For industries like power generation and petro-chemical, vibration analysis has historically been the technique of choice for monitoring the condition of large, critical pieces of rotating equipment.
Conversely, the fleet industries have relied upon oil analysis to make effective maintenance decisions. It is common for industries such as primary metals, pulp & paper, etc., to use both techniques.
In general, vibration analysis and oil analysis are the most effective techniques for monitoring the health of machinery. The two techniques are natural allies due to the complementary nature of their respective strengths. Unfortunately, the two techniques are rarely combined to form an effective union.
Vibration analysis activities typically reside in the condition monitoring or vibration monitoring group, while oil analysis usually resides with the lubrication team.
Making matters worse, the oil analysis program usually consists of submitting occasional samples to a laboratory in exchange for results that frequently look more like chemistry than machine condition monitoring. And, too often, oil analysis is used to schedule oil changes while equipment condition assessments are left primarily to vibration analysis.
This is changing in many organizations. For example, the Palo Verde Nuclear Generating Station in Arizona (see article this issue) made a dramatic change in their approach to condition monitoring.
They combined vibration analysis and oil analysis into a common group, brought their oil analysis on-site and began working as a team. Their results have been remarkable. In an assessment of bearing defects detected by technology, they found that oil analysis was responsible for 40% of the defects found, vibration analysis was responsible for 33%, and both techniques converged on the remaining 27% of the defects found.
The loss of either technology would have reduced their detection resolution and their ability to control the root causes of machine failure.
In research conducted at Monash University, Melborne, Australia, the correlation between oil analysis and vibration analysis was found to be generally very good. However, there are instances when one technique indicates a fault while the other shows no change or even a contradictory result.
For example, in applications where sliding wear is prevalent, one might detect increasing rates of wear generation and decreasing rates of vibration. This is caused by what the researchers termed a “lapping” effect.
Essentially, the sliding wear slowly hones the surfaces smooth, reducing the overall vibrations until the point at which looseness and mechanical vibration are induced. The effect is intensified by the presence of abrasive dirt.
Conversely, the Australian researchers found that vibration analysis very effectively identifies the presence of a fractured gear tooth, but because the size of the debris generated is so large, wear particle analysis is ineffective.
The debris falls to the bottom of the sump, never finding its way into a sample bottle until it is oxidized and leeches into the oil, a process that could take months. The Australian researchers concluded that both techniques are required to effectively monitor and diagnose the condition of plant machinery because each technique evaluates different and complimentary symptoms.
An example in which both techniques are required to effectively solve a problem is the case of a gearbox with increasing vibration at the gear mesh frequency. Inspection of the particle count and ferrous percentage revealed an increase in both categories, improving confidence that a problem existed. It was not until the oil’s viscosity trend was assessed, however, that the true nature of the problem was revealed.
A drop in viscosity from 220 cSt at 40°C to 70 cSt at 40°C was observed. A review of the work history showed that the oil was changed two weeks earlier. In all likelihood, the oil change was performed using the wrong oil leading to the wear and vibration. Without the combination of condition monitoring technologies, the root of the problem may have gone undetected.
In general, we can make the following conclusions about combining oil analysis and vibration analysis in detecting and analyzing machine faults:
1. Both techniques are required to control the root causes of machine failure.
2. Often, one technique serves as the leading indicator of machine failure while the other serves as the confirming indicator.
3. Oil analysis is generally stronger in detecting failures in gearboxes, hydraulic systems and reciprocating equipment.
4. Vibration analysis is generally stronger in detecting failures in high-speed journal bearing systems.
5. Vibration analysis is often better at localizing the point of failure depending on the application.
6. Oil analysis is often stronger in determining which wear mechanism is inducing failure.
7. Both techniques are required to effectively determine the root cause of failure.
8. Correlation between oil analysis and vibration analysis is very good, but there are contrary instances.
In conclusion, oil analysis and vibration analysis are natural allies in achieving machine reliability. They offer complementary strengths in controlling the root causes of machine failure and in identifying and understanding the nature of abnormal conditions.
Success depends on making changes in the organization to foster the development of condition monitoring and machine diagnostic generalists in lieu of technology specialists. A carpenter goes to the site with all the tools necessary to complete the job.
While it may be possible to cut a board with the claw of a hammer, the carpenter is more likely to draw his saw, a more effective tool for the task. We in condition monitoring must view technologies as enabling tools. We need the right tools in our bag to complete the job of ensuring machine reliability.
Ref:
Johnson, Maxwell, Arizona Public Service.
Troyer, Drew, Enteract Conference 1998
Mathew, J., Stecki, J.S., Comparison of Vibration and Direct Reading Ferrographic Techniques in Application to High-Speed Gears Operating Under Steady and Varying Load Conditions, 1986