Earthquake Engineering

Dynamic Eigenvalue problem

Stiffness Eigenvalue
Mass Matrix [M] problem of free
Matrix [K]
vibration

Eigenvalues (frequency) and eigenvectors (Mode shapes) are the

dynamic characteristics of the structures that are independent of
applied loads. It is like the stiffness of the structure which is a static
characteristic and independent of applied load (for linear system)
 Stiffness Matrix [K] for Shear Frames (approximate solution)

Assuming shear frame, infinitely rigid External force

beams (zero slopes at columns-beam
connection

Story stiffness = ∑ 𝑘 /

12 𝐸𝐼
𝑘 = 𝑟𝑖𝑔𝑖𝑑 − 𝑟𝑖𝑔𝑖𝑑
ℎ

𝑘 = 𝑟𝑖𝑔𝑖𝑑 −hinged Columns Floors

Floors
internal internal
forces forces
 Stiffness Matrix [K] for Shear Frames (approximate solution)
Coefficient Kij is defined as the force acts along ith DOF to maintain unit displacement at jth DOF while zero displacement for
the rest of DOFs.

Floor resisting force

Coefficient Ki1 is defined as the force acts along

ith DOF to maintain unit displacement at 1st DOF
while zero displacement for the rest of DOFs

Floor external load Coefficient Ki2is defined as the force acts along ith
DOF to maintain unit displacement at 2nd DOF
while zero displacement for the rest of DOFs

Coefficient Ki3is defined as the force acts along ith

DOF to maintain unit displacement at 3rd DOF
while zero displacement for the rest of DOFs
4
Mass Matrix [M]

mi=Wi/g

5
 Eigenvalue Problem
Definition: In the dynamic analysis, the eigenvalue problem is the solution of the dynamic
equilibrium equations under the free vibration state (no external forces; Homogeneous
equations).
{y} is the
Floor max. amplitude displacement
vector of n DOF
Structure frequency
The solution of the typical single degree of freedom (yi) is expressed as follows: a4
Floor displacement
a3
𝒚𝒊 = 𝒂𝒊 𝒔𝒊𝒏(𝝎𝒕) 𝟐
𝒚̈ 𝒊 = −𝝎 𝒂𝒊 𝒔𝒊𝒏(𝝎𝒕
Where ai is the max. amplitude of the yi degree of freedom, ω, is the natural frequency,. a2

Accordingly the above equilibrium equations can be rewritten as following:

a1
𝑲 𝒂 − 𝝎 𝑴 {𝒂} = {0}
Structure vibration mode
 Eigenvalue Problem

variable
For a system of “n” DOFs (n independent unknown displacements components, for
example floors x, and y displacements and the rotation which describe the deformed
configuration of the building), to ensure non trivial solution for {a}, the determinate of
the overall matrix ([K]-ω2[M]) has to be zero, resulting in an equation of order n in ω2 .
The task is to get the n roots (of ω2)….this is what is called eigenvalue problem
The solution of the eigenvalue problems (finding the roots of the above free vibration
equilibrium equations) mainly depends on trial approach to search for all roots.
a4
a3

a2

a1

7
 Eigenvalue Problem

Since there are n eigenvalues (ωi), n corresponding eigenvectors a4=Φ43

{a} can be determined, usually denoted by {φ}. This can be
achieved by substituting with each ωi in the above equation, and a3=Φ33
assume arbitrary value (for example a unit value) for any
unknown a to get the rest of as giving {φi}. The matrix a2=Φ23
containing all eigenvectors in columns called modal matrix [Φ].
Accordingly, since assuming any arbitrary value for one a1=Φ13
unknown to get the rest of unknowns as relative to the assumed
value, there will be infinite number of solutions for {a} as we
have infinite number of assumptions. So, the point is to get the 3rd Mode, ω3
relative values of the unknowns with respect to the firstly
assumed unknown. Therefore, only n-1 independent unkown are
present.
8
 Orthogonality Properties of the Mode Shapes

One of the most important properties of the mode shapes matrix (modal matrix)
[] (which contains n eigenvectors in its columns) is the following

  T
[ M ][  ]  [ M * ] Convert the original stiffness matrix [K] into
Diagonal modal stiffness matrix [K*]

 T [ K ][  ]  [ K * ]
Where the mass matrix [M], and the stiffness matrix [K] are transferred into diagonal matrices , namely
[M*], and [K*], which are called modal mass and modal stiffness matrices, respectively, and are
expressed as follows

9
Orthogonality Properties of the Mode Shapes
M1* 0 0 0 0 M 1*  m1112  m 2 212  .......  m n n21
 *  𝑀∗ = {𝜑 } [M]{𝜑 }
 0 M 2 0 0 0 which is the 1st modal mass
[M * ]   0 0 M 3* 0 0
M 2*  m1122  m 2 222  .......  m n n22
 
  𝑀∗ = {𝜑 } [M]{𝜑 }
 0 0 0 0 M n*  which is the 2nd modal mass.


Generally, M*j  m112j  m222j  ....... mnnj2  miij2 ∗

the element mii (ith row, and ith column of the mass matrix [M*]) is equal to {φiT}[M]{φi}, where the
order of {φiT} is 1*n, and the order of {φi} is n*1, and [M] is n*n, accordingly {φiT}[M]{φi} is a scalar
number.
0 𝒊
𝑻
𝒋

10
 Orthogonality Properties of the Mode Shapes

 K1  1 M 1
* 2 *
Since  K1* 0 0 0 0
 
( 𝑲 − 𝝎 𝑴 ){𝒂} = {0} Multiplying by {ϕi}T then {ϕj}
0 K 2* 0 0 0  K *
2   2
2
M 2
*

{φi}T[K]{φj}=ωi2{φ}iT [M]{φ}j = 0, if i ≠ j [K * ]   0 0 K 3* 0 0 
 Kj j M j
* 2 *
otherwise =ωi2∑𝒏𝒌 𝟏 𝒎𝒌 𝝋𝟐𝒌𝒊 = ωi2Mi* 
 
0 0 0 0 K n* 


Element kii (ith row, and ith column of the stiffness matrix [K*]) is equal to φiT[K]φi, where the order of φiT is 1*n,
and the order of φi is n*1, and [K] is n*n, accordingly φiT[K]φi is a scalar number.
Therefore, the modal matrix [] can transfer the corresponding mass and stiffness matrices into diagonal matrices.
Moreover it is convenient to normalize eigenvectors by mass matrix so that the modal mass matrix became unity
matrix ( i.e. M*j=1) as shown next example.
11
 Proof of the Orthogonality of the mode shapes
After solving eigenvalue to get the values of ω

For convenience, the case of two degrees of freedom (a1, and a2) will be adopted for explaining the
orthogonality of the modes shapes. The dynamic equations for two-story shear building is expressed as
following:

(k1+k2)a1-k2a2=ω2m1 a1
-k2a1+k2a2=ω2m2a2

Analogous to the static equilibrium equations, the right hand side can be considered to be
external loads (because K*displacement=external load, and they have units of force)

12
 Proof of the Orthogonality of the mode shapes
Since for the free vibration homogenous equations two ω2 (for the 2DOF) can be
found to satisfy the zero determinate, namely ω12, and ω22, resulting in two set of
eigenvectors {φ}1 and {φ}2. where {φ}1 = (φ11 φ21) and {φ}2 = (φ12 φ22).
Accordingly, two cases of loading can be considered as following:
Case1: (k1+k2)φ11-k2 φ 21= ω12 m1 φ 11
-k2 φ 11+k2 φ 21= ω12 m2 φ 21
Case 2: (k1+k2) φ 12-k2 φ 22= ω22 m1 φ 12
-k2 φ 12+k2 φ 22= ω22 m2 φ 22
Φ11 and φ 21 are equivalent to a1 and a2 in the previous equation using ω1, similarly,
φ12 and φ 22 are equivalent to a1 and a2 in the previous equation using ω2.

Adopting Maxwell-Betti theorem that is “work done by 1st case of loading

by deflection of the 2nd case of loading is equal to work done by 2nd case of
loading due to deflection due to the 1st case of loading”.
P1*δ12=P2*δ21 13
 Proof of the Orthogonality of the mode shapes

m1 ω12 φ11 φ12 + m2 ω12 φ21 φ22 = m1 ω22 φ12 φ11 + m2 ω22 φ22 φ21

≠ 0 as they are independent roots

(ω12- ω 22)( m1 φ11 φ12 + m2 φ21 φ22 )=0

𝟏𝟐
𝟏𝟏 𝟐𝟏
𝟐𝟐

{φ}iT [M] {φ}j,= 0 where i≠j

14
 Proof of the Orthogonality of the mode shapes

Regarding the stiffness matrix [K], the equilibrium (eigenvalue) equation can be rewritten in terms of a jth
normalized eigenvector {φ}j, as following:
[K]{φ}j=ωj2[M]{φ}j
Multiplying both side by {φ}Ti, results in the following:

{φ}iT[K]{φ}j=ωj2 {φ}iT [M]{φ}j = 0 for i≠j

M j *ω j 2 for i=j
Applying for all eigenvectors,
 K1* 0 0 0 0 
  K 1
*
  1
2
M *
1

 T [ K ][  ]  [ K * ] 0
[K * ]   0
K 2*
0
0
K 3*
0
0
0
0

 K *
2   2
2
M *
2
 
 Kj j M j
* 2 *

0 0 0 0 K n* 
 15
 Proof of the Orthogonality of the mode shapes
Normalizing the eigenvectors
Since the values of the eigenvectors are just relative values which satisfy the homogeneous equations, the following
normalized (for the eigenvector) can be incorporated as following:

𝝋𝒊𝒋
∅𝒊𝒋 = {Φ} is the normalized vector of {φ}
{ } 𝑴 {𝝋}𝒋

For diagonal mass matrix [M], the above normalized particular eigenvector ∅𝒊𝒋 can be expressed as following:
𝝋𝒊𝒋
∅𝒊 =
∑𝒏 𝟐
𝒌 𝟏 𝒎𝒌 𝝋𝒌𝒋
Accordingly,
{φ}iT[M]{φ}j= 0 for i≠j
1 for i=j
Therefore, applying the above normalization process for all eigenvectors, the following overall modal
orthogonality can be expressed as follows:

[φ]iT[M][φ]j= [unity matrix]…only diagonals is 1, otherwise=0

[φ]iT[K][φ]j= ωi2 for i=j, and 0 for i≠j 16
 Eigenvalue Problem
Example

The stiffness and mass matrices of the shown two story shear frame can be found to be as
follows:
161554 −124626 2 0
𝑲 = , 𝒂𝒏𝒅 𝑴 =
−124626 124626 0 1
The value 161554 ton/m=2*(18463+62313), where 18463=12EI/(63), and 62313=12EI/43,
and 124626=2*12EI/43. Mass of each story was obtained by dividing weight by 10, as an
approximation.
Equilibrium equation of free vibration is as follows:

161554 −124626 y1 2 0 𝑦1̈ 0

+ =
−124626 124626 y2 0 1 y2̈ 0
Or, in a matrix form

𝑲 𝒚 + [𝑴](𝑦̈ ) = (0)
17
Example (cont.)

Where y1, and y2 are displacement of first and second story, respectively.
Substituting (𝑦̈ ) = -ω2 (y)

161554 − 2𝝎 −124626 y1 0
=
−124626 124626 − 𝝎 y2 0
Or in a matrix form

𝑲 − 𝝎𝟐 𝑴 𝑦 = (𝟎)

( 𝟐 𝟐 - 2=0

2ω4-410806ω2+4602188928=0
Solving for ω2, results in ω12, and ω22 are 11891, and 193511, accordingly
ω1, and ω2 are equal to 109 rad/sec, and 439.9 rad/sec.
18
Example (cont.)

Mode shapes (eigenvectors),

Substituting for ω12=11891, in the above homogeneous equations, while assuming y1=1, results
in y2=1.105. also, Substituting for ω22=193511, in the above homogeneous equations, while
assuming y1=1, results in y2=-1.809.
Accordingly, the modal matrix [Φ] is expressed as following:

19
 Example (cont.)
Normalized eigenvectors
Since there are infinite number of solutions (when assuming unit values) it is convenient to
choose specific values by the following normalization process.

𝒚𝒊𝒋
𝝋𝒊𝒋 =
∑𝟐𝒌 𝟐
𝟏 𝒎𝒌 𝒚𝒌𝒋

For the first mode (j=1); ∑𝟐𝒌 𝟐

𝟏 𝒎𝒌 𝒚𝒌𝟏 = 𝟐 ∗ 𝟏𝟐 + 𝟏 ∗ 𝟏. 𝟏𝟎𝟓𝟐 = 𝟏. 𝟕𝟗𝟒𝟕

Therefore, φ11=1/1.7947=0.55718, φ21=1.105/1.7947=0.61569

For the second mode (j=2); ∑𝟐𝒌 𝟐

𝟏 𝒎𝒌 𝒚𝒌𝟐 = 𝟐 ∗ 𝟏𝟐 + 𝟏 ∗ 𝟏. 𝟖𝟎𝟗𝟐 = 𝟐. 𝟐𝟗𝟔𝟏𝟗

Therefore, φ12=1/2.29619=0.4355, φ22=-1.809/2.29619=-0.7878

20
 Example (cont.)

Accordingly, the normalized modal matrix [Φ] is expressed as follows:

𝟎. 𝟓𝟓𝟕𝟏𝟖 𝟎. 𝟒𝟑𝟓𝟓
𝜱 =
𝟎. 𝟔𝟏𝟓𝟔𝟗 −𝟎. 𝟕𝟖𝟕𝟖
The normalization process results in the following:

𝟎. 𝟓𝟓𝟕𝟏𝟖 𝟎. 𝟔𝟏𝟓𝟔𝟗 𝟐 𝟎 𝟎. 𝟓𝟓𝟕𝟏𝟖 𝟎. 𝟒𝟑𝟓𝟓 𝟏 𝟎

[𝜱]𝑻 𝑴 𝜱 = =
𝟎. 𝟒𝟑𝟓𝟓 −𝟎. 𝟕𝟖𝟕𝟖 𝟎 𝟏 𝟎. 𝟔𝟏𝟓𝟔𝟗 −𝟎. 𝟕𝟖𝟕𝟖 𝟎 𝟏
Also,

𝟎. 𝟓𝟓𝟕𝟏𝟖 𝟎. 𝟔𝟏𝟓𝟔𝟗 𝟏𝟔𝟏𝟓𝟓𝟒 −𝟏𝟐𝟒𝟔𝟐𝟔 𝟎. 𝟓𝟓𝟕𝟏𝟖 𝟎. 𝟒𝟑𝟓𝟓

[𝜱]𝑻 𝑲 𝜱 =
𝟎. 𝟒𝟑𝟓𝟓 −𝟎. 𝟕𝟖𝟕𝟖 −𝟏𝟐𝟒𝟔𝟐𝟔 𝟏𝟐𝟒𝟔𝟐𝟔 𝟎. 𝟔𝟏𝟓𝟔𝟗 −𝟎. 𝟕𝟖𝟕𝟖

𝟏𝟏𝟖𝟗𝟎 𝟎 𝝎𝟐𝟏 𝟎
= =
𝟎 𝟏𝟗𝟑𝟓𝟎𝟐 𝟎 𝝎𝟐𝟐

21