6 concrete steps to take your predictive maintenance program from idea to launch.
We wrote this guide for anyone who’s planning to implement a predictive maintenance program. Whether you’re in maintenance, operations, innovation or management, you can use the tools here from any angle.
You already know why you want to implement a predictive maintenance program: most commonly, it’s to raise uptime, lower costs, reduce risk and extend asset lifetime. But knowing why doesn’t help you with how. That’s where this guide comes in. It’s designed to walk you through the concrete steps from idea to implementation. You’ll still have to do the heavy lifting yourself—from gathering the numbers to talking with vendors to persuading reluctant colleagues, but our hope is that the worksheets here will equip you to do all of that, quickly and well. We’re rooting for you! Before we jump into the worksheets, let’s look at some definitions and principles that will guide you throughout.
Data + analysis + forecasting = predictive
By regularly collecting machine health data, you can identify patterns that help you predict future failure, early enough to prevent it. This is what we mean by predictive. The strategy has been around for a very long time—since at least 1939, when the first research paper suggested using machine vibrations to provide advance warning of failure. Predictive maintenance (PdM) is usually contrasted with running to failure (reactive / corrective maintenance) and with time-based (preventive) maintenance, which works to a schedule. Predictive maintenance is largely synonymous with condition-based maintenance, and it can include prescriptive, where recommended actions are included with the system’s alerts.
AI + IIoT = 4.0
As technology gives us better tools, our ability to reach the ultimate target grows. Recent advances in artificial intelligence (AI) and the industrial internet of things (IIoT) now let us analyze massive amounts of data in near real-time, which make it possible to find developing faults earlier, to automate root cause analysis, and to apply the technique at scale. That’s what makes it predictive maintenance 4.0.
Full-fledged predictive maintenance 4.0 affects the whole company and its culture. That’s a big change. And like all big changes, it makes sense to start small and scale upon success. Starting small lets you work out details while they are at a manageable level, and exposes the places where your organization will need to add skills or change its culture for predictive maintenance to work. It also helps convince your company’s critics to give predictive maintenance a try, by lowering both the barrier to starting and the perceived loss of time, effort and money if it doesn’t work out. That said, don’t start too small. Some critical mass is required to see real benefits.
The only way to know if predictive maintenance 4.0 (or any other project) helps your company is to have clearly worded goals plus hard data that tells you whether or not you’ve met them. At every step of your journey to predictive maintenance 4.0, ask yourself what you want to accomplish, and how you’ll measure whether you’ve done so. Concrete objectives backed by hard evidence of results are also the single best way to turn the critics who initially resisted your pilot into advocates for scaling it up.
Do your homework.
Guides like this one can help you organize your thoughts, but no guide, however thorough, can pick the right predictive maintenance system for your specific setup. Be prepared to spend time listening to colleagues, researching technologies and vendors, and even revisiting earlier steps in the process to refine your choices as you learn more.
Make it a team effort.
The best way to kill a great project is to go it alone. No one likes to be told someone else has decided what’s best for them, without bothering to ask for input. What’s more, your company has a gold mine of valuable information stored in those biological archives called brains. Colleagues and departments who are consulted in the planning phase of a predictive maintenance project are not only more likely to support it; their input will also help you craft better goals and ways to evaluate them.
No matter how fabulous you know predictive maintenance 4.0 can be for your company, you can’t make it happen alone. In this step, you’ll list all the people who need to love the idea of this project, or at least think they might grow to love it, and what it will take to convince them. Then, as you walk through the following steps, you’ll use this information to help focus, inform and quantify your predictive maintenance pilot to give it the greatest chance of succeeding.
Tip 1: List only the must-have people, and only their #1 driver.
You’re going to show how the predictive maintenance project will tangibly address each driver you list, which can blow up fast. Most of the time, if you can show someone that your idea will help fix their biggest problem, they’ll jump on board. And once they’re enthusiastic, they’ll help convince their teammates.
Tip 2: Make those drivers specific.
Common goals like “improve efficiency” and “lower costs” may be accurate, but they’re too generic to help you here. For each person on your list, ask yourself what specific problem they most want to solve.
Tip 3: Don’t forget the human side.
It’s harder to quantify than time and money, but just as important. Is anyone afraid of being replaced by technology? Historically slow to change? Are there interpersonal or inter-team hostilities that, however unrelated, will get in the way? You won’t confront these problems head-on, but you need to know what canvas you’re working on when you paint the predictive maintenance picture for each colleague.
Recommended reading: The business case for predictive maintenance (pdf download)
We’ve listed three examples in the worksheet to help you get started.
Chances are you already had specific machines in mind when you first picked up this guide. Nothing wrong with that, but they may not be the assets that will best help you prove the worth of predictive maintenance 4.0 to the rest of the company. In this step, you’ll combine hard data on your company’s bad actors with the value drivers from step 1 to select a group of machines for your pilot.
Tip 1: Choose 15-50 assets.
You need enough machines in the pilot to see results, but not so many that it seems hard to justify, implement or evaluate. Once the value has been proven, you can scale up and branch out to include more (and more kinds of) equipment.
Tip 2: Use in-house expertise.
Your maintenance technicians and machine operators know firsthand which equipment is most prone to fail and takes the most time to repair. If your company has performed a criticality analysis before, use that to inform your choices. (If not, check this recommended article to help get you started.)
Tip 3: Require past data.
You’ll need it when it comes time to evaluate your results. Choose assets with history going back at least 6 months for metrics that can answer the value drivers from step 1. We’ve put some examples in the worksheet below to get you started.
Hot strip mill case study
A hot strip mill has 300 coilers and rollers. The mill’s computerized maintenance management system (CMMS), installed two years ago, contains full maintenance records for all 300 assets. In consultation with the maintenance and operations teams, the company picks 25 rollers for the mill’s predictive maintenance 4.0 pilot: the critical ones in the runout table after the last finishing stand.
We’ve listed three examples in the worksheet to help you get started.
As in step 2, you may already have some specific failures in mind. In this step, you’ll see how well they mesh with the value drivers from step 1, and see what other failures might be worth including, or leaving out.
Tip 1: Use data to make your decision.
In step 2 you picked critical assets for which you have at least 6 months of CMMS data. Use that to identify the failures whose elimination would have the greatest impact on the value drivers you listed in step 1.
Tip 2: Choose something for everyone.
In step 6, you’ll evaluate the project’s results against the drivers you listed in step 1. Make sure you’ll have fodder to convince everyone on that list.
Tip 3: Pick all that you need, but no more.
Predictive maintenance 4.0 depends on high-quality data from a physical signal such as vibration, current or temperature. These signals carry different information, which means there’s no one-size-fits-all condition monitoring technology. The fewer failure modes your predictive maintenance project must detect, the more chance the system you choose in step 4 will perform with flying colors.
Recommended reading: The condition monitoring comparison guide (pdf download)
We’ve listed three examples in the worksheet to help get you started.
Now that you know what machines and failures you want to monitor, it’s time to shop around for a provider. Here, you’ll compare vendors on both your high-level requirements and the nitty-gritty financial picture.
Tip 1: Your predictive maintenance is only as good as your data.
Unless you have high-quality data from existing sensors, you’ll need to pick a vendor whose offering includes hardware (rather than a platform that analyzes your existing data). Many OEMs are starting to include high-quality sensors in their equipment (such as current and voltage sensors built into VFDs), so your newer machines might already be equipped.
Tip 2: Don’t forget your existing suppliers.
If you already use an industrial digital platform like Schneider’s EcoStruxure or IBM’s Maximo, you may have access to condition monitoring technology built into that system.
Use the first worksheet (step 4a) to narrow the field.
Comparing vendors on your high-level requirements will let you quickly narrow down the list, before you spend time diving into the details of their offerings. You should be able to find most of this information on the vendor’s website.
Use the second worksheet (step 4b) to engage with your shortlist.
The second worksheet is a detailed value assessment you can use to quantify vendors in detail. The points here will serve as valuable input for your conversations with them.
Now that you’ve gotten a close look at several condition monitoring systems, it’s time to revisit the previous worksheets. Maybe there’s a failure mode you included in step 3 that your chosen system doesn’t handle well. You could select a second vendor to take up that slack, but that will make your pilot more complex and expensive. Instead, see if there’s a different failure mode you could use to replace that one, or maybe you don’t even need it, because the remaining failure modes already cover all the value drivers from step 1.
Similarly, you might have learned that your chosen system is great at something you didn’t include earlier, because you were keeping things simple. Now is the time to add in any such nice-to-haves where your chosen system shines. (Just make sure the new items are also SMART: specific, measurable, achievable, relevant and time-bound.)
It’s also possible that after learning about the systems that are out there, you’ve decided there’s a better set of assets you could choose for your pilot. The great thing about this process is that it’s not too late to go back to the drawing board. No money has been spent, but a lot has been learned, and with just a few more hours of effort, your predictive maintenance plan will be 10x better.
You’ve done all the prep work, and now you’re ready to roll! You’ll work closely with your vendor here to install and commission the new system. It may surprise you, but once your system is up and running there’s a stumbling block ahead. In short, this new initiative to help your company work smarter may seem to make things worse before it makes them better. Your task now is to take your company through this journey, when the short view says the project is making things harder, but remember that it will lead to large long-term wins.
Tip 1: Remind your stakeholders that everything’s on track.
In step 1, you listed the people whose approval was crucial in bringing predictive maintenance 4.0 to your company. Remind them that right now, you’re gathering data, and it’s too soon to judge how the new system will affect uptime, effort and cost once it’s an established part of your maintenance strategy.
Tip 2: Use the incoming data to raise confidence.
Make a point of telling your stakeholders about any immediate wins, such as when the system prevents a failure that would have gone unnoticed without it. It’s not a bad idea to send out a regular update on how the system is performing. Especially if predictive maintenance 4.0 is shifting your company’s whole approach to maintenance, use concrete results to reinforce the message that the unfamiliar changes now (which may feel like more effort, instead of less) will lead to large wins in efficiency later.
Six to twelve months have passed. Because you set up your predictive maintenance project right, you’ve now got hard, quantified evidence that it works. You’ve made it through the valley of worse-before-better, and everyone who needs to sign off in order to scale up can see in black and white that this project is helping to solve their most pressing problems.
If the worksheets in this guide have been useful to you, then use them again to scale up your predictive maintenance 4.0 efforts with new assets, failure modes and systems that best fit them.