For each of your assets and equipment, there is a small, sweet spot where you’re doing enough maintenance to avoid unscheduled downtime but not so much that you run the risk of over-maintenance.
Predictive maintenance can help you find that spot and then stay on target.
The best place to start is with some basic definitions, and then work up from there.
What is predictive maintenance?
One way to think of this maintenance strategy is as the result of a combination of developments in both hardware and software. What makes it possible is cheaper, more reliable sensors and faster, more accurate algorithms.
On the most basic level, predictive maintenance is where you add sensors to your assets and equipment for a constant stream of real-time data that you then combine with a lot of historical data on maintenance, repairs, and associated MRO inventory. Once you have all your data together, you feed it into specialized software that tells you which maintenance tasks you need to do and when you need to do them.
What are the differences between other maintenance strategies and predictive maintenance?
An alternative title for this section could be, But isn’t all maintenance sort of “predictive” anyway? And the answer is yes, sort of.
To see understand why, it helps to think about the weather.
Predictive vs preventive maintenance
With preventive maintenance, you’re predicting the future, but you’re doing it mostly using reliable averages. So, you know that one of your pumps runs about 10,000 cycles before you need to inspect it for cracks in the seals and replace the lubricant. You’re never going to be one hundred percent sure when it might stop working without maintenance, but you have a rough idea, so you can schedule your PMs for just before then, adding a bit of a margin for safety.
It would be roughly the same as predicting the weather in Alaska in the middle of January. You don’t know exactly how cold it’s going to be, but even without looking out the window, you still put on a jacket and boots. It’s winter in Alaska. On average, it’s really cold.
Predictive vs condition-based maintenance
With condition-based maintenance, you’re still predicting the future, but now you’re working with both historical and real-time data. If you have a sensor on that pump so you know exactly how much it’s vibrating, you can use that to decide if it needs maintenance or not. In the past, it’s always been fine at frequency X. But you also know that it went up to frequency Y before the last time it failed. Putting what you know from the past with what you know right now, you can predict if the pump is fine or sliding toward a failure.
Jumping back to the weather analogy, you’re now looking out the window to get a better understanding of current conditions. Clear sky? Are those clouds in the distance? Looks overcast, and the last time it did, it rained. It’s not raining right now, but you know to take an umbrella.
With predictive maintenance, you’re looking at the future the same way experts forecast the weather. They’re not just thinking about the average temperatures in any given season. And they’re not just looking out the window.
Instead, they’re feeding a ton of historical information and current readings into complex modelling software that produces accurate predictions. And with the right combination of temperature readings and satellite images, it’s like having a crystal ball: seven-day forecasts are on average 80% accurate. For five-day forecasts, it’s the accuracy increases to 90%.
What are the advantages of predictive maintenance?
Another way of asking this question is “What would be the advantages of having superpowers and being able to see the future?”
It’s a long list of obvious advantages, but there’s also a lot of overlap with all the other maintenance strategies. Just like all the others, predictive maintenance helps you avoid costly, frustrating unscheduled downtime by helping you first find and fix small issues before they have a chance to develop into big problems. It’s a way of keeping all your assets and equipment as far up the P-F Curve as possible, where you have a lot of time to plan and schedule everything.
And with all that extra time, you can make sure you have the right MRO inventory and teams in place to do everything properly. You can also schedule the work for when the asset or equipment is already set to be offline.
But unlike the other maintenance strategies, with predictive, you can worry less about over-maintenance. And yes, you do have to worry about overdoing maintenance.
The first problem with over-maintenance is that it’s an overall waste of your time and your techs’ talent. The maintenance team is doing work it doesn’t need to, wasting their time and your inventory. If the team changes out a fan belt too early, you lose all the value that was still in the belt.
But you’re doing more than just throwing out value. You’re also brining in risk. Every time the team does maintenance work, you run the risk of them doing it wrong. A lot of the time, the risks are small, but they are always there.
Back to the fan belt example. How much of the asset or equipment are the techs leaving exposed when they change the belt? What if they drop a tool, damaging a different part? Or, what if they don’t properly close everything back up, leaving the guts vulnerable to dust and debris?
Predictive maintenance helps you schedule just the right amount of maintenance, so you catch problems early while avoiding lost value and added risk.
What are the disadvantages of predictive maintenance?
The big disadvantage is that it costs a lot more than other maintenance strategies, both to get everything set up and then to run it.
You need to invest in sensors, software, and sophistication. You need the sensors to collect the data, the software to crunch the numbers, and the people with the right training to understand and leverage the reports.
But those costs make a lot of sense when you see them in terms of return on investment. In the short term, the all-time cheapest maintenance strategy to set up and run is just doing nothing. But in the long term, it’s also the most expensive.
Predictive maintenance can be the opposite. It costs more to set up and run, but if you’re using it on the right asset and equipment, in the long term, you save time, money, and frustration.
The idea of having to match it to the right assets and equipment is a good segue to the next disadvantage, which is that it only makes sense for a specific subset of your assets. In fact, because of the associated costs, you can only use it on assets that are both critical and have failure modes that you can cost-effectively predict. For everything else, it doesn’t make sense.
What is the future of predictive maintenance?
All the things that go into predictive maintenance are set to become more widespread and cheaper as part of the general trends subject to Moore’s Law, which states that the number of transistors on a microchip double about every two years while the cost of computer’s drops by half.
About every two years, you get roughly twice the processing power at half the cost.
But that doesn’t mean predictive maintenance is set to eventually make sense for everyone and every asset. Some assets and equipment are always going to be cheaper to maintain with reactive, on-demand or preventive maintenance. It’s the same as the future of flying cars. No matter how cheap or easy they eventually become, for some locations, it’s always easier to just walk.
If you’re looking to move from paper- or spreadsheet-based work order management you’re next best step is implementing a reliable, cloud-based CMMS solution.
But if predictive maintenance is a better possible fit, there’s Archibus by iOFFICE + SpaceIQ.
Predictive maintenance helps the maintenance team do enough work on assets and equipment to maximums uptime without driving up the costs and risks connected to over-maintenance. The maintenance strategy depends on a combination of historical data, real-time data streams, and sophisticated software to accurately predict failures well in advance, giving the team the time it needs to schedule resources and bring in the necessary parts and materials. Although it shares many of the standard advantages of the other popular strategies, predictive maintenance helps leads and techs do only the work that’s necessary, avoiding both the waste and risk of accidental damage that can occur during regular maintenance inspections and tasks. The main challenge is matching the strategy to the right assets. Because of the connected costs, both start-up and ongoing, you can only see an ROI when applying the strategy to assets that are both critical and have failure modes you can cost-effectively predict. That means that even though Moore’s Law predicts prices for sensors and software should continue to drop, predictive maintenance only ever makes sense for a limited subset of assets.