Predictive Monitoring Can Save Money in Many Ways

by Andy Oram
October 6, 2015

This article originally appeared on the International Manufacturing Technology Show site.

In a recent television ad, a man came across a piece of paper that said, “You will have a heart attack today.” The ad was mocking the idea that we could predict our future. Yet an increasing number of sensors do give us the data to figure out when something will go wrong—in our bodies, our manufactured products, and our infrastructure itself.

At the most recent Solid Conference in San Francisco, I talked to General Electric staff about the massive monitoring project they undertake on windmills, now one of their major product lines. A typical windmill is outfitted with two to four thousand sensors, each reporting one tiny statistic that can indicate whether degradation is taking place. Analytics then make sense out of these data points and inform owners whether maintenance needs to take place.

This is cutting-edge technology, but predictive maintenance has been available for cars in years. Haven’t you ever seen one of the yellow warning lights come on in your car? I’ve found, unfortunately, that the problem usually boils down to a problem with the sensor, not a problem with the car. But such warnings save people from disastrous break-downs, I’m sure.

Preventative maintenance requires a major research effort. Researchers have to figure out:

• The problems that cause products to break down. What joint is most likely to break? What surface is most likely to disintegrate after years of abuse from the weather?
• The signs of upcoming failure you can detect. Is it a softening somewhere? An increase in the angle of a beam? A loose fastener? (After the Big Dig was finished in Boston, improper glue in ceiling anchors led to the fall of a concrete block that killed a passenger in the tunnel.)
• The sensors to use and where to place them. The data must also be collected by a network and shipped back to the manufacturer.
• Algorithms that determine how close you are to needing maintenance.

Everything has to be tied together into an effective alerting system. And you have to test the system in many different environments, because the type of use determines likely patterns of degradation. A windmill in the desert suffers different kinds of failure from a windmill on a high plain.

But according to the report Predictive Maintenance: A World of Zero Unplanned Downtime, that’s only the first of three layers of business activity you need in order to take advantage of predictive maintenance.

The second layer involves your organization. Where can you get the sensors and materials you need to install a system? How is the system built into each product? Can you be sure someone will take the necessary action once a problem is identified?

The third layer involves making the system work financially. You wouldn’t want to spend more on maintenance than you’d have to pay on failures. Of course, some failures are so catastrophic that almost anything you spend on maintenance is worthwhile. And as the subtitle of Predictive Maintenance indicates, the cost of a failure may not be the repair so much as the time you go without use of the product.

Furthermore, predictive maintenance can lead to other efficiencies because you gain vast amounts of new data about how your products or infrastructure are running. Some systems accept feedback from a human so that, through machine learning, the system can improve over time.

We have a well-known infrastructure crisis in the US. A well-considered investment in predictive maintenance can reduce the staggering costs of keeping our roads, bridges, trains, and electrical grid in good shape. It’s just as useful on a smaller scale as well.

This work is licensed under a Creative Commons Attribution 4.0 International License.

Author’s home page
Other articles in chronological order
Index to other articles