Data analysis is critical to understanding problems and decision-making. Analysis that is not followed up by an action is likely a wasted opportunity, but not always.
An upstream oil company had critical water injection facilities. Water is lifted by injecting gas at high pressure into specially designed wells. The gas is compressed by gas turbines driving centrifugal compressors through a reduction gearbox.
Multiple sensors are arrayed on the gas turbines and compressors, but only bearing temperatures are monitored on the gearbox.
Vibration, as measured by a vibration monitoring system on the gas turbine and compressor, increased suddenly one evening. The increase wasn’t large enough to trip the system. It steadily remained abnormal, but not increasing.
Further analysis of the waveform and confirmation from handheld data collection instruments pointed to the main gear of the gearbox. The frequency was shaft rotation speed (1X) for the gear. The tooth speed frequency hadn’t changed noticeably. The first conclusion was that the large gear had lost part of a tooth. Gearbox bearing temperatures were normal.
A lube oil sample from the gearbox sump showed the oil was still in good condition and that wear particles were in the normal range.
The question was how long the gearbox would last. Replacing the gearbox would take several days and the unplanned lost water injection wasn’t acceptable.
Taking the gas turbine down and placing it on a barring gear to avoid shaft bowing would require more than an hour to observe all of the gear through the inspection port.
A mechanic suggested using a flexible rod with a strong magnet on the end to “go fishing” in the sump. This was deemed unsafe as the gear would be turning an increment at time, so it was still dangerous and the magnet would be attracted to the gear and the housing.
The Do decision was to do nothing until the gas turbine compressor train came down for a full inspection in two years, if the vibration didn’t increase.
Two years later, the gearbox cover was removed and a chunk of gear tooth two inches long and half an inch wide was found in the sump. Later analysis would indicate a manufacturing defect caused the failure. It was a relatively clean break that caused little debris and no secondary damage.
The gearbox was changed out, but it’s likely the old box could have run forever with that damaged tooth.
The decision to do nothing, based on a thorough analysis, was the right one.