Contamination is without doubt the most common problem affecting the reliability of industrial machinery. A study conducted by the National Research Council of Canada found that on average, 82 percent of wear problems are directly attributable to particle-induced failures such as abrasion, erosion and fatigue.
Of those problems not directly associated with particles, water - typically the second most failure-inducing contaminant - plays just as serious a role in promoting premature failure, resulting in corrosion and hydrogen-induced wear like blistering and hydrogen embrittlement.
A more recent informal survey of more than 30 plants including power generation, paper, food, chemical, cement, textile and other manufacturing industries found, not surprisingly, that dust, process contamination, wear debris and moisture, are the most common contaminants.
To effectively control wear by controlling contaminants, it is vital to design an oil analysis test slate that will provide early warning of impending problems related to both wear debris and common contaminants.
However, in choosing test slates, it is important to be aware of both the strengths and weaknesses of the various tests, and of how to apply the available oil analysis tools in the most effective manner. For example, corrosive wear typically generates a significant number of very small particles.
In this instance, looking at spectrometric data using atomic emission spectroscopy offers an excellent early warning tool because this technique is sensitive to small particles - typically five to eight microns and smaller, depending on instrumentation.
By contrast, other wear mechanisms may generate particles in different size ranges. For example, adhesive wear in a sliding contact situation, such as above and below the pitch line of a spur gear, might initially generate particles in the 10 to 20 micron range, and become rapidly larger as the problem progresses from incipient to severe wear.
In this case, atomic emission spectroscopy may not be the best choice to detect the ongoing problem; rather a combination of particle counting, ferrous density and microscopic analysis may be more effective for determining the problem.
In many instances, particle sizes start small and grow progressively larger as the problem increases in severity. Take, for example, a contact fatigue problem. Particle-induced contact fatigue typically starts by generating small five- to 10-micron platelets and spheres.
These platelets and spheres usually show up in advance of any significant change in vibration signature. As the fatigue problem progresses, micropitting of the bearing surfaces causes the platelets and spheres to turn to small chunks. These micropits, and the corresponding wear debris found in the oil sample, rapidly increase in size as macropits on the bearings surface cause a concurrent increase in large particles (greater than 20 to 30 microns) in the oil.
Because of the limited effectiveness of emission spectroscopy under these circumstances, many oil analysis users have turned to particle counting to fill in the missing data at larger particle size ranges. While particle counting is an excellent tool for this purpose, particle count data also holds another key - the ability to quantify particles into different size ranges. This quantification is often referred to as the particle size distribution.
Particle size distribution is as important to oil analysis as frequency distribution is to vibration analysis. In vibration analysis, the overall vibration level is easy to measure and can be instructive, but it is not sufficient in determining the true extent of a problem or its root cause. This is primarily because the amplitude of an overall measurement is dominated by low-frequency sources.
To overcome this difficulty, vibration analysts use fast Fourier transform (FFT) vibration analyzers to study vibration peaks over an appropriate range of vibration frequencies, allowing them to focus on bearing defect frequencies to seek out and identify bearing failure and other related problems.
In the same way, spectrometric oil analysis data by atomic emission (AES) is easy to obtain and measure but is not sufficient in determining the true extent of the problem. AES is sensitive up to the five to eight micron range, but is insensitive to larger particles. This is why industrial oil analysts often use particle size distribution, ferrous density and microscopic wear debris analysis.
It is fairly common to see this contamination level in new oils taken directly from bulk storage. In this case, the oil sample includes 4.2 ppm v/v of contaminants with peak size at 5 mm.
Figure 1. Typical New Oil Sample
CSI has recently developed an improved method for analyzing particle count data and particle size distributions. The CSI 5200 Trivector particle counter determines data in up to eight size ranges. These data are then used to estimate the concentration in parts-per-million and the size corresponding to the maximum number of particles in the sample. Examples of some particle size distribution plots obtained by this method are shown in Figures 1, 2 and 3.
The premise behind calculating the ppm by volume is to use the particle count data in the different size ranges to interpolate between data points. In this way, the number of particles in each one-micron window can be estimated. Based on these estimates, the total volume of particulate material can be determined, either as a total or in different size categories, by assuming the equivalent spherical diameter of the particle measured by the particle counter.
The principal value of calculating particle size distributions is to provide additional information on the types of particles and possible wear mechanism present. For example, Figure 1 shows the typical particle size distribution of new oil taken from bulk storage.
The ISO code for this sample was 21/20/15, fairly high for new oil. However, by reviewing the particle size distribution, which shows a particle size maximum around five microns, this high particle count can simply be assigned to generally unclean new oil allowing the correct maintenance action - in this case, filtering the oil prior to use.
This oil sample has 8.7 ppm v/v solids greater than 4 mm . In this instance, atomic emission spectroscopy would do a good job quantifying contaminants and wear debris. The peak concentration corresponds to particles around 4 mm.
Figure 2. In-service Oil from an Automatic Transmission
Contrast the plot in Figure 1, with the plot shown in Figure 2, which illustrates particle size distribution for fluid from an automatic transmission. The particle count distribution reflects what is expected to be normal wear from this type of equipment. The high concentration of particles in the sub-five micron range indicates normal rubbing wear, and therefore, there is no immediate need for maintenance activity.
This sample contains almost no particles in the detectable range for spectrometric analysis (less than five to eight microns). Based on SEM-EDX data, the large particles were found to be silicon (dirt) particles around 24 mm in size and large ferrous particles around 40 mm in size. The total dust and wear in this case is 116 ppm v/v. Without particle count and particle size distribution data, this problem may have been missed completely if spectrometric wear metal data was relied upon as the sole means of detecting the problem.
Figure 3. In-service Oil from a Roots Blower
The third example, illustrated in Figure 3, most clearly demonstrates the value of evaluating the particle count distribution. In this example taken from a roots blower, a high overall particle count was observed, just like in Figures 1 and 2. However, based on the particle count distribution,
in conjunction with complementary tests such as scanning electron microscopy - energy dispersive X-ray (SEM-EDX), it can be determined that this component is in need of immediate attention. Large dirt particles, and more significantly, large ferrous material, indicate a serious mechanical problem exists.
Particle counting has proven to be an effective tool in detecting active machine wear. Used in conjunction with particle size distribution and parts per million by volume, it can also be an excellent tool in differentiating between possible root causes.
The parts-per-million by volume (ppm v/v) measurements determined here should not be confused with parts-per-million by weight (ppm w/w) measurements determined from elemental analysis. In spectrometric oil analysis, data is reported in the detectable size range (less than five to 10 microns) as micro grams (µg) of material per gram (g) of oil sample.
There is a fundamental difference in reporting ppm w/w and either particle count data or the ppm v/v in different size ranges. To understand this, consider a cube of pure silicon 60 x 60 x 60 microns in size. Using the size of this particle and the density of silicon (2330 kg/m3), the mass of this cube can be calculated to be approximately 0.5 µg.
Now, let’s drop the cube into 1 ml of oil. For a typical oil with a specific gravity of 0.90, 1 ml of oil weighs 0.9 g. So what is the concentration of silicon in the oil? The answer is obvious: 0.5 µg (Si)/0.9 g (oil) or approximately 0.56 ppm w/w compared to 0.022 cm3 (Si) / 1.00 cm3 (oil) or approximately .22 ppm v/v. The relationship between w/w and v/v calculation is simply equal to the ratio of specific gravity values.
But what is the particle size distribution and ISO code for the cube of silicon in the oil? Assuming the oil is totally clean, we now have one 60 x 60 x 60 µg cube in 1 ml of oil, so the number of particles in the oil is one. Obviously, having only one particle in an oil sample is unrealistic. So let’s take the 60 x 60 x 60 cube and break it up into 1,000 equal pieces. Now we have 1,000 6 x 6 x 6 micron cubes, so what is the ppm w/w?
Obviously nothing has changed; we still have a total weight of 0.5 µg, or 0.56 ppm w/w. But what is the particle count? One-thousand particles of six microns distributed in 1 ml of oil results in a greater than six micron ISO range code of 17, significantly higher than when one 60-micron cube is dropped into the oil.
As a rule of thumb, if you take one ppm w/w of a material such as silicon and disperse it according to a normal particle size distribution, such as that shown in Figure 1, the average ISO cleanliness rating will be close to ISO 18/16/13.
This example serves to illustrate not only the difference between ISO particle count data and ppm w/w, but also why particle counting is significantly more sensitive to contamination ingress than spectrometric data, particularly when base fluid cleanliness levels are kept very low (ISO 16/14/11 and cleaner).
Note: Ray Garvey is the tribology solutions manager for Emerson Process Management CSI. His certifications include PE, CLS, and OMA1. He is an inventor named on seven patents, five of which pertain to industrial oil analysis.