You are on page 1of 10

Back to main page

7th IFToMM-Conference on Rotor Dynamics, Vienna, Austria, 25-28 September 2006

FEM-Simulation and stability analyses of high speed rotor systems

Jonas Fischer, Jens Strackeljan

Institut fr Mechanik,
Otto-von-Guericke-Universitt Magdeburg,
D-39106 Magdeburg, Germany
jonas.fischer@mb.uni-magdeburg.de; jens.strackeljan@mb.uni-magdeburg.de

ABSTRACT

To obtain a safe design of high speed rotor systems it is essential to work with simulation tools like analytical
programs, multi-body-system software or FEM-software. One tool, which enables an individual adaptation for
special rotor systems due to its open source code is the FEM software FERAN. It was developed by the authors
and is introduced in this paper.
Simulation results of high speed lab centrifuges, as an example for a challenging rotor system, are presented
in this paper. The effects of internal damping and Coulomb friction are investigated via time integration of a
run-up. Furthermore the active modification of mechanical properties to alter the natural frequencies during run-
ups were computed with FERAN.

KEYWORDS
High Speed Rotor Systems, FEM-Simulation, Stability, Internal Damping, Nonlinearities

1 INTRODUCTION
Often advanced high speed centrifuges are designed as universal devices where one motor drives a wide
range of different rotors. Large-volume rotors with masses up to 20 kg, for particular blood bank centrifuges
even 60 kg are used in laboratories. In general high speed centrifuges operate with speeds up to 30000 rpm.
Efforts are undertaken to obtain a huge tolerance of unbalance, for avoiding time consuming taring of samples.
Auto-balancing devices or drive concepts, where the mechanical concept allows high unbalances are one
solution.
In general the development of new concepts for high speed centrifuges and the improvement of existing high
speed centrifuges require simulation tools which offer the possibility of special adaptation.
The FEM program FERAN, which was developed during the last years at the University Clausthal and
Magdeburg is well suited to determine eigenvalues, oscillations, stresses and bearing loads of rotor systems.
Due to the open program code, FERAN can be linked with algorithms to conduct automated parameter studies
or optimization. High speed rotor systems are governed by many parameters like bearing stiffness and damping,
the length and width of the shaft and the moments of inertia, which can be altered to enhance the dynamic
properties.
Stability analysis are essential for a safe design and no serious manufacturer should offer a lab centrifuge
without experimental stability tests. The effects of internal damping and friction between rotating parts, which
occurs when rotating parts are slackly attached to each other can provoke self excited vibrations. Those
phenomena can be simulated by an dynamic simulation.
Nonlinearities in rotor systems are caused e.g. by clearance in ball bearings, oil films in a journal bearing or
dry friction between the shaft and mounted elements. Some of these nonlinearities can be included in the

1 Paper-ID 184
dynamic simulation with FERAN. The effect of Coulomb damping is such an nonlinearity, which is discussed
later on.
In order to facilitate the passing of critical speeds with high speed centrifuges, the active change of natural
frequencies during run-up by diminishing the bearing stiffness were simulated with FERAN. The results
introduced at the end of the paper give a first glance.

2 THEORETICAL BACKGROUND
In the following the theoretical background of the FEM-software FERAN which is coded in C++ will be
presented only compendiously.

2.1 The fundamental equations


The dynamic behaviour of a rotor system is described by the discrete differential equation in the inertial
coordinate system.
&&} + [[C ] + [G ]]{D& } + [ K ]{D} = {F ext } + {F nl } ,
[ M ] {D (1)

with stiffness matrix [K], damping matrix [C], gyroscopic matrix [G] and mass matrix [M]. Vector {D}
consists of the nodal displacements, vector {Fext} of the linear external and {Fnl} of the non-linear internal and
external forces.
To obtain equation (1), the principle of linear momentum is first stated in a body fixed coordinate system and
subsequently transformed in the inertial coordinate system. The internal forces {Fint}=[K]{D} in a coordinate
system rotating with the frequency of revolution are transformed into the inertial coordinate system.

The shaft of the rotor system is discretised by beam elements. Beam elements can reflect deformations due to
bending loads, torque loads and axial forces. For the shape function of axial deformations, a first order
polynomial is used. The bending deformations are described by cubic shape functions including transverse shear
deformation optionally, according to Timoshenkos beam theory ([2]). The shape functions of torque loads are
computed in reference to Saint-Venants theory.

In FERAN the mass properties of the beam element are described by a consistent mass matrix. The consistent
mass matrix has the advantage, that it is positive definite. Thus problems which occur with semi- or indefinite
matrices during eigenvalue calculations are avoided. The drawback of consistent mass matrices is, that the
efforts to handle them are much greater than with lumped mass matrices.
Damping that influence structural dynamics can be categorized in viscous damping, hysteresis damping or
Coulomb damping. Of these kinds of physical damping only viscous damping is easy to represent in dynamic
equation. If Coulomb damping is included, the dynamic equation gets nonlinear. Viscous damping, Coulomb
damping and to some extent hysteresis damping can be considered in simulations with FERAN.

2.1.1 The gyroscopic matrix


If high revolutions and huge deflections occur in a rotor system, and if the shaft incorporates rotors with
a huge polar mass moment of inertia, it is necessary to take into account gyroscopic effects. Due to this effects
the natural frequencies change as a function of the rotational speed . The forward whirling mode increases as
the rotational speed increase, the backward whirling mode decrease. Resonance occurs with unbalance loads,
when rotational speed and natural frequency of the forward whirling mode are equal.
If we assume a disk mounted on an elastic shaft with the inclination angles {D} = {yS zS }T as degrees of
freedom, the moments in respect to the mass centre of the rotor are calculated by transforming the Eulerian
gyrokinetic equation from a rotor fixed coordinate system to the inertial coordinate system:

M yS J AS 0 &&yS 0 J PS &yS
= + (2)
M zS 0 J AS &&zS J PS 0 &zS

JAS is the axial mass moment of inertia of the rotor. The gyroscopic matrix, which consist of the polar mass
moment of inertia JPS , effects that the degrees of freedom are coupled. In FERAN, for each element, the axial
mass moment of inertia is located in the mass matrix. The gyroscopic matrix can be stated as lumped or as an
consistent matrix, similar to the mass matrix. In FERAN the former method was used.
If an rigid rotor is considered in the FEM model, the properties are allocated to the corresponding node of the
shaft.

2 Paper-ID 184
2.1.2 Boundary conditions and constraints
Each node of the used beam element consists of 6 degrees of freedom. If the degrees of freedom of any node
represent motions, which are prevented by supports, the corresponding rows and columns in the global matrices
have to be eliminated.
The developed FEM-software can take into account different bearings with different stiffness and damping
properties. At each node of the shaft, two spring elements and two dampers in each rotational and translational
direction can be included.
The upper and lower shaft bearing of centrifuges are often coupled rigidly by the stator. Under consideration
of constraints, which connect the movements of the corresponding nodes, the behaviour of the stator can be
simulated.

2.2 Calculation of eigenvalues


As an approach for the homogenous solution of the differential equation (1) the function
{Dh (t )} = {D h }e pt (3)

is chosen. By introducing this into equation (1) and with some calculations we obtain the standard
eigenproblem
([ K U ] p [ I ]){D U } = {0} (4)

with the unit matrix [I]. The matrix [KU] is an asymmetrical matrix. During the last years, different solvers
for eigenproblems were developed, like the Jacobi method, the Householder-QR method, the determinant search
method, the Sturm sequence, the inverse iteration or the Lanzcos method. The suitability of an algorithm for a
given task depends on the size and structure of the matrices, the numbers of eigenvalues which should be
considered and how skilfully algorithms are coded. As the rotor system is described by few elements, the arrays
of the eigenproblem are comparative small. Thus a method which calculates all eigenpairs can be used. The
matrix [KU] is asymmetric. Hence the application of the Householder-QR method is expedient. In FERAN a
readily modified Householder-QR algorithm of Wilkinson ([16]) is implemented.

2.3 Steady state calculation of deflections, forces and stresses


In the steady state case one considers vibrations that continues at constant amplitude and at the frequencies of
the forcing function after initial transients have disappeared because of damping. To compute this harmonic
response of the rotor system an unbalance load {Fiext} with constant speed of revolution is applied to the structure
at node i:
{Fi ext } = mU rU 2 {0 cos(t + ) sin(t + ) 0 sin(t + )lU cos(t + )lU }
T
(5)

The load is splitted in real and imaginary part. Assuming


{D p (t )} = {D p }e pt (6)

as solution for the inhomogeneous part of the linear equation of motion (1), one gets
[ K ( j)]{D p } = {F } . (7)

[K(j)] is the complex dynamic stiffness matrix. By separating imaginary and real part, a system of
equations is computed, which can be solved by the Gauss algorithm ([11]). Nonlinearities can not be considered
by this calculation.
Forces are computed by equation (7), whereby the rows and columns deleted due to the boundary conditions
have to be added to the complex dynamic stiffness matrix [K(j)].
The calculation of stresses is carried out regarding the continuous deformations according to the basic
theories of torsion, beams and bars.

2.4 Numerical Integration of the equation of motion


The response history of the rotor-system due to unbalance loads can be calculated with direct time integration
methods using finite difference approximations of time derivatives. In FERAN different explicit and implicit
methods are applied.

3 Paper-ID 184
The trapezoidal rule is used in FERAN as an implicit method ([1]). This method is unconditional stable, but
it can not be used for nonlinear systems unless in each step the nonlinear equation
[ K eff ] n +1 {D}n +1 {F nl ({D}n +1 )} = h({D}n ,{D& }n ,{D
&&} )
n
(8)

is solved by a Newton algorithm to obtain {D}n+1. The matrix [Keff] consists of


1
[ K eff ]n +1 = [ M ]n +1 + [C ]n +1 + [ K ]n +1 (9)
t 2 t
and h({D}n,{dD/dt}n,{dD/dt}n) is a function depending on the degrees of freedom of time step n, the
matrices [C]n+1 and [M]n+1 and the external linear force vector {fext}n+1 at time step n+1.
The 3rd order Runge-Kutta method used in FERAN is suited to solve nonlinear problems but the solution
becomes unstable, if the structure has huge eigenfrequencies even if Runge-Kutta methods of higher order are
used. As a simple rule, time steps used for calculation should not exceed t = T/, with 1/T the uppermost
eigenfrequency.

If the structure consists of huge eigenvalues the system of differential equations that has to be solved is stiff.
Efficient integration methods for stiff problems should be implicit in character to obtain the required stability
properties. Beneath the implicit-direct-integration methods, modified Rosenbrock methods can be used. They are
best suited when the differential equation of motion is nonlinear because the Newton iteration in each step, to
solve equation (8) is avoided.
One algorithm used in FERAN is the linear implicit Runge-Kutta method, introduced in Steihaug and
Wolfbrandt ([12]).

3 EXAMPLES
3.1. High speed lab centrifuge
At the Institute of Applied Mechanics, University Magdeburg, measurements were carried out on different
high speed centrifuges. In addition these centrifuges were simulated with the developed software FERAN. The
results for one of those centrifuges are presented in the following.

The measured eigenvalues coincide well with the simulation results of FERAN (figure 1). In the simulation
model the connection between roller bearing and shaft is considered to be stiff. In reality, the shaft bearing
connection is softer. To take into account the effects of the actual softer connection, the shaft diameter has to be
reduced in the simulation model in respect to the original diameter. Regarding the high speed lab centrifuge a
shaft diameter reduction of 15 % leads to a good compliance between measurement and simulation results.

350

300
eigenfrequency [Hz]

FERAN simulation
250
measurement
200

150

100

50

0
0 25 50 75 100 125 150 175 200
frequency of revolution [Hz]
Figure 1: Campbell diagram of the high speed centrifuge. Simulation results of FERAN in comparison with
measurement values.

4 Paper-ID 184
Figure 2: FEM-simulation model of a high speed centrifuge.

Figure 2 shows the FEM-simulation model of the centrifuge. The shaft is modelled with seven elements. The
rigid rotor is attached to node 8 as well as the mass of unbalance mu with the radius of unbalance ru. The bearing
nodes L12 and L22 are connected with constraints. They represent the stator. The spring elements k12 and k22 and
the damping elements c12 and c22 specify the properties of the roller bearings. The foot springs of the stator are
modelled with two torsion springs. In the picture, only one of them, torsion spring k11 is displayed. Furthermore,
damping properties are taken into account by additional damping elements.

The parameters used for the simulation are: shaft diameter 20 mm between node 1 and 2, 13 mm between
node 2 and node 7 and 22 mm between node 7 and 8, shaft length 98 mm (68 mm between node 1 and 2). Rotor:
axial moment of inertia JAS = 0.0063 kgm, polar moment of inertia JPS = 0.0104 kgm (both in respect to node
8), mass mS = 1.93 kg. Stator: moments of inertia JPStator JAStator 0.01 kgm, mass mStator = 5 kg. Foot springs:
stiffness k11t = 9000 N/m (translation), k11r = 180 Nm (rotation), damping c11t = 75 Ns/m, c11r = 0.06 Nms,
k21t = 150 N/m. Roller bearings: k12t =k22t = 6010 N/m. Unbalance load: mass mu = 0.003 kg radius ru = 80 mm.

The high speed centrifuge which was object of our research work is offered by the producer with a wide
range of different rotors. Once the model is created in FERAN, the behaviour of each rotor can be computed
with minor expense. Furthermore the effects of varying the geometrical, stiffness and damping properties of the
centrifuge can be studied easily.

3.2 Linear and nonlinear internal damping


The shaft rotor connection of high speed centrifuges operating with
different rotors is designed in such a manner that the connection is easily
released for changing the rotor and that the rotor is attached firmly during
operation. Nevertheless small motions can occur in-between of rotor and
shaft due to incorrect operation, manufacturing errors or constructive
errors. Thus damping or friction is inserted in the rotating system.
Internal damping is damping between the rotating parts of a rotor
system. It stabilises the system in the speed range lower than the major
critical speed and destabilizes it in the post critical range.
To show the effects of internal damping, one can regard a model with
damping elements between rotor and shaft (figure 3). The shaft diameter is
d = 6 mm, the length of the shaft is l = 60 mm, the mass of the rotor is mS
Fig. 3: model of internal damping. = 9.1 kg with an axial moment of inertia of J = 0.0219 kgm and a polar
AS
moment of inertia of JPS = 0.035 kgm. The mass of unbalance is mU =
0.01 kg with a radius of unbalance rU = 80 mm.
Concerning the damping force {Fid}in the rotating coordinate system, the following equation is stated:
{F id }' = ci {D& }' . (10)

indicates that the term is given in a rotating coordinate system. A transformation in the inertial coordinate
system and introducing this in equation (1) results in:

5 Paper-ID 184
&&} + [[C ] + [C ] + [G ]]{D& } + [[ K ] + [C~ ]]{D} = {F ext } + {F nl }
[ M ] {D (11)
a i i

~ ~
with [Ci ] a skew-symmetric matrix. Thus the matrix [[ K ] + [C i ]] changes its properties with altering
speed of revolution . Equation (11) is used for the calculations in FERAN. Parameters for the internal damping
are cit = 638 Ns/m in translational and cir = 0.57 Nms in rotational direction and for the outer damping cat = 250
Ns/m and car = 0.23 Nms.

160 40

damping of forw. whirl 1 [%]


backward whirl 1
without internal
140 forward whirl 1 30 damping
eigenfrequency [Hz]

120 backward whirl 2 with internal


forward whirl 2 20 damping
100
80 run-up 10
60 0
40
-10
20
0 -20
0 20 40 60 80 0 10 20 30 40 50 60 70 80
frequency of rev. [Hz] frequency of rev. [Hz]
Figure 4: Campbell diagram of a rotor system with internal damping (left) and damping of the first forward
whirl (right).

In figure 4, one can see that the internal damping reduces the external damping after the first critical speed.
The sign of the damping becomes negative after a certain speed. Self excitation gets possible.
In figure 5 the results of a run-up of this rotor system are shown. The simulation was computed by a 3rd order
Runge-Kutta time integration with and without internal damping. It is obvious, that the destabilizing effects of
internal damping can only be observed, when a dynamic analyse is simulated over the whole speed range, e.g. a
run-up. The evaluation of the steady state deflections is added for comparison.
The magnitude of the rotor displacement in figure 5 and followings is calculated by v(t ) 2 + w(t ) 2 with
displacements v(t) and w(t) transverse to the shaft direction. Hence one can observe the beat frequency in the
case of internal damping in the overcritical speed range in figure 5. The beat frequency is the difference between
frequency of revolution and first forward whirl, the frequencies of rotors displacements v(t) and w(t).

time [s]
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5
0,7
magnitude of rotor displacement [mm]

with internal damping,


0,6 steady state

with internal damping, time


0,5
integration

without internal damping,


0,4
time integration

0,3

0,2

0,1

0
0 10 20 30 40 50 60 70 80
frequency of revolution [Hz]
Figure 5: Run-up of a system with internal damping compared to one without internal damping.

6 Paper-ID 184
Nonlinearities in rotor dynamic systems can be analyzed with FERAN, too. As an example, the effects of dry
friction which can be approximated with Coulomb damping are simulated.

We assume that the dry friction acts between an elastic shaft and a rotor, which is slackly attached to the
shaft. If the elastic shaft whirls with an angular velocity that is different to the speed of revolution of the system,
the fibres of the shaft surface are compressed and stretched. Relative movement emerges. An elastic restoring
force works from the tension side to the compression side. Dry friction derogates the motion of the fibres. Thus
additional compression and tension occurs, which leads to an additional force vector {Fcd} in the direction from
tense to compressed surface, according to a model of Kimball and Loewe. In the post critical speed range the
force vector points in the same direction as the whirling motion and acts as a destabilising force.
The following equation can be used to describe this Coulomb damping force {Fcd} in the rotating coordinate
system ([17]):
{D& }' (12)
{F cd }' = [ Fc 0 ]
| {D& }' |

For a 4 degrees of freedom rotor with {Dt} = {v(t) w(t)}T as rotors deflections and {Dr} = {y(t) z(t) }T as
rotors inclination angles splitting equation (12) in a translational and a rotational part is expedient.
{D& t }' {D& r }'
{Ft cd }' = Fct 0 {Frcd }' = Fcr 0 (13)
| {D& t }' | | {D& r }' |
Transforming equation (13) into the inertial coordinate system yields for the translational Coulomb damping
force transverse to the shaft direction Fycd for example:
v (t ) + w(t )
Fycd = Fct 0 (14)
| (v (t ) + w(t ) ) 2 + ( w(t ) v (t ) ) 2 |
The Coulomb damping force {Fcd} is introduced in the right hand side of equation (11). The identical model
as for the simulation of linear internal damping (figure 5) with outer damping cat = 250 Ns/m and car = 0.23 Nms
is used to obtain figure 6, whereas linear internal damping is not considered. Values for the parameters Fct0 and
Fcr0 are difficult to determine in praxis. The values used for the simulation were only assumed.
time [s]
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5
1,2
magnitude of rotor displacement [mm]

steady state
1

without coulomb
damping
0,8
Fct0 = 5.2 N

0,6 Fct0 = 20.8 N

Fct0 = 41.6 N
0,4

0,2

0
0 10 20 30 40 50 60 70 80
frequency of revolution [Hz]
Figure 6: Effects of Coulomb damping during a run-up of a rotor system with Fcr0=0.

After passing the first critical speed at 33.4 Hz the rotor oscillates in figure 6 with the frequency of the first
forward whirl due to Coulomb damping added by a vibration with the frequency of revolution, similar to the case
of linear internal damping. But the effects of Coulomb damping occur immediately after passing the critical
speed. The magnitude of rotor displacement does not increase with increasing speed of revolution.

7 Paper-ID 184
3.3 Alternating spring stiffness during run-ups
In particular auto-balancing is an opportunity to create rotor systems with high tolerance of unbalance in the
post-critical speed range. One drawback of auto-balancing are high deflections during the passage of critical
speeds ([9]).
Researches are started to diminish this deflections not only in auto-balancing devices by changing the
mechanical properties specifically during run-ups e.g. by increasing damping values or by diminishing the
bearing stiffness. First calculations of run-ups of the centrifuge mentioned in paragraph 3.1. with altering
stiffness properties are undertaken with FERAN. A verification of the simulation results with measurements has
not been accomplished yet but is in preparation.

The effects of an incentive reduction of the bearing stiffness to 10 % of the original value in the post-critical
speed range are demonstrated in figure 7. An overshoot occurs as the bearing stiffness preventing this motion is
suddenly diminished. The first backward whirl is the natural frequency in this speed range with the lowermost
damping (figure 8, right). As a result, an excitation of the first backward whirl can be clearly observed in figure
7, with decreasing frequency from 0,98 Hz to 0,68 Hz. This coincides with the Campbell diagram (figure 8, left).
The steady state deflection of the rotor is shown by the dashed line in figure 9. The steady state deflection
reduces about 15 % due to the change of the stiffness parameters.
time [s]
0 2 4 6 8 10 12 14
0,25

0,2 steady state


stiffness shift
0,15
rotor displacement v(t) [mm]

time integration
0,1

0,05

-0,05

-0,1

-0,15

-0,2

-0,25
0 20 40 60 80 100 120 140
frequency of revolution [Hz]
Figure 7: Run-up of a centrifuge. The bearing stiffness is reduced in a single step from 100 % to 10%. Steady
state simulation as reference.

To avoid excitations due to incentive reductions of the stiffness, a continuous reduction of the bearing
stiffness is to aspire. This can be carried out theoretically with an arc tangent function.
0 .9
s ( ) = arctan( 0.1 ( f as )) + 0.55 (15)
2
The frequency fas is defined as an average switching frequency. At the average switching frequency fas the
derivation of the arc tangent function gets maximal. In further simulations function s() of equation (15) was
multiplied with the bearing stiffness reducing it smoothly from 100 % below fas to 10 % above fas. Thus altering
of the stiffness properties in the post critical speed range does not lead to those intense vibrations.
For the time integration of the stiffness shift in figure 10 an average switching frequency of fas = 3.98 Hz was
chosen. This value is slight below the first critical frequency in the unchanged configuration (figure 9, right).
Applying this value and equation (15) a reduction of the maximum rotor deflection up to about 30 % was
achieved in the calculations compared to a simulation with an unchanged bearing stiffness of 100 %. The impact
to the natural frequencies in respect to the unchanged case (figure 9, left) can be observed in figure 9, right.
Both the trapezoidal rule and the linear implicit Runge-Kutta method were used to obtain figures 7 and 10.

8 Paper-ID 184
30 60
backward whirl 1
forward whirl 1
25 50
eigenfrequency [Hz]

backward whirl 2
forward whirl 2

damping [%]
20 run-up 40 backward whirl 1
forward whirl 1
15 30 backward whirl 2
forward whirl 2
10 20

5 10

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
frequency of rev.[Hz] frequency of rev.[Hz]
Figure 8: Campbell diagram (left) and damping (right) of the centrifuge with a reduction of the bearing stiffness
in a single step from 100 % to 10 % in the post critical range.

30 30
backward whirl 1 backward whirl 1
forward whirl 1 forward whirl 1
eigenfrequency [Hz]

eigenfrequency [Hz]
25 25
backward whirl 2 backward whirl 2
20 forward whirl 2 20 forward whirl 2
run-up run-up
15 15

10 10

5 5

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
frequency of rev. [Hz] frequency of rev. [Hz]
Figure 9: Campbell diagram of the reference centrifuge with 100 % bearing stiffness (left) and of the one with
reduction by an arc tangent function in the speed ranger lower than the first critical speed (right).

time [s]
0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5
0,25
magnitude of rotor displacement [mm]

steady state, const. stiffness 100 %


0,2
steady state, const. stiffness 10 %

0,15 constant stiffness

stiffness shift

0,1

0,05

0
0 10 20 30 40 50 60 70 80 90 100
frequency of revolution [Hz]
Figure 10: Run-up of a centrifuge with constant bearing stiffness compared to a run-up with diminishing bearing
stiffness over the speed range. Steady state simulation of 10 % and 100 % constant bearing stiffness as reference.

9 Paper-ID 184
4 CONCLUSION
During our investigations on high speed centrifuges FERAN turned out to be an expedient tool for simulating
eigenfrequencies, steady state deflections, bearing loads and dynamic run-ups. The open code offers a couple of
possibilities to integrate further features like system optimisation ([3], [15]).
Efforts are in preparation to approve the calculated phenomena of linear and nonlinear internal damping and
the computed effects of altering bearing stiffness during run-ups via measurements. The improvement of high
speed centrifuges by active altering of damping and stiffness properties during run-ups in combination with auto-
balancing devices are tasks for further research activities.

REFERENCES
[1] Cook, Malkus, Plesha, Witt (2002): Concepts and applications of Finite element analysis. 4th edition,
John Wiley & Sons, Madison.
[2] Cowper (1966): The Shear Coefficient in Timoshenko's Beam Theory. ASME Journal of Applied
Mechanics, vol.33 No 2, pp.335-340.
[3] Feng, Strackeljan(2002): A pareto Optimization with EA in Application of Centrifuge Design.
Intelligent Technologies: Theory and Applications, Ed. P. Sincak, IOS Press, 2002.
[4] Fischer (2005): Erstellung eines FEM-Programms zur Schwingungsberechnung von Rotorsystemen.
Diplomarbeit, Institut fr Mechanik, Clausthal.
[5] Gasch, Nordmann, Pftzner (2002): Rotordynamik. 2. Auflage, Springer Verlag, Berlin Heidelberg New
York.
[6] Gldner (1991): Hhere Festigkeitslehre - Band 1: Grundlagen der Elastizittstheorie. 3. Auflage,
Fachbuchverlag Leipzig.
[7] Link (1984): Finite Elemente in der Statik und Dynamik. Teubner Verlag, Stuttgart.
[8] Magnus, Popp (1997): Schwingungen. 5. Auflage, Teubner Verlag, Stuttgart.
[9] Ryzhik, Sperling, Duckstein (2004): Non-Synchronous Motion Near Critical Speeds in a Single-Plane
Auto-Balancing Device. Technische Mechanik, Band 24, Heft 1.
[10] Schfer (1994): Erhhung der Laufsicherheit schnelldrehender Rotorsysteme. Dissertation, Technische
Universitt Clausthal.
[11] Schwarz (1993): Numerische Mathematik. Teubner Verlag, Stuttgart.
[12] Steihaug, Wolfbrandt (1979): An Attempt to Avoid Exact Jacobian and Nonlinear Equations in the
Numerical Solution of Stiff Differential Equations. Mathematics of Computation, vol. 33, no. 146, pp.
521-534.
[13] Stelzmann, Groth, Mller (2000): FEM fr Praktiker - Band 2: Strukturdynamik. Expert Verlag,
Renningen.
[14] Strackeljan (2002): Einsatzmglichkeiten von Softcomputing- Methoden zur Auslegung, Optimierung
und berwachung von Rotorsystemen. Habilitationsschrift.
[15] Strackeljan, Feng, Behr (2001): Design Optimization of High Speed Laboratory Centrifuges using
Evolutionary Algorithms. ISCORMA-1, Proc. 1th. Conference on Stability Control of Rotating
Machinery, South Lake Tahoe.
[16] Wilkinson, Reinsch (1971): Lineare Algebra. Springer Verlag, Berlin Heidelberg New York.
[17] Yamamoto, Ishida (2001): Linear and Nonlinear Rotordynamics. John Wiley & Sons.

10 Paper-ID 184

You might also like