Professional Documents
Culture Documents
Tribology International
journal homepage: www.elsevier.com/locate/triboint
art ic l e i nf o a b s t r a c t
Article history: Two dynamic models of the dual clearance squeeze film damper (DCSFD) were established. One model
Received 24 August 2012 was based on the general Reynolds equation and the other model was based on the Reynolds equation of
Received in revised form conventional squeeze film damper. To validate the two models, vibration responses of DCSFD were
28 April 2013
obtained with different exciting frequencies, elastic supports and exciting forces by experiments. Based
Accepted 15 May 2013
on the experiment, a dynamic model including DCSFD was established. Excellent consistency of
Available online 23 May 2013
numerical solution was achieved between two different DCSFD models. Good consistency was achieved
Keywords: between the experiment and simulation. These observations provide useful models for further
Squeeze film damper studying DCSFD.
Floating-ring
& 2013 Elsevier Ltd. All rights reserved.
Reynolds equation
0301-679X/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.triboint.2013.05.010
188 H.L. Zhou et al. / Tribology International 66 (2013) 187–193
Floating ring
x
Journal
Elastic support
2. DCSFD model
Two kinds of DCSFD models were built. Outer film of one model
was based on the Reynolds equation of conventional SFD; inner
film was based on the general incompressible Reynolds equation.
It was named as model I. Outer and inner films of the other model
were all based on the Reynolds equation of conventional SFD.
It was named as model II. Therefore, model I and II were different
in inner film modelling.
Damper journal
DCSFD
Face A
Force sensor
Face B
Flexible link Transfer component
Damper journal
Shaker
where mB is the mass of damper journal; mF is the mass of 4.2. The test rig without DCSFD
floating-ring; ka is the stiffness of the elastic support; C r is the
viscous damping coefficient; (xB, yB) and (xF, yF) are the displace- The test rig without DCSFD was installed on two different
ments of damper journal and floating ring in the fixed coordinate supporting A and B. In the experiment, the excitation loads were
system; F is the amplitude of shaker forces applied to DCSFD; f SIx 5 N and 10 N with frequencies ranging from 10 to 100 Hz. Let f SIx
and f SIy are the oil film forces generated by inner film in the x- and and f SIy equal to zero, the displacement response of the damper
y-directions, f SOx and f SOy are the oil film forces generated by outer journal can be obtained by solving the first two equations of
film in the x- and y-directions. Eq. (5) with the Runge–Kutta method. Thus, the simulation results
The lubrication system included a delivery pump which sup- of the test rig without DCSFD can be obtained.
plied 32# hydraulic oil. The kinematic viscosities of oil were Figs. 8 and 9 depict the displacement response of damper
measured in Ref. [21] with different temperatures; the test results journal with different excitation frequencies in x- and y-direction,
were listed in Table 1. when the rig without DCSFD is installed on the supports A and B
H.L. Zhou et al. / Tribology International 66 (2013) 187–193 191
respectively, where the curves with dots denote the experimental denote the error between the model I and model II, E1 ¼
results; the curves with circles denote the simulation results. jðmodel I−model IIÞ=model IIj100%; Let E2 denote the error
As can be observed from Figs. 8 and 9, the experimental results between the model I and experimental results, E2 ¼
and simulation results are accordant. It can be seen from Fig. 8 that jðexperiment−modelIÞ=modelIj100%; Let E3 denote the error
there is a maximum of the displacement response in both between the model II and experimental results, E3 ¼
experimental and simulation results. The corresponding excitation jðexperiment−modelIIÞ=modelIIj100%. Figs. 10 and 11(c, d) depict
frequency is equal to 50 Hz, which is the resonance frequency of E1 , E2 and E3 .
the test rig installed on support A. It can also be seen from Fig. 9 It can be seen from Fig. 10(a, b) that the results calculated by
that there is a maximum of the displacement response in both the model I and II are accordant. The maximal value of E1 is
experimental and simulation results. The corresponding excitation 3.8462%; the minimal value of E1 is close to zero. Thus, excellent
frequency is equal to 80 Hz, which is the resonance frequency of consistency of numerical solution is achieved between two differ-
the test rig installed on support B. ent DCSFD models. As can be observed from Fig. 10(a, b), there are
For simplicity, the test rig can be seen as one degree of freedom some differences between simulation and experimental results.
system. Therefore, the resonancepfrequency
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi of the system can be The maximal value of E2 is 27.7778%; the minimal value of E2 is
obtained by the formula, ωn ¼ ka =mB =ð2πÞ. When the test rig close to zero; the maximal value of E3 is 25%; minimal value of E3
without DCSFD was installed on support A, ωn ¼ 47.094 7Hz. The is close to zero. The error may come from the simplified boundary
resonance frequency is close to 50 Hz, which is consistent with the condition, machining error, etc. However, most of the values of E2
experimental and simulation result. Similarly, when the test rig and E3 are small, and the general trends of simulation and
without DCSFD was installed on support B, ωn ¼ 82.6068 Hz. The experiment are accordant. Thus, good consistency is also achieved
resonance frequency is near to 80 Hz, which is consistent with the between experiment and simulation.
experimental and simulation results. To further validate the Model I and II, support B, an elastic
Therefore, the experimental results, simulation results and support with greater stiffness, was adopted. As can be observed
simplified formula results are accordant. from Fig. 11, the excellent consistency of the numerical solution is
achieved between model I and model II; good consistency is also
4.3. The test rig with DCSFD achieved between the experiment and simulation. Therefore, the
models established in the paper are further validated.
In order to validate the two DCSFD models built in the paper, a All calculations in the paper were performed in Matlab on a
test rig was installed on support A and B. Figs. 10 and 11(a, b) standard desktop pc with a 2.83 GHz quad core processor. The
depict the displacement responses of damper journal with differ- computation time of model I and II is listed in Table 3. It can be
ent excitation frequencies in simulation and experiment. Let E1 seen that the computation time of model I is 10 times more than
the model II. Thus, model II can save more computation time.
Table 1
However, the model II is based on the Reynolds equation of
Viscosity-temperature relationship of 32# hydraulic oil. conventional SFD. It is hard to consider the deformation of floating
ring in model II. Therefore, the model I provides the groundwork
Temperatures 20 30 40 48 57 76 85 95 to further perfect the model of DCSFD.
Viscosities 7.322 4.5702 2.7921 1.9002 1.3038 0.8385 0.7415 0.5715
(10−2 Pa s)
5. Conclusion
Fig. 8. Test rig installed support B without DCSFD (a) response of journal in x-direction and (b) response of journal in y-direction.
192 H.L. Zhou et al. / Tribology International 66 (2013) 187–193
Fig. 9. Test rig installed support A (a) response of journal in x-direction and (b) response of journal in y-direction.
Fig. 10. Test rig installed support B (a) response of journal in x-direction, (b) response of journal in y-direction, (c) error analysis in x-direction and (d) error analysis in
y-direction.
Fig. 11. Test rig installed support B. (a) response of journal in x-direction, (b) response of journal in y-direction, (c) error analysis in x-direction and (d) error analysis in
y-direction.
H.L. Zhou et al. / Tribology International 66 (2013) 187–193 193