Bill Bliem believes he’s on the cusp of the holy grail of truck fleet equipment managers: true predictive maintenance.
Bill is senior vice president of fleet services at NFI, a former HDT Truck Fleet Innovator, a member of our Editorial Advisory Board, and is at the forefront of a lot of trucking technology, from electric trucks to artificial intelligence.
“In the 37 years I’ve been in this business, we’ve been talking about measuring uptime and downtime, and nobody’s ever been able to really do it,” he said. But now, he says, he’s able to track the lead-in time to a shop, the average time in a shop, the lead-out time from a shop until it’s dispatched, and the utilization of a truck based on available hours in a day – all thanks to a new machine-learning tool NFI has been developing with a company called Noodle.ai.
As the case study on Noodle.ai’s website explains it, “Noodle.ai's Fleet Health AI app creates a digital twin of NFI’s fleet, mapping every incident, trip, breakdown, or other unplanned incident, then feeding that data into our machine learning models. This enables us to predict asset and component-level failure. On top of that, NFI gains insights into asset acquisition/disposition and preventive maintenance effectiveness.”
Bliem told me the goal is to have the program up and running by the end of this summer. So far, he said, the system is only using data coming out of NFI’s TMT asset maintenance software and its TMW dispatch software. “Going forward we’ve got [Noodle] hooked up with Daimler trying to get them hooked up to collect ECM data, not just fault codes, and we’re combining all this data.
“Everybody like me, we want data, we want data, we want data – then no one knows what to do with all that data. NFI’s the first one that knows what to do with this floodgate of data coming in.”
Bliem explained what he hopes this effort will lead to. “Say you have a 2017 Freightliner day cab tractor that has 240,000 miles on it. My vision is that we’re going to be able to sit here at corporate and say there’s a risk assessment that’s yellow on this truck, and we’re going to spit out a custom PM sheet to bring that truck in and inspect so many items. We send that specific PM sheet to that shop or vendor and say, ‘This is what you need to inspect on this truck for this PM. Get it in and get it done.’ And the next inspection may not be for another 60,000 miles depending on what the risk of that truck is. “
Obviously trucks will still need to be brought in for required yearly DOT inspections and to change the oil (right now oil change intervals are at 60,000 miles.)
In Bliem’s scenario, he said, “the first PM on a truck might not be till 60,000 miles. Why are we brining it in at 30,000 miles if nothing ever goes wrong?”
How they got here
The project was the outgrowth of a continuous improvement project NFI did on its preventive maintenance inspection process.
NFI cut the number of PMs by decreeing that no PM inspections would be done before a minimum of 28,500 miles. It cut about 40 minutes out of the time it did to do the PM inspections through the continuous improvement process, doing things such as creating special PM bays.
“While we were doing that, I starting talking about, we need custom PM sheets. The owner’s manual for your car tells you at 15,000 miles you need to check these three things and at 40,000 miles you need to check these 5 things and replace this thing. I said we need that for our trucks. We’ve been doing inspections the same way my entire career. Yes, we’ve tweaked some of the checkmarks on our PM sheets, but we’ve been checking things the same way we did 50 years ago – and trucks are a lot different today.”
He asked his staff – a group with some 150-200 years worth of collective maintenance experience – how many time they have had to replace a one-way check valve in the air brake system.
“Out of five guys, I have one guy that thinks he remembers one. Yet we spend 15 minutes inspecting these every 30,000 miles. Why?”
“Why can’t we predict when a part’s going to fail and set it up for an inspection at 20% before we think it’s going to fail? We still have to inspect all the DOT safety items, of course, but we don’t have to inspect the alternator every 30,00 miles if we don’t think it’s going to fail until 300,000 miles. So that was how it started.”
He talked with NFI’s IT department, and with a company called Uptake that’s doing fault code predictive analysis. “Which is great, but we wanted more than fault codes,” Bliem said. “To us, a fault code is a result – and we want a cause. We want to know before you have a fault code, right?”
It turned out Noodle.ai was already working with NFI to develop more efficient warehouse operations. So Bliem’s department started working with Noodle last fall on some data collection projects and has been seriously working on the predictive maintenance project since the first of the year.
Noodle has a program designed, and NFI is working with them to validate and fine-tune it. “Right now we’re just going through all the systems.”
The question is, he explained, what parts do you use a six-digit system code and when do you dig deeper into the nine-digit part codes (in the VMRS reporting system)?
Initial data on rear brake failures, for instance, “didn’t look right,” Bliem said, and that was because the data was using the system code, combining all the failures from slack adjusters and S-cams and brake shoes and the like. “So you have to go down and analyze and decide whether you do those separately or roll them into a component. So that’s the complications we’re going through right now. So we can get a good risk assessment.”
Bliem says improving uptime will allow NFI to improve fleet efficiency to the point where it won’t need as many trucks to do the same job.
“The ultimate goal, it’s the maintenance holy grail, is reduced breakdowns and to never see the truck between inspections. We should never have a non-scheduled repair. That’s the ultimate goal.”