Você está na página 1de 8

Tribology International 36 (2003) 807814

www.elsevier.com/locate/triboint

A FEM approach to simulation of hydrodynamic-pad thrust bearing assemblies


D. Markin, D.M.C. McCarthy , S.B. Glavatskih
University of Technology, 971 87 Lulea , Sweden Division of Machine Elements, Lulea

Abstract

Finite-element method (FEM) modelling is applied to analysis of the performance of hydrodynamic-pad thrust bearing assemblies. A 3D model of the bearing assembly that includes the bearing pad and shaft is used to assess the inuence of operating conditions on bearing parameters such as temperature and oil lm distributions across the pads. The model is rst applied to the investigation of a spherically pivoted-pad. Through comparison with results from experiments carried out on just such a bearing, good correlation between the model and experimental results is found for maximum oil lm temperature, pressure distribution and thickness. The model is then applied to the examination of a bearing having spring-supported babbitt pads. The effect of different oil types on a spring-supported thrust bearing is analysed. Further application of the model to investigate the same spring-supported pad, this time with a resilient surface coating, is discussed. 2003 Elsevier Ltd. All rights reserved.
Keywords: Tilting-pad bearings; Synthetic oil; Finite-element method; PTFE

1. Introduction Thrust bearings are a major component in the turbine assemblies of hydroelectric power stations. Friction in the oil lm separating bearing thrust surfaces and the shaft contributes a considerable amount to power losses, reducing the efciency of the power plant. Through the introduction of synthetic oils to replace existing mineral oils, these losses can be reduced. It has been shown that ester base oil ISO VG46 can be an efcient substitute for mineral oil ISO VG68 in medium sized pivoted-pad thrust bearings [1]. In large spring-supported thrust bearings, thermal deformations are greater, so additional theoretical analysis is required before such recommendations can be made. Substantial amounts of oil are used for lubrication and cooling purposes in such large spring-supported bearings, potentially posing a considerable risk to the environment should they be released, whether through accident or system failure, into the waterways. Certain synthetic oils, having minimum adverse environmental
Corresponding author. Tel.: +46-920-492-854; fax: +46-920491-047. E-mail address: donald.mccarthy@mt.luth.se (D.M.C. McCarthy).

impact whilst retaining the essential properties for lubrication and cooling, can provide an attractive alternative to commonly used mineral oils. The objective of this study is therefore to develop a working nite-element method (FEM) model of the entire shaft and bearing assembly that can be applied to thrust bearing analysis. In doing so, the inuences of bearing design and oil type on the operating characteristics can be charted and the capabilities of the software assessed. Two different bearing types, spherical pivot and spring-supported, are investigated in this study. The addition of pad facing materials to the thrust surfaces provides an extra factor inuencing the performance in the case of the spring-supported bearing.

2. Modelling In this study, analysis of bearing operations is made through carrying out simulations on models created using commercially available FEM software [2]. FEM is useful since more complicated, 3D model geometries can be generated than is the case with FDM. Although there are a number of software packages available, we chose this particular option in order to assess its effectiveness in this application.

0301-679X/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0301-679X(03)00097-5

808

D. Markin et al. / Tribology International 36 (2003) 807814

Nomenclature a l k m h3 Tbath n H Hc y cv cm QV in QV out QV R E r Tin, TR, thermal expansion coefcient (K1) heat ow per unit area (W m2 K1) thermal conductivity (W m1 K1) viscosity (mPas) oil lm thickness at outlet (m) oil bath temperature (C) Poissons ratio () convection coefcient (W m2 K1) convection coefcient at trailing edge, outer corner (W m2 K1) hot oil carry over fraction () specic heat per unit volume (J m3 K1) specic heat per unit mass (J kg1 K1) oil ow at the inlet node (m3 s1) oil ow at the outlet node (m3 s1) reservoir oil ow to inlet node (m3 s1) Youngs modulus (Nm2) density (kg m3) Tout temperatures at the inlet, reservoir and outlet nodes (C)

In the rst stage, the model of the pivot-mounted pad is created in order to establish whether it is possible to accurately describe a tilting-pad thrust bearing using this software. Calibration of the model is carried out through comparison with experimental data recorded during tests reported in Ref. [1], including information relating to the use of four different oils. A second model is created to investigate a large bearing with spring-supported pads. This model allows for the inclusion of a surface facing. Actual pad dimensions are given in Table 1. Both models are simplied through rotational symmetry to permit concentration on only one individual thrust pad. The bearing pad and the shaft segment are modelled by means of 3D solid elements, making allowance for thermal and mechanical deformations in both components. The temperatures calculated are used for determination of the thermal displacements and stresses. A 2D lm element is used for representation of the oil lm. Calculated values belong to the ctitious midsurface plane located halfway between the top and botTable 1 Dimensional details of the two pad types Support type Outer radius (mm) 114.3 Inner radius (mm) 57.15 No. of pads Pad Pad angle () thickness (mm) 50 28.58

Spherical pivot, 60% offset Spring support

tom surfaces of the lm. This mid-surface is comprised of ctitious nodes having pressure and temperature values associated with the nearest pair of top and bottom surface nodes. The mid-surface temperatures are determined through a discrete form of the oil lm energy equation. Using a discrete form of the Reynolds equation, pressure is calculated based upon the surface velocity specied for the lm top surface, pressure boundary conditions, the deformed lm thickness and the temperature and pressure dependent viscosity. The pressure boundary conditions for a lm surface are specied so that the pressure around the oil lm edges is equal to zero. No consideration is made for any squeeze effect due to motion in the normal direction. The pressure distribution and the specied surface velocity determine the shear stresses and the volume ow of the oil. A calculation ow chart illustrating data exchange during simulation of a bearing is shown in Fig. 1. The lm element is considered to be non-linear and full Newtonian equilibrium iterations are always performed. Each iteration consists of calculation of the increments in pressure, temperature and displacement. The applied force/moment tolerances govern the convergence since a small change in pressure or temperature usually means a relatively large change in forces and moments. Boundary conditions for the convective heat ow are dened by specifying the temperature distribution over the oil lm inlet. The inlet node temperature is found using the following equation [3]: Tin
V cRQV R TR coutyQoutTout V cinQin

475

312.5

12

23.5

61

(1)

D. Markin et al. / Tribology International 36 (2003) 807814

809

Fig. 1.

Data ow chart.

y is the portion of the volume ow at the outlet nodes that is carried over to the ow at the inlet nodes. The remaining portion of the inlet ow is taken from the reservoir node. This has to be specied within the model. The lm element transports heat by means of conduction in the transverse direction driven by the temperature difference between the top and bottom surface nodes and the ctitious mid-surface nodes. The heat conducted to the top and bottom surfaces is transported further by conduction within the pad and the shaft adjacent to the oil lm. Since the lm element is 2D, verication of the code is required. Due to availability of a complete set of experimental data for a pivoted-pad bearing, the rst model prepared is for this type. The entire model assembly is shown in Fig. 2. The body of the pad consists of several blocks. This allows for setting of various characteristics such as thermal boundary conditions on pad surfaces for different parts of the pad. An irregular mesh is used in order to achieve better coincidence between the pad geometry and node location. Thermal boundary conditions set for the pad include heat transfer by convection to the ambient medium and the surrounding oil and air temperatures. The model geometry differs somewhat from that of the actual pad in order to permit simplication of the model through reduction of the number of nodes and elements. However, the area of the bearing load carrying surface has remained unaltered. In the model, the lack of material at the outer corners of the underside (in-cuts) is not considered. Such an omission does not signicantly affect stress or temperature distributions within the pad because deciencies are located in non-critical regions having low values for temperature and pressure. Fig. 3 shows an example of the contour map output

Fig. 2. Model assembly.

from the model illustrating distribution of temperature within the pad. Since the pivot contact area is small and can be assumed to be a point, the spherical pivot is modelled as a single point located on the underside face of the pad. The angle of inclination of the pad is free to change in all directions. The spring-supported pad model is built in the same way as for the pivoted-pad bearing. An isotropic, thermo-elastic steel material is used for the pad and the

810

D. Markin et al. / Tribology International 36 (2003) 807814

Fig. 3. Temperature distribution in the pad at 1500 rpm, 2.0 MPa with mineral oil ISO VG46.

shaft segment. An additional layer added on the top of the pad simulates the 2 mm thick facing. The software package used does not allow for handling of cavitation or sub-ambient pressures in the simulations. Hence it was not possible to simulate such a material as PTFE. However, it was possible to investigate the effect of a resilient material, having the same thermal properties as PTFE, on the bearing operating characteristics. The properties of this resilient material are shown in Table 2. More details are available elsewhere [4].

3. Results and discussion 3.1. Pivoted-pad bearing A typical loadspeed combination for the pivoted-pad bearing (1 MPa at 1500 rpm) is chosen in order to identify thermal boundary conditions such as convection coefcients, hot oil carry over factor and oil bath temperature. The tuning procedure includes variation of these parameters in order to nd the best correspondence between experimental and theoretical values. Experimental data referred to in this study were obtained from bearings mounted in a test rig as described in Ref. [5]. The value of H for the pad is varied between 100 and 500 W m2 K1 whereas, for the outer corner near the
Table 2 Material properties Parameter E (Nm2) n 0.3 0.3 0.3

trailing edge, Hc is varied between 100 and 1000 W m2 K1. The value for y is taken in the range 0.70.9 and Tbath in the range 5060 C. Other conditions applied include the oil ow rate to the bearing being kept constant, as is the temperature value for oil supplied to the bearing in the oil bath, maintained at 50 C for all simulations. Similarly, ISO VG46 mineral oil is used in all simulations. The characteristics of this oil are shown in Table 3. Increasing Hc gives a lower rate of circumferential temperature rise through the diminution of temperature at the trailing edge. Tbath and y have an equal effect on temperature levels: a decrease in their values gives lower calculated temperatures. y has a more pronounced inuence on lm temperatures. The best agreement for calculated and measured temperature distribution is observed for the following combination of the calibration variables: Tbath 50 C; H 100 W m2 K1; Hc 500 W m2 K1; y 0.9. These values are then used in other simulations. Computed and measured temperature results are plotted in Fig. 4. Tmid is the theoretical and T4, T5 and T6 are the experimental temperatures at the pad mid-width. In the experiments, thermocouples were positioned approximately 3 mm below the pad working surface. Theoretical temperatures are shown for the corresponding plane. T75/75 and T10 are, respectively, the theoretical and experimental temperature at the 75/75 location. This is dened as a point at 75% of the radial width from the inner radii of the pad at a position, circumferentially, 75% around from the inlet edge. An additional temperature, Toil is included in the diagrams. This theoretical temperature describes thermal conditions along the pad radial centre line, at the interface between the oil lm and the pad. Some deviation between simulated and experimental temperature values is still visible, especially at the leading edge. All simulations showed relatively good agreement between theoretical and experimental values for oil lm thickness. Calculated values for power loss are lower than the experimental values. This is explained by the fact that estimation of churning losses is not included in the model. Fig. 5 shows variation in h3 and bearing power loss for the compete bearing for different load combinations at 1500 rpm.

a (106 K1) 12 23 170

k (W m1 K1) 47 55 0.24

cm (J kg1 K1) 500 ~230 1050

Steel (pad backing and 2.1 1011 shaft) Babbitt 5.2 1010 Resilient material 5.0 109

D. Markin et al. / Tribology International 36 (2003) 807814

811

Table 3 Oil characteristics At 40 C r m cv (106) k ISO VG46 mineral 855 39.5 1.74 0.13 ISO VG46 PAO 819 39.8 1.82 0.145 ISO VG46 ester 906 42.7 1.87 0.173 ISO VG68 mineral 861 60.7 1.74 0.13

Fig. 4.

Temperature proles at 1500 rpm/2.0 MPa. Fig. 6. Pressure prole for different oils at 1500 rpm/2.0 MPa.

Fig. 5.

Power loss/minimum lm thickness at 1500 rpm.

Fig. 6 shows oil lm pressure distributions in the circumferential direction at the 25% and 75% positions of the pad width, measured radially from the pad inner radius. Pressure proles are well predicted by the model

therefore only one load case is considered in order to illustrate this. To verify against experimental ndings and TEHD results [1] of how oil type affects bearing operating parameters, simulations are carried out for two load cases, 0.5 and 2.0 MPa, and two rotation speeds, 1500 and 3000 rpm. Properties of the oil used are shown in Table 3. Fig. 7 shows temperature variations for different oils for a shaft speed of 1500 rpm and specic bearing load of 2 MPa. The results show similar temperature distributions with oil type as in the experiments. Ester base oil VG46 gives slightly higher temperatures than poly-olen (PAO) oil VG46 and mineral oil VG46, which have approximately similar temperature proles. The highest temperature is observed for mineral oil VG68. Theoretical power loss and minimum lm thickness for a complete bearing with different oils at a shaft speed of 1500 rpm and two bearing loads are shown in Fig. 8. The tendency is the same as in the experiments. Comparison of calculated and measured data show that variations as well as absolute values of h3 are well predicted by the current model.

812

D. Markin et al. / Tribology International 36 (2003) 807814

Fig. 7.

Temperature prole for different oils at 1500 rpm/2.0 MPa.

Fig. 8. Power loss and minimum lm thickness for different oils at 1500 rpm.

3.2. Spring-supported pad The inuence of pad facing is the rst thing to be examined. Oil used in this instance is mineral oil ISO VG46. Operating conditions are assumed as follows: specic bearing load 3.8 MPa, rotational speed 600 rpm. Additionally, thermal conditions are: T bath = 55 C, y = 0.9, H = H c = 100 W m2 K1. The value of H for shaft surfaces exposed to air is assumed to be 10 W m2 K1 [6]. It is further assumed that the temperature of air coming into contact with the shaft inside the bearing housing is the same as that of Tbath. Results of simulations for bearings with facings are shown in Fig. 9. From these, it can be seen that maximum lm temperature for the resilient material is 7.2 C higher than that for babbitt. Additional results show that maximum temperature of the pad backing decreases from 104.2 to 68.4 C. This is an important feature since thermal deformations of the pad are reduced, consequently increasing bearing load capacity.

The lower maximum pressure provides an additional improvement. A decrease in maximum lm pressure by 10% indicates that the allowable maximum bearing load can be increased. Calculated power loss for the resilient material facing is 41.1 kW, 2.5 kW less than that for the babbitted bearing. An increase in oil lm temperature and a corresponding decrease in power loss are explained by lower heat conductivity of the resilient material. Minimum lm thickness is only slightly decreased, by 0.9 m, an acceptable result. Similar data were reported in Ref. [7] for a spring-supported PTFE composite faced bearing. The same lm thickness and 30% lower maximum pressure were measured in the PTFE composite bearing compared to the babbitted bearing. It can therefore be assumed that a further reduction in Youngs modulus can provide lower maximum pressure whilst retaining the same lm thickness. No signicant difference in bearing operating parameters is observed either for pads with babbitt facing or those without any facing material. The differences in oil lm temperature and thickness are within 1 C and 0.4 m, respectively. In addition to the utilisation of a resilient facing, bearing performance may be further improved by choosing synthetic oils. The same oils are used in these simulations as in the case with the pivoted-pad bearing. Tbath is taken to be equal to 35 C. The data obtained are assembled in Table 4. Analysis of the results shows that variation in minimum oil lm thickness and power loss with oil type is the same as in the pivoted-pad bearing. There is only a small difference in calculated minimum lm thickness for ester base oil VG46 and mineral oil VG68. It should be noted that gures for power loss quoted in Table 4 relate purely to shearing losses and do not make allowance for churning as this was not included in the models. It is known that churning losses depend on the viscosity of the oil in the oil bath: higher viscosity leads to higher losses and vice versa. If churning is factored in to the results given in Table 4 for shearing power losses, then the total losses for the oils will differ noticeably. Hence total losses will be lower for the oils with lower viscosity grade. Film thickness is also often an important factor when assessing the suitability of oils, primarily from the point of view of bearing safety. Ester base oil provides a thicker oil lm than mineral or PAO oils of similar viscosity grade. Therefore the ester base oil can thus be judged to be an efcient replacement for the mineral oil of higher viscosity grade.

4. Conclusions A working FEM model of the entire shaft and bearing assembly has been developed and applied to thrust bearing analysis. Thermal and mechanical deformations in

D. Markin et al. / Tribology International 36 (2003) 807814

813

Fig. 9. Table 4 Results for different oils Tbath=35 C, y=0.9

Distribution of temperature and pressure (same parameters in both cases).

ISO VG46 mineral 10.86 25 92.1 88.6 87.2 52.8

ISO VG46 PAO 10.85 26 92.4 89.0 87.7 56.4

ISO VG46 ester 10.85 28 93.6 90.4 89.0 61.2

ISO VG68 mineral 10.38 30 92.0 87.7 86.4 62.4

Maximum lm pressure (MPa) Minimum lm thickness (m) Maximum lm temperature (C) Maximum facing temperature (C) Maximum pad backing temperature (C) Power loss (kW)

the bearing assembly as a whole were allowed for. Results from the simulation showed general agreement with those for the experiments. However, in order to increase the accuracy and capabilities of the model, 3D treatment of the lm element is necessary together with taking cavitation into account. Based on the results obtained, the following conclusions can be drawn: FEM simulations show that ester base oil VG46 can efciently replace mineral oil VG68 in middle and large sized tilting-pad thrust bearings without sacricing bearing safety. However, additional tests are required for other bearing designs or for centrally pivoted-pads.

Utilisation of a resilient facing in large spring-supported bearings allows noticeable improvement in the bearing performance. Compared to babbitted bearings, maximum lm pressure and power loss are reduced whilst minimum lm thickness remains essentially the same. In addition, the pad backing temperature is much lower, leading to lower pad thermal crowning and ultimately higher load carrying capacity. The introduction of a babbitt pad facing in the FEM model has negligible effect on operating parameters.

814

D. Markin et al. / Tribology International 36 (2003) 807814

Acknowledgements This research was supported nancially by the Swedish Institute (Svenska Institutet) through the provision of a scholarship to D. Markin and by Alstom Power. Additional assistance from Gunnar Larsson at SOLVIA Engineering AB and Tomas Arvidsson at Alstom Power is also gratefully acknowledged. References
[1] Glavatskih SB, Fillon M, Larsson R. The signicance of oil thermal properties on the performance of a tilting-pad thrust bearing. J Tribol 2002;124(2):37785.

[2] SOLVIA Finite Element System software package. SOLVIA Engineering AB, Sweden. [3] Ettles CM. Hot oil carry over in thrust bearings. Proc Inst Mech Eng 1969-1970;184(Part 3L):7581. [4] Markin D. Simulation of spring-supported thrust bearing with use University of Technology, Lulea , of FEM. Masters Thesis, Lulea Feb 2002. [5] Glavatskih SB. Laboratory research facility for testing hydrodynamic thrust bearings. Proc Inst Mech Eng 2002;216(Part J):105 16. [6] Ali El-Saie YMH, Ferner RT. Three-dimensional thermoelastohydrodynamic analysis of pivoted pad thrust bearingsPart 2: application of theory and comparison with experiments. Proc Inst Mech Eng 1988;202(C1):5162. [7] Uno S et al. Overview of recent tendencies in thrust bearings for hydrogenerators. Jap J Tribol 1997;42(2):20521.

Você também pode gostar