Professional Documents
Culture Documents
201710149
This contribution compares two approaches for modelling seal fluid forces in a rotor system with compliant seal support: on
the one hand the non-linear ODEs of the Muszynska model are used. On the other hand Hirs’ PDEs for turbulent lubrication
flow are numerically integrated and an ODE coefficient model is fitted to this data. Simulations are performed with both fluid
models and the results are compared: they show not only quantitative differences in the stability behaviour of the rotor-system
but also qualitative differences in the bifurcation behaviour.
c 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
The subscript R indicates rotor-related and the subscript S seal-related variables and parameters. Ff (∆r) is the fluid force
being dependent on ∆r = rR − rS . In this contribution two modelling approaches for the fluid forces Ff are compared: the
parametric Muszynska-model [4]
mf 0 D̄ 2τ̄ Ωmf K̄ − mf τ̄ 2 Ω2 τ̄ ΩD̄
Ff M (∆r) = − ∆r̈ − ∆ṙ − ∆r, where (2)
0 mf −2τ̄ Ωmf D̄ −τ̄ ΩD̄ K̄ − mf τ̄ 2 Ω2
1 ∂H
+ ∇ · (HU ) = 0 and (5)
RΩ ∂t
UZ 1
p|θ,Z=L,t = 0, Uθ |θ,Z=0,t = 0.5 and p|θ,Z=0,t = ∆p − ρ(1 + ξ)UZ2 |θ,Z=0,τ . (6)
Uz0 2
Here, ρ is the density, H is the relative sealing gap, R is the rotor radius, U = (Uz , Uθ )T is the fluid velocity, p is the pressure
and n0 , m0 are empirical constants (c.f. [7]), Rc is the circumferential Reynolds number, L is the seal length and ξ is the
∗ Corresponding author: e-mail baeuerle@uni-kassel.de, phone +0049 561 804 2169
∗∗ e-mail hetzler@uni-kassel.de phone +0049 561 804 2868
pressure loss coefficient at the entry of the seal (c.f. [6]). Numerical integration of these equations is done with C OMSOL
M ULTIPHYSICS
R
. The parameters of the following ODEs are then fitted with a linear least square approach to the FE data
which gives
Ff H = −Mf H (∆r, Ω)r̈ − Bf H (∆r, Ω)ṙ − Kf H (∆r, Ω)r. (7)
Note that according to FE simulations all matrices including Mf H turned out to be dependent on ∆r and Ω. Also, the results
are sensitive to parameter changes and show an apparent whirl frequency dependence which has not yet been investigated.
80 5 2 Results
CS-border Mus.
CS-border Hirs Figure 3 shows the results from eigenvalue computa-
4
60 NCS-border Mus. tion, where the stiffness ratio κ2 = ccRS has been plotted
NCS-border Hirs against the dimensionless rotor angular speed η. Three
3 interesting things can be concluded from this: Compar-
cR
cR
cS
cS
κ2 =
References
[1] A. Messenger, R. Williams, G. Ingram, S. Hogg, S. Tibos and J. Seaton, ASME Turbo Expo 2015: Turbine Technical Conference and
Exposition, Montreal, Canada (ASME, New York, 2015), pp. V008T26A031–V008T26A031.
[2] L. SanAndrés and A. Anderson, J ENG GAS TURB POWER 137.5, 052504 (2015).
[3] S. Bäuerle and H. Hetzler, Proceedings in Applied Mathematics and Mechanics, (Wiley Online Library, Berlin, 2015), pp. 241–242.
[4] A. Muszynska, J SOUND VIB 110.3, pp. 443–462 (1986).
[5] S. Li, Q. Xu and X. Zhang, NONLINEAR DYNAM 47.4 pp. 321–329 (2007).
[6] D. W. Childs, J LUBRIC TECH-T ASME 105.3, pp. 429–436 (1983).
[7] GG. Hirs, J FLUID STRUCT 95.2, pp. 137–145 (1973).
[8] S. Bäuerle and H. Hetzler, Proceedings in Applied Mathematics and Mechanics, (Wiley Online Library, Berlin, 2016), pp. 257–258.