Lateral Natural Frequencies of Rotor Shaft Systems

via the Finite Element Method

P. H. Mathuria
Email address: [email protected].
Institute Name: Indian Institute of Technology, Bombay.

Abstract
Two analytical methods, the Transfer Matrix method (TMM) and the Finite Element method
(FEM), are discussed for evaluating the lateral natural frequencies of a shaft rotor system.
Hypothetical cases are considered for evaluating fundamental and higher frequencies. A
MATLAB program is developed for the Finite Element method. It may be useful for evaluating
the frequencies of the system's higher modes as well as the fundamental frequency.

Introduction
Consider a rotor shaft system in which the operating speed matches the frequencies of the
higher mode. In this case, it is necessary to evaluate the higher frequencies of the system. It is
necessary to know the higher frequencies while taking the rotor through a critical speed to an
operational speed. To avoid failures of shafting, the general practice in the design of rotors is to
determine the bending critical speeds as well as higher frequencies. For most rotors, it is the
fundamental mode which falls in the running speed zone. There are several methods of
calculation of the critical speed:
1.
2.
3.
4.

Dunkerley method
Rayleigh method
Transfer Matrix method
Finite Element method

The first two methods are suitable for estimating the fundamental frequency by hand calculation.
The transfer matrix method and FEM would require the use of computers. The third and fourth
methods are also useful for evaluating the higher frequencies. In this paper, higher frequency
methods are discussed, specifically the Transfer Matrix method and Finite Element method.

Transfer Matrix Method

In a manner similar to Holzer method, Myklestad and Prohl [1] developed highly successful
method of computation for the bending critical speeds of a shaft. Consider an n mass system,
each mass representing a gear, a disk, or a flywheel etc. All these masses are taken as
lumped with their gyroscopic inertia neglected. We adopt a root searching technique to
determine the natural frequency . The algorithm of the program for TMM is as follows:
TMM Algorithm
1. Take the trial value of frequency .
2. Assign the element in the matrix with different variable.
3. Form the mass and field matrix.
4. Arrange the matrix in the order of multiplication and multiply it.
5. Apply boundary condition to overall transfer matrix.
6. Find the determinant value. If it is zero, or nearly zero, then this will be natural
frequency, otherwise repeat the same procedure for different values of till the
determinant approaches to zero.
7. After evaluating the first mode, repeat the same procedure by starting the frequency value
higher than the evaluated one.
Another method for evaluating the fundamental frequency, as well as higher frequencies, is the
Finite Element Method.

Finite Element Method

The main procedures in the finite element analysis are [2]
1. Read input data and allocate proper array sizes.
2. Calculate element matrices and vectors for every element.
3. Assemble matrices and vectors into the system matrix and vector.
4. Apply constraints to the system matrix and vector.
5. Solve the matrix equation for the eigenvalues.

w
w1

w2

EI

2
x

Figure 1: Beam Element

Input Data
The major input parameters needed for the finite element analysis program are:

the number of total nodes in the system

the number of total elements in the system
coordinate values of every node in terms of the global coordinate system
types of every element
information for boundary condition
element properties (modulus of elasticity, moment of inertia, length, area, poison's ratio,
etc.)

Element Matrices for Stiffness and Mass

The beam element is used for analysis of the rotor shaft system. A typical beam element is
shown in Figure 1. w1 and w2 are the vertical displacements at points 1 and 2, respectively.
1 and 2 are the slopes at point 1 and 2, respectively. l is length of beam element. The
derivation of the stiffness and mass matrices is given in Appendix A.
The following are the stiffness matrix and mass matrices used for evaluating the natural
frequencies.

The stiffness matrix [ K ] is

6l 12 6l
12

2
6l 2l 2
EI 6l 4l
[ K] =
l 3 12 6l 12 6l

2
6l 4l 2
6l 2l

(1)

The consistent mass matrix [ M ] cons is

22l
156
22 l
4l 2
m
[ M ] cons =
13l
420 54

2
13l 3l

54
13l
13l 3l 2
156 22l

22l 4l 2

(2)

The lumped mass matrix [ M ] Lumped is

m / 2
0
[ M ] Lumped =
0

0
0
0
0
0 m/2
0
0

0
0
0

(3)

The diagonal mass matrix [ M ] diag is

0
0
0
m 2
0
2
ml 78
0
0
[ M ] diag =
0
0
m 2
0

2
0
0
ml 78
0

(4)

Assembly of Element Matrices

Once these matrices are computed, they are assembled into the global system stiffness and mass
matrices.

Application of Constraints
The boundary condition of the simply supported shaft rotor system is applied to the global mass
and stiffness matrices, i.e. the deflection at each simply supported node point is zero.
Solving for the eigenvalues
Once the system matrices are modified than simply using following command line in MATLAB
software, we can evaluate the eigenvalue of the system.
= sqrt ( eig ( K , M ))
This two methods are demonstrated by solving the following three problems.

Problem - I
A uniform shaft with two disks and supported bearings is shown in Figure 2. Evaluate the
natural frequencies of the system.

Problem - II
A stepped shaft with one disk and supported bearings is shown in Figure 3. Evaluate the
natural frequencies of the system.

Problem - III
A uniform shaft with three disks and supported bearings is shown in Figure 4. Evaluate the
natural frequencies of the system.

Solution
The Transfer Matrix Method and modified computer program are used to obtain the natural
frequencies shown in Table 1. The table also shows the natural frequencies using FEM.
6, 4, and 4 beam elements are considered for evaluating the frequencies of problems I,
II and III, respectively. Typical FEM models for the above three problems are shown in
Figures 5, 6 and 7. In these figures, Roman letters denote the number of elements. The length
of each element is also given. All problems are evaluated using consistent, lumped and diagonal
mass matrix. The results are shown in Table 2. Table 1 shows a comparison of the Transfer
matrix method and the Finite element method using a lumped mass matrix.

m1 = 30 kg

m 2 = 50 kg

25

125

250

250

Figure 2: Example 1. Two disk and shaft system

All dimensions are in mm unless otherwise specified.

m1 = 90 kg
31.75

25

250

250

Figure 3: Example 2. Step shaft and disk system

All dimensions are in mm unless otherwise specified.

m 2 = 400 kg
m1 = 100 kg

m 3 = 200 kg

100

500

500

500

Figure 4: Example 3. Three disk and shaft system

All dimensions are in mm unless otherwise specified.

500

II

62.5

62.5

III

IV

125

125

VII

125

125

The dimensions are in units of mm.

Node point
Rotor mass added in the node
Element

Figure 5: FEM model for example 1

II

125

125

III

IV

125

125

Node point
Rotor mass added in the node
Element

Figure 6: FEM model for example 2

II

500

500

III

IV

500

500

Node point
Rotor mass added in the node
Element

Figure 7: FEM model for example 3

Table 1: Comparison of natural frequencies using transfer matrix method with FEM
Problem
Transfer Matrix Method
Finite Element Method with
Lumped Mass Matrix
fn 1(Hz)
fn 2 (Hz)
fn 3 (Hz)
fn 1 (Hz)
fn 2 (Hz)
fn 3 (Hz)
I
18.34
78.06
*
18.42
77.24
873.36
II
24.26
*
*
24.31
906.09
1320.00
III
16.52
88.87
167.52
15.70
81.23
151.67
III
15.54**
79.24**
147.85
15.70
81.23
151.67
* - unable to evaluate using developed program due to divergence problem.
** - shaft mass is added into rotor mass for transfer matrix method.
1,2,3 - denotes first, second, and third mode frequencies respectively.

10

Table 2: Comparison of natural frequencies using FEM with different mass matrix
Problem
No.

Consistent Mass Matrix

fn
1(Hz)
I
II
III

18.41
24.32
15.70

fn 2
(Hz)

fn 3
(Hz)

77.31
1026.18
81.42

1009.03
1804.68
153.77

Lumped Mass Matrix

Diagonal Mass Matrix

fn 1
(Hz)

fn 2
(Hz)

fn 3
(Hz)

18.42
24.31
15.70

77.24
906.09
81.23

873.36
1320.00
151.67

fn 1
(Hz)
18.41
24.29
15.69

fn 2
(Hz)

fn 3
(Hz)

77.18
1162.11
80.84

1113.09
1526.99
150.88

1,2,3 - denotes first, second, and third mode frequencies respectively.

The corresponding mode shapes of the above examples are shown in Figures 8, 9, and 10. The
mode shapes are plotted as relative amplitude versus length of beam.

Conclusion
Both the Transfer Matrix method and the Finite Element method are suitable for evaluating the
fundamental as well as higher frequencies. The Transfer Matrix method gave very similar results
to the Finite Element model with lumped mass for the fundamental frequency, as shown in Table
1.
The Transfer Matrix method for problem II was only able to obtain the fundamental frequency.
It was unable to converge at higher frequencies. When the number of elements is increased
from 2 to 4, the values of higher frequencies is obtained, as shown in Table 1. The Transfer
Matrix method and FEM give nearly same values of the frequencies when the mass of the shaft
is negligible.
Problem III gives a better idea of neglecting mass and adding mass on the system with TMM.
The effect of mass matrix in FEM is also observed in Table 2. For higher modes, TMM takes
more computation time due to iterations, but that is not case with FEM. If engineers take more
elements in the same system, then it will take more time for computation otherwise it is faster
than TMM. Sometimes more elements are required for increasing precision of eigenvalues to
achieve exact solution and mode shapes.
The Rayleigh Ritz method also gives the higher frequencies, but it requires judgement of the
mode shape. It is tedious and time consuming.

11

FIRST MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200

300

400

500

600

400

500

600

400

500

600

x (mm)

SECOND MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200

300
x (mm)

THIRD MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200

300
x (mm)

Figure 8. Mode Shape of Example 1

12

FIRST MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200

300

400

500

300

400

500

300

400

500

x (mm)

SECOND MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200
x (mm)

THIRD MODE

Displacement

0.10
0.05
0
-0.05
-0.10

100

200
x (mm)

Figure 9. Mode Shape of Example 2

13

FIRST MODE

Displacement

0.4
0.2
0
-0.2
-0.4

500

1000

1500

2000

1500

2000

1500

2000

x (mm)

SECOND MODE

Displacement

0.4
0.2
0
-0.2
-0.4

500

1000
x (mm)

THIRD MODE

Displacement

0.4
0.2
0
-0.2
-0.4

500

1000
x (mm)

Figure 10. Mode Shape of Example 3

14

Appendix A
Stiffness Matrix
A two-node beam element having four degree of freedom is shown in Figure 1 [3]. Rotation
is assumed to be small, so that = dw / dx.
Four degree of freedom define a cubic lateral displacement field,
w = [ N][ w1 1 w2 2 ] T = [ N]{ d}

(5)

where,
[N] = assumed shape functions is corresponding to cubic lateral displacement and are given
in Table 3,

1
2
w1
w2

= rotation at node 1,
= rotation at node 2,
= lateral displacement at node 1,
= lateral displacement at node 2,

15

x=0

x=l

1
(a)

1
(b)

1
(c)

1
(d)

Figure 11:
Typical shape functions of a cubic fitted to ordinates and slopes at x = 0 and at x = l

16

Table 3:
Typical shape functions of a cubic fitted to ordinates and slopes at x=0 and at x= l.
Figure
Shape function
At x = 0
At x = l
Ni
Ni , x
Ni
3

3x 2 2 x
+ 3
l2
l

2x 2 x 3
+ 2
l
l
3
2
2x
3x
N3 = 2 + 3
l
l
2
x3
x
N4 =
+ 3
l
l

11a

N1 = 1

11b

N2 = x

11c
11d

Ni , x

For constant EI, the element stiffness matrix given by

6l 12
12

4l 2 6l
EI 6l
T
[ K ] = [ B ] EI [ B ]dx = 3
l 12 6l 12

2l 2 6l
6l

[ B] =

6 12
[
N
]
=
2 + 3
l
dx 2
l
d2

6l
2l 2

6l

4l 2

4 6 x 6 12x 2 6x
+

+ 2
l l2 l2 l3
l l

(6)

(7)

Mass Matrix
A mass matrix is a discrete representation of a continuous distribution of mass. A consistent
mass matrix is defined by
22l
156
22l
4l 2
m
T
[ M ] cons = [ N] [ N]dV =
13l
420 54

2
13l 3l

54
13l
13l 3l 2
156 22l

22l 4l 2
(8)

17

It is termed consistent because [N] represents the same shape functions as are used to
generate element stiffness matrix. A simpler and historically earlier formulation is the lumped
mass matrix, which is obtained by placing particle masses M i at nodes i of an element, such

that mi is the total element mass. Particle lumps have no rotary inertia unless rotary
inertia is arbitrarily assigned. A lumped mass matrix is diagonal, but a consistent matrix is
not. Historical lumped mass matrix is given by Lumped mass matrix,

m / 2
0
[ M ] Lumped =
0

0
0
0
0
0 m/ 2
0
0

0
0
0

0
(9)

Another scheme is HRZ Lumping scheme.

The HRZ scheme is an effective method for
producing a diagonal mass matrix. It is recommenced for arbitrary elements. The idea is to use
only the diagonal terms of the consistent mass matrix, but to scale them in such a way that the
total mass of the element is preserved. The specific procedure steps are as follows.
1. Compute only diagonal coefficients of the consistent mass matrix.
2. Compute the total mass of the element, m
3. Compute a number s by adding the diagonal coefficients M ii associated with translation
d.o.f (but not rotational d.o.f, if any) that are mutually parallel and in the same direction.
4. Scale all the diagonal coefficients by multiplying them by the ratio m/s, thus preserving the
total mass of the element.

As an example, consider a beam element, and follow the steps.

1. Diagonal element of consistent mass matrix,
156 0

2
m 0 4l
[ M] =
420 0
0

0
0

0
0
0

156 0

0 4l 2
0

18

(10)

2. Total mass element is element is m,

m 420
3. s = M (1,1) + M( 3,3) = m (156 + 156) = 312 m. Hence
=
420
420
s 312

4. Hence Diagonal mass matrix [ M ]diag ,

5.
[ M ]diag =

m
[m]
s

(11)

0
0
0
m 2
0
2
ml 78
0
0

[ M ]diag =
0
0
m 2
0

2
0
0
ml 78
0

(12)

References
[1] J. S. Rao. Rotor Dynamics. Wiley Eastern Ltd., 1983.
[2] Y. W. Kwon and H. Bang. The Finite Element Method using MATLAB. CRC press,
1997.
[3] R. D. Cook, D. S. Malkus, and M. E. Plesha. Concepts and Application of Finite
Element Analysis. John Wiley and Sons, 1989.

19