Hi, I'm Jen from The Career Force. I've worked in analytics for 15 years and currently run my own analytics agency, as well as teach other people about analytics and how to get into analytics jobs. Today, I'm happy to participate in 365 Data Science Use Cases - a great initiative to highlight excellent use cases for data across all different industries.

There are so many great use cases for data but one of my personal favorites is vehicle predictive maintenance.

We also made a video on the topic - you can check it out below or just scroll down to keep on reading.

Why Is Vehicle Data Important?

A lot of people don't realize just how much current vehicles are driven by software and data. And I don’t mean just electric vehicles or self-driving vehicles. The typical vehicle on the road that has been made even in the last 10 years has hundreds of thousands of lines of code behind the scenes that are helping power it, control its emissions and monitor what's happening with the vehicle for safety and multiple other reasons.

When it comes to commercial vehicles that are transporting all of our goods - for example, the tractor trailers you might see on the highway - if a vehicle has a problem, it is a much bigger issue than if your car won't start in the morning. The latter might be inconvenient for you, but a commercial vehicle, a tractor trailer, that is broken down, won't be able to deliver products. They might hold up multiple stages of the delivery process. And we’ve seen this happen during breakdowns, when stores can’t get supplies and many places are out of stock. There are production issues, but there are also delivery issues with getting things moved quickly enough to the right places.

What Are Telematic Systems?

For many years, on both the consumer and the commercial level, vehicles have come equipped with some sort of telematic systems. Telematic systems, or, vehicle tracking devices, that allow us to read, store and communicate telemetry data. These telematic systems capture information like:

location
speed
fuel consumption
all the sensors on the vehicle
what sort of readings the sensors are having
any sort of fault with a sensor
the specific sensor fault or issue

Telemetry Data and Vehicle Diagnostics

In the past, this information was really treated as a retroactive supporting vehicle diagnostic tool. Let's take commercial vehicles as an example.

So, in the past, let's say 10 years ago, if a truck or a tractor was having an issue, they would go into their local dealership or a dealership nearby. They would wait in line. Sometimes they would wait for days to be seen. There would be diagnostics, which could take many hours if there is a complicated issue. And the telemetry data was a supporting factor. As mentioned, it was one of the clues that were used to help diagnose what was happening. Then, once that diagnostic was made, the repair could be identified so the truck could be fixed and back on its way. And then parts need to be shipped because the dealership doesn't have them in stock. They need to wait longer and maybe a week or two later – the truck is fixed.

Why Is Predictive Maintenance Important?

Maybe this doesn't seem like that big of a deal when you're talking about a big fleet of trucks. For instance, a company like FedEx has tens of thousands of vehicles operating at any time. But this is a very carefully orchestrated process that ensures every vehicle is in use as much as it can be. And things are organized in a way that the time frames flow well and everything gets delivered on time.

Moreover, many trucks are also operated by small fleets. They might have a dozen trucks or less or individual owner operators where they have one vehicle and that one vehicle is their entire livelihood. So not being able to do your job, and not having any money coming in for two weeks due to an issue is a massive problem.

This whole process doesn't have to be entirely reactive, though, and this is where predictive vehicle maintenance really starts to be impactful.

How Does Predictive Maintenance Work?

Instead of just focusing on what's happened in the past as a way to support a repair, right now we can read this information remotely off of the vehicle. We can see what different issues there are, if the vehicle has been running in conditions that may be creating an issue where there's a high engine speed or a high load on the vehicle. These can typically point to earlier failure of a part than what a typical usage would suggest should happen.

By combining all of this remotely-read telematics data, along with other data we have about what vehicles have experienced issues and how we've repaired them in the past, we can actually start predicting what's going to happen and get ahead of it.

So, instead of the inconvenience of being broken down on the side of the road, instead, someone is able to contact the driver, contact the company and say: “You're having an issue, you have a little bit of time, but we want to get you scheduled to get this fixed.”

Essentially, we use machine learning to be able to do this prediction. We assess what factors typically lead up to a specific type of failure and primarily focus on failures that leave the truck inoperable, that is, it can't work or it won't run at an acceptable speed until it gets repaired. Then, we can take that information. And here's almost an entirely separate data use case in itself. We can look at all of the dealerships in the area that the vehicle is or is traveling to. We can look at their stock of different parts and identify how busy they are.

So that not only is the driver contacted and told you have this issue that needs to get fixed, but they proactively know what the issue is. They're given an appointment at a dealership along the way that is available to look at that truck right away. And they make sure that that dealership has the parts necessary to do the work. So, a repair that in the past, could have taken several days if the part wasn't in stock and the diagnostics were difficult, now can be done in the matter of a few hours. That's still inconvenient, but it's a massive shift forward and it's a much less inconvenient problem than in the past.

What Are the Benefits of Predictive Maintenance?

Predictive vehicle maintenance isn't just convenient. It can really save millions of dollars, especially for larger companies over even relatively short time periods. And even one truck with one problem can cost thousands of dollars beyond the actual repair cost because most companies have delivery time frame agreements in place, where if the delivery window is missed, there are penalties to pay. With data, and the help of machine learning, these problems can be largely avoided and anticipated and really make the process smoother for everyone.

If this topic sparked your curiosity and you want to gain deeper understanding of machine learning, check out the 365 Data Science course that's perfect for beginners, Machine Learning in Python.

Jen Hood

Articles from This Author

365 Data Use Cases: Vehicle Predictive Maintenance with Jen