Are You Doing Too Much Predictive Maintenance?

Predictive maintenance (PdM) is more correctly called “condition-based maintenance” (CBM). It checks on the condition of an item or working fluids and verifies whether they are “good” or “bad.” If the condition is “good,” then the item continues in use to the next inspection or check. If it is “bad,” then it is replaced or restored. We call it “predictive” because we are predicting, on the basis of its existing condition, whether or not the item will last until the next inspection.

If the task frequencies are chosen correctly, you should be able to detect deteriorating condition and avoid the consequences associated with the pending failure. However, those conditions can deteriorate quite quickly relative to the useful life of the item and they arise randomly at any time. To detect deterioration, we need to consider the time it takes for deterioration to occur. Since that time is normally short, it forces us to monitor conditions quite frequently. Usually, you will find yourself doing a lot of condition checks before you ever find a deteriorating condition. It is therefore easy to get the sense that you are checking too often.

Only a small portion of your condition checks should reveal deterioration worthy of action, and when it does, you better act on it right away.

It is easy to spot when you are doing too little checking – the failures you are supposed to detect will be happening without being detected.

But how do you know if you are checking too often?

Cost and risk play a role here. If the cost of checking is greater than the costs associated with the failure, then it is not worth checking. You may also be checking too much if you’ve never experienced a failure before some threshold age, then after that the failure can appear randomly.

If you can trigger your checks to occur after that age, then you can save considerable effort and money. If your failures have low consequences, then you can probably accept allowing more of them and reduce checking frequency. But if consequences are severe, then that would be a mistake. If the failures are age or usage related, especially if the failures are tightly clustered about a particular age, then condition monitoring may not be your best option – consider PvM.

Those are just a few ideas, but bear in mind that only a thorough analysis like reliability-centred maintenance can truly help you with this one.

James Reyes-Picknell, P.Eng. – Principal Consultant – Conscious Asset