Real time simultaneous in-line wear and lubricant condition monitoring

303

Wear, 123 (1988) 303 - 312

REAL TIME SIMULTANEOUS IN-LINE WEAR AND LUBRICANT CONDITION MONITORING P. W. CENTERS and F. D. PRICE Lubrication Branch, Aero Propulsion Laboratory, Air Force Wright Aeronautical Laboratories, Wright-Patterson Air Force Base, OH 45433 (U.S.A.) (Received February 25,1987;

revised August 25,1987

; accepted November 20,1987)

Summary An in-line complete oil breakdown rate analyzer (COBRA) and quantitative debris monitor (QDMR) were installed in the oil system of an aircraft turbine engine lubricant simulator. The in-line COBRA readings correlated well with those of a standard laboratory model COBRA and total acid number of the deteriorating lubricant. Increases in QDMR particle counts corresponded with increases in trace iron content which reflected abnormal wear occurring in the simulator. Analytical and direct reading ferrographic data confirmed the rapid increase and eventual modest decrease in wear debris concentration. After test completion it was determined that a mainshaft bearing had experienced abnormal wear. Thus for the first time successful capability for real time simultaneous in-line wear and lubricant condition monitoring was demonstrated.

1. Introduction Several diagnostic methods are in use today for monitoring the condition of military aircraft gas turbine engines and lubricants. Generally, oil samples are analyzed in the laboratory using several methods. The information from laboratory analysis is invaluable in determining whether an engine is experiencing abnormal wear or whether the lubricant is degraded to a point beyond which severe corrosion and coking can occur. However, oil sampling and laboratory analysis are manpower intensive and, frequently, the time required for data generation and evaluation is unacceptable. Equipment readiness would be more quickly established and maintenance costs reduced if in-line monitors were developed to eliminate some or all of the required laboratory analyses. One instrument that has been used with success in monitoring turbine engine lubricant condition is the complete oil breakdown rate analyzer (COBRA) [I]. It has been documented that COBRA readings trend closely with a basic indicator of lubricant condition, total acid number (TAN) [2]. 0043-1648/88/$3.50

@ Elsevier Sequoia/Printed

in The Netherlands

304

An in-line COBRA has been developed and results of an initial evaluation are reported here. A commercial in-line inst~me~t used today in limited applications for diagnosing turbine engine wear is the quantitative debris monitor (QDMR), a “smart” chip detector which detects ferrous debris particles (greater than 250 pm) in the oilstream. Large and total counts of particles are summed continuously for digital presentation and evaluation, In the laboratory, the trace metal concentration of an oil is determined using methods of the spectrometric oil analysis program (SOAP), using either emission or atomic absorption spectrometric techniques, Much smaller debris particles can be detected by these methods but large particles are not efficiently analyzed. As such, QDMR particle counts may not correlate well with trace iron concentration [ 33 in some wear situations. With that knowledge, a QDMR sensor was evaluated concurrently with the in-line COBRA to demonstrate simultaneous in-line engine and lubricant health monitoring capabilities. 2. Experimental details A test stand aircraft turbine engine bearing and lubricant simulator [4] was used to stress a polyol-ester-based turbine engine lubricant to predict the performance of that lubricant in turbine engine applications. The simulator consists of the number 4 and 5 bearing compartment of a 557 gas turbine engine driven and heated electrically, so that with proper air flow, realistic turbine engine conditions are simulated. It is computer controlled for unattended operation and automatic~ly samples the lubricant at 5 h intervals. Collected oil samples were sent to a laboratory for determination of viscosity at 40 “C, TAN, trace metal concentration for 14 different metals, and COBRA analysis. Analytical ferrograms and direct reading ferrographic ratings for selected samples were obtained. None of the laboratory information was available until after the simulator test was completed. An in-line COBRA unit and a QDMR unit were installed on the simulator prior to the test. The in-line COBRA unit was mounted in a bypass line at the exit of the shell and tube lubricant cooling heat exchanger. The metal tubing delivering the lubricant to the COBRA detector resulted in lubricant temperatures being reduced by air cooling. The QDMR was located initially at the scavenge pump outlet. The QDM” was moved downstream after about 115 test h because abnormally high particle counts were being recorded. It was believed that the high counts might be the result of air bubbles and churning at the pump outlet. In-line COBRA and QDMR values were recorded at 5 h intervals. 3. Results and discussion TAN and laboratory COBRA readings increased rather consistently with test time as shown in Fig. 1. This is to be expected since these indica-

40

0 TEST

TIME

tHR)

Fig. 1. Lab COBRA readings

TEST (0),

viscosity

TIME

(HR)

(0) and TAN (A) of lubricant during test.

Fig. 2. In-line COBRA readings during test.

tions of oil condition have been shown to be related previously [2]. Readings from the in-line COBRA unit also increased during the test as shown in Fig. 2, though the increase in the in-line unit’s readings was not as steady as the rise in laboratory COBRA readings. The apparent stability of readings at 120 - 150 test h followed by a large jump at 155 test h reflects the inability of this particular instrument to register values above about 80 on the 0 - 100 scale. After some delay in noting the instrument’s limited capability, the scale was changed to the 0 - 200 scale. Nevertheless, there appears to be a linear correlation between the in-line COBRA readings and the laboratory COBRA readings as shown in Fig. 3. There was no apparent temperature effect; because of the placement of the in-line COBRA, readings were taken at 100 + 5 “F, which is well within the instrument’s capability, while laboratory COBRA readings were taken at room temperature. Additionally, Fig. 4 shows that the in-line COBRA readings increased proportionately to TAN, similar to laboratory COBRA readings which increase as TAN increases [ 21. QDMR particle counts increased dramatically early in the test, as did the trace metal concentrations of iron, copper, silicon and silver in the lubricant as determined by SOAP analysis. Trace iron concentration and QDMR counts are plotted us. time in Fig. 5; trace copper, silicon and silver concentrations are plotted us. time in Fig. 6. The trace iron concentration of

306

160

-

s

I

40 LAB

Fig. 3. Correlation 160

I

60

120

COBRA

160

of in-line COBRA

I

I

:

0

READING

and lab COBRA readings.

I

I 0

160

0

0

140

-0

-

&co 0

Q)

-

% 4 6

120

-

100

-

iz 5 s g

60-

8

z

aB_

% 60

0

m 0

40-cfpo=J 8 20

0

-0

0

I 0.2

I 0.1 TOTAL

ACID

I 04

I 0.3 NUMBER

Fig. 4. In-line COBRA

0.5

~Q/KOH/Q)

vs. TAN.

the oil increased dramatically for the first 155 h of the test. The trace silver concentration peaked 15 h into the test, followed by large increases in trace silicon and copper concentration. The dramatic increases in QDMR counts and trace concentration of iron, copper, silicon and silver were related to abnormally high wear found

24

Fig. 5. QDMR counts

and trace iron concentration

(0),

(A) during test.

40 30

20

10 9 0 7 6 5 4 3

2

IL

'0

I

50

I

I

100 TEST

TIME

150

d

I

200

250

(HRS)

Fig. 6. Trace wear metal concentration

during test:

0,

copper; A, silicon; 0, silver.

308

in post-test inspection of the number 5 bearing inner race and separator. The bearing is shown in Fig. 7. The separator is made of silicon bronze coated with silver; the remainder of the bearing is SAE 52100 steel. In similar simulator tests, bearing separator and race wear have been insignificant. In this test, however, not only was the solid lubricant silver coating worn away, but also substantial wear of both the separator and the bearing lands was found. The wear of the separator and inner race are shown in Figs. 8 and 9.

Fig. 7. Number

5 bearing.

Fig. 8. Wear of number

5 bearing separator.

Fig. 9. Wear of number

5 bearing

inner race

The wear of the inner race was asymmetric; the depth of the wear tracks varied from less than 25 pm at one point to greater than 125 pm at the diametrically opposite point, indicating possible separator imbalance resulting in periodic infractions between the separator and the inner race. The high trace silver concentration of the oil at 15 h, followed by sharp increases in trace copper and silicon concentrations, indicates that the abnormal wear began occurring very early in the simulator test. It is believed that eventually enough material was worn away to allow sufficient clearance between the bearing separator and the inner race land. Then, the loss of

309

debris to cavities within the lubrication system became greater than the generation of wear debris, resulting in an overall shift in equilibrium to a lower concentration of wear debris in the lubricant. The trace iron concentration and QDMR counts increased in approximately the same proportion, as can be seen in Fig. 10. These data do not indicate that QDMR counts correlate with trace iron concentration in all wear situations, because the QDMR detects large particles, and trace iron analysis generally measures very small particles and dissolved metal. The data do indicate that both QDMR and SOAP analyses were able to detect the abnormally high wear which occurred in this test. 32

I

I

I

I

I

0

0 26

24

0

0

o^ ‘0 z

20 6)

z

0

$

16

B a

12

0

0*

0 0 0

I -

0 )0

0

I 30

I 60 _

I SO

I

I

I

I

120

150

180

210

0

Fe fpvmf

Fig. 10. Total QDMR counts us. iron wear metal concentration.

Ferrograms of several oil samples diluted 9:l of original concentration were prepared. Scanning electron micrographs of several key ferrographic entry deposits are presented in Fig. 11. The ferrograms show that a high level of severe wear debris was present in the lub~cation system after only 10 test h. Increasing amounts of bronze and steel rubbing and severe wear particles were present up to 155 test h. The 185 test h ferrogram shows that less debris was present than at 155 test h. Wear debris exhibited on the ferrograms included large quantities of normal ferrous rubbing wear (about 2 - 7 pm in major dimension and less than 1 pm thick), a smaller number of large severe wear particles (30 40 pm X 5 - 10 pm thick) and a few cutting wear particles. The observed debris suggest a wear mechanism in which hard abrasive residue from debris built-up in the simulator was trapped in the silver coating in the inner

(a)

(d) Fig. 11. Scanning efectron micrographs of ferrographic entry deposits: (a) 0 h; (b) 10 h; (c) 155 h; (df 185 h.

raceway. Eventually, the surface roughness and the loss of both the silver coating and some inner race material led to wear resulting from sliding of the separator under periodic heavy loading against the inner race. Direct reading ferrographic results [5] of the samples presented in Fig. 11 are given in Fig. 12. Again, samples were diluted 92 of original concentration. The same trends found in the analytical ferrograms were observed in the direct reading values, Small and large particle concentrations and the wear severity index all increased up to 155 test h, at which point the wear moderated. Evidently, clearances in the worn area of the number 5 bearing became large enough at 155 test h for wear to decrease.

4. Conclusions (1) The in-line COBRA unit successfully monitored the condition of the lubricant on a real time basis. The in-line COBRA readings correlate well with laboratory COBRA readings and TAN.

311 1C

0 0

10'

l(r

5 Ei

3

10' 0

!s

0 A

2i

A

0 A

,

101

10

1

0

Fig. 12. Direct reading ferrographic results during test.

(2) The QDMR detected the generation of wear within the simulator. Increases in trace iron concentration corresponded with QDMR particle count increases. Ferrography and trace metal analysis of silver, silicon and copper aided in identifying the number 5 bearing as the source of the abnormal wear. The wear of the bearing was confirmed on visual inspection. (3) Analytical and direct reading ferrographic data confirmed the rapid increase and subsequent modest decrease in wear debris concentration during the test. (4) The in-line COBRA and the QDMR were used simultaneously for the first time to monitor successfully the condition of the lubricant and to assess wear condition in a flowing lubricant system. (5) The potential for using both monitors in-line has been demonstrated. It is believed that the monitors can be used in a wide variety of applications either independently or as components of an integrated diagnostics system. References 1 H. A. Smith, Complete oil breakdown rate analyzer (COBRA) for identifying abnormal operating engines, Znt. Oil Analysis Workshop, Pensacola, FL, 1983.

312 2 P. W. Centers

and H. A. Smith, COBRA analysis of laboratory degraded synthetic turbine engine lubricants, J. Synth. L&r., 1 (1984) 176. 3 J, B. Schrand and P. W. Centers, Experience with a quantitative debris monitor for determining test cell turbine engine health, ht. Oil Analysis Workshop, Pensacola,

FL, 1983. 4 B. B. Baber,

M. L. Valtierra and J. E. Eichelberger, Development of the automated AFAPL engine simulator test for lubricant evaluation, AFWAL-TR-81-2022, 1981 (Wright-Patterson Air Force Base, OH, U.S.A.). 5 D. Scott, P. McCullagh and G. Mills, Condition monitoring of gas turbine by ferrographic trend analysis, Proc. Int. Confi on Fundamentals of Tribology, Massachusetts Institute

of Technology

Press,

Cambridge,

MA, 1970,

pp. 869

- 874.