Oil Analysis Best Practices

Proactive and Predictive Strategies For Setting Oil Analysis Alarms And Limits In the past, users of oil analysis have relied almost exclusively on the commercial laboratory to set and enuncia enunciate te data data alarms alarms.. This This has put put an unreal unrealist istic ic burde burdenn on the labs labs to under understa stand nd information about user-equipment they have never seen. Likewise, the goals and obectives of the user with respect to reliability and maintenance may not be fully understood. !sually, this leaves the lab with no alternative other than to use standard default alarms. "hen these one-si#e-fits-all alarms alarms are used, used, many many of the opport opportuni unitie tiess and obect obective ivess of a moder modernn condit condition ion-ba -based sed maintenance program are missed. In recent years, with the advent of sophisticated user-level oil analysis software, many site oil analysis technologists are taking responsibility for setting alarms and limits independent of the lab. The lab, in turn, is being asked to only deliver accurate and timely oil analysis results, leaving interpretation and exception reporting to the user. "ith the user being familiar with the lubricants, machines, historical problems, and general general reliability goals, the most proper and effective limits can then be set. $owever, a critical ingredient is a sufficient level of training in oil analysis by the user. What are Oil Analysis Alarms and Limits? "hen plant-level oil analysis software is employed, one of the main benefits is the funneling down %or filtering& of the amount of data actually viewed and analy#ed. This data reduction goal is essential and relates to the tactical goal of exception reporting, reporting, that is, the viewing of only the data that is out of compliance with acceptable trends or levels. 'll other data is held and managed in a datab database ase for future future refere reference nce and and use. use. The limits, limits, when when prope properly rly set, set, give give confid confiden ence ce that that confor conformin mingg fluids fluids are, are, in fact, fact, okay okay and and that that non-c non-conf onform ormin ingg condit condition ionss have have been been caught caught proactively.

(I)!*+  Illustration of Upper and Lower Oil Analysis Limits

ecause the goals and expectations vary considerably from one organi#ation to another these limits are best set by those charged with the responsibility of machine reliability and long fluid life. They are also influenced by machine and application specific considerations. (or instance, where the gear box loads and duty cycle are low in one application, they may exceed rated levels in another. The availability of spares and standby equipment may influence the decision with equal magnitude. 'nd, the age of the machine or lubricant may impact the placement of limits in certain cases. asically, a limit or alarm is a strategically placed trip wire that alerts you to an abnormal condition. /ome data parameters have only upper limits such as particle counts or wear debris levels. ' few data parameters apply a few lower limits like T0, additive elements, flash point, and (TI* %additive&. 1ther data parameters use both upper and lower limits. These might relate to important chemical and physical properties of the lubricant such its viscosity %see (igure &. The computer trends the data and normally no alarm is reported as long as the limits are not exceeded. "hen an exception exists the software is designed to alert and direct corrective action in response to the deleterious result. To insure maximum benefit from oil analysis, careful thought needs to go into the type of limit used and its setting. Some Limits are More Proactive 2roactive limits are designed to alert users to abnormal machine conditions associated with root causes of machine and lubricant degradation. They are keyed to the proactive maintenance philosophy of setting targets %or standards& and managing the lubricant conditions to within the targets. ' strategic premise is that these conditions are controlled to levels that are improvements over past levels and that these become goals. est results occur when progress towards achieving these goals are charted conspicuously by the maintenance organi#ation. These types of limits are referred to as goal-based, see (igure 3.

(I)!*+ 3 Goal-Based Limits Are Set For Certain Data Parameters To Proati!ely Impro!e "a#ine And Lu$riant %elia$ility

In order to set a goal-based limit a level of machine reliability is identified. 'gain, this needs to be an improvement over previous levels. If, for instance, we use particle count as the parameter then  we need to select a target cleanliness which is a marked improvement from before. The target cleanliness becomes the limit. In the example, if previously contaminant levels were averaging about I/1 456 for a hydraulic system, a limit set at 653 would be a goal-based improvement. ' life extension of three times %7T(& would be expected based on controlled field studies. If, on the other hand, we set a limit of 456 our effort is downgraded to the detection of maor faults only. )oal-based limits of this type can be applied to particle counts, moisture levels, glycol levels, fuel dilution, T'0, and other common failure root cause conditions. 'nother similar type of proactive limit relates to the progressive aging of the lubricant or hydraulic fluid. (rom the moment the oil is first put into service its physical and chemical properties transition away from the ideal %i.e., those of the new formulated oil&. (or some properties the transition may be extremely slow but for others it can be abrupt and dynamic. Limits keyed to the symptoms of lubricant deterioration are referred to as aging limits, see (igure 8. They are designed to signal the need for a well-timed condition-based oil change and are usually pegged to the depletion of additives and the thermal5oxidative degradation of the base oil.

(I)!*+ 8 9 A&in&-Limits Are 'eyed To De&radin& Additi!e Base-Sto( Properties In order to properly set aging limits the new lubricant must be analy#ed to reveal its physical and chemical properties. This will become the oil:s base signature. ;are must be taken to insure that the new oil is analy#ed using the same test procedures and instruments that will be used to analy#e the used oil. !nder no circumstances should these base signature properties be simply lifted off the lubricant:s spec sheet as provided by the supplier. Instruments and test methods vary substantially and new oils are approved if their properties fall within a tolerance range. (or instance, I/1 viscosity grades vary plus5minus ten percent from a nominal center point %grade&. This means that a 6 centistokes %c/t&. This is too much variance for =4 centistokes to serve as a proper baseline. $owever, if the actual viscosity of

a specific lubricant was measured and found to be =? c/t then a precise baseline is now available against which the used oil can be trended. 'ging limits can be effectively applied to such parameters as T'0, T0, viscosity, *1T, emission spectroscopy for additive elements, (TI* %for oxidation, nitration, @ sulphation&, and dielectric constant. 'ging limits often follow trendable data patterns, i.e., they trend steadily in the direction of the limit. The actual time to limit might be predicted by linear or non-linear regressionA a feature in some oil analysis software products. (or instance, the following equation might be applied to estimate the remaining useful life of a turbine oil using *otating omb 1xidation Test %*1T&9

Lower aging limits are generally set for additive depletion, *1T life, and T0. !pper limits are set for dielectric constant, T'0, and (TI* for oxidation, sulphation, and nitration trends. 'nd both upper and lower limits should be set for viscosity. The table in (igure ? shows some example limits for both goal-based and aging parameters.

(I)!*+ ? )*ample of Goal-Based Limits Other Limits are More Predictive 2redictive limits are set to signal the presence of machine faults or abnormal wear conditions. They are aligned with the goals of predictive maintenance, i.e., the early detection of machine failure symptoms as opposed to failure root causes %proactive maintenance&. In oil analysis, a proper predictive limit set to the correct parameter has many advantages over other predictive maintenance technologies. /pecifically, it offers reliable incipient fault detection, spanning a wide range of machine failure modes. It is seer-like in that it has the ability to forecast a future event. 's compared to vibration analysis for instance, the time-based detection window using ferrous density analysis has been demonstrated to exceed 6 times for common gear boxes failures. *ate-of-change limits are generally identified as predictive. These are set to a property that is being progressively introduced to the oil, such as wear debris. The add rate %change& might be calculated per unit time, for instance ppm iron per BB hours on oil. "hen the parameter:s value is plotting against time the rate-of-change %add rate& equates to the current slope of the curve. 's an

alternative to representing rate-of-change, slope can be quantified by dividing rise by run for a fixed period of time %see (igure 6&. The linear trends also points to the approximate time interval remaining before a level-type limit is exceeded. !nlike level limits however, rate-of-change limits ignore the absolute value of a data parameter, emphasi#ing instead the speed %rate& at which the level is changing. This can be best illustrated by an example. /uppose for a given machine, such as an industrial gear unit, iron is typically introduced to the oil at roughly 6 ppm per month. The first month after an oil change the iron shows 6 ppm, the expected level. 'fter the second month the iron shows B ppm, again this is expected %6 C 6&. The same holds for the end of the third month when 6 ppm is reported. The add rate of 6 ppm per month remains uninterrupted. $owever, by the end of the fourth month the unexpected result of 6B ppm is obtained. The computer software shows this as a critical. The reason is not due to the fact that 6B ppm in a gear lube signals abnormal wear. In fact this is a rather common iron concentration in gear oils. $owever, the alarm is responding to the rate at which the iron concentration changed in the last month of service %86 ppm instead of the expected 6 ppm&. $ad rate-of-change limits not been applied this exception might not have been reported.

(I)!*+ 6 Usin& %ate-of-C#an&e Limits +rise,run "ust Be Calulated For T#e Trendin& Data Parameter The use of rate-of-change limits is well suited for wear debris analysis but can be used for other parameters as well. +xamples of where it is commonly applied include particle counts, elemental  wear metals, ferrous density analysis %D*, 2E FB, "ear 2article 'naly#er, ferrous ;ontam-'lert&, T'0, and *1T. It should be noted that multiple limit strategies can be used for single test parameters, i.e., where rate-of-change limits are applied so too are level limits %aging, proactive, and statistical&.

Statistical Limits are Predictive as Well (or many years statistical limits have been used successfully in oil analysis. The practice requires the availability of a certain amount of historical data on the target parameters %see (igure =&. ' population standard deviation is calculated. !pper limits are then set relating to the number of standard deviations %sigmas& above the sample population average. 7any analysts put a caution at one or two sigmas and a critical at two or three sigmas. "hen one sigma is exceeded this means that the value from the test result exceeds =4 percent of historical samples. ' result that alarms at two sigmas exceeds F6 percent of historical data. Three sigma exceedance corresponds to of FF.> percent of the database. 7any commercial laboratories have large repositories of data spanning numerous machine types and models. These data permit the relatively easy calculation of national averages and corresponding population standard deviations %sigmas&. In some cases the data can be conveniently sorted according to industry and application. These same databases can be used to assist in the setting of rate-of-change limits as well. Typical applications of statistical limits include elemental analysis of wear metals, ferrous density analysis %D*, 2E FB, etc.&, and other common predictive oil analysis measures. It is well known that many machines exhibit highly individual characteristics. They might trend high or low when compared to national averages. Data from a machine that is a low reader might not alarm early enough when national averages are used as the statistical base %false negative&. Likewise, when a high reader is encountered %potential false positive&, it may be well advised to adust the statistical limit accordingly, or simply rely more heavily on rate-of-change limits.

(I)!*+ = Statistial Limits Are 'eyed To .istorial Oil Analysis Trends Dealing With Data Noise Data noise can mask or distort the target data parameter %and trend& often making it nearly invisible to detection. 'nd, when data noise exists it can inhibit the ideal placement of certain alarms and

limits. ' machine that normally has a high level of wear particles in its lubricating oil is a good example. These particles do not represent current wear activity but are an accumulation of historical wear, possibly going back many months. This is a common situation when course filters or no filters are in use. This high concentration of wear particles constitutes noise for an oil analysis program.

(I)!*+ > Clean Fluids .elp Pro!ide Impro!ed Fault Detetion Sensiti!ity "hile the target signal %data& is current wear levels, these particles may be extremely difficult if not impossible to measure when they are mixed indiscriminately with historical wear debris. This equates to low signal-to-noise ratio. ;rudely stated the fault signal is getting lost in the sauce. This situation is illustrated in (igure >. "hen particle concentrations are controlled to an I/1 F5= fault detection is poor. y comparison, the I/1 85B fluid translates to high-resolution detection, i.e., high signal-to-noise ratio. ' similar problem occurs with infrared spectroscopy when weak absorption signals are lost due to inaccurate reference spectra and the presence of interfering materials in oil. Summary "ith the current trend of users taking control of the their oil analysis programs there has been a surge of interest in education. This has recently lead to an /TL+ %/ociety of Tribologist and Lubrication +ngineers& committee being formed to offer oil analysis certification levels. "hile oil analysis education is often aimed at data interpretation it is no less important in the area of limit setting. In fact, limit setting and data interpretation are co-mingled activities. "hen oil analysis limits are properly set and the correct tests are performed at the right frequency, data interpretation is easy and efficient. The strategic use of goal-based and aging limits enables proactive maintenance to be carried out at the highest level. Likewise, when rate-of-change and statistical limits are deployed, the benefits of early fault detection are achieved. The combination of these limit-setting strategies affords the broadest and most effective protection for the plant equipment and its lubrication assets.

Clean Oil Sampling How to Sample Oil Without Opening the Bottle 2roper oil sampling does not have to be a difficult or unpleasant experience. In fact, when the best sampling methods %and hardware& are employed the experience should be ust the opposite. There are many important lessons to learn when it comes to pulling a sample. 7any run contrary to intuition. 7ost of them have to do with insuring that what enters the bottle is both rich in information and remains undisturbed by the sampling process itself. It is this later concern that is addressed by this article.

7odern oil analysis programs typically include some tests that can be influenced by environmental contaminants entering the oil %bottle& during the sampling process. Tests of greatest risk include particle counting, elemental spectroscopy, and total acid number. In situations where there is considerable dust in the environment at the time the sample is pulled, a concerted effort needs to be made to insure that this dust does not contaminate the oil. +xamples of high-risk situations5environments include mine sites, construction sites, primary metals industries, foundries,  windy outdoor conditions, and sample points close to the ground. +xperiments on the influence of environmental dust on particle counts have shown a marked effect. It is not unusual for an oilGs I/1 code to increase 3-8 range numbers when a bottle is left open ust a few minutes. (or instance, the actual oil sampled might be a rather clean I/1 85B but after exposing the sample bottle to atmosphere it can show an I/1 =58. The amount of dirt needed to accomplish an I/1 =58 is only about  ppm. *ecently, a new method called Hclean oil sampling has emerged that greatly simplifies the process and minimi#es the risk of dirt entering the bottle. It involves the use of common #ip-lock sandwich bags and sampling hardware such as vacuum pumps and probe devices. elow is an outline description of this procedure9 Step One 1btaining a good oil sample begins with a bottle of the correct si#e and cleanliness. The topic of bottle cleanliness will be discussed in greater detail in a future issue of 2racticing 1il 'nalysis. $owever, it is understandable that the bottle must be at a known level of cleanliness and that this level should be sufficiently high so as not to interfere with expected particle counts.

/ome people relate this to a signal-to-noise ratio, i.e., the target cleanliness level of the oil %signal& should be several times the expected particle contamination of the bottle %noise&. (or more information on bottle cleanliness refer to I/1 8>33. StepTo efore going out into the plant with the sample bottles place the capped bottles

into very thin #ip-lock sandwich bagsA one per bag. Jip each of the bags such that air is sealed into the bag along with the bottles. This should be done in a clean-air indoor environment in order to avoid the risk of particles entering the bags along with the bottles. 'fter all of the bottles have been bagged, put these small bags %with the bottles& into a large #ip-lock bag for transporting them to the plant or field. /ampling hardware such as vacuum pumps and probe devices should be placed in the large bag as well. Step Three 'fter the sampling port or valve has been properly flushed %including the sampling pump or probe if used& remove one of the bags holding a single sample bottle. "ithout opening the bag, twist the bottle cap off and let the cap fall to the side within the bag. Then move the mouth of the bottle so that it is away from the #ip-lock seal. Do not un#ip the bag. Step Four Thread the bottle into the cavity of the sampling device %vacuum pump or probe&. The plastic tube  will puncture the bag during this process, however, try to avoid other tears or damage to the bag %turn the bottle, not the probe or pump, while tightening&. If a probe device is used, it is advisable to break a small hole in the bag below the vent hole with a pocket knife. This permits air to escape during sampling. Step Five The sample is then obtained in the usual fashion until the correct quantity of oil has entered the bottle. 0ext, by gripping the bottle, unscrew it from the cavity of the pump or probe device. "ith the

bottle free and still in the bag, fish the cap from the bottom of the bag onto the mouth of the bottle and tighten. Step Si! "ith the bottle capped it is safe to un#ip the bag and remove the bottle. ;onfirm that the bottle is capped tightly. The bottle label should be attached and the bottle placed in the appropriate container for transport to the lab. Do not reuse the #ip-lock bags. "ene#its o# $lean Oil Sampling This simple procedure effectively permits samples to be obtained without exposing the fluid or the bottle to the atmosphere or surface contamination. ' clean sample can even be obtained with dirty hands. There are no expensive materials to purchase and the technique can be applied to a large number of sampling situations. (or best results practice a few times with a spare bottle until the technique is perfected.

What Particles Mean and Why They Need to  be Monitored and Controlled With the idespread use o# plant%level particle counters& maintenance organi'ations are increasingly more sophisticated and s(illed in the management and control o# oil cleanliness) This has led to the discovery o# a host o# ne tactics and practices in com*ining the particle counter ith other important onsite oil analysis tools and methods+ many unheard o# ,ust a #e years ago) ;ontamination can be defined as any unwanted substance or energy that enters or contacts the oil. ;ontaminants can come in a great many forms, some are highly destructive to the oil, its additives, and machine surfaces. It is often overlooked as a source of failure because its impact is usually slow and imperceptible yet, given time, the damage is analogous to eating the machine up from the inside out. "hile it is not practical to attempt to totally eradicate contamination from in-service lubricants, control of contaminant levels within acceptable limits is accomplishable and vitally important. 2articles, moisture, soot, heat, air, glycol, fuel, detergents, and process fluids are all contaminants commonly found in industrial lubricants and hydraulic fluids. $owever, itGs particle contamination that is widely recogni#ed as the most destructive to the oil and machine. This explains why the particle counter is the most widely used instrument in oil analysis today. 'nd, the central strategy to its success in reducing maintenance costs and increasing machine reliability is proactive maintenance. Proactive Maintenance)  "hile the benefits of detecting abnormal machine wear or an aging lubricant condition are important and frequently achieved with oil analysis programs, they should be regarded as low on the scale of importance compared to the more rewarding obective of failure avoidance. This is achieved by treating the causes of failure, not simply the symptoms. 'nd,is is the foundation of the popular practice known as proactive maintenance. In fact, the only effective  way to obtain simple solutions to complex machine maintenance problems is through proactive maintenance. "henever a proactive maintenance strategy is applied, three steps are necessary to insure that its benefits are achieved. /ince proactive maintenance, by definition, involves continuous monitoring and controlling of machine failure root causes, the first step is simply to set a target, or standard, associated with each root cause. In oil analysis, root causes of greatest importance relate to fluid contamination %particles, moisture, heat, coolant, etc.&. $owever, the process of defining precise and challenging targets %e.g., high cleanliness& is only the first step. ;ontrol of the fluid:s conditions within these targets must then be achieved and sustained. This is the second step and often includes an audit of how fluids become contaminated

and then systematically eliminating these entry points. 1ften better filtration and the use of separators are required. The third step is the vital action element of providing the feedback loop of an oil analysis program. "hen exceptions occur %e.g., over target results& remedial actions can then be immediately commissioned. !sing the proactive maintenance strategy, contamination control becomes a disciplined activity of monitoring and controlling high fluid cleanliness, not a crude activity of trending dirt levels. (inally, when the life extension benefits of proactive maintenance are flanked by the early warning benefits of predictive maintenance, a comprehensive condition-based maintenance program results. "hile proactive maintenance stresses root-cause control, predictive maintenance targets the detection of incipient failure of both the fluid:s properties and machine components like bearings and gears. It is this unique, early detection of machine faults and abnormal wear that is frequently referred to as the exclusive domain of oil analysis in the maintenance field. Managing Particle $ontamination) There is no single property of lubricating oil that challenges the reliability of machinery more than suspended particles. It would not be an exaggeration to refer to them as a microscopic wrecking crew. /mall particles can ride in oil almost indefinitely and because they are not as friable %easily crumbled& as their larger brothers, the destruction can be continuous. 7any studies have shown, with convincing evidence, the greater damage associated  with small particles. /till, most maintenance professionals have misconceptions about the si#e of particles and the associated harm caused. These misconceptions relate to the definition people apply to what is clean oil and what is dirty oil. 'nd, it is this definition that influences the setting of appropriate target cleanliness levels for lubricating oils and hydraulic fluids. The process is not unlike a black box circuit. If we want a change to the output %longer and more reliable machine life& then there must be a change to the input %a lifestyle change, i.e., improved cleanliness&. (or instance, itGs not the monitoring of cholesterol that saves us from heart decease, instead it:s the things we do to lower the cholesterol. Therefore the best target cleanliness level is one that is a marked improvement from historic levels. "hile there are numerous different methods used to arrive at target cleanliness levels for oils in different applications, most combine the importance of machine reliability with the general contaminant sensitivity of the machine to set the target. The *eliability 2enalty (actor and the ;ontaminant /everity (actor are arrived at by a special scoring system that is included with the Target ;leanliness )rid. There are many expensive ways to achieve clean oil but experience has taught us the wisdom of contaminant exclusionKtreating the cause not ust the symptom. y effectively excluding the entry of contaminants and promptly removing contaminants when they do enter, the new cleanliness targets are frequently achieved. ;oncerns that filtration costs will increase are not often reali#ed due to the greater overall control, especially from the standpoint of particle ingression. Particle $ounting-The ./nvisi*le. Filter) +ngineers learn that controlled systems are those that have feedback loops. In proactive maintenance this is the monitoring step, i.e., particle

counting. If this is done on a frequent enough basis, not only is proactive maintenance achieved but also a large assortment of common problems can be routinely detected. 's such, particle counting is an important catch all type test. ecause of the obvious value, it is not uncommon to find organi#ations testing the cleanliness of their oils as frequently as weekly. The activity of routine particle counting has a surprising impact on achieving cleaner oils. "hen the cleanliness of oil:s are checked and verified on a frequent basis a phenomena known as the invisible filter occurs, which is analogous to the saying, what gets measured gets done. ecause a great deal of dirt and contamination that enters oils often come from the careless practices of operators and craftsmen, the combined effect of monitoring with a modicum of training can go a long way toward achieving cleanliness goals. The following are common proactive and predictive maintenance uses of an onsite particle counter9 Proactive Maintenance0 . *outinely verify that in-service oils are within targeted cleanliness levels. 3. ;heck the cleanliness of new oil deliveries. 8. Euickly identify failed or defective filters. ?. ;onfirm that seals and breathers are keeping contaminants out. 6. ;onfirm that systems are properly cleaned and flushed after repair. =. ;onfirm that new machines are cleaned and flushed before use. >. Identify the use of dirty top-up containers and poor maintenance practices. 4. Identify the timing for filter cart use. Predictive Maintenance0 . Identify early-stage abnormal machine wear with quick confirmation by repeating. 3. Identify the location5source of abnormal wear by multi-point isolating methods. 8. . 1il analysis data interpretation 4. (iltration and contamination control F. "ear particle analysis machine fault detection 1nce these fundamentals are in place, oil analysis can move forward enthusiastically, beginning  with the development of its mission and goals. 'nd, instead of indifference to oil analysis exceptions, rapid-fire corrections are carried out and measures are taken to preempt their reoccurrence. In time, unscheduled maintenance becomes rare and oil analysis exceptions are few as the ideali#ed machine operating environment becomes controlled. (inally, as the many elements of oil analysis and proactive maintenance merge together into a cohesive maintenance activity, the benefits should not be allowed to go unnoticed. !nlike many applications of new technology, proactive maintenance seeks non-events as its goal and reward. These non-events include oil that continues to be fit-for-service, machines that don:t break down, and inspections that don:t need to be performed. This quiet existence is the product of a highly disciplined activity but, at times, can be misunderstood by the casual observer as unneeded. Therefore, the close association of the activities of proactive maintenance with the benefits of proactive maintenance must be measured, monitored, and displayed for all to view.