The opportunity of IoT and big data analytics for the manufacturing industry has created new opportunities managing production and the supply chain. However, in terms of a single application of big data analytics, no opportunity has seen as much traction as predictive maintenance. In a recent PWC report, Digital Factories 2020, predictive maintenance tops the list of connectivity technologies and big data analytics in terms of expected increase of use in five years.
While this chart shows responses from one region, global respondents share this view, partly because of the results that early adopters have seen, including 25% – 30% reduction in maintenance costs and 70% – 75% reduction of breakdowns. Committing to a predictive maintenance program can yield staggering results and increase competitiveness.
A Comparison: Predictive Maintenance and Preventative Maintenance
The concept behind predictive maintenance is straight-forward: “Instead of waiting for equipment to break or maintaining it on schedule, predictive maintenance occurs only when necessary based on hard data.” Maintenance schedules tend to be fixed or based on the expected lifecycle of an asset. The preventative maintenance process ensures that assets are repaired or replaced during scheduled maintenance. Of course, an expected lifecycle is only an estimate, so one of two negative results can happen: if the asset is maintained too early, then costs are incurred with unnecessary downtime or the additional costs of parts that are replaced too often; if the asset is maintained too late, then the production process can be interrupted or quality suffers.
On the other hand, predictive maintenance monitors the asset’s conditions (“condition monitoring”) to identify anomalies. If the anomalies show up within a window before the asset fails and with enough lead time for corrective action, then maintenance can be performed for that asset and situation. In this way, predictive maintenance augments existing preventative maintenance schedules to override standard maintenance procedures in high-value use cases, like line stoppages.
Condition monitoring and predictive maintenance are big data solutions. Assets collect condition data for traditional process control. With the addition of predictive maintenance, big data analytics is required for analyzing across assets on detailed data and often with more condition data based on new sensors. Developing a plan for starting or enhancing predictive maintenance requires the right internal skills or data analytics consulting. Either way, consider these best practices as you embark on a predictive maintenance program.
Collect and Perform Big Data Analytics
To identify opportunities, you need to make sense of the data. That means considering big data solutions for managing large volumes of data in a flexible way. As noted, new sensors bring new data. For example, for one company, a lift truck had 10 monitoring conditions; today that number is 72. In a much larger example, an oil rig may have 30,000 sensors. When you consider all of the data being captured, only big data solutions can enable predictive maintenance.
Yet traditional techniques only use a small percentage of available data. In the oil rig example, “only 1 percent of the data are examined. That’s because this information is used mostly to detect and control anomalies—not for optimization and prediction, which provide the greatest value.” Big data analytics is required for techniques like deep learning because a complex combination of readings most often provides the signature of an anomaly. You probably already account for the simplest issues.
Select Opportunities for Feasibility and Benefits
While big data analytics helps discover complex conditions that lead to failures, your ability to address an issue and the benefit of performing the maintenance guide how you prioritize your predictive maintenance projects.
Asset managers, of course, need to consider priorities and manage conflicts between safety, uptime, costs, productivity, and other categories. Within these categories, there may be different risk levels, like those for the assets most critical for uptime, for example.
Fortunately, big data analytics enhances your overall picture of assets operating within a holistic system. “To get the most out of a predictive maintenance system, it all needs to be tied together.” With new data, organized for analysis, the most critical assets can be confirmed. In addition, if the signal for a failure occurs without enough time to correct the issue, then big data analytics can be deployed to improve preventative schedules. Just-in-time predictive maintenance is then reserved for the anomalies that can be addressed as they occur.
Getting Help – Data Analytics Consulting
With predictive maintenance or any big data solutions for manufacturing, you may need data analytics consulting. Syntelli Solutions can also help with demand management and other applications of data analytics in the supply chain. Let us know if we can help.