Wear debris generation in a tribosystem is a result of various interrelated factors. Debris formed from surface interaction is a valuable source of information on wear mechanism and mode.
Particle shape, texture and color can be used for wear analysis. Image recognition is used to solve the problem of debris analysis and classification for further use as a database of an integrated monitoring service.
Wear monitoring is considered for providing long-term operation, with no failures, at optimum friction performance of a tribosystem. Wear monitoring tools based on an analysis of debris accumulation in lubricated machinery are reviewed in the following discussion.
A typical curve of wear in a tribosystem vs. time is presented in Figure 1.
Figure 1. Time-dependent wear of tribosystem:
I – running-in;
II – stationary wear;
III – severe wear
The wear can be divided into three stages: running-in, stationary and severe. The first stage, running in, is relatively short. The second stage, stationary wear, should occur over a prolonged period of time at a minimal (mild) wear rate. Severe wear results from continuous surface interaction or lubricant degradation.
This third stage can and should be prevented by maintenance. The stages of wear combined with a variety of wear modes make wear simulation and prediction difficult. This problem requires a systematic approach.
Surface damage features and wear debris are the objects most often examined when considering the wear mode and mechanism. The shape, texture and other properties of debris provide valuable information on the mode and mechanism of wear.
These properties can also give information on the occurrence of severe wear and the necessity of maintenance. For this reason, wear debris analysis has attracted a lot of attention in recent tribological research.
Wear debris analysis is based on analysis and understanding of the relationship between the morphology of particles, character of surface damage and the state of lubricant. It allows one to obtain qualitative information on the type of surface damage (fatigue, cutting, etc.), severity of loading, local temperature and the quality of lubricant.
The interpretation of the information is challenging because tribological behavior and the morphology of wear debris are not clearly defined and require expertise for interpretation of data. Nevertheless, the debris analysis is recognized as a valuable source of information about machine performance.
According to commonly accepted concepts, wear particles can be divided into six main morphological types: rubbing, laminar, fatigue chunk, cutting, spherical, and dark and red iron oxides.
The term morphology usually includes such particle features as shape, texture and color obtained from their images. The analysis can be performed by detecting the specific particle types in wear debris samples and using a practical understanding of the particle features and symptoms to form an opinion about the underlying wear condition for the tribosystem.
Researchers and engineers are concentrating on the development of automated methods of wear debris analysis. The common approach in these methods uses an image acquisition device and standardized rules for representing the particle description and choosing an appropriate classification method.
A variety of methods are used for wear debris image acquisition. Images can be obtained by a scanning electron microscope (SEM), transmission electron microscope (TEM), or confocal and laser scanning microscopy. Optical microscopy is still the most popular method. The images from a microscope are typically captured on film. The photos are then scanned and stored in a personal computer (PC) for digital image processing.
To improve the efficiency of wear debris analysis, the authors developed a special attachment to an optical microscope and image-processing software. The device allows one to capture an image and deliver it to a PC (Figure 2).
Figure 2. Wear debris image capture system: 1) optical microscope
with CCD digital camera; 2) additional color monitor; 3) computer;
4) display; 5) color printer
It consists of an optical microscope equipped with a CCD digital color camera, a PC with peripherals and an image capture board. An additional color monitor can be used as an auxiliary visualization device. The system allows one to observe images on a PC display and capture them as single frames or as video sequence.
The shape of a wear particle is usually described in terms of the area, perimeter, form, etc. but these features are too general and the most valuable information on the particle shape can be lost. Moreover, the object is not described uniquely; there are many patterns with the same areas, perimeters and form factors but with different shapes.
When using the features based on Fourier transform, the initial contour of wear particles is presented by its signature. The signature is a function of particle radii R vs. contour length L.
Surface texture can be considered as a result of different spatial relations between surface elements. On the local level, the various characteristics of these relations correlate to the various morphologies of asperities; on the global level, various mutual spatial arrangements or texture patterns can be revealed.
Color is an important feature in wear debris analysis. If the shape and texture allow one to differentiate the wear particles according to their prehistory of formation, color may help to define debris composition. The current methods of color characterization are based on the fact that color can be specified in terms of three numbers representing the amount of the three primary colors added together.
Wear debris can be classified into a set of predefined morphological types. Each particle is differentiated by its shape, texture and color. The purpose of classification is to construct a decision-making method which distinguishes the particles.
At present neural net, production rules, fuzzy logic and statistical methods are used for solving this problem. There is no reason to use only one of these methods because the particles are classified by different feature sets.
For example: cutting and spherical particles are easily distinguished from each other and from laminar, fatigue and severe sliding ones only by shape features. Consider that some types of debris can be classified by texture parameters only. This is the purpose for using the three-stage classification procedure (Figure 3).
Figure 3. Wear debris classification:
SP – spherical particle; CT – cutting; LM – laminar;
FC – fatigue chunk; SS – severe sliding;
RO – red oxide; DO – dark oxide;
UN – unrecognized; ST – steel; CA – copper alloys
In the first stage, particles are differentiated by shape using neural net and features based on Fourier descriptors. The trained neural net classifies particles into three types: spherical, cutting and all others. In the second stage, particles recognized as “others” are classified as laminar, fatigue or severe sliding particles using the texture features.
Finally, all the particles are differentiated by color features into steel and copper alloy particles (Figure 4). Particles not recognized in the second stage are divided into red and dark oxide particles.
Debris classification data can be used in developing the databases and electronic atlases for further use in integrated monitoring systems.
Figure 4. Condition-monitoring systems based on wear debris analysis
The wear monitoring systems based on wear debris analysis can be divided into two classes, off-line and built-in devices.
Off-line systems are usually employed in laboratory conditions, utilizing periodic oil sampling with subsequent analysis. The information gathered through laboratory study using various methods (spectroscopy, ferrography, light scattering, flow ultramicroscopy, activation methods, etc.) is sufficient for reliable evaluation of the machine.
However, laboratory analysis requires expensive equipment and skilled operators. Time intervals between sampling and receiving the results on the machine condition are often unacceptably long. For this reason, portable field devices which can be used in service and repair centers have been developed.
Built-in: An alternative to off-line analysis is an increased interest in inexpensive built-in instruments which allow the monitoring of machine operation in real time. General requirements of built-in devices include sufficient information at a low cost, simplicity of mounting and service, and high reliability.
The commercial built-in devices are currently available on magnetic, electromagnetic instruments and those based on measuring difference in pressure of accumulated particles. The devices differ from one another by their range of particle sizes, type of detected particles, sensitivity to mass concentration of particles, and the method by which they’re installed in the lubrication system.
In-line devices provide for analysis at direct installation in the main oil line. Online devices must be mounted in a complementary oil line. Each apparatus finds application in the machines where it is the most efficient.
The development of real-time monitoring methods, such as built-in devices that provide continuous tracking of wear rate and machine severity, is a promising direction in wear debris diagnostics. Such devices make it possible to evaluate machine condition and predict their wear based on characteristics such as wear particle concentration and particle size distribution.
An online Opto-Magnetic Detector (OMD), developed at the Metal Polymer Researcher Institute of National Academy of Sciences of Belarus in cooperation with Korea Institute of Science and Technology and used for wear monitoring of heavy machinery in industry is an example of such monitoring techniques. Machine evaluation with OMD is based on oil analysis (Figure 5).
Figure 5. Evaluation of machine behavior with Opto-Magnetic Detector
The OMD measures variations in the optical density of used and clean oil and their differences under the effect of magnetic field. Optical density can characterize total contamination of the oil by oxidation and aging products, contaminating dust and wear debris. It can be a measure of contamination of the used oil compared to clean oil, defined as a total contamination index.
Variation in optical density of the used oil under the effect of magnetic field is proportional to ferrous wear debris concentration in the oil. It is explained by the effect of the field on the particles in the optical cell, which results in the particle removal from the optical axis to periphery of the cell.
Change in optical density can be related to concentration of ferrous debris in the oil. Rolling bearings of the turbine and gearbox, with higher rate of debris generation in the latter, were potential sources of wear debris in the air compressor. Because these components were made of steel, the oil contamination by ferrous debris can be a measure of wear of the components defined as a wear rate index.
OMD has been applied in the wear monitoring of an air-compressor system at Pohan Iron and Steel Company, a metallurgy plant in South Korea for two years. This was used in conjunction with vibration and temperature control detectors placed in several places (Figure 6).
Figure 6. Air-compressor system: V,
T – vibration and temperature sensors;
OMD – optical-magnetic detector
Figure 7 presents the data on monitoring the compressor during the 30 days before a serious failure occurred due to breakdown of the steel collar preventing the extraordinary slippage of the pinion in the gearbox in the axial direction. Figure 7a shows that the signal related to gearbox shaft displacement and acceleration has increased only a few days before the breakdown.
Temperature control of both bearings of the shaft hasn’t given any warning data as seen in Figure 7b. At the same time, oil contamination and wear rate indices (Figure 7c) gave an early warning three weeks in advance of failure and an emergency warning at least 10 days before the failure caused the system to stop, resulting in a significant economic loss.
The case study illustrates the efficiency of an integrated approach to machine maintenance in industrial conditions. Such an approach should combine both direct and indirect observation of the machine behavior.
Condition monitoring based on detection of debris in lubricant can be an efficient tool controlling machine condition. It should integrate the diagnostic devices, on line monitoring systems, data on machine operation and the knowledge accumulated in database.