I’m sure you’ve heard the old joke about the night two guys were walking down the street and came upon a frantic man searching the sidewalk near the streetlight. He said he was looking for his car keys.
The two guys immediately set to help him and looked around for the keys. After a couple of minutes, one guy asked him, “Where exactly do you think you lost them?” He answered, “Over there,” pointing toward the darkness. Incredulously, the guy asked him, “Why are you looking here? “ The man answered, “The light is better here!”
Are we handling our PM tasks in the same way?
Most preventive maintenance isn’t wrong – it’s just based on what we can measure, not what’s actually failing.
Let’s look at the mechanical world all around us. Since we are maintenance and reliability professionals, let’s focus on the specific world within our machines and other assets. Elements inside the mesh, clank, roll, brake, catch, sense, restrict, squeeze, and drive other parts to do useful work for us. Even something as simple as a truck brake has a dozen or more of these interactions to operate and stop the truck.
Reliability occurs when all the elements function as intended. Lack of reliability comes in a few flavors, such as when brakes do not affect the speed, slow the truck but not quickly enough, or grab and won’t let go.
Maintenance is simply the activity we do to keep it reliable or return it to reliability. So far, so good, everyone with me?
The Hidden Gap Between What We Measure and What Fails
There are a finite number of ways an asset can fail to function. The description of the cause of asset failure is referred to as the failure mode. A failure mode is a cause of failure or one possible way a system can fail. Any reasonably complex machine has dozens of additional failure modes.
Here is where we can go off the rails.
Maintenance is what we do to prevent or detect a failure mode when it is occurring. All the activities we do to detect failure modes are approximate. In other words, since we can’t directly see the failure mode, we look for something we can see and measure that! We then, through engineering, wild assed guessing, or experience, try to correlate the thing we measure with the thing we can’t measure.
Between engineering and educated guessing lies the fragile bridge we call preventive maintenance.
To investigate, let’s take a moment to shrink down to a molecular level. We agree that excessive vibration in a bearing is a bad thing (generally). We can readily measure the frequency and intensity of the vibration. Inside the bearing (we will simplify this to make the point), bits are breaking off because the strain from the dynamic shock, which occurs when the bearing bounces up and down, exceeds the yield limit of the steel, causing a piece of the steel race to break off.
We can’t easily read the strain, which measures the cause of the failure, but we can measure the vibration. We can easily read an accurate vibration. This accuracy leads us to feel confident about an outcome, when in fact, the indication we are sensing is only an approximation of what is happening.
Why Our PM Tasks Are Based on Approximations
Almost all inspections, whether conducted by human senses, sophisticated sensors, or instruments, measure convenience rather than the cause.
When smart people create PM task lists, they mentally develop lists of possible failure modes. They know some are much more likely, some are rare, and some of each have dire (safety or environmental) consequences. Then, they match up the failure modes with inspections that will detect them early enough to intervene before they occur. The other choice is to identify a task that will postpone the failure mode (such as tightening a nut when the failure mode is a loose bolt).
Some PM tasks don’t prevent failure – they buy us time before it happens.
Since the task often involves examining something we hope is related to the failure mode, there is considerable leeway in the timing of the result. That is half of the problem.
To complicate the problem, each asset has dozens or hundreds of failure modes. Each of which is in various states of decay. The measurements we are using for each are an approximation.
When that smart person chooses an interval for PM, they may be considering the dozens of failure modes of different components of the machine. Each pattern is different. The length and speed of decay are different. Each one has different consequences of failure.
Designing PM Intervals Is as Much Art as Science
Conscientious PM designers plot the various failure modes, evaluate the consequences of failure, and assign the PM frequency accordingly. Of necessity, the interval is an average, often a weighted average or sometimes an arbitrary one. Given that you can imagine why PMs don’t always “work” in our favor. You may also gain insight into why there is so much art in the development of useful PMs.
