The majority of posts you read on condition monitoring are all about detecting the onset of failure. Using condition monitioring in this way could also be described as the traditional approach to CBM.
The principles are pretty straightforward.
1st - Define the failure mode you are looking for and work out what signs it exhibits that it is starting to fail. (Be careful not to confuse cause and symptom, of course.)
2nd - Determine the time between detection and functional failure.
3rd - Work out if this will either provide a more cost-effective approach to what currently exists, (including thoughts on probability of detection) or whether this alone will reduce the likelihood of failure to a more tolerable level.
All great stuff, all things I agree with, and all things that I use regularly.
But, it is not wothout its problems and limitations ....
First, the science behind P-F intervals is still pretty nascent. There is a long way to go and any anecdotal evidence out there that you find will tell you that conservative estimates are a great start, but they also tend to either be overly conservative or not conservative enough!
Second, general thinking and approaches don't actually factor in the probability of detection issue.
Third, and most importantly, it tends to overlook some exceptionally useful applications for condition monitoring. (This is the essential point, actually.)
As well as predictors of failure, CM technologies can help us to determine the actual condition of the assets right now.
If there is a lot of vibration, that might be because of ambient vibration. How can we reduce that?
In fact, it could be due to misalignment due to plynth settling or poor installation. This is another one that can be eliminated totally with no cash and a bit of attention.
If there is a lot of heat, that might be because of ambient heat. How can we reduce that?
If current draw is regularly spiking or being driven to above-normal load areas, then we could be looking at operational or even design issues. How can we fix that?
Oil analysis can show us grit and foreign particles far earlier than it shows the results from them. How can we use this to STOP them from getting in at all?
You get the picture.
CM for failure detection is great; fantastic, actually. But it's still very much an unperfected science and there is a long way to go there.
But don't overlook CM for operating practices and asset operating conditions ... the vast majority of early life failures on simple AND complex assets is caused by relatively simple oversights, missed opportunities and often some form of human error leading to this.
