The Tornado is the Royal Air Force's (RAF) front line combat aircraft and is powered by two Rolls-Royce RB199 engines. The engine is a highly complex triple spool system generating up to 16,000 lbf of thrust. As a consequence of high g-loads, high end-thrust, and high cyclic loads, rolling contact fatigue wear is a natural bearing failure mode.
The buried location and hostile operating environment of the bearing chambers and poor signal transmission paths preclude vibration analysis as an effective condition monitoring tool. As such, wear debris monitoring (WDM) is the principal method used to detect the onset of bearing failure.
However, in the past the techniques employed proved both technically undiscerning and unresponsive to operational demands. To protect its aero-engine assets and eliminate wasteful maintenance costs the RAF recently invested in state-of-the-art Scanning Electron Microscopy (SEM/EDX) technology.
The Problem
The RAF has experienced difficulties in monitoring the in-service integrity of the engine oil-wetted components, specifically No 4 thrust bearing. The worst case of bearing failure results in seizure of the turbine, while the effects of secondary damage potentially threatens airworthiness.
Potential reliability problems exist with the increasing exposure of the gearboxes; both of these components are supported by condition-based maintenance.
Degradation of engine components caused by direct mechanical wear, or caused indirectly by suspended debris, demands highly discriminative and swift analysis in support of the flying task. The WDM and diagnostics are challenged by debris migration and cross contamination between the engine's oil sub-systems, i.e., the bearing chambers.
In an effort to find a solution, the RAF procured an Energy Dispersive X-Ray Fluorescence spectrometer (EDXRF) to supplement its existing wear debris analysis techniques of magnetrometry and optical microscopy.
After 24 months of in-service experience with the EDXRF it was decided that an investigation into other optional technologies should be undertaken. The outcome was the decision to explore the use of Scanning Electron Microscope Energy Dispersive X-ray analysis (SEM/EDX) as a single-point replacement for all past wear detection methods including the EDXRF spectrometer.
It was found that using the SEM/EDX acquiring meaningful spectra took only a fraction of the time taken by EDXRF. And, comparison of x-ray intensities with those from "standard" materials enables precise measurement of multi-element concentrations.
Requirements and Application of the SEM/EDX by the Royal Air Force
After a 6-month trial, the RAF procured the LEO's digital SEM/EDX (upgraded to LEO JetSCAN). While maintaining laboratory standards and continuity, JetSCAN's automated routines maximize sampling throughput, making it practical for field use.
Use of standard hardware, firmware and software resulted in a customized system while maintaining full SEM/EDX versatility. The overall control program is initialized through a simple sequence of "windows" operations. And, quantitative and qualitative data that is gathered from each individually scanned particle is processed by the computer algorithm to provide engine-specific analyses.
Known engine-failure characteristics have been established from the Failure Mode Effects Criticality Analysis (FMECA) and case studies. Limits of general engine behavior are used to train the software to identify normal, marginal, and abnormal wear conditions and alarm accordingly. The JetSCAN SEM/EDX provides the following additional wear particle monitoring capabilities:
• Automated analysis of 100% of chip specimens with full compositional (x-ray) analysis of 1500 individual particles (one full engine set - 5 chip detectors) with diagnostic recommendations within 15 minutes.
• Qualitative and quantitative analyses of multi-element materials for particles >5 microns, discriminating between active (deleterious) and benign wear debris; full particle characterization, and automatically de-linking of overlapping particles (software tweezers).
• An accurate wear-origin diagnostic capability using X-ray composition as the primary analysis, backed-up by morphological analysis of particles.
• A consistent statement on the conditional health of the engine, discriminating between a range of damage and failure mode symptoms with >90% confidence.
• Human/machine interface variability; auto control to enable concurrent analysis and hands-off functionality.
• Simple and environmentally acceptable means of sample preparation, without polishing and coating.
• User friendly operation and on-the-job training
The diagnostic software module provides:
• Use of historical data to create engine specific wear data, trending and distinguishing between critical and non-critical engine components using three alarm thresholds of normal, marginal and abnormal.
• Normal operation and different failure modes used to establish limits of wear generation for certain materials (alert/rejection criteria).
• Capability of automatically alerting the engine module of component specific problems based on trend, knowledge rules and fault signatures.
Sample Preparation, Processing, and Analysis
Following their removal from the engines, magnetic plugs are sent in for analysis. Solvent cleaning is carried out first to remove organic residues from the plugs. Then, all captured wear debris is transferred directly onto an adhesive, electrically conductive medium suitable for SEM use.
Debris is manually manipulated into the desired field of view within the central region of the tab. Particle spread is checked using a low powered optical microscope.
The overall sampling process is seamless from identification to detector analysis to diagnostic interpretation. Particle identification is accomplished by means of Image Analysis (IA) thresholding of a Back-Scattered Electron (BSE) SEM image (carbon grey level = 0, particles 70-250), see Figure 1.
The SEM possesses sufficient electron optical resolution at low magnifications to enable the collection of back-scattered electrons (BSE) and generate x-ray information from individual particles.
The custom-calibrated BSE imaging allows thresholding of the required engine surface metals only. If the system generated just one critical wear material type, the threshold process could screen for that material only. However, in this case, all of the engine's primary material surfaces require analysis, and several bear close elemental resemblances.
Yet, as shown in Figure 1, it is possible to eliminate materials that are less than the mean gray level of the lightest materials requiring analysis. For the RB199 engine this enables the routine elimination of nickel-rich abrasive contaminants which originate from air and oil labyrinth seal debris.
Image processing of particle features gives individual, morphological measurements, i.e., surface area (microns squared), peripheral distance, aspect ratio, compactness, edge roughness, and shape (wear type) to aid the fault type identification.
The process ensures that each particle feature is separated from surrounding features and that noise is eliminated from the resultant image. A wear mode classifier has additional potential.
Chemical filtering is the third process and is deployed to enhance discrimination between active and benign wear debris. The system filters out benign materials and then classifies remaining critical alloy types. The classification criteria accounts for expected material manufacture tolerances, material homogeneity, x-ray dispersion and statistical process variability.
Ten RB199 transmission materials are classified according to their failure characteristics and multiple combinations of materials can be addressed. JetSCAN's x-ray analyzer is configured to classify automatically all main line/gearbox bearing materials, gear, seal, shaft, housing, and spacer materials.
Analysis is not size dependent, as the electron beam scans a number of chords over each particle position. Sequential processing of up to 20 wear particle chip plugs is achievable.
Fault derived data is imported into a Rolls-Royce diagnostic package, where processing is based on a rules-based risk assessment. It also provides appropriate maintenance recommendation.
JetSCAN's automated process takes less than one hour to analyze four full engine sets of samples, i.e., about 5,000 particles. It achieves data correlation of < 2% variability. Agglomeration of wear debris specimens is inevitable, however, after 2-years of use, analyses showed that the agglomeration occurrences had no impact on analytical precision.
JetSCAN In-Service Experience
Eliminating the airborne contaminants from analysis reduced sampling process time by 90%. Instead it only highlights major wear-related problems that were difficult to characterize with past techniques.
Afterwards depot maintenance personnel validate the maintenance limits of each wear phase. Validation has proven that individual particle material composition and characterization is the key to effective diagnostics/prognostics, not overall bulk material properties. In fact, particular machine failure modes may only generate a few particles.
As mentioned, residual coatings and particle agglomerations have been the prime cause of unclassifiable results. These results are treated as novel, occurring at a level of < 2% of the total specimen content. If unclassified results exceed this threshold an alert is triggered, indicating poor sample preparation or a new unexpected failure is suspected.
Specific Case - No. 4 Bearing
The Rolls Royce engine No. 4 bearing gives little notice of impending failure. It is subjected to very high stresses, as much as 40 kN of thrust pushing on twenty-two, 14 mm balls mounted in a cage and running on a pair of split inner rings within an outer track.
The bearings run up to 19,400 rpm and are cooled by an oil jet. Fatigue is the natural process of bearing wear-out. As seen by the particle image in Figure 2, the fatigue particle is characterized by edge roughness, flaking and cracking.
Because bearing diagnostics is primarily based on elemental composition, JetSCAN is programmed to identify the following:
Initial Installation. Initial debris wear rate is high, as engine components bed-in, reducing to a known stable nominal condition, before built engines are passed off. It typifies the initial wear-in slope of a bathtub curve.
Normal Wear. Wear rate and particle generation is steady and at a benign level, but progressively leads to the condition of incipient failure. The nominal engine flight hours sampling frequency is exercised, i.e., the flat section of a typical bathtub curve.
Marginal Wear. Wear rate increases to a level that demands a closer watch and goes to reduced sampling intervals--initial upward gradient of the bathtub wear-out slope. The RB199 engine uses two-phases of reduced-interval sampling since high-g maneuvers can disturb old and finer lodged debris, which presents an analytical nuisance and skews diagnostics.
Abnormal Wear. Precipitous wear-rate increases exponentially up the wear-out slope, above safe levels for engine integrity. Mechanical instability due to the wear reaches the impending failure stage, i.e., the final phase.
Implementation and Field Performance
Since 1997, six JetSCAN (SEM/EDX) units have been deployed and have processed over 8,500 engine samples. Since deployment, no catastrophic bearing failures have occurred. Prior to SEM/EDX, five such incidents occurred annually.
Engine conditions alarmed by SEM/EDX diagnostics undergo depot maintenance, where incipient bearing failure has been confirmed in all cases. The high number of false and premature engine rejections prior to SEM/EDX, has been significantly reduced.
Post Project Evaluation
A £84 million ($136 million) cumulative life cycle cost savings is projected from the SEM/EDX units.
Primary Aim - Engine Integrity. In-flight shutdowns (No. 4 bearing) reduced by 90% - £2.8 million ($4.5 million) annual operational and logistic savings.
Secondary Aims- Logistical/Operational Efficiency
• 85% reduction in inconsequential engine rejections (50% reduction in all rejections)- £4 million ($6.5 million) annual cost savings.
• The system's rapid response (full diagnosis <1 hour from receipt) - £1.25 million ($2 million) annual operational cost savings.
Conclusion
The RAF took advantage of air-cooled x-ray detector technology to achieve improved reliability for harsher non-laboratory operating environments. It was prudent to invest in a technology that fully satisfied the need and application, offering both operational and logistical benefits. The JetSCAN is an unequivocal success and the RAF has acquired a truly 21st Century diagnostic tool.
The application of advanced electron probe microscopy for routine wear debris monitoring exceeded logistic and maintenance objectives. The transfer of laboratory-based equipment into front-line maintenance has been proven.
Its discerning technique has safeguarded engine integrity, enhanced Tornado aircraft operational availability and airworthiness and increased diagnostic confidence. It is fully supportive of conditional health maintenance, demonstrating the true value in developing diagnostics/prognostics.
Acknowledgements:
1. Data Systems & Solutions (Mr Nicholas Farrant)
2. LEO Electron Microscopy (Mr Allister McBride)
3. Institution of Diagnostic Engineers, Leicester UK
