Proceedings of COBEM 2003 17th International Congress of Mechanical Engineering

COBEM2003 - 1134 Copyright © 2003 by ABCM November 10-14, 2003, São Paulo, SP

TRANSVERSE VIBRATION ANALYSIS OF ROTATING SHAFTS

Victor Fernando Deorsola Sacramento

Instituto Tecnológico de Aeronáutica – ITA
Div. de Engenharia Mecânica-Aeronáutica, 12228-900, São José dos Campos – SP Brazil
[email protected]

João Carlos Menezes

Instituto Tecnológico de Aeronáutica - ITA
Div. de Eng. Mecânica - Aeronáutica, 12228-901, São José dos Campos - SP – Brazil
[email protected]

Abstract: The aim of this work is to study transverse vibration of rotating shafts, with an approach that makes
possible, during design, to prevent unstable vibration at working speed. The Finite Element Method is used and
the motion equations are obtained from Lagrange equations which describe the motion in two transverse
planes. A local rotating coordinate system is used and for the shaft's motion matrix equation a gyroscopic
matrix has to be taken in account. The beam elements have two nodes and six degrees of freedom in each node,
three displacements and three rotations. As a result, each element produces twelve differential equations. The
Mathematica® software was used to obtain the mass, the gyroscopic, and the stiffness matrices, by integration
of the kinetic and the potential energies, and by differentiation of the energies according to Lagrange
equations. With the matrices obtained, a computer software written in Fortran accomplishes the assembly of the
global matrices of the shaft, and simulates the shaft’s response, according to established dimensions, material,
angular velocity. This software has been successfully tested taking a classical case as a reference. Comparison
of the results obtained with the analytical model, that doesn’t include unbalancing effects, reveasl a very good
agreement. These results are presented in this work.
Keywords. Mechanical Vibrations, Rotating Shafts, Gyroscopic effect

1. INTRODUCTION

Design of rotating equipment operating smoothly under various conditions of speed, has been a
challenge to designers, due to the fact that operating speeds are increasing every day and equipments have to
work satisfactorily in low and high speeds.
Designers have to obtain the critical speed of a system, and compare with the operating range, to
make sure that the motion will be stable and low-level amplitude of vibration prevails.
Under certain conditions, a rotating shaft can perform a motion with very high amplitude of vibration.
This motion, can lead to the destruction of the system. The aim of this work is to study transverse vibration of
rotating shafts, with an approach that makes possible, during design, to preview unstable vibration at working
speed. The subject is most interesting in the design of aeronautic turbines shafts.
The Finite Element Method is used, adopting beam elements with two nodal points and six degree of
freedom for each node. The motion equations are obtained from Lagrange’s equations, and describe the motion
in two transverse planes. A local rotating coordinate system is used and shaft's matrix equation of motion
presents a characteristic gyroscopic matrix.
Each node of the finite element has six degrees of freedom, three displacements and three rotations. As
a result, for each element, twelve differential equations have to be accounted for. To obtain the mass, gyroscopic
and stiffness matrix, the software Mathematica® was used for the integrations of the kinetic and potential
energy, and for the differentiation required by Lagrange’s equations. A Fortran computer software was written,
to assembly the shaft element matrices into global matrices, and to simulate the shaft response, for several
dimensions, materials, angular velocities and unbalanced loads.
To obtain the time domain response, the integration procedure of Newmark was adopted. This software
was tested, and numerical results were compared to a classical case reported by the literature. Results reveal a
good agreement with the analytical solution of the problem.
2. MATHEMATICAL DEVELOPMENT

Consider a vibrating shaft rotating about axis X with constant angular velocity Ω relative to the inertial
axes XYZ. A set of axes xyz rotating with the shaft is adopted according to Figure (1). The displacement vector
can be written in the vector form as
r r r
u = uy ⋅ ey + uz ⋅ ez (1)
r r
where ey and ez are unit vectors according to the rotating xyz axes and uy and uz are the corresponding
displacement, according to Figure 2.

Figure 1 – Rotating elastic shaft in bending simply supported at both ends

Figure 2 – Displacements and unit vectors for the rotating coordinate system

Differentiating the preceding equation with respect to time, one may obtain the velocity vector of a
typical point on the shaft as
 r
∂u  ∂uy  r  ∂uz  r
= − uz ⋅ Ω ey +  + uy ⋅ Ω  ⋅ ez (2)
∂t  ∂t   ∂t 
The kinetic energy T of the shaft can be stated as (Meirovitch, 1986 and 1997)

T = T transversal + T longitudinal + T rotational (3)

where
l 2
1  ∂ux 
2 ∫0
Tlongitudinal = ⋅ m ⋅  dx (4)
 ∂t 
where m= mass per unit length of the shaft and

ym = y + e ⋅ cos(α ) (5)

zm = z + e ⋅ sin( α ) (6)

where y and z are geometric center coordinates of the cross sectional area of the shaft, e is the distance from
geometric center to gravity center, an α is the initial angular position of the gravity center
l 2
1  ∂ux 
Tlongitudinal = ⋅ ∫ m ⋅   dx (7)
2 0  ∂t 

where ux=longitudinal displacement

l 2
1  ∂θ 
2
0
∫
Trotacional = ⋅ Ip ⋅   dx
 ∂t 
(8)

where Ip = mass polar moment of inertia of the cross-sectional área of the shaft.
The potential energy V is considered as

V = V transversal + V longitudinal + V rotational (9)

where
l 2
1  ∂ 2u 
2 ∫
Vtransversal = ⋅ E ⋅ Ia ⋅  2  dx
0
 ∂x 
 
(10)

where E = elasticity modulus and Ia = area momentof inertia of the cross - sectional area of the shaft
l 2
1  ∂ux 
Vlongitudinal = ⋅ ∫ E ⋅ A ⋅   dx (11)
2 0  ∂x 
where A = Cross - sectional area
l 2
1  ∂θ 
2 ∫
Vrotational = ⋅ G ⋅ Jp ⋅   dx
0
 ∂x 
(12)

where G = shear modulus and J p = polar moment of inertia of the cross - sectional areaof the shaft
 Figure 3 – Nodal displacements for a two noded beam finite element

A two noded beam element, with six degrees of freedom for each node, was adopted for the finite
element model, as seen in Figure 3.
To obtain the displacemets along the element, the following shape funtions should be used (Meirovitch,
1986):
 3x 2 2 x 3   x 2x 2 x3   3x 2 2 x 3   x 2 x3 
uy( x ) = 1 − 2 + 3  ⋅ uy1 +  − 2 + 3  ⋅ l ⋅ β 1 +  2 − 3  ⋅ uy 2 −  2 − 3  ⋅ l ⋅ β 2 (13)
 l l   l l l   l l   l l 
 3x 2 2 x 3   x 2x 2 x3   3x 2 2 x 3   x 2 x3 
uz (x ) = 1 − 2 + 3  ⋅ uz1 −  − 2 + 3  ⋅ l ⋅ γ 1 +  2 − 3  ⋅ uz 2 +  2 − 3  ⋅ l ⋅ γ 2 (14)
 l l  l l l   l l  l l 
 x   
x
ux ( x ) = 1 −  ⋅ ux1 +   ⋅ ux 2 (15)
 l l
 x  x
θ ( x ) = 1 −  ⋅ θ 1 +   ⋅ θ 2 (16)
 l l

Replacing Eqs. (13), (14), (15) and (16) into Eqs. (4), (7), (8), (10), (11) and (12), one may obtain the
total kinetic and potencial energies for the element, according to Eqs. (3) and (9). Equations (3) and (9) can,
finally, be expressed as a function of the nodal displacements ux1, θ1, uy1, γ1, uz1, β1, ux 2, θ2, uy2, γ 2, uz 2 and
β2 . To accomplish the integrations involved in the equations of kinetic and potential energies dependent on the
variable x, software Mathematica® was used. In order to obtain the equations of motion for the finite element,
Lagrange’s equations were adopted (Meirovitch, 1986 and 1997):

∂  ∂T  ∂T ∂V
 − + = Qi (17)
∂t  ∂q&i  ∂qi ∂qi

Where

T = element's kinetic energy

V = element's potential energy
qi = generalized coordinates
Qi = generalized forces

To accomplish the differentiations involved in Lagrange’s equations, the software Mathematica® was
utilized. Taking the nodal displacements ux1, θ1, uy1, γ1, uz1, β1, ux 2, θ2, uy2, γ 2, uz2 and β2 as generalized
coordinates in Lagrange’s equations, the equation of motion can be derived and expressed in the matrix form:

[m]⋅ {q&&} + [c]⋅ {q&}+ [k ]⋅ {q} = {Q} (18)

 Where

[m] = mass matrix

{&q&}= acceleration vector
[c] = gyroscopic matrix
{q&} = velocity vector
[k ] = stiffness matrix
{q} = displacement vector
{Q} = unbalancing forces vector

For the matrix equation of motion, the following matrices were obtained:

 lm lm 
3 0 0 0 0 0 0 0 0 0 0 
6
 Il Il 
0 0 0 0 0 0 0 0 0 0 
 3 6 
 13lm 11l 2 m 9lm − 13l 2 m 
0 0
35 210
0 0 0 0
70 420
0 0 
 
0 11l 2 m l 3m 13l 2 m − l 3m
0 0 0 0 0 0 0 
 210 105 420 140 
 13lm − 11l 2 m 9lm 13l 2 m 
0 0 0 0 0 0 0 0 
 35 210 70 420 
0 − 11l 2 m l 3m − 13l m2 3
l m 
0 0 0 0 0 0 0
[m] =  lm 210 105
lm
420 140 

 0 0 0 0 0 0 0 0 0 0 
6 3 
0 Il Il
0 0 0 0 0 0 0 0 0 
 6 3 
 9lm 13l 2 m 13lm − 11l 2 m 
0 0 0 0 0 0 0 0 
 70 420 35 210 
0 − 13l 2 m − l 3m − 11l 2 m l 3m
0 0 0 0 0 0 0 
 420 140 210 105 
 9lm − 13l 2 m 13lm 11l 2 m 
0 0 0 0 0 0 0 0 
 70 420 35 210 
 13l 2 m − l 3m 11l 2 m l 3m 
 0 0 0 0
420 140
0 0 0 0
210 105 

(19)
 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 
 
0 − 26lmΩ 11l 2 mΩ − 9lmΩ − 13l mΩ 
2
0 0 0 0 0 0 0
 35 105 35 210 
 − 11l 2 mΩ 2l 3 mΩ − 13l 2 mΩ − 1l 3 mΩ 
0 0 0 0 0 0 0 0 
 105 105 210 70 
0 26lmΩ 11l 2 mΩ 9lmΩ − 13l 2 mΩ 
0 0 0 0 0 0 0
 35 105 35 210 
 − 11l 2 mΩ − 2l 3 mΩ − 13l 2 mΩ 1l 3 mΩ 
0 0 0 0 0 0 0 0 
[c] = 0 0
105
0
105
0 0 0 0 0
210
0
70
0 0 0

 
0 0 0 0 0 0 0 0 0 0 0 0 
 − 9lmΩ 13l 2 mΩ − 26lmΩ − 11l mΩ 
2
0 0 0 0 0 0 0 0
 35 210 35 105 
 13l 2 mΩ − 1l 3 mΩ 11l 2 mΩ 2l 3 mΩ 
0 0 0 0 0 0 0 0 
 210 70 105 105 
0 9lmΩ 13l mΩ
2
26lmΩ − 11l 2 mΩ 
0 0 0 0 0 0 0
 35 210 35 105 
 13l 2 mΩ 1l 3 mΩ 11l 2 mΩ − 2l 3 mΩ 
0 0 0 0 0 0 0 0 
 210 70 105 105 

(20)

[k ] = [k1] + [k 2] (21)

 AE − AE 
 l 0 0 0 0 0 0 0 0 0 0 
l
 GJ − GJ 
 0 0 0 0 0 0 0 0 0 0 
 L L 
 0 5040 IE 2520 IEl − 5040 IE 2520 IEl 
 0 0 0 0 0 0 0 
420l 3 420l 3 420l 3 420l 3
 2520 IE 1680 IEl − 2520 IE 280 IEl 
 0 0 0 0 0 0 0 0 
 420l 2 420l 2 420l 2 140l 2 
 0 5040 IE − 2520 IEl − 5040 IE − 2520 IEl 
 0 0 0 0 0 0 0
420l 3 420l 3 420l 3 420l 3 
 − 2520 IE 1680 IEl 2520 IE 280 IEl 
 0 0 0 0 0 0 0 0 
[k1] =  − AE 420l 2 420l 2
AE
420l 2 140l 3 
 0 0 0 0 0 0 0 0 0 0 
 l l 
 − GJ GJ 
 0 0 0 0 0 0 0 0 0 0 
 L L 
 0 − 5040 IE − 2520 IEl 5040 IE − 2520 IEl 
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 2520 IE 280 IEl − 2520 IE 1680 IEl 
 0 0 0 0 0 0 0 0 
 420l 2 140l 2 420l 2 420l 2 
 0 − 5040 IE 2520 IEl 5040 IE 2520 IEl 
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 − 2520 IE 280 IEl 2520 IE 1680 IEl 
 0 0 0 0 0 0 0 0 
 420l 3 140l 3 420l 2 420l 2 
(22)
 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 
 
0 − 156l 4 mΩ 2 − 22l 5 mΩ 2 − 54l 4 mΩ 2 13l 5 mΩ 2 
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 − 22l 4 mΩ 2 − 4l 5 mΩ 2 − 13l 4 mΩ 2 l 5 mΩ 2 
0 0 0 0 0 0 0 0 
 420l 2 420l 2 420l 2 140l 2 
0 − 156l mΩ
4 2
22l mΩ
5 2
− 54l 4 mΩ 2 − 13l mΩ 
5 2
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 22l 4 mΩ 2 − 4l 5 mΩ 2 13l 4 mΩ 2 l 5 mΩ 2 
0 0 0 0 0 0 0 0 
[k1] = 0 420l 2 420l 2 420l 2 140l 2 
 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 
 
0 − 54l 4 mΩ 2 − 13l 5 mΩ 2 − 156l 4 mΩ 2 22l 5 mΩ 2 
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 13l 4 mΩ 2 l 5 mΩ 2 22l 4 mΩ 2 − 4l 5 mΩ 2 
0 0 0 0 0 0 0 0 
 420l 2 140l 2 420l 2 420l 2 
0 − 54l mΩ
4 2
13l mΩ
5 2
− 156l 4 mΩ 2 − 22l 5 mΩ 2 
0 0 0 0 0 0 0
 420l 3 420l 3 420l 3 420l 3 
 − 13l 4 mΩ 2 l 5 mΩ 2 − 22l 4 mΩ 2 − 4l 5 mΩ 2 
0 0 0 0 0 0 0 0 
 420l 2 140l 2 420l 2 420l 2 

(22)

The displacement vector in the matrix equation is

{q} = [ux1 θ1 uy1 β1 uz1 γ1 ux2 θ2 uy 2 β2 uz 2 γ 2]T (23)

Then unbalancing force vector is

{q} = [0 0 Qy1 Qβ1 Qz1 Qγ1 0 0 Qy 2 Qβ2 Qz 2 Qγ 2]T (24)

Where, considering the eccentricity e constant along the shaft

Qy1 = elmΩ 2 cos( α ) / 2 (25)
2 2
Qβ1 = el mΩ cos( α ) / 12 (26)
2
Qz1 = elmΩ sin( α ) / 2 (27)
2 2
Qγ1 = −el mΩ cos( α ) / 12 (28)
2
Qy 2 = elmΩ cos( α ) / 2 (29)
2 2
Qβ2 = −el mΩ cos( α ) / 12 (30)
2
Qz 2 = elmΩ sin( α ) / 2 (31)
2 2
Qγ 2 = el mΩ cos( α ) / 12 (32)
A Fortran computer software was written, to assembly the shaft element matrices into global matrices,
and simulates the shaft response, for several dimensions, materials, angular velocities and loads. To obtain the
time domain response, the integration procedure of Newmark was adopted.
This software has been successfully tested taking a classical case as a reference. Comparison of the
results obtained with the analytical model and the numerical results reveal a very good agreement.

3. RESULTS

An analytical solution of a rotating elastic shaft symply supported at both ends has been presented by
Meirovitch, 1997. The free vibration solution was obtained in the time domain for the local rotating set of axes.
In this work, Meirovitch’s solution is used as a reference for the finite element model test. Therefore, some
conditions were estabilished for comparison. Since Meirovitch didn´t included unbalancing efects, the present
results are for free vibration.
A constant cross-section steel shaft 400mm long and 25mm diameter has been analysed. For the shaft
simply supported at both ends and the finite element model having 20 elements, it has been found that the first
natural lateral frequency of the non-rotating shaft is 1986.1 rad/s. For both models, results were obtained for the
rotating speed at 50%, 100%, 200%, and 400% of the first natural frequency. These results are presented in local
coordinates for the central node of the shaft, acording to Figures (4), (5), (6) and (7).

Analytical Result Numerical Result

2.00E-06 2.00E-06

1.50E-06 1.50E-06

1.00E-06 1.00E-06
uz DISPLACEMENT

uz DISPLACEMENT
5.00E-07 5.00E-07

0.00E+00 0.00E+00
-2.00E-06 -1.50E-06 -1.00E-06 -5.00E-07 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06 -2.00E- -1.50E- -1.00E- -5.00E- 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06
06 06 06 07
-5.00E-07 -5.00E-07

-1.00E-06 -1.00E-06

-1.50E-06 -1.50E-06

-2.00E-06 -2.00E-06

uy DISPLACEMENT uy DISPLACEMENT

Figure 4 – Local trajectory of the shaft for the rotating speed equal to 50% of the first frequency

Analytical Result Numerical Result

2.00E-06 2.00E-06

1.50E-06 1.50E-06

1.00E-06 1.00E-06
uz DISPLACEMENT

uz DISPLACEMENT

5.00E-07 5.00E-07

0.00E+00 0.00E+00
-2.00E-06 -1.50E-06 -1.00E-06 -5.00E-07 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06 -2.00E- -1.50E- -1.00E- -5.00E- 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06
06 06 06 07
-5.00E-07 -5.00E-07

-1.00E-06 -1.00E-06

-1.50E-06 -1.50E-06

-2.00E-06 -2.00E-06

uy DISPLACEMENT uy DISPLACEMENT

Figure 5 – Local trajectory of the shaft for the rotating speed equal to 100% of the first frequency
 Analytical Result Numerical Result

2.00E-06 2.00E-06

1.50E-06 1.50E-06

1.00E-06 1.00E-06
uz DISPLACEMENT

uz DISPLACEMENT
5.00E-07 5.00E-07

0.00E+00 0.00E+00
-2.00E-06 -1.50E-06 -1.00E-06 -5.00E-07 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06 -2.00E- -1.50E- -1.00E- -5.00E- 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06
06 06 06 07
-5.00E-07 -5.00E-07

-1.00E-06 -1.00E-06

-1.50E-06
-1.50E-06

-2.00E-06 -2.00E-06

uy DISPLACEMENT
uy DISPLACEMENT

Figure 6 – Local trajectory of the shaft for the rotating speed equal to 200% of the first frequency

Analytical Result Numerical Result

2.00E-06 2.00E-06

1.50E-06 1.50E-06

1.00E-06 1.00E-06
uz DISPLACEMENT

uz DISPLACEMENT

5.00E-07 5.00E-07

0.00E+00 0.00E+00
-2.00E-06 -1.50E-06 -1.00E-06 -5.00E-07 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06 -2.00E- -1.50E- -1.00E- -5.00E- 0.00E+00 5.00E-07 1.00E-06 1.50E-06 2.00E-06
06 06 06 07
-5.00E-07 -5.00E-07

-1.00E-06 -1.00E-06

-1.50E-06 -1.50E-06

-2.00E-06 -2.00E-06

uy DISPLACEMENT uy DISPLACEMENT

Figure 7 – Local trajectory of the shaft for the rotating speed equal to 400% of the first frequency
4. CONCLUSION

The present work presents a finite element model which allows a vibration analysis of a rotating shaft.
One can notice that the mass, stiffness and gyroscopic matrices for the rotating velocity iqual to zero, are those
classical matrices for the nonrotating case. The present results show additional matrix elements in the stiffness
matrix and the occurance of the gyroscopic matrix when rotation is considered. It also can be observed that the
angular velocity Ω does not affect the mass matrix. One particular aspect of the matrix equation of motion is the
anti-symmetric characteristic of the giroscopic matrix, which has to be taken into account for the correct solution
of the matrix equation.
For the purpose of corroborating the model, numerical results for some cases were compared to results
obtained from a analytical solution proposed by Meirovitch, 1997, which predicts the response of a constant
cross section of a rotating shaft symply supported at both ends. For comparision, four cases were devised. For
the same shaft, results were obtained for angular velocities Ω iqual to 50%, 100%, 200% and 400% of the first
natural frequency of the same nonrotating shaft. Local trajectory of the shaft for the four rotating speeds are
presented by Figures (4), (5), (6) and (7), considering both models. These results show a very good agreement
for the first three cases and a poorer match for the 400 % case.
One has to consider also the qualitative aspects of the results, which does not show any tendency of
resonance, as it should be, taking into account that the free motion of the shaft has been analysed.
A more general model for the rotating shaft coud also be considered, which is not the purpose of this
work. One can also considerer the shaft suporting discs of considerable inertia and the effect of a umbalancing
forces produced by these parts. At present, the model allows the consideration of a concentrated and
considerable mass in any point of the shaft, neglecting the rotating inertia. It can be done by simply altering the
geometrical parameters of the finite element at the point where the concentrated mass is to be considered.

5. ACNOWLEDGMENTS

The auhors ackowledge the support received for this work from Brazilian Aeronautic Technologic
Instituct (ITA).

6. REFERENCES

Meirovitch, L. “ Principles and Techniques of Vibrations”, Prentice Hall, United States, 1997

Meirovitch, L., “Elements of Vibrations Analysis”, McGraw-Hill, Singapore, 1986

Inman, D. J.,”Engineering Vibration”, Prentice Hall, United States, 1996

Gunter Jr, E. J., “Dynamic Stability of Rotor Bearing Systems” , NASA, United States, 1966

Brito, M. S. , "Um Modelo Vibratório de Eixos de Turbina", TG, ITA, 1999

Bismarck-Nasr, M. N., "Finite Elements in Applied Mechanic", Maher N. Bismarck-Nasr, Brasil, 1993

Sacramento V. F. D. e Menezes, J. C., “Análise por Elementos Finitos da Influência do Desbalanceamento na

Vibração Transversal de uma Viga de Euller-Bernoulli Rotativa” - 2º Congresso Temático de Dinâmica e
Controle da SBMAC, DINCOM 2003, São José dos Campos – SP, agosto de 2003