JP Venezia has a problem.
The assistant director of maintenance at Venezia Bulk Transport has trucks shutting down on highways, seemingly at random, with emission-related issues.
Venezia Bulk Transport, in Limerick, Penn., operates 500 trucks in the Mid-Atlantic region.
Modern diesel engines are programmed to derate their speed and torque when malfunctions, typically with exhaust after treatment systems, negatively impact emissions.
Venezia does not receive advance warning of when engine shutdowns are imminent – at least not yet.
The costs of “towing trucks all over” are high and the fleet’s customer service takes a hit when these unexpected events happen, he says. Many of its customers expect deliveries within a one-hour time window of appointments.
During the morning’s General Session, Venezia learned of a new predictive maintenance tool that he believes is “on the front edge” of technology. “It spoke to me,” he says.
Venezia Transport runs a mixed fleet of trucks. All Peterbilts in its fleet come standard with PeopleNet Mobile Gateway (PMG) devices that connect to real-time data from the engine. The devices wirelessly transmit data to a remote diagnostics portal to view notifications when diagnostic trouble codes (DTCs) appear.
A truck could already be derating when a DTC appears, however.
With the new product, announced as TMT Predict.Fault Code, users like Venezia will be notified up to three days in advance of an imminent derating event. With this information, fleets can take the necessary actions to avoid breakdowns.
Trimble companies PeopleNet, TMW Systems and Trimble Transportation Mobility Analytics (TTM Analytics) jointly developed the product. PeopleNet has PMG devices in Peterbilt and Kenworth trucks that power the OEMs’ remote diagnostics services. TMW Systems has a fleet maintenance system called TMT. The TTM Analytics (formerly Vusion) division of Trimble transforms the data captured by the PMG devices into predictive information fleets can act on within TMT Fleet Maintenance.
Anne Hunt, a data scientist at TTM Analytics, says she used a statistical learning method called “random forest” to create the model. The method successfully identified the paths that dozens of signal variables – temperature, pressure, velocity, torque, throttle, etc. – follow to reach eight types of “red lamp” fault codes for imminent breakdowns.
When data follows the red lamp paths for de-rate events, the model predicts with more than 90 percent accuracy an event will occur. Information about this and other probable failures appear in a dashboard in the TMT Fleet Maintenance software.
Maintenance was not the only breakthrough predictive model announced at the conference. Thomas Fansler, president of TTM Analytics, spoke of a driver retention model that uses mobility data from PeopleNet and enterprise data from TMW Systems to predict the specific factors that will determine turnover for each driver in a fleet.
After testing a proof of concept of this model with three fleet customers, Fansler says the predictions for which of their drivers would quit (within one week of the event) were more than 94 percent accurate.
“Without data science, you’re just another person with data,” Fansler said. “These analytics take complex business problems and transform our ability to solve them and manage them effectively.”
Presented By: Electronic Funds Source