Free Vibration of MDOFSystems

CEE 4491A Structural Dynamics II Free Vibration of MDOF Systems
Free Vibration of Multi-Degree of Freedom Systems
Simple Multi-degree of Freedom Systems:
𝑥 𝑥
𝑘 𝑘
𝑚 𝑚
Consider two masses 𝑚 and 𝑚 , with mass 𝑚 connected to a fixed spring with stiffness
𝑘 , with a second spring with stiffness 𝑘 connecting the two masses. The position of
mass 𝑚 is given by distance 𝑥 , while the position of mass 𝑚 is given by distance 𝑥 .
The equation of motion for mass 𝑚 can be written as

𝑚 𝑥̈ + 𝑘 𝑥 − 𝑘 (𝑥 − 𝑥 ) = 0
or, collecting terms
𝑚 𝑥̈ + (𝑘 + 𝑘 )𝑥 − 𝑘 𝑥 = 0
Similarly, the equation of motion for mass 𝑚 can be written as
𝑚 𝑥̈ + 𝑘 (𝑥 − 𝑥 ) = 0
or, collecting terms again
𝑚 𝑥̈ − 𝑘 𝑥 + 𝑘 𝑥 = 0
Writing the two equations in matrix form we obtain
𝑚 0 𝑥̈ 𝑘 +𝑘 −𝑘 𝑥 0
+ 𝑥 =
0 𝑚 𝑥̈ −𝑘 𝑘 0
The first 2 x 2 matrix is the mass matrix, 𝑀,
𝑚 0
𝑀=
0 𝑚
while the second 2 x 2 matrix is the stiffness matrix 𝐾,
𝑘 +𝑘 −𝑘
𝐾=
−𝑘 𝑘
The motions of the two masses are coupled through the off-diagonal terms in the stiffness
matrix, which means that the motion of one mass is not independent of the other.
8
To solve the set of equations we start by assuming that the motion of each mass is
governed by simple harmonic motion, which means that the motion of each mass can be
described by a function of the form
𝑥 =𝐴 𝑒 𝑥 =𝐴 𝑒
where 𝐴 ≠ 𝐴 . Differentiation of both functions to find 𝑥̈ and 𝑥̈ , followed by
substitution into the equations of motion then yields
(𝑘 + 𝑘 ) − 𝑚 𝜔 𝐴 − 𝑘 𝐴 𝑒 =0
(− 𝑘 𝐴 + (𝑘 − 𝑚 𝜔 )𝐴 )𝑒 =0
The non-trivial solution is found by setting the expressions within the brackets to zero,
that is
(𝑘 + 𝑘 ) − 𝑚 𝜔 𝐴 − 𝑘 𝐴 = 0
− 𝑘 𝐴 + (𝑘 − 𝑚 𝜔 )𝐴 = 0
This can be written in matrix form as
𝑘 +𝑘 −𝑘 𝑚 0 𝐴 0
−𝜔 =
−𝑘 𝑘 0 𝑚 𝐴 0
or in shorthand form as
[𝐾] − 𝜔 [𝑀] {𝐴} = {0}
Once again, we have two possibilities for a solution. Either
𝐴 0
=
𝐴 0
which is a trivial solution, or
𝑘 +𝑘 −𝑘 𝑚 0 0
−𝜔 =
−𝑘 𝑘 0 𝑚 0
To find a solution to the latter requires calculating the determinant of the resulting 2 x 2
matrix, and finding the values of 𝜔 that will yield a value of zero.
(𝑘 + 𝑘 ) − 𝑚 𝜔 −𝑘
= (𝑘 + 𝑘 ) − 𝑚 𝜔 (𝑘 − 𝑚 𝜔 ) − (−𝑘 )(−𝑘 )
−𝑘 𝑘 −𝑚 𝜔
= 𝑚 𝑚 𝜔 − (𝑘 + 𝑘 )𝑚 + 𝑘 𝑚 𝜔 + (𝑘 + 𝑘 )𝑘 − 𝑘 =0
The result is a quadratic equation for 𝜔 , which can then be solved to find the two values
of 𝜔 which represent the natural frequencies of vibration for the two masses.
In general, for a system with 𝑛 degrees of freedom, we will have 𝑛 natural frequencies
which can be solved for by finding the values of 𝜔 for which the determinant of an 𝑛 x
𝑛 matrix is equal to zero, that is
|[𝐾] − 𝜔 [𝑀]| = 0
9
If we expand the determinant calculation as above, this generally involves finding the 𝑛
roots of an 𝑛-degree polynomial.
An easier way of obtaining the natural frequencies is to note that if we use matrices this
can be recast as an eigenvalue problem, where the eigenvalues represent the natural
frequencies at which our system will vibrate. Letting 𝜆 = 1⁄𝜔 , and noting that the
identity matrix 𝐼 is defined by a matrix with 1’s on the leading diagonal, and 0’s
everywhere else, such that
1 ⋯ 0
𝐼= ⋮ ⋱ ⋮
0 ⋯ 1
we can rewrite our requirement for the determinant of the matrix [𝐾] − 𝜔 [𝑀] to be
zero as
|[𝐾] [𝑀] − 𝜆[𝐼]| = 0
before solving for the eigenvalues 𝜆 of the matrix [𝐾] [𝑀] .
Once the natural frequencies have been determined from the eigenvalues, they can then
be used to calculate the relative ratios of the displacement for each mass at each natural
frequency by substitution back into the original equations of motion. These ratios
correspond to the mode shapes associated with each natural frequency. For our two mass
system, mode shape 1 is given by
𝐴 𝑘 𝑘 −𝑚 𝜔
= =
𝐴 (𝑘 + 𝑘 ) − 𝑚 𝜔 𝑘
while mode shape 2 is given by
𝐴 𝑘 𝑘 −𝑚 𝜔
= = .
𝐴 (𝑘 + 𝑘 ) − 𝑚 𝜔 𝑘
If the initial displacements of the two masses correspond to the relative displacements of
one of the mode shapes the system will vibrate at the corresponding natural frequency.
On the other hand, if the initial displacements of the two masses are random, the motion
of the system will be some combination of the two mode shapes and frequencies
superimposed on top of each other.
In the case where we are using matrices, the mode shapes are given by the eigenvectors
of the matrix [𝐾] [𝑀] .
10
Solving for Eigenvalues and Eigenvectors using MATLAB:
When using MATLAB, the eigenvalues and eigenvectors can be solved for in one step
using the following form of the function eig :
[V,D]=eig(A)
where D is an 𝑛 by 𝑛 matrix containing the eigenvalues on the leading diagonal, and V is
an 𝑛 by 𝑛 matrix containing the eigenvectors, with each column of V corresponding to an
eigenvector. Input matrix A is equal to the matrix [𝐾] [𝑀] , which can be written as
inv(K)*M for input to function eig.
The function diag can be used to extract the leading diagonal of matrix D to a 𝑛 by 1
vector:
D=diag(D)
before using the following to calculate the corresponding natural frequencies:
omega = sqrt(D.^-1)
11
Example:
𝑥 𝑥
𝑘 𝑘 𝑘
𝑚 𝑚
Determine the natural frequencies and mode shapes of the system shown above,
assuming that masses 𝑚 and 𝑚 have identical masses of 10 kg, and that spring
constants 𝑘 , 𝑘 , and 𝑘 are also identical with a value of 100 N/m.
12
13
14
Multi-degree of Freedom Frame Structures:
𝑧 𝑧
𝜃 𝜃
𝑥 𝑥
𝑧 𝑧
𝜃 𝜃
𝑥 𝑥
A frame structure can be visualized as an assemblage of beams, columns, and walls that
are interconnected at nodal points or nodes, where the displacements of the nodes define
the number of degrees of freedom. In general, each node in a planar two-dimensional
structure has three degrees of freedom (two translational and one rotational), while in the
three-dimensional case this becomes six degrees of freedom (three translational and three
rotational). By way of example, the simple structure shown above has four nodes, each
with three degrees of freedom, which yields a total of 12 degrees of freedom for the
structure (four vertical displacements, four horizontal displacements, and four rotational
displacements).
For low-rise structures the axial deformation of the columns under the action of gravity
loads are negligible, which means that the vertical displacements can ignored. For high-
rise structures, gravity loads will affect the stiffness of the columns with height in the
building, but once again if the axial displacements are small relative to the horizontal
displacements of the structure they can be ignored.
Axial deformations of the beams supporting the floors of the structure can also be
ignored. This means that the horizontal displacements of the nodes on a given floor in the
structure will all be the same which means that they can be modelled by a single
displacement. This allows the number of degrees of freedom to be reduced, which in the
case of the simple structure shown above means that the number of degrees of freedom
can be reduced from 12 to six (two horizontal displacements, and four rotational
displacements) as shown below. In general, the reduction of the number of degrees of
freedom will depend on the number of nodes and bays in the structure.
15
𝜃 𝜃
𝜃 𝜃
Simple Shear Structures:
The simplest type of frame structure to consider is one in which the beams and floor
structure are infinitely stiff, that is 𝐸𝐼 , ≫ 𝐸𝐼 . In this case rotational
displacement of the individual nodes in the frame is not possible, and we only have to
consider the horizontal displacement of the individual floors, with the number of degrees
of freedom being equal to the total number of floors in the structure. Although the mass is
distributed throughout the building, we treat it as though it were a series of lumped
masses at floor level, which is a reasonable assumption for multistory buildings where the
mass is typically concentrated at floor level.
The motion of each floor, or node, is governed by Newton’s second law, which means
that in order to solve for the natural frequencies and mode shapes we must satisfy the
matrix equation
[𝑀]{𝑥̈ } + [𝐾]{𝑥} = 0
where [𝑀] is the mass matrix, [𝐾] is the stiffness matrix, and {𝑥̈ } and {𝑥} are the
accelerations and displacements of the individual floors respectively. The mass matrix
[𝑀] simply contains the individual floor masses along the leading diagonal of the matrix,
with zeroes elsewhere, that is
𝑚 ⋯ 0
[𝑀] = ⋮ ⋱ ⋮
0 ⋯ 𝑚
where 𝑚 , 𝑚 , … , 𝑚 are the individual floor masses.
To obtain the stiffness matrix [𝐾] we need to consider the stiffness of the columns
supporting the individual floors, noting that in general the resulting matrix will not only
include terms along the leading diagonal, but also in the off-diagonal positions which
leads to cross-coupling between the motion of the individual floors.
Recall that for a fix-ended column with modulus 𝐸, second moment of area 𝐼, and length
ℓ subjected to deflection Δ the following applies:
16
Δ
𝑃
𝑀
12𝐸𝐼
𝑃= Δ
ℓ
ℓ 6𝐸𝐼
𝑀 =𝑀 = Δ
ℓ
𝑀
𝑃
Note that the moments 𝑀 and 𝑀 need only be considered when rotation of a node is
possible. For our simple shear structure where no rotation is possible we need only
consider the horizontal shear force 𝑃 required to cause a displacement Δ.
The stiffness associated with each floor can be obtained by considering what happens
when an individual floor is given a unit displacement, while holding the displacement of
all other floors to zero. In the general case this requires applying additional loads to the
floors immediately above and below the floor being considered to hold the displacements
of these floors to zero, which then leads to cross-coupling terms appearing in the stiffness
matrix [𝐾].
Once we have obtained the mass and stiffness matrices we can then solve for the
eigenvalues and eigenvectors of the matrix [𝐾] [𝑀] to obtain the natural frequencies
and mode shapes for free vibration of our simple shear structure.
17
Example:
𝑚 = 20000 kg
𝑥
ℎ = 3.0 m 𝑚 = 15000 kg
𝑥 𝑚 = 12500 kg
ℎ = 3.0 m 𝑚 = 7500 kg
𝑥
ℎ = 3.0 m 𝐼 = 113 × 10 mm
𝑥 𝐼 = 76.6 × 10 mm
𝐼 = 76.6 × 10 mm
ℎ = 4.5 m
𝐼 = 52.7 × 10 mm
Determine the mass and stiffness matrices for the simple shear structure shown above,
before using the resulting matrices to calculate the natural frequencies and mode shapes
of the structure. The structure has the indicated properties, with a single bay in the
direction pointing into the page. The value of 𝐸 for all columns is equal to 200 GPa.
18
19
20
Normalization of Mode Shapes:
The resulting eigenvectors simply define a set of multiplicative factors that describe the
mode shapes of the structure, and any set of values that are a multiple of a particular
eigenvector can be used to describe the mode shape associated with a specific natural
frequency. We often normalize the mode shapes to standardize their elements associated
with various degrees of freedom. This is typically done by either normalizing each mode
by the largest value, positive or negative, so that the largest element in the mode shape is
unity, or by normalizing each mode by the value associated with a particular degree of
freedom, such as the top floor of a multistory building. In this course we will use the
former method.
21
Frame Structures with Shear and Rotation:
In the case where the beams and floor structure are not infinitely stiff, we must account
for both horizontal translation and rotation of the nodes. This is done by considering the
horizontal and rotational displacements separately. For the horizontal displacements the
determination of both the mass and stiffness coefficients that appear in the corresponding
mass and stiffness matrices is done as before, albeit with the inclusion of the necessary
moments to yield rotational displacements of zero.
For the rotational displacements it is necessary to consider the forces and moments
required to cause an angular rotation of 𝜃 for a fixed-ended column or beam with
modulus 𝐸, second moment of area 𝐼, and length ℓ:
𝜃
𝑃
𝑀 6𝐸𝐼
𝑃= 𝜃
ℓ
2𝐸𝐼
ℓ 𝑀 = 𝜃
ℓ
4𝐸𝐼
𝑀 = 𝜃
𝑀 ℓ
𝑃
An underlying assumption is that the columns and beams meeting at a particular node
remain at right angles to each other, and that they all undergo the same angle of rotation.
The stiffness associated with each node can be obtained by considering what happens
when an individual node is given a unit rotation, while holding the rotation of all other
nodes to zero. In the general case this requires applying additional loads to the nodes
surrounding the node being considered to hold the rotation of these nodes to zero, which
once again leads to cross-coupling terms appearing in the stiffness matrix [𝐾].
Finally, the mass coefficient associated with rotational displacement of a node is the mass
moment of inertia, rather than the actual mass associated with the node. In general, the
effect of the mass moment of inertia of individual nodes on the dynamics of most
practical structures is negligible, and so the mass coefficients associated with rotational
displacements are simply set to zero.
22
Example:
𝜃 𝜃 𝑥
𝑚 = 20000 kg
ℎ = 3.0 m 𝜃 𝜃 𝑥 𝑚 = 15000 kg
𝐼 = 113 × 10 mm
ℎ = 4.5 m 15 m
𝐼 = 76.6 × 10 mm
Determine the mass and stiffness matrices for the structure shown above, with the
indicated degrees of freedom, before calculating the natural frequencies and mode
shapes. The structure has the indicated properties, with a single bay in the direction
pointing into the page. The value of 𝐸 for all columns and beams is equal to 200 GPa,
while the moment of inertia 𝐼 is the same for the columns and beams at a particular level.
23
24
25
26
Static Condensation:
Static condensation allows us to eliminate those degrees of freedom that have no mass
associated with them if we are dealing with a structure where we have both horizontal
and rotational displacements. If this is the case we can eliminate the degrees of freedom
associated with the rotational displacements provided that there are no external moments
being applied to these degrees of freedom, which is usually the case for both earthquake
and wind loading.
For a multi-degree of freedom structure with no damping subjected to a set of external

time-varying loads {𝑃(𝑡)} that are being applied in the horizontal direction only we can
write the following partitioned set of matrix equations
[𝑀] 0 {𝑥̈ } [𝐾 ] [𝐾 ] {𝑥} {𝑃(𝑡)}
̈ + =
0 0 𝜃 [𝐾 ] [𝐾 ] {𝜃} 0
where [𝑀] is the mass matrix associated with the horizontal displacements, [𝐾 ] and [𝐾 ]
are the stiffness matrices associated with the horizontal and rotational displacements
respectively, while [𝐾 ] and [𝐾 ] are the stiffness matrices containing the cross-
coupling terms between the horizontal and rotational displacements. We can also note
that the stiffness matrix [𝐾 ] is the transpose of stiffness matrix [𝐾 ], that is
[𝐾 ] = [𝐾 ]
The lower part of the partitioned set of matrix equations yields the result
[𝐾 ]{𝑥} + [𝐾 ]{𝜃} = 0
which then allows the rotational displacements {𝜃} to be expressed in terms of the
horizontal displacements {𝑥} using
{𝜃} = −[𝐾 ] [𝐾 ]{𝑥} = −[𝐾 ] [𝐾 ] {𝑥}
This result can then be used to replace the rotational displacements that appear in the
upper part of the partitioned set of matrix equations to give
[𝑀]{𝑥̈ } + [𝐾 ]{𝑥} + [𝐾 ]{𝜃} = {𝑃(𝑡)}
[𝑀]{𝑥̈ } + [𝐾 ]{𝑥} − [𝐾 ][𝐾 ] [𝐾 ]{𝑥} = {𝑃(𝑡)}
[𝑀]{𝑥̈ } + [𝐾 ] − [𝐾 ][𝐾 ] [𝐾 ] {𝑥} = {𝑃(𝑡)}
or
[𝑀]{𝑥̈ } + [𝐾 ]{𝑥} = {𝑃(𝑡)}
where [𝐾 ] is the condensed stiffness matrix defined as
[𝐾 ] = [𝐾 ] − [𝐾 ][𝐾 ] [𝐾 ]
or, alternatively
[𝐾 ] = [𝐾 ] − [𝐾 ][𝐾 ] [𝐾 ]
27
Example:
Determine the condensed stiffness matrix for the example on page 23, before calculating
the natural frequencies and mode shapes and comparing them to the values obtained
when the rotational degrees of freedom were included.
28
29
30
Notation:
Before considering some of the properties of the mode shapes, we first consider the
notation to be used. In particular we note that we can write both the natural frequencies
(eigenvalues) and mode shapes (eigenvectors) in compact matrix form. The 𝑖th natural
frequency, 𝜔 , is associated with the corresponding mode shape {𝜙 }, which is a column
vector. We can assemble the individual mode shapes into a single square matrix, where
each column of the matrix defines a particular mode shape. For an 𝑛-degree of freedom
system this can be written as
𝜙 ⋯ 𝜙
[Φ] = 𝜙 = ⋮ ⋱ ⋮
𝜙 ⋯ 𝜙
where the matrix [Φ] is called the modal matrix. Similarly we can write the 𝑛 natural
frequencies 𝜔 as a diagonal matrix [Ω ], where
𝜔 ⋯ 0
[Ω ] = ⋮ ⋱ ⋮
0 ⋯ 𝜔
Mode Shapes and Free Vibration:

If a multi-degree of freedom structure vibrates at one of its natural frequencies then the
time-varying deflection of the structure will be given by a function of the form
{𝑥 (𝑡)} = 𝑞 (𝑡){𝜙 }
where {𝑥 (𝑡)} is a vector containing the time-varying displacements of the masses for
mode 𝑖, 𝑞 (𝑡) is a simple harmonic function describing the time variation of the
displacements, and {𝜙 } is the mode shape associated with natural frequency 𝜔 . The
function 𝑞 (𝑡) can take any form that is compatible with simple harmonic motion, i.e.
𝑞 (𝑡) = 𝐴 𝑒 𝑞 (𝑡) = 𝐴 sin(𝜔𝑡) 𝑞 (𝑡) = 𝐴 cos(𝜔𝑡)
In the more general case where the structure is vibrating at a frequency that is not equal
to one of its natural frequencies the time-varying displacements will be given by a
combination of contributions from all the mode shapes, that is
{𝑥(𝑡)} = 𝑞 (𝑡){𝜙 }
31
Orthogonality of Modes:
One of the most important properties of the mode shapes is that they are orthogonal,
which means that the modes are independent of each other.
Mode 1 Mode 2 Mode 3 Mode 4
-1.0 0.0 1.0 -1.0 0.0 1.0 -1.0 0.0 1.0 -1.0 0.0 1.0
Recall that each natural frequency and mode shape must satisfy the relationship
[𝐾] − 𝜔 [𝑀] {𝜙 } = {0}
or, in compact matrix form
[𝐾] − [Ω ][𝑀] [Φ] = {0}
The above equation can be rearranged so that for the 𝑖th natural frequency and mode
shape it becomes
[𝐾]{𝜙 } = 𝜔 [𝑀]{𝜙 }
Now multiply this equation by the transpose of the jth mode shape, that is 𝜙 . This
yields
𝜙 [𝐾]{𝜙 } = 𝜔 𝜙 [𝑀]{𝜙 }
We can write a similar expression for the 𝑗th natural frequency and mode shape, that is
[𝐾] 𝜙 = 𝜔 [𝑀] 𝜙
before multiplying by the transpose of the 𝑖th mode shape {𝜙 } , as before, to obtain
{𝜙 } [𝐾] 𝜙 = 𝜔 {𝜙 } [𝑀] 𝜙
If we now take the transpose of this equation, that is
{𝜙 } [𝐾] 𝜙 = 𝜔 {𝜙 } [𝑀] 𝜙
and noting that
[𝐴][𝐵] = [𝐵] [𝐴]
the equation becomes
𝜙 [𝐾] {𝜙 } = 𝜔 𝜙 [𝑀] {𝜙 }
32
Finally, since both the mass and stiffness matrices are symmetric about the leading
diagonal, that is [𝑀] = [𝑀] and [𝐾] = [𝐾] , the equation can be rewritten as
𝜙 [𝐾]{𝜙 } = 𝜔 𝜙 [𝑀]{𝜙 }
Compare this equation to the one that we obtained by multiplying the equation for the 𝑖th
mode by the transpose of the 𝑗th mode shape
𝜙 [𝐾]{𝜙 } = 𝜔 𝜙 [𝑀]{𝜙 }
𝜙 [𝐾]{𝜙 } = 𝜔 𝜙 [𝑀]{𝜙 }
The left hand side of both equations is identical, as is the term 𝜙 [𝑀]{𝜙 } that appears
on the right hand side. Equating the right hand sides of the equation we get
𝜔 𝜙 [𝑀]{𝜙 } = 𝜔 𝜙 [𝑀]{𝜙 }
or
𝜔 −𝜔 𝜙 [𝑀]{𝜙 } = 0
Since 𝜔 ≠ 𝜔 , the only way this equation can be satisfied for 𝑖 ≠ 𝑗 is if
𝜙 [𝑀]{𝜙 } = 0
This defines the orthogonality condition for mode shapes {𝜙 } and 𝜙 using the mass
matrix [𝑀]. The corresponding condition when the stiffness matrix [𝐾] is used is
𝜙 [𝐾]{𝜙 } = 0
Noting that the deflected shape described by the mode shapes do not change with time,
the time-varying displacements associated with a particular mode shape can be described
mathematically by a function of the form
{𝑥 (𝑡)} = 𝑞 (𝑡){𝜙 }
where 𝑞 (𝑡) is a simple harmonic function describing the time variation of the
displacement. The corresponding accelerations are then
{𝑥̈ (𝑡)} = 𝑞̈ (𝑡){𝜙 }
and the associated inertial forces are
[𝑀]{𝑥̈ (𝑡)} = 𝑞̈ (𝑡)[𝑀]{𝜙 }.
The work done by the inertial forces of the 𝑖th mode shape on the displacements of the
𝑗th mode shape is equal to
([𝑀]{𝑥̈ (𝑡)}) 𝑥 (𝑡) = (𝑞̈ (𝑡)[𝑀]{𝜙 } ) 𝑞 (𝑡) 𝜙 = {𝜙 } [𝑀] 𝜙 𝑞̈ (𝑡)𝑞 (𝑡)
where we have made use of the fact that [𝑀] = [𝑀] . Noting that the orthogonality
condition requires that
{𝜙 } [𝑀] 𝜙 = 0
33
the work done by the the 𝑖th mode shape on the displacements of the 𝑗th mode shape is
therefore zero.
34
Example:
Show that the mode shapes of the four-storey simple shear structure example considered
on page 18 of the notes are orthogonal using both the mass and stiffness conditions for
orthogonality.
35
36

Free Vibration of MDOFSystems

Uploaded by

Free Vibration of MDOFSystems

Uploaded by

CEE 4491A Structural Dynamics II Free Vibration of MDOF Systems

Free Vibration of Multi-Degree of Freedom Systems

Simple Multi-degree of Freedom Systems:

The equation of motion for mass 𝑚 can be written as

Solving for Eigenvalues and Eigenvectors using MATLAB:

Multi-degree of Freedom Frame Structures:

Simple Shear Structures:

Normalization of Mode Shapes:

Frame Structures with Shear and Rotation:

For a multi-degree of freedom structure with no damping subjected to a set of external

Mode Shapes and Free Vibration:

Mode 1 Mode 2 Mode 3 Mode 4

You might also like