University of Engineering & Technology,

Peshawar, Pakistan

CE-412: Introduction to Structural Dynamics and

Earthquake Engineering

MODULE 9
MODAL RESPONSE SPECTRUM ANALYSIS

Prof. Dr. Akhtar Naeem Khan & Prof. Dr. Mohammad Javed
drakhtarnaeem@[Link] mjaved@[Link]

1
 Modal decoupling of the EOMs

It is already known that the equations of motion for a MDOF with

lumped mass system and undergoing only lateral displacement can
be written as:

[m]{u}+ [c ]{u}+ [k ]{u} = {p(t)}

CE-412: MODULE 9 (fall 2015) 2

 Modal decoupling of the EOMs
Let [Φ ] be the modal matrix (matrix of mode shapes) in which the
nth column is the nth mode shape of vibration (i.e. each column
represents a particular mode shape).
Recalling the results of Prob M8.2

1.000 1.00 1.000  0.446 1.000 0.802 

[Φ ] = φ1 φ2 φ3 = 1.802 0.445 − 1.247 = 0.804 0.445 − 1.000
2.243 − 0.802 0.555  1.000 − 0.802 0.445 

{u} = [Φ ]{q} Where {u} is displacement vector and {q} is

the modal amplitude vector.
{u} = [Φ ]{q}
{u} = [Φ ]{q}
CE-412: MODULE 9 (fall 2015) 3
 0. 5q1
1.00 q1

0.804

a b c d ? (ϕ & q)
0.446
q1 ( t ) = A1Cosωn1t + B1Sinωn1t
ϕ1 *
u1=ϕ1*q1
q1 0.50q1

0.402q1
0.804q1

0.446q1 0.223q1

a b c d
?
CE-412: MODULEu9 (fall
=ϕ 2015)
*q 1 1 1
 Modal decoupling of the EOMs
Substituting values of {u}, {u} & {u} from slide 3 in the:
[m]{u}+ [c ]{u}+ [k ]{u} = {p(t)}
[m][Φ ]{q}+ [c ][Φ ]{q}+ [k ][Φ ]{q} = {p(t)}
Pre-multiply both sides by [Φ ]T

[Φ] [m][Φ]{q}+ [Φ] [c][Φ]{q}+ [Φ] [k ][Φ]{q} = [Φ] {p(t) }

T T T T

[M ]{q}+ [C ]{q}+ [K ]{q} = {P(t)}

Where [M ] = Modal mass matrix
[K ] = Modal stiffness matrix
[C] = Modal damping matrix
{P( t )} =CE-412:
Modal MODULE
(applied) 9forces vector
(fall 2015) 5
 Modal decoupling of the EOMs
Mode shapes are orthogonal (i.e., each mode shape is independent
of others) as shown below
0.802 0.445
1.000
Third mode shape

0.445
0.804 1.00

0.446 1.00 0.802

First mode shape Second mode shape

Mode shapes being normalized by taking greatest floor term taken as 1

CE-412: MODULE 9 (fall 2015) 6

 Modal decoupling of the EOMs

A matrix [A ] said to be orthogonal if [A ]T [A ] = [I] where [I] is

an identity matrix in which diagonal terms are 1 and off diagonal
terms are 0 and therefore det [I]=1. [M ] , [K ] and [C] are diagonal
matrices (i.e., matrices in which off diagonal terms are zero)
m3
m3
k3
m2
m2 m1 0 0 
m1 k2 [m] =  0 m2 0 

 0 m 3 
m1 0
k1

CE-412: MODULE 9 (fall 2015) 7

 Modal decoupling of the EOMs
Since [M ] , [C] and [K ] are diagonal matrices so the N
coupled equations replaces by N uncoupled equation for SDOF
systems
M q + C q + K q = P (t)
n
 n n
 n n n n

Where Mn= Generalized mass for the nth natural mode

Kn= Generalized stiffnes for the nth natural mode
Cn= Generalized damping for the nth natural mode
Pn(t)= Generalized force for the nth natural mode

CE-412: MODULE 9 (fall 2015) 8

Modal decoupling of the EOMs

The equation given on previous slide is for nth mode of MDOF of

order N. All the independent equations for N modes in matrix form
can be written as
Mq + Cq + Kq = P(t)

Where M= Diagonal matrix of the generalized modal masses

K= Diagonal matrix of the generalized modal stiffnesses
C= Diagonal matrix of the generalized modal dampings
P(t) = Column vector of the generalized modal forces Pn(t)

CE-412: MODULE 9 (fall 2015) 9

Modal analysis for earthquake forces
The uncoupled equations of motions for earthquake excitations can

[M ]{q}+ [C ]{q}+ [K ]{q} = {Peff (t)} = [Φ] {peff (t )}

be written as
T

Where {p (t)} = −[m]{}

eff
ι u (t ) g

{ι}= Influence vector (refer slide 15 for details) of size Nx1.

{ι} = {1}Nx1 for structures where the dynamic degrees of freedom
are displacements in the same direction as the ground motion

  { }
⇒ [M ]{q} + [C ]{q} + [K ]{q} = Peff (t) = −[Φ ] [m]{}
T ι ug ( t )

CE-412: MODULE 9 (fall 2015) 10

Modal analysis for earthquake forces

[Φ ] [m]{}
ι = {L}
T L and m has
Replacing same units

⇒ [M ]{q}+ [C ]{q}+ [K ]{q} = {Peff (t)} = −{L}ug (t )

For nth mode M n qn + Cn q n + K n qn = − Lnug (t )

Mn 2ζ n M nωn Kn Ln
qn + q n + qn = − ug (t )
Mn Mn Mn Mn
Ln
or qn + 2ζ nωn q n + ωn qn = −
2
ug (t )
Mn

CE-412: MODULE 9 (fall 2015) 11

 Modal Participation factors

The term Ln/Mn has been given the name of participation factor for
the nth mode and is represented by Γn (capital Greek alphabet for

{φn } [m]{}
Gamma)
Ln
T
ι
Γn = =
M n {φn } [m] {φn }
T

Γn is usually considered a measure of the degree to which the nth

mode participates in the response. This terminology is misleading,
however, because Γn is not independent of how the mode is
normalized, nor a measure of the modal contribution to a response
quantity.
The magnitude of the participation factor is dependent on the
normalization method used for the mode shapes.
CE-412: MODULE 9 (fall 2015) 12
Participation factors
Once the modal amplitudes {q} have been found the
displacements of the structure are obtained from
{u} = [Φ ]{q}
The above mentioned equation to determine modal displacements
cancel out the effect of normalization carried out to calculate q
(slide 11).
The displacements associated with the nth mode are given by

{un (t)} = [φn ]{qn (t)}

CE-412: MODULE 9 (fall 2015) 13

 Effective weight of structure in n th mode, Wn
Effective weight of structure in nth mode=
{φn }T [m]{}{
ι φn }T [m]{}
ι Ln Ln 2
Wn = g= (L n g ) = Γn (L n g) = g
{φn }T [m]{φn } Mn Mn
It shall be noted that the sum of the all effective weights for an
excitation in a given direction ( i.e. for a given {ι}) should equal
the total weight of the structure. Note, this may not be the case
where rotational inertia terms also exist in the mass matrix.
Many building codes require that a sufficient number of
modes be used in the analyses such that the sum of the effective
weights is at least 90% of the weight of the structure. This
provides a measure on the number of modes required in the
analysis.
CE-412: MODULE 9 (fall 2015) 14
 Influence (Direction) vector {}
ι
{ι} = Influence vector= {1}Nx1 for structures where the dynamic degrees of
freedom are displacements in the same direction as the ground motion

u3 u2

u2 u3
u1
u1

Direction of EQ is Direction of EQ is
horizontal horizontal
1 1 
   
{ι} = 1 {ι} = 1 
1 0
   
CE-412: MODULE 9 (fall 2015) 15
Base Shear Force in the structure in the n th node, Vbn

The base shear in nth mode can be determined using relation

2
 An  Ln  An 
Vbn = Wn   = g 

 g  Mn  g 
Where Wn = Effective weight of structures in nth mode

CE-412: MODULE 9 (fall 2015) 16

 Distribution of nodal (joint) forces in the structure
from the base shear

In many design codes the first step is to compute the modal base
shear force and this is then distributed along the structures (shown
on next slide) to each degrees of freedom.
The distributed loads are assumed to give the same displacements
in the structure as those generated by the exciting base shear.

{ f n } = [m]{φn }
Vbn
Ln

CE-412: MODULE 9 (fall 2015) 17

Distribution of nodal (joint) forces in the structure
from the base shear
m3n
f3n

≡
m2n
f2n

m1n
f1n

Vbn Vbn distributed along the

Base shear acting in structure in nth mode
nth mode, Vbn

Vbn = f1n + f 2 n + f 3n = ∑ f n
CE-412: MODULE 9 (fall 2015) 18
Response spectrum model analysis: Example
A 3 story R.C. building as shown below is required to be
designed for a design earthquake with PGA=0.3g, and its elastic
design spectrum is given by Fig 6.9.5 multiplied by 0.3). Carry
out the dynamic analysis by using the above mentioned design
spectrum. Take: m3

 Story height = 10ft

k3
m2
Total stiffness of each story = 250 kips/in.
 Weight of each floor = 386.4 kips
m1 k2

k1

CE-412: MODULE 9 (fall 2015) 19

Mass and stiffness matrices
m=W/g = (386.4k) /(386.4 in/sec2)= 1.0 kip-sec2/in

1 0 0  500 − 250 0 

[m] = 0 1 0  [k ] = − 250 500 − 250
0 0 1  0 − 250 250 

500 − ωn 2 − 250 0 
 
[k ] − ωn [m] =  − 250
2
500 − ωn
2
− 250 
 0 − 250
2
250 − ωn 


CE-412: MODULE 9 (fall 2015) 20

 Natural frequencies

[ ]
Setting det [k ] − ωn [m] = 0
2
yields following values

ω = 7.04 rad / sec

n1

ω = 19.69 rad / sec

n2

ω = 28.49 rad / sec

T = 0.89 sec
n3

n1

T = 0.32 sec
n2

T = 0.22 sec
n3

CE-412: MODULE 9 (fall 2015) 21

Mode shapes
Normalized coordinates of first mode shape

[[k ] − ω n1
2
[m]]{φ1} = 0
500 − ωn12 − 250 0  φ11 
  
⇒  − 250 500 − ωn1 − 250  φ21  = 0
2

 − 250 250 − ωn1  φ 

2
 0  31 

ω = 49.56 and φ = 1
2
Substituting
n1 11

CE-412: MODULE 9 (fall 2015) 22

 Mode shapes
450.44 − 250 0  1 
 
⇒  − 250 450.44 
− 250 φ  =0
  21

 0 − 250 200.44 
φ 31


First row gives 450.44 − 250φ21 = 0 ⇒ φ21 = 1.80
Second row gives − 250 + 450.44φ − 250φ = 0
21 31

⇒ −250 + 450.44(1.80) − 250φ = 0 ⇒ φ = 2.24

31 31

φ11  1.00  1.00/2.24  + 0.45

       
⇒ {φ1} = φ21  = 1.80  = 1.80/2.24  = + 0.80
φ  2.24 2.24/2.24 + 1.00 
 31       
CE-412: MODULE 9 (fall 2015) 23
Mode shapes ω = 387.7
2

n2

500 − 387.7 − 250 0  φ 

 
12

⇒  − 250 500 − 387.7 − 250  φ  =0

  22

 0 − 250 250 − 387.7  φ 32



112.3 − 250 0  1 
  
⇒ − 250 112.3 − 250 φ  =0
  22

− 137.3  
 0 − 250 φ 32 
⇒ φ = 0.45 φ12  + 1.00 
22    
φ =
 22   + 0.45 
& φ = −0.80 φ  − 0.80 
 32   
32

CE-412: MODULE 9 (fall 2015) 24

Mode shapes
φ31  + 1.00  + 0.80
     
Similarly φ23  = − 1.25  = − 1.00 
φ  + 0.56 + 0.45
 33     
1.00 + 1.00 + 1.00 
[Φ ] = φ1 φ2 φ3 = 1.80 + 0.45 − 1.25 
2.24 − 0.80 + 0.56
0.45 + 1.00 + 0.80
= 0.80 + 0.45 − 1.00 
 
1.00 − 0.80 + 0.45
CE-412: MODULE 9 (fall 2015) 25
Modal mass, Mn and participation factor,Γn

Ln {φn } [m]{}
T
ι
Γn = =
M n {φn } [m] {φn }
T

For structure given in problem{ι}={1}

Ln {φn } [m]{1}
T
∴ Γn = =
M n {φn } [m] {φn }
T

L1 {φ1} [m]{1}
T
Γ1 = =
M 1 {φ1} [m] {φ1}
T

CE-412: MODULE 9 (fall 2015) 26

Modal mass, Mn and participation factor,Γn
φ11 
T
1 0 0 1
  0 1 0 1
L1 = {φ1} [m]{1} = φ21 
T
  
φ 
 31  0 0 1 1
T
0.45 1 0 0 1
  0 1 0 1
L1 = 0.80   
1.00  0 0 1 1
 
1

L1 = 0.45 0.80 1.00 1
1

2 2
L1 = 2.25 kip - sec /in. = 27 kip - sec /ft
CE-412: MODULE 9 (fall 2015) 27
Modal mass, Mn and participation factor,Γn
M 1 = {φ1} [m] {φ1}
T

T
0.45 1 0 0 0.45
  0 1 0 0.80
M 1 = 0.80   
1.00  0 0 1 1.00 
 
0.45
 
M 1 = 0.45 0.80 1.00 0.80
1.00 
 
M 1 = 1.84 kip - sec 2 /in. = 22.1 kip - sec 2 /ft

CE-412: MODULE 9 (fall 2015) 28

Participation factor, Γn
L1 2.25
Γ1 = = = 1.22
M 1 1.84

L2 0.65
Similarly Γ2 = = = 0.36
M 2 1.84

L3 0.25
Γ3 = = = 0.14
M 3 1.84

CE-412: MODULE 9 (fall 2015) 29

Effective weight of structure participating in n th mode, Wn

{φn } [m]{}{
T
ι φn } [m]{}
T
ι Ln
2
Wn = g= g
{φn } [m]{φn }
T
Mn
g = 386.4 kip - sec 2 /in
W1 =
L1
2
g=
( 2.25)
2
* 386.4 = 1063.1 kips
M1 1.84

W2 =
L2
2
g=
(0.65)
2
* 386.4 = 88.7 kips
M2 1.84

W3 =
L3
2
g=
( 0.25)
2
* 386.4 = 13.1 kips
M3 1.84
CE-412: MODULE 9 (fall 2015) 30
 Mass of the structure participating in n th mode , PMn
Wn
=
Participating mass of the structure in nth mode= PM *
n
W1 1063.1 W
PM = = = 0.917 = 91.7%
*
1
W 3 * 386.4
W2 88.7
PM 2 = = = 0.077 = 7.7%
*

W 3 * 386.4
W3 13.1
PM 3 = = = 0.0113 = 1.13%
*

W 3 * 386.4
⇒ ΣPM = 1.00
Most of the code requires that such number of modes shall be
considered so that ΣPM≥ 0.9. In our case, indeed, the consideration
of just the first mode would have been sufficient as PM1≥ 0.9

CE-412: MODULE 9 (fall 2015) 31

 Base shear in n th mode, Vbn

Ln
2
 An   An 
Vbn = g   = Wn   Values of A for each Tn can be
Mn  g   g  determined from Fig. 6.9.5 given
on next slide
Mode 1: Tn1=0.89 sec
 A1   1.8 1
Vb1 = W1  . = 1063.1*  g *  * 0.3 = 645.0 kips
 g   Tn1 g
 A2  2.71g
For Tn2 = 0.32 sec : Vb 2 = W2   * 0.3 = 88.7 * * 0.3 = 72.1 kips
 g  g
 A3  2.71g
For Tn3 = 0.22 sec : Vb 3 = W3   * 0.3 = 13.1* * 0.3 = 10.7 kips
 g  g

CE-412: MODULE 9 (fall 2015) 32

Tn1 = 0.89 sec
Tn 2 = 0.32 sec
Tn 3 = 0.22 sec 0.22 sec

0.32 sec

0.89 sec

CE-412: MODULE 9 (fall 2015) 33

Nodal forces acting on the structure in n th mode, fn

First mode { f n } = [m]{φn }

Vbn
Ln
 f1n  φ1n   f11  φ11 
  Vb1     Vb1  
 f 2 n  = [m]φ2 n  ⇒  f 21  = [m]φ21 
 f  L1 φ   f  L1 φ 
 3n   3n   31   31 
 f11  12 0 0  0.45 129.0 
  645      
 f 21  =  0 12 0  0 .80  =  229.3 
 f  (2.25 *12)  0 0 12 1.00  286.7
 31      

CE-412: MODULE 9 (fall 2015) 34

 Nodal forces acting on the structure in n th mode, fn

Second mode { f n } = [m]{φn }

Vbn
Ln
 f12  φ12  12 0 0  + 1.00 
  Vb 2   72.1    
 f 22  = [m]φ22  =  0 12 0   + 0.45
 f  L2 φ  (0.65 *12)  0 0 12 − 0.80
 32   32    
 f12  + 110.8
   
 f 22  = + 49.8 
 f  − 88.6 
 32   

CE-412: MODULE 9 (fall 2015) 35

 Nodal forces acting on the structure in n th mode, fn

Third mode { f n } = [m]{φn }

Vbn
Ln
 f13  φ13  12 0 0  + 0.80
  Vb 3   10.7    
 f 23  = [m]φ23  =  0 12 0   − 1.00 
 f  L3 φ  (0.25 *12)  0 0 12 + 0.45
 33   33    
 f13  + 34.2
   
 f 23  = − 42.8
 f  + 19.3 
 33   

CE-412: MODULE 9 (fall 2015) 36

Nodal forces acting on the structure in n th mode, fn

129.0 + 110.8 + 34.2

{ f n } = { f1n f 2n 
f 3n } = 229.3 + 49.8 − 42.8 
286.7 − 88.6 + 19.3 

CE-412: MODULE 9 (fall 2015) 37

Nodal forces acting on the structure in n th mode, fn
286.7 k 88.6 k 19.3 k

i1 i2 42.8 k i3
229.3 k 49.8 k

j1 j2 j3
129.0 k 110.8 k 34.2 k

72.0 kips 10.7 kips

645.0 kips

Mode 1 Mode 2 Mode 3

CE-412: MODULE 9 (fall 2015) 38

 Combination of Modal Maxima
The use of response spectra techniques for multi-degree of
freedom structures is complicated by the difficulty of combining
the responses of each mode.
It is extremely unlikely that the maximum response of all
the modes would occur at the same instant of time.
When one mode is reaching its peak response there is no way of
knowing what another mode is doing.
The response spectra only provide the peak values of the
response, the sign of the peak response and the time at which the
peak response occurs is not known.

CE-412: MODULE 9 (fall 2015) 39

 Combination of Modal Maxima

Therefore {u}max ≠ [Φ ]{q}max

and, in general {u}max ≤ [Φ ]{q}max

The combinations are usually made using statistical methods.

CE-412: MODULE 9 (fall 2015) 40

 Combined Response ro

Let rn be the modal response quantity (base shear, nodal

displacement, inter-storey drift, member moment, column stress
etc.) for mode n .The r values have been found for all modes (or
for as many modes that are significant).
Most design codes do not require all modes to be used but many do
require that the number of modes used is sufficient so that the sum
of the Effective Weights of the modes reaches, say, 90% of the
weight of the building. Checking the significance of the
Participation Factors may be useful if computing deflections and
rotations only.

CE-412: MODULE 9 (fall 2015) 41

Absolute sum (ABSSUM) method

The maximum absolute response for any system response quantity

is obtained by assuming that maximum response in each mode
occurs at the same instant of time. Thus the maximum value of the
response quantity is the sum of the maximum absolute value of the
response associated with each mode. Therefore using ABSSUM
method N
ro ≤ ∑ rno
n =1
This upper bound value is too conservative. Therefore, ABSSUM
modal combination rules is not popular is structural design
applications

CE-412: MODULE 9 (fall 2015) 42

 Square-Root-of-the Sum-of-the-Squares (SRSS)
method
The SRSS rule for modal combination, developed in [Link]’s
PhD thesis (1951) is

r ≅ ∑r
o
( )
N

n =1
no
2
1/ 2

The most common combination method and is generally satisfactory

for 2-dimensional analyses is the square root of the sum of the
squares method. The method shall not be confused with the root-
mean-square of statistical analysis as there is no denominator.

CE-412: MODULE 9 (fall 2015) 43

Square-Root-of-the Sum-of-the-Squares (SRSS)
method

This method was very commonly used in design codes until about
1980. Most design codes up to that time only considered the
earthquake acting in one horizontal direction at a time and most
dynamic analyses were limited to 2-dimensional analyses.

CE-412: MODULE 9 (fall 2015) 44

Three Dimensional Structures
In three-dimensional structures, different modes of free-
vibration in different directions may have very similar natural
frequencies.
If one of these modes is strongly excited by the earthquake at a
given instant of time then the other mode, with a very similar
natural frequency, is also likely to be strongly excited at the same
instant of time. These modes are often in orthogonal horizontal
directions but there may be earthquake excitation directions where
both modes are likely to be excited.
In these cases the Root-Sum-Square or SRSS combination
method has been shown to give non-conservative results for the
likely maximum response. In such cases some other methods such
as CQC, DSC are used

CE-412: MODULE 9 (fall 2015) 45

 Modal combination of responses
Consider nodes i & j of the frame for which [Link]
analysis was carried out on previous slides
Ai1 Ai2 Ai3
Using SRSS method
Mi1 Mi2 Mi3
i1 i2 i3 A i = A j = A i12 + A i 2 2 + (− A i3 )2

M i = M i12 + M i 2 2 + (− M i3 )2

( )
j1 j2 j3
Mj1 Mj2
M j = M j12 + M j2 2 + − M j3 2
Mj3

Aj1 Aj2 Aj3

Mode 1 Mode 2 Mode 3

CE-412: MODULE 9 (fall 2015) 46

Caution
It must be stressed that what ever response item r that the
analyst or designer requires it must be first computed in each mode
before the modal combination is carried out.

If the longitudinal stress is required in a column in a frame, then

the longitudinal stress which is derived from the axial force and
bending moment in the column must be obtained for each mode
then the desired combination method is used to get the maximum
likely longitudinal stress.

It is NOT correct to compute the maximum likely axial force and

the maximum likely bending moment for the column then use these
axial forces and bending moments, after carrying out their modal
combinations, to compute the longitudinal stress in the column.

CE-412: MODULE 9 (fall 2015) 47

Home Assignment No. M9
A 3 story R.C. building as shown below is required to be
designed for a design earthquake with PGA=0.25g, and its elastic
design spectrum is given by Fig 6.9.5 (Chopra’s book) multiplied
by 0.25). It is required to carry out the dynamic modal analysis by
using the afore mentioned design spectrum . Take: m3
• Story height = 10ft
k3
•Total stiffness of first 2 stories = 2000 kips/ft. m2

• Total stiffness of top floor = 1500 kips/ft

m1 k2
• Mass of first 2 floors = 5000 slugs
• Mass of top floor = 6000 slugs k1

CE-412: MODULE 9 (fall 2015) 48