Modern Strategies for Trending Additive Depletion and Contamination Using FT-IR

The application of Fourier Transform Infrared (FT-IR) spectroscopy in lubricant condition monitoring has rapidly increased. Initially, this was driven by mobile equipment operators using FT-IR spectroscopy for crankcase lubricants. In that application, FT-IR spectroscopy monitors common contaminants, breakdown products, and additive package components. However, FT-IR analysis can be applied to a much larger variety of mechanical and fluid systems, such as gear boxes, hydraulic systems, turbines, and other systems found in the manufacturing environment.

STRATEGIES FOR MONITORING CONTAMINANTS BY FT-IR
The spectral interpretation methodology is a major factor of importance in a condition-monitoring lab. Originally, an interpretation method would be developed for a specific system, such as water in a crankcase lubricant. Once developed, it would then be applied to every sample received by the laboratory, based on the assumption that an analysis methodology developed for one would equally apply to all. Then the analysis method fails to detect the fault in some samples that differ from the original system. From this, FT-IR analysis in general is presumed to fail, when in fact, the method used to interpret the data has been improperly applied.

Using the analogy to metal analysis by emission spectroscopy, a wear metal spectrometer may be configured to monitor bearing wear levels from the lead readings. However, if fluid from a machine using silver alloy based components is tested, monitoring the lead emission line will not be useful in indicating bearing wear. Like wear metal analysis, the proper regions of the spectrum must be monitored to detect the fault signatures of interest.

In FT-IR based condition monitoring, both the bulk matrix and the interactions between the matrix and analyte need to be considered. An example of this is shown in Figure 1. Here three different lubricants, a synthetic ester–based lubricant, a high detergent/dispersent crankcase lubricant, and an EP (Extreme Pressure) additive gear lubricant, all spiked with 1000 PPM of water, are presented. The area highlighted in blue indicates the optimal integrated band area for measuring water contamination in each. The different measurement regions are necessary because water reacts differently in the different systems. Different reactions result in the maximal water response appearing in different locations.

An illustration of how applying a methodology developed for one system to a different system creates misleading results is presented in Figure 2. Here, water measurement trends are presented from a gearbox that is lubricated with an oil containing an extreme pressure (EP) additive. The blue line shows the response for the water in EP additive oils, and is contrasted to the red line showing the measurement optimized for water in crankcase oils. As can be seen, the "Water in EP Oil" measurement generates a strong response from a water contamination problem, while the measurement optimized for crankcase oils only shows a slight deviation from the normal trends.

TO CALIBRATE OR NOT TO CALIBRATE?
While the example presented in Figure 2 shows a significant deviation from the normal response of water in an EP additive oil, the next question usually becomes "how much water is present"? Correlating the infrared response to a physical concentration, or calibration method, involves running a series of known standards and relating the infrared response to the physical concentrations. In the examples presented above, it would be a "relatively simple" process of blending water into a series of used oils known to be uncontaminated by water. While simple in concept, reliably blending low levels of water into oils becomes a significant test of the fluid handling and mixing skills of the laboratory.

Most condition monitoring laboratories running metal analysis spectroscopy are familiar with running periodic calibration standards. Running these standards is used to correct day-to-day variances in the emission characteristics of the plasma and other variances of the spectrometer. However, once a consistent and reproducible sampling method has been set up on a FT-IR, such daily calibration standards are unnecessary, as an FT-IR spectrometer is inherently self-correcting through measuring a background spectrum. This is a spectrum collected on the instrument (ideally including the sampling cell) without any sample present. Subsequent samples collected are then ratioed to this background, correcting for any variances in the source intensity or alignment effects.

As running calibration samples to correct day-to-day instrument variances is unnecessary, the original question of correlating the infrared response to a physical concentration remains. While a series of calibration standards can be prepared to generate a concentration reading instead of a simple infrared response, this can involve a significant amount of additional work for the laboratory. This entails not only preparing the calibration standards, but also documenting standard preparation, tracking standards, ensuring consistent standards quality, and so on. In addition, a proper calibration set for systems that are more complex should include not only the component of interest, but also be blended using a large set of used lubricants. This is necessary to take into account all the interactions and reactions with other components and contaminants which may be found in the fluid. Some laboratories have successfully generated such complex calibrations to predict physical properties like TAN, TBN and viscosity. However, this involves running 150 calibration samples or more, and using complex mathematical techniques such as PCR/PLS (Principle Component Regression / Partial Least Squares). Such mathematical techniques are beyond the scope of this article.

Ultimately, many laboratories are choosing to establish a simpler relationship, using a few prepared standards at or near their previously established alarm limits, and setting an alarm level based only on the infrared response. In the example figures above, an infrared reading of 100 for the Water in EP Oil correlates to a prepared water concentration of 1000 ppm. Thus, the alarm level for the “Water in EP Oil” infrared measurement would be triggered when the integrated area measurement reached 100. This example shows that an alarm would not be triggered on the absolute level, but maintenance action was still triggered based on the rate-of-change of this reading. Establishing such a relationship in a simpler manner preserves the reliability of the infrared measurements without adding in additional sources of error and headaches from standard preparation, storage, tracking, and so on.

ESTABLISHING ALARM LIMITS
While establishing alarm limits based on prepared standards is conceptually simple, in practice it becomes difficult due to sample handling concerns. Handling such compounds is difficult, and many lubricant manufacturers are not willing to prepare such standards on demand for individual laboratories. However, alarm limits can be established for these components in the same manner as alarm limits are established for wear metals. Using simple statistics and statistical distribution profiles, such limits can be easily established. Figure 3 presents an example of a distribution profile of the infrared integrated band area measurements for a phosphate-based antiwear component. This distribution profile was developed from 920 samples taken from a series of gearboxes of the same type. Just as in wear metal alarm limit determination, it is important that the same system, or systems with similar characteristics and sump size, is used in the statistical analysis. From the numerical results, the average infrared integrated band area across all these systems is 13.6, with a standard deviation of 1.8. Typically, caution flags are set when the measured value drops below two standard deviations, or 2 sigma, from the mean, and warning raised when the value drops below 3 sigma from the mean. Using this, a caution flag would be set when the antiwear reading drops below 10, and an alarm flag raised when it drops below 8.

An example showing the trend for the antiwear integrated band area is presented in Figure 4, in conjunction with the infrared water response. Note the drop in the antiwear reading at the start of this trend. This is not unexpected, as new lubricants, like new or rebuilt machines, will go through an initial "run-in" process (preferred to the phrase "break-in"). Components of new oil will undergo rapid change as they react with contaminants and breakdown products remaining in the system, and other additives adhere to metal surfaces. After this initial drop from 24 to 16 in the antiwear response, the level remained constant in next sample checked in early December. Sometime between December 4th and the next sample checked on December 22nd, an event occurred which caused the antiwear reading to drop to zero.

Examining the water response from this system immediately indicates the cause of this component loss: severe water contamination. Note that in this example, the severity of the water contamination (nearly 1%) not only adversely affects the operation of the machine, but has removed or destroyed the antiwear component. In this case, simply replacing the additive package, or completely changing the oil, would not itself correct the root cause of the loss. All factors monitored should be used in the process of diagnosing and correcting machinery and lubricant problems.

BASIC TRENDING ANALYSIS
While establishing alarm limits through statistical analysis is the preferred route, judgements on fluid quality and performance can also be established by monitoring multiple factors measured by FT-IR. In Figure 5, an example trend is presented from a petroleum-based hydraulic fluid. Here, the infrared response for degradation products of petroleum oxidation and nitration are presented, along with the infrared response of a hindered phenolic antioxidant. Antioxidants, such as hindered phenolics, are preferentially oxidized to protect the base fluid. Note that in this example, no attempt has been made to convert the infrared response numbers to a physical concentration. The large number of different partially oxidized and nitrated hydrocarbons produced makes preparation of known pure standards impossible. While such calibration could be prepared for a simple antioxidant as in this example, this too is unnecessary. As can be seen from the trends, as the infrared response of the antioxidant drops, response from the oxidation and nitration products increases. Once the antioxidant infrared response has dropped to zero, the response of oxidation and nitration products has almost doubled. While simply replacing the antioxidant might be an option, this would not reduce or "rebuild" the oxidized products back to the starting hydrocarbons. Here, the trends taken together, indicate that the fluid has probably reached the end of its useful life, and should be corrected by replacement or reclamation to remove the various degradation products.

CONCLUSIONS
FT-IR spectroscopy is a powerful and useful tool in monitoring lubricant and hydraulic fluid systems. While originally developed for monitoring "classic" diesel crankcase oils, proper application of consistent and appropriate integrated band area measurements expands the application to other systems such as synthetic ester fluids and EP additive fluids. Coupling FT-IR analysis with standard statistical analysis techniques and trending analysis gives today's condition monitoring laboratory a fast, flexible, and effective tool in monitoring lubricant quality, and diagnosing related problems across a wide range of mechanical and fluid systems.