The Rise of PdM and Why It’s So Hard to Get It Right
Predictive maintenance (PdM) is a proactive approach to addressing machine failures before they become a problem. By analyzing huge volumes of machine data, PdM allows maintenance staff to understand the health of machines and optimize their activities accordingly in order to prevent unplanned downtime and eliminate sudden failures. But, despite its obvious benefits, most PdM deployments are doomed to failure. In this post, we’ll look at the rise of PdM, and why so many vendors - and their customers - continue to get it wrong.
Predictive maintenance - then and now
Predictive maintenance has been around for a lot longer than you might think. During the Second World War, scientist CH Waddington observed that a plane’s rate of failure or repair tended to be at its highest immediately after an inspection or maintenance session. Known as the “Waddington effect”, this phenomenon resulted in the adjustment of maintenance processes to correspond with a plane’s physical condition and the frequency of its use, with adjusted inspection cycles based on analysis of the resulting data. In short, it was the beginning of PdM.
A lot has changed since then, of course. The Fourth Industrial Revolution and the advent of Industry 4.0 have seen technological breakthroughs occurring at a such a rate that they’re “disrupting almost every industry in every country”. These breakthroughs have led to significant improvements in sensor, network, data acquisition, and storage technologies which, along with the access to a wealth of computing power and data made available by recent advances in AI technology, have seen PdM become increasingly applicable to wider industry.
Today, just as it was almost 80 years ago, the key benefit of PdM remains its capacity to inform decisions. Responsible for overseeing many machines across one or more sites, maintenance professionals are extremely busy people. By providing them with a better understanding of the ongoing health of their machines, a PdM solution can help them to make better use of the limited time and resources available to them.
So, given its heritage and the clear advantages it offers, why has it been so hard for so many to achieve success in PdM?
Three common mistakes
The truth is, many vendors have jumped aboard the PdM bandwagon despite having little appreciation of what is, essentially, a very unique domain. Some have simply tried to “super-charge” legacy monitoring tools, while others have applied conventional data science approaches to a problem space that is far from conventional. Without the necessary understanding of exactly what a PdM system is, and just how it works, many new solutions won’t even make it to the marketplace. As a result, few businesses will achieve any real success at scale.
Ultimately, much of this lack of understanding - and subsequent PdM success - comes down to three fundamental mistakes that vendors and their customers will often make time and time again.
- As we’ve seen, the concept of PdM is nothing new. Techniques such as condition monitoring, maintenance credits, and prognostics have been in existence for some time now. But a lack of ability to scale these techniques beyond just critical machines has meant their deployment has been largely limited to just critical machines.
- PdM is not, as some believe, a Big Data problem in which there are millions of data points and labels on which to train models. A factory environment is highly dynamic and noisy, with a range of variables including machine maintenance, different production speeds, and even behavior of different machine operators. And, of course, every machine is unique. Despite this, however, many organizations will still take a classic data science approach to PdM.
- It’s important to remember just how busy maintenance professionals are. If a PdM system’s user experience doesn’t reflect this, there’s a risk that it won’t engage with its target users. All the valuable information and insight it generates will be ignored, and an organization’s investment in the system itself will be wasted.
Senseye understands PdM
Senseye has spent more than 150 person-years of research and development time exclusively on PdM. That’s why we understand how PdM works. Importantly, we understand how PdM improves the day-to-day operation of maintenance engineers. And it’s because of that understanding that our PdM solutions are successful. It’s why we know how to deploy it at scale, it’s why we don’t just throw data scientists at it, and it’s why the user experience matters.
Our next post will explore how we’ve applied the lessons we’ve learned to everything we do. Until then, you can find more detail on where PdM came from and why vendors - and some buyers - aren’t getting it right in our white paper “Senseye in Depth: Why is Predictive Maintenance so hard?”.