Professional Documents
Culture Documents
Rotor Vibrations
Rotors with three or more fluid-film bearings (or fluid seals) have "redundant"
supports, and, therefore, interdependent bearing loads that are generally unknown
both in magnitude and direction. The steady-state bearing eccentricities and the
dynamic stiffness and damping coefficients of the bearings are, therefore, also un-
known since both are functions of the bearing loads. Thus, the dynamic behavior of
J. L. Nikolajsen multibearing rotors generally cannot be predicted with good accuracy without access
Staffordshire University, to a procedure for calculating the steady-state bearing loads and eccentricities. This
School of Engineering, paper outlines such a procedure in terms of both the influence coefficient method,
Stafford, ST18 OAD, the transfer matrix method, and the finite element method. Radial bearing misalign-
England
ment and flexibility of the bearing back-up structures are accounted for. Once the
eccentricities are available, the bearing stiffness and damping coefficients can be
calculated in the usual way and used to predict critical speeds, instability threshold
speed, and rotor response to imbalance. A numerical example is presented that
illustrates some of the nonlinear effects of bearing support redundancy, notably the
large variations in instability threshold speed with radial bearing misalignment.
The example shows how the method can be used to determine the level of bearing
misalignment that leads to optimum rotor stability. It is concluded that no simple
guide lines exist by which optimum stability can be achieved. Neither perfect bearing
alignment nor equal load sharing between bearings necessarily lead to optimum
stability.
Contributed by the International Gas Turbine Institute and presented at the 41ST The Influence Coefficient Method. Figure 1 shows the
International Gas Turbine and Aeroengine Congress and Exhibition, Birmingham,
United Kingdom, June 10-13, 1996. Manuscript received by the ASME Head-
steady-state equilibrium position J of the journal within the
quarters January 31, 1996. Paper No. 96-GT-373. Associate Technical Editor: flexibly supported bearing. Three forces, acting on the jour-
J. N. Shinn. nal, are in static equilibrium: P (the bearing's share of the
Journal of Engineering for Gas Turbines and Power JULY 1998, Vol. 120 / 635
Copyright © 1998 by ASME
y z
Fig. 2 Shaft displacement components
by the standard flexibility influence coefficient method for stati- 2 < ( < « - 1 (4)
cally indeterminate beams, see, for example, Argyris (1958).
The influence coefficients for a shaft with variable cross section «-i n-i
can be derived as outlined in Vernon (1967). They are listed dyi = ~ X ayQyj = - X aij[Fj sin (0, - fa) - PyJ],
inNikolajsen(1978). j=2 ;=2
Nomenclature
av = flexibility influence coefficient F = hydrodynamic fluid-film force x, y = radial displacements
for simply supported shaft = ra- h = bearing (or seal) housing radial z = axial bearing (or seal) loca-
dial deflection of station i due displacement tion
to unit force at station j / = second moment of area of shaft e = Elc = bearing (or seal) eccentricity
ayi = back-up structure flexibilities at cross section ratio
bearing (or seal) no. i K, k = seal stiffness (direct and cross- 6, 4> = angular displacements
c = radial bearing (or seal) clear- coupled) 0 = direction angle of E
ance / = length of shaft element 4> = bearing attitude angle
d = elastic shaft deflection n = number of bearings and seals H = oil viscosity
e = radial bearing (or seal) mis- N = total number of shaft stations fi = shaft rotational speed
alignment P = external force on journal
E = bearing (or seal) eccentricity or Q = internal shaft elastic restoring
elastic modulus of shaft mate- force
rial
I
™V*'
Jvji+,
+ CiCj sin Oi + ayiFi sin (0,- — # , ) v; vt
n-l
+ X a f f [F ; sin (0j - ft) - Pyj] Fig. 3 Discrete shaft model for transfer matrix analysis
X (z, - z,)/(z„ - zi) + eyi = 0. (6) Also, the hydrodynamic bearing forces FX(E) and Fy(E) can
be expressed in terms of the housing flexibilities ax and ay, and
Moment equilibrium of the external forces for the entire shaft the bearing misalignments ex and ey as
about the end bearing positions results in
Fx = (ex - x')lax
I [Fj cos (0j - ft) - Pxj](zn -z,) =0 (7)
;=i Fy = (.ey ~ y')lay
from which
I [Fj sin (0y - <£,) - Pyj](z„ - zj) = 0 (8)
j=i
axFx
I [Fy cos (6j - </>,) - Pxj](zj -z,) =0 (9) y' = ey- ayFy. (13)
J =2
Journal of Engineering for Gas Turbines and Power JULY 1998, Vol. 120 / 637
where the eight top elements of the last column of the matrix resulting transfer matrix will look similar to Eq. (18), with
are nonzero at the bearing and seal locations only. Shear flexi- nonzero elements in the last column resulting only from the last
bility has been omitted in Eq. (16) for clarity, but can easily transfer as the bearing is reached. Thus, the two first lines of the
be included in the usual manner. If seals are present and the matrix equation represent two nonlinear equations from which
seal forces are considered linear, then further simplification can improved values of the eccentricity components at bearing no.
be achieved by substitution of Eq. (15). 1 can be derived, for example by Newton-Raphson iteration.
The effect of bearing and seal support flexibility and radial This procedure in repeated between bearings no. 1 and 2, etc.
misalignment is included, as before, by substitution of Eqs. (12) to calculate improved values of all the bearing eccentricities;
and (13), giving (vi) go back to (ii) until convergence.
where An through A22 are the four 4 x 4 submatrices outlined Numerical Example
inEq. (16). The three methods described above have been programmed
The basic solution procedure goes as follows: and included in rotordynamic programs based on the IC, FE,
and TM methods, respectively. They have been thoroughly
(i) choose initial values of eccentricity components Ex and tested and found to produce consistent results. One of the test
Ey at all the bearing locations; cases is presented here, for the rotor-system shown in Fig. 4,
(ii) make a transfer across the entire shaft, using Eq. (17), to illustrate the analysis procedure and to provide some insight
to get into the static and dynamic behavior of multibearing rotors.
[EA The rotor system shown in Fig. 4 consists of two lumped
Ey fill By. masses carried by a massless shaft supported by four identical
6 plain journal bearings. The shaft and the bearing stiffnesses are
of the same order of magnitude. The bearings were realistically
<^ 0 dimensioned with regard to mean bearing pressure, minimum
> = Vx (18) oil-film thickness, oil flow, friction loss, and temperature rise.
V"
y. A damped free vibration analysis was carried out using the
Vy
My My IC method, extended for eccentricity calculation, as described
Mx Mx
I 1J N . 0 0 0 0 0 0 0 0 1 1 J
(iii) Substitute the boundary conditions (usually zero forces E
and moments at the ends) to get Y § 001657 m
d 01
Ex E=2.05-l01'N/m-
-X
&x- g--G=CWE-
10
(19)
<P C=7.5«r?n
u=3.92-1(J2Ns/m
(iv) calculate {Ex, Ey, 9, $} 1 from the linear equation system
(19); 0.8 m
(v) use the corresponding full state vector {Ex, Ey, 9, <f), 0,
0, 0, 0}i in a transfer from station no. 1 to bearing no. 1. The Fig. 4 Sample multibearing rotor
above. The Sommerfeld Number, the attitude angle, and the Conclusions
usual eight bearing stiffness and damping coefficients for the
bearings were determined from graphs given in Woodcock A complex relationship has been shown to exist between
(1971). The bearing loads are generally not vertical for a radial bearing misalignment and stability of a multibearing ro-
multibearing rotor so it becomes necessary to perform a coordi- tor. Neither perfect alignment nor equal load sharing between
nate transformation of the bearing coefficients as outlined for bearings can be relied upon to give optimum stability. Numeri-
example in Nikolajsen (1973). cal analysis is necessary to predict the "misalignments" and
the tolerances on alignment that will ensure an adequate stability
The calculated steady-state journal positions as functions of
margin. Bearing support flexibility can improve the stability of
speed are shown in Fig. 5 for selected misalignments e of the
badly aligned rotors. However, it could reduce the stability of
middle double-bearing relative to the two end bearings. The
a well-aligned rotor by lowering the natural frequency in which
instability threshold speed QT is also indicated. The arrows
the rotor goes unstable. The complex relationship between mis-
indicate the direction of motion of the journal centers for in-
alignment and threshold speed, illustrated here, suggests that
creasing speed. The points on the loci correspond to speeds that
improper alignment could be responsible for some of the unex-
are continually doubled from a low of 500 rpm through 1000
plained cases of turbomachinery instability which occur from
rpm, 2000 rpm, etc. up to 128,000 rpm to get traces of the
time to time. It is recommended that misalignment effects be
entire loci, although in practice fiT sets an upper limit for the
investigated in such cases using one of the methods outlined
speed.
above. These methods are useful for all types of fluid-film bear-
In Fig. 5 ( a ) , the vertical misalignment has been chosen such ings and seals whose dynamic characteristics vary with eccen-
that the total static bearing loads on all four bearings are equal tricity ratio.
at all speeds. In Fig. 5(b), the misalignment is such that the
middle journals are concentric in the bearings at zero speed.
This leads to the unusual situation of increasing eccentricity for
increasing speed (below 1000 rpm). The loci of Figs. 5(c) and
(d) result when horizontal misalignment equal to the bearing
clearance is also added.
Figure 5 shows that both the bearing loci and the instability
threshold speed depend in a complex manner on the magnitude
and direction of misalignment. For example, Figs. 5(c) and
5(d) suggest that, for this rotor, misalignment in the negative
^-direction results in better stability than the same misalignment
in the positive x-direction. However, if the x-misalignments are
increased to ±4.28 X 10~4 m, then the threshold speeds become
10,795 rpm and 10,141 rpm respectively, thus reversing the
trend.
In Fig. 6, the instability threshold speed has been plotted as
a function of vertical misalignment of the middle bearing pair.
Optimum stability occurs at a misalignment of about -0.25 X
10~4 m (vertically down). This corresponds to neither equal
load sharing nor perfect alignment, although it is close to both Fig. 6 Rotor stability chart
Journal of Engineering for Gas Turbines and Power JULY 1998, Vol. 120 / 639