What are We Wanting to Solve with Predictive Maintenance?

William A.J. Mercier, Director of Maintenance Planning & Programs, JetBlue

Over the past decade, aviation operators and type certificate holders have developed an insatiable appetite for predictive maintenance solutions. From high dollar software platforms, long term consultancies and OEM contracts, there are some robust paths forward to obtaining predictive maintenance. Many firms will be pining over the next decade to reach this goal, I would candidly ask myself (or my firm) two questions before I head down this path. 1.) What is the actual problem you are trying to solve? 2.) What is the expected outcome in solving this problem?

These questions will offer a wide range of answers. Maybe you want fewer delays and cancels, increased reliability of the platform, more predictable production throughput, or reduced maintenance costs. Predictive maintenance is an approach to advancing all of these complex problems, but it's incorrect to describe it as an approach to solving these complexities. The distinction is important—and necessary to understand.

Take a look at common threads in many predictive maintenance practitioners’ box of tools. You don’t see many Pareto analyses or MSG-3 evaluations or failure modes/effects analyses. To be fair in my argument, one could ask if these are necessary. What you will find are platforms that comb through an intense amount of data to find answers more from the elimination of complexity, rather than identifying the actual maintenance or reliability issue (or root, cause if you will).

I am far from saying that predictive maintenance is a white whale we will never catch. On the contrary, it can be achieved at many levels and it can add great value to any organization. With most operators and many other aviation firms neglecting or outsourcing their data management, a decrease in understanding of their current maintenance and reliability programs is growing. However, this means there are many steps that can be taken immediately that could help a company pursue any of the above goals stated. Taking some basic steps upfront could greatly increase the success of a predictive maintenance platform you would wish to deploy.

This can be a tough fact to face. A firm must understand the “why” for predictive maintenance to be successful. Deeper questions could then be: “Why is the data trending this way?” Or even, “Why am I doing this maintenance task at all?” Or, “what is this task supposed to accomplish?” You have an incredible collection of data on a component or system, but understanding why you are doing anything should be paramount in getting to your answer. These questions can also help improve your performance before even investing a penny into a predictive maintenance platform.

Another Achilles heel is thinking that the OEM “requires” or says “we should do it” is an understanding of the task. Understanding the actual purpose of a task or function of a component’s system will help you understand your data, which will help you understand all of your maintenance tasks on that component and its system. Conceptually, MROs and operators interface well with each other and with the OEMs. Many MROs and operators participate in annual meetings and symposiums, listen to presentations on why or why we should not escalate tasks, change tasks, add tasks, or delete tasks. All key conversations.

While important meetings to participate in, seldom do operators give that scrutiny to their own situation. How often do we interface with our own maintenance program and operational performance at this level? Many feel that managing all of that data and analyzing so many systems is too complex, and by definition, unmanageable. But if we do not understand our data and our programs now, how do we expect to understand the predictive maintenance solutions from a big AI platform?

“We must all remind ourselves periodically that while predictive maintenance is primarily data-driven, it should not a completely mathematical framework. It needs to be a data-driven business philosophy.”

Relearning some fundamentals will make an operation more efficient and more competitive before you even get to predictive maintenance. MROs and operators can sometimes manage systems in ways that cause periodic chaos (and therefore, to cause undesirable behaviors). They’ll then scratch their heads over how to manage the mess in front of them. We will zoom in on the components and view them in isolation. We look closely at the symptoms and start to hypothesize. Sometimes we see a component demonstrating a problem behavior, and we optimize our maintenance program around our hypothesis. We then walk away thinking we have solved the issue and the aircraft will fly more reliably.

Their conclusion (correctly) is that they can't manage the process as a whole, just as you can’t solve a problem you can’t see. While I cannot emphasize enough the importance of having command over your data, the same can be said for your maintenance and reliability programs. What often gets lost in all of this conversation, is the “Why.” “Why is this component driving itself off wing?” “Why can’t we figure out how to increase its system’s reliability?” We sometimes make quick assumptions around the component or system, without ever having asked if that particular system is performing the way it is intended. The good news is that it can start today. It may seem daunting or impossible with your headcount or software platforms, but the data you need and information on your program are all at your fingertips.

The most daunting task will be solving these issues with answers that are sustainable, and that will prevent the system from descending into chaos again. We must all remind ourselves periodically that while predictive maintenance is primarily data-driven, it is not a completely mathematical framework. It needs to be a data-driven business philosophy. It can be an amazingly effective approach to wrangling complex systems so that they can operate seamlessly. Those who will have the most success with it will invest in maintaining and management of their data. They will take the time to understand the picture the data paints at the component, system and program levels. They will be the ones with the near-flawless operation that everyone will be asking for the ingredients of their “secret sauce.”

Operators and MROs are intelligent organizations, and so are the Type Certificate Holders and OEMs. However, many struggle with data management, their data collection process, or both. They struggle to get good samples of data, let alone quality data. Many depend on outside organizations to help solve these issues. This struggle is real for everyone.

If operators saw the need to invest in data and program management, they would be leading the conversation of predictive maintenance, not depending on it to be solved by someone else. Even in the age of the aircraft talking to satellites and software programs, we still do not have a holistic view of how many systems are performing. Operators also need to treat maintenance as a critical contributor to operational performance, especially when trying to eliminate service disruptions. Prioritizing resources and investments are essential to make this successful, and often lead to improvements before you start down the predictive path. Once these improvements and understandings occur, operators must push to improve their data environment to develop consistently effective predictions. Once this challenge is met, operators will have to protect and standardize that data, which is where you truly see the operational and financial benefits.