Setting Oil Analysis Alarms: General Considerations

It is worth reviewing the two types of alarming methodologies that are employed in oil analysis.

Absolute alarms

These are alarms based on manufacturer’s recommendations and/or lubricant supplier technical bulletins. These alarms generally define working ranges or condemnation limits and are most applicable to lubricant and contamination condition. Extensive research is conducted to arrive at these limits, and they provide a good starting point for any analysis program. Absolute alarm limits matter greatly when warranties on new equipment are at issue. Failure to comply to the recommendations is often viewed as justification for not honoring such warranties.

Statistical Alarms

Manufacturer’s guidelines for alarm limits or general standards are disadvantaged in that they are based on average operational and performance situations, which may not accurately reflect the actual conditions of a specific machine. This is particularly applicable to machine condition. Statistical alarms limits are based on gathering a small sampling of data from equipment, analyzing the distribution of that data, and using this statistical characterization to set specific alarm limits. statistical trend analysis allows the identification of the equipment in greatest need of attention, thus allocating maintenance in an efficient way.

Combining Absolute and Statistical Alarms

Effective oil analysis exception management relies on the combination of both types of alarms. The following illustration is an example of the alarm combination. The condemnation limit is the absolute alarm. Statistical trending, taking into account variability based on the sampling, contamination, make-up oil etc. will develop the standard deviations. Departure from this normal variability signals genuine problems occurring. This is the earliest possible time to take action and head off problems. Failing this, as the trend approaches it’s warning limit, action such as changing or cleaning the oil, or inspection of the unit is required.

Fig 1: The idealized graph shown is an example of how absolute and trendline alarms are used together. The test used could be iron content, viscosity or other parameter. The normal result variability range takes into account minor variations caused by analytical accuracy, sample homogenity, etc. as the trend approaches the warning limit, Some action must be taken, either cleaning the oil or inspecting the equipment.

Establishing statistical alarms which provide the earliest possible warning without false alarms is a difficult task. Factors such as adding or changing oil, filter changes, sampling technique can distort results. However if the following foundation steps are employed, a very good oil analysis and lubrication management program will be the result.

1. Know your Equipment

Many engineers and technicians do not know their equipment from a lubrication viewpoint. Information such as oil wetted component metallurgy, equipment loading conditions and environment are very important, and contribute a great deal in solving problems shown up by oil analysis. Find the equipment maintenance manuals or call the project engineer who installed the machine. Keep the information close at hand in your file, and provide this information to the oil analysis laboratory.

2. Read your Equipment Specifications

Many manufacturers of equipment have published specifications for lubricants and their maintenance. All manufacturers of turbine systems and diesel engines have recommended guidelines with absolute limits for degradation, wear metals and contamination. Many manufacturers of compressors and hydraulics have similar notes. Put these limits into your trending system and let the lab know. You now have absolute alarms, and you stay within any warranties expressed by the equipment provider. The laboratory is happy because it has alarms specific to that piece of equipment.

3. Assess your lubricant

Oil analysis programs frequently find incorrect lube types and/or grades in equipment. A common problem recently is the use of new advanced synthetic lubricants, with different baseline properties than the original mineral based oil specified in the maintenance manual. Consult with your lubricant vendor and have a product application guide handy. Advise the manufacturer of the differences and ask your lubricant vendor to consult with them if there is a problem. Send the lab samples of all your new oil, so that a baseline analysis is performed. Laboratories generally do not charge for this analysis. The values obtained are your baseline for trending lubricant properties.

4. Understand Oil Analysis testing

Be familiar with the tests that have been very briefly described in this paper, and of any other tests as they are necessary. It is important to understand that all test methods are subject to some normal, predetermined variability, depending on the test. Viscosity testing has a normal variation of only 0.5%. In contrast, the total base number (TBN) test has a variation of 15%. A reported TBN value of 7.0 could actually represent a value of between 5.95 and 8.05. The accuracy ranges have been defined for the tests. Consult with the laboratory to account for variation in statistical trending analysis. The value of the analytical test as an effective alarm trigger decreases as the test variation increases.

5. Sample Carefully

Poor sampling can cause distortion of several hundred percent in sample data. Follow established guidelines for sampling, and sample from the same point each time. Air, oil and filter changes and conditions can result in distorted alarms. Similarly, not recording or reporting topping up and/or bleed & feed operations have the effect of artificially changing wear, contamination and additive levels without any real change in equipment or lubricant condition. These events must be recorded so as to take account of variations to trendlines.

6. Start logging operation time

Statistical trending programs are not scientifically valid without including operation time. A standard trend alarm for wear metals and contaminants is to signal an alert when an increase of 10 ppm over a 10 hour period is found in any of the spectrochemical analysis readings. Engine users have traditionally tracked time, in order to determine drain intervals, and schedule overhauls. Oil trend data for diesel engines is generally highly developed, with manufacturers specifying absolute limits and trend alarms. Rotating machinery users generally do not have tachometers fitted to equipment, and in cases of turbines, hydraulics and large gearboxes, the oil is present since installation, and the units are running continuously. Time based sampling the most common and effective method in these cases, especially when performed at regular intervals. Avoid random sampling, as their predictive value can be suspect. In all cases, record the date when sampling.

7. Establish Baseline trends

Taking one sample from a piece of equipment that has had no previous history will not give you information about the wear trend of the machine, and it is difficult to set alarms at this point. A monthly analysis for three consecutive months will establish a good wear trend. At this stage, a decision may be made on the sampling frequency of that equipment (with criticality factored in) based on the data received. This is an example of how the program manages both the equipment and itself by exception, thereby reducing the effort needed to maintain it by busy personnel.