How Predictive Maintenance Tech Works Before Parts Fail
July 14, 2020—As we've seen before, fleet managers' experiments in telematics and connective technology will provide a blueprint for the tech's rollout to the fleet of regular vehicle owners.
Recently, Garrett announced the launch of its Early Warning System (EWS) predictive maintenance software for fleets. If implemented in the light vehicle market, this could be a crucial tool for repairers.
Because Garrett works in both commercial vehicle aftermarket fleet and light passenger vehicle programs, ADAPT reached out with some questions about how their platform works. Answers came courtesy of communications director Mike Stoller.
Garrett EWS boasts the monitoring of 700 health indicators. Can you talk about how a modern vehicle is able to produce that data and how that's different from older vehicles?
First, modern vehicles are much better equipped from a sensor, ECU and controls perspective than older ones.
Second, these new era vehicles can provide much more data, including that which is dictated by regulations (pollution, safety, energy efficiency and right to repair laws). And as you consider we are moving toward more autonomous and connected features, modern vehicles need to have more control and better insights into the way they behave on the road. Even though older vehicles which are still running are not equipped in the same manner, Garrett EWS is still capable of extracting information from the small set of signals offered by older models.
Garrett EWS is based on physics and AI, therefore capable to detect those fine changes in the behavior of the vehicle and report out when or why a failure is likely or imminent.
How do predictive maintenance systems mingle both internal data (from a single vehicle) and external data (aggregated from a bunch of connected vehicles) to reach maintenance solution?
In Garrett EWS (and not in any other systems we have seen) the models delivering insights are based on physics and they are calibrated and running on each vehicle, at an individual level. That means that every vehicle that runs in our platform has its own individual set of models and calibration data, depending on the driver behavior, street conditions, vehicle condition. We are indeed comparing the performance of our models against a large amount of aggregated data insights in order to validate the models.
It's interesting to see commercial fleets explore predictive maintenance with telematics. Would the system look any different for everyday, light-duty vehicles?
No, the same system/platform can and will be used by any vehicle that can bear a dongle on its ports (or directly connected on the CAN buses).
How does Garrett envision repair shop participation in an environment with stronger predictive maintenance data? Are there models for subscription or member-based groups, or would vehicle owners be sharing data with shops?
Yes, there are subscription models for repair shops and they can build their own monitored customer fleet to improve their level of service.
Today we work directly with a very large number of garages to understand the real-life issues, like what’s most important to them and to validate the performance of our models in Garrett EWS against what they see.
Together, we are establishing partnerships to understand what actual failures in the field look like, and we are using those experiences to feed back into our physics based models insights
Most relevant to repair shops is the fact that we are can collect data remotely, thus allowing us to offer insights from the moment when the vehicle begins to be driven, not just in the service bay “at cold” condition.