In manufacturing, keeping machines running at peak efficiency and minimizing downtime has a major impact on the company’s overall costs and output. Predictive maintenance (PdM) can help manage and optimize maintenance tasks in real-time to ensure equipment is maintained at the optimal times.
While maintenance is vital to preventing costly equipment breakdowns, running maintenance early or too often can increase costs and downtime. Predictive maintenance can help manufacturers find the right balance to extend the life of their equipment while increasing overall efficiency.
If you’re considering implementing PdM into your manufacturing, this article will provide you with an overview of the benefits of predictive maintenance in manufacturing, what is involved in implementation, and discuss recent examples and advancements to help you get started.
Predictive maintenance is a proactive approach to maintenance that aims to detect and solve issues with equipment performance before they occur. The strategy involves collecting data from sensors to constantly monitor and analyze equipment conditions and then make predictions about its future performance.
PdM was first applied in the manufacturing industry in the 1990s and has gained more traction in recent years due to the introduction of the Industrial Internet of Things (IoT), machine learning, big data, and cloud computing. These technologies provide more data points using more affordable sensors and the growth in machine learning and cloud computing means predictive models are more accessible and affordable to manufacturers.
In a recent survey of manufacturing industry experts, 80% agreed that predictive maintenance is essential for the manufacturing industry and will gain additional strength in the future.
To understand where predictive maintenance fits, here are the current most common maintenance strategies in manufacturing:
Predictive maintenance has countless benefits for manufacturing from a cost, performance, and safety perspective. Here are some of the most compelling benefits:
A predictive maintenance development project will involve four basic parts. Here is an overview of what this project involves:
One major decision to be made in any predictive maintenance development project is which approach to take to make predictions:
This PdM approach collects data through condition monitoring systems and then sends alerts when specific rules have been activated. This rule-based AI system requires cross-department collaboration between product, engineering, and customer service to understand the direct and indirect causes of equipment breakdown.
Once the cause-and-effect relationships are understood, a virtual model can be created which outlines the behaviors and interdependencies between the different IoT elements. For example, if the temperature of Machine A increases above X degrees, send an alert.
This approach delivers some automation but mainly relies on the team’s understanding of which events to monitor and the correct response to specific alerts.
Machine learning algorithms can be built which take all the data collected from the sensors and work based on a probabilistic approach. Using data generated from IIoT sensors historically and in real-time, ML models can determine equipment’s normal behavior and automatically detect anomaly data and events.
The main benefit of machine learning for predictive maintenance is that it can find correlations the maintenance teams may have missed and they can dynamically adjust to new data and make sense of what’s happening in real time.
With the growth of technology like IIoT and machine learning, predictive maintenance has seen increased adoption. Even more recent advancements and trends are causing lots of hype around PdM and creating even more compelling reasons for manufacturers to consider the approach.
Companies have been benefiting from PdM for decades, but the last few years have seen some incredible advancements and allowed manufacturers to see even greater results. Here are a few global manufacturers that have implemented predictive maintenance into their operations:
During its manager presentation at the Leading Reliability 2021 conference, Frito-Lay reported that predictive maintenance helped the company reduce planned downtime to 0.75% and unplanned downtime to 2.88%.
In one example, the PepsiCo subsidiary used PdM to help prevent the failure of a PC combustion blower mower. Had the company not received early warnings due to PdM, the failure could have caused the shutdown of the entire department. Another example involves increased acid levels in oil samples which could have led to downtime for their entire Cheetos Puffs production.
The packaging and paper manufacturer implemented predictive maintenance to avoid abnormal shutdowns of its plastic extruder machine. During the PAW Industry Virtual Conference, Mondi revealed that a single failure of this equipment could cost the company as much as €50,000 in cleanup and lost revenue. The manufacturer estimates that predictive maintenance has helped them save up to €80,000 in operating costs and waste generated by the machine.
The alumina product manufacturing facility implemented predictive maintenance in 2019 and has since saved $900,000 in bearing purchases and reduced downtime. In terms of performance, the manufacturer’s grease completion rate improved from 67% in 2019 to 92% in 2021 thanks to PdM.
The benefits of predictive maintenance for manufacturing are clear. The only question most manufacturers ask is whether PdM can be implemented into their current equipment and systems, and how long it will take to see a return on the investment. Thanks to recent advancements, these cost and technology barriers are much lower, and nearly every manufacturer can see PdM as a viable option.
If you’re looking to get started with predictive maintenance for your company, reach out to NineTwoThree. We work with manufacturers to build machine learning models and data strategies to make the most of PdM. Our team has a strong background in manufacturing and even holds an IoT sensor patent or two.