77-405 '

ON THE USE OF ATTACHMENT MODES IN SUBSTRUCTURE COUPLING FOR DYNAMIC ANALYSIS+

Roy R. Craig, Jr." and Ching-Jone Chang*>

U The University of Texas at Austin
Austin, Texas

I Abstract
X
- Substructure physical coordinate
vector
Substructure coupling, or component mode
s y n t h e s i s , may be employed i n t h e s o l u t i o n of dynam- Substructure l a b e l s
i c s problems for complex s t r u c t u r e s . Although Lagrange m u l t i p l i e r vectors
numerous substructure coupling methods have been
devised, l i t t l e a t t e n t i o n has been devoted t o Substructure generalized s t i f f n e s s
methods employing attachment modes. I n t h e present and mass matrices
paper t h e various mode sets (normal modes, con-
Substructure eigenvalue, and diagonal
s t r a i n t modes, attachment modes) w i l l be defined. matrix of eigenvalues
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

A generalized substructure coupling procedure w i l l

be described. Those substructure coupling methods 5s A generalized coordinate defined by
which employ attachment modes w i l l be described i n Eq. ( 5 2 )
d e t a i l . One of t h e s e methods is shown t o lead t o
A modal vector
r e s u l t s (e.g., system n a t u r a l frequencies) compkr-
a b l e t o , o r b e t t e r t h a n , those obtained by, t h e Matrix of modal vectors. The sub-
Hurty method. s c r i p t ( s ) give t h e i d e n t i t y of t h e
modal matrix.
Nomenclature System eigenvalue
A,B,C Constraint c o e f f i c i e n t matrices Amplitude of harmonic motion vector
---__
~'~'~'~'f
J,P,R,u,v,B } S e t s of physibal coordinates
Introduction
d .e Subscriots s i w i f v i n z deDendent and
- I - .

linearly-independent coordinates Since 1 9 6 5 when Hurty presented h i s paper on

substructure coupling, numerous methods f o r coup-
f Substructure physical force vector
l i n g substructures f o r dynamic a n a l v s i s have been
fa
Force vector used i n defining
attachment modes
- oresented. I n 1972 Benfield. e t . a i .
a number of t h e s e methods and came t o t h e conclu-
surveved

sion t h a t methods based on free-interface normal

G Pseudo f l e x i b i l i t y matrix
- modes, HOU'S method and Goldman's method, were
d i s t i n c t l y i n f e r i o r t o methods employing fixed-
F l e x i b i l i t y matrix of e l a s t i c modes
Ge
i n t e r f a c e normal modes or loaded-interface normal
Residual f l e x i b i l i t y matrix modes. Although Bamford 11 and MacNeal had
Ge
i,j Subscripts f o r i n t e r i o r and junction previously discussed attachment modes, none of t h e
coordinates methods surveyed by Benfield, e t . a l . incorporated
attachment modes. Rubin 6 extended MacNeal's
k ,m Substructure s t i f f n e s s matrix and technique f o r incorporating attachment modes, which
mass matrix he c a l l e d r e s i d u a l modes, and developed an imwoved
kl Term defined by Eq. ( 6 9 ) component modes model. Hintz presented a thor-
ough discussion of methods f o r generating s t a t i c a l -
k2 Term defined by Eq. ( A - 1 1 ) l y complete mode s e t s and introduced t h e "method of
K.M System s t i f f n e s s matrix and mass c o n s t r a i n t modes" and t h e "method of attachment
matrices modes." Using t h e Benfield truss of Ref. 9 , Hintz
L conducted convergence s t u d i e s on s e v e r a l v a r i a t i o n s
Lagrangian function defined by Eqs.
of t h e s e two b a s i c methods and showed t h a t t h e
(49) and (51) "method of attachment modes with unconstrained
ml Term defined by Eq. (73) normal modes" produced system frequencies of accur-
acy campemable t o those produced by t h e Hurty
P Frojection matrix defined by Eq. (24) method. Hintz's method gave b e t t e r accuracy on
P Substructure generalized coordinate lower modes; Hurty's gave b e t t e r accuracy on higher
vector modes but a l s o gave very good accuracy on lower
modes.
Qs Generalized force component
9 System generalized coordinate In a r e c e n t paper Craig and Chang presented a
generalized coupling procedure which permits a l l
S Transformation matrix defined by s u b s t r u c t u r e coupling methods t o be e a s i l y
Eq. (59) described. They a l s o surveyed t h e majority of
System k i n e t i c energy and p o t e n t i a l methods presented i n t h e l i t e r a t u r e up t o 1975.
enerm
+This work was supported by NASA under Grant NSG 1268. Dr. R.W. F r a l i c h of NASA Langley
Research Center was Technical Monitor.
*Associate Professor, Aerospace Engineering and Engineering Mechanics, Member AIAA
""Graduate Student, Engineering Mechanics

89
 -
In the present paper the generalized coupling modes are determined. The normal modes are
procedure is reviewed and those methods which obtained from eigenvalue equations of the form
employ attachment m@es are described in detail.
The relationship of the methods of MacNeal and (k - ,Az m) $ = 0 (4)
Rubin, as originally derived, to the generalized
Ritz-based coupling procedures is demonstrated. and it will be assumed that the modes are normal-
Convergence studies are presented which show that ized with respect to m, i.e.,
the methods of Hintz and Ruhin produce system fre-
quencies of accuracy comparable to the accuracy of T
en m en = Inn , @
: k Qn =' 'nn (5)
frequencies obtained by the Hurty method. A poten-
tial advantage of the Rubin method is the fact that
all interface degrees-of-freedom are eliminated where @ is a matrix whose columns are the compo-
from the final set of system coordinates. nent no&al modes. While the complete normal mode
set will be identified by a subscriptn, as in Eq.
Definition of Mode Sets (S), the normal mode set is usually truncated. The
truncated normal mode set will be labelled Ok (for
The mode sets to be employed in the present work kept modes).
have been defined previously in References 1 and 2 .
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

The most general type of component is one which is Fixed-interface normal modes were employed in
connected to adjacent components by redundant conjunction with constraint modes by Hurty and
interfaces. Figure 1 illustrates a simple truss Craig and Bampton 4. The emphasis in this paper,
component of this type. however, will be upon free-interface normal modes
and hybrid-interface normal modes, since these are
employed when attachment modes are used.

coordinates Free-interface normal modes are obtained from

the eigenvalue equation corresponding to Eq. ( 2 1 ,
i.e.,

Figure 1. An Unrestrained Truss Component

with Redundant Interface
The equation of'motion (undamped) for a compo- Hence, there are N = Ni + Nj free-interface
4
nent may be written in the form modes including Nrp rigid-body modes if the compo-
nent is externallv unrestrained.
m!i + kx = f (1)
or

m.. m..
[ii11 m i j [i
X.
+ [ii
k.. k..
11 I}]:I
x.
E

The physical cocrdjnates-constitute a set which

will be called P. I and J are subsets corresponding
(2)

to interior coorqinates and junction (interface)

coordinates respectively.
Figure 2. A Truss Component with
The physical coordinates, x, may be represented Partially-Restrained Interface
in terms of component generalized coordinates, p,
by the coordinate transformation Figure 2 shows a component subjecsed to a temporary
restraint on the coordinate set F , which is com-
x = 4 p (3) prised partially of junction coordinates and par-
tially of interior coordinates. Let B be the
The transformation matrix @ consists of p r e -
selected component modes of the following types: complement of F in P .
The resulting eigenvalue
normal modes of free vibration of the component, equation may be written
constraint modes, and attachment modes. The term
"modes" is used here in the sense of Ritz trial (kbb - A* mbb) $ = 0 (7)
vectors. Procedures for generating each of the
above classes of modes will now be described.
Constraint Modes and Rigid-Body Modes
Normal Modes
Let the physical coordinates be partitioned into
Component normal modes may be classified as a set C and its complement V A constraint .
d
fixed-interface normal modes, free-interface normal mode is defined by imposing a unit displacement on
modes, or hybrid-interface normal modes, depending one coordinate in the ? set and zero-displacement
on whether all, none or part of the interface coor- on the remaining coordinates of that C set.
dinates are restrained when the component normal Thus, the set of constraint modes is defined by

90
 t h r e e sets: E , Aand , where t h e E
set
provides r e z t r a i n t a g a i n s t r i g i d body motion and
($)
where t h e A set c o n s i s t s of physical coordinates
v where u n i t forces are t o be applied. Then t h e
attachment modes a r e defined by

where Rcc i s t h e s e t of r e a c t i o n s a t coordi-

nates. From t h e t o p row-partition
(14)
@VC = - k w kvc (.g )

The 5 s e t must be a t l e a s t s u f f i c i e n t t o nrevent

rigid-body motion so t h a t kw w i l l be nonsingular.
The c o n s t r a i n t modal matrix i s From t h e t o p two row-partitions of Eq. (14)
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

(10)
(15)

If a component has Nr rigid-body degrees-of-

freedom, then an R s e t of coordinates may be The attachment modes a r e thus
employed t o restrai! t h e component a g a i n s t r i g i d -
body motion. Let U be t h e complement-of _R i n
? . Then, if Eq. ( 8 ) i s w r i t t e n with C = R ,
= 0, t h e c o n s t r a i n t modes so obtained w i l l be (16)
rigid-body modes f o r t h e component, i . e . , 'a E

The rigid-body modes form t h e modal matrix

where Qur
Qr =

is determined by
[I] (12)

Then,

Qw = - -'
kUu kur
(13)

A l t e r n a t i v e l y , rigid-body modes may be defined

from kinematics alone or may be determined as a That is, attachment modes may be thought of as
subset of t h e f r e e - i n t e r f a c e normal modes of an columns of a f l e x i b i l i t y matrix.
unrestrained or p a r t i a l l y - r e s t r a i n e d component.
The attachment modes defined by E$. (14) a r e
Attachment Modes and Inertia-Relief Modes s t a t i c d e f l e c t i o n shapes r e s u l t i n g from applica-
Let be a subset of P . An attachment mode
t i o n of u n i t f o r c e s on coordinates i n t h e A s e t ,
and measured r e l a t i v e t o a prgscribed s t a t i c a l l y -
is defined as t h e " s t a t i c deflection" of t h e compo-
nent which r e s u l t s when-a u n i t force i s exerted on
determinate c o n s t r a i n t s e t , R . An a l t e r n a t i v e
manner of defining attachment modes f o r [Link]-
one coordinate of t h e A s e t while t h e remaining
s t r a i n e d component leads t o t h e so-called p t i a -
coordinates i n 1 a r e force f r e e . References 1
r e l i e f modes. These were used by MacNeal and
and 2 have defined attachment modes f o r both ',
r e s t r a i n e d and unrestrained components. I n t h e
l a t t e r case these references f o l l o w MacNeal
Rubin i n defining attachment modes f o r a Struc-
t u r e whose e x t e r n a l forces a r e e q u i l i b r a t e d by
and
Rubin

cussed by Hintz '.

and t h e i r significance i n contributing t o
a statically-com l e t e i n t e r f a c e mode s e t w a s d i s -

Let Eq. ( 3 ) be w r i t t e n
rigid-body i n e r t i a forces. These a r e c a l l e d
i n e r t i a - r e l i e f modes. x = (Pr Pr + 0, P, (19)
ci
Here w e w i l l first present t h e general d e f i n i - where Q? contains a complete s e t of e l a s t i c
t i o n of attachment modes of an unrestrained compo- deformation modes and where
nent, as given by Hintz l. Then we w i l l define
i n e r t i a - r e l i e f modes. Let P be divided i n t o

91
 -
T Then, with Cr chosen t o s a t i s f y Eq. (301, Eq. ( 2 9 )
mrT m mr = Irr , me m me = lee (20)
becomes
T -
ma = P ma (31)
If Eq. (19) is s u b s t i t u t e d i n t o Eq. (l), t h e
r e c u l t i n g equation of mo'tion f o r rigid-body modes
Thus, i n e r t i a - r e l i e f attachment modes which a r e
is
orthogonal t o rigid-body modes can he obtained by
Pr
" = 4; f (21) combining Eqs. ( 2 7 ) and ( 3 1 ) t o g i v e

The r e s u l t i n g rigid-body displacement i s

ia = (PT G P ) fa (32)
xr E 'r Pr (22)
From t h e above discussion it is c l e a r t h a t
attachment modes can be defined f o r an unrestrained
The applied forces, f , a r e reacted by r i g i d -
component by use of Eq. (le), which r e q u i r e s only
body d'Alembert i n e r t i a f o r c e s so t h a t t h e unre-
f l e x i b i l i t y c o e f f i c i e n t s . To generate i n e r t i a -
s t r a i n e d component i s subjected t o an e q u i l i b r a t e d
r e l i e f modes, however, r e q u i r e s knowledge of t h e
f o r c e system
mass matrix m , as required by Eq. (24).
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

fe = f-mxr = P f (23) I t i s i n s t r u c t i v e t o consider t h e i n e r t i a -

where r e l i e f mode s e t defined by Eq. (32) from another
P = I - m er r mT (2U) viewpoint. If Eq. (19) i s s u b s t i t u t e d i n t o Eq.
(l), t h e r e r e s u l t s
Equations (1$) and ( 2 3 ) may be used t o define
i n e r t i a - r e l i e f modes Pe + nee Pe = f (33)
where
, A: ,...
kww kwa

[aw kaa

krw kra km
kwr

kar] ]I[ 'wa

= [aw
'rw
'ww w
'a

'ra
Paa
'wr

'rr
pa]
Owa Aee = diag. (Al2 Gel

If t h e i n e r t i a of e l a s t i c modes is neglected and a

(34)

pseudo-static d e f l e c t i o n , xe , i s c a l c u l a t e d , then

(25)
- *-l T
Pe E ee @ e
and
where t h e Ora r e s u l t s from t h e f a c t t h a t t h e
forces Ia a r e balanced by i n e r t i a f o r c e s as a Xe = ce f
r e s u l t of zq. ( 2 3 ) . The i n e r t i a - r e l i e f attachment
where
mode s e t
Ge = 9 A-' OT
e ee e
-
Thus, Ge i s t h e f l e x i b i l i t y matrix associated
(26)
with e l a s t i c deformation. From t h e way i n which
Eq. (32) was derived, Ge can a l s o be expressed by

is measured r e l a t i v e t o E
a l s o be w r i t t e n i n t h e form
. Equation ( 2 5 ) can

- Linear Independence of Mode S e t s ; Truncation of

ma = G P fa (27)
Mode S e t s
where G has t h e s p e c i a l form'
In previous s t u d i e s of substructure coupling
(component modal synthesis) many d i f f e r e n t c-ina-
t i o n s of mode s e t s have been employed. Hurty and
Craig and Bampton used c o n s t r a i n t modes, with
C = J , and f i x e d - i n t e r f a c e normal modes. These
mode s e t s a r e l i n e a r l y independent s i n c e t h e con-
s t r a i n t modes a r e defined by u n i t displacement of
i n t e r f a c e coordinates while t h e normal modes a r e
where Eq. (17) is used t o define t h e submatrices based on a fully-constrained i n t e r f a c e . Hou's
in G. study employed only f r e e - i n t e r f a c e normal modes,
and again t h e s e form a l i n e a r l y independent s e t .
References 5 and-6 remove t h e rigid-body mode
contribution from Pa t o define an i n e r t i a - r e l i e f On t h e o t h e r hand, when attachment modes are
mode set IPa .
This may be done as follows. Let employed i n a modal s y n t h e s i s technique, t h e
question of l i n e a r independence of modes arises.
For example, i f a complete s e t of free-interface
4
= sa t mr cr (29) modes is supplemented by attachment modes', t h e
latter a r e l i n e a r l y dependent on t h e former. The
and choose Cr so t h a t "method of c o n s t r a i n t modes" and t h e "method of
IPT m Q-a
attachment modes," described by Hintz both
= O (30)
r employ mode sets which r e q u i r e a truncation of

92
 e l a s t i c modes i n order t o render t h e complete mode
s e t s l i n e a r l y independent.

Equation (37) leads t o a convenient method f o r

v making attachment modes and noma1 modes l i n e a r l y
independent. Equation (37) may be w r i t t e n
and (44)

Ee =
*e
1 1
ms ms
T
(39) rKa 0 1
s=1

Rodden ', f o r example, discussed t h e use of mea-

sured modes i n deribing s t r u c t u r a l influence coef-
f i c i e n t s using a mode-superposition approach
similar t o Eq. (39). The substructure coordinates forming p a r e not
independent. The equation o f displacement compati-
If Ee i s derived from Eq. (38), t h e contribu- b i l i t y a t the interface, i.e.,
t i o n of t h e kept normal modes can be subtracted
= x.6
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

leaving a r e s i d u a l f l e x i b i l i t y matrix xa (45)

1 1
Ge I: PT G P - mk \k-1
mkT (40) leads t o a c o n s t r a i n t equation which may be w r i t t e n
T
Thus, if Ge i s s u b s t i t u t e d f o r P G P i n Eq. Bp = 0 (46)
( 3 2 ) , t h e r e s u l t i n g attachment modes w i l l be
l i n e a r l y independent from t h e rigid-body modes and
a l s o from t h e e l a s t i c modes i n Qk provided, of When attachment modes (or i n e r t i a r e l i e f modes)
course, t h a t Nk + Na 2 Ne .The r e s u l t i n g s e t of a r e employed, t h e generalized displacements associ-
r e s i d u a l i n e r t i a r e l i e f modes is ated with t h e attachment modes can be i n t e r p r e t e d
as forces by v i r t u e of t h e way attachment modes a r e
'a Ge fa (41) defined, e.g., Eqs. (14) and (41). A s w i l l be
shown l a t e r i n t h e discussion of s p e c i f i c methods,
it i s possible t o enforce e x p l i c i t l y t h e i n t e r f a c e
force equilibrium equation
Coupling of Substructures

In t h i s section we w i l l describe a general sub-

f1 + !f = 0 (47)
../ s t r u c t u r e coupling procedure. Then we w i l l provide a t those i n t e r f a c e coordinates f o r which attachment
d e t a i l s on s e v e r a l s p e c i f i c methods. The general modes are defined. I n terms of p I Eq. (47) may
coupling procedure was introduced i n Ref. 2 . be w r i t t e n as

Let t h e system be composed of two substructures, Ap E: 0 (48)

a and 6 , which have a common (redundant, i n
general) i n t e r f a c e . Let From t h e Lagrangian function

1.T * 1 T T T
L = 7 P P p - ~ pK p t n A p t v B p (49)

where il and v a r e Lagrange m u l t i p l i e r vectors.

Equation (49) can be s i m p l i f i e d by introducing
(42)

f = , Q = [j , and p = [l]
Then Eq. (49) may be w r i t t e n

The d e r i v a t i o n of system equations of motion

w i l l be based on Lagrange's equation of motion with L
1 *T
= -2 p
. 1. T
! l p - - 2p
T
r p + o c p (51)
undetermined m u l t i p l i e r s . To t h a t end, expressions
f o r k i n e t i c energy, p o t e n t i a l energy and equations
The system equations of motion can be obtained by
of c o n s t r a i n t a r e required. The system k i n e t i c
applying Lagrange's equation i n t h e form
energy and p o t e n t i a l energy a r e given by

1 .T * 1 *T *
T I -x
2
mx = ~p u p
and (43)
d
U = -12 x T k x = T1 P T ~p where Qs is t h e generalized f o r c e associated with
5, , which can r e f e r t o e i t h e r ps or as For.
t h e substructure coupling problem with forces
r e s p e c t i v e l y , where
exerted only a t t h e i n t e r f a c e , Q, i s determined

93
 from t h e v i r t u a l work expression From Eq. (60), however, ST CT 5 0 , so Eq. ( 6 1 )
T reduces t o
6W = (6x3 f? + E T f B.
(ax.) T T
(53) M i + K q = S @ f
j l 1 1 (63)
v
Using t h e transformations @ and S of s p e c i f i c
= sp PUT fa t 8pBT PET f B = 6pT mT f methods, it may be shown i n each case t h a t Eq. (63)
reduces t o

When Eqs. (51) and ( 5 3 ) a r e s u b s t i t u t e d i n t o Eq.

M i + K q = 0 (64)
( 5 2 ) , t h e r e r e s u l t s t h e system equation of motion because of t h e i n t e r f a c e equilibrium equation
T T
!Jp+Kp = @ f + C 0 (54) f?
1
+ fS
1
= 0 (65)

together with t h e c o n s t r a i n t equation

c p = 0 (55) Description of S p e c i f i c Coupling Procedures

Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

It should be noted t h a t Eq. (54) i s a s e t of To i l l u s t r a t e t h e use of attachment modes i n

cqupled system equations due t o t h e presence of t h e substructure coupling, four s p e c i f i c procedures
Lagrange m u l t i p l i e r term. This is i n d i s t i n c t i o n w i l l be described: Hintz's "method of attachment
t o similar uncoupled equations presented by various modes with unconstrained normal modes," Rubin's
authors, e.g., Refs. 1, 7 . method ( a s derived by Craig and Chang), MacNeal's
method ( a s derived by Craig and Chang), and a
Reference 2 shows t h a t p r a c t i c a l l y a l l substruc- hybrid method. The two-component planar t r u s s
t u r e coupling methods solve t h e coupled s e t of employed by Benfield, e t . a l . w i l l be used t o
equations, Eqs. ( 5 4 ) and ( 5 5 ) , by introducing a i l l u s t r a t e t h e s e methods.
l i n e a r transformation of t h e form
Hintz's Method of Attachment Modes with
P = s q (56) Unconstrained Normal Modes

r]
Let p be p a r t i t i o n e d i n t o dependent coordinates, Hintz's method employs i n e r t i a r e l i e f modes,
pd , and ( l i n e a r l y ) independent coordinates, p~ , plus unconstrgined component normal modes. He
and l e t Eq. ( 5 5 ) be p a r t i t i o n e d accordingly, i . e . choosEs t h e A s e t t o be t h e i n t e r f a c e coordinate
s e t , J , and l e t s t h e component be supported a t
any R s e t n o t on t h e interface. Thus, a force
is applied a t each junction point while-the compo-
4 rdd 'del = (57) nent i s supported a t an R subset of I Equa- .
t i o n ( 2 5 ) determines t h e i n e r t i a r e l i e f modes with
PI? rj t E = i and = 3 .The unconstrained compo-
nent normal modes include N i - Nr e l a s t i c modes
Wt and Nr rigid-body modes.
C.
Figure 3 shows an fi s e t which might be used
i n solving f o r attachment modes of a t r u s s
component.

where q E pL . Then

When Eq. ( 5 8 ) is s u b s t i t u t e d i n t o Eq. (571, t h e r e

results
Figure 3 . Temporary'Supports f o r Obtaining
c s t 0 (60) Attachment Modes of a Truss
Component.
Equations ( 5 4 ) and ( 5 6 ) can now be combined
i n order t o derive a s e t of system equations i n The vectors and matrices employed i n H i n t z l s
terms of t h e independent coordinates, q Thus . method of attachment modes with unconstrained
normal modes a r e :
4 M i t K q = ST P T T T
f + S C (61)
wnere
T
M = S U S , K : ST r S (62)

94
 combinations of t h e modes employed i n t h e "method
of c o n s t r a i n t modes."

Rubin's Method as Derived by Craig and Chang

W
I n t h e Appendix t o t h i s paper t h e o r i g i n a l
method used t o d e r i v e Rubin's method i n Refs. 6 and
10 is given. Rubin apparently f a i l e d t o recognize
t h e r e l a t i o n s h i p of h i s method t o Rayleigh-Ritz
procedures s i n c e he s t a t e d 6, "The focus here is on
a new form of component-mode representation, not
based upon a Rayleigh-Ritz procedure.. .".
The f a c t
t h a t , i n numerical examples, both Rubin's method
and MacNeal's method always l e d t o upper bounds on
a l l frequencies obtained l e d t h e present authors

' = I 'aa 'ar 'ak I t o suspect t h a t t h e s e methods were, i n some fashion,

based on a Rayleigh-Ritz procedure. That r e l a t i o n -

1 s h i p w i l l now be demonstrated.
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

Lora 'rr 'rk

(66) Equations (4O)and(41) a r e used t o d e f i n e a s e t
of i n e r t i a - r e l i e f modes, 4, , for t h e e n t i r e s e t
of junction coordinates, i . e . , A = J .
In Hintz's
method, as described by Eqs. ( 6 6 ) , attachment mode
generalized coordinates p: and p g a r e employed,
and Eqs. ( 5 6 ) and (66e) a r e used t o eliminate pg
from t h e f i n a l s e t of system coordinates, q . This
leaves N j attachment mode coordinates pg i n q.
The present method i s based on t h e observation
(e.g., see Eq.(A-l)) t h a t when attachment modes
a r e defined i n terms of t h e r e s i d u a l f l e x i b i l i t y
matrix by Eq. (4l), t h e attachment mode coordinates
0 0 k:1 0 0 may be i d e n t i f i e d with junction forces by t h e
S - equations
i-~
+a --B +a --6 *a _--E $8 --B U
- fq , and p! = f B. (67)
aa aa 'ad ar *ab ak 'aa ar >*aa ak Pa 1

lo 0 0 : 1 0
Since fy and f @ are r e l a t e d t o each other by t h e
equation of intergace equilibrium, Eq. (47) or
Eq. ( W ) , Eqs. ( 6 7 ) imply N . c o n s t r a i n t s t h a t a r e
not present i n Hintz's methoi. ,

The following s e t of vectors and matrices may

be employed, t h e n , t o define a method which leads
t o a system equation of motion which is i d e n t i c a l
where $2: Z (;tal-' and where Sa is given by t o t h a t obtained by using a form of Rubin's compo-
Eq. ( 2 7 ) . nent modes model 6 . This r e l a t i o n s h i p is discussed
f u r t h e r i n t h e Appendix of t h i s paper.
Note t h a t
1
+ :N + N E + Nfl + N E ) modes a r e
(N-
used i n t h i s method, s o Nj i n t g r f a c e freedoms
a r e required i n addition t o component normal modes.
and r i g i d body modes. This is similar t o t h e s i t u -
a t i o n pertaining in t h e Hurty method and t h e
Craig-Bampton method
are retained.
',
where i n t e r f a c e coordinates

Equations ( $ 2 ~ ~ 1(,6 5 ) (66c), and (66e) may be

combined t o show t h a t S? mT f vanishes so t h a t ,
f o r Hintz's method described here, Eq. (63)
reduces t o t h e system equation of motion given i n
Eq. ( 6 4 ) .

Since t h e attachment modes i n Eq. ( 6 6 ~ )a r e not

made orthogonal t o t h e rigid-body modes or kept 6.
normal modes, t h e system matrices K and M w i l l kl 'jk
be f u l l y populated.
I -a
I 0
kk
In addition t o t h e "method of attachment modes,''
described a "method of
s -
4 6.
described above, Hintz
-k
c o n s t r a i n t modes". A s observed by Hintz, t h e kl 'qk 1 ik
u
"method of attachment modis" and t h e "method of
c o n s t r a i n t modes" a r e equivalent. This r e s u l t s 8
Ikk
from t h e fact t h a t t h e "method of attachment modes"
emplojrs modes which can be shown t o be l i n e a r

95
 where MacNeal's Method as Derived by Craig and Chang

In the preceding section it was shown that the

v transformation matrices and S in Eq. ( 6 8 )
It is of interest to note what happens when lead to the same system equation of motion which
Eqs. (68) a r e substituted into Eq. ( 5 4 ) with the would result from application of Rubin's component
aid of Eq5. (50). Equation (54) can be written in mode method. In the Appendix it may be seen that
the form "MacNeal's method," Eq. (A-71, and "Rubin's method,"
Eq. (A-13), have the same stiffness matrix, while
u
!laa
..a
Pa a
f Kaa Pt =
a
*ja (fg + V) .( II
MacNeal's method has a simpler system mass matrix.
It can be said that Rubin's method employs a COn-
sistent system mass matrix, while MacNeal's method
uses a system stiffness matrix based on 4 as
p: + AEk :p = mjkUT (f? + u) given in Eq. (68~)but uses a system mass matrix
1
(70) based on the transformation
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

Thus, MacNeal's method retains the contribution of

The presence O t h e Lagrange multipliers in these static (inertia-relief) modes only in the system
equations represents a distinction between these stiffness matrix. The system mass matrix contains
equations for the coupled system and similar equa- contributions only from the kept normal modes,
tions which previous authors have written for the
uncoupled components, e.g., Eq. (16) of Ref. 1. k
' '

The resulting system mass and stiffness

Using Eqs. (68) it can be shown that for ST6'Tf matrices are
to vanish from Eq. (63) it is necessary that Eq.
( 6 5 ) be satisfied. Thus, force equilibrium enters La, o q Kaq
explicitly in Eq. (68d) and also implicitly in
reducing the system equation of motibn to the form
given in Eq. (64).
The system mass and stiffness matrices have the
~

forms where the K terms m e given in Eqs. ( 7 2 ) .

MaE p KaE A Hybrid-Interface Method

In this paper the name hybrid-interface sub-

= [Ea MEJ ' and = LEa K 6 4 (71) structure coupling method refers to a coupling
procedure employing both constraint modes and
attachment modes which are associated with the
where interface coordinates. Figure shows a possible
choice of statically-determinate constraints, R ,
M~~ = lEk + maT
ik
m rn?
1 ik which include a portion of the interface coordi-
nates.
-
J
.
-.
(72) II
I
I
\ E
I
I

Figure 4. A Component with Hybrid Interface

Coordinates

Corresponding to the I?. interface coordinates

The constraints implied by Eq. (67) have thus led there will be defined Nr rigid-body modes.
to a set of system equations in which Only the kept Attachment modes will b$ dezined for the remailing
>.
normal mode coordinates remain. interface coordinates A = J - R , while W 5 I ,
The rigid-body modes are defined by Eq. (11) and
the attachment modes by Eq. (14) or Eq. (41).
Fybrid-interface modes are defined relative to the
R constraints by Eq. ( 7 ) . The coordinates

96
 ~

p; and prB may be i d e n t i f i e d w i t h t h e physical

displacements a t t h e R coordinates. Pii kia kir7
(77)
Thus, t h e vectors and matrices defining a
c-' hybrid-interface method a r e

Then t h e atrachilient modes are obtained from Eq.

(14) w i t h W E I .
From Eq. ( 7 7 ) and t h e top row
p a r t i t i o n of Eq. (14) we obtain

(76)
Qia = - ky? k
ii ia aa
(78)

li
The c o n s t r a i n t modes can be obfained r e l a t i v e t o
t h e i n t e r f a c e c o n s t r a i n t s e t R by expanding
Eq. ( 8 ) t o give
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

r-ik i c ki] =
'ir ail kci kcc kcr (79)

' = 'aa 'ar 'ak kri klrc krr Orc Rrc

Ora '
I
T Ork
where, i n t h e present c a s e , = .i From the .
first row-partition o f Eq. ( 7 9 ) is obtained
a B
r.a @ar 'ik "aa
E = '0.
1C
= - kii-1 kic (80)

Equations ( 7 8 ) and ( 8 0 ) can be combined t o give

[il
m.l a = m.i c 'aa
7 Hence,
0 0 0
Iza
0 0 0
= ' [i] (81)
;1
0 'a E 'aa Icc 'aa
0 0 ;'k
s = - S ,a
'aa aa 0
ma; aa
8ak ) Ora Orc

0 0 0
I& Thus. t h e attachment modes defined on = 9 - E A
E a r e i i n e a r c+ications of t h e c o n s t r a i n t modes
0 0 0
Ikk - defined on C = A .
Numerical Examples
where 2
B -1
(rnaa) .
Some numerical convergence s t u d i e s were made t o
compare t h e system n a t u r a l frequencies obtained by
The f i n a l number of coordinates i s N4, + N: ..+ N,.,.B t h e methods presented i n t h i s paper. Results
obtained using t h e method o f Craig and Bampton are
I t i s of i n t e r e s t t o note t h a t t h e above method is
presented i n order t o compare attachment mode meth-
very s i m i l a r t o Hurty's method 3. To show t h i s we
ods with a c o n s t r a i n t mode method. The Benfield
w i l l show tha? t h e attachment modes employed above
f o r t h e s e t A = J - R can be expressed as a
truss was used i n t h e s e s t u d i e s and t h e r e s u l t s
a r e presented i n t h e same graphical form used by
l i n e a r combination of-cons_trajnt modes defined on
t h e same s e t , i . e . , C = J - R H u r t y l s method
uses normal modes obtained with t h e e n t i r e i n t e r -
. Benfield, e t . a l . , i n order t o f a c i l i t a t e compani-
son with other methods included i n t h a t survey.
face J constrained, while t h e method proposed
I t may be noted i n Figs. 5-7 t h a t t h e number of
here employs modes defined r e l a t i v e t o a s t a t i c a l l y
determinate subset of 3 . modes of a s p e c i f i e d accuracy (O.l%, 0.5%, o r 1%)
is s i g n i f i c a n t l y higher fop t h e Rubin method than
f o r other methods, and t h a t t h i s is e s p e c i a l l y t r u e
Let t h e s t i f f n e s s matrix be p a r t i t i o n e d as
when t h e minimum number of degrees of fieedom is
follows:
r e t a i n e d i n t h e system model. I t appears t h a t t h e
advantage of t h e Rubin method over Hintz's method
lies i n t h e f a c t t h a t intepface coordinates a r e
reduced out i n t h e former. Hintz's method produces
more modes of 0.1% accuracy than does t h e Craig-
Bampton (Hurty) method, but l o s e s t h i s advantage

97
 when l e s s s t r i n g e n t accuracy requirements ( O . S % , Results and Conclusions
. 1%) a r e imposed. The consistent mass matrix of
Rubin's method appears t o give it a d i s t i n c t advan- Three c l a s s e s of modes which may be used i n cam-
t a g e over MacNeal's method for t h e present problem. ponent mode s y n t h e s i s a r e defined. A general
c/ substructure coupling procedure based on t h e
Lagrange m u l t i p l i e r technique i s presented. Four
*.e< mh,a"Ll-eny) s p e c i f i c substructure coupling procedures which
employ attachment modes a r e described. These a r e :
P Hintz's method of attachment modes with uncon-
s t r a i n e d normal modes, Rubin's method, MacNeal's
method and a hybrid-interface method. In t h e case
of Rubin's method and MacNeal's method, t h e
approach used i n deriving t h e s e methods i n t h e
present paper i s q u i t e d i s t i n c t from t h e methods
o r i g i n a l l y used, so t h e o r i g i n a l approach i s
b r i e f l y swmnarized i n t h e Appendix. It i s shown
t h a t both Rubin's method and MacNeal's method can
be viewed as Rayleigh-Ritz based methods. Numeri-
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

c a l convergence s t u d i e s a r e presented t o serve as a

b a s i s f o r comparing t h e four attachment mode meth-
ods presented i n t h i s paper and, i n a d d i t i o n , t h e
method of Craig and Bampton. These s t u d i e s show ,
t h a t Rubin's method produces system frequencies
which have accuracy comparable t o , or b e t t e r than,
t h e accuracy of frequencies obtained by t h e Craig-
Bampton method. The methods of Hintz and MacNeal
did n o t , i n general, produce b e t t e r r e s u l t s than
t h e Craig-Bampton method. The hybrid method pro-
duced r e s u l t s comparable t o those of t h e Craig-
Bampton method.

It appears t h a t f u r t h e r study of t h e attachment

mode methods described i n t h i s paper i s warranted
s i n c e it has been shown t h a t t h e s e methods a r e as
accurate a s methods based on c o n s t r a i n t modes.

Appendix: Methods of MacNeal and Rubin

In Ref. 1 0 substructure coupling equations

based on t h e component mode models proposed by
MacNeal and Rubin were presented. These w i l l
be reviewed here and compared with t h e r e s u l t s
presented i n Eqs. (75) and (71) r e s p e c t i v e l y .

Consider harmonic motion, i . e . , f = ? cos u t

and p = p cos u t . A s shown i n Refs. 6 and 1 0 , t h e
response t o harmonic e x c i t a t i o n can be approximated
by .
2 = @
k
5k t Q
a
i.] (A-1)

where ma is given by Eq. (41). The f i r s t term i n

Eq. (A-1) includes t h e i n e r t i a of t h e kept modes;
t h e second term r e f l e c t s only t h e f l e x i b i l i t y of
t h e attachment ( r e s i d u a l ) modes. The uncoupled
equations of motion f o r t h e kept modes a r e

When harmonic motion i s considered, t h e s e t of

equations of motion i s given by

( - u 2 1 k k t Akk) 6, = QT
ik
P.1 (A-3)

When Eq. (A-1) is r e s t r i c t e d t o t h e junction coor-

q i n a t e s , it becomes

x.1 = 0.
ik k
t 0. 6
ia j
z
(A-4)

Equations (A-3) and (A-4) a r e t h e two fundamental

equations f o r a substructure.

The coupling of substructures a and 6

98
 r e q u i r e s compatibility of i n t e r f a c e displacements k2 as given by Eq. (A-12). Reference 10 gives
and equilibrium of i n t e r f a c e f o r c e s , i . e . , a d d i t i o n a l d e t a i l s of t h e d e r i v a t i o n of Eqs. (A-7)
and (A-13).
-a
X.
= ;5 (A-5)
1 1 References
??1 t $51 = 0 (A-6)
1. Hintz, R.M., "Analytical Methods i n Component
When Eqs. (A-3) and (A-q)-are w r i t t e n for each sub- Modal Synthesis," AIAA Journal, V . 13, n. 8,
s t r u c t u r e , and f ? and f 6 a r e eliminated through Aug. 1975, pp. 1007-1016.
1 1
t h e use of Eqs. (A-5) and (A-61, t h e r e r e s u l t s a 2. Craig, R.R., Jr. and Chang, C-J., "A Review of
system equation Substructure Coupling Methods f o r Dynamic Analy-
sis," Advances i n Engineering Science, V . 2 ,
NASA CP-2001, Nov. 1976, pp. 393-408.
3 . Hurty, W.C., "Dynamic Analysis of S t r u c t u r a l
Systems Using Component Modes," A I A A Journal,
V . 3 , n. 4, April 1965, pp. 678-685.
Jr., and Bampton, M.C.C.,
4. Craig, R . R . , "Coupling
=
Downloaded by UNIV OF TEXAS AT SAN ANTONIO on September 7, 2014 | [Link] | DOI: 10.2514/6.1977-405

The system s t i f f n e s s and mass matrices here a r e o f Substructures f o r Dynamic Analyses,"

seen t o he those describing MacNeal's method, i . e . , Journal, V. 6, n. 7 , J u l y 1968, pp. 1313-1319.
Eqs. ( 7 5 ) .
5. MacNeal, R.H., "A Hybrid Method of Component
Mode Synthesis," J. of Computers and S t r u c t u r e s ,
Rubin suggested an improved representation by
V . 1, n. 4 , Dec. 1971, pp. 581-601.
replacing Eq. (A-1) by
6 . Rubin, S . , "ImDroved Component-Mode Representa-
X.1 = m.
ik
Pk + (m.i a + 0 2 :m m ma) 2:j (A-8) t i o n for S t r u c t u r a l Dynakc Analysis ,"-E
Journal, V. 13, n. 8 , Aug. 1975, pp. 995-1006.
or
7. Hou, S . , "Review of Modal Synthesis Techniques
2.
1
= [Link] Pk + ( a j a + w2 uaa) f..
1
(A-9)
and a New Approach," The Shock and Vibration
B u l l e t i n , N. 40, Pt. 4, Naval Research Labora-
A d e t a i l e d d e r i v a t i o n of Eq, (A-9) is presented i n t o r y , Washington, D . C . , Dec. 1969, pp. 25-39.
Ref. 1 0 . When Eqs. (A-3) and (A-9) a r e w r i t t e n for
substructures a and and Eqs. (A-5) and (A-6) e
are used t o eliminate f? and ?? t h e r e r e s u l t s
8 . Rodden, W.P., "A Method for Deriving S t r u c t u r a l
Influence Coefficients from Ground Vibration
1 Tests," A I A A J o u r n a l , V. 5, n. 5 , May 1967,
t h e system equation
pp. 991-1000.
v
* :;* ]k?'2 [-';:
'Z?k] ] F\ 1 (A-101
9. Benfield, W . A . , Bodley, C.S. and Morosow, G . ,
"Modal Synthesis Methods," Symposium on Sub-
s t r u c t u r e Testing and Synthesis, NASA, Aug.1972.
I
L .- Ik;'2'
- '
.*I& 'Z%ll
.'A I J
1'I1 1 -. R.R.. Jr.. and Chans.
10. Craie. _.C-J.. "Substructure
Coupling f o r Dynamic Analysis and Testing," t o
be published as a low-number NASA CR.
where
k2 = (4:
ia
+ e'
ja
+ w2 uaa
a + w '
uaa) (A-11)
11. Bamford, R.M., A Modal Combination Program for
Dynamic Analysis of S t r u c t u r e s , Tech. Memo.
33-290, J e t Propulsion Laboratory, Pasadena, CA,
Equation ( A - 1 0 ) is t h e general form of Rubin's J u l y 1967. '
method when applied t o substructure coupling. Note
t h a t it i s not i n standard eigenprohlem format
s i n c e w2 appears i n kz .
Rubin suggested a f u r t h e r approximation based
on a MacLaurin s e r i e s expansion of t h e right-hand
s i d e of Eq. (A-11). This r e s u l t s i n k2 being
approximated by

k2 = k1 - w 2 m 1 (A-12)

where kl i s given by E q . (69) and m l by Eq.

(73). When Eq. (A-12) is s u b s t i t u t e d i n t o Eq.
(A-lo) t h e r e s u l t i n g system equation can be mitten

The system s t i f f n e s s and mass matrices i n t h i s case

a r e Seen t o be those given i n Eqs. (71) and (72).
Thus, "Rubin's method a s derived by Craig and
Chang" i s t h e s p e c i a l form of Rubin's method which
incorporates t h e MacLaurin s e r i e s expansion of

99