It was Kenneth Calhoun who first educated me on the powerful potential predictive maintenance could have for fleet maintenance operations and vehicle uptime. This was a decade ago. Calhoun and I sat down after a long day at the American Trucking Associations’ Technology & Maintenance Council meeting. Over a couple of beverages, he painted a compelling picture of fleet operations in a near-future where every single maintenance schedule would be driven by data.
Calhoun, who today is fleet optimization manager for Altec and a past Technology & Maintenance Council chairman, assured me that there would be sensors on every component on a truck. Over time, fleets would take incoming telemetry off of trucks and build their very own database. This information would be combined with other relevant data, such as information from completed work orders logged into the system, proprietary benchmarking, and inventory flow.
The foundation of this database would be the fleet’s own, unique operating parameters: the make, model, and specs of its trucks; the applications they run; even the routes and the geographic conditions. The result would be the fleet’s very own maintenance signature: A blend of real-time vehicle monitoring with data collected on a daily basis, combined with historical maintenance records.
Using this information, every component’s useful life could be accurately predicted well in advance of actual failure, Calhoun predicted. Fleet managers would know, for example, that a certain steer tire running on certain long-haul routes would be ready to be pulled at 155,000 miles. The same would hold true for water pumps, wheel bearings, driveshafts and any other component you could name — provided it was fitted with sensors and transmitting data back to the fleet.
But the data was even more powerful than that, Calhoun told me. Imagine, for example, that a supplier delivers a bad batch of coolant reservoir tanks to your truck OEM’s assembly line during a production run of vehicles. And, he said, for whatever reason, these tanks develop hairline cracks at around 130,000 miles. In that scenario, sensors on the trucks, combined with advanced machine learning and artificial intelligence information, would recognize the issue, and automatically alert technicians working on the affected coolant systems to replace the tanks before a failure occurred.
In this future, Calhoun said, vehicle downtime would largely be regulated to freak occurrences that are just part of life on the road for trucks — a rock through a radiator, a bolt through a tire, a component with an undetected manufacturing fault. Every time a truck came in for scheduled maintenance, the system would automatically flag components that were nearing the end of their useful lives so the technicians could go ahead and replace them. Downtime would eventually transform into a relatively rare event almost always caused by external factors beyond the control of any kind of maintenance management system.
How close are we today to Calhoun’s vision of predictive maintenance?
Information Layers from Global to Granular
“Predictive maintenance empowers a fleet with greater insight into when vehicles will likely need general and unique maintenance operations performed,” explains Jake Schell, associate product manager for Mitchell 1’s Commercial Vehicle Group. “Having the knowledge predictive maintenance provides makes planning service, technician staffing and skill levels, parts inventory and cost easier to establish for the future months and even years ahead.”
While it’s your fleet and your data that informs those maintenance decisions, there are even more information layers being added as well.
“Your data, your trucks, and your applications make up your unique maintenance timetable,” he says. “But, with Mitchell 1, predictive maintenance can also be driven by global maintenance and repair information from our customers around the world. This provides a general foundation as to when specific service operations will likely need to be carried out on a year, make, and model of truck. When combined with your proprietary data, it creates a completely unique maintenance schedule based on all of that pertinent information.”
Maintenance and repair data is collected and compared over time, Schell says. In reviewing the data, it becomes possible to see when failures may occur. With the information, a fleet has the ability to monitor potential part/system failures more closely, as well as plan for repair before a breakdown.
“In addition,” he notes, “the predictive data makes it possible for fleets to see how regular maintenance schedules need to be adjusted to meet the needs for the fleet.”
Millions of Data Points
Predictive maintenance is still in the relatively early stages of industry-wide adoption, according to Jonathan Gravell, executive vice president of business development for Pressure Systems International.
PSI, which started life as an automatic tire inflation system, was acquired by Clarience Technologies last year. The combination of PSI’s TireView Live predictive maintenance system was a natural fit with Clarience’s other fleet telematics and predictive maintenance systems, such as Road Ready.
“I agree with Ken’s early assessment of this technology,” Gravell says. “He certainly nailed the possibilities. And forward-thinking fleets have been adopting predictive maintenance systems at a steady place since those early days.”
But many components on trucks still lack sensors, he says, which has held the adoption of predictive maintenance systems back somewhat. As a result, there’s still a sort of patchwork feel to the technology.
“Everybody has their own system,” he explains. “All the OEMs have their own systems, for example. And then you’ve got another specialized system from pretty much any supplier you can think of. So, while there are moves being made toward more integrated systems, that’s another process that is still in its early stages.”
Those limitations aside, Gravell says the sheer power of predictive maintenance systems is beyond doubt. PSI TireView is now collecting 7 million data points from tires in fleet operations every single day.
“I’m at a point now that I can confidently walk into any fleet in the country and tell them to test our system for six months. And at the end of that time, they’ll see a reduction in tire-related roadside service calls somewhere between 70% and 80%,” he says.
“We can look at a fleet running from, say, San Antonio to Chicago with 100 psi in their tires, as recommended,” he explains. “And we know that if that pressure is maintained for most of the life of the tire, they can expect to get 100,000 miles out of it.
“However, if they run the same route, but with tires that are underinflated by 20% most of the time, they’re only going to get 85,000 miles out of those tires before they’re used up. The information we’re gathering is so reliable that we’re within a 5,000-mile margin of error for pulling an end-of-life tire. And that’s just really remarkable, in my opinion.”
Sensors + Telematics + Algorithms and AI
Tires have proven to be a predictive maintenance application where many fleets have seen very tangible results. But we are seeing more and more sensors on trucks, and the data they provide, coming into play for predictive maintenance. Trailer telematics systems can feed data into a predictive maintenance platform from sensors on everything from the trailer brakes to the transport refrigeration unit. Even fifth wheels — Fontaine Fifth Wheel’s new SmartConnect collects and feeds back data relevant for predictive maintenance.
And there are more companies trying to offer a way to turn all the data coming from those sensors into meaningful insights fleets can use for predictive maintenance.
Pitstop predictive analytics software, for instance, takes in over 10 billion data points on battery and brake health, fuel anomalies, tire health, engine air flow and more, according to the company. It uses proprietary algorithms, machine learning, and artificial intelligence to analyze sensor data and diagnostic trouble codes and predict potential vehicle failures weeks in advance.
Uptake is another provider of predictive maintenance technology. It inked a deal earlier this year with Daimler Truck North America where, with customer consent, DTNA will facilitate streaming the necessary data to Uptake Fleet. From there, by analyzing information from subsystem sensors and work orders, Uptake can predict vehicle problems in advance of a fault code and recommend corrective actions before they lead to costly repairs or breakdowns.
Diagnostics software provider Noregon teamed up with CalAmp to offer fleets remote vehicle diagnostics and predictive maintenance capabilities. Combining real-time data insights from CalAmp fleet management software, edge computing, and cloud platform services, with predictive algorithms in Noregon's TripVision remote diagnostic software, lets fleets get preemptive alerts about vehicle issues before critical failures occur. Predictive health scores in TripVision can even estimate the mileage and time for when a fault will escalate.
Acting on the Data Stream
All the data in the world doesn’t do one bit of good unless you’re in a position to use it effectively. While artificial intelligence shows potential to help get real insights from huge amounts of data, at this point, at least, you still need people.
That means putting internal resources in play to review the information and act on it.
“My guess is about 20% of fleets today are going all-in on this technology,” Gravell says. “You’ve got another 60% that know this is coming and they understand its potential. But they don’t know where to start. You’ve got another 20% of fleets that think it’s completely overwhelming and they just not interested in trying to begin using it.”
Truck telematics have proven to be so powerful that people initially went overboard with them, Gravell says.
“It all comes down to time and resources,” he says. “It’s a two-way street. We have to provide actionable data. That’s critical. Because if a fleet manager has data coming off of several different systems, and he’s got 150 emails in his inbox every morning, he’s going to delete the dashboard and tell you to pull your sensors off his assets.
“What we want is to give that fleet manager two emails in the morning — and they’re both flat tires. If we can do that, it’s an eye-opener for them.”
To achieve that goal, Jean-Sébastien Bouchard, executive vice president of sales and a co-founder of Isaac Instruments, suggests focusing on what Isaac calls “The Five Vs of Big Data:”
Once you have those five elements in play, he says, you can begin to use that data in ways that were unimaginable just a few years ago. Bouchard boils this down to three “impact levels.”
Low-level productivity impact: Raw data collection, data-driven operational reports and standard event reporting.
Mid-level productivity impact: Cloud-sourced big data, data source integration and sharing, and descriptive analytics
High-productivity impact: Predictive analytics and prescriptive analytics
Once you have that architecture in place and functioning, Bouchard says, you can begin using the data to answer the following questions:
- What is happening?
- What actually happened?
- Why did it happen?
- What will happen?
- What do we want to happen?
At that point, he says, you have all the information necessary to begin making accurate predictions about anything from driver behavior to component failures and can put policies in place to manage them.
Another problem, as noted, is all the different system and dashboards that are in the market now. But Gravell says the marketplace is already driving technology suppliers to integrate systems across multiple platforms.
“I am really optimistic on this front,” he says, citing the example of Clarience’s Fusion dashboard, which can integrate Clarience data with tire data not only from PSI, but also from component suppliers and truck makers.
“All we need is permission, and then we simply add it to the dashboard. It’s very simple and easy to plug in the systems you want — and unplug them if you find something you like better later on.
“The one thing I also tell prospective fleet customers is that they need to put human resources in charge of all this data,” Gravell adds. “I think a lot of fleet managers think all of this information will come directly to them and they’ll be solely responsible for acting on it. And there’s far too much information coming off the trucks for that to be an effective plan.”
Gravell recommends breaking up the data by system or components and assigning each one to a fleet maintenance manager.
“You want to have a guy in charge of tires, a guy in charge of freight efficiency, someone monitoring engine systems, exhaust, trailers, whatever,” he says. “That way, no one is overwhelmed.”
As with most new technologies, predictive maintenance is slowly working its way into fleet operations. Over time, more truck components will receive sensors. And individual fleet data steams will begin to paint operational pictures that will give fleet managers more insight than ever thought possible into how their trucks are used and how to extend their lives. This process has already begun, and it’s accelerating as more components, trucks and fleets plug into the data streams coming from their fleet assets.