Rotordynamic Modeling

and Analysis 6
6.1 Analytical Models and Essential Components
The analytical prediction of the rotor dynamic behavior and bearing performance depends
heavily on accurately modeling the physical system and understanding the assumptions
and limitations applied in the modeling and analytical tools employed. The modeling of
complicated rotating machinery, however, relies on sound engineering judgment and
practical experience. The modeling process transforms the complex physical system into
a representative, hopefully simple, mathematical model. The connections between the
physical system and the mathematical model must be understood well enough, so that the
results obtained from the mathematical analysis can be verified and fully utilized in the
design process. Figure 6.1-1 shows several industrial rotating machines, from simple
rotors to complex rotor assemblies with static structures, in simplified mathematical
models that allow for various rotordynamic analyses. The complete system under
consideration may contain both the rotating assembly and non-rotating structures. For
rotordynamic study, the primary interest is in the dynamics of the rotating component.
Therefore, the majority of the degrees-of-freedom (DOFs) under study is in the rotating
component; however, the importance of the flexible supports and soft foundation must be
considered and included in the model if necessary. The assumptions and simplifications
made in the component equations, which are then used to form the system governing
equations, must be fully understood to properly use these component equations and
interpret the analysis results.

341
342 CHAPTER 6: ROTORDYNAMIC MODELING AND ANALYSIS
 6.1 ANALYTICAL MODELS AND ESSENTIAL COMPONENTS 343

Figure 6.1-1 Mathematical models for various rotating machinery

A finite element “station” along the shaft, as shown in Figure 6.1-2, has six (6)
DOFs, three translational displacements (x,y,z) along the (X,Y,Z) axes and three
rotational displacements ( θ x ,θ y ,θ z ) about the (X,Y,Z) axes, with the Z-axis being the
spinning axis. The term station, rather than the conventional term node used in general
finite element analysis, is commonly used in the model of rotordynamics because of the
alternate meaning that node has in the vibration mode shapes of rotordynamics.
344 CHAPTER 6: ROTORDYNAMIC MODELING AND ANALYSIS

Figure 6.1-2 Coordinate system and DOFs at a typical finite element station

For lateral vibration, the motion of each finite element station is described by four
DOFs: two translational displacements (x,y) in the X and Y directions, respectively, and
two rotational (angular) displacements (θx,θy) about the X and Y axes, respectively. For
torsional vibration, the motion of each finite element station is described by a rotational
displacement (θ z ) about the spinning axis (Z). For axial vibration, the motion of each
finite element station is described by a translational displacement (z) along the spinning
axis (Z). Therefore, the motions at a typical finite element station are described by the
following displacement vectors:

T
Complete motion: q(6x1) = { x, y, z , θ x , θ y , θ z }
T
Lateral vibration: qL = { x, y, θ x , θ y }
Torsional vibration: qT = {θ z }
Axial vibration: q A = { z}

In the design of rotor systems, the lateral, torsional, and axial vibrations are generally
de-coupled and considered separately. In this text, mainly lateral vibration is addressed
since it is more design involved and critical in the design phase of rotating machinery.
Torsional and axial vibrations are commonly dealt with during the selection of drivers or
driven units and couplings, after all the individual equipment has been designed. For
integrally-geared rotating machinery, the lateral, torsional, and axial vibrations are
coupled together through the gear meshes and thrust collars. That is, the torsional and
axial excitations influence the lateral vibration and vice versa. Also, once the motions are
coupled, the system is no longer isotropic in lateral vibration. It can be highly asymmetric
due to introduction of the coupled lateral-torsional-axial vibration effects. Although this
kind of coupled analysis is available in some commercial rotordynamic software, an
understanding of the fundamentals is still the key to a successful design. In designing the
system, one should focus on each rotor assembly design based on the de-coupled
vibrations. Once all the rotor assemblies are designed, a detailed analysis with coupled
motions can be considered and the components can be fine-tuned to achieve a better
design if necessary. In this chapter, only the lateral vibration is considered. The coupled
lateral-torsional-axial vibrations are discussed in Chapter 8.
 6.1 ANALYTICAL MODELS AND ESSENTIAL COMPONENTS 345

A cutaway diagram and rotating assembly for a six-stage centrifugal compressor are
shown in Figure 6.1-3. The rotordynamics mathematical simulation model for this six-
stage compressor is presented in Figure 6.1-4. It contains four essential components in
the modeling of a typical rotor-bearing system: the rotating shafts with distributed mass
and elasticity, rotating disks, bearings, and the most common synchronous excitation –
mass imbalance. These four basic components are the most common ingredients in the
rotordynamic model.

Figure 6.1-3 Six-stage centrifugal compressor

Figure 6.1-4 Computer simulation model for rotordynamic analysis

6.1.1 Rotating Shaft Elements

A rotating shaft with distributed mass and elasticity is the most essential component in
the rotordynamics model. The rotating shaft is made up of numerous shaft segments with
various cross-sections called elements or sub-elements in the finite element formulation.
The most common types of rotating shaft elements are the cylindrical element with
constant diameters and the tapered (conical) element with linearly varied diameters along
the shaft axis. For other types of non-uniform cross-section elements, they can always be
approximately modeled with these two basic element types. For very complex elements,
which are difficult to model, the elemental matrices can be obtained from experiments.
Each element can possess several sub-elements and levels (layers), which allows for
346 CHAPTER 6: ROTORDYNAMIC MODELING AND ANALYSIS

reasonable flexibility in modeling shafts with geometric and material discontinuities, as

Figure 6.1-5 Rotating shaft elements

Figure 6.1-6 shows a finite element model for a high-speed compressor. It contains a
shaft with a bolted-on impeller and thrust collar. In the finite element formulation, the
DOFs at the finite element stations are the so-called active (or master) DOFs, which are
kept in the assembled equations of motion and variables to be solved directly, while the
DOFs at the internal sub-elements are considered the dependent (or slave) DOFs, which
are condensed out before the assembly process. The condensation technique commonly
used in finite element formulation is based on Guyan reduction (Guyan, 1965). Once the
displacements of the finite element stations (master DOFs) have been solved from the
assembled system equations of motion, the displacements of the sub-elements (slave
DOFs) can be easily calculated by utilizing the condensation matrix previously used in
the reduction procedure.
In the past, the use of sub-elements was strongly encouraged when modeling large
complicated rotor systems with limited computational power and memory storage. This
saved tremendous computational time with a minimal loss of accuracy in the results.
However, due to the rapid advances in computational speed and computer memory, the
advantage of using sub-elements is not as significant today as it was before, especially
when performing linear analysis. In the rotor preliminary design stage, the use of sub-
elements provides convenience when design modifications are anticipated because it
allows for only local modification without affecting the entire model or station numbers.
When performing nonlinear time-transient analysis with large complex rotor systems, the
use of sub-elements is still highly recommended due to the intense computational
requirements and numerical error accumulation in the large model.
 6.1 ANALYTICAL MODELS AND ESSENTIAL COMPONENTS 347

Figure 6.1-6 Finite element model for a high-speed compressor

Occasionally, there is a need to model the shaft element with different diameters for
the kinetic energy and strain (potential) energy calculations, as shown for element 2 sub-
element 2, and element 3 sub-element 1, in Figure 6.1-6. Due to the sudden change in
diameters, the entire diameter of the pinion section contributes kinetic energy (mass) but
not strain energy (stiffness). Therefore, the original geometric diameter is used for the
kinetic energy calculation, but a reduced diameter is used for the strain energy
calculation. Another example with different mass and stiffness diameters is the electrical
motor windings, which contribute kinetic energy but little strain energy.
The lateral motion of a rotating shaft element (or sub-element) can be described by
the displacements of the two end-points, and each end-point possesses four DOFs, as
shown in Figure 6.1-7. As stated earlier, each element can possess several sub-elements
and levels (layers), which allows for more flexibility in modeling systems with geometric
and material discontinuities. Each element can also be modeled with the different
diameters used in the kinetic and strain energy calculation. For the rotating shafts, only
the DOFs at the “stations” are kept and the DOFs at the sub-elements are condensed out
before the system assembly process.

Figure 6.1-7 Coordinates for a typical finite shaft element

In rotordynamics, the motions of the lateral vibration are very small in comparison to
the shaft diameter (i.e., on the order of 10-3). The small shaft deformation assumption is
practical and leads to linearized equations of motion for the rotating shafts and disks.
The equation of motion for a rotating shaft element (sub-element) with a constant
rotational speed Ω is:

M eqɺɺe + ΩG e qɺ e + K e q e = Q(8x1)
e
(6.1-1)

where q e contains the displacements at both the left and right ends of the element under
consideration. For purely lateral vibration, q e is an (8×1) displacement vector:

T
= ( qL | qR ) = ( xL , yL ,θ xL ,θ yL | xR , yR ,θ xR ,θ yR )
e T
q(8x1) (6.1-2)

If the torsional and axial vibrations are also considered in the analysis, the displacement
vector q e becomes a (12×1) vector:

T
x1) = ( q L | q R ) = ( xL , yL , z L , θ xL , θ yL , θ zL | xR , yR , z R , θ xR , θ yR , θ zR )
e T
q(12 (6.1-3)

Again, only the lateral vibration is considered in this chapter and the coupled lateral-
torsional-axial vibrations are discussed in Chapter 8.
The matrices Me, ΩGe, and Ke are the typical elemental mass/inertia, gyroscopic, and
stiffness matrices. The mass/inertia matrix Me, derived from the kinetic energy, is a
positive definite symmetric matrix; the conservative gyroscopic matrix ΩGe, derived from
the rotational kinetic energy, is a real skew-symmetric matrix; the stiffness matrix Ke can
be a general real matrix, which contains a symmetric elastic matrix derived from the
strain energy of the shaft element and the non-symmetric and non-conservative stiffness
from the axial torque and gravity for vertical rotors along the spinning axis. The force
vector Qe is the generalized force vector, which contains all the excitations acting at the
shaft element. The details of these matrices and generalized force vector are documented
in Nelson (1976, 1980) and Chen and Gunter (2005) and not repeated here.

6.1.2 Rotating Disks

A rotating component, which is either attached to the shaft or an integral part of the shaft
with relatively short axial length and large diameter (e.g., compressor impellers,
turbine wheels, thrust collars, balancing rings, couplings, and oil slingers), contributes
mainly kinetic energy with negligible strain energy; it can be modeled as a disk when
studying its effects on rotor dynamics. Due to the sudden change in the kinetic energy
distribution from a disk, the disk is always located at a finite element station in the shaft.
The motion of the disk is described by the motion of the shaft, to which the disk is
attached. Typically, the disk model falls into four types: rigid disk, flexible disk, offset
rigid disk, and offset flexible disk, as illustrated in Figure 6.1-8.
 6.1 ANALYTICAL MODELS AND ESSENTIAL COMPONENTS 349

Figure 6.1-8 Rotating disk models

For small vibrations, the lateral motion of a rigid disk, as shown in Figure 6.1-9, can
be described by four DOFs − two translational motions (xd,yd) in the X and Y directions
and two rotational (angular) motions (θxd,θyd) about the X and Y axes, respectively. The
kinetic energy of a spinning disk due to small vibrations ( xd , yd ,θ xd ,θ yd ) , excluding the
1 
pure spinning energy  I p Ω 2  , is given by:
2 

1 1 1
T=
2 2
( )2
(
md ( xɺd2 + yɺd2 ) + I d θɺxd2 + θɺyd2 + ΩI p θɺxdθ yd − θ xdθɺyd ) (6.1-4)

Figure 6.1-9 Displacements of a rigid disk

For a rigid disk, the disk is rigidly connected to the shaft and remains normal to the
shaft at all time during vibration, if disk skew does not exist. Therefore, rigid disk
motion can be described by the generalized displacements at the rotor (shaft) station, to
which the disk is attached, and no additional DOFs are introduced other than the existing
DOFs at the rotor station. The disk generalized displacements qd are the same as the
generalized displacements at the rotor station q, to which the rigid disk is attached.
T T
q d = ( xd , yd , θ xd ,θ yd ) = q = ( x, y,θ x ,θ y ) (6.1-5)

i.e., xd = x, yd = y, θ xd = θ x , θ yd = θ y
350 CHAPTER 6: ROTORDYNAMIC MODELING AND ANALYSIS

Hence, no additional DOF is introduced for a rigid disk. The governing equation of
motion for a rotating rigid disk (4DOF), with a constant rotational speed Ω in the fixed
reference frame, can be derived from the Lagrange’s equation:

M d qɺɺ + ΩG d qɺ = Q(4d x1) (6.1-6)

The (4×1) rigid disk generalized displacements are replaced by the generalized
displacements at the rotor station, to which the rigid disk is attached. The mass/inertia
matrix Md, derived from the kinetic energy, is a positive definite symmetric matrix; the
conservative gyroscopic matrix Ω Gd, derived from the rotational kinetic energy, is a real
skew-symmetric matrix that cross-couples the two rotational DOFs (θx, θy). Qd is the
generalized force vector, which contains all the excitations acting at the disk center of
mass and the gravity loading due to disk mass. Again, the derivations of the above
matrices and vector are well documented in previous publications (Nelson, 1976, 1980;
Chen & Gunter, 2005), and are not repeated here.
Figure 6.1-10 shows the disk motion for a rigid disk and a flexible disk. For a rigid
disk assumption, the disk is considered to be rigidly connected to the shaft and remains
normal to the shaft at all time during the vibration. However, for a flexible disk, the disk
generalized rotational displacements can be different from the rotor station to which it is
attached.

Figure 6.1-10 Rigid and flexible disk motions