Trucking Along with Big Data and the IoT
Millions of trucks hit American roads every day, delivering essential goods like food, medicine, and the new iPhone 6s. But trucks also break down, which delays deliveries and hurts the profits of truck operators. Now manufacturers like International Truck are tapping into the potential of the Internet of Things (IoT) and big data analytics to minimize disruption to this essential service.
Two years ago, International Trucks–a subsidiary of Navistar (NYSE: NAV)–launched a program to take advantage of big data and IoT technologies. Called OnCommand Connection, the program is designed to help its customers and itself to make the best use possible of the reams of data flowing off today’s modern truck fleet.
“Every new truck that we’re building as of this summer will have a telematics device option on it,” says Andy Minteer, International Truck’s director of IoT analytics and machine learning. “The device itself plugs into the engine and reads data from other components in the truck system. They’re essentially all wired together.”
The telematics devices gather data on everything from engine speed and truck speed to coolant temperature and brake wear. The data is collected at 15 to 60 second intervals, and submitted wirelessly across cellular networks to (in most cases) an Amazon Web Services repository, where International Truck and its customers can access it.
International Truck then loads the truck data into its Hadoop cluster (it’s a Cloudera customer) where it’s analyzed using various machine learning algorithms written in R, Python, and SAS. The company uses the analytic programs primarily to improve its own internal testing mechanisms and, based on that improved testbed, to engage customers in predictive maintenance activities.
“There’s already a high level of testing, but this happens to be a more precise statistically-based, real-world testing,” Minteer says. “The customer is going to get a better quality truck, and not for more money.”
Since the launch of the OnCommand Connection program, International Truck customers have enrolled more than 150,000 trucks in the program, and that number is expected to exceed 200,000 by 2016, Minteer says. By that time, the CDH cluster should have more than 1 petabyte of data.
The trucking industry is more complex than the automotive industry, and International Truck alone deals with about 20,000 different possible truck configurations, depending on which combination of chassis, engine, transmission, cooling systems, and brake system the customer picks.
As each additional piece of truck data comes in through the IoT, it helps International Truck better predict how different truck configurations and part combinations impact each other in the real world.
“We’ve developed an algorithm that can do a prediction in all those 20,000 combinations to be able to detect and alert the customer when one of these fault codes indicates that we need to address something,” Minteer says. “It’s a combination of alerts. We also give them a real time view of their own fleet.”
The end result is less downtime. This system enabled one International Truck customer to proactively brings one of its long-haul trucks in to have an engine component serviced before its regularly scheduled maintenance. Having a truck in the shop costs a fleet up to $1,000 per day in lost revenue, so the potential to maximize uptime is a big incentive for trucking outfits.
Most trucking outfits are already operating their trucks in the high 90 percent range, he says, but there is room for improvement. “We should see an improvement in uptime immensely,” he says, adding that one customer in Mississippi saw a 35 percent improvement in vehicle uptime just six months after adopting the OnCommand system.
In addition to improving its own internal testing and giving customers predictive maintenance, the OnCommand system can also help customers make better decisions when ordering trucks from International.
According to Minteer, the algorithms can help customers pick the ideal configuration of truck depending on the history of other customers who had similar configurations and used the truck in a similar way.
“If we go back to the original example of cluster analysis, we’re looking for which groups of operations kind of bunch together in a similar usage case,” he says. “One of them, it turns out, is customers who drive a lot of miles, but another one is customers who spend a lot of time idling.”
Long-haul truckers in the heart of our country often keep their engines idling at night to power the heater or the air conditioner (North America may not be the biggest continent, but it sure exhibits its share of climate extremes.)
Certain truck configurations could lead to extra wear and tear on a given part, Minteer says. “That could lead us to design decisions to offer two options,” he says. “You’d make sure the customer knows whether they really need a component, like an extra heavy-duty cooling system, or that anybody operating under those conditions does have it. They’re going to have a better experience and it’s going to cost us less and we’ll have fewer issues.”
But the impacts of data analytics on IoT data don’t stop there. Minteer sees the possibility of using analytics to help boost fuel economy. Fuel costs can account for up to 40 percent of the total cost of running a truck, and there are ways to tweak the equation here and there.
One of those is changing how the engine runs. “The calibration settings themselves can have big impact on fuel economy,” Minteer says. “I can see the point where we can recommend that if they just change a setting in their calibration, they can save X percent in fuel economy, and it can all happen automatically.”
How the driver operate the vehicle is another factor Minteer is considering. “You can have instant feedback to the driver of the vehicle if they’re operating in the range that they’re expected to be,” he says. “It could be less Big Brother, and more coaching and trying to help the driver not get in trouble with their boss.”
At the end of the day, trucking play an essential role in our daily lives. But there’s room for improvement in the $225-billion industry, and those who figure out how to best leverage big data will gain an advantage.