(Adams) Theoretical Background
(Adams) Theoretical Background
(Adams) Theoretical Background
• modal superposition
• component mode synthesis,
• mode shape orthonormalization
• kinematics of markers on flexible bodies
• applied forces
• flexible body equations of motion
Copyright
2003
c MSC.Software Corporation. All rights reserved.
2003BFLXTH-01
Printed in the United States of America.
All product names are trademarks of their respective companies.
1
1.1 History of flexible bodies in ADAMS
The first attempt to automatically interface with Finite Element Method
(FEM) software was in a product called ADAMS/FEA. In ADAMS/FEA
the FEM software used Guyan Reduction to automatically condense the
entire set of FEM degrees of freedom (DOF) to a reduced number of DOF.
In the Guyan reduction method, a set of user-defined master nodes are
retained and the remaining set of slave nodes are removed by condensation.
Only stiffness properties are considered during the condensation, and
inertia coupling of master and slave nodes are ignored. This is why Guyan
reduction is sometimes referred to as static condensation.
Guyan reduction condenses the large, sparse FEM mass and stiffness
matrices down to a small, dense pair of matrices, with respect to the master
DOF.
The challenge in ADAMS/FEA was to represent the master nodes using
PART elements and an NFORCE element. While the condensed stiffness
could be captured correctly by the NFORCE, the dense, condensed mass
matrix from the Guyan reduction did not always lend itself to being
represented by an “equivalent” lumped mass matrix. The goals of matching:
• total mass
• center-of-mass location
• moments of inertia
• natural frequencies
2
In 1996 an alternative modal flexibility method was introduced in a product
called ADAMS/Flex. Rather than being based on ADAMS primitives like
PART and NFORCE elements, ADAMS/Flex introduced a new inertia
element, the FLEX BODY.
=1∗ −2∗
3
with a much smaller number of modal DOF. We refer to this reduction in
DOF as modal truncation.
Equation 1.1 is frequently presented in a matrix form
u = Φq (1.2)
where q is the vector of modal coordinates and the modes φi have been
deposited in the columns of the modal matrix, Φ. After modal truncation
Φ becomes a rectangular matrix. The modal matrix Φ is the
transformation from the small set of modal coordinates, q, to the larger set
of physical coordinates, u.
This raises the question: How do we select the mode shapes such that the
maximum amount of interesting deformation can be captured with a
minimum number of modal coordinates? In other words, how do we
optimize our modal basis?
4
The Craig-Bampton method [?] allows the user to select a subset of DOF
which are not to be subject to modal superposition. These DOF, which we
refer to as boundary DOF (or attachment DOF or interface DOF), are
preserved exactly in the Craig-Bampton modal basis. There is no loss in
resolution of these DOF when higher order modes are truncated.
The Craig-Bampton method achieves this with a very simple scheme. The
system DOF are partitioned into boundary DOF, uB , and interior DOF,
uI . Two sets of mode shapes are defined, as follows:
Figure 1.1: Two constraint modes for the left end of a beam that has attach-
ment points at the two ends. The figure on the left shows the constraint mode
corresponding to a unit translation while the figure on the right corresponds
to a unit rotation.
5
Figure 1.2: Two fixed-boundary normal modes for a beam that has attach-
ment points at the two ends.
The relationship between the physical DOF and the Craig-Bampton modes
and their modal coordinates is illustrated by the following equation.
( ) " #( )
uB I 0 qC
u= = (1.3)
uI ΦIC ΦIN qN
where
ΦIC are the physical displacements of the interior DOF in the constraint
modes
ΦIN are the physical displacements of the interior DOF in the normal
modes
6
stiffness transformation is
" #T " #" #
T I 0 KBB KBI I 0
K̂ = Φ KΦ =
ΦIC ΦIN KIB KII ΦIC ΦIN
" #
K̂CC 0
= (1.4)
0 K̂N N
while the mass transformation is
" #T " #" #
T I 0 MBB MBI I 0
M̂ = Φ MΦ =
ΦIC ΦIN MIB MII ΦIC ΦIN
" #
M̂CC M̂N C
= (1.5)
M̂CN M̂N N
where the subscripts I, B, N and C denote internal DOF, boundary DOF,
normal mode and constraint mode, respectively. The caret on M̂ and K̂
denotes that this is generalized mass and stiffness.
Equations 1.4 and 1.5 have a few noteworthy properties:
• M̂N N and K̂N N are diagonal matrices because they are associated
with eigenvectors.
• K̂ is block diagonal. There is no stiffness coupling between the
constraint modes and fixed-boundary normal modes. (See reference
[?] for details.)
• Conversely, M̂ is not block diagonal because there is inertia coupling
between the constraint modes and the fixed-boundary normal modes.
7
dynamic content. However, the raw Craig-Bampton modal basis has certain
deficiencies that make it unsuitable for direct use in a dynamic system
simulation. These are:
These problems with the raw Craig-Bampton modal basis are all resolved
by applying a simple mathematical operation on the Craig-Bampton modes.
The Craig-Bampton modes are not an orthogonal set of modes, as
evidenced by the fact that their generalized mass and stiffness matrices K̂
and M̂, encountered in equations 1.4 and 1.5, are not diagonal.
By solving an eigenvalue problem
Nq∗ = q (1.7)
8
The effect on the superposition formula is
M M M
∗
φ∗i q∗
X X X
u= φi qi = φi Nq = (1.8)
i=1 i=1 i=1
9
Figure 1.3: Constraint mode with an unknown frequency contribution
3. Although the removal of any mode constrains the body from adopting
that particular shape, the removal of a high-frequency mode such as
the one depicted in figure 1.4 is clearly more benign than removing
the mode depicted in figure 1.3. The removal of the latter mode
prevents the associated boundary node from moving relative to its
neighbors. Meanwhile, the removal of the former mode only prevents
boundary edge from reaching this degree of “waviness”.
10
• Flexible marker kinematics
• The marker measures, (for example DX, WZ, PHI, ACCX, and so on)
that appear in expressions and user-written subroutines require
information about position, orientation, velocity, and acceleration of
markers
Position
where
~x is the position vector from the origin of the ground reference frame to the
origin of the local body reference frame, B, of the flexible body.
11
Figure 1.5: The position vector to a deformed point P 0 on a flexible body
relative to a local body reference frame B and ground G.
~sp is the position vector of the undeformed position of point P with respect
to the local body reference frame of body B.
x is the position vector from the ground origin to the origin of the local
body reference frame, B, of the flexible body, expressed in the ground
coordinate system. The elements of the x vector, x, y and z, are
generalized coordinates of the flexible body.
12
sp is the position vector from the local body reference frame of B to the
point P , expressed in the local body coordinate system. This is a
constant.
G B
A is the transformation matrix from the local body reference frame of B
to ground. This matrix is also known as the direction cosines of the
local body reference frame with respect to ground. In ADAMS,
orientation is captured using a body fixed 3-1-3 set of Euler angles, ψ,
θ and φ. The Euler angles are generalized coordinates of the flexible
body.
where Φp is the slice from the modal matrix that corresponds to the
translational DOF of node P . The dimension of the Φp matrix is
3 × M where M is the number of modes. The modal coordinates qi ,
(i = 1 . . . M ) are generalized coordinates of the flexible body.
13
Velocity
Orientation
14
The orientation of marker J relative to ground is represented by the Euler
transformation matrix, GAJ . This matrix is the product of three
transformation matrices:
G J
A = GAB BAP PAJ (1.18)
where
G B
A is the transformation matrix from the local body reference frame of B
to ground.
B P
A is the transformation matrix due to the orientation change due to the
deformation of node P .
P J
A is the constant transformation matrix that was defined by the user
when the marker was placed on the flexible body.
The matrix BAP requires more attention. The direction cosines for a vector
of small angles, θp , are
1 −θpz θpy
B P
A = θpz
1 −θpx
= I+θ
θ̃p (1.19)
−θpy θpx 1
where the tilde denotes the skew operator
0 −az ay
a × b = az
0 −ax
b = ãb = −b̃a (1.20)
−ay ax 0
Angular velocity
15
1.3.2 Applied loads
The treatment of forces in ADAMS distinguishes between point loads and
distributed loads. This section discusses the following topics:
• Distributed loads
fz tz
QM
16
force on the translational coordinates is obtained by transforming FK to
global coordinates.
QT = GAK FK (1.24)
where GAK is given in Eq. 1.18. The generalized translational force is
independent of the point of force application.
An applied torque does not contribute to QT .
Generalized Torque: The total torque on a flexible body, due to F~ and
T~ is T~tot = T~ + p~ × F~ , where p~ is the position vector from the origin of the
local body reference frame of the body to the point of force application.
The total torque, can be written in matrix form, with respect to the ground
coordinate system as:
Ttot = GAK TK + p
e GAK FK (1.26)
17
T G K
TI = GAB A TK (1.29)
and then projected on the mode shapes. The force is projected on the
translational mode shapes and the torque is projected on the angular mode
shapes
QF = ΦTp FI + Φ∗Tp TI (1.30)
where Φp and Φ∗p are the slices of the modal matrix corresponding to the
translational and angular DOF of point P , as discussed in section 1.3.1.
Note that since the modal matrix Φ is only defined at nodes, point forces
and point torques can only be applied at nodes.
Distributed loads
Mẍ + Kx = F (1.31)
Here M and K are the FEM mass and stiffness matrices for the flexible
component, and x and F are, respectively, the physical nodal DOF vector
and the load vector.
Equation 1.31 is transformed into modal coordinates q using the modal
matrix Φ
ΦT MΦq̈ + ΦT KΦq = ΦT F (1.32)
This modal form of the equation simplifies to
18
where M̂ and K̂ are the generalized mass and stiffness matrices and f is the
modal load vector.
The applied force is likely to have a global resultant force and torque.
These show up as loads on the rigid body modes and are treated in
ADAMS as point forces and torques on the local reference frame, as covered
in the previous section. The global resultant force and torque are not
discussed further.
The projection of the nodal force vector on the modal coordinates
f = ΦT F (1.34)
19
Residual forces and residual vectors
f = ΦT F (1.38)
There is one special case of force truncation that deserves mention. This
case is best illustrated by considering a FEM node with incomplete
20
stiffness, as found on solid elements or shell elements. Applying a load to
this node leads to a singularity in the FEM analysis. When Craig-Bampton
modes are generated for this model, they will share a common attribute —
the mode shape entry for this DOF is zero in all the modes. Consequently,
any attempts in ADAMS to apply a load in this DOF will fail, because the
load does not project on any of the modes and the structure will appear
infinitely stiff. It is recommended that no loads be applied in ADAMS that
could not have been applied in the FEM software.
Preloads
21
Figure 1.3.2 illustrates the force-deformation relationship of the process
described above. The undeformed state is defined by operating point O. As
the body deforms, it is brought through a non-linear path to a deformed
state A. A linear model of the body at O, such as might have been defined
by an ADAMS flexible body, would incorrectly have predicted an operating
point at B rather than at A. Note further, that if the body is linearized at
A, and a modal description exported to ADAMS in the form of a preloaded
flexible body, a limited range of validity must also be observed. Fully
unloading the ADAMS flexible body would bring it to operating point C,
which is not correct.
A preload is applied in ADAMS in the same way modal loads described in
the previous section are applied, except that the preload is not under the
user’s control. The preload cannot be disabled or scaled because it is
considered an immutable property of the flexible bodies with an associated
deformed geometry. Only one preload can be defined for any given flexible
body.
A preload is an internal load and as such only operates on the modal
coordinates. There is no global resultant force. In other words, there is no
22
load on the rigid body DOF. If this were otherwise, the flexible body would
have a tendency to accelerate itself, which would be counterintuitive.
Unless the external load that gave rise to the preload is reapplied within
ADAMS, the preloaded flexible body will recoil. If the flexible body
originated from a linear finite element model, it will recoil to its undeformed
shape. If the body came from a non-linear analysis, the effect will be more
like that described in figure 1.3.2. If the body is constrained to other bodies,
this tendency to recoil will cause the body to push on the other bodies.
where
23
The Lagrangian is defined as
L=T −V
The velocity from Eq. 1.15 can be expressed in terms of the time derivative
of the state vector ξξ̇
h i
vp = I −GAB (sp g
+ up )B G B
A Φp ξξ̇ (1.42)
We can now compute the kinetic energy. The kinetic energy for a flexible
body is given as
1Z 1X
T = ρ vT vdV ≈ mp vpT vp + Gω BT G B
P Ip ω P (1.43)
2 V 2 p
where mp and Ip are the nodal mass and nodal inertia tensor of node P ,
respectively. Note that Ip is often a negligible quantity which arises when
reduced continuum descriptions, i.e. bars, beams, or shells, are employed in
your flexible component model. Lumped masses and inertia may also
contribute to this term.
24
Substituting for v and ω and simplifying yields an equation for the kinetic
energy in ADAMS’ generalized mass matrix and generalized coordinates.
1 T
T = ξξ̇ M(ξξ) ξξ̇ (1.44)
2
For clarity of presentation we partition the mass matrix, M(ξξ), into a 3 × 3
block matrix
Mtt Mtr Mtm
MTtr Mrr Mrm
M(ξξ) = (1.45)
T T
Mtm Mrm Mmm
where the subscripts t, r and m denote translational, rotational, and modal
DOF respectively.
The expression for the mass matrix M(ξξ) simplifies to an expression in nine
inertia invariants.
Mtt = I 1 I (1.46)
h g3 i
Mtr = −A I 2 + I j qj B (1.47)
I3
Mtm = AI (1.48)
h h i i
Mrr = BT I 7 − I 8j + I 8T
j qj − I 9ij qi qj B (1.49)
h i
Mrm = BT I 4 + I 5j qj (1.50)
Mmm = I 6 (1.51)
The explicit dependence of the mass matrix on the modal coordinates is
evident. The dependence on orientation coordinates of the system comes
about because of the transformation matrices A and B.
The inertia invariants are computed from the N nodes of the finite element
model based on information about each node’s mass, mp , its undeformed
location sp , and its participation in the component modes Φp . The discrete
form of the inertia invariants are provided in Table 1.1.
25
N
X
I1 = mp (scalar)
p=1
XN
I2 = mp sp (3 × 1)
p=1
XN
I 3j = mp Φp j = 1, . . . , M (3 × M )
p=1
N
mp sfpΦp + IpΦ0p
X
I4 = (3 × M )
p=1
XN
I 5j = mpφ
gΦ
pj p j = 1, . . . , M (3 × M )
p=1
N
T
mpΦTp Φp + Φ0 p IpΦ0 p
X
I6 = (M × M )
p=1
N
mp sfp T sfp + Ip
X
I7 = (3 × 3)
p=1
XN
I 8j = mp sfpφ
g
pj j = 1, . . . , M (3 × 3)
p=1
XN
I 9jk = mpφ
gφ g
pj pk j, k = 1, . . . , M (3 × 3)
p=1
26
In the elastic energy term, K is the generalized stiffness matrix which is, in
general, constant. Only the modal coordinates, q, contribute to the elastic
energy. Therefore, the form of K is
Ktt Ktr Ktm 0 0 0
K = KTtr Krr Krm = 0 0 0 (1.53)
T T
Ktm Krm Kmm 0 0 Kmm
The damping forces depend on the generalized modal velocities and are
assumed to be derivable from the quadratic form
1
F = q̇T Dq̇ (1.56)
2
which is known as Rayleigh’s dissipation function. The matrix D contains
the damping coefficients, dij , and is generally constant and symmetric.
27
In the case of orthogonal mode shapes, the damping matrix can be
effectively defined using a diagonal matrix of modal damping ratios, ci .
This damping ratio could be different for each of the orthogonal modes and
can be conveniently defined as a ratio of the critical damping for the mode,
ccr
i . Recall that the critical damping ratio is defined as the level of damping
that eliminates harmonic response as seen in the following derivation.
Consider the simple harmonic oscillator defined by uncoupled mode i.
mi q̈i + ci q̇i + ki qi = 0 (1.57)
where mi , ki and ci denote, respectively, the generalized mass, the
generalized stiffness, and the modal damping corresponding to mode i.
Assuming the solution qi = eλt , leads to a characteristic equation
m i λ2 + c i λ + ki = 0 (1.58)
which has the solution
q
−ci ± j 4mi ki − c2i
λ= (1.59)
2mi
The critical damping of mode i, is the one that eliminates the imaginary
part of λ q
ccr
i = 2 ki m i (1.60)
Defining ci as a ratio of critical damping introduces the modal damping
ratio, ηi , which is referred to as CRATIO in the ADAMS dataset.
ci = ηi ccr
i (1.61)
The solution to Eq. 1.57 is
q √
ki
−ηi mi
t j (1−η 2 ) ωi t
qi = e e (1.62)
q
where ωi = mkii is the natural frequency of the undamped system. This
solution ceases to be harmonic when ηi = 1, which corresponds to 100% of
critical damping.
28
Constraints
ξ ,ξξ̇,ξξ̈ the flexible body generalized coordinates and their time derivatives
29
λ Lagrange multipliers for the constraints
30