HOT SPOTS IN TURBOEXPANDER BEARINGS: CASE HISTORY, STABILITY

ANALYSIS, MEASUREMENTS AND OPERATIONAL EXPERIENCE

Joachim Schmied Josef Pozivil Joseph Walch

DELTA JS, Switzerland Cryostar SAS, France Cryostar SAS, France

ABSTRACT synchronous vibration, it moves on the shaft surface, thus causing

Vibration induced hot spots in radial fluid film bearings can a phase angle and amplitude change of the synchronous vibration
cause spiral vibrations. The phenomenon is known as Morton in a polar plot. The vibration signal in the polar plot then has the
effect. Various authors have described the phenomenon of spiral appearance of a spiral.
vibrations caused by rubbing in general [1 - 4] and specifically Although the phenomenon is known for as long as more than
by the differential viscous shearing in fluid film bearings [5 - 8]. 80 years [1], it has taken quite a while until it became clear, that
The history of the case of a turbo-expander for the cryogenic the phenomenon can also be caused in oil lubricated bearings. In
industry is described. The investigated machine has a single 1987, it was suspected and some theoretical evidence was given,
relatively stiff rotor with two overhung impellers. The rotor is but the hot spot in the bearing was not measured [5]. The
supported on two tilting pad bearings, as they are commonly phenomenon in the bearing was clearly experimentally verified
used for these applications. It was designed in line with API 617 by temperature measurements on the rotor only in 1996 [6].
7th edition [11]. During internal factory testing excessive Further researchers then contributed to the understanding of the
vibration at high speed suddenly developed, in spite of very phenomenon [7, 8].
high damping and comfortable separation margins of the In [6], the phenomenon was observed in a compressor with a
relevant rotor vibration modes. The vibrations evolution in a large overhung coupling mass causing a typical overhang
polar plot had the appearance of a spiral, as they are typically vibration mode. Other machines prone to the phenomenon due to
observed in case of vibration induced hot spots. The tilting pad such vibration modes are gear compressor pinions, which have
bearings were suspected as the most likely cause for the hot overhung impeller masses. Several problems due to hot spots
spots. In parallel to the tests, hot spot stability analyses with a with this type of machine are described in [9]. In the conclusion
rotor dynamic model of the turbo-expander were carried out. of this paper, which is written by a rotating equipment specialist ,
The applied method, introduced by Schmied [5] allows the who has experience with many compressor vendors, it says:
handling of general rotor systems. The hot spot model is based ―There is solid evidence, that bearing effects can cause thermal
on the theory of Kellenberger [3] using a thermal equation spiral vibration, even with tilting pad bearings.‖ Gear compressor
between the shaft’s thermal deflection and the shaft pinions additionally have the effect of a load dependence of the
displacement at the hot spot location. The analytical results rotor dynamic behavior due to the varying mesh force causing
gave evidence of the instabilities and confirmed the fluid-film different bearing loads.
bearings as the source of the hot spot. The model was used to
assess changes for the solution of the problem. The following CASE HISTORY
measures were studied: Reduction of the oil viscosity, stiffening Cryogenic expansion turbines are used in air separation
of the rotor and reduction of the bearing width. The final plants, in the natural gas processing industry, for energy recovery
successfully implemented solution was a combination of lower in pressure let down stations as well as in organic Rankine or
bearing width and reduced viscosity. Apart from this thoroughly Kalina cycles.
studied case, an overview of further turbo-expanders regarding The dual expansion turbine/booster compressor loaded unit
their hot spot behavior is given. as shown in Fig. 1 is used to provide refrigeration for the tail gas
purification/treatment process in an ethylene processing plant.
The two turbines and the two compressor stages are operating in
INTRODUCTION series.
Hot spot induced spiral vibrations are observed in various
types of rotating machinery, e.g. compressors, water turbines,
steam turbines and turbo-generators. The hot spots are typically
generated by friction due to soft rubbing of the shaft to
stationary (non-rotating) parts. Such parts can be labyrinth
seals, seal rings, brushes and slip rings in electric machines and
even oil lubricated bearings. The hot spot on the shaft surface
arises if the rubbing is caused by a synchronous vibration. Since
the hot spot thermally bents the shaft and thus changes the

1
 Adjustable nozzles are located around the expander wheel to
control the gas flow by changing the flow area.
A 3D backwards lean blades compressor wheel is employed
as a loading device, to recover the mechanical energy from the
expander.
A vane diffuser is provided at the compressor wheel
periphery to optimize the booster head-flow characteristics and
to maximize its efficiency.
The stiff shaft is supported in the bearing carrier by two
tilting pad bearings with multi lobe axial face for the thrust load.
The radial bearings have 5 pads. They are loaded by the rotor
weight of about 120 kg, which yields a specific bearing load of
about 0.24 N/mm2. The load direction is on pad.
Figure 1. Dual turboexpander 2-TC 400/90 with the oil unit and The bearing support structure is relatively stiff compared to
control panel assembled in the workshop the oil film and can therefore be regarded as rigid.
The bearing carrier also includes the labyrinth seals and
The turbo-expanders in this case are very robust stiff shaft supports the expander nozzle assembly and booster diffuser. The
machines of proven design, for operation up to rather high wheels are connected to the shaft by a Hirth radial spur tooth
maximum continuous operating speed (MCOS) of 18’600 rpm. coupling and tightening bolt. The assembly of the bearing carrier
At this MCOS speed the journal bearing peripheral velocity is with nozzle assembly, diffuser and shrouds is called plug-in unit.
about 88 m/s. Expander and compressor housings are assembled on each side
During the internal testing of the turbo-expanders, the of the flanged bearing carrier (see Fig. 2).
measured shaft vibration started to rise suddenly just above the
nominal speed. The sudden vibration increase was identified as Expander Compressor
a spiral vibration as described in [5]. The bearings were
considered as the most likely source for a hot spot. Hot spot
stability analyses were performed to confirm these test findings,
and different remedies were considered for the machines based
on these analyses. Two described remedies as described below
were implemented and extensively tested at speeds up to 22’000
rpm (bearing journal velocity 104 m/s), a speed which is much
higher than the machine trip speed of 19’500 rpm.
After the successful internal testing the machines were
given free for factory acceptance test (FAT) witnessed by
customer. The performance and mechanical behavior of both
machines were accepted by the customer without reservations.
Figure 2. Cross section of the turboexpander TC 400/90
DESCRIPTION OF THE MACHINE
The expansion turbine is of the radial inflow type with
advanced aerodynamic design maximizing the isentropic
ROTOR DYNAMIC BEHAVIOR (ANALYSIS AND TEST)
efficiency. The machine is of a cartridge concept, which allows
a quick replacement of the machinery bearing carrier including
Calculated basic rotor dynamic behavior
the rotating parts without interference with the expander cold
The Campbell diagram (natural frequencies and damping
insulated process pipes.
ratios as a function of speed) of the shaft is shown in Fig. 3
together with the mode shapes of the forward whirling modes.
The speed range from the nominal speed to maximum continuous
speed in the diagram is grey shaded. There are two forward and
backward whirling rigid body modes below nominal speed with a
very high damping (damping ratio >20%). The forward whirling
1st and 2nd bending modes are above the speed range with a
comfortable separation margin. The backward whirling 1 st
bending mode coincides with the speed at the maximum
continuous speed. However, it cannot be excited by unbalance or
thermal bending.

2
 Test bed measurements of original configuration
During the internal performance test of the machines also the
mechanical machine behavior was tested up to the trip speed of
19’500 rpm, following the API 617 test requirements [11]. The
vibrations were monitored utilizing the Bently Nevada ADRE
system. Figure 5 shows the typical test bed computer mimic for
1xn the automated performance data monitoring and acquisition.

Figure 3. Campbell diagram and mode shapes

Although the shaft has two large overhung masses there are
no typical overhang modes thanks to the stiff design. The 1 st
bending mode has its largest deflection at the impellers; Figure 5. Test bed computer mimic
however, the vibration node is outside the bearing.
The response to an unbalance with a magnitude of 4 times The measured vibration of the tested machine had a
the API residual unbalance (165gmm) [11] is shown in Fig. 4. dominant synchronous (1xN) component. The ADRE vibration
The unbalance distribution as indicated in the figure excites the polar plot (all plots show only this 1xN component) had the
1st bending mode. The unbalance magnitude is based on the typical spiral vibration appearance, as shown in Fig. 6 for the
complete rotor weight, not just on the overhung portions as for MCOS speed of 18’600 rpm. The rotation direction of the spiral
overhang modes. The evaluation of the response according to vibration was forward (co-rotational). The time period for 360°
API is also shown in the figure. It can be seen, that the phase angle change was around 5 minutes.
separation margin (SM) of 55.5% is by far larger than the
required margin of 25.6%, which depends on the amplification
factor (AF) and damping, respectively. The unbalance
sensitivity of 0.437 (amplitude limit Al / amplitude in the
critical Ac) is smaller than the required 0.5, although it is not an
issue because of the large separation margin. The API
evaluation is automatically done in the program MADYN 2000
[10], which has been used for the analysis of the basic behavior.

Figure 6a. Spiral vibration in polar plot – expander side

Figure 4. Unbalance response plot with API evaluation Figure 6b. Spiral vibration in polar plot – compressor side

3
A hysteresis between run up and run down can be seen in the later than position 1. It is obvious, that for each position, there is
vibration versus speed plot (Bode plot), Fig. 7. It is a typical one point on the shaft surface, which is subject to maximum
behavior for a spiral vibration. friction due to the maximum velocity gradient (red profile). The
opposite point of the shaft is subject to the minimum friction due
to the minimum gradient (blue profile). In case of a synchronous
orbit it is always the same point on the shaft, which is subject to
maximum friction (red dot) and minimum friction (blue dot).
Thus a differential heating and a hot spot occurs.

Figure 7. Vibration hysteresis in Bode plot

Several additional internal mechanical running tests were

Figure 8. Hot spot mechanism in the bearing and the real
performed with higher and lower oil temperature, with higher
circular orbit as tested
and lower oil pressure and thus flow and also with increased
and decreased radial bearing clearance. Some of these tests For the present case of tilting pad bearings with very low
brought little improvements in the shaft vibration behavior, but load the case of a centered circular orbit is close to reality as
did not eliminate the hot spot phenomena and the spiral shown in Fig. 8. The estimation of the ratio of added to
vibration. Therefore it was decided to perform extensive eliminated heat according to appendix B is among others based
theoretical hot spot stability analyses prior to envisage rotor and on this assumption.
bearings modifications. In case of an elliptic off-centered orbit differential heating
will also occur. It may be less since it is not always the same two
HOT SPOT STABILITY ANALYSIS opposite points, which are subject to maximum and minimum
The method introduced by Schmied [5] to calculate the hot friction. On the other side the maximum friction can be higher,
spot stability based on the hot spot model of Kellenberger [3,4] since the minimum clearance can be lower. Balbahadur and Kirk
is described in detail in appendix A. The method can be used [8] claim, that the centered circular orbit is the worst case.
for any rotor system and heating mechanism (also see [12]). However, there is no real evidence from practical experience for
Several heating locations can be considered in one model at this, since hot spot induced spiral vibrations have also been
once. The stability threshold is calculated as a function of the observed in case of highly loaded bearings with elliptic off-
ratio of added to eliminated heat. The added heat is assumed to centered journal orbits.
be proportional to the speed and shaft deflection at the hot spot The thermal deflection of the shaft due to 1oC temperature
location, the eliminated heat proportional to the thermal difference in the cross section at the bearing on the compressor
deflection. Thus the stability threshold is a function of the ratio side is shown in Fig. 9.
p /q with p as the proportionality factor for the added heat, q
for the eliminated heat and as the speed.
The proportionality factors depend on the heating
mechanism and must be derived accordingly. The heating
mechanism in a bearing is briefly described in the following
paragraph. The corresponding proportionality factors can be
estimated with some simplifications as described in appendix B.
If the ratio of added to eliminated heat is close to the
threshold or even above the threshold, the rotor can be regarded
as sensitive to bearing hot spots.

Hot spot mechanism in the bearing o

In Fig.8 the shaft is shown in the bearing for a centered, Figure 9. Thermal deformation due to 1 C temperature
difference in the compressor bearing
circular orbit in two positions denoted as 1 and 2 together with
the velocity profile of the oil film. Position 2 is a quarter period

4
Hot spot stability chart (original design) thrust collars and by increasing the diameter of the central
The hot spot stability chart for the original configuration is section.
shown in Fig. 10. Hot spots in both bearings were considered in The reduction of the heat input into the shaft can be
the analysis, which work together and excite the rotor. accomplished by reduction of the tilting pad bearing to shaft
Therefore two thermal modes (see appendix A) appear. They contact area, by reduction of the oil viscosity, or by increasing
represent the response to the thermally bent rotor. The threshold the bearing clearance.
of the 1st thermal mode (black line) is determined by the 1st The heat elimination rate can be increased by higher lube oil
bending and has its minimum at its critical speed. The threshold flow, or by bearing modifications. The bearing modification can
of the 2nd mode (green line) is determined by the 2nd bending. It include asymmetric tilting pads, cooling groove in the pads, or
is not important in the speed range of interest. pads made of material with higher conductivity e.g. bronze.
The estimated heat ratio (see appendix B) is the red curve In view of the delivery constrains of tilting pads made of
in the chart. It is the ratio of the differential heat generated in bronze the following two configurations have been selected for
the bearing due to losses (which is assumed to enter the shaft) to further testing:
the differential heat eliminated from the shaft by heat transfer to 1. modified stiffer rotor
the oil. Due to the simplifying assumptions, which do not 2. bearings with reduced pad width and in addition
consider the detailed bearing geometry such as the bearing utilizing lower oil viscosity.
preload and the gaps between the pads, a second heat ratio line The influence of the bearing parameter changes on the ratio
(dotted) is drawn assuming only 50% of the input according to of added to eliminated heat can be seen in Eq. (22) in appendix
Eq. (22) in appendix B. The stability threshold is between the B. The influence of the lower viscosity is obvious, since the ratio
ratio with 100% and 50% heat input. The present and many is proportional to it. The influence of a reduced bearing width
other experiences prove that in this case there is a large risk to cannot be directly seen, since it does not appear in the equation.
run into a hot spot problem, i.e. to get a spiral vibration with However, the parameter (thermal deflection per temperature
increasing amplitude. difference in the bearing cross section) is roughly proportional to
the width. Thus the ratio is also proportional to the width.
The increase of the bearings clearance was not considered,
4.00E-02
because it did not bring any significant spiral vibration reduction
3.50E-02
during the already performed foregoing testing, in spite of the
3.00E-02
1.Hop Spot Mode
fact that the heat ratio is a square function of the clearance. Most
unstable
2.50E-02
unstable
2.Hot Spot Mode
Bearing
probably the lower heat ratio was compensated by a lower
threshold due to a reduction of the 1st bending mode frequency
p* /q

2.00E-02 Bearing (50% heat input)

Nominal Speed
1.50E-02 MCOS and a larger vibration deflection in the bearing.
+ -
1.00E-02 The above described modifications 1 and 2 were
+ -
5.00E-03 implemented and internal testing performed to confirm the
0.00E+00
10000 15000 20000 25000 30000
theoretical results.
n [rpm]

Modified stiffer rotor

Figure 10. Hot spot stability chart for the original configuration The original rotor was equipped with replaceable thrust
collars. The modified rotor was without replaceable thrust collars
The direction of the hot spot precession is also indicated in and its diameter was increased in the central part. The original
the chart. ―+‖ means the hot spot and thus the phase angle and modified rotors are shown in Fig. 11 and 12.
change in the polar plot move in the same direction as the rotor
speed, ―-― means in the other direction. In the speed range of
interest the direction is ―+‖, which agrees with the measurement
in Fig. 6.

ANALYZED AND TESTED MODIFICATIONS

Based on the hot spot stability analysis the following three Figure 11. Original rotor
remedies were considered for the elimination of the hot spot
phenomena:
- Install modified rotor with higher stiffness
- Reduce the heat input into the shaft in the bearing area
- Increase the heat elimination in the bearing area
Within the existing bearing carrier the rotor can be
modified into stiffer execution by elimination of the replaceable
Figure 12. Modified rotor

5
 The Campbell diagram for the 1st forward whirling bending The vibration levels for the modified rotor could be accepted
mode of the original and stiffened rotor is shown in Fig. 13. As according to the API 617 criteria. But because of the eliminated
expected the stiffer rotor increases the separation margin. The thrust collars and the beginning hot spot phenomena on the
damping is reduced. However, due to the large separation compressor side above the MCOS, this variant was not
margin this should not deteriorate the rotor dynamic considered as an acceptable solution.
performance. From this behavior can be concluded, that the actual heat
ratio at MCOS is close to the threshold in Fig. 14, i.e. it
corresponds to about 80% heat input according to the estimation
in appendix B (Eq. (22)).

Figure 15. Vibration in polar plot – expander side

st
Figure 13. Campbell diagram for the 1 forward whirling
bending mode

Hot spot stability chart for the modified rotor

The stability chart of the stiffened rotor is shown in Fig. 14
together with the stability threshold line of the original rotor as
dashed lines. The slight improvement of the threshold can be Figure 16. Spiral vibration in polar plot – compressor side
attributed to the higher critical speed, which moves the
minimum threshold to a higher speed.

4.00E-02
1.Hop Spot Mode
3.50E-02 2.Hot Spot Mode
Bearing
3.00E-02
Bearing (50% heat input)
2.50E-02 1.Hot Spot Mode Stiffer Rotor
2.Hot Spot Mode Stiffer Rotor
p* /q

2.00E-02
Nominal Speed
1.50E-02 MCOS

1.00E-02

5.00E-03

0.00E+00
10000 15000 20000 25000 30000 35000
n [rpm]

Figure 14. Hot spot stability chart for the modified rotor Figure 17. Vibration hysteresis in Bode plot – compressor side

Test bed measurements for the modified rotor Modified bearings

The ADRE vibration records for the modified rotor are The original tilting pad radial bearings were of a robust
shown below in Fig. 15-17. It can be seen, that the vibration on design with a standard width of 0.6D. The contact area to the
the expander side was behaving exceptionally well without any shaft can be reduced by shortening the pad, by narrowing the
reservations. However on the compressor side, a slight thermal bearing width, or by combination of both measures. It was
bow by hot spot and spiral vibration was identified in the Bode decided to narrow the bearing width from 55 to 40 mm, also in
plot, which shows the vibration hysteresis above the MCOS. view of the very low bearing loads. The existing original tilting
pads and the bearing body were modified as shown in Fig. 18.

6
Figure 18. Original 0.6D bearing and modified 0.4D bearing Figure 20a. Vibration in polar plot – expander side

In addition to the bearing width reduction also the oil

viscosity was reduced from nominal 46 cSt to 32 cSt oil. The
rotor dynamics was recalculated for the thinner oil and found to
be adequate. There was almost no change except a slightly
reduced damping.
The two measures reduce the heat input into the shaft
approximately by the following factor:
40/55 x 32/46 = 0.50
This means that both modifications reduced theoretically
Figure 20b. Vibration in polar plot – compressor side
the heat input into the shaft to some 50% of the original value.
It was important to confirm this assumption by hot spot stability
calculation and chart and by testing the new configuration.

Hot spot stability charts for modified bearings with

reduced oil viscosity
The stability chart with the bearings with reduced width
and the lower viscosity is shown in Fig. 19. The reduced
bearing width practically did not change the threshold line, but
the heat ratio now is considerably lower. The heat ratio line for
100% heat input is clearly below the threshold in the speed
range of interest. Therefore a stable system can be expected.
Figure 21. Vibration hysteresis in Bode plot

4.00E-02 Decided final configuration

3.50E-02
The final configuration of the turboexpander rotor and
3.00E-02
unstable 1.Hop Spot Mode
bearings was decided after the internal testing to be the execution
2.50E-02
unstable
2.Hot Spot Mode
with reduced bearings width and with the reduced lube oil
p* /q

Bearing
2.00E-02

1.50E-02
Bearing (50% heat input)
Nominal Speed
viscosity. With this execution both turbo-expanders and their
1.00E-02
+ - MCOS
spare plug-in units were successfully FAT tested with the
5.00E-03
+ - customer and all accepted without reservations.
0.00E+00
10000 15000 20000 25000 30000
n [rpm] EXPERIENCE WITH OTHER TURBO-EXPANDERS
Figure 19. Hot spot stability chart for the reduced bearing After this new exhaustive experience with the vibration
width with reduced oil viscosity induced hot spots, based on predictive analysis and testing, the
unexplained historical cases from the years 1996 and 2001 were
Test bed measurements for modified bearings with analyzed and the unexpected excessive vibrations during FAT
reduced oil viscosity were re-evaluated. Both cases with still available ADRE
With the modified bearings and the oil viscosity reduced vibration records could be clearly identified as the spiral
from nominal 46 cSt to 32 cSt the test results did not show any vibration problem (see Fig. 22 and 23).
more the spiral vibration and vibration hysteresis. With the two Both cases happened on turbo-expanders with similar shaft-
described measures the shaft thermal bow due to the hot spot bearings configurations as shown in Fig. 2 and 11. The first case
phenomenon and the spiral vibration were completely (Fig. 22) was identified by slight labyrinth rubbing marks on
eliminated up to the tested speed of 22’000 rpm. The test results shaft and was corrected by increasing the labyrinth clearance.
are shown in Fig. 20 and 21. The second case was tested up to 104 m/s bearing journal

7
velocity (Fig. 23) and was corrected during FAT by clearance, the lube oil properties (viscosity) and can be
modification of the bearing clearance. But the vibration influenced by the same. The described advanced hot spot
problem solving was a kind of a try and error solution, the stability analysis can be utilized as the analytic tool to eliminate
predictive analytical method was missing. unforeseen and possible vibration problems already during the
design stage. This way the time consuming vibration trouble
shooting during the internal testing, the FAT and the start up of
high speed rotating machines can be avoided.
The turbo-expanders as described in this paper are today
successfully operating in the foreseen speed and load range.

ACKNOWLEDGEMENTS
The authors thank Cryostar SAS for allowing publication of
this document.

Figure 22. Vibration in polar plot for a case in 1996 REFERENCES

[1] Newkirk, B. L., 1926, ―Shaft Rubbing. Relative Freedom of Rotor
Shafts from sensitiveness to rubbing Contact When Running
Above Their Critical Speeds,‖ Mech. Eng. (Am. Soc. Mech.
Eng.), 48(8), pp. 830-832.
[2] Dimarogonas, A. D., 1973, ―Newkirk Effect, Thermally Induced
Dynamic Instability of High Speed Rotors,‖ Proc. ASME Int. Gas
Turbine Conference, Washington D.C., ASME 73-GT-26.
[3] Kellenberger, W., 1978, ‖Das Streifen einer rotierenden Welle an
einem federnden Hindernis – Spiralschwingungen,― Ing.-Archiv,
47, pp. 223-229.
[4] Kellenberger, W., 1980 ―Spiral Vibrations Due to the Seal Rings
Figure 23. Vibration in polar plot for a case in 2001 in Turbogenerators Thermally Induced Interaction Between Rotor
and Stator,‖ ASME J. Mech. Des., 102, pp 177-184.
In another case of integral gear expander-compressor [5] Schmied, J., 1987, ―Spiral Vibrations of Rotors,‖ Proc. 11th
Biennal ASME Design Engineering Div. Conference, Vibration
machine, the predictive hot spot stability analysis was
and Noise, DE-Vol. 2, Rotating Machinery Dynamics, Boston
performed during the design stage to eliminate any possible MA, ASME H0400B, pp. 449-456.
vibration problems in advance. The analysis identified the [6] De Jongh, F. M., and Morton, P. G., 1996, ―The Synchronous
possible vibration induced hot spot problem and therefore the Instability of a Compressor Rotor due to Bearing Journal
bearing-shaft configuration was correspondingly and adequately Differential Heating,‖ ASME J. Eng. Gas Turbines Power, 118,
adjusted to improve the heat elimination in the bearing area. pp. 816-824.
During the FAT and also during full load operational test no [7] Keogh, P. S., and Morton, P. G., 1993, ―Journal Bearing
vibration problems were identified. Differential Heating Evaluation With Influence on Rotor Dynamic
Behaviour,‖ Proc. R. Soc. London, Ser. A 441, pp. 527-548.
[8] Balbahadur, A. C., and Kirk R. G., 2004, ―Part I—Theoretical
CONCLUSIONS Model for a Synchronous. Thermal Instability Operating in
For high speed rotors, even with favorable rotor dynamics Overhung Rotors,‖ Int. J. Rotating Machinery, 10(6), pp. 469–
evaluation according to the API rules, the hot spot stability 475.
analysis as described in this paper is a proven method to [9] Carrick, H. B., 1999, ―Integrally Geared Compressors and
determine possible unpredicted vibration problems. As a rule of Expanders in the Process Industry,‖ Proc. Seventh European
thumb, all stiff rotors running below the bending critical speed Congress on Fluid Machinery for the Oil, Petrochemical and
with a journal bearing peripheral velocity above 80 m/s are Related Industries, The Hague, IMechE.
prone to vibration induced hot spots. For more flexible rotors [10] Schmied, J., 2007, ―MADYN 2000 Documentation,‖ Delta JS,
this threshold speed can be even lower. In this case overhang Zurich.
[11] API 617, 2002, ―Axial and Centrifugal Compressors and
modes instead of the bending modes influence the appearance Expander-compressors for Petroleum, Chemical and Gas Industry
of the stability chart. Typically the threshold in the overcritical Services,‖ 7th ed., API American Petroleum Institute.
range then is very low in contrast to the present case, where this [12] Eckert, L., and Schmied J., 2008, ―Spiral Vibration of a
is the case in the subcritical range of the 1st bending mode. The Turbogenerator Set: Case History, Stability Analysis,
reason for this is the location of the vibration node, which is Measurements and Operational Experience,‖ ASME J. Eng. Gas
outside the bearing here and inside the bearing in a typical Turbines and Power, 130, 012509 (10 pages).
overhang mode. [13] Klement, H. D., 1997, ―MADYN Program System for Machine
The vibration induced hot spot behavior in bearings Dynamics, Version 4.2,‖ Ingenieurbüro Klement, Darmstadt.
depends on the shaft geometry, the bearing design, the bearing

8
NOMENCLATURE If the rotor is thermally deformed the equation of motion for
c Specific heat capacity coordinates x relative to the static position of the undeformed
f Vector of external excitation forces rotor is
m Mass of the bearing journal
p Factor for added heat Mx (D G)x Kx K R xT f( t ) (2)
q Factor for eliminated heat
t Time where KR is the stiffness matrix of the rotor alone (without
u Bearing peripheral speed pedestals and journal bearings) and xT is the vector describing
x Translational deflection coordinate the thermal deformation. The thermal deformation xT is assumed
x Vector of coordinates to depend linearly on the thermal deflections xT,HS of the shaft at
xT Thermal translational deflection coordinate a reference location. Typically it is the location of the hot spot
xT Vector of coordinates describing the thermal deflection (HS).
A Area of the bearing journal surface surrounded by oil
D Damping matrix xT T xT, HS , (3)
G Gyroscopic matrix T
xT,HS ( xT,h , xT,v ) (4)
K Stiffness matrix
KR Stiffness matrix of the free rotor where xT,h and xT,v are the thermal translational deflections in
M Mass matrix horizontal and vertical direction at the location of the hot spot.
P Bearing power loss The matrix T is derived from the thermal deformation of the
T Transformation matrix rotor (see Fig. 9), which is determined by a static calculation with
 Heating efficiency
Q thermal loads: for the first column of T for a temperature
Real part of the eigenvalue gradient in horizontal direction and for the second column for a
Heat transfer coefficient temperature gradient in vertical direction.
Relation between thermal deflection and cross The modeling of the hot spot phenomenon according to the
sectional temperature difference theory of Kellenberger is based on the following thermal
Radial bearing clearance equation:
Factor for heat dissipation
x T,HS
~ p ~ HS q ~ (5)
Eigenvalue x x 
T, HS
Imaginary part of the eigenvalue 
Q Q
Oil viscosity
Factor for heat input with ~
x T,HS as the thermal deflection at the location of a hot
Cross sectional temperature difference
T
Rotor speed in [rad/s] spot and ~ x HS ~
xh , ~
xv as the vector of the translational

 is the
shaft displacements at the location of the hot spot. Q
APPENDIX A
added heating efficiency and Q  is the dissipated heating
CALCULATION METHOD FOR THE HOT SPOT efficiency.
STABILITY Equation (5) was derived in detail by Kellenberger (see [3]).
The equation of motion of the Finite Element model of a The coordinates in Eq. (5) are rotating in contrast to the
multi bearing rotor has the form coordinates in Eqs. (1), (2) and (3) which are stationary.
Equation (5) implies the following:
Mx (D G)x Kx f(t ) (1) The change of cross sectional temperature difference is
with the mass matrix M, the damping matrix D, the gyroscopic proportional to the change of thermal deflection,
matrix G and the stiffness matrix K. The matrices D and K are the differential heat input across the shaft cross section is
non-symmetric due to the damping and stiffness coefficients of proportional to the shaft displacement and the speed,
the journal bearings and dependant on rotor speed. the differential dissipated heat across the shaft cross section
is proportional to the thermal deflection.
The proportionality factors for the added and dissipated heat
are p and q. Kellenberger has applied the method to simplified
shaft models. The method was extended for the use on general
shaft systems (see [5]).

9
 Transforming Eq. (5) into stationary coordinates xT,h, xT,v, The additional two eigenvalues are a conjugate complex pair
xh, xv yields
T i . (9)

They will be called „thermal― eigenvalue in the following. The

1 0 xT,h p 0 xh imaginary part of the thermal eigenvalue is almost equal to .
0 1 xT,v 0 p xv The interpretation of the thermal eigenvector (so-called hot spot
, (6) mode) is derived in [5]:
q xT ,h 0 The difference between and indicates how fast the
q xT ,v 0 thermal bow moves along the shaft respectively at which speed
the spiral is traced. The period to complete 360°-turn of the
spiral is
or by using matrices and vectors
T 2 / . (10)
I x T,HS P xHS Q xT,HS 0. (7)
The direction of the revolution of the unbalance response
Substituting (3) into (2) and extending (2) by (7) yields vector is as follows:
, forward (co-rotational, same direction as the
M 0 x (D G) 0 x rotor spin)
0 0 xT,HS 0 I x , backward (counter- rotational, opposite

  
  T,HS direction as the rotor spin).
A B
(8) The real part of the thermal eigenvalue T indicates whether
R
K K T x 0 the spiral increases or decreases in magnitude:
P Q x 0 0 , increasing magnitude (unstable behavior)
   T,HS 0 , decreasing magnitude (stable behavior).
C To receive the stability threshold the heat input (i.e. the
where P is a 2 x N matrix (with N as the dimension of Eq. (2)), factor p) is varied until the real part becomes = 0.
which has the coefficients of P at the columns of the The method can also be applied for more than one hot spot
translational coordinates xHS of the hot spot. by extending Eq. (8) accordingly. The number of additional
To receive Eq. (8) the original Finite Element model of the ―thermal‖ eigenvalue pairs corresponds to the number of hot
rotor represented by Eq. (1) must be extended by introducing spots.
the two extra degrees of freedom xT,h and xT,v. The additional
coefficients must be added to the global matrices. This APPENDIX B
possibility is provided for the input of the program MADYN
[13]. ESTIMATION OF THE RATIO OF ADDED TO
From Eq. (8) one could calculate the time history of x by a ELIMINATED HEAT FOR A FLUID FILM BEARING
time step method. The polar plot of the time history of each The ratio of the added to eliminate heat is estimated with the
coordinate would be a spiral either increasing or decreasing in help of the following simplifying assumptions:
magnitude. This calculation however would require larger 1. The journal orbit is a synchronous circle.
computational effort, since the period of a spiral to complete a 2. The radial clearance around the circumference is constant.
360°-turn is very long for realistic examples. That is why the 3. The circumferential distribution of the added heat due to the
computation would require the simulation of quite a long period vibration is a sinus shape. The maximum is at the high spot
of time until one could judge whether a spiral increases or of the shaft.
decreases in magnitude. 4. For the eliminated heat distribution the same applies as for
As described in [5] this information can be easier extracted the added heat.
from the complex eigenvalues of Eq. (8). Their calculation 5. The axial distribution of the temperature in the journal is
requires little computational effort. constant.
Equation (8) has 2(N+1) eigenvalues. Since realistic values 6. The journal ends are adiabatic.
for p and q are very small a set of 2N eigenvalues are 7. The complete sinus shaped friction power due to the
practically the same as those of Eq. (1) representing the rotor vibration enters the shaft (not the oil).
system. These assumptions are best fulfilled for an unloaded
cylindrical bearing.

10
 The thermal equation for the cross sectional temperature The added heat can be estimated as
difference in the journal is: . 2 x (20)
. Q P
(11)
with
with the bearing power loss P
3 A
, (12)
mc u2 A , (21)
P
3 .
Q (13)
2mc the viscosity , the peripheral speed u and the radial bearing
and m as the mass of the journal, c as the specific heat capacity, clearance .
A as the surface area of the journal surrounded by oil, as the (20) and (21) substituted into (19) yields:
.
heat transfer coefficient and Q as the added heat. p u2
. (22)
The thermal equation according to Kellenberger has the q 2
following form (also see A5):
.
xT p x qxT (14)

with x as the radius of the circular journal orbit and xT as the

thermal deflection.
The relation between xT and is:

xT , (15)

where has to be determined from the rotor model by assuming

a temperature difference across the journal. For the thermal
deflection any reference location on the rotor could be taken.
The ratio of added to eliminated heat depends on this location,
but the stability threshold will adjust accordingly. Typically the
location at the hot spot is taken (also see Eq. (3)).
(15) substituted into (14) yields:
. p
q x. (16)

Comparing (16) and (11) considering (12) and (13) yield:

3 A
q , (17)
mc
p 3 .
x Q. (18)
2mc
(17) and (18) yield:
.
p Q
. (19)
q 2x A

11