CE 809 - Structural Dynamics

Modal Analysis for Forced Vibration Response of MDF Systems

Semester – Fall 2021

Dr. Fawad A. Najam Prof. Dr. Pennung Warnitchai

Department of Structural Engineering Head, Department of Civil and Infrastructure Engineering
NUST Institute of Civil Engineering (NICE) School of Engineering and Technology (SET)
National University of Sciences and Technology (NUST) Asian Institute of Technology (AIT)
H-12 Islamabad, Pakistan Bangkok, Thailand
Cell: 92-334-5192533, Email: [email protected]
 Concept of Generalized Coordinates

𝑢1
𝑢1
Consider a discrete system of N DOF 𝑢2 𝑢2
where its physical displacements are defined by . (an Nx1 vector)
𝒖= . 𝑢3

.
𝑢𝑖 is the physical displacement at the ith floor. 𝑢𝑁

2
The displacements of the NDOF system can also be defined by a different
set of coordinates:

𝑞1
𝑞2
𝒒= . (an Nx1 vector) =
.
𝑞𝑁
× 𝑞1 (𝑡)
where the relation between 𝒒 and 𝒖 is linear
+
𝑢1 𝑡 𝜙11 𝜙12 𝜙1𝑁
. × 𝑞2 (𝑡)
𝜙21 𝜙22 𝜙2𝑁
𝒖= . = . 𝑞1 𝑡 + . 𝑞2 𝑡 + ⋯ . 𝑞𝑁 𝑡
. . . . +
𝑢𝑁 𝑡 𝜙𝑁1 𝜙𝑁2 𝜙𝑁𝑁
× 𝑞3 (𝑡)

𝒖 = 𝝓1 𝑞1 𝑡 + 𝝓2 𝑞2 𝑡 + ⋯ + 𝝓𝑁 𝑞𝑁 𝑡
3
 Concept of Generalized Coordinates

𝒖 = 𝝓1 𝑞1 𝑡 + 𝝓2 𝑞2 𝑡 + ⋯ + 𝝓𝑁 𝑞𝑁 𝑡

𝝓𝑖 = a vector defining vibration shape (dimensionless constant)

𝑞𝑖 𝑡 = a time-varying function
𝒒 = a vector of generalized coordinates

4
 Concept of Generalized Coordinates
𝒖 = 𝝓1 𝑞1 𝑡 + 𝝓2 𝑞2 𝑡 + ⋯ + 𝝓𝑁 𝑞𝑁 𝑡
In matrix form,
𝒖 𝑡 = 𝚽 .𝒒 𝑡
Where

𝚽 = 𝝓1 𝝓2 . . . 𝝓𝑁 Transformation matrix (𝑁 × 𝑁)

For one-to-one linear transformation, 𝝓1 , 𝝓2 , . . . , 𝝓𝑁 must be a set of independent vectors

(then the inverse transformation 𝒒 = 𝚽 −1 𝒖 is possible)

A special case: 𝚽 = 𝐈 (identity matric), then 𝒒 = 𝒖

If 𝚽 is properly chosen, the equation of motion in terms of 𝑞𝑖 𝑡 are uncoupled
(This will be discussed later – modal analysis)
5
Equations of Motion in Normal Coordinates: Uncoupled Equations

6
Equations of Motion in Normal Coordinates: Uncoupled Equations

Free vibration of an undamped N-DOF system: 𝑴𝒖(𝑡) + 𝑲𝒖(𝑡) = 𝟎

𝑁

General Solution: 𝒖(𝑡) = 𝝓𝑖 𝑎𝑖 sin 𝜔𝑖 𝑡 + 𝜃𝑖

𝑖=1

where 𝝓𝑖 = mode shape and 𝑎𝑖 sin 𝜔𝑖 𝑡 + 𝜃𝑖 = time-varying function

Forced vibration of an undamped N-DOF system: 𝑴𝒖(𝑡) + 𝑲𝒖(𝑡) = 𝑸(𝑡) (23)

(The effect of damping will be considered later)

In analogy to the case of free vibration, 𝑁

let’s express the motion in the case of forced vibration as 𝒖(𝑡) = 𝝓𝒊 𝑞𝑖 (𝑡) (24)
𝑖=1

where 𝝓𝒊 is the 𝑖 𝑡ℎ mode shape and 𝑞𝑖 (𝑡) is the time-varying function.

Equations of Motion in Normal Coordinates: Uncoupled Equations
𝑁

𝒖(𝑡) = 𝝓𝒊 𝑞𝑖 (𝑡) (24)

𝑖=1

In matrix form: 𝒖(𝑡) = 𝚽 𝒒(𝑡) (25)

𝑞1 (𝑡)
𝑞2 (𝑡) 𝒒(𝑡) is a vector of “new” generalized coordinates
where 𝒒(𝑡) = ;
⋮ 𝚽 is the modal matrix (used as a transformation matrix).
𝑞𝑁 (𝑡)

Since 𝚽 consists of N-independent vectors (orthogonal vectors), 𝚽 is non-singular and 𝚽 −𝟏 exists.

Hence 𝒒(𝑡) = 𝚽 −1 𝒖(𝑡) (26)

Eq. (25) and Eq. (26) say that 𝒖(𝑡) and 𝒒(𝑡) are one-to-one convertible by a linear transformation.
Equations of Motion in Normal Coordinates: Uncoupled Equations

Introducing Eq. (25) into Eq. (23): 𝑴 𝚽 𝒒 + 𝑲 𝚽 𝒒 = 𝑸(𝑡) (27)

Pre-multiply Eq. (27) by 𝚽 𝑻 ; 𝚽 𝑻 𝑴 𝚽 𝒒(𝑡) + 𝚽 𝑻 𝑲 𝚽 𝒒(𝑡) = 𝚽 𝑻 𝑸(𝑡) (28)

Employing orthogonality condition, 𝜇1 ⋯ 0 𝑞1 𝜇1 𝜔1 2 ⋯ 0 𝑞1 𝑅1 (𝑡)

⋮ ⋱ ⋮ ⋮ + ⋮ ⋱ ⋮ ⋮ = ⋮
(Eqs. (21) and (22))
0 ⋯ 𝜇𝑁 𝑞𝑁 0 ⋯ 𝜇𝑁 𝜔𝑁 2 𝑞𝑁 𝑅𝑁 (𝑡)
𝑅1 (𝑡) (29)
𝑅2 (𝑡)
where 𝑹(𝑡) = = 𝜱𝑻 𝑸(𝑡) (30) Eq. (29) is the equations of motion in 𝒓 coordinates.
⋮
The equations are uncoupled.
𝑅𝑁 (𝑡)
So 𝒒(𝑡) is called “principal coordinates” or “classical modal coordinates” or “normal coordinates”.
Equations of Motion in Normal Coordinates: Uncoupled Equations

𝜇1 ⋯ 0 𝑞1 𝜇1 𝜔1 2 ⋯ 0 𝑞1 𝑅1 (𝑡)
⋮ ⋱ ⋮ ⋮ + ⋮ ⋱ ⋮ ⋮ = ⋮ (29)
0 ⋯ 𝜇𝑁 𝑞𝑁 0 ⋯ 𝜇𝑁 𝜔𝑁 2 𝑞𝑁 𝑅𝑁 (𝑡)

Eq. (29) can be written as 𝜇𝑖 𝑞𝑖 (𝑡) + 𝜇𝑖 𝜔𝑖 2 𝑞𝑖 (𝑡) = 𝑅𝑖 (𝑡), 𝑖 = 1,2, … , 𝑁 (31)

where 𝑅𝑖 (𝑡) = 𝜙1𝑖 𝑄1 (𝑡) + 𝜙2𝑖 𝑄2 (𝑡) + ⋯ + 𝜙𝑁𝑖 𝑄𝑁 (𝑡) (32)

Eq. (31) is identical to the Equation of Motion of a SDOF system. Therefore the solution 𝑞𝑖 (𝑡) is:
𝑡
𝑞𝑖 (0) 1
𝑞𝑖 (𝑡) = sin 𝜔𝑖 𝑡 + 𝑞𝑖 (0) cos 𝜔𝑖 𝑡 + 𝑅𝑖 (𝜏) sin 𝜔𝑖 𝑡 − 𝜏 𝑑𝜏 (33)
𝜔𝑖 𝜇𝑖 𝜔𝑖
0
Homogenous solution,𝑤ℎ𝑒𝑟𝑒 𝑞𝑖 (0) and
Duhamel integration of 𝑅𝑖 𝑡
𝑞𝑖 (0) are initial modal displacement and
velocity of the 𝑖 𝑡ℎ mode.
Equations of Motion in Normal Coordinates: Uncoupled Equations

𝑞1 (0)
𝒒(0) = ⋮ = 𝚽 −1 𝒖(0)
𝑞𝑁 (0)
(34)
𝑞1 (0)
𝒒(0) = ⋮ = 𝚽 −1 𝒖(0)
𝑞𝑁 (0)
Equations of Motion in Normal Coordinates: Uncoupled Equations

𝚽 −𝟏 can be easily obtained by the following relation:

−1
𝜇 [𝜇] = 𝑰

𝜇 −1 𝚽 𝑇 𝑴 𝚽 = 𝑰 = 𝚽 −1 𝚽

∴ 𝚽 −1 = 𝜇 −1
𝚽𝑻𝑴 (35)

1
⋯ 0
𝜇1
𝝁 −1 = ⋮ (36)
where ⋱ ⋮
1
0 ⋯
𝜇𝑁
easier than the performing the matrix inversion of 𝚽 directly
Equations of Motion in Normal Coordinates: Uncoupled Equations

Once 𝑞𝑖 (𝑡) (for 𝑖 = 1, 2, 3, 4, 5, … , 𝑁 ) are evaluated,

𝒖(𝑡) can be formulated and

𝒒𝑡 𝒖(𝑡) = 𝚽 𝒒(𝑡)  structural response

transform
Using modal orthogonality, we can
extract 𝑞𝑛 𝑡 [the 𝑛𝑡ℎ mode’s
contribution in 𝒖 𝑡 ], if the
combined response 𝒖 𝑡 is given.

Orthogonal Filter Method of Modal

Decomposition
Example: Forced Vibration Response of a Uniform Cantilever Beam

A uniform cantilever beam idealized as a three-degree-

of-freedom system.

Lumped mass approximation of a uniform cantilever beam

𝑚1 = 𝑚 6 𝑚 2 = 𝑚3 = 𝑚 3

0.1667 0 0
Equations of motion: 𝑴 = 𝑚 0 0.333 0
𝑢1 (𝑡)
𝑴𝒖(𝑡) + 𝑲𝒖(𝑡) = 𝟎 𝒖 𝑡 = 𝑢2 (𝑡) 0 0 0.333
𝑢3 (𝑡)
𝐸𝐼 43.12 −98.30 73.18
𝑲 = 3 −98.30 269.97 −281.16
𝐿
73.18 −281.16 489.72
 By Eigen analysis (𝑲𝝓𝑖 = 𝜔𝑖 𝑴 𝝓𝑖 ), we obtain:

1.0
𝐸𝐼
𝜔1 = 3.36 𝝓1 = 0.54
𝑚 𝐿3 0.162

1.00
𝐸𝐼
𝜔2 = 18.9 𝝓2 = −0.707
𝑚 𝐿3 −0.731

𝐸𝐼 1.0
𝜔3 = 47.2 𝝓3 = −1.592
𝑚 𝐿3
+2.224

Figure: Normal modes of vibration for the idealized beam

 1.0 1.0 1.0
Modal matrix 𝚽 ∶ 𝚽 = 0.54 −0.707 −1.54
0.162 −0.731 0.333
Free Vibration of Beam:
Now let’s consider a free vibration of beam. Assume, for example, at 𝑡 = 0
𝑢1 (0) 3.0 𝑢1 (0) 0
𝒖 0 = 𝑢2 (0) = 2.0 and 𝒖0 = 𝑢2 (0) = 0
𝑢3 (0) 1.0 𝑢3 (0) 0

Initial modal displacement (at 𝑡 = 0) :

0.611 0.660 0.198 3.0 +3.353

𝒒 0 = 𝚽 −1 𝒖 0 = 0.326 −0.461 −1.477 2.0 = −0.421
0.063 −0.731 0.279 1.0 +0.068

Initial modal velocity (at 𝑡 = 0) : 𝒒(0) = 𝚽 −1 𝒖(0) = 𝟎

The governing equation of motion of 𝑞 (𝑡):
𝑖
𝑞𝑖 (𝑡) + 𝜔𝑖2 𝑞𝑖 (𝑡) = 0 𝑖 = 1, 2, 3

The solution is: 𝑞𝑖 (𝑡) = 𝑞𝑖 (0) cos 𝜔𝑖 𝑡 𝑖 = 1, 2, 3

𝑞1 (𝑡) 3.353 cos 𝜔1 𝑡

That is: 𝒒 𝑡 = 𝑞2 (𝑡) = −0.421 cos 𝜔2 𝑡
𝑞3 (𝑡) 0.068 cos 𝜔3 𝑡

Then, the modal coordinate 𝒒(𝑡) is transformed back to 𝒖 𝑡 :

1.0 1.0 1.0 3.353 cos 𝜔1 𝑡

𝒖(𝑡) = 0.54 −0.707 −1.54 −0.421 cos 𝜔2 𝑡
0.162 −0.731 0.333 0.068 cos 𝜔3 𝑡

The displacement at the beam’s tip: 𝑢1 (𝑡) = 3.353 cos 𝜔1 𝑡 − 0.421 cos 𝜔2 𝑡 + 0.068 cos 𝜔3 𝑡
The displacement at the beam’s tip:

𝑢1 (𝑡) = 3.353 cos 𝜔1 𝑡 − 0.421 cos 𝜔2 𝑡 + 0.068 cos 𝜔3 𝑡

𝑢2 (𝑡) = 0.54(3.353 cos𝜔1 𝑡) + 0.707(0.421 cos𝜔2 𝑡) − 1.54(0.068 cos𝜔3 𝑡)

𝑢3 (𝑡) = 0.162(3.353 cos𝜔1 𝑡) + 0.731(0.421 cos𝜔2 𝑡) + 0.333(0.068 cos𝜔3 𝑡)

Forced Vibration of Beam:

We shall now examine “forced vibration” of the beam.

𝑄1 (𝑡) A step force

𝐹𝑎

𝑄1 (𝑡) 𝐹𝑎
Governing Equations of Motion: 𝑴𝒖(𝑡) + 𝑲𝒖(𝑡) = 𝑸(𝑡) where 𝑸(𝑡) = 0 = 0
0 0

Equations of Motion in modal coordinates: 𝜇𝑖 𝑞𝑖 𝑡 + 𝜇𝑖 𝜔𝑖 2 𝑞𝑖 𝑡 = 𝑅𝑖 𝑡 , 𝑖 = 1,2,3

0.273 0 0
where 𝜇 = 𝚽𝐓 𝑴 𝚽 = 0 0.511 0 𝑚
0 0 2.660
 𝑄1 (𝑡) 𝜙11 𝐹𝑎
Modal force: 𝑹 𝑡 = 𝚽 𝐓 𝑸(𝑡) = 𝚽 𝐓 0 = 𝜙12 𝐹𝑎
0 𝜙13 𝐹𝑎

The 𝑖 𝑡ℎ equation of motion: 𝜇𝑖 𝑞𝑖 𝑡 + 𝜇𝑖 𝜔𝑖 2 𝑞𝑖 𝑡 = 𝑅𝑖 𝑡 = 𝜙1𝑖 𝐹𝑎 𝑖 = 1, 2, 3

Let’s assume that 𝒖(0) = 𝟎, 𝒖(0) = 𝟎 hence 𝒒(0) = 𝟎, 𝒒(0) = 𝟎

𝜙1𝑖 𝐹𝑎
𝑞𝑖 𝑡 = 𝑖 = 1, 2, 3
2 1 − cos(𝜔𝑖 𝑡)
The solution is
𝜇 𝑖 𝜔𝑖

(see “the response to step force of SDF system”)

 𝜙1𝑖 𝐹𝑎
The solution is 𝑞𝑖 𝑡 = 2 1 − cos(𝜔𝑖 𝑡) 𝑖 = 1, 2, 3
𝜇𝑖 𝜔𝑖

Then, the displacement at the tip of the beam is:

𝜙11 𝐹𝑎 𝜙12 𝐹𝑎 𝜙13 𝐹𝑎

𝑢1 (𝑡) = 𝜙11 2 1 − cos(𝜔1 𝑡) + 𝜙12 2 1 − cos(𝜔2 𝑡) + 𝜙13 2 1 − cos(𝜔3 𝑡)
𝜇1 𝜔1 𝜇2 𝜔2 𝜇3 𝜔3
or,

𝐹𝑎 𝐿3
𝑢1 (𝑡) = 0.328 1 − cos(𝜔1 𝑡) + 0.005 1 − cos(𝜔2 𝑡) + 0.0002( 1 − cos(𝜔3 𝑡 )
𝐸𝐼

Contribution of the ≫ Contribution of the ≫ Contribution of the

1𝑠𝑡 mode 2𝑛𝑑 mode 3𝑟𝑑 mode
Some Additional Notes on the Modal Analysis Method:

The method is

a) very efficient for the analysis of dynamic responses of linear structures,

b) able to be greatly simplified by considering only a few important modes (lower modes):
𝑁 𝑘

𝒖 𝑡 =𝚽𝒒 𝑡 = 𝝓𝒊 𝑞𝑖 𝑡 ≈ 𝝓𝒊 𝑞𝑖 𝑡 , (𝑘 ≪ 𝑁)
𝑖=1 𝑖=1

If the next higher mode ( 𝑘 + 1 ) make no significant

contribution to the responses (𝒖(𝑡)), it may be assumed
𝑞1 (𝑡)
that sufficient modes have already been included. 𝑞2 (𝑡)
⋮
𝒖 𝑡 ≈𝚽𝒒 𝑡 , where 𝒒 𝑡 = 𝑞𝑘 (𝑡)
0
negligibly small.
⋮
0
c) not applicable for non-proportional damping systems.

𝚽 𝑻 𝑪 𝚽  non-diagonal matrix
In that case, complex modal analysis is needed (a more complicated procedure).

d) not applicable for non-linear structures

For example, Response of a building to a severe earthquake shaking is nonlinear.

Damping Matrix:
Modal Analysis Summary:
 Discrete Systems Continuous Systems
Classical
Modal
Analysis
Thank you