As my wife will attest, technical people tend to be loners, preferring to rely on cold, hard facts and instrument readings instead of the squishy, poorly expressed opinions of the average human being.
Combine that with an innate aversion to social situations in general, and it’s amazing we talk with each other at all, much less on a regular basis.
But optimizing maintenance operations and solving complex problems requires accurate information, both current and historical, from multiple sources, and often the only way to get it is through others.
For example, at the recent Noria Lubed, Reliable and Lean Conference, among the many excellent sessions was “Multiple technologies help find imminent failure,” presented by Mark Kingscade, Allied Reliability; Matt Spurlock, Noria and Gregg Wegner, Cargill. The session focused on the incipient failure of a gearbox in a Cargill plant.
The new box was installed in December 2003, and vibration, oil analysis, ultrasonic and thermography readings were taken at various intervals thereafter. In October 2005, Kingscade noticed a slight increase in vibration amplitude of the input shaft bearing at 35 to 70 times the input speed, which suggested a failing bearing cage. The vibration levels were well below the alarm limits, and by themselves weren’t sufficient evidence for concern, but when he compared them to previous readings, they definitely indicated a recent change for the worse.
The gearbox is critical to operations, located in a hazardous area and takes 10 days downtime to remove and replace. A new gearbox is “very expensive” and has a six-month lead time.
Unwilling to recommend an unplanned outage based on vibration readings alone, Kingscade consulted with Wegner and Spurlock, who reviewed the oil analysis data for the machine. The data indicated somewhat high and varying iron content, again well below alarm limits.
In an effort to understand the variation, Spurlock and Wegner reviewed the maintenance history. The variations in iron readings coincided with oil changes and an off-line filtration treatment (higher before, lower after). Spurlock, Kingscade and Wegner agreed that the gearbox is failing, probably at the input shaft bearing. They decided to continue operations but closely monitor the box.
Subsequent oil and vibration analysis, along with more frequent oil maintenance, have allowed the plant to continue to operate with confidence that the gearbox will survive without significant collateral damage until a planned shutdown this month.
This particular gearbox failure was invisible to thermography, with any rise due to the failure apparently masked by normal variation in operating temperature. Ultrasonic readings rose 16 dB, a strong indicator of impending failure.
The session presenters made an excellent case for correlating results from multiple condition-monitoring technologies to detect, diagnose and project the remaining life of a failing critical component.
The example also shows the need for (and power of) human collaboration. Kingscade’s vibration readings alone were not enough to convince himself, much less operations management, that failure was incipient, nor could they point to the best plan for dealing with the problem. Those could happen only when Kingscade, Spurlock and Wegner pooled their information, knowledge and experience.
Finally, it’s clear that the technologies themselves would benefit from better coordination. The vibration and oil analysis readings never rose to alarm levels -- what if Kingscade had overlooked the rise in vibration? The oil analyses only indicated a problem when they were carefully reviewed in the context of the vibration readings and maintenance history.
Is it wise for plants to rely so heavily on the sensitivity and diligence of people like Kingscade, Spurlock and Wegner to spot and analyze incipient failures of critical equipment? The case history suggests it would be wise to use historian software and statistical analysis to help us monitor equipment and trigger alarms.
It also suggests the wisdom of taking frequent readings, making historical as well as current readings from multiple condition-monitoring technologies readily available, and cross-referencing those results with complete, accurate, detailed maintenance records.
Then, what if we had some software that would look at all of it and tell us what to do so we wouldn’t have to deal with people?
Until then, we better keep talking.
