How to launch a successful predictive maintenance program
An approach that keeps people at the center of the digitalization project has a better chance of success
- Define key terms such as “digital” and establish a core purpose to focus and motivate a data-driven organization.
- The status quo needs to be assessed and an organizational approach to improvement needs to be in place before plotting a pilot program.
- Historical data and experienced staff need to be used to choose the right critical assets for a pilot program.
- The pilot program needs to be small enough in scale that initial failure won’t prevent starting again.
- The successful pilot program can become the basis for scaling throughout a facility or even an entire enterprise.
- Think big, start small and scale fast toward a zero-downtime, zero-waste manufacturing mindset.
A predictive maintenance program (PdM) anticipates the future state of physical assets and makes timely and better-informed maintenance decisions. PdM — as well as the idea of Industry 4.0 — depends on achieving a convergence of information technologies (IT) and operational technologies (OT). We believe bringing together people, process and technology is the key to a successful predictive maintenance program. If you look at Industry 4.0 tools in general and at predictive maintenance specifically, one of the key metrics it can impact is the overall equipment effectiveness (OEE) of an operation, which includes uptime, performance speed, and quality, as well as labor productivity.
Digitalizing and implementing such tools has immediate impact on providing real-time data analytics and prediction to the maintenance and quality function, as well as significantly improving labor productivity. It shifts the mindset from a reactive, potentially manually driven operation to a more predictive, prescriptive and proactive operation (see Figure 1).
But it’s not just about improving uptime. It’s also about linking to other systems within the operation, especially maintenance systems, spare parts systems, maintenance scheduling and labor allocation, so management can achieve predictive insights. Spare parts on hand can be reduced, which also reduces overhead for the business. More proactive scheduling and even predictive spare parts replenishment is possible. The vision is driving toward near-zero downtime and near-zero waste within the manufacturing space and directly impacting the business in a positive and profitable way.
So, where to start? How does one set up a successful predictive maintenance program that can lead to a direct impact on the business bottom line?
Define “digital,” establish a core purpose
A PdM or any other digital program can’t be launched in an organization without knowing exactly what digital means. A critical first step is attaining a common definition of “digital” for the organization.
One favorite for the authors is: “Digitalization is a nearly instant, free and flawless ability to connect people, devices and physical objects anywhere.” Admittedly, it’s a broad, simple definition, but it’s all about connectivity and sharing information. It’s at the core of data management and helps create program identity.
With a common definition in place, the next task is to define the core purpose of the initiative: establishing the “why.”
We’ve all been around initiatives where the “why” wasn’t clearly established, and they all fail. It’s important to make it clear what the purpose is — and what it isn’t. The goal is not implementing new technology because there is new technology to implement. Rather, it’s solving real business issues. It’s creating new core competencies and integrating business functions that will improve the bottom line.
We’re applying technology and newer processes into mainstream manufacturing where they hadn’t been before. At the end of the day, this is a different way of thinking. If the bottom-line goals of the program are not constantly kept in view of all its participants, the long-term success of the program will remain elusive.
Assessing before executing
Consider these first steps for digital transformation of an enterprise. With that initial framework in mind, Let’s look at using PdM as a specific transformation worth pursuing.
After establishing the core purpose for stakeholders, identify targeted outcomes and digital focus areas that suit that purpose. This calls for assessment of the status quo. It’s done with a team that includes the plant manager (or an appointed deputy) and managers of various departments including production, maintenance, IT and others — all of whom know their systems inside and out.
The team’s initial task is to document the processes and systems currently in use to assess what works well and where there are gaps in knowledge and capabilities. This includes a look at infrastructure and identifies critical work centers and data collection points.
The assessment considers and benchmarks a facility’s digital transformation readiness, including a gap analysis, which not only documents the status quo of processes and technology but specifically delineates how near or far that status quo is from digital readiness.
Out of that assessment will come a pilot project, a production test bed that will deploy technology and practices needed to fill in some of those assessment-identified gaps in a demonstrable way. During the pilot project, the evaluation phase begins. This step establishes a center of excellence (CoE) for studying and training stakeholders in the project’s processes and technologies.
After all of this, you can begin to execute a transformation — the actual final step is scaling the digital innovations throughout the rest of the plant and eventually the rest of the enterprise in a phased rollout supported by the CoE (see Figure 2).
Our advice for any initial transformation is always “Think big, start small and scale fast.” You want to think about having a predictive maintenance capability everywhere in the organization and having a world-class OEE, a huge improvement for the global company — that’s thinking big. But you must start small, with a pilot project small enough that if it fails for some reason the first time out, the loss of time and resources can be absorbed. The team can absorb what’s been learned and start again. Then, when success does come, it can be scaled from a single asset to a line, to a plant, to a second plant and then quickly across the enterprise.
Pick the right asset for a pilot predictive maintenance program
Proving the capability of a PdM program depends on a successful pilot project, so it is imperative to use data to select the right assets for the project. Analysis can be performed at varying levels:
- Downtime history of production equipment
- Quality data on critical components
- Frequency and types of failure modes.
However, modern or primitive the method has been for tracking such data — computerized maintenance management system (CMMS) or enterprise asset management (EAM) software or operator notes entered into a logbook — the historical data is pertinent to the evaluation. But just as pertinent is the experience of maintenance and operations personnel who need to be a part of the discussion.
A data-driven four-quadrant approach can lead the way to choosing the right asset for a PdM pilot. In this method, past downtime events are charted on a graph with event frequency on one axis and length of downtime on the other, with the chart then divided into quarters (see Figure 3). The first quadrant is for failures that are more frequent, with longer downtimes (and therefore highly impactful and costly); these may be considered candidates for redesign. In the second quadrant falls high-frequency failures with minimal downtime — a problem possibly best addressed by keeping an inventory of spare parts on hand. The third quadrant is for failures of short duration and of minimal frequency, probably best dealt with through regular maintenance. But the fourth quadrant, charting less-frequent failures of longer, more costly downtime, is where to find a good candidate for a PdM pilot program.
Collect, visualize and analyze the data
Once the best asset for a pilot program has been chosen, it’s time to launch the program. Make the asset PdM-ready by installing any needed sensors and ensuring appropriate software applications are in place for collecting, visualizing and analyzing the asset’s production data. For the general methodology for PdM execution, refer to Figure 4. The steps include data collection, signal processing, feature extraction, fault diagnosis, trending/prediction and health assessment.
There is more to be said about each of these steps, and about the challenges of scaling from the pilot to the enterprise, that will be addressed in future articles appearing in the pages of Plant Engineering. In the meantime, a case study of a successful PdM project that was featured in an earlier issue of this publication may be found here.
Final thoughts on a predictive maintenance program
We leave you with two last points about making this first move into predictive maintenance, and perhaps a first move into Industry 4.0 technologies in general.
First, businesses don’t do digital because it’s cool. They do it because it solves business problems and help achieve targeted business outcomes, eventually at a massive scale.
Second, it’s not only data, but people that make this work. Never embrace digital technology to the extent that everything pertaining to the legacy standpoint is eliminated. The organization likely has experts who know a machine or know a process from front to back and have been using it for many years. The digital data now collected and used is of paramount importance, but the insights gained from experts — on things for which there is no data or were never thought to measured — can include priceless information that helps validate the results. These people are valuable assets and an important part of any digital strategy.