Posted under Blog by Mike Davey - Content Marketing Specialist
October 21, 2019 (Updated February 24, 2020)
Predictive maintenance is a data-driven approach to maintaining equipment and other assets. The strategy depends on condition-based monitoring of assets. Unlike preventive maintenance, which plans and schedules maintenance activities based on usage or time statistics, predictive maintenance uses sensors to determine when certain conditions occur that indicate maintenance is needed to ensure optimal running.
Asset data lies at the heart of any predictive maintenance program, and data drives its success. All maintenance strategies share similar goals: keep assets running at capacity for the maximum amount of time.
Before we get into how predictive maintenance works, let’s look at how at how reactive and preventive maintenance strategies try to reach that goal.
Reactive maintenance seeks to increase uptime by never performing maintenance until it’s absolutely necessary, typically because the asset has broken down. This saves money in the short term, but it often carries high hidden costs as there is no way to control when the asset will stop functioning.
Preventive maintenance is a strategy that tries to ensure necessary maintenance is completed at regular intervals of either time or use.
This ensures less unplanned downtime, but at the cost of performing maintenance when it may not be needed. For example, an oil change may be “required” every 1,000 cycles, but the machine really only needs the oil changed every 1,200 cycles. This means 200 wasted cycles every time you change the oil.
Predictive maintenance aims to eliminate the disadvantages of both reactive and preventive maintenance by using , also known as condition-based (CBM). You can think of it as listening to what your assets are telling you. For a more in-depth look at CBM, check out “The Importance of Condition Based Monitoring.”
With a predictive maintenance program, since the asset is being monitored, there should be clear warnings before the equipment goes offline. Maintenance is performed when sensors indicate that it’s truly needed, rather than on a fixed schedule.
Our blog, “Reactive vs Preventive vs Predictive Maintenance,” goes into more detail on the differences between these types of maintenance strategies.
Successful predictive maintenance uses asset data to accurately forecast when equipment is mostly likely to fail. In turn, this gives you the power to prevent the failure from ever occurring by scheduling the right maintenance tasks.
Certain things must be in place for a successful using : to gather and collect , IIoT to ensure the is distributed when and where it’s needed, and finally, a method to ensure that is translated into action. All of these come together to develop a . Let’s take a closer look at each of those items.
The type of sensor used depends on the type of , thermographic imaging, acoustic , speed gauges and thermometers. can also be used to provide of fluid levels and other conditions. This is commonly referred to as CBM. being monitored and what situations indicate that a failure will likely occur. Some common examples of use
CBM can be either continuous or periodic, depending on the type of sensor and the asset being monitored.
Regardless of the type of sensor used, the goal is to monitor the equipment for certain warning signs that a breakdown is likely to occur.
For example, historical data from the equipment may show that a breakdown is likely to occur within the next 72 hours once a certain bearing reaches a critical temperature threshold. A sensor assigned to monitor this bearing triggers when the event occurs.
Sensors serve another purpose in terms of predictive analytics. By continuously monitoring the assets, they can gather the data needed to build better predictive models. When it’s done properly, this means the models used to determine maintenance needs will get better over time.
In terms of predictive maintenance, IIoT is how the data gets from the asset itself to a place where it can be analyzed. Then, on to the maintenance department so they can make decisions about what is required.
IIoT is extremely important when it comes to gathering data, analyzing it, and supplying that analysis to people or other systems.
Our example from earlier looked at an overheating bearing, one likely to cause an asset failure if not dealt with in 72 hours. The sensors detect that the heat threshold has been reached. This triggers an event, and the data is sent via IIoT to a central repository. From there, the report makes its way to the maintenance department. What happens next is largely dependent on your processes.
In cases where maintenance isn’t set up to deal with these situations, what may happen is nothing at all. The reports come in, but they’re either ignored or pushed aside for work that seems more pressing in the moment. You can have the most , on hand, a well-run warehouse, and great and IIoT in place, but none of that will help prevent if the aren’t set up to work together.
A predictive maintenance strategy will only succeed if you update your processes to leverage its advantages. For example, rather than having the sensor trigger an alert, you could alter your processes so that certain alerts automatically generate the necessary work orders and pass them along to the scheduler.
Request a Demo for more information on how we can help you to update your processes to take advantage of a predictive maintenance strategy.
A properly executed predictive maintenance strategy will increase asset utilization and uptime, while simultaneously decreasing your maintenance spend. Assets are less likely to break down and maintenance activities are only performed when they are needed.
According to a report prepared by Deloitte, “Predictive Maintenance and the Smart Factory,” organizations that moved to predictive maintenance have realized quantifiable benefits. These included a reduction in maintenance planning time between 20 and 50% and reduced overall maintenance costs of 5 to 10%.
The same internal analysis by Deloitte recognized several other benefits, such as a reduction in material costs and a decrease in inventory carrying costs. These contributed to the overall reduction in maintenance costs.
The Deloitte study also found that organizations that fully implemented predictive maintenance strategies experienced increased asset uptime and availability ranging from 10 to 20%.
An independent report by the US Department of Energy shows similar numbers, but also notes that the advantages of moving to predictive maintenance are even greater if the facility had been relying on reactive maintenance. The report states that moving to a predictive maintenance program has an average return on investment of 10 times.
The most obvious disadvantage of a predictive maintenance strategy is its high start-up cost. Sensors, IIoT systems, and hardwired or wireless data transmission may have fallen in price, but the costs can easily mount up if you’re trying to digitalize an entire factory, even a small one.
One way to hold these costs down is to target only the most critical assets. Targeting just these assets will obviously cost less to roll-out and it’s a great way to show the return-on-investment. Show just how much money can be saved in a year by preventing failure on one or two critical assets, and you’ll find that it’s easier to get the budget needed to continue.
The other possible disadvantages of predictive maintenance usually have to do with data, specifically the interpretation and analysis of that data.
Let’s take another look at our example of the overheating bearing. Remember that we have 72 hours from the time it reaches that temperature threshold to replace the bearing or risk an asset failure.
We may be looking at the bearing’s temperature not because it will break when it gets too hot, but because the excessive heat shows that it’s wearing out. The heat is a symptom, not a cause.
What happens if the location is much colder or hotter than assumed by whoever set that condition? Much hotter, and you could see a situation where the bearing just keeps sending out that “too hot” code no matter how many times maintenance replaces it. Much colder, and maybe the temperature never triggers the sensor, even when that bearing is about to fail.
As we mentioned in the beginning, data lies at the heart of predictive maintenance. Gathering and recording the data is just the start. Successful predictive maintenance relies on how the data is interpreted and analyzed.
Request a Custom Demo and let us show you how we can help you roll out a predictive maintenance strategy.