By David Hicks, Jason Richardson, John Alford, Ralph Canale
Predictive analytics for maintenance applications has been described as one of the first clear killer applications for the Internet of things (IIoT). McKinsey & Company estimates that predictive analytics applied to maintenance applications could save U.S. manufacturers more than $630 billion per year by 2025. By leveraging industrial Internet of things (IIoT) technology, they say that better predictive maintenance could reduce factory equipment costs 10-40% per year, cut downtime by 50% and, by extending machine life, shrink capital equipment investment requirements 3-5%.
Such forecasts are a big reason why predictive analytics for maintenance applications (see definition “What is Predictive Analytics for Maintenance Applications” further down) is a hot topic among manufacturing executives. It’s also why there are a whole host of solutions providers ready to sell you such solutions.
It is our opinion, when it comes to using predictive analytics for maintenance applications, that many manufacturers are chasing the latest silver bullet not unlike the lights-out factory promises of years past. As always, we advise clients not to rush into something your team does not fully understand and is not fully prepared to make work.
As we’ve written previously, there is a maintenance maturity curve. At the bottom end is a wait-until-it-breaks, 100% reactive approach to maintenance. Waiting for breakdowns to happen can be expensive both in terms of labor and unexpected downtime that leads to lost production, missed shipments, and dissatisfied customers. Achieving a higher level of maintenance excellence requires the right capabilities, leadership and process discipline.
Bain & Company describes this maintenance maturity path as a pyramid. At the bottom is the break-fix, run-to-failure mindset. Above that are preventive tactics following time-based, usage-based and condition-based approaches to maintenance scheduling. Moving up the maintenance excellence curve to more predictive work, according to Bain, can reduce breakdowns 70-75%, cut downtime by 35-45%, and shrink maintenance costs 25-30%.
To achieve those benefits, we argue, you have to start with the fundamentals. Predictive maintenance practices have to be applied on top of effective preventive measures. Installing, monitoring and collecting data from thousands of sensors—and setting up alerts when readings go out of spec—won’t fix any of your maintenance issues until your preventive maintenance program is solid and your team is truly prepared to apply more advanced technology.
As Bain points out: “Prediction and prescription are the buzzwords in industrial analytics. But what if your infrastructure is not ready for complex solutions that aim to predict problems before they happen? What if you’ll get satisfactory payback from a simpler improvement, such as a shift from reactive maintenance to real-time monitoring of assets and processes?”
An effective manufacturing preventive maintenance program starts with fundamentals. When setting maintenance priorities, for example, a simple analysis of breakdown history and the machine performance data that is already available without adding any new sensors can direct maintenance attention to where it’s most needed.
If you’ve adopted a computerised maintenance management system (CMMS), make sure it is being utilised. Too many of these systems are not used.
Maintenance people follow human nature. They often focus on the PMs that are easy to do and put off the more difficult and more time-consuming jobs. It’s critical that the more-involved PMs are completed correctly and on schedule as well.
Traditional predictive maintenance tools have long been used to monitor machine condition and optimize preventive maintenance schedules.
Gathering such data and samples helps optimise preventive maintenance by determining whether maintenance needs to be done quarterly, or if every six months would suffice.
Planned and unplanned downtime can cost a single manufacturing operation somewhere between 5-20% of productive capacity every year. This more than justifies the need for manufacturing executives to pay more attention to maintenance and asset condition. This is especially true if these areas have been neglected for some time—which we’re finding with many of our clients—or if your unplanned downtime is rising or at the high end of that range.
Predictive analytics is a powerful tool that has been successfully used for years, mostly to prevent failures in hyper-critical operations. For most manufacturing applications however, predictive analytics should be used as an additional layer of insurance—when it can be cost justified—for preventing breakdowns on top of a well-planned and executed maintenance program.
As noted above, predictive maintenance is not new. Various monitoring devices—handheld infrared spectrometers, vibration meters, oil analyzers, and so on—have long been used to monitor machine condition to head off future breakdowns.
Using data and analytics for predictive maintenance applications is not new either. The technology is widely used in applications where any unplanned downtime is extremely disruptive, dangerous or costly. In transportation this includes airliners, ships and trains. Manufacturing applications include oil rigs and refineries, mines, power stations and auto manufacturing, where an unexpected line stoppage can cost hundreds of thousands of dollars.
What’s new is the decreasing costs of sensors, monitoring and cloud-based data storage. So-called industrial internet of things (IIoT) technology is opening up more opportunities for applying predictive maintenance in manufacturing. Many such operations are already collecting much of the relevant machine and production data, but it’s not being used.
Predictive maintenance promises to optimise the productivity tradeoff between lost production time for maintenance and the risk of breakdowns. It starts with sensors that monitor and record machine activity and conditions (such as temperature, pressure, vibration and energy usage) of key components. Points of potential failure include hydraulic systems, pumps, valves, and motors. Advanced analytical algorithms then compare real-time performance to historical or test data, sounding an alert when significant deviations or negative trends signal that a failure is imminent.
Ultimately, predictive analytics can reduce manufacturing maintenance resources, time and costs by scheduling maintenance work only when it’s needed. This compares to a time-based preventive maintenance schedule or a reactive approach that waits for a machine to breakdown.
Key Takeaways
Don’t miss industry expert insights.
Join a community committed to excellence.
