Uptime may seem like an overused buzzword, but the reality is that fleets are constantly trying to improve their vehicle intelligence in order to keep their trucks out of the shop and on the road.
We’ve long heard about moving from preventive maintenance to predictive maintenance, where developing problems are identified long before they turn into on-road, expensive failures. There has been a great deal of development in this area, and a growing role for artificial intelligence and machine learning in identifying potential problems.
One company that is using AI and machine learning in an effort to help fleets is Uptake, which approximately five years ago started offering what it calls “predictive maintenance powered by artificial intelligence and machine learning software.” Its first foray was in rail, but it has since branched out into the trucking, energy, mining, manufacturing, and heavy equipment industries.
Jay Allardyce, Uptake’s chief product officer, explains that Uptake’s software uses raw data from a vehicle gained through its telematics devices, as well as “contextual data,” such as weather, traffic, and operating conditions, that tell the software more about each vehicle. “We also have a series of data science models in the software that are trained on fleet data to detect multiple specific failures, like [diesel particulate filters] to predict when they will fail.” The company says it currently has approximately 20 different data science models in its software that focus on detecting problems in a range of vehicle systems.
Time is important when it comes to identifying problems. It’s not helpful to know just a few seconds beforehand that a failure is imminent, because there is no time to take action. Allardyce says Uptake’s lead times in alerting the fleet manager of a problem range from one month to 10 days. “Having enough lead time to take an action is our goal,” he adds.
The system alerts the fleet manager about an impending problem so he or she can decide whether to pull the truck off the road immediately, direct the driver to the nearest shop, allow the driver to complete the route and deal with the problem later, or disregard the alert. The action by the fleet goes back into the AI model to help it “get smarter” and learn about the fleet preference, so the next time the problem develops, the software will know whether to send an alert.
Allardyce says what sets Uptake apart is the fact that it delivers actionable insights. “Descriptive analytics summarizes data to provide insight into past performance. It answers questions like ‘What happened/is happening?,’ which results in insights like ‘the engine is running hot,’ ‘hard braking occurred last week,’ or ‘truck X used 1,500 gallons of fuel last month.’”
Ultimately addressing problems before they result in a breakdown should not only help reduce maintenance costs, but also improve operating efficiency. With its ability to learn, artificial intelligence might just be the tool that more accurately predicts a pending problem — and does so in enough time for the fleet to take meaningful action.