Tag: predictive maintenance

OM in the News: Using AI to Keep Trucks on the Road

“In the trucking industry, few things will sour a manager’s mood like a mechanical failure disabling an 18-wheel rig in the middle of a big delivery,” writes The Wall Street Journal (March 12, 2019). But if mechanics can predict when a pump or cable or other component is about to fail, they can avoid having a truck stuck on the side of the road.

NFI Industries Inc., a $2 billion N.J.-based company, is using artificial intelligence to anticipate when the truck components in its 2,200 tractors and 9,700 trailers need adjusting or replacing. By predicting maintenance and reducing malfunctions, NFI expects to reduce truck maintenance and repair costs by 7%, or $1.5-$2 million a year.

NFI’s data is taken from truck sensors, odometers, speedometers, repair logs, temperature logs and other sources. The information collected includes truck ages, route distances, payload weights, weather conditions, driving conditions, and even the braking and accelerating styles of individual drivers. That data is analyzed by Noodle.ai, a San Francisco startup that pushes the information through a supercomputer nicknamed The Beast. Noodle.ai’s machine learning technology synthesizes the disparate bits of data to determine when a $100,000 rig needs an oil change, a filter replacement, a brake adjustment or a new set of tires.

As a result, NFI is jettisoning a sacrosanct industry ritual: regular truck maintenance and mandatory oil changes every 30,000 miles. Instead, the company is switching to less frequent tuneups, as prescribed by AI, that are based on a truck’s age, wear, driving conditions and a host of other factors. NFI’s trucks break down about twice a year, on average. The company expects predictive maintenance to reduce those mishaps to 1.5 breakdowns a year per truck. Among the surprising insights AI has produced: NFI had been procuring a truck model from a manufacturer that offered a $10,000 purchase incentive per truck. But over a lifespan of five to six years, that truck model was costing NFI about $25,000 more in maintenance and repair than other trucks.

Classroom discussion questions:

  1. What is the difference between predictive and preventive maintenance?
  2. What is the role of AI at NFI?

OM in the News: Using Machine Learning to Keep the Beer Flowing

Anheuser-Busch uses this sensor to pick up ultrasonic sounds coming off conveyor belt and motors.

The world’s largest beer maker is using low-cost sensors and machine learning to predict when motors at a Colorado brewery might malfunction, reports The Wall Street Journal (Jan. 24, 2019).  The Anheuser-Busch plant was the first among the company’s 350 beer facilities to test whether wireless sensors that can detect ultrasonic sounds—beyond the grasp of the human ear—can be analyzed to predict when machines need maintenance. “You can start hearing days in advance that something will go wrong, and you’ll know within hours when it’ll fail. It’s really, for us, very practical,” said the VP.

The installation at the brewery cost just $20,000. Since the system was deployed, it has predicted pending equipment failures and prevented unscheduled production-line halts, and more than $200,000 in product loss. (The Colorado plant employs 580 people and ships 225 truckloads of Budweiser, Bud Light and other beer brands each day).

Sensors have been used for predictive maintenance in the past, but they were unable to transmit information in real time. Advances in processing data at the edge of the network, referred to as edge computing, enable companies to collect and analyze real-time sensor data from machines. Machine learning refers to the subset of AI that allows computers to act “intelligently” without being explicitly programmed. Algorithms can increase the accuracy of predictions based on large amounts of historical and real-time sensor data.

Organizations that own wind turbines or jet engines are expected to save about $1 trillion a year as a result of predictive maintenance techniques. Sound-based predictive maintenance is becoming more important for companies, as there has been a wave of retirements among workers who were tasked with listening to machines to identify potential breakdowns. The price of internet-of-things sensors is expected to fall to 26 cents on average by 2024, from 46 cents in 2018.

Classroom discussion questions:

  1. What is predictive maintenance?
  2. How does this differ from “breakdown maintenance?”

OM in the News: A Primer on Predictive Maintenance

“Nearly everyone in manufacturing, from equipment manufacturers to processing plants, commonly face the challenge of keeping their fleet, machinery, and other assets working efficiently, while also reducing the cost of maintenance and time-sensitive repairs,” writes Industry Week (Dec. 6, 2018).  So it is crucial to identify the cause of potential faults or failures before they have an opportunity to occur. Emerging technologies such as the Industrial Internet of Things, data analytics, and cloud data storage are enabling more vehicles, industrial equipment, and assembly robots to send condition-based data to a centralized server, making fault detection easier, more practical, and more direct. By proactively identifying potential issues, companies can deploy their maintenance services more effectively and improve equipment up-time.

Using AI to identify anomalous behavior, the information derived from the equipment sensors can be turned into meaningful and actionable insights for proactive maintenance of assets, thereby preventing incidents that result in asset downtime or accidents. Known as predictive maintenance (a topic we have added to Chapter 17 in our new edition, due out Jan. 1st), this added intelligence enables organizations to forecast when or if functional equipment will fail so that its maintenance and repair can be scheduled before the failure occurs. As industrial customers become increasingly aware of the growing maintenance costs and downtime caused by the unexpected machinery failures, predictive maintenance solutions are gaining even more traction.

Predictive maintenance is also a step ahead of preventive maintenance. As maintenance work is scheduled at preset intervals, maintenance technicians are informed of the likelihood of parts and components failing during the next work cycle and can take action to minimize downtime. In addition to the advantages of controlling repair costs, avoiding warranty costs for failure recovery, reducing unplanned downtime and eliminating the causes of failure, predictive maintenance employs non-intrusive testing techniques to evaluate and compute asset performance trends.

Classroom discussion questions:

  1. How do preventive maintenance and predictive maintenance differ?
  2. What technologies are allowing predictive maintenance to spread?