Professional Documents
Culture Documents
Introduction sleeve at 120 deg spacing along the bearing circumference induce
mechanical preloads. Their prediction showed that three shims
GFBs are widely used in high-power-density turbomachinery
increase the direct stiffness and damping coefficients, thus
because they can operate reliably at high speeds in a compact
improving the rotordynamic stability of a simple rotor-GFB sys-
unit. As supports for high speed rotating machinery, GFBs offer
tem test rig. Heshmat and Walton [8,9] first applied GFBs to mul-
greater stability, and load capacity while incurring a lower mainte-
tishaft rotating systems. They analyzed and tested two
nance cost than gas rigid bearings.
combinations of multishaft systems (rigid–rigid and rigid–flexible
Over the last few decades, GFBs have been advanced consider-
coupled shaft) connected using a flexible coupling. They provided
ably through research efforts. Heshmat et al. [1] analyzed GFTBs
considerable insight into the range of stiffness needed for coupling
using the finite difference method. They solved the Reynolds
when used in rigid–flexible coupled rotor systems supported on
equation for a compressible fluid flow and predicted the load
GFBs. Lee et al. [10] conducted speed-up and coast-down tests of
capacity of GFTBs. Agrawal [2] reported the chronological pro-
a coupled motor–compressor shaft system. The two shafts are
gress of the development of GFBs and their applications in turbo-
coupled using a bellows-type flexible coupling for isolating the
machinery. The author insisted that a GFB was one of the
dynamics from each shaft. The test results show that the peak
promising bearings for applications to missile engines in light of
amplitude at the critical speed decreases and the onset speed of the
its many applications particularly to air cycle machines (ACMs).
peak amplitude is delayed slightly with decreasing radial clearance.
Dellacorte and Valco [3] proposed the “Rule of thumb” (ROT) for
Thus far, however, studies have not reported on the rotordynamic
a quick estimation of the GFB load capacity. The authors used
behavior of multiple shaft systems supported on GFBs.
experimental test data obtained by various researchers since 1965
To address this deficiency, the present study develops a rotordy-
to show that the load capacity of GFBs was, in general, linearly
namic model to predict the behavior of a 225 kW motor–generator
proportional to the rotor surface speed. Dellacorte et al. [4] pro-
with multiple rotors supported on GFBs. The natural frequencies
vided a comprehensive review of the design, fabrication, and per-
and mode shapes are analyzed using the rotordynamic model and
formance test of first-(Generation I) and second-(Generation II)
predicted GFB dynamic coefficients. During the experiments, at
generation bump-type GFBs. Heshmat [5] developed double-
speeds of up to 60,000 rpm, corrections and adjustments such as
bump-layer GFBs that offered significantly increased load
multiple shaft alignments and bearing dimension resizing were
capacity. The author reported an ultimate load capacity of 673.5
required for enabling operation at the maximum speed without
kPa (97.7 psi) for test GFB. Kim and San Andres [6] developed a
excessive subsynchronous motions.
simple model for GFBs with upper (soft) and lower (stiffer) bump
layers. The authors predicted larger direct stiffness and damping
coefficients for double-bump-layer GFBs. The authors [7] also
predicted the dynamic performance of GFBs with mechanical pre- Motor–Generator System
loads. They showed that three shims in contact with a bearing
Figure 1 shows a schematic view of the 225 kW (300 HP) class
PMS motor–generator system supported by GFBs. The motor
1
Corresponding author. shaft is radially supported by two identical GFJBs and axially by
Contributed by the Structures and Dynamics Committee of ASME for publication
in the JOURNAL OF ENGINEERING FOR GAS TURBINES AND POWER. Manuscript received
a pair of GFTBs positioned about the thrust collar. On the genera-
December 4, 2014; final manuscript received January 15, 2015; published online tor shaft, similar bearings and an identical arrangement are
February 25, 2015. Editor: David Wisler. employed. To enable torque measurement, the alternator of the
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 092505-1
C 2015 by ASME
Copyright V
(a) GFJB
the changes in rotor speed and assumes that the flexible coupling
is isotropic. This assumption essentially simplifies the transfer
matrix and minimizes the computational time while maintaining
the predictive accuracy.
Figure 5 shows the results of the predicted free–free analysis
and experimental data from the impact test. The first and second
natural frequencies (or rigid modes) occur at 27 Hz (1638 rpm)
and 185 Hz (11,114 rpm), respectively. The third and fourth natu-
ral frequencies (first and second flexural modes) are predicted to
occur at 1149 Hz (68,933 rpm) and 1335 Hz (80,081 rpm), respec-
tively. However, test results indicate that the third and fourth natu-
ral frequencies (first and second bending modes) actually occur at
1130 Hz (67,800 rpm) and 1320 Hz (79,200 rpm), respectively.
This gives marginal error values of 11.6% and 8.33%, respec-
tively. This indicates the predictive accuracy of the model. None-
theless, because the rated operating speed of the PMS
motor–generator system is 60,000 rpm (1000 Hz), i.e., 13% criti-
cal margin below the first flexural mode, the rotor assemblage can
be considered rigid–rigid coupling.
Figure 6 shows the Cambell diagram, which is representative of
Fig. 7 (a) Mode shape of coupled motor–generator system
a function of rotating speed and bearing rotordynamic coeffi- for each N.F. (N.F.1–4) and (b) mode shape of coupled motor–
cients. The actual systems include negative stiffness effects of the generator system for each N.F. (N.F.5–6)
permanent magnet motor and generator. It is assumed that these
effects are ignored because negative stiffness effects of the perma-
nent magnet are lesser than those of GFBs. The rotordynamic Figure 7(a) shows the mode of an assembled rotor from the first
coefficients are calculated using an in-house code by the perturba- to the fourth natural frequency. The first four natural frequencies
tion analysis of the Reynolds equation. Damping analysis is per- cause strains of the shafts about the flexible coupling. Actually,
formed using an advanced rotordynamic tool. The plots of these because the rated operating speed of the coupled rotor–generator
against the rotor speed give the Cambell diagrams. The first four system is 60,000 rpm, a driving motor should quickly pass
natural frequencies (152, 159, 185, and 249 Hz) are rigid modes over this region (9000–20,000 rpm) at a speed up velocity of
of each shaft and denote a rigid critical range (9148–14968 rpm). 5000 rpm/s to prevent very large vibrations at the coupling critical
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 092505-3
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 092505-5
behavior of the test apparatus and facilitate the successful attain- are calculated using an in-house code based on the perturbation
ment of the design speed (60,000 rpm). analysis of film reaction forces by the Reynolds equations. The
predictions indicate that the force coefficients of the three
d ¼ lnfxðtÞ=xðt þ nTÞg=n shimmed GFJBs are consistently larger than those of the original.
This is evidently true from the reduction in film clearance. How-
Figures 16 and 17 show that the frequency of the subsynchro- ever, their disparities become progressively smaller with increas-
nous motion is one-half of N.F.4 (fourth critical speed). Further- ing journal speed. Moreover, the magnitude of the direct spring
more, the one-half frequency of N.F.4 results from the force (Kxx) that acts in the static load (or rotor weight) direction is
supersynchronous excitation (2). Therefore, the subsynchronous larger and more responsive to changes in journal speed than the
motion in the waterfall graph is associated with N.F.4. Figure 17 coefficient (Kyy) in the normal direction. Although the cross-
shows the logarithmic decrement graph of the system for different couple stiffness (Kxy, Kyx) of the three shimmed GFJBs are
rotor speeds. Equation (1) defines the logarithmic decrement. The higher than those of the original, their destabilizing strength is suf-
initial GFJB (bearing width ¼ 60 mm, without three shims) shows ficiently minimized by the corresponding larger direct damping
significant subsynchronous motions at excitation frequencies with coefficients (Cxx, Cyy). The synergic effects of improved damping
peak amplitudes near 120–130 Hz at a rotor speed of 35,040 rpm and the comparatively higher direct stiffness of the modified
(584 Hz) for a logarithmic decrement of 0.158. For the modified GFJB enable the test machine to operate at the rated speed with
GFJB with a bearing width of 45 mm, the onset of subsynchronous enhanced stability. Quantitatively speaking, the modification
motions is delayed to 43,200 rpm (672 Hz) and the amplitude of the improves the values of Cxx and Cyy by 33.05% and 42% of the
logarithmic decrement similar to that for a rotor speed of 35,040 rpm original GFJB at 35,040 rpm, respectively. This improvement in
(584 Hz) of the initial GFJB. Furthermore, the analysis result of the damping delays the onset speed of subsynchronous motions and
logarithmic decrement increases from 0.158 to 0.201 at 35,040 rpm, reinforces rotordynamic stability. Note that the stiffness and
which is 21.4% larger than the initial GFJB. Furthermore, the loga- damping effects of the force fields generated by the PMS motor
rithmic decrement increases from 0.106 to 0.151 at 43,200 rpm. and generator are one order of magnitude smaller than those of
the bearing; and therefore, they are considered negligible.
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 092505-7