Politecnico di Milano

M.Sc. in Mechanical Engineering

The Finite Element Method in Structural Dynamics

Mechanical System Dynamics

Prof. Roberto Corradi
Introductory comments

In the previous lectures, vibration of continuous systems was considered, focusing on

the fundamental analytical models which govern
■ transverse vibration of stretched strings
■ bending vibration of slender beams
■ axial vibration of bars
■ torsional vibration of shafts

All these analytical models

■ share the following hypotheses
 the continuous system can be assumed one-dimensional
 the axis of the string/beam/bar/shaft is rectilinear
 the cross-section of the string/beam/bar/shaft is constant (in shape and size)
 the material is homogeneous and linear elastic
■ result in PDE

An analytical solution to the problem of computing the natural frequencies and the
corresponding mode shapes can be found only in a limited number of cases, which
correspond to specific boundary conditions. 2
Introductory comments

In the more general case of structures which are characterized by

■ complex geometrical configurations
■ parts made of different materials
■ arbitrary combinations of boundary conditions
the only possibility is to make use of numerical methods.

The most popular and versatile one is the Finite Element Method (FEM).
This method consists in an approximate but very general formulation, which can be
applied to develop the numerical model of any kind of structure.

The Finite Element Method is based on the idea of dividing the structure in a certain
number of small portions (finite elements). In each element, some remarkable points
(nodes) are identified and a set of independent coordinates (nodal coordinates) is
selected to describe the displacement of each node. The motion of any other point
within the element is then expressed as a function of the nodal coordinates, through
proper shape functions.

3
Introductory comments

1D elements

3D elements

2D elements 4
Introductory comments

The Finite Element Method allows discretizing the continuous system, whose
vibration is then no more described through PDE. In their place, ODE represent the
dynamics of the discretized structure. As a consequence, any kind of finite element
model will result in a finite number n of degrees of freedom, and in a corresponding
set of n 2nd order ordinary differential equations.

The systematic nature of FEM makes it

■ general and widely applicable
■ particularly suitable for implementation in a computer software
■ easy to be integrated with CAD software, for both model development and
results visualization

Moreover, the matrix formulation of the motion equations allows easy and efficient
numerical solution of structural dynamics problems, by means of standard
algorithms available in mathematical libraries.

5
FE modelling of plane beam structures

In the following, the fundamentals of the Finite Element Method will be illustrated in
detail, by making reference to the case of plane beam structures.

First, one single beam finite element will be considered, so as to introduce

■ nodes and nodal coordinates
■ shape functions for axial and bending deformation
■ formulation of the mass and stiffness matrices, for one single beam element

Then, reference will be made to an entire beam structure, thus illustrating the
procedures for
■ developing the structure model, through proper coordinate transformation and
matrix assembling
■ accounting for boundary conditions and for concentrated/distributed loads
■ introducing damping into the model
■ performing structural dynamics analysis (free and forced vibration)

6
 Beam finite element (for in-plane dynamic or static structural analysis)
yjL
ELEMENT k xjL
qjL  xiL 
w(x,t)
yiL NODE j  L
xiL  yi 
qiL u(x,t)   L

x k   iL 
L
yL
NODE i
x
 xj 
Lk  yL 
xL  Lj 
 j 
L
x k = 6 nodal coordinates (time varying, in the dynamic case)

u(x,t) = axial displacement of the beam section at a distance x from the left node i
w(x,t) = transverse displacement of the beam axis, in correspondence with the
cross-section at a distance x from the left node i 7
 Shape functions – axial deformation

yjL
ELEMENT k xjL
qjL
w(x,t)
yiL NODE j
xiL
qiL u(x,t)
NODE i
x Lk u (x , t )  a  bx

 Shape functions allow the motion of any section along the beam element to be
expressed as a function of the nodal coordinates
 Shape functions are formulated in the element local reference system
 In the specific case of axial deformation of a beam element, the shape functions
are assumed to be linear functions of x, in that they represent the element
8
static deformation for loads concentrated at the nodes
Shape functions – axial deformation

u (x , t )  a  bx
By imposing that the axial displacement at the boundaries be equal to the
corresponding nodal coordinates, the constants a and b can be computed:

 a  x L

 u (0, t )  xiL (t ) 
i

   x L
 x L
 x 
  1  L 
L j i
 k
u ( L , t ) x j (t ) b
 Lk  k

 xiL (t )   0 
  L x L  L   
x  yi (t ) 
u (x , t )  1   xi  x j f u (x )   0 
 Lk  Lk   L
i (t )
  x 
x k (t )   L 
L
 
 x j (t )   Lk 
u (x , t )  f u (x ) x (t )
T L
 y L (t )   0 
 Lj 
k
 9 
 j (t )   0 
Shape functions – axial deformation

 x  x 
f u (x )  1   0 0 u varies linearly with x
T
0 0
 Lk  Lk 
u
x  is constant along
x 1i
L
u (x , t ) x 0
L
j
x the beam element

1 The two linear functions

x x   represent the axial
Lk
displacement of any cross-
section along the beam
xiL  0 u (x , t ) x Lj  1 element when a unit axial
displacement is imposed
at the left/right extremity,
1
x x  while keeping the other
Lk one fixed.
Lk 10
 Shape functions – bending deformation

yjL
ELEMENT k xjL
qjL
w(x,t)
yiL NODE j
xiL
qiL u(x,t)
NODE i
x Lk w(x , t )  a  bx  cx 2  dx 3

 Shape functions allow the motion of any section along the beam element to be
expressed as a function of the nodal coordinates
 Shape functions are formulated in the element local reference system
 In the specific case of bending deformation of a beam element, the shape functions
are assumed to be cubic functions of x, in that they represent the element
11
static deformation for loads concentrated at the nodes
Shape functions – bending deformation

w(x , t )  a  bx  cx 2  dx 3
By imposing that the cross-section displacement/rotation at the boundaries be equal
to the corresponding nodal coordinates, the constants a, b, c, d can be computed:
 w(0, t )  yiL (t )  a  yiL
 
 w b   L
 i (t )
L
 i
 x 
 x 0 3 L 3 L 2 L 1 L
   c   2 yi  2 y j  i   j
 w( Lk , t )  y j (t )
L
 Lk Lk Lk Lk
 w  2 L 2 L 1 L 1 L
   j (t )
L
d  3 yi  3 y j  2 i  2  j
 x x  Lk  Lk Lk Lk Lk

w(x , t )  f w (x ) x k (t )
T L

12
 Shape functions – bending deformation

w
yi = 1 yi = 1 qi = 0
w(x , t )  f w (x ) x (t )
T L
k yj = 0 qj = 0

 0  x
 3 2  w
 x  x   yi = 0 qi = 1
 2    3    1 
  L k   Lk 
 qi = 1 yj = 0 qj = 0
  3 2

 x 
 L    2      x  x x
 k  Lk   Lk  Lk   w

f w (x )    yi = 0 qi = 0
 0  yj = 1
 3 2  yj = 1 qj = 0
 x  x  
  2    3   
 k
L  k
L x
 
  x 3  x  2   w
 L  qj = 1 qi = 0
   
yi = 0
k  
   Lk   Lk   
    yj = 0 qj = 1

L T x
x   x
L
k i
L
y i
L
 i
L
x L
j y L
j   j
13
 Shape functions

u (x , t )  f u (x ) x k (t )
T L

w(x , t )  f w (x ) x k (t )
T L

 Axial deformation and bending deformation are decoupled

 The function u(x,t), which represents the axial displacement of the generic cross-
section along the beam element, is expressed as a linear combination of the two
linear shape functions for axial deformation, xiL(t) and xjL(t) being the coefficients
of this linear combination

 The function w(x,t), which represents the deflection of the beam axis due to
bending deformation, is expressed as a linear combination of the four cubic shape
functions for bending deformation, yiL(t), qiL(t), yjL(t) and qjL(t) being the
coefficients of this linear combination
14
Mass matrix of the beam finite element

1 Lk  u   w  
2 2

Under the assumption of slender beam: Tk   m       dx 

2 0  t   t  

1  u   u  1  w   w 
T T

    m   dx   
Lk Lk
 m  dx
2  t   t 
0 2  t   t 
0

u
u (x , t )  f u (x ) x (t )   f u (x ) x k (t )
T L T L

t
k

w
w(x , t )  f w (x ) x (t )   f w (x ) x k (t )
T L T L

t
k

Tk  ( x k )  f u (x )m f u (x ) dx   f w (x )m f w (x ) dx  ( x k )
1 L T  Lk T Lk T L

2  0 0 

Tk  ( x k ) [ M k ]66  [ M k ]66  ( x k )  ( x k ) [ M k ]( x k )
1 L T u w L 1 L T L L
2 2 15
 Mass matrix of the beam finite element

Lk Lk
[M ]   f u (x )m f u (x ) dx   f w (x )m f w (x ) dx
L T T
k 0 0

 13 mLk 0 0 1
6 mLk 0 0 
 0 13 11 2 9
 420
13
mL2k 
 35 mLk 210 mLk 0 70 mLk

 0 11
mL2 1
mL3
0 13
mL2
 140
1
mL3k 
[M k ]   1
L 210 k 105 k
1
420 k

 6 mLk 0 0 3 mLk 0 0 
 0 9
70 mLk
13
mL2
0 13
35 mLk  210
11
mL2k 
 3 
420 k

 0  420
13
mL2k  1401
mL3k 0  210
11
mL2k 105 mLk 
1


Note that not only the length Lk, but also the mass m per unit length may be specific
for the considered element k. The subscript k is omitted, just to simplify the notation.
16
 Stiffness matrix of the beam finite element
2 2
1  u 
Lk 1  w Lk
2
  EA   dx   EJ  2  dx 
(shear deformation
Vel , k
2 0  x  2 0  x  is neglected)

T T
1 Lk  u   u  1 Lk   w    2w  2
    EA   dx    2  EJ  2  dx
2 0  x   x  2 0  x   x 
u
u (x , t )  f u (x ) x (t )   f 'u (x ) x k (t )
T L T L

x
k

2w
w(x , t )  f w (x ) x k (t )   x
T L T L
f '' ( ) x (t )
x 2 w k

 ( x k )  f 'u (x ) EA f 'u (x ) dx   f ''w (x ) EJ f ''wT (x ) dx  ( x k )

1 L T  Lk T
Lk L
Vel , k
2  0 0 

 ( x k ) [ K k ]66  [ K k ]66  ( x k )  ( x k ) [ K k ]( x k )
1 L T u w L 1 L T L L
Vel , k 17
2 2
Stiffness matrix of the beam finite element

Lk Lk
[K ]  
L
k f 'u (x ) EA f ' (x ) dx  
T
u f ''w (x ) EJ f ''wT (x ) dx
0 0

 L1k EA 0 0  L1k EA 0 0 
 
 0 12
Lk3
EJ 6
Lk2
EJ 0  12
L3
EJ 6
Lk2
EJ 
 
k

 0 L
6
2 EJ
4
Lk EJ 0  L62 EJ 2
Lk EJ 
[ K kL ]   1 k k

  Lk EA
1
0 0 Lk EA 0 0 
 0  12 3 EJ  L62 EJ 0 12
EJ  L62 EJ 
 L k k Lk3 k 
 0 6
2 EJ
2
EJ 0  L62 EJ 4 
Lk EJ
 Lk
Lk
k 
Note that not only the length Lk, but also the axial and bending stiffness EA and EJ
may be specific for the considered element k. The subscript k is omitted, just to
simplify the notation. 18
 Procedure for developing the FE model of a structure

1) Mesh generation

2) Definition of the global and local reference systems

3) Removal of external constraints and introduction of corresponding constraint forces

4) Energy functions formulation in the local nodal coordinates of each element

5) Coordinate transformation from the local to the global reference system

6) Matrix assembling, for the entire structure

7) Model of the structural damping

8) Re-introduction of external constraints and matrix partition

19
Step 1: mesh generation

Reference example

The structure is modelled through beam elements. Nodes are conventionally

numbered by Arabic numerals (from 1 to 6), while Roman numerals are used for
elements (from I to V). Note that the numbering of nodes and elements is arbitrary
and does not have any consequence on the results of the calculation.
20
Step 1: mesh generation

When generating the mesh of a plane beam structure, the following two criteria
shall be adopted:
1) a node shall be positioned in correspondence with any kind of discontinuity
(i.e. a change of the section shape/dimension or of the beam material
properties, the intersection of two or more beams with different axis orientation,
the occurrence of a concentrated mass/spring/damper)
2) for assigned geometry of the beam cross-section and material properties, the
length of any beam finite element shall not exceed a certain limit value, which
depends on the maximum driving frequency Wmax of the forces acting on the
structure
reference formula (pinned-pinned beam)
2
  EJ k
 (1)
k   k(1)  Wmax  Lk  Lmax
 Lk  mk
in other words, each single finite element shall be working in its
quasi-static region 21
Step 2: definition of the global and local reference systems

22
Step 3: removal of external constraints and introduction of corresponding
constraint forces

Note that constraints will be reintroduced at the 8th and last step of the procedure
23
Step 4: energy functions formulation in the local nodal coordinates
of each element

1 L T
T   Tk Tk  ( x k ) [ M kL ]( x k )
L
The dissipation function is
k 2 not considered at this stage.
1 L T L L
Vel  Vel , k
Damping will be introduced
Vel , k  ( x k ) [ K k ]( x k ) later, at the 7th step of the
k 2
procedure.
 W   Wc   Wext   Wc    Wext ,k
k

 Wc  H1 x1G  V1 y1G  M 11G  H 6 x6G  V6 y6G  M 66G   xTc R
 H1 
V 
 1
 M1 
R 
 H6 
 V6 
  24
 M 6 
 Step 4: energy functions formulation in the local nodal coordinates
of each element

 Wext    Wext ,k
k

CASE 1 – external forces concentrated at the nodes

 same procedure as for the constraint forces

CASE 2 – external forces concentrated in a point internal to a beam element

yjL
F
Fy qjL
xjL  Wext ,k  Fx u (x )  Fy w(x )
yiL Fx
xiL  u (x )  f Tu (x )  x kL  ( x kL )T f u (x )
qiL
x  w(x )  f Tw (x )  x kL  ( x kL )T f w (x )

Wext ,k  ( x kL )T  f u (x ) Fx  f w (x ) Fy   ( x kL )T F kL, conc 25

 Step 4: energy functions formulation in the local nodal coordinates
of each element

CASE 3 – external forces distributed along a beam element

b px (x )  p(x ) cos( b )
p y (x )  p(x )sin( b )
yjL
p(x) Lk
qjL
xjL  Wext ,k   px (x ) u (x )dx 
0
Lk
yiL
xiL   p y (x ) w(x )dx
0
qiL
x  u (x )  f Tu (x )  x kL  ( x kL )T f u (x )
Lk  w(x )  f Tw (x )  x kL  ( x kL )T f w (x )

 ( x )  f u (x ) px (x )dx   f w (x ) p y (x )dx   ( x k )T F26k , distr

 Lk Lk
 Wext ,k L T L L
k  0 0 
Step 5: coordinate transformation from local to global reference system

This step is needed to express all nodal displacements with respect to the common
global reference system.
The coordinate transformation is based on the following linear relationships (a
generic node i for a certain element k is considered):

x1. _
— xiG cosak+yiG smak
-

yliL = —xl.G sin 05k + i cos ark

8? = 3f
U cos 05k sin 05k
£f=[1ar]£? m1= —sinak cosak
O 0

xj = [/L
k 1xG.
L G
The same for node ji:: xL. xj

 x
L
  x
G
 [k ] [0]  G
xk   L 
L i
xk   G 
G i
x 
L
 x    k k
x
G

 [0] [k ]
k k
 x j   x j  27
Step 5: coordinate transformation from local to global reference system

The coordinate transformation introduced in the previous slide is substituted into

the expressions of the kinetic energy, the elastic potential energy and the virtual
work of the external forces, for each single beam element.
Therefore, the 6x6 mass matrix, the 6x6 stiffness matrix and the 6x1 column
matrix of the input forces for each single beam element, in the global reference
system, are obtained through simple matrix products.

1 L T L L 1 G T 1 G T G G
Tk  ( x k ) [M k ]( x k )  ( x k ) [ k ] [ M k ][ k ]( x k )  ( x k ) [ M k ]( x k )
T L G

2 2 2
1 L T L L 1 G T 1 G T G G
 ( x k ) [ Kk ]( x k )  ( x k ) [ k ] [ Kk ][ k ]( x k )  ( x k ) [ K k ]( x k )
T L G
Vel , k
2 2 2

 Wext ,k  ( x kL )T F kL  ( xGk )T [ k ]T F kL  ( xGk )T F Gk

F k  F k , conc  F k , distr
L L L L
Fk accounts for both concentrated and distributed forces:
28
Step 6: matrix assembling, for the entire structure

The global nodal coordinates of the entire structure are collected in a column
matrix x . The order assigned to the nodal coordinates is arbitrary, provided that
constrained coordinates be kept separate from the free ones.

 x2G 
 G
 y2 
2G 
yG  G
 x3 
 xF 
x  
 G   xC 
5 
xG  xG 
 1 
 
 G
6  29
Step 6: matrix assembling, for the entire structure

The assembling of the mass, stiffness and external forces matrices for the entire
structure can be performed by means of proper extraction matrices.

x   EI  x
G
For example, for element I: I
yG

xG

30
Step 6: matrix assembling, for the entire structure

In this way, the mass/stiffness matrices of one single beam element can be
expanded from 6x6 to nxn (from 6x1 to nx1 in the case of the external forces matrix),
n being the total number of nodal coordinates in the entire structure model (for the
considered reference example, n=18). For example, in the case of element I:
1 G T G G 1 T
TI  ( x I ) [ M I ]( x I )  x [ EI ]T [ M IG ][ EI ]x
2 2
[EE1 [[0] [o] [o] [o] [I] [o]
[I] [0] [0] [0] [0] [0] ]

I
[M12] .-0 .-0
IOI 10110110119110:

IOI 16230339110210:
IGIZOIZOISOIZOI

[0] IOIZOIZOISOIZOI
[0]
[EI]T[Mf][EI]=
[0]
[M112]
31
_ [0]
I
Step 6: matrix assembling, for the entire structure

The expanded matrices of all the elements are then summed up, to obtain the
aggregate structure matrices:

1 G T G G 1 T  1 T
T  Tk   ( x k ) [ M k ]( x k )  x  [ Ek ] [ M k ][ Ek ]  x  x [ M ] x
T G

k k 2 2  k  2

1 G 1 T  1 T
Vel  Vel , k   ( x k )T [ K kG ]( x k )  x  [ Ek ]T [ K kG ][ Ek ]  x  x [ K ] x
G

k k 2 2  k  2

 W   Wext   Wc    Wext ,k   Wc   ( xGk )T F Gk   xTc R external

k k forces

 G  FF 
 W   x  [ Ek ] F k    xT [ EC ]T R   xT F
constraint
F 
T T

  forces
 k   C
F R 
Note that a suitable extraction matrix [EC] is introduced here for the constrained
coordinates also.
32
 Step 6: matrix assembling, for the entire structure

[MII] [MIII]  x2G 

 G
x2 x3 x4 x5 x1 x6  y2 
2G 
x2 [MI22] [MI21]
 G
[MIV]  x3 
   xF 
x  
 G   xC 
x3 5 
 xG 
x4  1 
 
 G
6 
x5 [MV55] [MV56]

yG
x1 [MI12] [MI11]

xG
x6 [MV65] [MV66]
33
Step 7: model of the structural damping

1 T
D  x [C ] x [C ]   [ M ]  b [ K ]
2
The damping matrix [C] is expressed as a linear combination of the mass and
stiffness matrices [M] and [K]. Therefore, if the modal coordinate transformation
x  [] q is adopted, not only [M] and [K] but the damping matrix [C] also will be
diagonalized and:

ci  bi
ci   mi  b ki  xi   
2mii 2i 2

Typically, the non-dimensional damping ratios xi for a given set of vibration modes
are obtained from experimental modal analysis. The procedure for estimating the
corresponding values of  and b is illustrated in the next two slides.

34
 Step 7: model of the structural damping

xi

Suppose that the damping ratios xi have been obtained for a set of vibration modes
(with natural frequenciesi), through proper modal testing and identification
procedures. The coefficients  and b can be estimated from a least-squares
minimization process. The figure above shows the x- curve which best fits the
experimental data set.
35
 Step 7: model of the structural damping

  b1  1 1 
 2  2  x1  2 2 x
 1  1     1   
         [ A]    B
    b    b 
bn  1 n  x n 
   xn
 2n 2  2n 2 

The linear system above, in the unknowns  and b, is over-determined (matrix [A]
being nx2). Therefore a least-square solution is required, which can be formulated
in the following way:

   
 
1
[ A] [ A]    [ A]T B
T
 b   [ A]T
[ A] [ A]T
B
b   
pseudoinverse matrix [A]+
36
Step 8: re-introduction of external constraints and matrix partition

1 T 1 T
T  x [M ] x Vel  x [ K ] x
2 2
1 T
D  x [C ] x  W   xT F
2
By applying Lagrange equations in matrix form, the following motion equations
are obtained:

[ M ] x  [C ] x  [ K ] x  F
The equations above correspond to the free structure, i.e. to the unconstrained
one. External constraints can be re-introduced by proper separation of free and
constrained nodal coordinates and consequent matrix partition:

 xF  free nodal coordinates (unknowns)

x 
 xC  constrained nodal coordinates
37
(assigned functions of time – zero in case of fixed constraints)
 Step 8: re-introduction of external constraints and matrix partition

[ M ] x  [C ] x  [ K ] x  F

[ M FF ] [ M FC ]  x F  [CFF ] [CFC ]  x F  [ K FF ] [ K FC ]  x F   F F 
[ M ] [ M ]  x   [C ] [C ]  x   [ K ] [ K ]  x    F  R 
 CF CC   C   CF CC   C   CF CC   C   C 

equivalent nodal forces which account for the assigned

 FF 
F  
time-dependent concentrated/distributed loads
 C
F  R  unknown constraint forces

 [ M FF ]x F  [ M FC ]xC  [CFF ]x F  [CFC ]x C  [ K FF ]x F  [ K FC ]x C  F F


 [ M CF ]x F  [ M CC ]x C  [CCF ]x F  [CCC ]x C  [ K CF ]x F  [ K CC ]x C  F C  R

38
Integration of the motion equations

The first one of the two matrix equations in the previous slide can be reformulated
in the following way:

[M FF ]x F  [CFF ]x F  [ K FF ]x F  F F  ([M FC ]xC  [CFC ]xC  [ K FC ]xC )

All the terms at the right side of the equation above are assigned functions of time.

The equation above can then be integrated to compute xF (t), which is given by the
sum of three contributions:
 free system response (homogeneous solution)

 system forced response to the external input forces FF (particular solution)

 system forced response to the motion xC imposed at the constraints

(particular solution)

The latter contribution will be zero in case of fixed constraints (xC =0).

Once that the solution xF has been computed, xF can be substituted into the second
matrix equation (see the previous slide), to calculate the constraint forces R . 39
Computation of the structure’s natural frequencies and mode shapes

The equations for free undamped vibration (analogous to those of any multi-dof
discrete system) are considered:

[ M FF ]x F  [ K FF ]x F  0

The structure’s natural frequencies and mode shapes can then be obtained by
computing the eigenvalues and the eigenvectors of matrix [MFF]-1[KFF] :

x F  Xe ˆ j0t  ( 2[ M ]  [ K ]) Xˆ  0
0 FF FF

( 2[ I ]  [ M ]1[ K ]) Xˆ  0
0 FF FF

The eigenvalues of matrix [MFF]-1[KFF] are the square natural frequencies 0i
2

The eigenvectors of matrix [MFF ]-1[K ˆ

X
(i )
FF] are the modal vectors

In general, the structure’s free response to assigned initial conditions will be given
by the superposition of the contributions of all its vibration modes. 40
 Computation of the steady-state structure’s response to harmonic input
1st contribution: external forces

[ M FF ]x F  [CFF ]x F  [ K FF ]x F  F F  F 01e jWt

x F 1  X 01e jWt  (W 2 [ M FF ]  jW[CFF ]  [ K FF ]) X 01  F 01

X 01  (W2[M FF ]  jW[CFF ]  [ K FF ])1 F 01

[G( jW)]  (W2[M FF ]  jW[CFF ]  [ K FF ])1  frequency response matrix

(it is a square matrix)
X 01  [G( jW)]F 01  x F1 (t )  Re( X 01e jWt )

In case of input with more than one harmonic contribution, the response to each
one of them can be computed separately and then summed up, to obtain the
aggregate response (superposition principle).

41
 Computation of the steady-state structure’s response to harmonic input
2nd contribution: motion imposed at the constraints

[M FF ]x F  [CFF ]x F  [ K FF ]x F  ([ M FC ]xC  [CFC ]xC  [ K FC ]xC )

[M FF ]x F  [CFF ]x F  [ K FF ]x F  (W2[M FC ]  jW[CFC ]  [ K FC ]) X C 0e jWt

[M FF ]x F  [CFF ]x F  [ K FF ]x F  F 02e jWt equivalent nodal forces (due to the

motion imposed at the constraints)

X 02  [G( jW)]F 02  x F 2 (t )  Re( X 02e jWt )

In case of both external input forces and motion imposed at the constraints, the
response to each one of the two contributions can be computed separately and
then summed up to obtain the aggregate response (superposition principle).

42
Computation of the steady-state structure’s response to harmonic input
2nd contribution: motion imposed at the constraints

Note that a frequency response matrix [GC(jW)] can be defined, which relates
the harmonic motion imposed at the constraints to the steady-state structure’s
harmonic response:

X 02  [G( jW)]F 02 F 02  (W2[M FC ]  jW[CFC ]  [ K FC ]) X C 0

X 02  [G( jW)](W2[M FC ]  jW[CFC ]  [ K FC ]) X C 0

X 02  [GC ( jW)] X C 0 note that [GC(jW)] is not necessarily a

square matrix (in general, it is not)

[GC ( jW)]  [G ( jW)](W 2[ M FC ]  jW[CFC ]  [ K FC ]) 

 (W 2 [ M FF ]  jW[CFF ]  [ K FF ]) 1 (W 2[ M FC ]  jW[CFC ]  [ K FC ])
43