Dear Digitally Yours, # 8, Predictive Maintenance?

Q: “Can you explain the different methods of Predictive Maintenance that is offered? Which one is better?”

(C.)

Thank you for the question C.

Predictive Maintenance at it’s most basic is attempting to prevent any critical or catastrophic failures by making sure you do maintenance beforehand.

Let’s say you eat lunch at your desk every workday. You are a messy eater and a lot of crumbs fall into your keyboard. In time, the crumbs will fill up your keyboard until finally most of the keys cannot be pressed.

?

Preventative maintenance would say that every day or week you would clean out all of the crumbs to avoid the keyboard not being able to work. Some times this will not be needed. For instance, maybe you were traveling that week, and no new crumbs fell into the keyboard. So WHY should you perform the preventative maintenance if it is not needed? Time and cost that are unnecessary based on the ACTUAL need.

?

Predictive maintenance would try to ‘predict’ when the amount of crumbs would get to a level that you could no longer use your keyboard and forecast the be time to clean out your keyboard WHILE IT STILL WORKS but right before it is forecast to fail. This could be sooner or later than the preventative maintenance timeframe depending on what you were eating for lunch every day. Eating string cheese might not fill up your keyboard in the same way that eating crackers would for instance.

In this way you maximize the return on investment for all of your maintenance time and money.

How to predict

?

Almost any ‘device’ you would buy today comes with a manual of some sort. The manufacturer has an inhouse set of data that says “______ is how long the average machine will run before _____ needs to be replaced.”

Taking a car for instance, in most cases, your brake pads, tires, engine oil filter, wiper blades, and other parts would NEED to be replaced long before your steering wheel.

Taking all of the cars in a particular model they have produced across the world they have a bell curve of all the owners of the car model and how and when failures happen.

People that use clean fuel, synthetic oil, run their engine until it gets up to optimal operating temperature before shutting it off, and do the regular maintenance as needed will be toward the higher end of the bell curve, for instance maybe 7,500 miles between oil changes instead of the manuals’ recommendation of every 5,000 miles.

?

People that use their car to only drive short distances so the car never gets up to operating temp, never change their oil (or check it!), who always use dirty fuel may need to replace their oil every 2,000 miles and show up on the lower end of the bell curve.

?

Because of this data in the testing phases of engines, the manual is written to the largest part of the bell curve to ensure that 95% of the customers with this engine can get the longest life which will allow for them to feel they got value for their money.

?

Improvement #1

After the manual has been written, sales ramp up and many, many cars are sold with this engine and manual. The manufacturer then gets the benefit of all the repair information from around the world where their cars are being used.

This information goes through their greatest minds to find the root cause of issues. This information is then used to refine the design of the engines and then update the manuals with new information.

These top minds then build databases of “If X happens, these are the most likely causes.”

Improvement #2

When you have a rotating smooth part that is properly secured and supported, there should be no vibration or movement perpendicular to the axis of rotation.

In the above picture, the smooth wheels should not vibrate up/down or front/backwards when turning. Whereas the bumpy parts of the cylinder will have a repeatable and predictable ‘vibration’ in the rotation.

If the smooth part starts to show a difference in the amount of vibration, this indicates a future failure is coming.

Some manufacturers use lasers to detect the smallest possible vibration in the rotating part. ?

?

Improvement #3

A newer version uses thermal imaging to detect slight variations in temperature as parts start to BEGIN failing.

The best example is bearings that try to minimize the friction of two moving parts. As the bearing needs more grease or lubrication the temperature starts to rise in the parts it is touching. So thermal imaging is very quick at identifying the potential issue BEFORE it fails and breaks things or people.

Now with thermal imaging cameras available for smartphones many people are able to start looking at where and how issues are rising.

Improvement #4

One of the newest methods replicates what those human experts used to impress us with, ‘sonic diagnostics’.

We have all heard our refrigerators, our computers, our cars, or other devices “properly run”. When something goes wrong you HEAR IT sound “wrong”.

Experts can usually tell you by the sound what is wrong with the device through their long experience.

Now with super sensitive microphones, the detection of any abnormality is available much quicker than a human ear could detect it.

?

Improvement #5

All of the above were then used against a database the experts had fed those root causes into allowing for a semi-intelligent diagnostics depending on which detection system they used. (knowledge-based, vibration, thermal, sonic, etc.)

?

Future/Current Improvement #6

Now AI is being used by multiple startups in order to detect maintenance issues on many or in some cases ANY equipment given a detection method. (And also many OEMs.)

One of the benefits of AI looking at the detected data is how quickly it can detect trends in the detection.

So which method is the best?

It depends.

If you have a factory floor with multiple machines, you could strategically place super sensitive microphones around the factory floor, but you NEED the baselines for each machine operating correctly in order to analyze new sounds against 'proper operation' sound.

If you don't have this information, then you need to select another option that will work for you.

Thank you again for the question C.

?

Don’t forget to submit any of your questions to me if you want an answer.

Stay Safe, and Stay Sane.