You are on page 1of 6

Advanced Strategies for Selecting Oil Analysis Alarms and Limits

Mark Barnes The Fluid Life Corporation 9321 48th Street Edmonton, AB Canada Traditionally, oil analysis users have relied upon their commercial oil analysis lab, lubricant supplier or equipment OEM to provide warning limits for routine oil analysis tests. While this approach has some merit in a predictive approach to oil analysis, particularly for mobile equipment where wear metals limits and other parameters are reasonably well defined, plant equipment oil analysis users and those wishing to take a proactive approach to oil analysis have started to set their own limits based on their own unique goals or equipment environment. In this paper, we review some of the methods available to set basic alarms and limits as well as discuss advanced strategies for data evaluation and interpretation.

Introduction
An oil analysis program without a series of goals is like a ship without a rudder, directionless and prone to going off track at any moment. These goals may be as simple as "finding a glycol leak in an engine before a catastrophic failure occurs" or they may be more sophisticated such as "increasing productivity and equipment uptime by 5% through improved filtration and fluid cleanliness." Whatever your motivation, a well planned oil analysis program needs a series of limits or alarms by which the success or failure of reaching these pre-defined goals can be gauged.

Statistical Analysis
Many labs perform statistical type calculations to determine caution and critical levels. For example, elemental wear metals limits are often set based on an average over several hundred or even thousand sample histories, with caution and critical levels set at one and two standard deviations higher than the average value. While this approach has merit for mobile equipment, particularly when the data set from which the calculation is performed is large, for smaller sample sizes or stationary industrial equipment, meaningful statistical analysis is difficult. For example, two manufacturing plants may use the same make and model of gear reducer, however, one may run intermittently, it may be loaded at only 50% of its maximum rated capacity or it may run outside in sub-zero temperatures. All of these factors will affect the wear rates making comparison between different components almost impossible. In this type of situation, it is usually better for the end user to set their own alarms and limits.

Basic Limits - The Status Quo


Most oil analysis users rely on their commercial oil analysis lab to provide flagging limits. These limits may be based on the labs statistical analysis of many tens of thousands of similar components, previous sample results from the component in question, OEM limits provided by the component manufacturer, or data supplied by the lubricant manufacturer. A good lab will consider all of the above when flagging data from each set of test results. Usually labs provide two (sometimes three) warning levels such as caution to indicate the onset of a problem requiring no immediate action, and critical to warn of a more serious problem condition. Critical and caution levels may be set as positive going limits, such an increase in wear metal concentration, negative going, such as decrease in total base number (TBN) or RBOT value, or both positive and negative such as viscosity where both an increase in viscosity due to for example oil oxidation, and a decrease caused by shearing may be expected.

Cumulative Percent
100 90 80 70 60 50 40 30 20 10 0 0
critical caution

402 ppm

220 ppm

(ppm)

100

200

300

400

500

600

700

range/ ppm 0-54 55-109 110-164 165-219

central value 27 82 137 192

total number of observations 42 35 14 6

% of total

cumulative %

35 30 12 5

35 65 77 82

540-594 595-649 650-677 total

567 622 677 --

1 1 1 118

1 1 1 100

98 99 100 --

applications, provide specific guidelines for wear metals, contaminants (such as dirt, soot and fuel) and lubricant properties. Often these limits are set using the same statistical analysis methods used by commercial oil analysis labs. While these limits provide good general guidelines for interpreting oil analysis data, they should be considered just that, guidelines.

Figure 1: Cumulative frequency distribution approach to setting alarms. Data shows typical iron wear (in ppm) for an diesel engine at 300 hours oil service (Ref: Noria Corp./KOWA)

Another approach which is often used is to look at the cumulative frequency distribution. This is illustrated in Figure 1. In this approach, a series of ranges are defined and the number of events or sample results for which fall inside each range is counted. The total number events in each range are then expressed as a percentage of the total number of observations, which in turn can be related to a cumulative percentage. Using this approach, cautionary limits are often set at 85% (i.e., the value below which 85% of all samples fall) and critical limits at 95%. Again, like all statistical approaches to setting limits, this method relies on having an appropriately large data set of equipment operating under similar conditions. Another factor which must be considered, particularly when looking at wear metal concentrations is to look at the wear rate, rather than the absolute value of a test parameter. For example, an engine may normally produce 25 ppm of iron during a 250 hour oil change interval. However, if this oil change interval is stretched to 300 hours and no other problems occur, it would be expected that the amount of iron in the sample might increase to 30 ppm, 1 ppm for every 10 hours of operation. Similarly, most components exhibit significantly higher wear rates when first installed due to break-in effects. If provided the correct information, most oil analysis labs will allow for these effects, however all too often, samples are sent to the lab without component hours, making the labs job of providing meaningful flagging limits almost impossible. In many circumstances, it may be more appropriate to set limits in terms of wear rates (i.e., 1 ppm/10 hours) rather than set a discrete number limit such 25 ppm.

For example, the author was questioned by a locomotive engine manufacturer about flagging limits applied to one of their engines in use by a major Canadian railroad company. The pertinent data for this case study is illustrated in Figure 2, showing the correlation between sodium concentration, measured to indicate a coolant leak, and lead, the main component of the engine bearing Babbitt overlay layer. The question that was posed by the engine manufacturer was "why did you flag sodium as unacceptable at 69 ppm when our cautionary limit for sodium is set at 200 ppm?". The answer to this question lies in the data illustrated in Figure 2. When sodium levels are relatively low (<25 ppm), the amount of bearing wear, as illustrated by the lead levels is also comparatively low. However, when the sodium values spikes at 69 ppm, only 35% of the OEM's "allowable" limit, the amount of lead generated has tripled. Obviously elevated lead levels are an indication of elevated wear and a corresponding reduction in bearing life.
date 99/05/26 99/06/26 99/07/18 99/08/29 99/11/02 sodium/ppm 21 24 17 69 6 lead/ppm 2 2 1 7 2

80

70 60 50

sodium (ppm) lead (ppm)

40 30

20 10 0
99/ 05/ 26 99/ 06/ 26 99/ 07/ 18 99/ 08/ 29 99/ 11/ 02

OEM Limits
Many equipment OEM's, particularly in mobile fleet
2

Figure 2: Correlation between sodium (from a coolant leak) and lead (Babbitt wear) for a locomotive engine (Ref: Fluid Life Corp.)

If this engine had been left unchecked the sodium levels would have continued to rise, along with lead production, significantly reducing the life expectancy of the engine bearings. It is conceivable that if this engine had been left to run until the sodium levels reached the recommended 200 ppm warning limit, the wear rate may be ten times higher than under normal operating conditions. Fortunately in this case, the customer heeded the lab's warning, inspected the engine and located a cooler leak. The take home message is that caution should be exercised when relying on OEM flagging limits. Again, it is far better to set your own limits based on your own data.

Lubricant Manufacturer Limits


Most lubricant manufacturers are able to provide limits for key fluid property tests such as viscosity, neutralization numbers or RBOT number. Just like OEM limits, fluid property test limits provided by a lubricant manufacturer should be considered guidelines. For example, a lube supplier may set a cautionary limit for total acid number (TAN) at 3.0, above this value, the oil is likely to thicken or may cause corrosive wear to occur. However, depending on the situation and metallurgy of the component, significant corrosive wear may have already occurred when the TAN value reaches 2.5. These types of limits can however be valuable in allowing the end user to base oil changes not just on guess work or the conservative estimates provided by OEM's but on hard fast scientific data. For example, the author was asked by a reliability coordinator at a large pulp mill whether the maintenance engineers decision to change oil in their turbo-generator system "because it hadn't been changed for 5 years" was justified. After conducting an RBOT test, it was found that the RBOT value, a measure of the degree of degradation of a lubricating oil was at 90% of the new oil value, indicating that the oil was in excellent shape. Based on one single test, and knowing the appropriate new oils specifications and limits from the lubricant supplier saved this mill $25,000 for an unnecessary oil change. However, just like OEM limits, it is unwise to put too much faith in lube manufacturers limits, without considering other factors such as wear rates and contamination levels.

what limits they should be setting for water in their stern propeller tubes. After checking with the OEM and lube supplier, nobody could supply reliable guidelines for sea water ingression limits. Discussing the problem with the engineer, it was determined not to be specifically the water, which in a stern tube is very difficult to eliminate, but the electro-chemical corrosive effects of salt and other minerals present in sea water that was the problem. Based on a trend analysis of this ships historical data, which is shown in Figure 3, it was deduced that by using sodium as a flag instead of water as an indicator of sea water ingression a limiting value of 20 ppm could be set; sodium concentrations below 20 ppm did not cause bearing wear, while the presence of sodium above this limiting value tended to cause white metal bearing wear, as indicated by the increase in lead levels. In this case, in the absence of reliable OEM, lubricant manufacturer or lab analysis flagging limits, the end user was able to determine an appropriate limit for predictive purposes based on his own historical data. Many oil analysis users rely on software to track and trend their oil analysis data. The more sophisticated software programs date Sodium lead/ available today allow /ppm ppm users to do both graphical trend analysis 97/03/22 14 0 97/04/20 8 0 as shown in Figure 3 and 97/05/23 8 0 basic statistical analysis 97/07/04 25 39 97/12/30 8 0 of their own data base to 98/03/02 6 0 determine alarms and 98/04/04 7 0 limits. These limits can 98/04/26 9 0 98/05/28 10 0 then be set as default 98/06/24 29 27 values, so that data 98/07/25 5 2 98/08/23 5 2 returned from the lab is
80 70

98/09/24 98/10/24 98/11/26

4 5 6

0 2 2

60 50

sodium (ppm) lead (ppm)

40 30 20

User Defined Limits


Once you are familiar with the test methods being used and has sufficient historical data, it is best to start setting you own limits since these limits can be set to reflect the real life conditions and reliability goals in place. For example, the author was asked by the chief engineer of a coastal ferry for information on
3

10 0
99/0 5/26 99/0 6/26 99/0 7/18 99/0 8/29 99/1 1/02

Figure 3: Trend Analysis of sodium and lead for starboard stern tube showing the effects of sea water ingression on white metal bearing wear (Ref: Fluid Life Corp.)

flagged according to the users own specifications. In the future, it may even be possible to allow user to compare their data not only with their own sample histories, but with "data warehouses" compiled by oil analysis labs, which allow end users to compare their data with a much greater data set, thus improving statistical accuracy.

The Importance of Understanding Oil Analysis Test Methods


One of the keys to understanding oil analysis flagging limits and applying them appropriately is to understand the tests to which these limits apply. Take for example the case illustrated in Figure 4. In this situation, oil samples were taken on a monthly basis from a natural gas engine. Based on the labs statistical analysis, in conjunction with OEM data, the iron caution limit was set at 25 ppm and critical limit at 95 ppm.A sample taken in April 97 was found to exceed the critical limit, however, based on pressure from production it was decided to continue in operation and re-sample the following month. The sample taken the following month in May 97 was found to have dropped below the cautionary iron limit.
date
400 350 300 250 200 150 100 50 0
97/ 01/ 01 97/ 02/ 03 97/ 03/ 02 97/ 04/ 01 97/ 05/ 02 97/ 06/ 04

5 microns reveals a very different story. By not acting when the iron level exceeded the caution point, the problem has been allowed to progress such that much larger particles are being generated. Since spectroanalysis using an ICP instrument is limited to sub 5 micron particles, the iron count by ICP drops, while the number of large particles caused by a more serious wear condition is increasing. Again if you plan on setting your own flagging limits, it is important to understand the test method used to which the limit applies and most importantly its limitations.

Predictive vs. Proactive Limits - Setting Goals


Whether you rely on your lab, equipment OEM or lube supplier to provide limits or even if you are currently setting your own limits, most oil analysis programs tend to focus on predictive data patterns such as "is the iron wear rate normal" or "is the viscosity of my oil still in grade". However, the maximum benefit can be achieved through oil analysis when a proactive approach adopted. This approach requires setting limits which reflect a certain maintenance goal, such as "to extend the meantime between engine overhauls" or "reduced the number of iron/ppm sediment servo-valve failures each mg/L year". Since most lubrication related problems are 3 20 2 30 associated with contami20 20 nation, goal based limits will 164 120 usually apply to tracking and 52 340 trending contamination. For 40 20 the purposes of discussion, we will focus on the two most common contaminants, namely particulate contamination and water.

Fe (ppm) Sediment

97/01/01 97/02/03 97/03/02 97/04/01 97/05/02 97/06/04

Particulate Contamination
Industry studies have revealed that between 75-85% of all failures are caused by particulate contamination. For this reason, many oil analysis users are starting to perform routine particle counting, either onsite or through their commercial oil analysis lab. Studies have also revealed that the degree of particle contamination can be correlated directly with the life expectancy of a component. Table 1 shows data from a study conducted by the British Hydrodynamic Research Association. In this study, 117 hydraulic systems from different types of equipment were compared and the average number of hours between mechanical breakdown correlated with fluid cleanliness.

Figure 4: Iron and sediment trend analysis for natural gas engine showing the relationship between large and small particle wear (Ref: Fluid Life Corp.)

To the uneducated observer, it may appear that the problem has miraculously gone away. However, the sediment test data, which shows the total amount of large particle contamination (in mg/L) greater than
4

ISO code average hours between breakdown

24/21 200

23/20 250

22/19 325

21/18 430

20/17 600

19/16 800

18/15 1050

17/14 1400

16/13 1900

15/12 2600

14/11 3800

13/10 5000

12/9 6500

Table 1: British Hydrodynamic Research Association (BHRA) study of fluid cleanliness on average breakdown hours for 117 hydraulic systems (Ref: BHRA/Noria Corp.)

Based on this study, it stands to reason that if your current machine cleanliness is 18/15, improving the cleanliness by even one ISO range code can have a significant impact on equipment life and reliability. Table 2 provides a means of assessing the theoretical life extension possible by improved fluid cleanliness and allows specific targets or goal based limits to be set. For example, as a maintenance engineer, your mandate may be to improve the reliability and life expectancy of a particular piece of hydraulic equipment by a factor of two. If your current average ISO cleanliness rating is 18/15, based on Table 2, your goal based limit for improved cleanliness should be set at 15/12. Of course setting the limit and doing particle counting doesn't change the cleanliness of your system, to achieve this improvement a lifestyle change such as an upgrade of air or oil filtration may be required, but the goal based limit provides a target by which the success of this lifestyle change can be measured.

When setting goal based limits, it is important to be realistic. For example, according to Table 2, by improving fluid cleanliness in the above example from 18/15 to 11/8, the life expectancy of this hydraulic system should go up be a factor of 7. However, the question which must be asked is how realistic is it that I can keep my oil as clean as 11/8. As with all aspects of oil analysis, theory should be tempered with real life common sense. Goal based limits should be considered moving targets. Lets assume using the above example, that by upgrading the breathers on this hydraulic unit, the fluid cleanliness improves from 18/15 to 15/12, resulting in the expected two fold increase in reliability. At this point, it may be possible, through upgraded oil filtration to improve the oil cleanliness to 14/11. From Table 2, this would results in a 50% increase in reliability. If the added expense of improving oil filtration to achieve another 50% increase in reliability makes economic sense, why not move the target cleanliness code lower? Goal based improvements should be considered a iterative process; continual changes and improvements should be made with lower and lower limits set until no further financial gains can be achieved.

Water contamination
In many industries, water contamination is the number one source of component failure. In rolling element bearings in particular, water contamination can have a serious impact on component life and equipment reliability. Figure 5 shows data from a study by Timken Bearing Co. illustrating the effects of water in oil on rated bearing life. As can be seen from Figure 5, keeping water contamination at 100 ppm or lower can result in a significant increase in component life. While this may not always be practical, the general rule of thumb in setting limits for water contamination is to keep it as low as reasonably possible, but certainly below the saturation point of the lubricant in use to prevent free water from being present.
5

Table 2: Life Extension table for improved fluid cleanliness (Ref: Noria Corp.)

Conclusions
Oil analysis limits provide us a way of measuring the success of our oil analysis program. By getting away from the premise that the equipment OEM or lubricant supplier knows best and using our own knowledge and historical data to determine our own limits, maximum benefit can be gained by using oil analysis as a predictive tool for condition monitoring. Going one stage further and setting limits and targets such as particle contamination or water concentration based on specific maintenance goals can provide the end user with a proactive means of significantly improving equipment reliability and production uptime.

Figure 5: The effects of water in oil on bearing life (Ref: Timken Bearing Co.)

Just like particulate contamination, goal based limits for water contamination can be set as a target for improved equipment reliability. Table 3 illustrates the effects of reduced water contamination on component life. Again, by setting a goal based limit using Table 3 and making a lifestyle change such as improving seals or deploying vacuum dehydration a proactive approach to equipment reliability can be made.
Table 3: Life Extension table for water contamination

(Ref: SKF/OSU/Noria Corp.)

You might also like