When you hear the words artificial intelligence or machine learning, you may think of IBM’s Watson, or the 2001 Steven Spielberg flick about a robotic boy. When you hear blockchain, you may think bitcoin. Yet these and other emerging data technologies are already being employed in trucking and logistics.
“I’ve been in the trucking world for a long time. Normally, trucking has not adopted things as fast” when it comes to technology, says Tim Leonard, chief technology officer and executive vice president at TMW Systems. But when it comes to things like deep learning analytics and blockchain, he says, “the transportation world is one of the leaders.”
Predictive analytics
We’ve been hearing for years now about the idea of predictive maintenance rather than preventive maintenance – the ability to analyze data and predict when failures are about to happen, allowing fleets to replace or repair components before they lead to a breakdown on the road.
Some fleets have already been using data from their own fleets to do predictive maintenance. As Randy Obermeyer, terminal manager in charge of equipment and maintenance for Batesville’s private fleet in Batesville, Indiana, explains, “There’s a lot of data that will tell you if you have an issue with that system before it becomes a major problem that’s going to leave the guy sitting on the side of the road.”
For instance, he says, “If you know the starter lasts for two years, you can proactively put a new one on every two years and not worry about it breaking down at 25 months and causing unnecessary downtime.”
But the next step brings in data beyond a fleet’s own maintenance records.
The amount of data is key, explains Renaldo Adler, TMW Systems’ principal of asset maintenance, fleets and service centers. The more historical information gathered, the better. A fleet of 1,000 trucks and lots of miles will have better information than a fleet with 20 or 30 trucks, he says.
That’s why big data is the key behind TMW Systems’ new predictive maintenance application, TMT Predict.Fault Code, the first tool to be offered under the company’s new TMT Predict Series of maintenance analytics solutions.
The system uses predictive models created using data from 80,000 vehicles that have traveled over 1 million miles. The data was analyzed by a team of scientists to identify patterns that occur before a part fails.
“OEMs are teaming up with us and finding hidden intelligence in maintenance expense, parts, labor, lost productivity,” Leonard explains. “So we can, within a 90% threshold, tell you vehicles that are going to have problems in the next few months.” Its research in developing the product discovered unreported fault codes, from low coolant to a serious problem with the antilock braking system.
While truck systems throw out hundreds of fault codes, Leonard says, traditional diagnostics systems don’t analyze all the data. “In the old world, no one could look at the thousands of data elements being read,” he explains. “We read the entire [CANbus] system, and match it to history of other trucks of those same miles. When you look at what we’re trying to do, with several hundred terabytes of data, we can look at all aspects of the error codes in brand new ways.”
Using the PeopleNet Mobile Gateway, TMW’s new application gathers 80 performance variables from the engine and aftertreatment system. Those variables are then transmitted to the cloud, where they are fed into eight models, which were developed by Vusion (a sister Trimble company), and then analyzed for indicators of possible failures.
When fault codes and other vehicle data indicate an increased probability of failure, a dashboard alert appears within the user’s TMT Fleet Maintenance software. The alert identifies the fleet’s assigned equipment number, vehicle identification number, probability of failure, diagnostic trouble code and description, the performance variables triggering the probability, and other key information, so fleets can decide on a course of action.
TMW is not the only company working on better predictive maintenance tools. For instance, Navistar has been piloting a new OnCommand Connection feature called Live Action Plans, which predicts when a part is going to fail before it does. Live Action Plans uses prognostic models that were developed using Navistar’s field service intelligence and algorithms based on big data analytics. The result is that when certain adverse conditions are identified on a vehicle, OnCommand Connection can give customer alerts about potential corrective actions, potential repair, the parts needed, and the training required to make the repair.
And maintenance isn’t the only arena where predictive analytics is coming into play.
The Journal of Commerce reports that chassis leasing companies, academics, and technology providers are developing predictive analytics to forecast intermodal chassis demand, which could mean fewer chassis shortages at ports. And TMW is working on developing such systems for driver management and freight network management.
Omnitracs offers predictive analytics around driver safety and retention issues. These tools look at patterns in data associated with drivers – not just safe-driving data such as speed or hard stops, but also the hours they’re working, are they happy in their work, hours of service, and more.
“Predictive analytics starts to pull all these things together and then make recommendations based on the patterns,” explains Brad Taylor, Omnitracs vice president of data/Internet of Things, identifying drivers likely to have a crash or to quit the company and giving the fleet suggestions for how to address those issues in conversations with the drivers.
From predictive to prescriptive
Taking predictive analytics a step further is the notion of prescriptive analytics, explains Jonathan May, director of business intelligence at McLeod Software.
“Predictive analytics is looking at historical trends and starts predicting out based on a level of confidence what’s going to happen,” he explains. “Customers love that — ‘Am I going to be able to buy some more trucks or trailers this year, or what’s my projected cash flow?’ Based on historical trends and where we are on a predictive model, it’s going to give you that answer.
“Once you have that, the next thing on the spectrum is, how can we get there? What do we have to do? We’re not there yet, but we’re trying to get there.”
Prescriptive analytics not only anticipates what will happen and when it will happen, but also suggests decision options to address the predicted issue and shows the implication of each decision option.
So in the area of maintenance, for example, predictive analytics may tell you that a certain component on a certain model of truck is likely to fail at a certain mileage. It’s up to the fleet manager to decide what to do with that information. Prescriptive analytics will tell you what to do about it.
The same concept applies to areas such as freight contracts and equipment investment.
For instance, Ben Wiesen, vice president, products and services, Carrier Logistics Inc., points out that today’s costing systems spit out a report telling you which customers you’re losing money on, and it’s up to the people at a carrier to decide to whether to “fire” that customer. “Tomorrow maybe what the systems will do is automatically start declining that customer’s load tenders,” he says.
May says McLeod is working on building a tool that would allow customers to ask questions such as, ‘What would be the impact if I were to buy five more trucks next year?”
“I kind of picture it as a slider on a report,” May says, which would then provide the expected impact on revenue, costs, lane coverage, and the like. “That’s something we‘re looking at building in, that ‘What if?’ capability.”
Chris Scharaswak, senior director of product development and innovation at Ryder, says many companies are pretty good at what he calls “publishing the news, which is, what does this data tell me, and creating reports. But our next step is to make sure we’re accumulating all this data and it’s normalized to a certain degree so we can do predictive and prescriptive analytics. We can predict based on trends and mathematical models if there’s going to be a failure on a component on a tractor, or predict certain times of the year that weather patterns in the Northeast cause certain delays, or when capacity is going to tighten in certain lanes and certain kinds of industries. So it’s those kind of things we can create predictive models around and then plan for.”
Predictive analytics, he says, defines what could happen, based on historical and forecast data. “Prescriptive really starts to define the best alternative solution to those central risks or the things that could happen. In certain industries, such as aerospace, they already have models like this. It’s something we’re definitely making investments in, and we know it will have huge impact on our ability to service our customers and mitigate risks and drive improvements in transportation networks and supply chains.”
Blockchain links it all together
One emerging technology that tracks data in a different way may help drive deep learning and predictive analytics further.
Blockchain is a combination of technologies that allow transactions between parties via a trusted, shared ledger. Each transaction is coded into a block, which becomes part of a chain of “blocks.” Entries or changes to the chain cannot be made without authorization of all participating members.
TMW, which plans to unveil a blockchain product this year, says integrating blockchain into freight contracts opens up a host of data. Leonard explains that once you create a blockchain ledger, both shipper and carrier can track in real time the commitments the shipper made, when the carrier delivers the freight, and more, offering “total transparency,” he says.
TMW has been running tests of its blockchain programs with select carriers, such as Dart, and says it has cut down to seven days from 21 days the time it takes to execute a transaction. Leonard says carriers could be paid almost immediately through smart contracts and blockchain.
In addition, the analytics made possible by the data in the blockchain allows for patterns to be discerned that can help shippers and carriers during contract negotiations. Carriers could see in near real time whether freight was weak or strong coming into or out of given markets.
Currently, Leonard says, “Trucking companies build contracts, get awards from shippers, and go to the next contract, and never keep those historical contracts,” at least not in a way that makes them easy to analyze. “In the big data world you have historical contracts – ‘for the last two years here’s everything we’ve done and bid on, and my analytics is telling me I am overpriced in this area, so can we work out a better deal.’”
TMW was a charter member of the Blockchain in Transportation Alliance, made up of members such as shippers, carriers, software and technology companies, is a forum for the development of blockchain technology standards and education for the freight industry.
A more recent BITA member is UPS. “Blockchain has multiple applications in the logistics industry, especially related to supply chains, insurance, payments, audits and customs brokerage,” said Linda Weakland, UPS director of enterprise architecture and innovation, in announcing UPS was joining the alliance. “The technology has the potential to increase transparency and efficiency among shippers, carriers, brokers, consumers, vendors and other supply chain stakeholders.”
Analytics at the edge
The increasing ability to share data via telematics and cloud computing means that data analytics no longer has to be a back-end office activity. Increasingly, deep learning, machine learning, and the beginning of artificial intelligence are going to allow data analytics in near real-time – in the vehicle.
Right now, telematics systems such as remote diagnostics or camera-based safety systems upload the data to some central hub for analysis. But with the help of machine learning and artificial intelligence, some say, new technologies will cut out the middleman and provide data analytics right on the truck, or “out on the edge, the place of interaction.”
Adam Kahn compares it to taking photos – at one time you had to send film out to be developed before you could see the picture. Today the photo is available immediately on your phone.
Kahn is vice president, fleet business, with Netradyne, which is applying edge analytics, deep learning and artificial intelligence concepts to in-cab, camera-based safety systems.
For instance, he explains, take hard braking, traditionally associated with risky driving. Instead of a fleet manager or a third party poring through hard-braking data and video to determine if a particular hard braking event was truly a problem, he explains, Netradyne’s Driveri camera-based system uses multiple inputs, including both sensors and camera data, and can tell on the spot if a driver had a hard brake because he or she was stopping to avoid hitting a pedestrian who just stepped out in front of the truck, rather than from risky behaviors such as following too closely.
“I think you’ll see the devices close to the point of interaction will get smarter and faster and get data from other sources and do their own analytics,” says Eric Witty, vice president of product management at PeopleNet. “Drivers, mechanics and other people not in the office will not only get data, but they’ll get better decision-making,” whether it be warnings, or driving guidance or maintenance guidance.
Another example, Witty says, are lane departure warnings. Right now, if a truck starts wandering from the lane, the driver is likely to get a warning — perhaps a rumble strip sound in the cab. But if those warnings start coming closer and closer together, they may well be a sign of fatigue. Edge analytics, he says, combine LDW with other available data such as speed, and tell the driver he or she is becoming fatigued and should pull over to rest, as well as alert the back office.
As we discuss many of these emerging technologies, inside of trucking and out, we hear about terms such as deep learning, neural networks, machine learning, and artificial intelligence. But you don’t have to understand these technical terms to benefit from them, says PeopleNet’s Witty.
“What it’s really all about is, I don’t have to know everything and do everything. I have to tell you what it is I’m trying to figure out and the problems I’m trying to solve, and the machine will figure it out and be able to provide guidance and the answer to the questions that you need, in near real time.”
0 Comments
See all comments