Normalizing Elemental Data to “Rate Of Wear”

There are many ways to monitor the condition of your machinery through oil analysis and some methods can be employed beyond simply evaluating the raw data.

It can often be difficult to consistently sample your component in precise and exact intervals, i.e., every sample taken at exactly 100 hours or 20,000 miles for example. Assuming that your target sampling interval is every 100 hours, but actual hours at the time of sampling are higher or lower than your target, we can still calculate the measured data to approximate a typical value at 100 hours.

Some laboratories will include “rate of wear” data with your oil sample. The value provided is a calculated index which can be useful in identifying a divergence from normal wear patterns. This will give us a normalized “rate of wear” value to correspond with your sampling interval. Even if your lab does not provide this data it’s easy to do on your own. This is achieved by applying the data to this formula:

Example below: the level of iron measured in parts per million for nine diesel engine samples.

The values listed in the orange row represent the normalized rate of wear as calculated by the above formula. Under the engine’s typical operating conditions the normalized rate of wear should ideally be fairly consistent and within a few parts per million of each other from sample to sample.

Once a large enough dataset has been acquired we can estimate the average normalized rate of wear, which in the above case calculated to 74 ppm. If we exclude sample number 4 as an outlier we can see that the values are approximately within +/- 10 ppm of the average value. Therefore, we should expect future samples’ normalized rate of wear to be close to that value.

Significant deviations from and inconsistencies with the normalized rate of wear may indicate a problem. Investigate other data points such as contamination, viscosity, changes in oil type or brand, as well as other factors such as load, operator habits, noise or other operational observations, etc., as they will commonly correlate with the deviation. If the deviation cannot be explained by typical causes, the component should be closely monitored. If the deviation in data is extreme, an inspection of the component may be required.

Grant Dawson - Laboratory Data Analyst, OMA I, MLA I, MLA II