Blogs

Predictive Maintenance is a Machine's Best “Friend”

Program Director for Predictive and Business Intelligence Solutions Marketing, IBM

As a proud, new father, I often find myself quite awake in the early hours of the day.

Sometimes I’m rocking the cradle, holding my girl, or just checking to see if she is breathing, and regardless, I inevitably can’t get back to sleep. 

A few days ago, as I was wide awake I turned on one of my go-to late night channels and an episode of Friends happened to be airing. It was the series finale (the one where Rachel is going to Paris for work, but Ross wants her to stay with him in New York).

During this episode, as Rachel is boarding the plane, Phoebe tells her that the plane doesn’t have a “Flangee.” Rachel doesn’t believe it, but her seatmate overhears the conversation and becomes hysterical causing the entire crew and cabin to deplane while the plane operators recheck for missing and failed parts. (Watch the scene here.)

Underlying this humor is a real problem for airlines (or any manufacturer) – how do you make sure “flangees” or other parts don’t fail? And, how do you ensure there is an early-warning system, one where you can predict when a part is going to fail?

More and more organizations are turning to Predictive Maintenance. These solutions help predict which asset or part of an asset is likely to fail or needs service, so organizations only replace parts or machinery when needed, not when they are “supposed to.”

It could be a machine on a manufacturing production floor, a pipe in a cities’ water system, a motor in a construction crane, a drill on an offshore oil drilling rig, or even a nozzle on beverage vending machines (Spoiler alert: you’ll learn more about the vending machines at the upcoming Information on Demand conference in Las Vegas in October)

A great example is DC Water, an organization that distributes drinking water and collects and treats wastewater for more than 600,000 residential, commercial and governmental customers in the District of Columbia.

They had and aging water and sewer infrastructure problem. Pipes had low reliability and lifespan and they were receiving numerous customer complaints. They utilized Predictive Maintenance to gain visibility to predict potential problems so their maintenance team could be proactive, rather than reactive. (Read the full case study.)

Predictive Maintenance is also very much aligned to Warranty Analytics. The warranty analysis process is typically the first post-sale interaction between manufacturers and customers, and can be a crucial first step in improving operational areas such as logistics, quality control and equipment maintenance.

Warranty data provides valuable insights that enable companies to identify the root cause of a claim, such as whether it is a delivery or production-related issue. Analysis of a firm’s data may reveal that a defective part is responsible for a high number of warranty claims, and that the defect is actually the result of a production quality issue.

Predictive Maintenance involves capturing all of the data that the organization has, from structured sensor and ERP data to unstructured maintenance logs on the factory floor. The embedded algorithms in the software then analyze the data, find hidden patterns, and predict when and which parts will fail.

The solution allows for integration of the predictive insights to a dashboard or other applications that enable decision-makers to act upon the information. With Predictive Maintenance, organizations will be able to reduce unplanned downtime, reduce MRO inventories, and improve productivity of maintenance resources.

Suffice to say, with Predictive Maintenance there would not have been any perceived issues with the “flangee” on Rachel’s plane, keeping the airline on schedule, and not allowing Ross to speak to Rachel with the additional time during the deplaning process.

Meaning, there may not have been a Ross and Rachel, and we would have been tortured with more episodes of our six favorite bumbling friends.

For more information: