An assessment of quantitative and qualitative ferrography

  • Published on

  • View

  • Download

Embed Size (px)


<ul><li><p>Wear, 126 (1988) 31 - 55 31 </p><p>AN ASSESSMENT OF QUANTITATIVE AND QUALITATIVE FERROGRAPHY </p><p>G. R. WAKEFIELD and H. LEVINSOHN </p><p>Defence Scientific Establishment, Ministry of Defence, Auckland Naval Base, Auckland (New Zealand) </p><p>(Received July 14, 1987;revised November 2, 1987; accepted February 2, 1988) </p><p>summary </p><p>The direct reading and analytical ferrographic processes have been examined in detail to evaluate their use as quantitative and qualitative failure prediction tools. Particular attention was paid to the quantitative aspects of ferrography and the broad assumptions on which they are based. </p><p>Experimental work has shown that, for the type of samples analysed, the repeatability of ferrographic data is inadequate for the ferrograph to be used in isolation as an engine health monitoring instrument. A number of minor alterations were made to the instrument to enhance its repeatability, and the quantitative results can now be successfully used in conjunction with both qualitative ferrographic information and data from other engine health monitoring techniques such as spectrometric oil analysis. </p><p>1. Introduction </p><p>A project was undertaken with the following objectives: (a) to deter- mine whether a spectrometric oil analysis program (SOAP) based on atomic absorption spectrometry (AAS) chemical analysis data trends could be correlated with the results obtained from ferrography, (b) to assess the repeatability of the ferrographic procedures and (c) to determine whether the ferrograph could be successfully used to characterize abnormal wear situations. </p><p>A duplex ferrograph analyser was used for this study [ 11. </p><p>2. Initial tests </p><p>Early tests showed that routine analytical ferrographic analysis of the oil from a particular model of turbopropeller engine used in some RNZAF aircraft-produced data from which detection of wear rate trends was difficult (Fig. l(a)). The severity of wear index (Appendix A) of consecutive samples </p><p>0043-1648/88/$3.50 @ Elsevier Sequoia/Printed in The Netherlands </p></li><li><p>Figure 1A </p><p>/ Figure 16 </p><p>5500 Ina 8500 </p><p>Engine hours Fig. 1. (a) Comparison of analytical ferrograph results with (b) corresponding SOAP-AAS data. </p><p>often followed a pattern of transient fluctuations with up to an order of magnitude of variation. A number of reasons for this phenomenon were suggested. </p><p>(a) The sampling procedure from the aircraft was incorrect. (b) There was a lack of sample homogeneity. (c) There was a lack of repea~bi~ty in the ferrographic process. (d) The values were actually a true indication of fluctuating wear debris </p><p>levels within a lubricating system. These hypotheses prompted the investigation of a number of factors, </p><p>in order to determine the reliabifity of ferrographic data. </p></li><li><p>33 </p><p>3. Initial comparison atomic absorption spectrometry-ferrography data </p><p>Plots of analytical and direct reading (DR) ferrographic data were compared with the SOAP-AAS data for the same samples (Figs. 1 and 2) in order to establish possible causes of erratic results. </p><p>From Fig. 1, it is apparent that it is not possible to correlate fluctua- tions in the severity of wear index with the iron concentration as determined by AAS. The AAS results showed a gradual increase in iron concentration between successive oil samples, as expected. However, the severity of wear index plot tended to show large random fluctuations with no definite trends. Unlike the AAS data, the ferrographic data did not change consistently when an engine had undergone an oil change. Plots of readings from both the large-particle (Dl) sensor and the small-particle (Ds) sensor of the DR ferrograph also did not correlate well with the corresponding SOAP results (Fig. 2). </p><p>4. The direct reading ferrograph: preliminary investigations and alterations </p><p>Early trials with the DR ferrograph showed that variations in the ambient light intensity directly affected the optical density readings. This was found to be caused by the leakage of stray light into the exposed ends of the precipitator tube. The interference was overcome by enclosing the photodetectors, precipitator tube and clamping assembly in a light-tight black polyethylene cover. </p><p>In order to facilitate continuous monitoring of sensor recordings as a sample flowed through the precipitator tube, a chart recorder terminal was connected to the DR ferrograph, enabling the output of the large or small particle sensor amplifier [2] to be recorded. </p><p>4.1. Physical measurements A series of physical measurements were carried out on the DR unit to </p><p>obtain a better understanding of the significance of the particle deposition patterns within the precipitator tubes and the implications of the Dl and Ds readings. The DR unit was dismantled and measurements were made to establish the positions of the particle sensors relative to the precipitator tube (Fig. 3). The DR readings could then be correlated with the observed debris deposition within selected tubes. </p><p>4.2. Magnetic field strength measurements A series of measurements was carried out to determine the magnetic </p><p>field strength along the length of the DR magnet. Measurements were carried out with a longitudinally sectioned precipitator tube in position in order to simulate the actual magnetic field within a tube during running conditions. Figure 4 shows the change in gaussmeter readings as the probe is moved along the magnet, up field from the start position. Any readings closer than </p></li><li><p>34 </p><p>P - ________---- </p><p>______________-__---------~., \ </p><p>/- \ / A , .\ / I - \. </p><p>Fig. 2. (a) Comparison of SOAP-AAS data with (b) and (c) corresponding DR ferrograph results. </p><p>4.5 mm from the extremities of the magnet were disregarded because of inaccuracies caused by the fact that the probe overlapped the ends of the magnet. By extrapolation the field strength in the vicinity of the large- particle sensor is estimated to be 2050 - 2100 G. The field strength in the vicinity of the small-particle sensor rwed from 2175 to 2225 G. As the magnet of the DR ferrograph is positioned parallel to the precipitator tube, it would be expected that the magnetic field strength along the precipitator tube would be constant. This is not the case, as Fig. 4 indicates. The field </p></li><li><p>35 </p><p>Cross section </p><p>I sensor sensor Datum line </p><p>(leaoqintgube,dge </p><p>clamping assembly) </p><p>Stab of magnet </p><p>Fig. 3. Position of photodetectors relative to magnet (DR). (Dimensions in millimetres.) </p><p>0 10 70 00 </p><p>Fig. 4. Variation in magnetic field strength along the analytical and DR magnets. </p><p>strength decreases at both ends of the magnet with a more pronounced effect towards the tube entry. These decreases in magnetic field strength are due to end effects at the extremities of the magnet which cause a lower- ing of the magnetic flux density above the magnet. The more pronounced </p></li><li><p>36 </p><p>decrease at the entry region has been attributed to the presence of the sensors which attenuate the magnetic field. Unlike the analytical ferrograph which provides an increasing magnetic field along the entire oil flow path by virtue of the inclined ferrographic slides, the DR unit provides increasing magnetic field only in the region of the sensors. </p><p>5. Repeatability trials </p><p>In order to examine the degree of repea~bility of the DR fe~o~aphic process, a series of trials was carried out on various oil samples contaminated with different types of debris. </p><p>Two different samples were tested for repeatability of readings. Sample A consisted of oil extracted from a helicopter main rotor gearbox. Sample B consisted of unused NATO O-156 (MIL-L-23699C) oil artificially contami- nated with iron microspheres to a concentration of 0.001%. Chemical analysis had previously shown the spheres to be 99 wt.% Fe and they had a diameter distribution ranging from 0.6 to 6.2 pm [3]. </p><p>The results obtained from the repeated testing of the above samples are shown in Table 1. </p><p>TABLE 1 </p><p>Direct reading ferrograph repeatability </p><p>Sample A Sample B </p><p>Dl RSD (%) 5.0 8 Ds RSD (%) 5.6 75 Is RSD (W) 11.3 - </p><p>Relative standard deviation (RSD) = standard deviation/mean) x 100%. Severity of wear index Is = D12 - Ds2, where Dl is the standardized large-particle sensor reading and Ds the standardized small-particle sensor reading. </p><p>Two significant factors arose from these tests. (a) The RSD of the severity of wear index value is higher than that for </p><p>individual Dl and Ds readings. It is important to note this large variation, particularly if the wear index is to be used in a practical engine health monitoring system. </p><p>(b) The repeatability for the artificially contaminated sample was significantly lower than for the used sample which had Dl values of a similar magnitude. There is no apparent explanation for this anomaly, The fresh oil used to prepare the synthetic samples was free of any detectable particulate contamination. </p><p>The effect of ultrasonic agitation of samples passed through the DR ferrograph was investigated. The specimens analysed were used NATO O-156 (MIL-L-23699C) from aircraft turbopropeller engines which usually have </p></li><li><p>31 </p><p>very low levels of wear debris contamination and produce DRs of less than 10. The trial involved analysing the samples by two different methods: the standard method of analysis; the standard method of analysis preceded by an ultrasonic agitation treatment. This involved soaking the sample at 65 + 5 C as in the standard routine. The sample container was removed from the oven, hand shaken for 10 s and immediately placed in an ultrasonic bath containing water which was preheated to a nominal 65 C. After ultra- sonic agitation for 45 - 60 s the specimen was removed from the bath and hand shaken for a further 10 s. Oil was then drawn off for analysis. Each specimen was individually treated immediately preceding the analysis. </p><p>It was clearly shown (Figs. 5 and 6) that ultrasonic agitation, in this case, improved the repeatability of both the Dl and the Ds readings. The presence of isolated unusually high readings could be attributed to a small quantity of relatively large non-magnetic particles which deposited on one of the two sensor locations. (Experience has shown that this effect is fre- quently encountered in analytical ferrography.) If the single high reading in both Fig. 5 and Fig. 6 are omitted, the following statistical values are obtained: from the standard procedure, </p><p>Dl average = 9.6 </p><p>Dl standard deviation = 2.6 </p><p>Ds average = 8.5 </p><p>Ds standard deviation = 3.2 </p><p>0 I 2 3 </p><p>dun m&amp;M </p><p>a 7 I </p><p>Fig. 5. DRs from a series of tests performed on a single used oil sample from an Allison T-56 turbopropeller engine. </p></li><li><p>Ds reading 0 DI reading m </p><p>! 2 3 1 5 </p><p>Run &amp;umb&amp; B 9 ,o </p><p>Fig. 6. As in Fig. 5 with ultrasonic agitation of sample prior to testing. </p><p>and, with ultrasonic agitation, </p><p>Dl average = 4.4 </p><p>Dl standard deviation = 0.6 </p><p>Ds average = 2.5 </p><p>Ds standard deviation = 0.7 </p><p>The ultrasonic agitation treatment has significantly improved the repeatabil- ity of the readings. A second notable fact arising from these results is that the values of both the Dl and the Ds readings have decreased significantly with the use of the ultrasonic treatment. This suggests that ultrasonic agita- tion has caused agglomerates of small particles to disperse. Agglomerates of small particles would have the physical behaviour of a large particle and be deposited in the vicinity of the large-particle sensor. However, the decrease in the Ds reading is not fully understood but could be a particle orientation effect. Ultrasonically treated samples left for a number of hours and then retested gave readings similar to the untreated samples. This indicated that the dispersion effect was temporary and the small particles had con- glomerated again. </p><p>Tests on a sample from a similar engine produced no change either in the repeatability or in the magnitude of the readings despite the ultrasonic treatment. Therefore the effects of ultrasonic agitation appear to be sample dependent. The size, shape and composition of the particles would all affect the response to this treatment. The benefit of ultrasonic agitation on DR samples is uncertain and, because this process is very time consuming, each application needs to be preceded by a comprehensive study. </p></li><li><p>39 </p><p>6. Particle deposition patterns </p><p>6.1. Method of tube preparation Particle deposition patterns within selected DR precipitator tubes were </p><p>examined by axially sectioning a number of specially prepared tubes. Three methods of tube preparation were utilized. </p><p>Method 1. Tubes were prepared for examination by slicing along section AA (Fig. 7) prior to use using a diamond cut-off wheel. The segment was then replaced using a silicone rubber cement. Readings of the Dl and Ds sensors were not taken because of the optical interference caused by the silicone cement. After testing, the tube and its contents were washed with fixer solvent and the tube was then carefully drained. When dry, the tube with the deposited debris in place was carefully removed from the DR unit and the sections were separated by splitting the silicone rubber joint. The segment was then ready to be examined under an optical microscope. An </p><p>A </p><p>Deposition i </p><p>zone - </p><p>A </p><p>1 Section removed </p><p>enable observation </p><p>Method 1 </p><p>Debris position </p><p>TD </p><p>Section BB </p><p>Method 2 </p><p>Fig. 7. Precipitator tube preparation. Section orientations for internal examination. </p></li><li><p>40 </p><p>external clean with an organic solvent was required to prepare the specimen for scanning electron microscopy (SEM). </p><p>Method 2. The selected sample was processed in a normal precipitator tube and the sensor readings were recorded. The fixer solvent remaining in the tube was drained carefully. When dry, the precipitator tube was carefully filled by injecting an epoxy encapsulating resin. Once the resin had set, the tube could be removed from the DR unit. The section over the sensors was cut from the precipitator tube using a diamond cut-off wheel. This section was mounted in a transparent medium (acrylic resin). The orienta- tion was such that polishing of the mounting revealed a vertical cross section through the debris deposit (Fig. 7, section BB). The polished sample was then examined under an optical microscope. </p><p>Method 3. A tube which had been impregnated with epoxy resin was placed in concentrated hydrofluoric acid for approximately 3 h. This dis- solved the glass tube but left the resin intact, allowing the exact position of the debris deposits to be measured. </p><p>6.2. Investigation of particle deposition patterns The first deposition patterns to be investigated were produced by </p><p>passing samples of iron microspheres (10 ppm in NATO O-156 lubricating oil) through presplit tubes. These were examined by SEM to observe the microsphere size distributions and positions (Fig. 8). When the three micro- graphs (Figs. 8(a), 8(b) and 8(c)) were compared, it was evident that a decrease in the average sphere size occurred along the flow path. The large- particle sensor aperture occupies a position between 4.75 and 7.50 mm from the datum line, as shown in Fig. 3. All the spheres in Fig. 8 ar...</p></li></ul>