Politecnico di Milano

M.Sc. in Mechanical Engineering

Rotordynamics:
bending critical speeds and rotor balancing

Mechanical System Dynamics

Prof. Roberto Corradi
Technical issues in rotordynamics

❑ bending vibration and bending critical speeds

❑ rotor balancing
❑ dynamic instability in fluid-lubricated journal bearings
❑ condition monitoring and diagnostics
❑ etc.

2
Preliminary considerations

Rotating machines are normally designed to be balanced, but some residual

unbalance may result from small imperfections in the manufacturing process,
misalignment of supports, etc.
Any unbalanced rotor generates centrifugal forces, when the rotor itself is rotating
at a given angular speed W.
For fixed W, these forces are constant in the rotor’s local reference system (i.e.
they rotate together with the rotor), while they result in harmonic functions of time
if projected on a global (fixed) reference system. Since the circular frequency of
these harmonic functions is equal to the rotor’s angular speed, unbalance forces
are referred to as synchronous forces.

Centrifugal forces in the local

(rotating) reference system:

FcL () = m () W2 ()

and in the global (fixed) one:

FcG () = FcL () e jWt 3

Preliminary considerations

Unbalance forces are dangerous in that they result in dynamic loads on the
bearings, possible fatigue damages, vibration and noise. All these problems
become particularly serious in correspondence with bending critical speeds.
Any rotor’s shaft is flexible and therefore any rotor possesses bending natural
frequencies. As a consequence, a resonance condition may occur whenever the
angular speed W of the rotor equals one of the rotor’s bending natural frequencies.
These particular values of W are called bending critical speeds.
For all the reasons mentioned above, proper balancing techniques need to be
applied, that consist in mounting on the rotor additional masses (with proper value
and position), so as to minimize the overall rotor’s unbalance.
Note that balancing procedures depend on whether the rotor can be assumed to be
rigid or flexible. While in the previous case the rotating speed W is small compared
to the rotor’s first bending natural frequency w1, in the latter one a significant
contribution of the rotor’s flexible vibration modes is to be expected.
In other words, the same rotor can be assumed rigid or flexible depending on its
operating conditions.
4
Simplified analysis of the bending vibration phenomenon

The Jeffcott rotor (2 dof model) W

Rigid disc mounted in the middle of a massless
flexible shaft on simple rigid supports (damping
and gravitational forces are neglected).
Dynamic equilibrium equations in the x-y plane: m
 −mxG − kx = 0  48EJ 
  k = 
−myG − ky = 0  L3  z = x+ jy G
xG = x +  cos Wt xG = x − W2 cos Wt 

yG = y +  sin Wt yG = y − W2 sin Wt S
mx + kx = mW  cos Wt
 2

 z

 my + ky = mW 2
 sin Wt
Due to the polar symmetry of the shaft’s cross- O
section, the two equations of motion are decoupled
k
and the same natural frequency w is obtained from both equations: w= 5
m
Simplified analysis of the bending vibration phenomenon

The two scalar equations in the previous slide are equivalent to the following
vector equation: 10
damped
8
mz + kz = mW2 e jWt undamped

|z0/|
6
4
iWt
By imposing z = z0 e the system’s 2

response for variable W can be obtained: 0

0 0.5 1 1.5 2 2.5 3

z0 mW2 (W w ) 2 0

= =
 k − mW 1 − (W w )2
2 -50

deg
-100

-150
W=w → bending critical speed -200
0 0.5 1 1.5 2 2.5 3

W/w

6
Simplified analysis of the bending vibration phenomenon

Point S describes a circumference with radius z which is centred in the origin O of

the x-y plane (i.e. the intersection of the plane with the undeformed axis of the shaft).
Point S moves along this circumference with an angular speed W which is the same
as the angular speed of the disc (synchronous whirl).
W<w z is small and the two vectors z and  are aligned
W=w (critical speed) z is large and vectors z and  are orthogonal
W>w vectors z and  are parallel, with opposite directions; when W>>w they
tend to have the same magnitude and the c.o.m. G moves towards O,
thus reducing the eccentricity and the consequent unbalance force
7
(self-balancing)
Balancing of rigid and flexible rotors

Obviously, in real applications the actual unbalance of the rotor is not known.
Therefore, balancing techniques need to rely on experimental procedures which are
based on the idea of measuring some effect of the unbalance (vibration or loads
transmitted to the bearings) and to process the collected data so as to identify the
value and position of the balancing masses.
When dealing with real rotating machinery, balancing masses are installed at more
than one section along the rotor (balancing planes) and vibration/force
measurements are taken in a set of measurement sections.
The number of balancing planes and measurement sections depends on whether
the rotor can be assumed to be rigid or flexible.
We already know that for a rigid rotor to be balanced its center of mass G needs to
belong to the rotation axis and the same axis needs to be a principal axis (i.e. the
resultant force and the resultant moment of the inertia forces associated with the
rotor’s unbalance are both zero). In the case of a flexible rotor, this condition
is no more sufficient to avoid bending vibration.

8
Measurement procedures

This measurement procedure applies to both

vibration and force measurements

t  2
= T=
T 2 W
t T 9
Balancing of rigid rotors: equivalent system of forces

resultant force and resultant moment equivalent system of forces

of the inertia (centrifugal) forces (planes 1 and 2 can be arbitrarily selected)
associated with the rotor’s unbalance

 Fi = F1 + F2

 M i = l1  F1 + l2  F2
Note that all forces/moments are constant since W is fixed and the local (rotating)
reference system is considered here. 10
 Balancing of rigid rotors based on force measurements

d2
1 and 2 are rigid dynamometric supports d1

 R1 + R2 + Fi = 0
 → 2 equations in the 2 unknowns Fi and M i
 d1  R1 + d 2  R2 + M i = 0

11
Balancing of rigid rotors based on vibration measurements

Influence coefficients (2 balancing planes and 2 measurement sections):

V1(u +m1 ) − V1(u ) V2(u +m1 ) − V2(u )

11 =  21 =
m1 m1
V1(u +m2 ) − V1(u ) V2(u +m2 ) − V2(u )
12 =  22 =
m2 m2
Once that the influence coefficients have been identified, the two balancing masses
can be computed by imposing the annulment of the overall vibration resulting from
the contributions of both the rotor’s intrinsic unbalance and the added balancing
masses:


 1
V (u )
+  m*
+ 12 2 = 0
m*

 (u )
11 1
→ 2 equations in the 2 complex

 2
V +  m*
21 1 +  22 2 = 0
m * *
unknowns m1 and m2
*

12
Balancing of flexible rotors

In the case of flexible rotors, we need to make sure that the generalized unbalance
forces be zero for all the vibration modes which may undergo resonance excitation,
including those corresponding to a non-rigid motion of the rotor. Therefore, more
than two balancing planes are normally required (nb>2). Moreover, the number of
measurement sections is usually greater than two (nm>2), since vibration is
normally measured at each bearing. The balancing of flexible rotors can be
performed according to the method of the influence coefficients.
u +m j
Vi − Vi u
 ij = i = 1, 2, nm j = 1, 2 nb
mj
V +   m* = 0
u
Balancing condition:

 11 12 1n  V1u   m1* 

     
b

   
  =   u

 V = u m* =  * 
 ij Vi   mj 
 
 nm 1  nmnb  Vnu  mn* 
 m  b 13
Balancing of flexible rotors

In flexible rotors, the influence of unbalance changes with the rotation speed, since
at different speeds different vibration modes can be excited. Therefore, different
values of both the influence coefficient matrix and of the vector of measured
vibrations need to be considered (for the ns speeds of interest), while the vector of
the balancing masses shall be the same at any speed.
V u ( W ) +  ( W )  m* = 0
 1  1 
 → complex-valued system (usually overdetermined)
 ...
 u
( ) ( )
of nm x ns equations in nb unknowns
V Wns +  Wns  m* = 0
 V u (W )    ( W1 )  
 1
  
 V ( W ) 
u
  ( W2 )  
V s +  s  m* = 0  s  = 
u u
Vs = 2
 
   
 
V s +  s  m* = 
u
 u
V ( )
Wns 

   ( )
W 
ns  

 s   s  m* = −  s  V s
H H u

Least-squares minimization of the residual 

( )
−1
 =   = min) yields:
2
m* = −  s   s   s 
H H H u
(i.e. 14
Vs