What is predictive maintenance?
Author: Shane Schick
Avoiding production line delays or shutdowns is critical to maintaining supply goals and optimum output. Imagine a failure to factory line equipment goes undetected and causes an outage. Unplanned downtime can happen more dramatically, of course, via a hurricane or other disaster. Regardless of the cause, predictive maintenance technology can represent a solid approach to help manufacturers avoid worst-case scenarios by taking action before they happen.
What is predictive maintenance?
Predictive maintenance is an approach that harnesses the power of technology to determine the most effective time to perform maintenance on machines. Also known as condition-based monitoring, this differs from traditional maintenance approaches that are either based on pre-determined intervals or an unexpected machine failure. Using a combination of artificial intelligence (AI), 5G edge computing and Internet of Things (IoT) in manufacturing, factories can leverage predictive capabilities to identify assets that might imminently fail.
Given the costs involved in expanding and improving their operations, manufacturers have an obvious need for their assets to be utilized as fully as possible. They may expect to replace parts or upgrade aging equipment regularly, but sudden breakdowns can have dire consequences. Such incidents could mean wasted time, lost revenue and damaged customer relationships.
The need to drive greater productivity, increase yields and gain a competitive advantage can drive manufacturers to make strategic use of their budgets. According to a recent survey from the National Association of Manufacturers (NAM), more than 65% said they would be spending on new equipment and technologies in 2023, despite economic concerns. Nearly 39% said they are also increasing investments in new and existing facilities.
Predictive vs. reactive and preventative maintenance
Before technology offered the prospect of predicting when machinery would most benefit from maintenance, manufacturers were typically limited to a few other approaches.
Preventative maintenance, for example, assumes some degree of asset failure within a manufacturing environment and attempts to plan ahead accordingly. This might involve establishing a schedule for replacing equipment or parts based on the time it was purchased, a recommendation from the equipment provider or anecdotal information.
While preventative maintenance can mitigate some business risks, it isn't based on near real-time data that can analyze performance trends within a manufacturing environment. In a sense, it's like an educated guess about when equipment failure might happen. It's not unlike risk-based maintenance, where assets are largely replaced based on historical data.
In the event of unplanned downtime, meanwhile, manufacturers usually resort to what may be described as reactive maintenance. At that point, it's up to the organization to effectively coordinate the people and replacement parts necessary to limit how long the unplanned downtime will last.
Reactive maintenance rests on two assumptions: that those with the skills and expertise to handle repairs or replacement are nearby, and that the necessary parts are readily available. This is often not the case. In fact, a survey from the Association of Equipment Manufacturers found that access to intermediate components for production is second only to workforce shortages as a cause of supply chain disruptions.
The bottom-line impact of equipment and machinery failure is considerable. A 2022 study from Siemens estimates that unplanned downtime will cost manufacturers $1.5 trillion, or 11% of their annual revenues.
Benefits to a predictive approach to maintenance
Predictive maintenance isn't just a way to avoid downtime. It also helps address a number of other common manufacturing industry priorities. These include:
When products are poorly produced, it can mean disappointed customers or even a recall. Substandard work can tarnish a manufacturer's reputation among customers and even potential new employees. This could explain why a 2022 study conducted by Microsoft found four out of five manufacturers consider overall operational effectiveness (OOE) a key performance indicator. A predictive approach to maintenance could ensure those quality standards remain high.
The most recent U.S. Census found that exposure to harmful substances or environments led to 798 worker fatalities in 2021, the highest figure since it began collecting this data in 2011. Even minor injuries suffered on the job can be serious, and manufacturers naturally want to help employees avoid them. Predictive maintenance can help contribute to safer work sites and ensure faulty equipment doesn't lead to safety risks.
Improved resource allocation
If technology-enhanced predictions can allow manufacturers to know when they need to order replacement parts or make repairs, it could mean cost savings that get funneled elsewhere. This includes areas that can lead directly to improved customer experiences and company growth. According to a 2023 manufacturing trend forecast from Euromonitor, for example, 62% of companies globally plan to increase investments in cloud computing over the next five years, while 50% plan to invest in AI, the IoT and production automation tools.
How to put predictive maintenance into practice: AI, 5G edge and IoT in manufacturing
Manufacturers can adopt a predictive approach to maintenance by deploying technologies that allow them to monitor the conditions of machinery or equipment, process and transmit data about what the technology observes, analyze the usage and health of an asset, and then alert employees and/or take automated action.
IoT in manufacturing
Companies can begin to leverage IoT in manufacturing by installing sensors directly on factory or plant assets, for example. 5G edge computing plays a critical role here because the low latency of 5G enables conditional data of an asset to be processed closer to where it is located so corrective actions can be taken more rapidly.
AI applications can pair with machine learning to identify common causes and the average timeline of when assets typically break down. The same applications can then look at weather conditions, run rates and other factors to predict the likelihood of future downtime so that production team members can be notified in advance when an asset may need to be repaired or replaced. If needed, replacement parts could be ordered in advance or appointments with third-party maintenance partners could be arranged.