With the advent of fine filtration of lubricating oil streams in aircraft engines, new methods and instrumentation are required to determine metals for engine wear monitoring. One approach is to flush out and analyze the particles retained in the filter. This technique, known as filter debris analysis, involves rinsing particles from the filter for collection on a membrane filter which is in turn analyzed to determine the composition and quantity of the debris. Energy Dispersive X-ray Fluorescence (EDXRF) analysis of such a sample provides unambiguous results for all elements above atomic number 10. Compilation of a significant database of analysis results and correlation of EDXRF filter debris analysis with the existing SOAP database would expand the usefulness of the technique.

This presentation provides an introduction to the technique of EDXRF, including an overview of excitation, emission and detection of the fluorescent X-ray spectrum. Also described are the physical and analytical characteristics of an EDXRF analyzer. Finally the application of EDXRF to analysis of filter debris is described, including typical spectra, data processing and analysis results.

INTRODUCTION

X-ray Fluorescence (or XRF) is an elemental analysis technique that has been used extensively for analysis of metals, but has seen only limited use for routine analysis of wear metals to this point. This work takes a new approach to wear metals monitoring involving analysis of particles collected from the oil filter rather than analysis of the oil.

The development of fine filtration systems has created a problem for those involved in analysis of engine oils for wear metals. These newer filters effectively remove so much of the particulate metal from the oil, that analysis for metals remaining in the oil does not reliably predict engine failure.

Without a replacement for the long used and reliable spectrometric oil analysis method, or SOAP, it has been necessary to increase the frequency of maintenance of these engines to avoid the possibility of catastrophic failure. However this solution is expensive and inefficient. This situation has lead to the search for new ways of measuring engine wear.

This paper starts with a brief review of the problems with the present SOAP method. Second is a review of the X-ray Fluorescence Analysis Method including generation and interpretation of the X-ray spectrum, the effects of sample thickness on response, and the types of XRF systems available today. Third is a description of the application of XRF to Filter Debris Analysis, including the data we have collected so far. The data shows that this method has excellent sensitivity for the elements of interest, results are obtained with a minimum of labor and expense, the analysis is non-destructive, and the analytical results have the potential to be correlated with existing historical SOAP data.

LIMITATIONS OF PRESENT SOAP METHOD

The oil filters used on F18 aircraft and other new systems have been improved to the point where particles larger than 1 um are virtually eliminated. As a result, oil taken from aircraft engines equipped with fine filtration systems does not show increases in metal content that in the past were used to evaluate wear and detect incipient failure.

Optical Emission Spectroscopy and Atomic Absorption do not effectively detect particles larger than 2 - 5 um. Even in an engine without a fine filtration system the largest particles, those of most concern are not detected by optical emission techniques but are detected by XRF.

Research done by others has indicated that XRF analysis of oil for wear metals can detect wear metals earlier than SOAP because of this ability to detect large particles in the oil.

BENEFITS OF FILTER DEBRIS ANALYSIS

Since the metals of interest by design are collected in the oil filter, this seems the logical place to find samples to analyze. In the case of engines with fine filtration, this is the only place to find significant concentrations of wear metals.

The oil filter collects the debris and separates it from the oil. This effectively acts to preconcentrate the sample. When flushed out and collected on a membrane filter in the form of a thin layer of small metal particles, the sample is ideal for XRF analysis.

This approach to wear metals analysis need not be limited to monitoring systems with fine filtration. Although conventional filters would not necessarily trap and hold the smaller particles, the method could be used to collect and analyze the larger particles. The large particles are of most concern as they do the most damage, and when ground up produce the smaller particles analyzed in the oil by OE.

WHY EDXRF ANALYSIS?

The reason for selecting XRF for the analysis of filter debris include characteristics of the XRF method, and the instrumentation. Because of the nature of X-ray Fluorescence, spectral data is acquired non-destructively without the need for acid digestion. For this reason, XRF is often the method of choice for analysis of metals and other solids that are difficult to dissolve.

The method is relatively fast, especially considering the number of elements determined. In our work, XRF analysis of 18 elements is accomplished in about 10 minutes.

Energy Dispersive XRF (EDXRF) systems tend to be very stable. In most cases it is unnecessary to re-calibrate for months at a time. They can be configured to be rugged and transportable for use outside laboratory environments. Our experience with this instrument is that it can be plugged in, turned on, and operated successfully after shipping by Fed X without being re-calibrated.

Components of the XRF Spectrometer

The type of system used for this work is an Energy Dispersive XRF (EDXRF) analyzer with a solid state silicon detector. This type of XRF system is called an Energy Dispersive because the detector is used to measure the energy of the X-ray photons from the sample without the need for a diffracting crystal.

Other types of XRF spectrometers include Wavelength Dispersive XRF (WDXRF) and EDXRF spectrometers using proportional counters. WDXRF systems have better resolution than EDXRF systems but are larger, more expensive, and not considered transportable. EDXRF spectrometers that use gas filled proportional counters do not have the resolution to separate emission lines of elements close in atomic number and are effective over a more limited energy range.

The components of the EDXRF spectrometer are the X-ray tube, detector, and a PC computer based analyzer.

Figure 5 - Block diagram showing the components of the EDXRF spectrometer.

EXPERIMENTAL

XRF Hardware

For this work the instrument used is a Spectrace Model 6000. Since this work was done, the Model 6000 has been replaced with the QuanX. The system is equipped with a silicon detector that resolves the primary emission lines of adjacent elements and can analyze all elements from sodium and up in atomic number. The detector is electrically cooled using a Peltier device, thus avoiding the need for liquid nitrogen used in most such detectors today. The heat sink for the detector and for the X-ray tube are air cooled, so the only utility requirement is for power. Listed below are some of the specifications and characteristics of the QuanX.

• X-ray tube: 50KV, 50 watts             
• X-ray detector: Si(Li) electrically cooled             
• Resolution 170eV             
• Transportable             
• Utility requirement: 1500 watts AC power             
• Compact Footprint

The principle advantages of a system like this are that there are few moving parts, making the system relatively stable and insensitive to vibration. Furthermore, the system is small enough to be moved in the trunk of a car, and has frequently been setup in the mobile laboratories used by environmental contractors.

Standards

Table 1 shows the elements analyzed, and the materials used as standards for calibration. All of the standards were pure metal and oxide thin films. Use of Fundamental Parameters software allows for standardless analysis of some of the elements.

Table 1 - Elements analyzed and standards used for calibration of each.

Excitation Conditions

One way that EDXRF performance is enhanced is through use of a filter in the primary beam of the spectrometer. Since the elements of interest occur over a wide energy range, three different analysis conditions are used. Figure 7 shows the three spectra acquired from a typical sample.

Each analysis condition uses a different combination of X-ray tube voltage and beam filter to tune the excitation for a particular group of elements. This provides the optimum spectrometer performance for all the elements of interest. The spectrometer automatically steps through the three conditions and compiles the results.

In the 7KV condition, you may note that the elements P, S, K, Ca, and Rh have peaks in the spectrum but are not analyzed. The Rh X-rays originate in the X-ray tube and are scattered by the sample into the detector, and so do not represent an element in the sample. The Ca may be a component of the membrane filter. P, S, and K were simply not considered. An indication of the sensitivity of the method is shown by the lead and niobium peaks you see in the 30KV spectrum. These elements are present at about 10 times the detection limit. Zirconium was only a trace in some of the samples and was not considered. Use of a very thick beam filter in the 50KV condition reduces the background, allowing for detection of cadmium at very low levels.

Figure 7 - Spectra of the three excitation conditions and concentration results for a typical sample.

Debris Samples and Sample Preparation

The filter debris was collected from used engine oil filters of the F18 aircraft. No data was available concerning the condition of the aircraft or filter such as the number of hours of service for the filter or the number of hours of service for the engine since the last overhaul. The filters were not pretreated in any way, they were sampled as received. After sampling, the filter could be returned for reconditioning and re-use.

The sample preparation process for this work involves placing the oil filter in solvent and using a sonicator to loosen the particles. The solvent used was Electron (Ecolink Inc.) a non-hazardous organic solvent. After 5 minutes, the oil filter is removed and the particles are filtered from the solvent using a membrane filter and a vacuum filtration setup. A 1 micron Poretics 1um polycarbonate membrane filter was used to collect the filter debris. The vacuum is run until the filter is dry.

The deposit is collected within a 34mm circle on the filter. When wet with the organic solvent, the debris tends to stick to the filter, but allowed to dry out, the material can flake and fall off. To improve the stability of the samples, a water base varnish was used to improve the adhesion of the samples to the filter. When the varnish dries, the sample is ready for analysis. The membrane surface can be sprayed with a water based varnish to attach the particles. The sample is now ready for XRF analysis. See Figure 8.

The method requires about 10 minutes of labor to prepare the sample. Real time for preparation of the sample is about 30 minutes. Most of that is spent in the final filtering step.

Figure 8 - The filter debris sample collected on a Poretics 1um polycarbonate membrane filter.

Evaluation of the Filter Debris sample

XRF analysis of the filter provides the mass per unit area of the debris and the concentration of the elements in the debris. These values are then converted to micrograms of the element per gram of solvent used, and expressed as ppm.

As previously mentioned it is important that the samples be less than infinitely thick in order to measure the mass of the debris. Based on the typical composition of the samples encountered, Fundamental Parameters software was used to model the intensity vs. thickness curve for all the elements of interest assuming a homogenous thin film sample. Results of this determination, shown in Figure 9, indicate that the filter debris samples were infinitely thin for Ti and all of the higher atomic number elements. Which means that we can accurately measure the loading of the filter for these elements.

Figure 9 - Theoretical plot of intensity as a function of debris loading on the filter. The plot shows that the response is linear for elements above silicon, and logarithmic for the lower elements.

For the lower atomic number elements Mg, Al, and Si, the debris samples range from infinitely thin to infinitely thick . Fortunately, this does not undermine the thickness determination as these are not major constituents of the sample.

Another presentation at this conference addresses the possibility of collecting the debris in separate particle size fractions. This would have the twofold benefit of reducing the thickness of the samples and making it possible to accurately analyze large particles.

RESULTS

The XRF analysis provides results in units of weight percent for the metals in the debris, and the mass of the debris in units of mg/cm². Using these values together with the volume of solvent used, and the area of the sample on the filter, we calculate the concentration of the metals in the solvent.

The data in Table 2 demonstrate the conversion of weight percent and loading values to ppm and gives the lower limits of detector calculated for the analysis.

Table 2 - Example of conversion of Wt% and loading values to ppm of the element in solution. LLD values are calculated using a 3 sigma uncertainty.

Conversion of the results from mass to concentration in this way is somewhat arbitrary, as the volume of solvent used is not connected to the volume of engine oil. However it produces results in a familiar and understandable form. Nevertheless these results are not directly comparable to results from the SOAP method at this time.

Of particular interest is the fact that the typical values were so much higher than the detection limits.

Analysis Results

The results in Table 3 show the range of results observed. The values in bold indicate the highest results for the given element. It is clear that a variety of materials are contributing wear particles to the filter debris, and at different proportions in different engines. Also of interest are the wide engine to engine variations.

Table 3 - These analysis results are for samples that were at or near the upper limit for one or more elements.

Unfortunately, given such a small data set, and no information about the history of these engines to correlate with the data, it is not possible to make judgments about the condition of the engines at this point. Another way to illustrate the data is with a frequency distribution diagram.

Frequency Distribution

In order to visualize the data being produced by this method, a frequency distribution chart was created. The data, shown in Figure 10, includes results for the 26 filters that were sampled and analyzed.

Figure 10 - Population distribution plots for six of the analyzed elements.

The analysis results for Al and Si, are relatively low, and appear to show a normal distribution. The data for Ti showed only one sample above 2ppm, and that sample was 4 times above the rest. Although the results for iron would be considered high by OE standards, the results are not directly comparable, and so are not necessarily indicative of a problem. The cobalt data is interesting mainly because this element is not currently analyzed as part of the spectrometric method.

CONCLUSIONS

Evaluation of the range of concentrations observed, and the frequency distribution data indicate several interesting facts. First is that there are significant differences in the proportions of the metals detected from one engine to the next. This seems to indicate that there are different parts contributing to the wear metal debris in one engine that in another. Second is that there is a wide range of concentrations of wear metals detected. This implies a wealth of information available when a larger body of data is analyzed.

The method clearly provides qualitative and quantitative analysis of wear metals for engines with fine oil filtration systems. Lower Limits of Detection (LLD) are more than adequate to measure the elements of interest. LLDs for the method are 10 to 100 times lower than the typical sample for most elements.

The overall analysis from start to finish takes as much as 40 minutes, but the technician is only actively involved for 10 minutes or less per sample. Analysis of filter debris by EDXRF is a fast, inexpensive, and sensitive method for determining wear metals generated by engine systems.

The solvent used does not contain chloro-fluoro carbons, and is not a hazardous waste liability. It can be disposed of along with used engine oil

XRF analysis does not consume or destroy the sample. The debris sample can be saved for archival purposes, examined by another analytical method after the XRF analysis, or saved for comparison with later samples from the same aircraft. The samples are quite small, and could practically be slipped into an envelope and stapled to the printout of the analysis and saved with other maintenance documentation for the engine.

The method can be used for (but is not limited to) analysis of wear metals from engines with fine filtration system

With minor modifications, this unit is field deployable because use of the electrically cooled detector eliminates the need for liquid nitrogen usually required for systems with equivalent performance. This type of spectrometer has proven to be rugged and reliable in field operations for environmental applications.

FUTURE WORK

The data we have so far is limited. The first step is to collect more data. Without a body of data large enough to perform statistical analysis, it is difficult to gauge with certainty the usefulness of the method. When that data exists, the really interesting job of looking for a correlation between analysis results and actual engine condition can start. The possibility of being able to correlate the EDXRF data directly with the existing SOAP data looms as a really exciting possibility.