These days, people across many industries are abuzz about ways artificial intelligence (AI) can benefit their operations. That’s certainly true in fleet management as well.
AI uses existing data to solve problems, and so does predictive maintenance. AI takes away the human guesswork involved in predictive maintenance, putting it instead on an equation that uses data from a vehicle’s maintenance history and telematics device.
Asset management services and software company AssetWorks just wrapped up a pilot of an AI predictive maintenance program. One of its customers, the city of Long Beach, California, tested the technology on more than 600 of its vehicles. A service company called Serco also tested the technology on four of its customers’ fleets.
Avoiding Data Overload with AI Insights
AssetWorks partnered with fleet maintenance software provider Pitstop for the pilot. Pitstop’s software has consolidated over 10 billion data points from large, mixed-vehicle fleets. Pitstop then uses advanced AI algorithms to generate predictive insights on critical vehicle components like battery health, brakes, fuel systems, tires, and engine airflow.
Pitstop’s data, paired with vehicle health data from specific vehicles’ telematics devices managed by AssetWorks software, helped the software predict when a vehicle might need to be serviced or repaired.
The technology helped Long Beach Fleet Services Manager Dan Berlenbach, CPFP, avoid what he calls data overload. Vehicles can generate hundreds of codes on their respective telematics devices in a given day, leading fleet managers to struggle to figure out what data is important.
“We've had this data flowing for years. Our customers can push their telematics data into our telematics cloud today,” AssetWorks Senior Industry Consultant for Fleet Technology Marc Knight explained. “But it's like drinking from a firehose. And what this initiative is able to do is to get it down to a trickle, so that customers can actually use it.”
Pitstop’s technology sifted through the data to allow Berlenbach and his team to know whether a code pointed to a potentially significant maintenance issue on a vehicle.
Cutting Down Diagnostic Time
When a vehicle comes to the fleet shop and a work order is created, the technician will see a list of available insights about that vehicle, ranked from minor to critical. Having that data allows technicians to cut down on diagnostic time, because they already know what’s going on under the hood.
“Right then and there, before they get started [working on the vehicle], they're able to see any pending messages or alerts insights off that data…it reduces the amount of diagnosis time. We think we can probably save about 20 minutes for a job or per event in diagnosis, by having that in the technician’s hand automatically,” Knight said.
Knowing what’s wrong with the vehicle before it even rolls into the shop also allows fleets to lessen downtime, which can lead to a multitude of benefits.
“Our goal is always to maximize uptime or reduce downtime for the customer. That way, you have the least impact to their mission, so they can continue to do what they need to do,” Berlenbach explained.
Technicians can also address the non-critical alerts while the vehicle is in the shop.
“We want to take care of everything possible at that time since we're already taking it away from [the customer] for a day or for hours, or whatever it might be. Let's get everything done at once, so that when we give it back to you, it won't break down between now and the next scheduled service,” Berlenbach added. “By getting to a predictive system of maintenance, we can catch these things before they happen. And that enables us to reduce the downtime and all its negative effects.”
Reducing vehicle downtime through a single visit to the fleet shop also reduces the need for a reserve fleet, Berlenbach explained.
When Serco installed the technology in its fleet customers’ vehicles, its customers discovered that a large number of their drivers had check engine lights on. In many cases, operators simply hadn’t told their fleet departments that their vehicles needed to be serviced.
The data pulled through Pitstop’s technology allowed their technicians to decide whether the vehicles with check engine lights needed to be brought in and serviced immediately.
This technology can also allow shops to work well with the technicians they have, as many continue to experience a labor shortage.
“If we can shorten the time it takes to do these jobs, we can open up the amount of time available to technicians to do what they need to do. We can get more work through our shops with fewer resources,” Knight said.
Think of it as another tool in your technician’s toolbox.
“It isn’t about replacing but working with a current workflow to make operations more efficient and smarter, with the same or limited resources,” Pitstop CEO and Founder Shiva Bhardwaj explained. “AI predictive analytics can significantly save fleets time by doing the previously manual work at a faster pace while addressing failures before they happen…allowing managers to plan their maintenance schedules accordingly.”
Slashing Vehicle Towing Costs
Keeping your fleet in working condition through predictive maintenance can also help you chisel away at another part of your fleet spending: your towing budget. A vehicle is less likely to break down when it has had all of its maintenance issues serviced, especially the not-so-obvious ones that the predictive maintenance software can point out.
In July, Pitstop released its official Long Beach case study results. Over the course of the pilot, Pitstop observed a total of 147 tow or road call events, resulting in an estimated cost of approximately $61,000, excluding repair expenses. Per year, this amounts to $651,940 in tow and road call costs.
Pitstop reported that much of this could have been avoided if the predictive analysis had been used to resolve the issues ahead of breakdown.
The city did end up seeing savings in one specific instance, thanks to Pitstop's prediction. In December 2022, dozens of hybrid vehicles were at risk of having dead batteries because they had been sitting in a lot for a while during the holidays.
When the fleet department was alerted that the batteries might die, Berlenbach was able to send someone out to start and run the vehicles for a little bit so that they wouldn’t all be dead when employees returned to work. The technology was at work, even on the vehicles that were sitting in a lot, unused.
“The ROI, we think, is anywhere from two to five times the cost, in just savings potentially to the fleet,” Knight said. “And that doesn't equate to the broader organizational savings of reduced downtime and the safety related to cargo and passengers have a breakdown.”
Delaying Maintenance Services
One way the pilot revealed the city could save money was by delaying services for operator cases when no fault code is present on a vehicle, therefore spreading out maintenance costs. For Long Beach, this approach applies to 24% of the vehicles in the fleet.
Over the course of the pilot, 941 unplanned service cases cost roughly $1,000 per visit. By electing to note perform these services on 24% of these visits during the three-month pilot, Pitstop determined the city could save up to nearly $78,000 per month.
It's important to note that any declined services would still need to be performed eventually; these services would instead be moved to preventive maintenance visits and spread out over a more extended period. The savings would primarily be derived from spreading out the maintenance costs.
Overall, Pitstop concluded that adopting a predictive, data-driven approach could provide Long Beach with an estimated $809,500 in cost savings per year.
Hesitant to Trust the Technology? Try it
Trusting technology over your own expertise of a vehicle can be difficult. Berlenbach’s advice? Test it out.
“Try it and see if it matches what your gut told you,” he said.
Based on feedback from its pilot, AssetWorks is targeting the release of a fully integrated AI predictive maintenance product later this year. For now, it’s offering a discounted subscription fee for early adopters who are interested in signing up to use the technology.
Pitstop’s models work better when they have more data.
“The more enrollment we have, the better the data comes in, the more accurate the models are. Bigger is better,” Knight explained.'
Editor's Note: This article was updated on July 13, 2023 with information from Pitstop's official case study report.
0 Comments
See all comments