Predictive diagnostics is here

Whether you run one truck or hundreds, one of the worst calamities you can face is a breakdown on the road. Even worse if you’re far from home and not within easy distance of a repair facility. Towing charges can easily reach thousands of dollars. Not to mention the cost of downtime, unhappy customers, and angry drivers.

I hardly need to say any of that, but what isn’t so obvious is that in this new age of technological wizardry, you can actually predict many kinds of component failures. No, not to the minute or the mile, but it’s getting closer all the time, like just days in some cases. It’s thanks to AI, or artificial intelligence, more particularly to machine learning, which is a subset of AI. I’ll let Columbia University’s Engineering department make the distinction:

“Artificial intelligence refers to the general ability of computers to emulate human thought and perform tasks in real-world environments, while machine learning refers to the technologies and algorithms that enable systems to identify patterns, make decisions, and improve themselves through experience.”

It all goes back to a famous moment in computer history when the phrase ‘machine learning’ was coined by an IBM engineer in research he was doing around the game of checkers. Essentially, he taught a computer – a tiny mouse by today’s standards – to play checkers better and better by continuous learning and then beating a human checkers master with an IBM 7094 computer in 1962. How humiliated must that poor human have felt?

I mean, that happened five years before Texas Instruments produced the first handheld calculator!

Evolving predictive diagnostics

Predictive diagnostics has intrigued me for many years, and I’ve had many discussions with engineers about the topic, knowing what it could mean, always asking when we’ll see it become a commonplace maintenance tool. The answer has always been soon, or more likely soonish.

Back in 2015 I talked to Friedrich Baumann, then senior vice president – aftermarket for Daimler Trucks North America (DTNA), asking about a timeline. “I don’t think we’re further than five or 10 years away,” he said. “I think it will be an iterative process,” by which he meant that development would be incremental, one advance leading to another and so on.

Will predictive diagnostics be 100% accurate, I asked? Likely not, he said, at least not in the early stages, but he figured that even if you replace a given component 10,000 km early, the money saved in disruption and lost downtime would very likely be worth the trade-off. I’m not so sure, and I think we’re now closer than that anyway.

In fact, that same day I talked to Zonar co-founder and then CEO Brett Brinton who said his company had done prototype testing that allowed “fairly precise” turbocharger failure predictions. While he wouldn’t elaborate on what sort of precision he meant, he said “It’s the next frontier… There’s a real return on that development.” DTNA and Zonar are partners in the telematics game.

Saving days on valves and batteries

Well, I think we’re getting a lot closer to what “fairly precise” can mean in at least some cases. Like down to five days for an EGR valve, seven days for batteries. That’s the experience of Allan Babin, director of safety, compliance and fleet at Cardinal Couriers in Mississauga, Ont., according to a testimonial on the Pitstop Connect website. He uses the company’s predictive diagnostics tool embedded in his Geotab dashboard. Last winter he claims to have reduced no-starts in his fleet by 75% by knowing pretty accurately when his trucks’ batteries would fail.

Now, rest assured that I don’t want this to be or even sound like a commercial for his company, but I had to have a chat with Pitstop Connect founder and CEO Shiva Bhardwaj after reading that. He’s a young Toronto guy who spent several years coming to understand how to exploit machine learning for truck maintenance before launching his company. Interestingly, unlike most other telematics entrepreneurs, he’s not your typical computer nerd. Well, I guess he’s that, but he’s also a car guy with real experience on the greasy side through his family’s auto shop.     .

“I’ve been a part of GB Autos & Trucks in Etobicoke, Ont., since I was 12 years old, all the way through my university years,” he says. “Even now, when I’m not traveling, I spend my Saturdays there. Over the years, I’ve been involved in a wide range of activities at the shop, from helping with tire changes, oil maintenance, and vehicle cleanup, to custom audio installations and electrical work. I even delved into tuning ECUs and spent countless hours building things like pocket bikes. I gained valuable experience by taking engines apart and putting them back together. I also became deeply engaged in the racing scene, as some of the mechanics at the shop frequently raced their cars, allowing me to become actively involved in that aspect as well.”

Bhardwaj says it makes no sense to do preventive maintenance by set mileage intervals when AI is 100% better and easier to manage, raising technician efficiency by at least 20%, chopping tow costs by 30%. A fleet of 500 trucks can save $150,000 a year, he says. And get this — he claims the return on investment is 644%. Pitstop is a monthly per-truck subscription service.

It’s not the only such game in town but nearly so, and the key here is that predictive diagnostics isn’t a “soonish’ thing any longer. Other outfits are already playing in this arena or will be in the very near future and the benefits to truck operators are there for the picking.

And remember, the thing about machine learning is that it’s all about getting smarter and smarter the more it soaks up and analyzes data. It can only get better.