How Machine Learning is Shaping Preventive Maintenance

Logistics Management reported that business maintenance and repairs were up by 42% in 2017, with 16% spending between USD 1 million (HKD 7.8 million) to USD 4.9 million (HKD 38 million). With most businesses already going digital, the percentage of companies automating their repairs is slowly rising, too—21% are already more than 50% automated.

In the U.K., businesses are already using continuous monitoring technologies like IoT, AI, and predictive analytics. They also consider predictive maintenance as the next big step for controlling maintenance costs. Information Age explained that predictive maintenance takes data from multiple valid sources. It then combines it with machine learning techniques. The information can then be used to anticipate equipment failure before it happens.

Once predictive systems point out any imminent failure, AI can help to take the next best action. This can range from automatically creating a work order to notifying the appropriate technicians or teams. This use of machine learning can help companies save a lot of money in maintenance. Trucks are one of the most fitting examples, especially in the U.K. On CDOTrends, we talked about how they are already considered computers on wheels. Telemetry systems and other technologies allow trucks to communicate data points like location, speed, engine status, and even fuel consumption.

Verizon Connect stated that the U.K. fleet companies use asset tracking that sends alerts if a vehicle needed replacement parts or to schedule a maintenance check. Meanwhile, companies in the U.S. use preventive maintenance within industrial production at factories, mines, and other utilities.

For instance, the U.S. startup company FogHorn collaborated with Energia Communication to help Japanese industrial electronics company DAIHEN find a solution that would allow DAIHEN's Osaka factory to speed up sensor data analysis from dozens of devices. The Osaka office was able to measure the condition of materials and reduce the need for manual monitoring using an RFID tracking system to track manufacturing and team efficiency. 

FogHorn also installed sensors and machine monitoring technology from Augury to monitor factory conditions and detect machine problems. Augury co-founder and chief executive officer Saar Yoskovitz told IoT World Today that through machine learning, they were able to detect severe bearing wear on some vital equipment. “The discovery enabled the brewery to address the problem during planned downtime. Because they are a 24/7 facility, they don’t have any room for unplanned downtime,” Yoskovitz said.

More countries are starting to shift from the traditional reactive maintenance approach to the predictive maintenance strategy, although slowly. Schneider Electric’s Ludovic Debuchy suggested that businesses should begin with an audit covering all of their infrastructure. The audit will help to gauge the current equipment performance. From there, organizations can proceed to install the right equipment and software that can collect data and integrate it with the relevant maintenance systems.

In Asia, AirAsia is one of the best examples of maximizing the data they gather for use not only in predictive maintenance but also operations management, ground operations, and customer service.

“There are several ways in which we gather and leverage data in our business,” said AirAsia deputy group CEO Aireen Omar. The airline collects data from more than 24,000 sensors and uses the data to prevent anything from going wrong.

Costs and efficiencies drive everything. “The quicker the aircraft gets back in the air, the more it benefits our customers and us from a cost perspective," Omar explained.

Megan Andrews authored this article. She is a freelance writer with a passion for technology and how it is changing the world. Her goal is to provide a comprehensive guide to the latest technological innovations. When she isn't writing, she can be found reading about the latest inventions.

The views and opinions expressed in this article are those of the author and do not necessarily reflect those of CDOTrends.