Event data recorders

In addition to gathering diagnostic and operational data from the vehicle, onboard computers can also monitor unusual events such as hard braking, hard turns, or rapid acceleration or deceleration. The systems then issue alerts to the driver and often to the back office as well.

J.B. Hunt receives real-time alerts when a driver hits the brakes hard. "We have seen a significant reduction in hard-braking events when fleet managers are getting that information," says Palmer, but the driver needs to receive feedback about an incident when it happens, not a week or a month later. "It has to be real time," she says.

At multinational carrier Schneider National, director of engineering Dennis Damman is testing a system that puts event data on a Web site where drivers can review it. "It drives insight," he says.

United Parcel Service uses event logging in its tractor fleet for accident analysis. The onboard event recorder watches revolutions per minute, engine speed, GPS location and other variables. Onboard computers may store activity data during the 60 seconds leading up to an accident and a short time thereafter. "You can see when they shifted, whether rpm's were in the proper range, whether the brake was applied," says Jack Levis, director of package process management.

Drowning in analytics data

With more vehicle data available than ever before, managers face the challenge of what to do with it. Schneider evaluates summary data only bimonthly. "We don't have time to look at all of the data that comes in [in real time], particularly when you're talking about 10,000 trucks," says Damman.

Analysts use software tools to scrutinize the summary data for critical cost and mechanical-failure trends. "We used to wait for the mechanics [to diagnose problems after the fact]. Now, with data-mining capabilities, you can react to a problem before it becomes a major issue and change it on the fly," says Damman.

Palmer agrees that the firehose of available data can be overwhelming. "Unless you have the processes and people [to use the data], it can be information overload to send all of the information from the ECM in real time. And it costs a lot," she says. J.B. Hunt doesn't transmit any diagnostic fault data. "If it's a critical fault, the engine protects itself, and we do regular enough maintenance that we're not looking for real-time alerts," she says.

But UPS can't get enough of it. The global package-delivery leader takes in all of the GPS, vehicle and event information the vehicle transmits, filters out what it doesn't need and analyzes the rest in an enormous IBM DB2 database. The operations research group at UPS includes mathematicians who go through this data to find what are known as "outliers" and correlations, using statistical packages and techniques such as clustering.

"We're just scratching the surface of what there is to find out in vehicles. We can predict a failure before it happens," says Levis. "One example had to do with an alternator. The precursor to the alternator going out was this change in voltage. [Operations research] found a failure and, by looking at many vehicles, asked, 'What was the outlier event that caused the failure?'" By monitoring alternator voltage levels, UPS was able to address the problem before vehicles failed in the field.

UPS runs its vehicles for 20 years. While the technology doesn't increase the number of years UPS can run its fleet, the company's use of analytics lets it operate the vehicles more efficiently and with fewer breakdowns, Levis says.