Lecture 7 - MDOF
Lecture 7 - MDOF
Lecture 7: MDOF
where;
𝐴 is a square matrix 𝑁 × 𝑁
{𝓍} is the eigenvector
𝜆 is the eigenvalue
Eigenvalues and Eigenvectors (cont…)
1. Determinant Calculation Method
In vibration analysis, the Eigenvalues correspond to natural frequencies of
vibration, and the Eigenvectors to the relative vibration amplitudes.
From Equation (1), it follows that:
[A] {x} − λ{x} = 0
or [A] {x}-λ[I]{x} = 0
Expanding;
2𝑥2 = 2.73𝑥1
𝑥1
∴ = 0.73
𝑥2
Normalize the vector to its first element 𝑥1 = 1
1
1 𝒙 =
1.365
A = LR (1)
where;
L = Unit Lower Triangular Matrix
R = Upper Triangular Matrix
Step 2:
A new Matrix 𝑨1 which is similar to [A] (i.e. 𝑨1 has the same eigenvalues as [A]), is obtained by multiplying LR in reverse order.
Hence 𝑨𝟏 = 𝑹𝑳 (2)
Eigenvalues and Eigenvectors (cont…)
This process is repeated to obtain further matrices, which are similar to [A] as
follows
𝑨𝒊 =𝑳𝒊 𝑹𝒊 ; 𝑨𝒊+𝟏 =𝑹𝒊 𝑳𝒊
2 −1 0
𝐴 = −1 2 −1
0 −1 2
Eigenvalues and Eigenvectors (cont…)
• Solution using L-R Algorithm
Step 1: decompose [A] = [L][R]
2 −1 0 1 0 0 2 −1 0
𝐴 = −1 2 −1 = −0.5 1 0 ∙ 0 1.5 −1
0 −1 2 0 −0.667 1 0 0 1.333
2.5 −1 0 1 0 0 2.5 0 0
𝐴1 = −0.75 2.167 −1 = −0.3 1 0 ∙ 0 1.876 −1
0 −0.889 1.333 0 0.476 1 0 0 0.857
Hence;
2.8 −1 0 1 0 0 2.8 0 0
𝐴2 = −0.56 2.343 −1 = −0.2 1 0 ∙ 0 2.143 −1
0 −0.408 0.857 0 0.19 1 0 0 0.667
Eigenvalues and Eigenvectors (cont…)
Step 7: Process is repeated further to obtain:
𝑛 𝑛
ii. 𝜆
𝑖=1 𝑖 = 𝑖=1 𝑎𝑖𝑖 Trace of the Matrix
𝑛
𝑖=1 𝜆𝑖 = 0.587 + 2.054 + 3.35 = 5.99
𝑛
𝑖=1 𝑎𝑖𝑖 = 2 + 2 + 2 = 6
Eigenvalues and Eigenvectors (cont…)
b). Power Method: Determining the Dominant Eigenvalue and eigenvector
The dominant eigenvalue is the highest valued root of the characteristic equation.
Let us illustrate this through an example.
1 2
𝑨= ; 𝐴 𝑥 =𝜆 𝑥 .
1 3
1
Assume an arbitrary value for 𝑥 , say 1𝒙 =
1 Not Greater than 1.0
1 2 1 3
Pre-multiply 1𝒙 by [A]: i.e. = = 1𝒙∗ say.
1 3 1 4
∗ 0.75
Normalize 1𝒙 to the largest element (4) 2𝒙 =
1
Eigenvalues and Eigenvectors (cont…)
Note:
• If this vector is equal to the previous one, it is the dominant
eigenvector, and the normalization factor (i.e. 4) is the dominant
eigenvalue.
• If 𝑛+𝑖𝒙 ≠ 𝑛𝒙 iteration process continues.
• Next iteration:
1 2 0.75 2.75
= = 2𝒙∗
1 3 1 3.75
0.735
∴ 𝟑𝒙 =
1 Normalise wrt 3.75
Eigenvalues and Eigenvectors (cont…)
Dominant
Eigenvalue
𝑨 𝟏𝒙 𝟏𝒙
∗
𝟐𝒙 𝟐𝒙
∗
𝟑𝒙 𝟑𝒙
∗
𝟒𝒙 𝟒𝒙
∗
𝟓𝒙
𝟓𝒙 = 𝟒𝒙
Eigenvalues and Eigenvectors (cont…)
Since 𝟓𝒙 = 𝟒𝒙 ∴ 𝜆 = 3.733 (dominant eigenvalue)
0.733
and 𝒙 = Corresponding eigenvector
1
Note:
With this iteration method, any arithmetic errors occurring
in the intermediate iteration will not affect the result.
Eigenvalues and Eigenvectors (cont…)
c). Power Method: Determination of the Lowest Eigenvalue
−𝟏
Thus, performing the same iterative process on 𝑨 , results in the
lowest eigenvalue being found.
1 2
Let us consider the same matrix 𝐴 = .
1 3
3 −2
𝐴𝐷𝐽[𝐴] −1 1
[𝐴]−1 = = Note:
𝐴 1 Adj[A] = Transpose of cofactor Matrix [A]
3 −2
∴ [𝐴]−1 =
−1 1
Eigenvalues and Eigenvectors (cont…)
Dominant
Eigenvalue
1. Perform similar iterations as before
𝐴 ∗ ∗ ∗ ∗
1𝑥 1𝒙 2𝑥 2𝒙 3𝑥 3𝒙 4𝑥 4𝒙 𝟓𝑥
𝟓𝒙 = 𝟒𝒙
Eigenvalues and Eigenvectors (cont…)
𝟓𝒙 = 𝟒𝒙 ; 𝝀 =3.726
1
and the associated eigenvector is =
−0.365
Matrix Decomposition Technique
If 𝑨𝑿 = 𝜆𝑩𝑿 (1)
where 𝑨 = 𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝑀𝑎𝑡𝑟𝑖𝑥
𝑩 = + 𝑣𝑒 𝑑𝑒𝑓𝑖𝑛𝑖𝑡𝑒 𝑀𝑎𝑡𝑟𝑖𝑥
𝜆𝑟 < 𝓁
𝜆𝑟+1 > 𝓁
Matrix Decomposition (cont…)
Example:
2 −1 𝑥1 1 0 𝑥1
𝑥2 = 𝜆 0 ; 𝜆 = 𝜔2
−1 1 1 𝑥2
A X B X
r=0
From 𝜆𝑟 < 𝓁
𝜆𝑟+1 > 𝓁 ; ⇒ 𝜆𝑜 < 𝓁 (Not Possible)
∴ 𝜆1 > 0
Take 𝓁 = 0.6
Matrix Decomposition (cont…)
𝐴 − 𝓁𝐵 = 𝐴 − 0.6𝐵
r=1
1.4 −1
=
−1 0.4
1 0 1.4 0 1 −0.71
=
−0.71 1 0 −0.31 0 1
∴ 𝜆1 < 0.6
𝜆2 > 0.6
More iterations will give upper & lower limits until the solution converges.
Properties of Vibrating Systems
1. Flexibility Matrix Approach
𝑥 = 𝜙 𝐹 (1)
Part (c)
𝐹2 1
= = (𝑘𝑡𝑜𝑡 )2 (1)
𝑥2 𝑥2
For 𝑥1 : The deflection at 1 due to a unit force applied at 2 is the same as would be produced by
the same force applied at 1.
1
i.e. 𝜙12 =
𝑘1
Since 𝑥3 = 𝑥2
1 1
𝜙32 = +
𝑘1 𝑘2
Flexibility Matrix Approach (cont…)
Part (d)
𝐹3 1
= = (𝑘𝑡𝑜𝑡 )3 ;
𝑥3 𝑥3
11 1 1
Where = + +
𝑘𝑡𝑜𝑡 3 𝑘1 𝑘2 𝑘3
1 1 1 1
∴ 𝑥3 = = + +
𝑘𝑡𝑜𝑡 3 𝑘1 𝑘2 𝑘3
1 1 1
Thus; 𝜙33 = + +
𝑘1 𝑘2 𝑘3
As in Part (c) above;
1 1 1
𝜙23 = + ; 𝜙13 =
𝑘1 𝑘2 𝑘1
Flexibility Matrix Approach (cont…)
1 1 1
𝑘1 𝑘1 𝑘1
1 1 1 1 1
∴ 𝝓= + +
𝑘1 𝑘2 𝑘2 𝑘1 𝑘2
1 1 1 1 1 1
+ + +
𝑘1 𝑘1 𝑘2 𝑘1 𝑘2 𝑘3
𝐾𝑖𝑗 −Stiffness coefficient i.e. the force which must be applied at point “𝒾” in order
to produce a unit displacement at point “j” and zero displacements at all other
points in the system.
Stiffness Matrix Approach (Cont…)
Stiffness Matrix Approach (Cont…)
Part (b)
Force acting on 𝑚1
𝐹1 = 𝑥1 𝑘1 + 𝑥1 𝑘2 = 1 𝑘1 + 𝑘2 .
∴ 𝑘11 = 𝑘1 + 𝑘2 etc.
𝑘1 + 𝑘2 −𝑘2 0
∴ 𝐾 = −𝑘2 𝑘2 + 𝑘3 −𝑘3
0 −𝑘3 𝑘3
Stiffness Matrix Approach (Cont…)
Note:
(i) The Stiffness Matrix is symmetric
(ii) 𝑘𝑖𝑗 = 𝑘𝑗𝑖 – Reciprocal theorem (Maxwell’s theorem).
(iii) Relationship between stiffness and flexibility matrices
𝑥 = 𝜙 𝐹
⇒ [𝜙]−1 𝑥 = 𝐹
But 𝐾 𝑥 = 𝐹
∴ 𝐾 = [𝜙]−1
2 DoF
• Revise single dof
• Introduce 2 dof based on single dof
Systems with several DOF
𝑴𝒙 + 𝑲𝒙 = 0 (1)
Equation (1) represents a set of n coupled equations
where; 𝑖𝑿 = 𝒊𝑿𝟐
⋮
𝒊𝑿𝒏
Note:
1). Each 𝜔 corresponds to a “natural mode frequency”, and
2). Each amplitude vector X corresponds to a normal mode shape.
Systems with several DOF (cont…)
2
For a particular 𝜔𝑖 and 𝒊𝑿; substitute into Equation (4) to get
{K−𝜔𝑖2 M} 𝒊𝑿 = 0
⟹ 𝑲𝒊 𝑿 = 𝜔𝑖2 𝑴𝒊 𝑿 (6)
Similarly, for a particular 𝜔𝑗2 and 𝒋𝑿 we get
𝑲𝒋 𝑿 = 𝜔𝑗2 𝑴𝒋 𝑿 (7)
Systems with several DOF (cont…)
From Equation (6);
𝑻 𝑻
𝑲 𝒊𝑿 = 𝜔𝑖2 𝑴 𝒊𝑿
⇒ 𝒊𝑿𝑻 𝑲 = 𝜔𝑖2 𝒊𝑿𝑻 𝑴𝑻
Post – multiply by 𝒋𝑿
𝒊𝑿
𝑻𝑲
𝒋𝑿 = 𝜔𝑖2 𝒊𝑿𝑻 𝑴𝑻 𝒋𝑿 (8)
𝑻 2 2 2
𝒊 𝑿 𝑴 𝒊 𝑿 = 𝑋
𝑖 1 𝑀1 + 𝑋
𝑖 2 𝑀2 + ⋯ 𝑖 𝑋𝑛 𝑀𝑛
= 𝐿2𝑖 𝑆𝑎𝑦 {Note: all quantities are positive}
𝒊𝑿
Let 𝒊𝒁 = be a normalized vector
𝐿𝑖
Then;
𝒊𝑿𝑴 𝒊𝑿 𝐿2𝑖
𝒊 𝒁𝑻 𝑴 𝑖𝒁 = = =1
𝐿2𝑖 𝐿2𝑖
𝑻
𝒊𝑿 𝑴 𝒋𝑿 0
and 𝒊 𝒁𝑻 𝑴 𝒋𝒁 = = =0
𝐿𝑖 𝐿𝑗 𝐿𝑖 𝐿𝑗
Systems with several DOF (cont…)
Let us now establish a normalized eigenvector matrix, called a “modal matrix”, Z
↑ ↑ ↑ ↑
𝒁 = 1𝒁 2𝒁 … 𝑟 𝒁 … 𝑛 𝒁
↓ ↓ ↓ ↓
𝑟𝑍1
where 𝑟𝒁 = 𝑟𝑍2
⋮
𝑟𝑍𝑛
𝜔12 ⋯ 0
Ω= ⋮ 𝜔𝑖2 ⋮
0 ⋯ 𝜔𝑛2
Systems with several DOF (cont…)
From Equation (13), pre-multiply by 𝒁𝑻 to get
𝒁𝑻 𝑲 𝒁 = 𝒁𝑻 M 𝒁 𝛀
But 𝒁𝑻 M 𝒁 = 𝑰
∴ 𝒁𝑻 𝑲 𝒁 = 𝐈 𝛀 = 𝛀 (14)
Systems with several DOF (cont…)
Decoupling the equations:
or 𝑞𝑟 + 𝜔𝑟2 𝑞𝑟 = 0 ; 1≤ 𝑟 ≤ 𝑛
Note:
From Equation (16),
𝒙= 𝒁𝒒 ⇒ 𝑴𝔁 = 𝑴𝒁𝒒
Pre-multiply by 𝒁𝑻
𝒁𝑻 M 𝒙 = 𝒁𝑻 M 𝒁𝒒 = 𝑰𝒒 = 𝒒
Systems with several DOF (cont…)
𝒒 = 𝒁𝑻 𝑴𝒙
𝑑𝒒 𝑻
𝑑𝒙
=𝒁 𝑴
𝑑𝑡 𝑑𝑡
Example 1: Systems with several DOF
For the system shown, determine the natural modes if:
• 𝑚1 = 1𝑡𝑜𝑛𝑛𝑒
• 𝑚2 = 1/2 𝑡𝑜𝑛𝑛𝑒
Example 1 (cont…)
Procedure:
1). Calculate the stiffness matrix (K)
2). Calculate the mass matrix (M)
3). Solve KX= 𝜔2 𝑴 𝑿
Example 1 (cont…)
2 −1 𝑋1 1 0 𝑋1
∴ 104 =𝜔 2
−1 1 𝑋2 0 0.5 𝑋2
2
Let 𝜆2 = 𝜔 104
2 − 𝜆 𝑋1 − 𝑋2 = 0 (1)
−𝑋1 + 1 − 0.5𝜆 𝑋2 = 0 (2)
2 − 𝜆 1 − 0.5𝜆 =1
2−𝜆 2 = 2
2−𝜆 = ± 2
Determination of Eigenvalues:
From Equation (1a),
𝑋1 1
= 2−𝜆
𝑋2
For 𝜆1 =2− 2
1𝑋1 1
𝟏𝑿 = =
1𝑋2
2
This is the 1st eigenvector or fundamental mode-associated with the lowest frequency
Example 1 (cont…)
For 𝜆2 = 2+ 2
2𝑋1 1
𝟐𝑿 = =
2𝑋2
− 2
This is the 2nd eigenvector (harmonic or higher harmonic)
Example 1 (cont…)
• Alternative Calculation
Example 1 (cont…)
𝑚1 𝑥1 + 𝑘1 𝓍1 − 𝑘2 𝓍2 − 𝓍1 = 0 (1a)
𝑚2 𝑥2 + 𝑘2 𝓍2 − 𝓍1 = 0 (1b)
∴ 𝑚1 𝑚2 𝜔4 − 𝑘1 + 𝑘2 𝑚2 + 𝑚1 𝑘2 𝜔2 + 𝑘1 𝑘2 = 0.
0.5𝜔4 − 2 × 0.5 + 1 104 𝜔2 + 108 = 0
0.5𝜔4 −2× 104 𝜔2 + 108 = 0
2×104 ± 2×104 2 −4×0.5×108
𝜔2 =
2×0.5
∴ 𝜔12 = 5857.86
𝜔22 =34142.14
⇒𝑓1 = 12.18 𝐻𝑧
𝑓2 = 29.4 𝐻𝑧 As before!
Example 1 (cont…)
∴ 𝑇1 = 0.082 𝑠𝑒𝑐
𝑇2 = 0.034 𝑠𝑒𝑐
Note:
𝑻 𝑴 X ≠ 0 = + 𝑣𝑒
𝟏 𝑿 𝟏
𝑿𝑻 𝑴 X ≠ 0 = + 𝑣𝑒
𝟐 𝟐
Forced vibration with many DOF
𝑴𝒙 + 𝑲𝒙 = 𝑷 (1)
𝑇
where 𝑷 = 𝑃1 𝑃2 … 𝑃𝑛 is the Dynamic Force Vector