You are on page 1of 10

Materials

& Design
Materials and Design 26 (2005) 680–689
www.elsevier.com/locate/matdes

Pressurised chamber design for conducting rolling


contact experiments with liquid refrigerant lubrication
Zulfiqar Ahmad Khan *, Mark Hadfield, Ying Wang
Sustainable Product Engineering Research Centre, School of Design, Engineering and Computing,
Bournemouth University, Poole House, Bournemouth BH12 5BB, UK

Received 4 February 2004; accepted 10 August 2004


Available online 15 September 2004

Abstract

A novel pressurised chamber was designed to construct a special purpose test rig to achieve saturated liquid state of refrigerants
for lubricated rolling contact fatigue experiments. Traditional bench testing using refrigerants as the lubricant is difficult due to the
gaseous phase at standard atmospheric conditions. Pressurising the concentrated contact test chamber is therefore necessary to eval-
uate at a liquid state and therefore simulate practical applications. The gas/liquid phase transitions have a significant influence on
the lubricant properties and hence wear mechanisms. It is necessary to modify the wear test conditions for the Hydrocarbon (HC)
and Hydrofluorocarbon refrigerants to obtain realistic simulation of refrigerator compressor tests. The chamber design and test rig
are described in this paper. A preliminary experimental study of the influence of the HC (R600a) on rolling wear of the silicon nitride
(Si3N4)/steel elements using the pressurised chamber is presented. Rolling fatigue test methods are adopted to measure the wear per-
formance of silicon nitride/steel bearing materials. In this case the rolling wear mechanisms of Si3N4 were measured using R600a
refrigerant lubrication.
 2004 Elsevier Ltd. All rights reserved.

Keywords: Pressurised chamber; Rolling contact; Ceramic bearings; Hydrocarbon; Refrigerant

1. Rolling contact fatigue test machines tions have shown that, except for very specialized condi-
tions of load or local chemistry, solid or liquid
Many rolling contact fatigue (RCF) test machines are lubrication is required to effectively utilize silicon nitride
in use to study the wear performance of materials and in wear applications. Dry, rolling wear rates on silicon
lubricants in rolling contact. In rolling contact, surface nitride were measured by Kato et al. [19,18]. Generally,
and subsurface initiated fatigue is observed due to re- wear occurs either by mechanical or chemical means and
peated loading. At critical number of fatigue cycles the has six principal, quite distinct phenomena that have
break-up of material surface results in the formation one thing in common, removal of material from the rub-
of small pits, also known as pitting or spalling. Several bing surfaces [6]. These types of wear mechanisms are
investigators, prompted by increasing application of sil- adhesive wear, abrasive wear, fatigue wear, impact wear,
icon nitride in bearing systems, have documented wear corrosive wear, electrical-arc-induced wear and fretting
and fatigue behaviour of these materials [23]. Investiga- and fretting Corrosion.

*
Corresponding author. Tel.: +44 1202 595 079; fax: +44 1202 595 1.1. Modified four-ball machine
314.
E-mail addresses: zkhan@bournemouth.ac.uk (Z.A. Khan),
mhadfield@bournemouth.ac.uk (M. Hadfield), ywang@bournemouth. The modified four-ball machine simulates deep
ac.uk (Y. Wang). groove ball bearing environment. This test configuration

0261-3069/$ - see front matter  2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2004.08.006
Z.A. Khan et al. / Materials and Design 26 (2005) 680–689 681

Collet Load
Load (L)

Collet
Upper ball

Lubricant
Lower ball Upper ball

Lower ball

Steel cup

Fig. 1. Loading configuration of modified four-ball machine. Fig. 2. Loading configuration of five-ball machine.

has three lower balls free to spin and revolve with an Load
upper fixed ball having one degree freedom and only
Ball or
translating spindle motion. The lower three balls revolve
Roller
in a cup that simulates the bearing outer race, while the
upper ball represents the inner bearing race. The lower
Ball Retainer
planetary balls representing the rolling elements within
the bearing. Loading geometry for a modified four-ball
machine is shown in Fig. 1. This machine has been very
Test piece
useful in the study of RCF performance of materials
subjected to various tribological conditions. The influ- Fig. 3. Loading configuration of ball-on-plate machine.
ence of lubricants on steel ball RCF was evaluated using
the machine [4,22,27]. The Institute of Petroleum com- loaded between a stationary flat and a rotating grooved
piled numerous papers [31] that describe various test re- washer. Rotating washers produce motion and transmit
sults, kinematics, ball dynamics using modified four ball load to the rollers/balls and the flat washer.
machine. This machine has been used to study the RCF
performance of hot pressed silicon nitride bearing mate- 1.4. Ball-on-rod machine
rials [29,28]. Recently, a modified four-ball machine has
been used to simulate the rolling contact in hybrid (cera- This machine has a cylindrical rod of 9.53 mm in
mic/steel) rolling element bearings [13–15]. diameter. A motor coupled in line with the rod drives
the rod. Combination of a lower and upper cup provides
1.2. Five-ball machine housing for three balls of 12.7 mm in diameter. Oil drip-
ping on the rod provides lubrication. Pre-calibrated
Fig. 2 shows contact geometry and rotating mecha- mechanical springs are used to apply load by driving
nism of a five-ball machine. The four lower balls are dri- the upper cup towards the lower. This machine was
ven in a race by an upper ball. This arrangement used to investigate rolling contact wear [3,8,10,23]. Ball-
simulates rolling and sliding of angular-contact ball on-rod machine is shown in Fig. 4.
bearings. This machine has been used to investigate
the fatigue life of high-speed rolling elements including 1.5. Disc-on-rod machine
hot pressed silicon nitride balls [9,26].
This machine is shown in Fig. 5. A long straight
1.3. Ball-on-plate machine cylindrical specimen 76 mm in diameter is held between
two discs 177.7 mm in diameter and 12.7 mm thick.
The ball-on-plate machine shown in Fig. 3 was used These discs can press against the specimen which is held
to investigate the RCF performance of silicon nitride in position by a spindle and can rotate up to 10,000 rpm.
[12,20]. This machine has a unidirectional thrust bearing The geometry of the machine enables to calculate the
consisting of three balls or roller positioned at 120 contact stresses as a function of load. The machine pro-
each. The balls/rollers are retained by a retainer and vides condition for rapid testing in nearly pure rolling.
682 Z.A. Khan et al. / Materials and Design 26 (2005) 680–689

Rod
specimen
Load

Ceramic
Balls roller

Driving
roller

Fig. 4. Loading configuration of ball-on-rod machine.


Fig. 6. Loading configuration for contacting ring.

determine the volume of wear. This machine was used to


Drip study the wear performance of ceramics [1,2,7].
lubrication
nozzle
2. Rolling contact fatigue testing methodology

2.1. Modified four-ball machine


Steel disc
The TE92 Microprocessor Controlled Rotary Trib-
ometer offers facilities to run high and low load tests
Test rod over a wide range of speeds, coupled with the latest con-
trol and instrumentation. The body of the machine is
Turnbuckle
cylindrical steel fabrication inside which is located the
and load test chamber, the loading piston and the drive spindle
cell and the associated bearings. The lower race assembly
is carried on the load piston. End load is applied to
the piston by means of a loading arm mounted at the
Fig. 5. Loading configuration for disc-on-rod machine. lower end of the cylindrical body. For high-speed tests
the lower test assembly, consisting of the high speed roll-
The specimen undergoes 12 · 105 stress cycles/h at ing contact race, the oil bath and heater block, are
10,000 rpm. Some researchers performed earlier work mounted directly on the load piston. For low speed test,
using disc-on-rod machine [5,24]. the assembly, consisting of the ball cup, oil bath, torque
arm and heater block, are mounted on a thrust bearing
1.6. Contacting ring machine located on top of the load piston thus allowing rotation
of the test assembly to allow torque measurement.
Contacting ring machine is shown in Fig. 6. This ma- The equipment consists of a bench-mounting test and
chine has two discs or rings that are rotated against each control interface. A microprocessor is connected to the
other. The outer contact surfaces for both the discs may controller interface. The core of the machine locates
be flat or one of the surfaces can be toroidal. The discs/ two rigid vertical columns, which ensures accurate posi-
rings are driven independently by separate motors or a tioning of a drive spindle with respect to the normal
gear mechanism between two discs may be employed loading axis. The test adapters are mounted on a cross
thus driving both discs at different speed. This helps roll- beam which is guided by linear bearings on the columns
ing contact with differing sliding/slippage at the surfaces. and loaded by a pneumatic actuator. The actuator
Load is applied through a compressed coil spring or assembly included an in-line force transducer measure-
dead weight mechanism. Dry or lubricated tests can be ment for direct feedback control of load. The test
conducted. The wear Profile on the contact surface helps adapters are mounted on a thrust bearing and rotational
Z.A. Khan et al. / Materials and Design 26 (2005) 680–689 683

movement is restricted by the strain gauge transducer, saturated liquid state of the refrigerant as fluid at oper-
thus giving direct friction torque measurements. The ating conditions by using the pressurised chamber. The
adapters also include integral electrically heated sup- tests were conducted in HC (R600a) and HFC
ports for tests up to 200 C. The software includes an (R134a) refrigerants as lubrication.
8-channel chart recorder simulation for data output to
a printer. At the end of the test the data may be manip- 2.3. Test rig modification
ulated for calculation or graphical presentation on the
integrated spreadsheet software programme. Tests are 2.3.1. Basic design
defined by a series of steps, each step containing load, The pictorial and the cross-sectional views of the
temperature and speed set-points, data recording and pressurised chamber are shown in Figs. 7 and 8, respec-
alarm level information. Set-points may be adjusted by tively. A coiled copper tube is fitted on the surface of the
step change or ramp. The test sequence is followed un- pressurised chamber. The cooling liquid is circulated in
less interrupted by the operator or an alarm. the tube with the help of a peristaltic pump. The cooling
circuit has a heat sink where the liquid is cooled down
2.2. Rolling contact fatigue testing with refrigeration and re-circulated on the pressurised chamber surface.
A schematic of the cooling arrangement is shown in
Refrigerant lubrication has a direct affect on the Fig. 14.
durability of the rolling element bearings within indus-
trial compressors. In addition, refrigerants have evolved
rapidly over the last decade due to sustainable develop-
ment legislation. Therefore, the tribology of the refriger-
ants [HC Hydrocarbon, (R600a, CH(CH3)3,2-methyl
propane (isobutane) and HFC tetrafluoroethane
(CH2FCF3, HFC-134a)] used in industrial application
systems must be studied. The refrigerant dissolved in
the oil brings about lubricating behaviour different from
that in an ambient environment, especially for boundary
lubrication [16,17]. The lubricants are easily deteriorated
in the presence of refrigerant because the two react
chemically [30,32].
Komatsuzaki and Homma [21] used a four-ball tester
to evaluate the extreme pressure (EP) properties of pure
refrigerants and oil refrigerant mixtures. They found
that R-134a in Polypropylene glycol gave higher rates
of wear than R-12 and R-22 in mineral oil. Although
the HFC-134a had showed some capability of forming Fig. 7. Isometric view of the pressurised chamber.
a fluoride layer on the sliding surface, it did not exhibit
good lubricating quality. The reason for that is the con-
dition for the HFC to form a protective layer during
sliding is more severe than in the actual compressors
[25]. The gas/liquid phase transition has a significant
influence on the wear mechanisms of traditional lubri-
cants and is therefore a useful experimental study [11].
It was necessary to simulate actual operating condi-
tion for rolling contact hybrid ceramic-steel ball bearing
with refrigerant as lubrication. A novel pressurised
chamber was designed and manufactured. Rolling con-
tact hybrid ceramic-steel tests can be performed with
refrigerant in liquid state without generating severe fric-
tional heat. TE92 Microprocessor Controlled Rotary
Tribometer offers facilities to run high and low load tests
over a wide range of speeds, coupled with the latest con-
trol and instrumentation. Test rig was fitted to the Ro-
tary Tribometer with novel pressurised chamber. The
pressurised chamber offers controlled conditions of tem-
perature(s) and pressure(s). It is now possible to get Fig. 8. Cross-sectional view of the pressurised chamber.
684 Z.A. Khan et al. / Materials and Design 26 (2005) 680–689

2.3.2. Cooling fin design


In this design fins were generated on the surface area
of the chamber. These fins help increase the contact
area. A continuous blow of cool air on the circumferen-
tial area of the pressurised chamber helps control the
pressure temperature conditions. The pictorial and the
cross-sectional views are shown in Figs. 9 and 10,
respectively.

2.3.3. Internal cooling design


This design is based on the concept of water cooling
jacket in the IC engines. The water jacket is designed for
the purpose of cooling the inside atmosphere by means
of running coolant. An entry and exit port helps the
coolant flow in and out from the pressurised chamber
body. The pictorial and the cross-sectional views are
shown in Figs. 11 and 12, respectively. Fig. 11. Isometric view of the pressurised chamber.

Fig. 12. Cross-sectional view of the pressurised chamber.


Fig. 9. Isometric view of the pressurised chamber.

2.4. Vapour phase commissioning of the test rig

The rig was constructed for the pressurised-chamber


to be fitted and pressurised with refrigerant. The
arrangement of the rig for pressurised-chamber is shown
in Fig. 13. The pressurised-chamber is fitted with two
clamps both on top and bottom of the chamber. The
top clamp has transparent polycarbonate plate to ob-
serve refrigerant state. The polycarbonate plate is
strengthened with an aluminium clamp having a circular
window to see through. The bottom clamp consists of
aluminium plate only. Both these clamps are bolted
together by eight studs. The circuit has a charging line
fitted with refrigerant supply. This line is fitted with
shut-off valve and a pressure gauge. A vacuum/discharge
line which can be connected to a vacuum pump when
the system needs to be vacuumed. The same line can
Fig. 10. Cross-sectional view of the pressurised chamber. be connected to recovery cylinder when the refrigerant
Z.A. Khan et al. / Materials and Design 26 (2005) 680–689 685

Fig. 13. Pressurised-chamber test rig.

is discharged. A compound gauge reads the pressure of Pressure (Bar, gauge) 5


the system. The pressurised-chamber is also fitted with a Observations
4 Theoratical
temperature gauge to read the temperature of the sys-
tem. The pressurised chamber was fitted to TE92 Micro- 3
processor Controlled Rotary Tribometer. A schematic 2
of the test rig is shown in Fig. 14. The test rig has three
1
systems, cooling system, refrigerant charging system and
vacuum system. 0
Fig. 15 shows visual observation of the saturation -10 0 10 20 30 40
conditions during the experiment for each specimen Temperature (Deg C)
relative to the theoretical conditions and shows that
Fig. 15. Saturation diagram for R600a.
the sate of the refrigerant is sub-cooled liquid.

Load (L)

Rotary
Chamber Tribometer
Upper Part Spindle
Coolant flow Coolant
Direction pipe
Pressure
Gauge
Heat
Exchange
Shaft
Seal Vacuum
Shut-off
Chamber Pump
valve
Lower Part
Collet Refrigerant
'O' rings Charging
Liquid Ceramic ball
refrigerant
Steel ball
Steel cup Thermocouple

Fig. 14. Schematic of rig for the high speed rotary tribometer.
686 Z.A. Khan et al. / Materials and Design 26 (2005) 680–689

3. Preliminary experimental results using the pressurised


chamber

During this test a novel pressurised chamber was


used to maintain the liquid state of the refrigerant
throughout the tests. The cooling system was em-
ployed to ensure lower operating pressures. Chamber
was vacuumed to 1 bar before pressurising with
refrigerant. Maximum chamber pressure was 3 bar
(gauge). Minimum chamber temperature was 5 C
to maximum 30 C. R600a was used in these tests.
Rolling contact fatigue tests were performed by using
High Speed Rolling Tribometer. The upper silicon ball
was held by means of a collet into the drive shaft of
the machine and brought into contact with three low-
Fig. 16. Light micrograph showing fatigue crack propagation.
er steel balls held together in a housing cup shown in
Fig. 14. Silicon ball was ground and polished to 12.7
mm diameter. The elastic modulus and Poisson ratio
of the ball are 310 GPa and 0.28, respectively. The
average roughness (Ra) of the silicon ball was 0.01
lm. The steel ball was carbon chromium steel with
12.7 mm diameter and a surface roughness (Ra) of
0.02 lm. The hardness of silicon ball was 1637 Hv
and that of steel ball was 839 Hv. The housing cup
with contact balls was then filled with pressurised
refrigerant. Preliminary test results are shown in Table
1.

4. Surface analysis of ceramic rolling elements

Fig. 16 shows light microscopic result of the test A. Fig. 17. Light micrograph failure due delamination.
This result was recorded at 4 h and 1 min. The shaft
speed was 2000 rpm and shaft load was 3 GPa. An obvi- Fig. 20 shows scanning electron micrographs for
ous crack development and propagation helps in delam- specimen C. The maximum contact pressure in this test
ination. Fig. 17 shows the failure in the lower indent as was 6 GPa with 2000 rpm spindle speed and R600a as
delamination. lubrication environment. The fatigue life of the speci-
Figs. 18 and 19 showing scanning electron micro- men was recorded to be 19 · 102 cycles while the time
graphs for specimen B. The maximum contact pressure to failure was 9 min. Fig. 21 shows surface fatigues dam-
in this test was 7 GPa with 2000 rpm spindle speed age for specimen D the maximum contact pressure was 6
and R600a as lubrication environment. The fatigue life GPa with R600a refrigeration. The fatigue life of the
of the specimen was recorded to be 83 · 102 cycles while specimen was recorded to be 13 · 103 cycles while the
the time to failure was 2 min. time to failure was 7 min.

Table 1
Test programme for preliminary experiments
Test Contact stress (GPa) Time to failure Indent load (kg) Shaft speed (rpm) Fatigue cycles
A 3 4 h 30 min 5 2000 5.39 · 105
B 7 2 min 5 3000 8.34 · 103
C 6 9 min 1 2000 1.93 · 104
D 6 7 min 1 2000 1.33 · 104
E 3 13 h 30 min 1 2000 1.61 · 106
F 3 6h C-crack 2000 7.19 · 105
G 6 0 h 42 min C-crack 2000 8.21 · 104
H 6 0 h 04 min C-crack 2000 7.71 · 103
Z.A. Khan et al. / Materials and Design 26 (2005) 680–689 687

Fig. 18. SEM showing surface fatigue damage. Fig. 21. SEM showing surface fatigue damage.

Fig. 19. SEM showing surface fatigue damage.

Fig. 22. SEM showing surface fatigue damage.

time to failure was 13 h and 30 min. Fig. 23 shows sur-


face fatigue damage for specimen F the maximum con-
tact pressure was 3 GPa with R600a refrigeration. The
surface had a pre-induced c-crack. The c-crack was posi-
tioned in the contact path as shown in Fig. 24(a). The
number of fatigue cycles to failure recorded were
72 · 104 and time to failure was 6 h. The maximum con-
tact pressure for specimen G was 6 GPa and spindle
speed was 2000 rpm. The test was carried out in liquid
state of R600a refrigerant. A C-crack was induced into
the specimen. The crack was positioned in the contact
path as shown in Fig. 24(a). The number of fatigue cy-
cles at which specimen failed were 82 · 103 and time to
Fig. 20. SEM showing fatigue spall. failure was 42 min.
Specimen H was tested in liquid R600a with a
C-crack. The maximum contact stress was 6 GPa.
Fig. 22 showing scanning electron micrographs for The crack was positioned in the contact path as
specimen E. The maximum contact pressure in this test shown in the Fig. 24(b). The fatigue life was 4 min
was 3 GPa with 2000 rpm spindle speed. The fatigue life and the number of fatigue cycles to failure recorded
of the specimen was recorded to be 16 · 105 cycles and were 77 · 102.
688 Z.A. Khan et al. / Materials and Design 26 (2005) 680–689

References

[1] Akazawa M, Kato K. Wear properties of silicon nitride in rolling-


sliding contact. Wear 1988;124:123–32.
[2] Akazawa M, Kato K, et al.. Wear properties of silicon nitride in
rolling contact. Wear 1986;110:285–93.
[3] Allen DL. Effect of composition and physical properties of silicon
nitride on rolling wear and fatigue performance. Tribol Trans
1994;37:410–14.
[4] Barwell FT, Scott D. Effect of lubricant on pitting failure of ball
bearings. Eng 1956:9–12.
[5] Baumgartner HR. Evaluation of roller bearings containing hot-
pressed silicon nitride elements. Brook Hill Publishing; 1974.
[6] Bhushan B. Introduction to tribology. Columbus: John Willey &
Sons, Inc.; 2002.
[7] Braza JF et al.. Silicon nitride wear mechanisms:rolling and
sliding contact. Tribol Trans 1989;34(4).
Fig. 23. SEM showing surface fatigue damage. [8] Burrier HI. Optimizing the structure and properties of silicon
nitride for rolling contact bearing performance 1996; 39: 276–
85.
[9] Carter TL, Zarretsky EV. Rolling contact fatigue life of a
crystallized glass ceramic, National Areonautics and Space
C-Crack Y Administration; 1960.
[10] Chao LY, Shetty DK. Reliability analysis of structure ceramics
subjected to biaxial flexure. J Am Ceram Soc
Contact
1991;74(2):333–34.
path [11] Ciantar C. Sustainable development of mechanical systems using
a X replacement environmentally acceptable refrigerants. DEC.
Bournemouth: Bournemouth University; 2001.
[12] Fujiwara T, Yoshioka T, et al.. Study on load rating property of
silicon nitride for rolling bearing material. J JSLE, Int Ed
‘A’ 1989:81–86.
[13] Hadfield M. Failure of silicon nitride rolling elements with ring
C-Crack Y crack defects. Ceram Int 1998.
[14] Hadfield M, Stolarski TA, et al.. Failure modes of ceramics in
rolling contact. Proc R Soc, Lond 1993.
Contact
[15] Hadfield M, Stolarski TA, et al.. Failure modes of ceramic
path elements with ring-crack defects. Tribol Int 1993;26:157–64.
a X [16] Hon-Ma Y, Komatsuzaki S. Problems in lubrication of refriger-
ating compressor lubricant. J JSLE 1983;28(12):887–90.
‘B’ [17] Itoh T. Experimental study of lubrication for refrigeration
compressors. Proc R Soc, Lond 1966;50(571):969–72.
Fig. 24. C-crack positioning in the contact path. [18] Kato K et al.. Wear properties of silicon nitride in rolling sliding
contact. Wear 1988:124.
[19] Kato K et al.. Wear mechanisms of ceramic materials in dry
rolling friction. ASME Trans 1986:108.
[20] Kikuchi K, Yoshioka T, et al.. Rolling contact fatigue life of
5. Conclusion ceramics for rolling bearing materials. J JSLE, Int Ed
1984:137–42.
[21] Komatsuzaki S, Homma Y. Antiseizure and antiwear properties
1. The design, manufacture and successful commission-
of lubricating oils under refrigerant gas environments. Source
ing of the pressurised chamber and test rig modifica- Lubr Eng 1991;47(3):193–98.
tion of the Rotary Tribometer provides an [22] Krivoshein GS. On evaluating the effect of lubricants on the
opportunity to conduct rolling hybrid (ceramic/steel) fatigue pitting of steel. Ind Lab 1960;26:405–7.
contact tests in sub cooled refrigerant environment. [23] Lucek W John. Rolling wear of silicon nitride bearing materials.
In: International gas turbine and aeroengine congress and
2. The liquid state of the refrigerant is important to sim-
exposition. CERBEC Ceramic Bearing Co, East Granby, USA,
ulate practical condition. Also lubrication film will Belg, GT165. New York: ASME; 1990. p. 7.
exist that satisfies the boundary lubrication [24] Lucek JW, Cowley PE. Investigation of the use of the ceramic
conditions. material in aircraft engine bearings, US Deparment of Navy;
3. Gaseous phase of the refrigerant may result in severe 1978.
[25] Mizuhara K, Matsuzaki T. The friction and wear behaviour in
wear during rolling contact fatigue of hybrid ceramic/
controlled alternative refrigerant atmosphere. Tribol Trans
steel contact. 1994;37(1):120–28.
4. Positioning of the crack in the contact track has influ- [26] Parker R, Zarretsky EV. Fatigue life of high-speed ball bearings
ence on the fatigue life. with silicon nitride balls. J Lubr Technol 1975:350–57.
Z.A. Khan et al. / Materials and Design 26 (2005) 680–689 689

[27] Scott D. Study of the effect of lubricant on pitting failure of balls. [30] Spauschus HO, Doderer GC. Reaction of refrigerant 12 with
Lubr Wear Convention, I Mech E 1963. petroleum oils. ASHRAE J 1961;2:65–9.
[28] Scott D, Blackwell J. Hot pressed silicon nitride as a rolling [31] Tourret R, Wright EP. Rolling contact fatigue: performance
bearing material – a preliminary assessment. Wear 1973;24:61–7. testing of lubricants. London: Heyden and Son Limited; 1977.
[29] Scott D, Blackwell J, et al.. Silicon nitride as a rolling bearing [32] Walker W, Levy SL. Stability of mixtures of refrigerants and
material – a preliminary assessment. Wear 1971;17:73–82. refrigerating oils. ASHRAE Trans 1962;68:360–89.