether contact point
jk
B
r
B of node j is on the
patch plane . The position vector can be written as k
jk
B
r
ij
k
ij
k
ij
p
i i jk
B
d u A R r n + =
(9)
where is the transformation matrix associated with the orientation
coordinates of link and
i
A
i
ij
p
u is the local position vector of node j in the
track chain link coordinate system. On the other hand, using scalar triple product
RecurDyn
0
0
0
3 31
2 23
1 12
n
B
n
B
n
B
u r r
u r r
u r r
or (10)
0
0
0
3 31
2 23
1 12
n
B
n
B
n
B
u r r
u r r
u r r
then the node j of link is in contact with patch element . i k
If the node j is in contact with patch plane , the contact force at the
contact node can be computed using the equation as
k
ij
k
ij
k
ij
k
ij
k
ij
k
d C d K F
&
+ =
(11)
where and C are, respectively, the stiffness and damping coefficients of
the contact force model at node
ij
k
K
ij
k
j of body on patch plane . Using the
expression for the contact force as defined by the preceding equation, the contact
force vector can be defined as
i k
ij
k
ij
k
ij
k
F n F = (12)
where is a unit normal vector shown in Fig. 3. The virtual work of the
contact force at the nodes is given by
ij
k
n
=
=
n
j
ij
k
i
k
W W
1
=
=
n
j
ij
k
ij
k
ij
k
d F
1
n | |
(
=
i
i
T
i
T
i
R
R
Q Q
(13)
where
=
=
n
j
ij
k
T
i
R
1
F Q
=
=
n
j
ij
k
T ij
p
i
T
i
1
)
~
( F u A Q
(14)
are the generalized contact forces associated with the Cartesian and orientation
7-15
coordinates of link , and i
ij
p
is the skew symmetric matrix associated with the
vector
ij
p
u . In order to evaluate the tangential component of these contact forces
for friction effect at each contact nodes, the smooth Coulomb friction model [6]
is employed in this investigation. Figure 8 shows the computer animation of
multibody tracked vehicle running over APG Profile IV test ground.
u
~
Note that the proposed element free finite contact node method have several
advantages such as, simple computer implementation, easy contact detecting
algorithm for irregular surface, independent contact coefficients, and distribution
of concentrated contact forces, however, in the penalty function approach used in
this contact force model the determinations of spring and damping coefficients
may be a black art. These coefficients may not correspond to familiar physical
properties that can be measured experimentally. Careful numerical calibration
process is necessary to obtain reliable model, accordingly.
Time = 0.0 sec Time = 9.0sec
Time = 3.0 sec Time = 12.0 sec
RecurDyn
left
7.6. FUTURE WORK AND CONCLUSIONS
The dynamic track tension for a high mobility tracked vehicle is investigated
in this paper. The three dimensional multibody tracked vehicle consists of the
hull, sprockets, road arms, road wheels, support rollers, and sophisticated
suspension systems of hydro-pneumatic and torsion bars. A compliant force
model is used to connect the rigid body track links. The tracked vehicle model
has 189 bodies, 36 pin joints and 152 compliant bushing elements and has 954
degrees of freedom. Various ground profiles are developed by using triangular
patch elements. Numerical results are validated against experimental results.
Numerical simulations have been carried out under various maneuvering
conditions and effects of several conditions are discussed . Numerical results
showed that the optimal track tension may not be necessarily 10 % of the total
vehicle weight as many track vehicle researchers have believed. Further studies
must be carried out to find the optimal track tension.
7-23
REFERENCES
[1] G. R. Doyle and G. H. Workman, 1979, ''Prediction of Track Tension when Traversing
an stacle'', Society of Automotive Engineers, 790416
[2] A.G. Galaitsis, 1984, "A Model for Predicting Dynamic Track Loads in Military
Vehicles", ASME, Journal of Vibration, Acoustics, Stress, and Reliability in Design, Vol.
106/289
[3] R.M. Trusty, M.D. Wilt, G.W. Carter, D.R. Lesuer, 1988, ''Field Measurement of Tension
in a T-142 Tank Track'', Experimental Techniques
[4] M. K. McCullough, and E. J. Haug, 1986, ''Dynamics of High Mobility Tracked
Vehicles'', ASME, Journal of Mechanisms Transmissions, and Automation in Design,
Vol.108, pp. 189-196
[5] Choi, J. H., Lee., H. C., Shabana, A. A., Jan. 1998, ''Spatial Dynamics of Multibody
Tracked Vehicles: Spatial Equations of Motion'', International Journal of Vehicle
Mechanics and Mobility, Vehicle System Dynamics, Vol. 29, pp. 27-49
[6] Lee, H. C., Choi., J. H., Shabana, A. A., Feb. 1998, ''Spatial Dynamics of Multibody
Tracked Vehicles: Contact Forces and Simulation Results'', International Journal of
Vehicle Mechanics and Mobility, Vehicle System Dynamics, Vol. 29, pp. 113-137
[7] C. Scholar and N. Perkins, 1997, "Longitudinal Vibration of Elastic Vehicle Track
System", SAE, 971090, International Congress and Exposition, Detroit, MI, Feb. 24-27
[8] J. Chung, J. M. Lee, 1994 ''A New Family of Explicit Time Integration Methods for
Linear and Non-linear Structural Dynamics'', International Journal for Numerical
Methods in Engineering, Vol.37, 3961-3976
[9] H. S. Ryu, D. S. Bae, J. H. Choi and A. Shabana, 2000 ''A Compliant Track Model For
High Speed, High Mobility Tracked Vehicle'', International Journal For Numerical
Methods in Engineering, Vol. 48, 1481-1502
[10] Changwon Proving Ground Construction Manual, 1996, Agency for Defense
Development, GWSD-809-960634
[11] Shabana A, 1989 ''Dynamics of Multibody Systems'', John Wiley & Sons, New York
RecurDyn
2 1
2
2
2
1
+ +
= n ]
(4)
METHODS OF FINITE CONTACT NODES FOR GROUND INTERACTIONS
Unlike wheel and surface contact, the interactions between track link and
ground are very complicated problems. This is because the track link has
8-9
irregular contact geometry and different material properties. Due to large number
of track links of each track subsystem, commonly used contact theory of
surface to surface interactions in finite element community can not be employed
for this work. Choi[17] suggested that element free finite contact nodes were
distributed on the contact surface of track link, which have their own stiffness
and damping characteristics. The relative indentations of each node were
monitored and positions are restored. The use of element free finite contact node
methods demonstrated clearly the computational efficiency for dynamic analysis
of track system. Based on the method developed by Choi[17], the interactions
between track link surface and triangular patch surface are developed in this
investigation.
Figure 7. Interaction between track shoe body and triangular patch element
Figure 7 shows the interaction between finite contact nodes of track link and
triangular patch surface. The perpendicular deformation scalar of contact
node of link on patch plane can be defined as
k
d
j i k
RecurDyn
1
= (5)
where is shown in Fig. 7 and unit vector is defined in Eq. (4). The
criterion of necessary condition for the contact to occur of node , which is not
sufficient, is
P 1
r
k
n
j
>
seperate d
contact d
ij
k
ij
k
: 0
: 0
(6)
If this conditions is satisfied, the position vector shown in Fig. 7 is used
to compute the node location whether contact point
jk
B
r
B of node is on the
patch plane . The position vector can be written as
j
k
jk
B
r
k k
ij
p
i i jk
B
d n u A R r
+ =
(7)
where is the transformation matrix associated with the orientation
coordinates of link and
i
A
i
ij
p
u is the local position vector of node in the
track link coordinate system. On the other hand, using scalar triple product if one
of the following conditions is satisfied
j
(8)
3 31
2 23
1 12
3 31
2 23
1 12
k
B
k
B
k
B
k
B
k
B
k
B
or
n r r
n r r
n r r
n r r
n r r
n r r
, then the node of link is in contact with patch element . j i k
8.4. EQUATIONS OF MOTION
In this investigation, the relative generalized coordinates are employed in
order to reduce the number of equations of motion and to avoid the difficulty
8-11
associated with the solution of differential and algebraic equations. Since the
track chains interact with the chassis components through contact forces and
adjacent track links are connected by compliant force elements, each track chain
link in the track chain has six degrees of freedom which are represented by three
translational coordinates and three Euler angles. Recursive kinematic equations
of tracked vehicles were presented by [8] and the equations of motion of the
chassis are given as follows :
) (
r
i
T r
i
T
q B M Q B q MB B &
&
& & = (9)
where and are relative independent coordinates, velocity
transformation matrix, and is the mass matrix, and is the generalized
external and internal force vector of the chassis subsystem, respectively. Since
there is no kinematic coupling between the chassis subsystem and track
subsystem, the equations of motion of the track subsystem can be written simply
as
r
i
q B
M Q
t t t
Q q M =
& &
(10)
where , and Q denote the mass matrix, the generalized coordinate
and force vectors for the track subsystem, respectively. Consequently, the
accelerations of the chassis and the track links can be obtained by solving Eqs.
(9) and (10)..
t
M
t
q
t
8.5. EXTENDED BEKKERS SOIL MODEL FOR MULTIBODY TRACK
SYSTEM
The interactions between track link and soil used in this investigation consist
of the normal pressure-sinkage and shear stress-shear displacement relationships.
Bekker[13] developed the bevameter technique to measure terrain characteristics
by the plate penetration and shear tests. He also proposed the equation for
pressure-sinkage relationship, given by
RecurDyn
k
ij
s k
ij
p
ij
F F t n F
+ =
(24)
where and t
c k f
&
&
&
+ =
(5)
where k and are the spring and damping coefficients which are determined
by an experimental method, respectively and the is time differentiation of .
The exponents m and generates a non-linear contact force and the
exponent yields an indentation damping effect. When the penetration is
very small, the contact force may be negative due to a negative damping force,
which is not realistic. This situation can be overcome by using the indentation
damping exponent greater than one.
c
&
1 2
m
3
m
9-9
The friction force is obtained by
n f
f f =
(6)
where is the friction coefficient and its sign and magnitude can be
determined from the relative velocity of the pair on contact position.
9.6. KINEMATICS AND EQUATION OF MOTION FOR THE RECURSIVE
FORMULAS
A contact search algorithm is proposed in the previous sections. The proposed
method makes use of the relative position and orientation matrix for a contact
pair. This section presents the relative coordinate kinematics for a contact pair as
well as for joints connecting two bodies.
Translational and angular velocities of the z y x frame in the
frame are respectively defined as
Z Y X
w
r
&
(7)
Their corresponding quantities in the z y x frame are defined as
=
w A
r A
w
r
Y
T
T
& &
(8)
where is the combined velocity of the translation and rotation. The
recursive velocity and virtual relationship for a pair of contiguous bodies are
obtained in [17] as
Y
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Y B Y
&
(9)
where denotes the relative coordinate vector. It is important to note that
matrices and are only functions of the . Similarly, the
1)i (i
q
1)i1 (i
B
1)i2 (i
B
1)i (i
q
RecurDyn
+ = q B Z B Z
(10)
If the recursive formula in Eq. (9) is respectively applied to all joints, the
following relationship between the Cartesian and relative generalized velocities
can be obtained:
q B Y
&
=
(11)
where is the collection of coefficients of the and B
1)i (i
q&
[ ]
T
1 nc
T T
2
T
1
T
0
=
n
Y , , Y , Y , Y Y K (12)
[ ]
T
1 nr
T
) 1 (
T
12
T
01
T
0
=
n n
q , , q , q , Y q
&
K
& & &
(13)
where nc and nr denote the number of the Cartesian and relative coordinates,
respectively. Since in Eq. (11) is an arbitrary vector in q&
nr
R , Eqs. (9) and (11),
which are computationally equivalent, are actually valid for any vector
such that
nr
R x &
x B X & =
(14)
and
1)i - (i 1)i2 - (i 1) - (i 1)i1 - (i i
x B X B X + =
(15)
where is the resulting vector of multiplication of and . As a
result, transformation of into is actually calculated by
recursively applying Eq. (15) to achieve computational efficiency in this
research.
nc
R X B x
nr
R x
nc
R Bx
Inversely, it is often necessary to transform a vector in G
nc
R into a new
vector in G B g
T
=
nr
R . Such a transformation can be found in the generalized
force computation in the joint space with a known force in the Cartesian space.
The virtual work done by a Cartesian force is obtained as follows.
nc
R Q
9-11
Q
Z W = (16)
where must be kinematically admissible for all joints in a system.
Substitution of
Z
q B Z = into Eq. (16) yields
* T T T
Q q Q B q W = = (17)
where . Q B Q
T *
T
= + =
&
( (18)
where the is the Lagrange multiplier vector for cut joints [18] in
m
R
and represents the position level constraint vector in
m
R . The and Q
are the mass matrix and force vector in the Cartesian space including the contact
forces, respectively.
M
The equations of motion and the position level constraint can be implicitly
rewritten by introducing v q= & as
0 a , v q F = ) , , (
(19)
0 q = ) (
(20)
Successive differentiations of the position level constraint yield
(21)
0 v v q
q
= = ) , (
&
0 v v v q
q
= = & &
& &
) , , (
(22)
RecurDyn
+ +
+ +
) (
) (
) (
=
) (
) (
, ,
,
, ,
) (
2 0
T
0
1 0
T
0
&
&
& &
&
&
(23)
where [ ]
T T T T T
, v , v , q x & = , , and are determined by the
coefficients of the implicit integrators and is an nr matrix such
that the augmented square matrix is nonsingular.
0
U
T
0
1
0
U m) (nr
The number of equations and the number of unknowns in Eq. (23) are the
same, and so Eq. (23) can be solved for . Newton Raphson method can be
applied to obtain the solution .
n
x
n
x
H x H
x
=
(24)
1,2,3,... i ,
i 1 i
= + =
+
x x x
(25)
0
0 U U 0
0 0 U U
0
0 0
0 0 0
F F F F
H
T
0
a v q
v q
q
q q q q
x
=
=
T
0 0
T
0 0
T
0
=
i
i
i
r
Y
&
(2)
(
=
i
i
r
Z
(3)
Their corresponding quantities in the
i i i
z y x frame are defined as
10-5
(4) (
i
T
i
i
T
i
A
r A
&
=
i
i
i
r
Y
&
(5) (
=
i
T
i
i
T
i
i
i
A
r A
r
Z
i
) 1 (
x
i
) 1 (
z
i
i
O
) 1 ( i
O
i
r
) 1 ( i
r
i i ) 1 (
s
A pair of contiguous bodies is shown in Figure 2. Body 1 i is assumed to
be an inboard body of body and the position of point is i
i
O
1) i(i 1)i (i 1)i (i 1) (i i
+ + = s d s r r
(6)
By using Eq. (5), the angular virtual displacement of body in its local
reference frame is
i
(7)
1)i (i 1)i (i
T
1)i (i 1) - (i
T
1)i (i i
+ = q H A A
where is determined by the axis of rotation and is defined as
i i ) 1 (
H
1)i (i
A
i
T
1) (i 1)i (i
A A A
=
(8)
Taking variation of Eq. (6) yields
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
1)i (i 1)i (i
1)i (i
T
1) i(i 1)i (i 1)i (i
1)i (i
T
1) (i
1)i (i
T
1) - i(i 1)i (i 1)i (i 1)i (i
1)i (i
T
1) (i
1)i (i
T
i
)
~
) ( (
)
~
~
~
(
1)i (i
+ +
+
=
q H A s A d A
A s A d s A
r A r
q
(9)
where symbols with tildes denote skew symmetric matrices comprised of their
vector elements that implement the vector product operation and denotes
the relative coordinate vector.
1)i (i
q
Combining Eqs. (7) and (9) yields the recursive virtual displacement equation
for a pair of contiguous bodies
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Z B Z
(10)
where
(
+
(
(
I 0
A s A d s I
A 0
0 A
B
)
~
~
~
( 1)i (i
T
1) i(i 1)i (i 1)i (i 1)i (i
1)i1 (i
1)i (i
1)i (i
T
T
(11)
(
(
+
(
(
1)i (i
1)i (i
1)i (i
T
1) i(i 1)i (i 1)i (i
1)i (i
1)i (i
1)i2 (i
~
) (
1)i (i
H
H A s A d
A 0
0 A
B
q
T
T
(12)
It is important to note that matrices and are functions of only
relative coordinates of the joint between bodies
1)i1 (i
B
1)i2 (i
B
(i 1) and . As a
consequence, further differentiation of the matrices and in Eqs.
(11) and (12) with respect to other than yields zero. The virtual
displacement relationship between the absolute and relative coordinates for the
whole system can be obtained by repetitive application of Eq. (10) as
i
(i 1)i1 (i
B
1)i
1)i2
B
(i
q
q B Z =
(13)
where is the velocity transformation matrix with relationship between
Cartesian and relative coordinates. The relationship between Cartesian velocity
B
10-7
Y q& and relative velocity can be derived in the same manner.
+
=
M
=
q B Y
&
=
(14)
10.3. EQUATIONS OF MOTION
The variational form of the Newton-Euler equations of motion for a
constrained multibody system is
0 Q Y M Z
=
) (
T T
&
(15)
where and Q are the mass matrix and general force vector in Cartesian
space, respectively.
M
Z must be kinematically admissible for all joints except
cut joints [12]. In the equation, and , respectively, denote the constraint
equations and the corresponding Lagrange multiplier in in which m is the
number of the constraint equations. Substituting the virtual displacement
relationship and acceleration relationships into Eq. (15) yields (see
Ref. [5])
m
R
q B q B &
&
& & + = Y
&
n * T *
R = + F 0 Q q M F
q
& &
(16)
where n is the number of generalized coordinates and the mass matrix
and force vector are defined as
*
M
*
Q
B M B
T
=
*
(17)
) (
*
q B M Q B Q
T
&
&
(18)
A recursive method has been proposed to compute Eqs. (17) and (18) in Ref.
[7].
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
10.4. ELIMINATION OF LAGRANGE MULTIPLIERS AND LINEARIZATION
OF THE EQUATIONS OF MOTION
The relative coordinates can be partitioned into dependent coordinates
and independent coordinates q such that the sub-Jacobian is well-
conditioned. Variational form of the cut constraint equations can be written as
q
D
q
I
D
q
0 q q
q q
= + =
I D
I D
(19)
The can be obtained from Eq. (19) as
D
q
I
1
D
I D
q q
q q
= (20)
By using the relationship in Eq. (20), is represented as
I
1
D
I D
q q
q q
=
I
q N q = (21)
where
(
=
I
N
q q
I D
1
(22)
Direct calculation of shows that is the null space of as
T T
q
N N
q
| | 0
I N
q
q
q q q
=
(
(
=
T
T
T 1 - T T T
I
D
D I
) (
(23)
As a result, pre-multiplication of Eq. (16) by gives
T
N
0 Q N q M N F = =
* T * T *
& &
(24)
where Lagrange multiplier term was eliminated since is the null space of . N
q
10-9
However, the equations of motion are dependent on not only the dependent variables ,
and but also independent variables , and q& . Taking variation of Eq. (24)
yields
*
F
q
*
| |
&
*
q
q
q
+
+
=
q
q
q
& &
&
F q
q& q& & q q& &
D
D D I
q+
&
0
+
=
I
0
0
0
I
F
q& &
*
q
& &
(
(
(
(
(
0
0
0
I
0
T
q& q
q F q F
q
= + =
& &
&
* *
(25)
Equation (25) can be rewritten in a matrix form as
0
q
q
q
F F F
q q q
=
& &
&
& &
* *
(26)
Variations of position, velocity and acceleration level constraints are
0 q
q
0
q q q
q q
q
=
&
& & &
&
&
(27)
Appending the trivial identity relationships for the variations of independent
coordinates, velocities and accelerations to Eq. (19) yields
(
(
(
(
(
I
I
I
q
q
q
0
0
0
I
0
0 0
q q q
q q
q
& &
&
& & &
&
(28)
Equation (28) is solved for the { } q& & and substituted into the
linearized equations of motion in Eq. (26) to yield the following linearized
equations of motion only in terms of the variations of independent coordinates,
velocities and accelerations:
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
(29)
| | 0
q
q
q
I
0 0 0
0 0 0
0 0 0
I
0
0 0
F F F
q q q
q q
q
q q q
=
(
(
(
(
(
(
(
(
(
I
I
I
1
* * *
& &
&
& & &
&
& & &
Direct comparison of Eq. (29) and the following linearized equations of
motion yields the M
)
, C
)
and K
)
matrices:
0 q K q C q M F
q
= + + =
I I I
*
*
& & &
(30)
10.5. NUMERICAL EXAMPLES
10.5.1. FOURBAR MECHANISM WITH A SPRING
Figure 3 shows a four bar mechanism with a spring. The system consists of
four revolute joints and one spring and their material properties are defined in
Table 1. As a result, three generalized coordinates, , and are defined
for the first three revolute joints and the remaining one revolute joint is defined
as a cut joint. The constraint equations are introduced from the cut joint.
1
Link
Link
Link
500
400
Cut joint
Figure 3 A four-bar mechanism with a spring
10-11
Table 1 Material property of bodies and a spring
Mass (kg)
Inertia Moment
(kg*mm^2)
Link A 7.707 161760.83
Link B 3.946 53005.79 Body
Link C 7.707 161760.83
Stiffness (N/mm) Damping (N*sec/mm)
Spring
10.0 0.0
Dynamic analysis of the mechanism is performed to obtain the time domain
response. FFT of the time response is performed to extract dominant frequency
domain response. Figures 4 and 5 show the time and frequency responses,
respectively.
The proposed linearization method is applied for the system. The dominant
frequency and corresponding mode shape are shown in Figure 6 and Table 2.
The frequency obtained from the proposed method and that obtained from FFT
analysis of the time domain responses are shown to be very close, which
validates the proposed method.
Figure 4 Angle of link C in time domain
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
Figure 5 Response in frequency domain
Figure 6 Mode shape of fourbar mechanism
Table 2 Undamped natural frequency and mode shape from the proposed method
Undamped Natural
Frequency (Hz)
Mode
1
2
3
5.040164E+00
5.773503E-01 -5.773503E-01 5.773503E-01
10-13
10.5.2. CANTILEVER BEAM DRIVEN BY A MOTION
The system characteristics of a rotating cantilever beam differ from those of
beam in a static state, because the stiffness of the beam is changed by a
centrifugal force due to the rotational motion. (see Ref. [13]). A cantilever beam
rotating with the angular velocity is shown in Figure 7.
Figure 7 A rotating cantilever beam
Length of the beam is 6.8 m, density of the material is 14705.88 kg/m
3
,
Young's modulus of the material is 7.010
8
N/m
2
. Area of the cross section is
0.002 m
2
, the moment of inertia 4.010
-7
m
4
. The beam is divided into 21
lumped mass and 20 beam elements. Figure 8 shows the lowest three natural
frequencies of the rotating beam. As the angular speed increases, the bending
natural frequencies are shown to be increased.
Figure 8 The relationship between angular velocity and natural frequencies
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
10.5.3 A SPRING SYSTEM WITH 2 D.O.F.
A spring model shown in Fig. 9 is a system with two D.O.F, and the system
has two masses, joints and spring elements. Their material properties, spring and
damping coefficients are shown in Table 3.
Figure 9 A spring model
Table 3 Material properties, spring and damping coefficients
Mass1 5 Kg
Mass2 3 Kg
Length of m1 300 mm
Spring coefficient (k1) 10 N/mm
Spring coefficient (k2) 20 N/mm
If the rotational angle is small, sin and the equation of motion of this
system can be derived as:
0 =
(
(
(
(
+
+
(
y
k k
l
k
l
k
l
k l
y
m
I
2 2
2 2
2
1
2
2
2
2 4
0
0
& &
& &
(31)
From Table 3, Eq. (31) can be replaced as:
(32)
0
20000 3000
3000 1350
3 0
0 15 . 0
=
(
+
(
y y
& &
& &
10-15
The characteristic equation of this spring system is derived from Eq (32).
0
3 20000 3000
3000 15 . 0 1350
=
(33)
Also, the analytic natural frequencies can be computed as:
) Hz ( 76 . 17 f sec) / rad ( 6 . 111 12455
) Hz ( 019 . 9 f sec) / rad ( 66 . 56 3211
2 2
1 1
= = =
= = =
(34)
Finally, the eigenvalues of this spring system is validated shown in Table 4.
Table 4 Eigenvalues of spring model
Undamped Natural Frequency (Hz)
Mode
number
RecurDyn/Eigenvalue Analytic solution
1 9.01862E+00 9.019
2 1.77621E+01 17.76
10.5.4 A CANTILEVER BEAM
Two cantilever beam models shown in Figs. 10 and 11 have a fixed-free end
condition and ten lumped masses. One is modeled by using ten beam force
elements and the other is modeled by using one flexible body of RecurDyn. The
flexible beam model is originally generated in ANSYS. The material properties
and geometry conditions of the beam are shown in Table 4.
Figure 10 Beam model using RecurDyn/Beam element
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
Figure 11 Beam model using RecurDyn/Flexible body element
Table 5 The material properties and geometry conditions of beam
Length 0.4 m
Mass 3.9888 Kg
Youngs
modulus
9
10 1 N/m
2
Inertia of area
8
10 215 . 1
m
4
Area 0.0018 m
2
In Ref. [14], the analytic natural frequencies of these beams are computed as:
4
2
1
875 . 1
AL
EI
=
,
4
2
2
694 . 4
AL
EI
=
,
4
2
3
855 . 7
AL
EI
=
(35)
By replacing Eq. (35) with Table 5, the natural frequencies can be computed as:
8601 . 3 2537 . 24 875 . 1
4
2
1
= = =
n
f
AL
EI
1929 . 24 0085 . 152 694 . 4
4
2
2
= = =
n
f
AL
EI
7477 . 67 6714 . 425 855 . 7
4
2
3
= = =
n
f
AL
EI
10-17
Finally, the eigenvalues of this beam model is validated shown in Table 6.
Table 6 Eigenvalues of cantilever beam model
Undamped Natural Frequency (Hz)
Mode
number
Beam element Flexible Body Analytic solution
1 3.84002E+00 3.84259E+00 3.8426
2 2.37455E+01 2.38154E+01 23.8154
3 6.55744E+01 6.60152E+01 66.0152
4 1.26483E+02 1.28016E+02 128.016
5 2.05481E+02
6 2.65264E+02
In addition, RecurDyn can show the mode shapes of the beam model through 3D
animation, as shown in Figs. 12 and 13.
(a) 1
st
mode shape (b) 2
nd
mode shape
(c) 3
rd
mode shape
Figure 12 The mode shapes of model using RecurDyn /Beam element
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
(1) 1
st
mode shape (2) 2
nd
mode shape
(3) 3
rd
mode shape
Figure 13 The mode shapes of model using RecurDyn/Flexible body
10.6. CONCLUSIONS
In this paper, a linearization method for constrained multibody systems is
proposed for the non-linear equations of motion employing the relative
coordinates. Null space of the constraint Jacobian is pre-multiplied to the
equations of motion to eliminate the Lagrange multipliers and to reduce the
number of equations. The set of differential equations are perturbed in terms of
all relative positions, velocities and accelerations. The position, velocity and
acceleration level constraints are perturbed to express the variations of all
relative positions, velocities and accelerations in terms of the variations of
independent positions, velocities and accelerations, which are substituted into the
perturbed equations of motion. The equations of motion perturbed with respect
to the , and q& finally become the corresponding equations perturbed with
respect to the , and . Eigenvalues and eigenvectors are then computed
from the equations of motion perturbed with respect to the , and . The
proposed method is implemented in a commercial program RecurDyn.
Numerical results obtained from the proposed method are in good agreement
with the results reported in the literature and obtained by other methods.
q q& &
q&
I
q
I I
q& &
I
q
I
q&
I
q& &
10-19
REFERENCES
1. Sohoni VN, Whitesel J. Automatic Linearization of Constrained Dynamical
Models. ASME. Journal of Mechanism, Transmission, and Automation in Design,
Vol. 108, pp 300-304, 1986.
2. Neuman CP, Murray JJ. Linearization and Sensitivity Functions of Dynamic
Robot Models. IEEE Transactions on Systems, Man, and Cybernetics, Vol. SMC-
14, No.6, pp.805-818, 1984.
3. Balafoutis CA, Misra P, Patel RV. ecursive Evaluation of Linearized Dynamic
Robot Models. IEEE Journal of Robotics and Automation, Vol.RA-2, No.3,
pp.146-155, 1986.
4. Gontier C, Li Y. Lagrangian Formulation and Linearization of Multibody System
Equations. Computers & Structures, Vol.57. No.2, pp.317~331, 1995.
5. Bae DS, Haug EJ. A Recursive Formulation for Constrained Mechanical System
Dynamics: Part I, Open Loop Systems. Mech. Struct. and Machines, Vol. 15, No.
3, pp.359-382, 1987.
6. Bae DS, Haug EJ. A Recursive Formulation for Constrained Mechanical System
Dynamics: Part II, Closed Loop Systems. Mech. Struct. and Machines, Vol. 15,
No. 4, pp. 481-506, 1987.
7. Bae DS, Han JM, Yoo HH. A Generalized Recursive Formulation for
Constrained Mechanical System Dynamics. Mech. Struct. & Mach., Vol. 27(3),
pp. 293-315, 1999.
8. Bae DS, Lee JK, Cho HJ, Yae H. An Explicit Integration Method for Realtime
Simulation of Multibody Vehicle Models. Computer Methods in Applied
Mechanics and Engineering, Vol. 187, pp. 337-350, 2000.
9. Bae DS, Han JM, Choi JH, Yang SM. A Generalized Recursive Formulation for
Constrained Flexible Multibody Dynamics. International Journal for Numerical
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
Methods in Engineering, Vol. 50, pp. 1841-1859, 2001.
10. Ryu HS, Bae DS, Choi JH, Shabana AA. A Compliant Track Link Model for
High-speed, High-mobility Tracked Vehicles. International Journal for
Numerical Methods in Engineering, Vol. 48, pp. 1481-1502, 2000.
11. Kim HW, Bae DS, Choi KK. Configuration Design Sensitivity Analysis of
Dynamics for Constrained Mechanical Systems. Computer Methods in Applied
Mechanics and Engineering, Vol. 190, pp. 5271-5282, 2001.
12. Wittenburg J. Dynamics of Systems of Rigid Bodies. B. G. Teubner Stuttgart,
1977.
13. Southwell R, Gough F. The Free Transverse Vibration of Airscrew Blades.
British A.R.C. Reports and Memoranda No. 766, 1921.
14. L. Meirovitch, Analytical Methods in Vibrations, MACMILLAN, 1967.
10-21
RecurDyn
/ Solver THEORETICAL MANUAL
LINEARIZED EQUATIONS OF MOTION FOR MULTIBODY SYSTEMS WITH CLOSED LOOPS
11
NONLINEAR DYNAMIC MODELING OF
SILENT CHAIN DRIVE
11.1. INTRODUCTION
Chain drives are widely used in the power transmission applications in the
automotive field for a long time because they are capable of transmitting large
power at high efficiency and low maintenance cost. However, the noise and
vibrations created by chain drives have always been major problems, especially
for higher speed, lighter weight, and higher quality. Noise and vibrations in
chain systems are largely caused by chordal(polygonal) action and impacts
between chain and sprocket. The links of the chain form a set of chords when
wrapped around the circumference of the sprocket. As these links enter and leave
the sprocket, they impart a jerky motion to the driven shaft by chordal action.
The chordal action causes chain span longitudinal and transverse vibrations.
Whereas, impact between sprocket and link excites high frequency vibration and
is a major source of noise in chain drives at high speeds. In order to minimize
such problems, silent chains are introduced in many camshaft drives of
motorcycle/automobile engines and the primary drive between the engine and
transmission, as well as in other high-speed applications. It is also used with the
object of increasing chain life. However, in spite of the widespread use of silent
chain drives, surprisingly little works have been published about their dynamic
analysis. This may be due to three major difficulties; the first is the complexity
of the contact algorithms among components, the second is small integration step
size resulting from the impulsive contact forces and the use of stiff compliant
elements to represent the joints between the chain links, and the third is the large
number of the system equations of motion to solve.
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
Camshaft Sprocket
(Idler Sprocket)
Crankshaft
Sprocket
Pivot Guide
Tensioner
Fixed Guide
Figure 1. Silent Chain Drive Model of Automotive Engine
Chen and Freudenstein [1] presented a kinematic analysis of chain drive
mechanism with the aim of obtaining insight into the phenomena of chordal
action, with the associated impact and chain motion fluctuation. Veikos and
Freudenstein [1] developed a lumped mass dynamic model based on Lagranges
equations of motion and showed chain drive dynamics and vibrations. Wang [3,
4] investigated the stability of a chain drive mechanism under periodic sprocket
excitations and studied the effect of impact intensity in their axially moving
roller chains. Kim and Johnson [5, 6] developed a detailed model of the roller-
sprocket contact mechanics that allowed the first determination of actual
pressure angles and a multi-body dynamic simulation. This investigation is based
on Kanes dynamic equations. Choi and Johnson [7, 8] investigated the effects of
impact, polygonal action, and chain tensioners into the axially moving chain
system and showed the transverse vibration of chain spans. Quite recently Ryu et
11-3
al [9] developed very detailed chain models including contact forces for links,
sprockets and idlers with special application to large-scaled civilian and military
tracked vehicles. There has been some design analysis in the view of dynamic
behaviors of silent chain in powertrain industry and commercial software [10].
However they showed some primitive dynamic analysis and design of silent
chain system because it has high frequency contact forces, speedy revolution and
large number of bodies.
The purpose of this work is to investigate and suggest the dynamic modeling
and analysis of silent chain drive mechanism with high speed revolution using
multibody dynamic techniques. In this investigation, numerical skills of
multibody chain dynamic analysis are employed and showed very good
agreement of physical phenomenon of silent chain system. Dynamic tension,
impact forces, and vibration of chain links are explored for the sake of
understanding dynamic behaviors of the chain system.
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.2. MULTIBODY MODELING OF SILENT CHAIN DRIVE
As shown in Fig. 1, in general a chain drive mechanism has four main
components, which are sprockets, chain links, guides and tensioner element. The
sprockets can be recognized as drive sprockets and idler sprockets. The chain
link can includes link plates, guide plates, and pins. The tensioner element
maintains stable tension during operation by adjusting pressure force to the chain
link system. While roller chain mechanism has engagements between pins and
sprocket, since silent chain mechanism engages between chain link teeth and
sprocket teeth, there is much less chodal vibrations and can transmit the power
more quietly. In this investigation the dynamic analysis and numerical modeling
techniques are presented by using multibody methods.
11.2.1 SPROCKET
The sprockets of the chain system are interacted by the introduced contact and
friction forces acting on between the chain and the sprocket teeth. The crank
sprocket of the system is driven by motion constraint. This motion constraint can
be constant or time dependent. In this investigation the sprocket is modeled as a
rigid body and attached on ground by revolute joint. The geometry of sprocket
teeth profiles consists of a series lines and arcs with different length and radii as
shown in Figure 2. The sprocket of silent chain is shaped more like a gear than
one of roller chain.
Figure 2. Geometry of Silent Chain Sprocket
11-5
11.2.2 SILENT CHAIN LINK
Roller chains, although having excellent wear and strength capability, are
inherently noisy and oscillatory. As a result, inverted-tooth chain mechanisms
were developed in order to reduce the forcing function of the noise-producing
mechanism. The difference in noise performance between silent and roller chains
can be attributed to the manner in which they engage and disengage the sprocket
teeth. After the sprocket tooth initially contacts the chain link, and as the
engagement proceeds, a combination of rolling and sliding motion occurs
between the tooth and link contacting surfaces. Such an engagement mechanism
effectively spreads the engagement time over a significant interval, thereby
minimizing tooth/link impact and its inherent noise generation.
A silent chain consists of several layers of links connected with pins. Since
there is no advantage for the modeling of pins and multi layer links as separate
components, in this investigation these multi layer links are treated as a rigid
body with mass and inertia property which takes into account the effects of the
pins. An individual silent chain link looks much different comparing to a roller
chain link. The geometry of link profile, which resembles a tooth, consists of
several lines and arcs in a complex arrangement as shown in Fig. 3. As used in
the roller chain from previous work, the connections between links are modeled
with bushings to account for the flexibility in this investigation. Though the
sprockets of the silent chain serve in the same function of the rolling chain
system, however, they are designed to engage specifically with the links of the
silent chain with different tooth contour as illustrated.
RecurDyn
/ Solver THEORETICAL MANUAL
Figure 3. Components of Silent Chain Link System
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.2.3 TENSIONER AND CHAIN GUIDE
In a chain drive system, the chain guide ensure that the chain remains on the
path, while tensioner try to keep constant tension of chain system. Usually the
chain guide directs the tight chain portion which runs from the driven sprocket to
the driving sprocket. Conversely, the chain arm directs the slack portion of the
chain which runs opposite (from the driving to the driven). The pivot guide also
serves to distribute the force on the chain from the hydraulic tensioner to
maintain certain level of chain tension. In this investigation hydraulic tensioning
force model is used which is offered from hydraulic tensioner manufacturer.
The chain guide and the chain arm are both modeled as separate rigid body
parts. The geometric profiles of the guides consist of a series arcs with different
radii. If desired, the chain guides can be modified so that they are constructed as
flexible bodies for the calculation of vibrations, stresses and bending moments,
etc.
11.2.4 EQUATIONS OF MOTION AND INTEGRATION
Since the chain system interacts with the frame component through the contact
forces and adjacent chain links are connected by compliant force elements, each
chain link in the chain system has six degrees of freedom which are represented
by three translational coordinates and three Euler angles. The equations of
motion of the frame structure such as sprockets that employs the velocity
transformation defined by Choi [9] are given as follows :
) (
r
i
r
i
q B M Q B q MB B
T T
&
&
& &
=
(1)
where and B are relative independent coordinates and velocity
transformation matrix of the engine chassis subsystem, and M is the mass matrix,
and Q is the generalized external and internal force vector of the frame structure
subsystem, respectively. Since there is no kinematic coupling between the frame
structure subsystem and chain subsystem, the equations of motion of the chain
subsystem can be written simply as
r
i
q
11-7
t t t
Q q M =
& &
(2)
where , and Q denote the mass matrix, the generalized coordinate
and force vectors for the chain subsystem, respectively. Consequently, the
accelerations of the frame structure components and the chain links can be
obtained by solving Eqs. (1) and (2).
t
q M
t t
Many different types of integration methods can be employed for solving the
equations of motion for mechanical systems. Explicit methods have small
stability region and are often suitable for smooth systems whose magnitude of
eigenvalues is relatively small. Contrast to the explicit methods, implicit
methods have large stability region and are suitable for stiff systems whose
magnitude of eigenvalues is large. In the model used in this investigation, a
contact between two bodies is modeled by compliance elements. Lumped
characteristics of the spring and damper must represent elastic and plastic
deformations, and hysterisis of a material. Such characteristics may include
artificial high frequencies which are not concern of a design engineer. Unless
such artificial high frequency is filtered, an integration stepsize must be reduced
so small that integration cant be completed in a practical design cycle of a
mechanical system. To achieve this goal, the implicit generalized-alpha method
[9, 11] has been employed to filter frequencies beyond a certain level and to
dissipate an undesirable excitation of a response in this investigation. One of the
nice advantages of the generalized-alpha method is that the filtering frequency
and dissipation amount can be freely controlled by varying a parameter in the
integration formula. As a result, the generalized-alpha method is the most
suitable integration method for integrating the equations of motion for stiff
mechanical systems.
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.3. CONTACT FORCE ANALYSIS
The contact collision algorithms for a silent chain drive used in this
investigation are composed of three main routines such as search routines for,
sprocket teeth and chain link contact, chain guide and chain link contact, and
side guide of chain link and sprocket contact. The contact positions and
penetration values are defined from the kinematics of components in searching
routines. Thereafter a concentrated contact force is used at the contacted position
of the contact surface of the bodies. A detailed discussion on the formulation of
the contact collision is represented in this section, respectively. Efficient search
algorithms should be considered seriously because there are large number of
chain link bodies and sprocket which take long time to search all the bodies
whether they are in contact or not.
11.3.1 STRATEGE OF CONTACT SEARCH
For the efficient search of the sprocket-chain link contact kinematics, the
contact search algorithm is divided by pre-search and post-search. In the pre-
search, bounding circle relative to sprocket center is defined. All of chain links
are employed to detect a starting link and ending link which has a possibility of
sprocket contact. Then, chain links from starting link are investigated the
engagement with sprocket valley. Post-search means a detailed contact
inspection for chain links in a bounding circle. Once a starting and ending link is
found at one time through pre-search prior to analysis, only detailed search is
carried out by using the information of starting link and ending link from the
next time step. There are four contact possibilities such as, arc-line, arc-point,
arc-arc and line-point contact for interaction between the sprocket teeth and
chain link.
11-9
11.3.2 LINE-ARC CONTACT
X
Y
Z
i
X
i
Y
i
Z
i
t
X
i
t
Y
i
t
Z
i
R
i
t
u
j
p
X
j
Y
j
Z
j
R
j
p
u
p
t
ij
k
u
a
j
p
Y
j
X
Link arc
Tooth line
Chain link
coordinate system
Sprocket
coordinate system
Global coordinate system
Tooth
coordinate system
Figure 4. Line-Arc Contact Kinematics
The contact conditions between the sprocket teeth line segment and the chain
link arc segment can be determined. A coordinate system is attached
to each of the sprocket surfaces shown in Fig. 4. The surfaces of the tooth line
are approximated by plane surfaces and the axis of each surface coordinate
system is assumed to be parallel to the tooth surface. The surfaces of the chain
link arc segment are approximated by plane surfaces and the axis of each arc
origin coordinate system is assumed to be directed to the starting arc point from
arc origin. The orientation of the tooth surface coordinate system with
respect to the global system is defined by
i
t
i
t
i
t
Z Y X
j
p
X
i
t
X
k
i
k
i i
t
A A A =
(3)
RecurDyn
/ Solver THEORETICAL MANUAL
where is the transformation matrix that defines the orientation of the
coordinate system of the sprocket and is the transformation matrix that
defines the orientation of the tooth line surface coordinate system
with respect to the sprocket coordinate system. The orientation of the link arc
coordinate system with respect to the global system is defined by
i
A
i
i
k
A
k
i
t
i
t
i
t
Z Y X
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
j
l
j j
a
A A A = (4)
where is the transformation matrix that defines the orientation of the chain
arc surface coordinate system with respect to the chain link
coordinate system.
j
l
A
l
j
p
j
p
j
p
Z Y X
The global position vector of the coordinate system of the tooth surface is
defined as
k
i
t
i i i
t
u A R r + = (5)
where
i
R is the global position vector of the coordinate system of the
sprocket and i
i
t
u is the position vector of point with respect to the origin
of the sprocket coordinate system .
t
i i i
Z Y X
The global position vector of the center of the chain link arc segment, denoted
as point p , can be defined as
j
p
j j j
p
u A R r + = (6)
where
j
R is the global position vector of the origin of chain link j , is
the transformation matrix of chain link
j
A
j and
j
p
u is the position vector of
point p defined in the chain link coordinate system .
j j j
Z Y X
The position vector of the center of the arc of chain link j with respect to the
origin of the tooth line surface coordinate system can be defined in the global
coordinate system as
i
t
j
p
ij
k
r r u = (7)
The components of the vector along the axes of the tooth line surface
coordinate system are determined as
ij
k
u
[ ]
ij
k
T
i
t
T
ij
z
ij
y
ij
x
ij
u u u u A u = = (8)
Necessary but not sufficient conditions for the contact to occur between the
chain link arc and the sprocket tooth line surface are k
k
ij
x
l u 0
(9)
11-11
p t
ij
z p t
w w u w w +
(10)
r
ij
y
u
(11)
where is the length of the tooth line surface , is half width of the
tooth and is half width of the chain link outer plate and
k
l
w
k
t
w
p
r is the radius of
the chain link arc. If the above conditions are satisfied, it has to be checked if
contact point is existed in the arc range for the next step.
g d =
ji
k
, where (12) ] [ h g f A =
i
k
[ ]
ji
k
T
j
a
T
ji
z
ji
y
ji
x
ji
k
d d d d A d = = (13)
) , ( atan2
ji
x
ji
y k
d d = (14)
a k
0 (15)
where is the opposite signed normal vector of the tooth line surface ,
is the angle of with respect to the link arc segment coordinate system
and is the angle of arc segment.
ji
k
d k
k
ji
k
d
a
ij
y
ij
u r = (16)
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.3.3. ARC-POINT CONTACT
X
Y
Z
i
X
i
Y
i
Z
i
t
X
i
t
Y
i
t
Z
i
R
i
t
u
j
Y
j
Z
j
R
j
p
u
p
t
ij
k
u
a
j
X
Link point
Tooth arc
Chain link
coordinate system
Global
coordinate system
Tooth
coordinate system
Sprocket
coordinate system
Figure 5. Arc-Point Contact Kinematics
There are two arc-point contact possibilities such as convex arc vs. point and
concave arc vs. point contact for arc-point interaction between the sprocket teeth
and chain link. Figure 5 shows a convex arc-point contact kinematics. The arc-
point contact conditions between the sprocket teeth and the chain link can be
determined. A coordinate system is located at the center point of the
sprocket arc surfaces.
i
t
i
t
i
t
Z Y X
The position vector of the point p of chain link j with respect to the center
point of the tooth surface coordinate system can be defined in the global
coordinate system such as in Eqs. (7) and (8)
Necessary but not sufficient conditions for the contact to occur between the
chain link point and the sprocket tooth surface are k
r u u
ij
y
ij
x
+
2 2
) ( ) ( (17)
p t
ij
z p t
w w u w w + (18)
where r is the radius of the sprocket arc segment, is half width of the
t
w
11-13
tooth and is half width of the chain link outer plate.
p
w
If the above conditions are satisfied, it has to be checked if contact point is
existed in the arc range for the next step.
) , ( atan2
ij
x
ij
y k
u u = (19)
a k
0 (20)
where is the angle of with respect to the sprocket arc segment
coordinate system and is the angle of arc segment.
k
ij
k
u
a
2 2
) ( ) (
ij
y
ij
x
ij
u u r + = (21)
11.3.4. ARC-ARC CONTACT
There are four arc-arc contact possibilities such as convex vs. convex, convex
vs. concave, concave vs. convex, concave vs. concave arc contact for arc-arc
interaction between the sprocket teeth and chain link. Since the radius and angle
of each arc are given at geometry, the contact kinematics between arcs can be
calculated by expanding arc-point contact logic. At the center of the arc a marker
is attached and X axis is fixed to the starting point of arc. The monitoring vector
between arc centers can be easily detected whether they are in contact boundary
or not using the arc angles with respect to the X axis of the marker. If the vector
is in contact boundary and the length between the centers of arcs is less than the
sum of the radii of arcs, they are considered in contact situation.
11.3.5 LINE-POINT CONTACT
The search kinematics of line-point contact is one of the most simple search
algorithms in contact analysis. An axis of marker can be attached on the line and
the vertical vector from the point to line can be evaluated whether the point is in
contact with line, respectively.
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.3.6 CONTACT FORCE MODEL
In the field of multi-body dynamics, one of the most popular approximation of
the dynamic behavior of a contact pair has been that one body penetrates into the
other body with a velocity on a contact point, thereafter the compliant normal
and friction forces are generated between a contact pair. In this compliant
contact force model, a contact normal force can be defined as an equation of the
penetration, which yields
(22)
&
n m
n
c k f =
where and are an amount of penetration and its velocity, respectively.
The spring and damping coefficients of and c can be determined from
analytical and experimental methods. The order of the indentation can
compensate the spring force of restitution for non-linear characteristics, and the
order n can prevent a damping force from being excessively generated when the
relative indentation is very small. As it happens, the contact force may be
negative due to a large negative damping force, which is not realistic. This
unnatural situation can be resolved by using the indentation exponent greater
than one. A friction force can be determined as follows.
&
k
m
n f
f v f ) ( = (23)
11.4. NUMERICAL STUDY OF AN AUTOMOTIVE SILENT CHAIN SYSTEM
Four cylinder DOHC (double overhead cam) engine valve drive mechanism is
employed for the sake of numerical verification of proposed methods as shown
in Figure 1. A silent chain drive system has 1 crankshaft sprocket, 2 camshaft
sprockets, 1 fixed guide, 1 pivot guide, tensioner element, and 135 chain links.
The crank sprocket of the system is rotated by motion constraint. Resistance
torque is applied at each camshaft sprockets. Hydraulic tensioning force model is
used which is offered from manufacturer.
Figure 6 shows the computer simulation model of automotive silent chain
11-15
system in computer graphic environment. The system consists of 143 rigid
bodies, 270 bushing force elements to connect chain link bodies, 4 revolute joints,
2 resistance torque and a hydraulic force element of tensioner. It has 815 degrees
of freedom.
Figure 7 and 8 demonstrate the trajectory and velocity of the chain link during
the cycle around the system when the engine runs 4000 rpm, respectively. The
X-Y trajectory of the links agrees the defined path of the chain motion and the
magnitude of link velocity with respect to system inertia reference frame reflect
the linear velocity of 4000rpm as clearly shown in Fig. 8. Figure 9 shows the
contact force between a chain link and the sprockets or the chain guides and
figure 10 shows the dynamic chain tension measured between chain links during
simulation. Since the hydraulic auto tensioner is attached on guide arm, the
dynamic tension of the chain is controlled not to have excessive or be loosened.
Dynamic analysis of the silent chain system is performed for 200 milli-sec. It is
found that the CPU simulation times is 4039 sec on a Pentium 1.8 GHz platform
personal computer. Note that since the numerical results from the proposed
methods are almost showing the real physical behaviors and dynamic
characteristics of the chain mechanism, the proposed methods using multibody
dynamic techniques can be valid and suitable for the design of the silent chain
system, accordingly.
Figure 6. Simulation Model of Automotive Silent Chain System
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
Figure 7. Trajectory of the Chain Link
Figure 8. Velocity of the Chain Link
11-17
Cam Sp.
Cam Sp. Crank Sp. Cam Sp.
Pivot
Guide
Pivot
Guide
Fixed
Guide
Figure 9. Contact Forces of the Chain Link at 4000 rpm
Figure 10. Dynamic Tension of the Chain Link at 4000 rpm
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
11.5. FUTURE WORK AND CONCLUSIONS
It is clearly proved in this investigation that using the multibody dynamic
simulation methods the dynamic analysis of silent chain mechanisms can be
achieved clearly. While previous works showed rough estimations of the silent
chain system, the proposed methods in this paper show the possibility of the
replacement of real prototype at early design stage. The presented three
dimensional silent chain consists of the driving sprocket, idle sprockets, pivot
guide, fixed guide, tensioner, and chain links. Pre and post contact search
algorithms are employed in order to increase the simulation speed significantly.
For the sprocket teeth and link teeth, guide and link contacts, line-arc, arc-point,
arc-arc, and line-point kinematic interactions are presented in this investigation.
A compliant force model is used to connect the rigid body chain links. The silent
chain model has 143 bodies, 4 pin joints, tensioner element and 270 compliant
bushing elements and has 815 degrees of freedom. The numerical study of
automotive silent chain system shows that the tendency of the chain motion and
tensions are close as real system and it shows the characteristics of silent chain
comparing to roller chain with less oscillation.
11-19
REFERENCES
1. C. K. Chen and F. Freudenstein, ''Towards a More Exact Kinematics of Roller
Chain Drives, ASME Journal of Mechanisms, Transmission, and Automation in
Design, Vol.110, No.3, 123-130 (1988)
2. N. M. Veikos and F., Freudenstein, "On the Dynamic Analysis of Roller Chain
Drives: Part1 and 2", Mechanism Design and Synthesis, DE-vol 46, ASME, NY,
431-450 (1992)
3. K. W. Wang, ''On the Stability of Chain Drive Systems Under Periodic Sprocket
Oscillations'', ASME Journal of Vibration and Acoustics, Vol. 114, 119-126
(1992)
4. K. W. Wang, et al, ''On the Impact Intensity of Vibrating Axially Moving Roller
Chains'', ASME Journal of Vibration and Acoustics, Vol. 114, 397-403 (1992)
5. M. S. Kim and G. E. Johnson, Advancing Power Transmission into the 21
st
Centrury, DE-vol. 43-2, ASME, NY, 689-696 (1992)
6. M. S. Kim and G. E. Johnson, Advances in Design Automation, DE-vol. 65-1
(B. J. Gilmore et al., eds), ASME, NY, 257-268 (1993)
7. W. Choi and G. E. Johnson, Vibration of Mechanical Systems and the History of
Mechanical Design, DE-vol. 63 (R. Echenpodi et al., eds), ASME, NY, 29-40
(1993)
8. W. Choi and G. E. Johnson, Vibration of Mechanical Systems and the History of
Mechanical Design, DE-vol. 63 (R. Echenpodi et al., eds), ASME, NY, 19-28
(1993)
9. H. S. Ryu, D. S. Bae, J. H. Choi and A. Shabana, ''A Compliant Track Model For
High Speed, High Mobility Tracked Vehicle'', International Journal For
Numerical Methods in Engineering, Vol. 48, 1481-1502 (2000)
10. Phased Chain System Quietly Transmits Power, Automotive Engineering,
RecurDyn
/ Solver THEORETICAL MANUAL
EFFICIENT CONTACT AND NONLINEAR DYNAMIC MODELING OF AUTOMOTIVE SILENT CHAIN DRIVE
Dec. (1995)
11. J. Chung, J. M. Lee, ''A New Family of Explicit Time Integration Methods for
Linear and Non-linear Structural Dynamics'', International Journal for
Numerical Methods in Engineering, Vol.37, 3961-3976 (1994)
12
DYNAMIC ANALYSIS AND CONTACT
MODELING FOR TWO DIMENSIONAL MEDIA
TRANSPORT SYSTEM
12.1. INTRODUCTION
Recently the media transport systems, such as printers, copiers, fax, ATMs,
cameras, film develop machines, etc., have been widely used and being
developed rapidly. Especially, in the development of those systems, the media
feeding mechanism for paper, film, money, cloth etc., is an important key
technology for the design and development of the media transport systems.
Tedious and iterative experimental trial and errors methods have been essential
way to determine kinematic mechanisms of parts dimensions, and materials, etc
for the media machine developers. Since the iterative trial & error methods are
truly inefficient, in order to shorten the time, reduce the cost, and improve the
machine performance, it has been absolutely required to develop the computer
simulation tool, which analyses the paper feeding and separation process.
Cho and Choi [1] developed a computational modeling techniques for two
dimensional film feeding mechanisms. The flexible film is divided by several
thin rigid bodies which are connected by revolute joints and rotational spring
dampers. The primitive computer implementation methods for contact search
algorithms are presented. Diehl [2, 5] presented the local static mechanics of
electrometric nip system for media transport system. The nonlinear finite
element method and experimental measurement techniques are used to
investigate the large deformable rollers. Several unique phenomena, such as
skewing sheet, etc., of nip feeding system are well described in this research.
Ashida [3] suggested the computer modeling techniques for the design and
analysis of film feeding mechanisms. The primitive dynamic analysis of two
dimensional film feeding models are presented by using commercial computer
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
program. The paper feed mechanism with friction pad system is investigated by
Yanabe [4] by using commercial nonlinear FEA program. It show the local
separation phenomenon between papers and roller, and proved very good
agreement with experimental measurements. Shin [6, 7] developed web
simulation and design tools using roll tensions. They show that the control of
tensions of each segment is the key design factors for web system.
In this investigation, a numerical modeling method and dynamic analysis of
the two dimensional flexible sheet for thin flexible media materials such as paper,
film, etc., and their roller and guide contacts are suggested by using multibody
dynamic techniques. Since the flexible sheet undergoes large deformation with
assumed linear material properties, the flexible sheet has been modeled as a
series of thin rigid bars connected by revolute joints with rotational spring
dampers force elements. It shows good visual appearance of the sheet under
severe bending conditions. An efficient contact search and force analysis
between sheet and rollers, and guides are developed and implemented
numerically. The sheet is fed by contact and friction forces when it contacts with
rotating rollers or guides. In order to detect a contact phase efficiently, the
bounding box method is used in this contact search algorithm. The method has
an advantage that the number of contact search can be smaller than conventional
methods for a system in which the position of rollers and guides are fixed on a
point of a base body. The proposed numerical models for media transport
systems will make it possible to confirm the potential problems of jamming by
given different sheet size, weight, stiffness, temperature, humidity extremes,
sheet velocity due to misalignment of drive-driven roller sets, and roller
velocities due to gap, wear or etc.
12-3
12.2. TWO DIMENSIONAL FLEXIBLE MULIBODY SHEET
In general, there are two methods to build a thin 2-D flexible sheet for
dynamic analysis. One is to employ beam element at discretized sheet body, and
the other is small rigid bar interconnected by revolute joint with rotational
spring-damper forces. In this investigation, the second method is used and
proposed the modeling techniques.
Figure 1 Modeling definition of a two dimensional flexible sheet
Several research works show that the most efficient way to model two-
dimensional approximation of the proper behavior of a sheet can be a series of
rigid bars connected by revolute joints and rotational spring-dampers as shown
in Figure 1 [1, 3]. The sheet is divided into a number of rigid bars with mass.
The mass and inertia moment of each rigid bar can be defined as follows
s s
t L m =
(1)
12
) (
2 2
s s
zz
L t
m I
+
=
(2)
where, is a sheet density per unit depth, is thickness, and is length
of each rigid bar. The leading body is connected to a ground by a planar joint to
guarantee an in-plane motion. The planar joint has one rotational and two
translational degrees of freedom. The body is connected to the body
by a revolute joint and rotational spring damper. The revolute joint has one
rotational degree of freedom between two rigid bars. The relative angle of
s
t
s
L
(i i ) 1 +
) 1 ( + i i
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
is directly integrated. The torque of the rotational spring-damper is computed as
following
) 1 ( ) 1 ( + +
=
i i i i
c k
&
(3)
s
s
L
t E
k
12
3
=
(4)
k c = (5)
where,
) 1 ( + i i
&
E and are the young s modulus and the structural
damping ratio of a sheet.
Figure 2 Contact geometry of two-dimensional sheet
The contact geometry of a sheet is described as a box and two circles as
shown Figure 2. The x-axis of the body reference frame of each rigid bar is
defined along longitudinal length direction and the y-axis is defined by right
hand rule. The mass center of each rigid bar is located at the center point of box.
In order to generate a continuous contact force, two circles are located on both
sides of the box. Even thought the proposed assumed method for flexible sheet
has an excellent visual appearance of the sheet under severe bending conditions,
this approach shows the lack of continuity between rigid bodies, which can cause
noise problems when the sheet is contact with rollers. It has also rigid leading
and trailing effect of the sheet. Problems can be overcome with introducing a
circular edge at leading and trailing points of each rigid bar.
There can be another approach to assume flexible sheet in dynamic analysis,
12-5
which employs a series of beam forces, and for the contact definitions, a rigid
bar can be attached simply. One of the advantages of this approach is a natural
definition of the flexible properties using the beam elements. However this
approach can cause problems with the contact definitions since it has possible
gaps and the lack of continuity between rigid contact bodies. The contact forces
on the edges of the rigid bodies are amplified as torques applied where the rigid
body is connected to the junction of two beams, and the rigid leading and trailing
edges of the sheet cause unnatural behaviors.
12.3. CONTACT FORCE ANALYSIS
In the field of multi-body dynamics, one of the most popular approximation of
the dynamic behavior of a contact pair has been that one body penetrates into the
other body with a velocity on a contact point, thereafter the compliant normal
and friction forces are generated between a contact pair. Figure 3 shows the
schematic diagram of contact force analysis used in this investigation.
Figure 3 Contact forces between a contact pair
In this compliant contact force model, a contact normal force can be defined
as an equation of the penetration [9], which yields
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
(6)
&
n m
n
c k f =
where and are an amount of penetration and its velocity, respectively.
The spring and damping coefficients of k and can be determined from
analytical and experimental methods. The order of the indentation can
compensate the spring force of restitution for non-linear characteristics, and the
order can prevent a damping force from being excessively generated when
the relative indentation is very small. As it happens, the contact force may be
negative due to a large negative damping force, which is not realistic. This
unnatural situation can be resolved by using the indentation exponent greater
than one. The phenomenon is very important for the case of sheet contact
interaction since it is very thin and light. A friction force can be determined as
follows.
&
c
m
n
n f
f v f ) ( =
(7)
where, and
n
f ) (v
are a contact normal force and a friction coefficient,
respectively.
12.3.1. KINEMATICS NOTATIONS
The coordinate system is the inertial reference frame and the single
primed coordinate systems are the body reference frames, and the double primed
coordinate system is the contact reference frame in order to define contact
conditions as shown in Figure 4. The orientation and position of the body
reference frame are denoted by and , respectively.
Y X
A r
12-7
Figure 4 Kinematic notations of a contact pair
12.3.2. SHEET AND ROLLER INTERACTIONS
In this investigation, two kinds of rollers are defined for the system. One is a
fixed roller with one rotational degree of freedom. The fixed roller is linked to
the ground with a revolute joint. The other is a movable roller, which has two
degrees of freedom for a translational and a rotational motion. The movable
roller is linked to rotational axis retainer (RAR) with a revolute joint and the
retainer is linked to the ground with a translational joint. The contact geometry
of rollers is described as a circle as shown in Figure 5
Figure 5 Definition of rollers
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
Two different interactions between roller and sheet are introduced in this
investigation. Since the proposed flexible sheet is constructed by linear part and
circular part, these are interactions between linear part and rollers, and circular
part and rollers, as clearly illustrated in Figure 6
Figure 6 Sheet and roller interaction
In the case of linear part contact with rollers, the contacted penetration is
determined as follows:
r
R =
y sr,
d
, (8)
) r (r A d
s r
T
s sr
=
where, is the orientation matrix of a rigid bar, and is the radius of a
contacted roller, respectively. The location of contact between rigid bar and
roller can be defined as follows:
s
A
r
R
=
0
2 / ) (
s y s
t sign
sr,
x sr,
c
d
d
s
, (9)
and
) d s ( A A s
sr sc s
T
r rc
= (10)
where, and t are the orientation matrix of a roller and the thickness of
the sheet. The relative velocity at the contact point can be determined as
r
A
s
12-9
( )
( )
sc r s rc r r
r s rc r r
s w A r s w A
s A r s A r
+
|
.
|
+
s
dt
~ ~
&
r
T
n
T
n
r u
u
=
\
|
=
d
&
&
(11)
c
T
n
d u
&
=
(12)
and tangential relative velocity is
c
T
d u
&
t t
v =
(13)
where, and
r
w
s
w are the angular velocities of a roller and a rigid bar with
respect to each body reference frame, and u and are the normal and
tangent vectors of relative position between rigid bar and roller, respectively.
n t
u
12.3.3. ROLLERS INTERACTIONS
A circle to circle contact is used to describe the interactions between circular
rollers. In this circle to circle contact, the positive normal direction is same in the
direction of the relative position vector between two roller center points. The
tangent direction vector is determined by the right hand rule. The relative
velocity and the contact forces at the contact point can be computed similarly as
the sheet and roller interactions.
12.3.4. SEEET AND GUIDE INTERACTIONS
Guide has three types. Commonly used sheet guides for media transport
machines can be divided into three different types, which are an arc guide with
radius and angle, a linear guide with two points, and a circle guide similar to a
roller. In order to avoid the complex contact detect algorithms. It is assumed that
the arc and line guide are interacted with the circular part of rigid bars of the
sheet. However, in the case of circle guide, both linear and circular parts of the
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
sheet are interacted with.
Figure 7 Sheet and arc guide interactions
As shown in Figure 7, the relative displacement between a circular edge of
rigid bar and arc guide can be determined as
g g g s s s gs
s A r s A r d + =
(14)
where, and are the center position and the orientation of the guide,
and the vectors of
g
r
g
A
g
s and s
s
are positions of the arc reference frame and the
circular edge center position with respect to each body reference frame,
respectively. If the vector is projected into the arc reference frame, the
resultant vector can be represented as follows
gs
d
gs gs
d C A d
g g
T
) ( = (15)
where, is the orientation matrix of the arc reference frame. The relative
angle between x-axis of the arc reference frame and the resultant vector of Eq.
15 is within an arc angle, which can be written as
g
C
g
) ( cos 0
1
gs
g
T
gs
d
f d
(16)
12-11
where, is the arc angle and is a constant unit vector of . If
the condition of Eq. 16 is satisfied, the penetration between circular part of sheet
and arc can be defined as follows
g
f 1 | |
T
0 0
g
g s gs
R t + = d
(17)
where, is a radius of the arc guide. The contact positions can be computed
as follows.
g
R
n g g
R u s
c
= (18)
c g g
T
c
c c
s C A A s
d s s
s
g
s s
gs g s
g
=
=
(19)
where, is the normal direction vector and determined
n
u
gs
gs
d
d
u
n
=
(20)
The tangent direction vector is determined by the right hand rule, and the
relative velocity at the contact point is defined as follows.
) (
~
) (
~
) ( ) ( (
gc g g g g s s s s
gc g g g s s s
dt
d
s C s w A r s s w A r
s C s A r s s A r d
g sc
g sc c
+ + + =
+ + + =
& &
&
(21)
where, and
g
w
s
w is the angular velocities of guide and a bar with respect
to each body reference frame, respectively. The contact forces at the contact
point can be computed similarly as described in the sheet and roller interactions.
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
Figure 8. Sheet and line guide interactions
The sheet and line guide interactions are clearly illustrated in Figure 8. If the
component of the vector
d
defined in the double primed line guide
reference frame is the range of guide length, simple circle and line contact
algorithm is used in this investigation. After definitions of penetration and its
derivative, the contact force is created to restitute each body as similar as
previous interactions between sheet and guides.
x
gs
12. 4. EQUATIONS OF MOTION
Figure 9 Kinematic relationships between rigid bars of the sheet
12-13
Since the multibody sheet system interacts with the roller and guide
components through the contact forces and adjacent rigid bars are connected by
revolute joint and rotational spring damper forces as shown in Figure 9, each
sub-rigid bar in the sheet system has one degree of freedom which is represented
by one rotational coordinates and the leading body has three free coordinates.
The equations of motion of the sheet system that employs the velocity
transformation defined by Bae [8] are given as follows:
) (
r
i i
q B M Q B q MB B
T T
&
&
& & = (22)
where , and are relative independent coordinates, velocity
transformation matrix, and Cartesian velocities of the media feeding system, and
is the mass matrix, and is the generalized external and internal force
vector of the media feeding system, respectively. The velocity transformation
matrix of the sheet is more explicitly as
r
i
q B q&
M Q
B
(
(
(
(
(
(
=
1)n2 (n 232 1)n1 (n 122 1)n1 (n 012 1)n1 (n
232 122 231 012 121 231
122 012 121
012
B B B B B B B
0 B B B B B B
0 0 B B B
0 0 0 B
B
L L L
M M M M M
L
L
L
where the recursive velocity and virtual relationship for a pair of rigid bars are
obtained [8] as
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Y B Y & (23)
and denotes the relative coordinate vector. It is important to note that
matrices and are only functions of the .
1)i (i
q
(i
B
1)i1 1)i2 (i
B
1)i (i
q
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
12. 5. NUMERICAL RESULTS
The proposed algorithm is implemented and a film-feeding problem is solved
to demonstrate the efficiency and validity of the proposed method.
Figure 10 Film feeding machine
The system has 29 degrees of freedom and consists of four fixed rollers and
three movable rollers, five line guides, one arc and circle guide and one sheet of
film shown in Figure 10. The sheet is modeled by using 20 rigid bars. The
density and Young s modulus of sheet are 2.2e-6( kg ) and 2250( ),
respectively. And the thickness and length of sheet are 0.5( mm) and 200( ),
respectively.
3
/ mm
2
/ mm N
mm
Figure 11 Slip between rollers and sheet
12-15
The film goes through a path while contacting the roller pairs. The
circumferential speed of each driving roller is 300( ). The slip velocities
between driving rollers and the sheet are shown in Figure 11. The path of first,
second and third segment bodies of the thin film are plotted as shown in Figure
12. The x and y axes of the plot are displacements measured in the directions of
x and y axes in the global reference frame, respectively.
sec / mm
Figure 12 Path of segmented bodies of film
The analysis was performed on an IBM compatible computer (PIII-933Mhz)
and took about 60 sec. per 1 sec. for simulation.
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
12. 6. CONCLUSIONS
The dynamics and modeling techniques of two-dimensional media transport
system is investigated in this paper. The flexible sheet is divided by finite
number of rigid bars. Linear motions are constrained in order to allow rotations
between the rigid bars of the sheet. Rotational spring damper force is applied for
the reflection flexible stiffness of the sheet. From previous empirical
measurements in manufacturing process effective stiffness and damping
coefficients are substituted in this investigation. Compliant contact force model
is used for the interactions between sheet rollers, and guides. Kinematics
notations of the contact search algorithms for the media transport system are
clearly represented. A simple film feeding example is represented in this
investigation and manufacture [3] confirms that simulation results have very
good agreement with experimental measurements. The media transport system
manufactures have rely on trial error techniques for the design of their core
mechanisms, however the proposed method by employing multibody dynamics
in this paper can reduce many difficulties at the early design stage.
12-17
REFERENCES
1. H. J. Cho, and J. H. Choi, 2001, 2DMTT development specification Technical
report, FunctionBay Inc.
2. Ted Diehl, 1995, Two dimensional and three dimensional analysis of nonlinear
nip mechanics with hyper elastic material formulation Ph. D. Thesis,
University of Rochester, Rochester, New York
3. Tsuyoshi Ashida, 2000, The meeting material of The Japan Society for
Precision Engineering Japan
4. http://www.yanabelab.nagaokaut.ac.jp
5. http://www.me.psu.edu/research/bension.html
6. http://www.engext.okstate.edu/info/WWW-WHRC.htm
7. Shin, K. H., 1991, Distributed Control of Tension in Multi-Span Web Transport
Systems , Ph. D. Thesis Oklahoma State Univ.
8. D. S. Bae, J. M. Han, and H. H. Yoo, 1999, A Generalized Recursive
Formulation for Constrained Mechanical System Dynamics, Mech. Struct. And
Machines, Vol. 27, No 3, pp 293-315
9. Lankarani H. M. and Nikravesh P. E., 1994, Continuous Contact Force Models
for Impact Analysis in Multibody Systems, Journal of Nonlinear Dynamics,
Kluwer Academic Publishers, Vol. 5, pp 193-207
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS AND CONTACT MODELING FOR TWO DIMENSIONAL MEDIA TRANSPORT SYSTEM
13
HYDRAULIC AUTO TENSIONER (HAT) FOR
BELT DRIVE SYSTEM
13.1. INTRODUCTION
The hydraulic auto tensioner is a device that automatically adjusts the tension
for engine belt drive system. By reducing the noise due to play that occurs if the
tension on the belt drive system is insufficient and by holding the tension
constant, an auto tensioner extends the product life of the belt drive system and
is an indispensable part for improving engine reliability [1]. It is important to
analyze and to predict the dynamic behavior and the characteristics of the
hydraulic auto tensioner for design of the system. At this, numerical simulation
models can provide significant advantages in early design stage referred in [2]
and [3]. A simple simulation technique of HAT is applied for the initial design of
belts and chains using commercial multibody software [7]. Figure 1 shows the
hydraulic auto tensioner system. The plunger is connected to the belt drive
system. The spring force and the hydraulic force of the pressure chamber create
the damping force and are balanced with the load that is from belt drive system.
The check ball has the function of the check valve for control the oil flow
through orifice between the plunger and the cylinder.
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
Plunger
Cylinder
Check Ball
Pressure Chamber
Plunger
Spring
Check Ball
Spring
Figure 1. Hydraulic auto tensioner system
Figure 2 shows the schematic diagram of the operating principle of HAT. As
the tension of the belt drive system decreases, and the pressure of chamber
decreases, the check ball moves down and the check valve opens. Afterward due
to the plunger moves up by the spring force and the plunger pushes the belt, and
then the tension of belt system is increased. As the tension of the belt drive
system increases, the plunger moves down by the load and the plunger pushes
the pressure chamber, and it leads that the pressure of chamber increases, finally
the check ball moves up and the check valve closes. As a result, the oil flows
through the leakdown and the plunger moves down slowly.
Tension Decreasing Tension Increasing
Oil
Flow
Figure 2. Operating principle
13-3
Since the tension of the belt drive system is oscillated over 200~300Hz, the
hydraulic auto tensioner must be a reciprocating hydraulic device that can
respond to frequencies up to 300Hz.
The multibody simulation model of the hydraulic auto tensioner is presented
in the following sections. The differential equations are used to describe the
function and damping characteristics of the hydraulic auto tensioner, and the
circle to curve contact model is used for the movement of the check ball. In this
investigation, the developed HAT model is tested numerically for multibody belt
drive system.
13.2. MULTIBODY SIMULATION MODEL
The hydraulic auto tensioner consists of cylinder, plunger and check ball. The
spring force and the damping force of the plunger relative to the cylinder balance
these bodies. The spring force is built up by the spring preload and the spring
rate multiplied by the spring stiffness. The damping force is a friction force and a
hydraulic force that is proportional to the relative velocity of plunger and
cylinder [3].
The schematic diagram of analysis model is shown in Figure 3. When the
plunger is loaded from belt drive system, the spring force and the hydraulic force
react against the motion of the plunger. The hydraulic force from the check ball
is ignored in this investigation since it is relatively small amount. The motion of
the plunger is assumed to have the parallel direction to the motion of the cylinder.
The check ball has the spring force and the hydraulic force from the plunger. The
motion of the check ball is also assumed to have the parallel direction to the
motion of the plunger. The check ball is contacted between plunger and retainer.
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
T
r
a
n
s
l
a
t
i
o
n
J
o
i
n
t
S
p
r
i
n
g
F
o
r
c
e
H
y
d
r
a
u
l
i
c
F
o
r
c
e
T
r
a
n
s
l
a
t
i
o
n
J
o
i
n
t
Check
ball
S
p
r
i
n
g
F
o
r
c
e
H
y
d
r
a
u
l
i
c
F
o
r
c
e
Plunger
C
o
n
t
a
c
t
Cylinder
Belt Drive System
Joint
Figure 3. Schematic diagram of analysis model
11.3. THE EQUATION OF MOTION
When the external load is forced to the plunger, the equation of the plunger
motion is following [3].
load ric c p
c p p hydraulic p p p
F f x x
x x K F x m
=
) sgn(
) (
.
& &
& &
(1)
where, , , and are the displacement of the plunger, and its first and
p
x
p
x&
p
x& &
13-5
second time derivatives, and , , , and are the mass of plunger, the
displacement of the cylinder, its first time derivative, and the stiffness coefficient
of the plunger spring, and , , and are the hydraulic force, the
friction force and the load form belt drive system, respectively.
p c
x
hydraulic
B
K
F
m K
c
ric
B
x&
f
B
hydraulic
x
p
load
F
BO
B
F
x
p
F
.
B B
x = & &
)
(
(
.
&
oil
&
c
x& )
p
p
x & ( V +
&
S =
In the case of check ball, since its motion is forced by the hydraulic force, the
spring force, and contact force, the equation of motion of the check ball can be
written as [3]
contact p
c B
F x
x m
+
) &
(2)
13.4. HYDRAULIC FORCES
The hydraulic forces that interact with the check ball and the plunger are
obtained from the pressure of the pressure chamber. The pressure is caused by
the volume variation of the pressure chamber and the oil flow rate. The volume
variation of the pressure chamber can be described by relative velocity between
plunger and cylinder. The rate change of the chamber volume is given by the
following equation [5].
p B air chamber
Q Q V V + + =
&
(3)
where is the compressed volume rate of pure oil and is the compressed
volume rate of air component in the oil. Q
oil
V
&
air
V
&
p
is leak oil flow rate out of the high
compression chamber at high pressure phase, Q
B
is the oil flow rate through
check valve, and S
p
is the effective area of hydraulic force, respectively.
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
13.4.1. OIL FLOW RATES THROUGH THE CHECK VALVE
Accordingly to the check ball moves between the plunger and the retainer, the
check valve opens or closes and the oil flows. When the check valve opens, the
oil flow through the check valve is shown in Figure 4.
Po
Pi
d
r
QB
Figure 4. Oil flow rate through check valve
As the resistance of the oil through the check valve depends on the orifice area,
in this investigation, the dynamic resistance is considered for the turbulent flow
of the oil flow through the check valve [4], which yields,
i o i o d B
P P
g
P P A C Q =
2
) sgn(
(4)
where, is discharge coefficient of check valve, A is the orifice area, g is
gravity acceleration, and is weight density of oil, respectively. Q
d
C
B
represents
the oil stream flowing rate through the opened check valve into or out of the
pressure chamber. The area of orifice is obtained such as
cos sin 2 d r A =
(5)
13-7
13.4.2. OIL FLOW RATES THROUGH THE LEAK
BETWEEN PLUNGER AND CYLINDER
As the pressure of chamber is different comparing to the air pressure, the oil
flow through the gap between plunger and cylinder is shown in Figure 5. The oil
flowing between the plunger and the cylinder is laminar flow. The oil speed is
faster than the plunger speed. As shown in Figure 5, variation of the oil speed is
fully depended on the pressure difference between the pressure chamber and
reserver, and it is not affected by the plunger speed. The oil flow rate between
the gap of plunger and cylinder, Q
p
, can be written as [4]
) (
12
2
3
i o
P
p
P P
l
h r
Q =
(6)
where is the coefficient of viscosity of oil.
P
o
P
i
Q
p
l
r
p
h
Figure 5. Oil flow rate through leak
As shown in the Figure 4 and 5, we can consider about the relationship
between the plunger speed and the flow rate. It is assumed that inflow does not
induce any outflow from the pressure chamber by considering compressibility,
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
and expansion and compression processes are isentropic. It is also assumed that
there is no cavitation caused by negative pressure. The volume of air in chamber
is obtained as.
0 .
1
air
i
o
air
V
P
P
V
= (7)
where is the ratio of specific heat and V
air.0
is the initial air volume. Since air
can be compressed, the volume rate is achieved by using the following equation,
such as
i
i
air
air
P
P
V
V
& &
= (8)
, and the volume of oil in chamber can be approximated by
p c p oil
S x x V ) (
(9)
In the case of oil, since it can also be compressed with high pressure, the
volume rate of oil is written as.
i
oil
oil
P
K
V
V
& &
=
(10)
where K is the bulk modulus.
The equations (4), (6), (8) and (10) are substituted into the equation (3), and
the differential equation for the pressure of the camber can be obtained,
accordingly;
13-9
+ +
=
K
V
P
V
P P
g
P P A C P P
l
h r
x x S
oil
i
air
i o i o i o
P
C P P
i
2
) sgn( ) (
12
2
) (
3
& &
p&
)
(11)
The hydraulic force to the plunger and to the check ball yield as
) (
. i o p hydraulic p
P P S F =
(12)
) (
. i o B hydraulic B
P P S F =
(13)
where S
B
can be obtained from Figure 4 as following.
(
2
cos = r S
B
(14)
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
13.5. CONTACK OF THE CHECK BALL
The contact analysis of the check ball employs the circle to curve contact
method [6] in this investigation. This method is very efficient algorithm in
contact detection and force generation of the check ball contact.
Figure 6. Concept of circle to curve contack
The candidate lines on the plunger body have been selected for the contact of
the check ball. For the candidate lines, it is necessary to compute the amount of
penetration to generate the contact forces, as shown in Figure 6.
The relative position of a check ball with respect to the contact reference
frame is obtained as follows.
pn
d
1
p cn pn
s d d =
(15)
where the vector
pn
d
is projected into the contact reference frame as
13-11
pn
d
T
p pn
C d =
(16)
where C
p
is the orientation matrix of the contact reference frame. The
penetration of the node into the patch is calculated by
pn
T
p
- d n = r
(17)
where is always positive. The
p
n is a normal vector of a line and a constant
vector with respect to the contact reference frame. Thus, the contact normal force
is obtained by
3
2
1 m
m
m
contact
c k F
&
&
&
+ =
(18)
where k and c are the spring and damping coefficients which are determined by
assumed numerical experiences, or experimental methods, respectively and the
is time differentiation of . The exponents and generates a non-
linear contact force and the exponent yields an indentation damping effect.
When the penetration is very small, the contact force may be negative due to a
large negative damping force, which is not realistic. This situation can be
avoided by using the indentation damping exponent greater than one.
&
1
m
2
m
3
m
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
13.6. BELT DRIVE SYSTEM
An automotive belt drive system is used for the simulation of HAT in order to
test numerically. This system is consisted of 5 pulleys and a belt system. A
continuous belt system can be modeled using series of a single body that has six
degrees of freedom and has a matrix (6x6) force element to connect the belt
bodies. Contact forces between the belt and pulleys are defined clearly.
Disturbance roller
Drive
Pulley
HAT
Sensing belt tension
Belt
Figure 7. Belt drive system
As shown in Figure 7, there are a drive pulley, a disturbance roller, four idle
pulleys, and an idle roller equipped with HAT.
13-13
13.7. NUMBERICAL RESULTS
The hydraulic auto tensioner must be a reciprocating hydraulic device that can
respond to frequencies up to 300Hz. When the reciprocating load is applied to
plunger with 300Hz, Figure 8 shows the result of the pressure in chamber and
Figure 9 shows the result the displacement of the check ball. The numerical
results show that the proposed modeling of HAT is acting to the reciprocating
load with 300Hz. As the load increases, the check valve closes and the oil flows
out only through the leak. As the load decrease, the check valve opens and the
oil flows in through the check valve.
Figure 8. Pressure in chamber [300Hz]
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
Figure 9. Displacement of check ball [300Hz]
The proposed modeling method of hydraulic auto tensioner is applied for the
belt drive system as shown in Figure 7. The drive pulley rotates with 100 rpm.
As the disturbance roller increases the belt length, the belt tension around HAT
increases such as shown in Figure 10. Due to the belt tension increases, the
pressure in chamber arises as shown in Figure 11 and the oil flows out through
the leak as shown in Figure 12. As a result, the plunger is pushed back and the
belt tension decreases. Figure 10 shows less increase of tension of the belt with
HAT comparing to without it.
13-15
Figure 10. Tension
Figure 11. Chamber pressure
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
Figure 12. Oil flow rate through leak
As the disturbance roller decreases the belt length, the tension around HAT
decreases as shown in Figure 13. Due to the belt tension decreases, the pressure
in chamber decreases as illustrated in Figure 14. Figure 15 shows the oil flow
rate through the check valve. As a result, the plunger is pushed to the direction
for increasing the tension by the plunger spring, and therefore the tension
increases. The tension drop can be quickly recovered with proposed HAT
element as shown in Figure 13.
13-17
Figure 13. Tension
Figure 14. Chamber Pressure
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
Figure 15. Oil flow rate through check valve
13-19
13. 8. CONCLUSIONS
In this investigation, in order to design automotive power transmitting system
at early design stage, modeling and simulation methods of HAT, which is
necessary component for the tension adjusting system, are presented. The
multibody simulation model is proposed using three rigid bodies, which are
plunger, check ball and cylinder. The plunger and the cylinder bodies can be
connected by constraints and mechanical force elements. The plunger and the
cylinder are interacted by hydraulic force and spring force. The forces between
plunger and check ball are modeled by contact, hydraulic, and spring forces. The
circle to curve contact analysis is employed for the plunger and the check ball
contact efficiently. The differential equations of motion of the components and
the hydraulic force equations are developed in this investigation. It can be
assured that the proposed HAT model is able to respond to frequencies up to
300Hz. The proposed methods of HAT are simulated in different ways,
component level simulation with reciprocating forces, and with automotive belt
system. Both numerical results show reasonable responses as expected. Though
it is necessary to be correlated by experimental results. Therefore the proposed
numerical method of HAT shows the possibility of simulation for automotive
power transmitting system, which has been challenging works for long period.
RecurDyn
/ Solver THEORETICAL MANUAL
HYDRAULIC AUTO TENSIONER (HAT) FOR BELT DRIVE SYSTEM
REFERENCES
1. http://www.ntn.co.jp/english/corp/news/news/20011001_2.html
2. NTN TECHNICAL REVIEW No. 61
3. NTN TECHNICAL REVIEW No. 67
4. Frank M. White, "Fluid Mechanics", 5th edition, McGraw-Hill International
Editions, 1999.
5. E. Sonntag , Richard, Claus Borgna, kke, and Gordon J. Van Wylen,
"Fundamentals of Thermodynamics", 5th Edition, John Wiley & Sons, Inc., 1998.
6. B. O. Roh, H. S. Anm, D. S. Bae, H. J. Cho, H. K. Sung, "A Relative Contact
Formulation for Multibody System Dynamics", KSME International Journal, Vol.
14, No. 12, pp. 1328-1336, 2000.
7. www.fev.com
14
DYNAMIC ANALYSIS OF CONTACTING SPUR
GEAR PAIR FOR FAST SYSTEM SIMULATION
14.1. INTRODUCTION
All Geared systems are commonly used in many mechanical power
transmitting systems, such as robot manipulator, automotive transmissions, etc.,
so as to transmit motion and power from one shaft to another. One of important
factors in the gear design is the dynamic transmission error, which gear vibration,
noise and other performance can be predicted by. When two mating gear is
operated, the dynamic transmission error is generated by gear dynamic forces.
These forces are caused by contact between meshing teeth. In other words,
contact mechanics between meshing teeth, considering backlash and tooth
geometric profile, is very important in the dynamic analysis of geared systems. A
lot of numerical and experimental works have been published about their
dynamic analysis. One of main topics in these studies is the conventional finite
element analysis. Traditional finite element methods are effective for calculating
quantities such as mesh stiffness, tooth deformations, and stress distributions
under static conditions. But it requires refined meshes to represent the tooth
contact and precise tooth surface shape for gear mechanics. Also, it takes
amazingly long time to analyze the dynamics effects of contacting gears.
Moreover, it is not suitable for analysis of entire system with the sets of gear
pairs as well as other components [1, 2]. Another topic is that concerning the
single degree of freedom(sdof) models of a pair of gears. It is because sdof
model can give relatively accurate results and computational efficiency despite
its simplicity. The sdof model approach in terms of entire system dynamic
analysis with gear pairs is desirable from research and design perspectives. In
sdof model, primitive approach is to model gear pairs with simple constraint or
force element using speed ratio, pressure angle and rotational angles. Gear
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
systems can be analyzed with fast computational time, but detailed inputs such
as tooth profiles and distance between gears are not considered directly because
it is not real gear teeth contact. More advanced approach is considered by contact
between teeth profile of gears. It enables designer not only to obtain gear contact
position and force exactly but also to simulate with entire system in various
operating conditions [3, 4].
A review of the mathematical models used in gear dynamics was given by
Ozguven and Houser [5], and T. Shing et al. presented an improved model for
the dynamics of spur gear systems with backlash consideration [6]. The torsional
vibration behavior was investigated experimentally by Kahraman and
Blankenship [7, 8, 9]. In the recent studies, a sdof model was proposed, which
considers a time-varying stiffness and backlash of the meshing tooth pairs with
similar formulations. However, most gear models in these numerical
investigations have been used the kinematic relations between the rotational
angles of each gear. It is not real contact model between bodies and needs some
limitation that gear shafts have no translational displacement.
The main purpose of this paper is to develop efficient contact algorithm
between meshing teeth in geared system for better understanding of the dynamic
behavior of entire system. Externally specified dynamic forces, or assumptions
about modeling the mesh forces by time-varying stiffness and static transmission
error are not required since dynamic mesh forces are obtained by contact
analysis at each time step. A simple spur gear pair modeled by using proposed
methods is compared and verified with the measurement results represented by
reference [7]. The dynamic modeling techniques are suggested and efficient &
fast dynamic analysis of a set of complex geared mechanical system is presented
in this investigation.
14.2. TOOTH PROFILE OF SPUR GEAR
The gear teeth profile is usually defined a special profile called an involute
curve for constant speed ratio. However, it is not efficient to use the exact
involute profile in the contact search algorithm because of its complexity of
contact search kinematics. In order to approximate the involute profiles, biarc
14-3
curve fitting method which is proposed by Bolton[11] is employed in this
investigation. The optimum biarc curve passing through a given set of points
along involute curve can be determined by this approximation technique. The
more arcs are used to describe the involute profiles, the less numerical error is
occurred in approximation, but the more calculation time will be required for
contact search of tooth profiles. Consequently, the real geometry of involute
tooth profiles in this investigation is represented by 5 arcs with different radii as
shown in Figure 1, since the error is acceptably small.
Fig. 1 Involute curves by 5 arcs
Arc segment Absolute error (mm) Relative error (%)
1 0.000229 0.00147
2 0.000349 0.00152
3 0.000388 0.00165
4 0.000409 0.00168
5 0.000461 0.00182
Table 1. Absolute and relative error
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
Table 1 shows the difference between exact involute curve and approximated
arc segment in spur gear with 24 teeth, 2 mm module, and 20 pressure angle.
Absolute error is an average distance between points on exact involute curve and
points on arc segment from gear center. Relative error is an average difference
percentage that absolute error is divided by average distance of points on
involute curve from gear center. Since the main purpose of the research is to
understand the dynamic behaviors of system with the gear pairs, these kinematic
errors might affect very small for the highly oscillating nonlinear dynamics of
gear system, accordingly.
14.3. EFFICIENT CONTACT SEARCH ALGORITHM
AND CONTACT FORCE MODEL
The contact algorithms for a gear pair are investigated in this section. The
contact positions and penetrated values are defined from the kinematics of
components in searching routines. Thereafter, a concentrated contact force is
generated at the contacted position of the contact surface of the bodies. A
detailed discussion on the formulation of the contact collision is represented in
this section.
14.3.1. ARC-ARC CONTACT
Since the radius and angle of each arc are given at geometry, the contact
kinematics between arcs can be calculated by contact logic. A marker is attached
at the center of the arc and X axis is fixed to the starting point of arc. The
monitoring vector between arc centers can be easily detected whether they are in
contact boundary or not using the arc angles with respect to the X axis of the
marker. If the vector is in contact boundary and the length between the centers of
arcs is less than the sum of the radii of arcs, they are considered as contact
candidate.
14-5
X
Y
Z
i
X
i
Y
i
Z
i
t
X
i
t
Y
i
t
Z
i
R
i
u
j
p
X
j
Y
j
Z
j
R
j
p
u
p
t
ij
k
u
j
j
p
Y
j
X
i
Pinion tooth
coordinate system
Pinion
coordinate system
Gear
coordinate system
t
Gear tooth
coordinate system
Global coordinate
system
Fig. 2 Arc-arc contact kinematics
The contact conditions between the gear tooth convex arc segment and the
pinion tooth convex arc segment can be determined as follows. A coordinate
system and is attached to each arc origin coordinate system
shown in Fig. 2. The surface of the gear tooth arc segment is approximated by
plane surfaces and the axis of each surface coordinate system is assumed to be
directed to the starting arc point from arc origin. The surface of pinion tooth arc
segment is approximated by plane surfaces and the axis of each arc origin
coordinate system is assumed to be directed to the starting arc point from arc
origin. The orientation of the gear tooth arc coordinate system with respect to
the global system is defined by
i
t
i
t
i
t
Z Y X
j
p
j
p
j
p
Z Y X
i
t
X
j
p
X
k
i
k
i i
t
A A A =
(1)
where is the transformation matrix that defines the orientation of the
i
A
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
coordinate system of the gear and is the transformation matrix that
defines the orientation of the gear tooth arc coordinate system with
respect to the gear coordinate system. The orientation of the pinion tooth arc l
coordinate system with respect to the global system is defined by
i
l
j
A
i
k
A
k
i
t
i
t
i
t
Z Y X
j
j
l
A
j
p
X
A +
i
+
r
j
Z
j j
p
A A =
(2)
where is the transformation matrix that defines the orientation of the
coordinate system of the pinion
j
A
and is the transformation matrix that
defines the orientation of the pinion tooth arc coordinate system
with respect to the pinion coordinate system.
l
j
p
j
p
Z Y
The global position vector of the center of the gear arc segment, denoted as
point , is defined as t
i
t
i i i
t
u R r =
(3)
where
i
R is the global position vector of the origin of the gear and
i
t
u is
the position vector of arc center point with respect to the origin of the gear
coordinate system .
t
i i i
Z Y X
The global position vector of the center of the pinion arc segment, denoted as
point p , can be defined as
j
p
j j j
p
u A R r =
(4)
arc center point defined in the pinion coordinate system . p
j j
Y X
The position vector of the center of the arc of pinion with respect to the origin
of the gear tooth arc can be defined in the global coordinate system as
i
t
j
p
ij
k
r u =
(5)
The components of the vector with respect to the gear and pinion tooth
ij
k
u
14-7
coordinate system are determined, respectively, as
[ ]
ij
k
T
i
t
T
i ij
z
i ij
y
i ij
x
i ij
u u u u A u = =
, , , ,
(6)
[ ]
ij
k
T
j
p
T
j ji
z
j ji
y
j ji
x
j ji
u u u u A u = =
, , , ,
(7)
Necessary but not sufficient conditions for the contact to be occurred between
the gear and pinion arc segment are
p t
i ij
y
i ij
x
r r u u + +
2 , 2 ,
) ( ) (
(8)
p t
ij
z p t
w w u w w +
(9)
where and r are the radius of the gear and pinion arc segment respectively,
is half width of the gear tooth and is half width of the pinion tooth.
t
r
p
t
w
p
w
If the above conditions are satisfied, it has to be checked if contact point is
existed in the arc range for the next step.
) , ( atan2
, , i ij
x
i ij
y m
u u =
,
) , ( atan2
, , j ji
x
j ji
y n
u u =
(10)
k m
0
, (11)
l n
0
where and are the angle of with respect to the gear and pinion
tooth arc segment coordinate system and and are the angle of gear and
pinion arc segment, respectively.
m
ij
k
u
k l
2 2
) ( ) (
ij
y
ij
x p t
ij
u u r r + + =
(12)
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
14.3.2. ARC-POINT CONTACT
The arc-point contact conditions between the gear and the pinion can be
determined. A coordinate system is located at the center point of the
gear arc surfaces.
i
t
i
t
i
t
Z Y X
The position vector of the point p of pinion j with respect to the center point
of the gear tooth arc is defined in the global coordinate system such as in Eqs.
(5) and (6).
Necessary but not sufficient conditions for the contact to occur between the
pinion point and the gear tooth are k
r u u
ij
y
ij
x
+
2 2
) ( ) (
(13)
p t
ij
z p t
w w u w w +
(14)
where r is the radius of the gear arc segment, is half width of the gear
tooth and is half width of the pinion tooth.
t
w
p
w
If the above conditions are satisfied, it has to be checked if contact point is
existed in the arc range for the next step.
) , ( atan2
ij
x
ij
y m
u u =
(15)
k m
0
(16)
where is the angle of with respect to the gear arc segment coordinate
system and is the angle of arc segment.
m
ij
k
u
k
2 2
) ( ) (
ij
y
ij
x
ij
u u r + =
(17)
14-9
14.3.3. CONTACT FORCE MODEL
In the field of multi-body dynamics, one of the most popular approximations
of the dynamic behavior of a contact pair has been that one body penetrates into
the other body with a velocity on a contact point, thereafter the compliant normal
and friction forces are generated between a contact pair. In this compliant
contact force model, a contact normal force can be defined as an equation of the
penetration, which yields
3
2
1
m
m
m
n
c k f
&
&
&
+ =
(18)
where k and c are the spring and damping coefficients which are determined,
respectively and the is time differentiation of penetrated value
&
. The
exponents and generates a non-linear contact force and the exponent
yields an indentation damping effect. When the penetration is very small,
the contact force may be negative due to a negative damping force, which is not
realistic. This situation can be overcome by using the indentation damping
exponent greater than one. The friction force is obtained by
1
m
2
m
3
m
n f
f f =
(19)
where is the friction coefficient and its sign and magnitude can be
determined from the relative velocity of the pair on contact position.
14.4. KINEMATICS AND EQUATION OF MOTION FOR
SYSTEM DYNAMICS USING THE RECURSIVE FORMULAS
Recursive formulas using relative coordinates are very useful for gear system
dynamic analysis since gears in geared systems are usually rotated to one axis
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
direction. This section presents the relative coordinate kinematics for a contact
pair as well as for joints connecting two bodies.
Translational and angular velocities of the body coordinate system with
respect to the global coordinate system are respectively defined as
w
r&
(20)
Their corresponding quantities with respect to the body coordinate system
are defined as
=
w A
r A
Y
T
T
&
(21)
where is the combined velocity of the translation and rotation. The recursive
velocity and virtual relationship for a pair of contiguous bodies are obtained in
[16] as
Y
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Y B Y &
(22)
where denotes the relative coordinate vector. It is important to note that
matrices and are only functions of the . Similarly, the
recursive virtual displacement relationship is obtained as follows
1)i (i
q
(i
B
1)i1 1)i2 (i
B
1)i (i
q
1)i (i 1)i2 (i 1) (i 1)i1 (i i
+ = q B Z B Z
(23)
If the recursive formula in Eq. (22) is respectively applied to all joints, the
following relationship between the Cartesian and relative generalized velocities
can be obtained:
q B Y & =
(24)
where is the collection of coefficients of the and B
1)i (i
q&
14-11
[ ]
T
1 nc
T T
2
T
1
T
0
=
n
Y , , Y , Y , Y Y K
(25)
[
T
1 nr
T
) 1 (
T
12
T
01
T
0
=
n n
q , , q , q , Y q & K & & & ] (26)
where nc and nr denote the number of the Cartesian and relative coordinates,
respectively. Since in Eq. (24) is an arbitrary vector in q&
nr
R , Eqs. (22) and
(24), which are computationally equivalent, are actually valid for any vector
such that
nr
R x &
x B X & =
(27)
and
1)i - (i 1)i2 - (i 1) - (i 1)i1 - (i i
x B X B X + = (28)
where is the resulting vector of multiplication of and . As a
result, transformation of into is actually calculated by
recursively applying Eq. (28) to achieve computational efficiency in this
research. Inversely, it is often necessary to transform a vector in
nc
R X B
G
x
nc
nr
R x
nc
R Bx
R into
a new vector in G B
T
= g
nr
R . Such a transformation can be found in the
generalized force computation in the joint space with a known force in the
Cartesian space. The virtual work done by a Cartesian force Q is
obtained as follows.
nc
R
Q Z W
=
(29)
where must be kinematically admissible for all joints in a system.
Substitution of
Z
q B Z = into Eq. (29) yields
* T T T
Q q Q B q W = =
(30)
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
where . Q B Q
T *
T
= + =
&
(
(31)
where the is the Lagrange multiplier vector for cut joints [17] in and
represents the position level constraint vector in
m
R
m
R . The and Q are
the mass matrix and force vector in the Cartesian space including the contact
forces, respectively.
M
14.5. NUMERICAL RESULTS
A spur gear pair system is analyzed for the sake of numerical verification of
proposed methods as shown in Fig. 3. The shafts of the two gears are assumed to
be rigid and the only the compliance of contact force between meshing teeth is
considered in this model. The gear pare model is composed of 2 spur gears, 2
revolute joints, and a gear contact element. Rotational dampers are used for
resistance torque at revolute joints. A gear is driven by steady torque of 10 Nm.
1 1
, r
2 2
, r
Revolute joint & Rotational spring
damper
Applied torque
Contact element
Fig. 3 Gear pair model
14-13
Gear/Pinion
Module 3 mm
Number of teeth 50
Pressure angle 20
Radius of pitch circle 75 mm
Radius of outside circle 78 mm
Radius of base circle 70.477 mm
Radius of root circle 71.25 mm
Tooth width 20 mm
Elasticity modulus
2 9
/ 10 200 m N
Density
3 3
/ 10 85 . 7 m kg
Center distance 150 mm
Table 2 Design parameters of gear and pinion
Table 2 shows design parameters of the spur gear sets which are the inputs of
numerical simulation. Dynamic analysis of a spur gear pair is simulated during
0.08 sec. Gear speed is increased up to 500 rad/sec (4800 rpm) almost linearly as
shown in Fig. 4(a). It is found that the CPU simulation time is just 15 sec on a
Pentium IV 3.0 GHz platform personal computer. Figure 4(b) demonstrates the
dynamic transmission error (DTE= ) with respect to time domain when
a gear is driven at the constant torque of 10 Nm. As rotating speed of gear is
increased, dynamic transmission error (DTE) is changed by gear teeth contact.
Figure 5(a) and 5(b) show the time-domain DTE around mesh frequency of 1900
Hz and 3000 Hz. Magnitude and waveform of DTE are different in each mesh
frequency. Magnitude of DTE is around 30 and 3 micro meter, respectively.
These results show similar magnitude and exact dynamic pattern as compared to
experimental measurement results (in the reference Fig. 6 and 7) introduced by
Blankenship and Kahraman [7]. The minor differences between the proposed
method and referenced [7] might be expected from the dimensions, measurement
settings and noises.
2 2 1 1
r r +
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
(a) Rotational velocity of driven gear
(b) Oscillating DTE with respect to time
Fig. 4 Rotational velocity and DTE
14-15
(a) DTE at the mesh frequency of 1900Hz
(b) DTE at the mesh frequency of 3000Hz
Fig. 5 Oscillating DTE time history
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
The key advantage of the proposed method is the fast & efficient system
simulation of geared multibody dynamic system without losing the system
dynamic characteristics caused by gear pair contacts and their flexibility. An
Engine system with multi gear sets is illustrated as another geared system
example model. The system has 4 degrees of freedom, which has 13 bodies, 6
revolute joints, one translational joint, 14 fixed joints, and 2 sets of contacting
spur gear pairs. Crankshaft in this model is rotated by gas force and gear sets are
driven by rotation of crankshaft as shown in Fig. 6. In order to examine the
effect of gear contact dynamics, the proposed gear contact force model is
compared by constraint coupler model which should be ideal solution but not
realistic. Figure 7 shows well the difference of output velocity from the final
gear between proposed method and conventional dynamic anaysis using
constraint only. Dynamic analyses of both models are performed for 0.01 sec. It
is found that the CPU simulation time is just 85 sec for the proposed method on
a Pentium IV 3.0 GHz platform personal computer.
Fig. 6 Engine model with multi gear set
14-17
Fig. 7 Rotational velocity in output gear
14.6. CONCLUSION
This research proposes an efficient implementation algorithm of spur gear
contact mechanisms for the fast system dynamic analysis. Externally specified
dynamic forces, or assumptions about modeling the mesh forces by time-varying
stiffness and static transmission error are not required since dynamic mesh forces
are obtained by contact analysis directly at each time step. Arc-Arc and arc-point
kinematic interactions are presented and a compliant force model is used in this
investigation. The relative coordinate formulation is employed to generate the
equations of motion. Two numerical examples, a simple spur gear pair and an
engine transmission system, are illustrated and simulated numerically in this
investigation. A simple spur gear pair model shows the validation of the
proposed method with measurement results illustrated by reference, and engine
transmission system shows the advantages of the proposed method, respectively.
Consequently it is possible to simulate the entire geared system dynamic analysis
without losing its important dynamic characteristics, such as vibration and noise,
etc., with reasonable CPU time as represented in this investigation.
RecurDyn
/ Solver THEORETICAL MANUAL
DYNAMIC ANALYSIS OF CONTACTING SPUR GEAR PAIR FOR FAST SYSTEM SIMULATION
REFERENCES
1. ANSYS User Manual, ANSYS Inc., PA., USA
2. ABAQUS User Manual, ABAQUS Inc., RI., USA
3. Kahraman, A. and Singh, R., ''Non-Linear Dynamics of a Spur Gear Pair'',
Journal of Sound and Vibration, Vol.142, No.1, pp.47-75, 1990.
4. Amabili, M. and Rivola, A., ''Dynamic Analysis of Spur Gear Pairs: Steady-State
Response and Stability of the SDOF Model with Time-Varying Meshing
Damping'', Mechanical Systems and Signal Processing, Vol.11, No.3, pp.375-390,
1997.
5. Ozguven, H. N. and Houser, D. R., ''Mathematical Models used in Gear
Dynamics a Review'', Journal of Sound and Vibration, Vol.121, pp. 383-411,
1988.
6. Shing, T., Tsai, L., and Krishnaprasad, P., "An Improved Model for the Dynamics
of Spur Gear Systems with Backlash Consideration ", ASME-PUNLICATION-
DE, Vol.65-1, pp. 235-244, 1993.
7. Blankenship, G. W. and Kahraman, A., Gear dynamics experiments, Part-I:
Characterization of forced response, ASME, Power Transmission and Gearing
Conference, San Diego, 1996.
8. Kahraman, A. and Blankenship, G. W., Gear dynamics experiments, Part-II:
Effect of involute contact ratio, ASME, Power Transmission and Gearing
Conference, San Diego, 1996.
9. Kahraman, A. and Blankenship, G. W., Gear dynamics experiments, Part-III :
Effect of involute tip relief, ASME, Power Transmission and Gearing
Conference, San Diego, 1996.
10. Parker, R. G. and Vijayakar, S. M. and Imajo, T., ''Non-linear Dynamic
Response of a Spur Gear Pair : Modeling and Experimental Comparisons'',
Journal of Sound and Vibration, Vol.237, pp. 435-455, 2000.
14-19
RecurDyn
/ Solver THEORETICAL MANUAL
11. Bolton, K. M., Biarc curves, Computer Aided Design, Vol.7, No.2, pp.89-92,
1975.
12. Parkinson, D. B. and Moreton, D. N., ''Optimal biarc curve fitting, Computer
Aided Design, Vol. 23, No.6, pp.411-419, 1991.
13. Ryu, H. S., Huh. K. S., Bae, D. S. and Choi, J. H., ''Development of a Multibody
Dynamics Simulation Tool for Tracked Vehicles, Part I : Efficient Contact and
Nonlinear Dynamic Modeling'', JSME International Journal, Series C, Vol.46,
No.2, pp.540-549, 2003.
14. Lankarani, H. M., "Canonical Impulse-Momentum Equations for Impact
Analysis of Multibody System", ASME, Journal of Mechanical Design, Vol. 180,
pp. 180-186, 1992.
15. Bae, D. S., Han, J.