As the amount of diagnostic data grows, product designers and marketers who plan for the future want to let maintenance personnel predict when failures are imminent. Prognostics is becoming feasible now that all types of information is available.
Once developers correlate sensor data with real-world wear patterns and failure modes, they can use that information to alert owners and operators when breakdowns are expected. Though implementations of this sort of swami-like forecasting are still in the future, some observers feel that today’s regulatory environment may soon turn it into commercialized technology.
"I think prognostics will appear on engines fairly quickly because of the emissions laws. We will have to be much more predictive than ever before," said Rick Hall, Product Planning Manager at CNH Construction Equipment. Strict regulations require engines to remain within tight parameters, he explained.
Along with their commercial contemporaries, the military is doing a lot of work in prognostics, hoping to improve operating time for vehicles in the field. The Tank-Automotive Research, Development, and Engineering Center (TARDEC) is one of the groups leading this effort.
“Condition-based maintenance is our term for prognostics that let us get to the point when we can efficiently tell when a vehicle is poised to break down so we can repair it and keep it running,” said Paul Skalny, Director of the National Automotive Center at TARDEC.
This research currently focuses on areas dubbed “low-hanging fruit,” which includes batteries, brakes, bearings, belts, and tracks, according to TARDEC Project Engineer Tom Udvare.
However, TARDEC cites a number of technology gaps that require further development before condition-based maintenance goes onto many vehicles. Sensors and sensor networks are critical for monitoring data, as well as inexpensive computing and data-acquisition systems.
While those technology gaps are being filled, equipment makers will be gathering data on their components. Historical data will be a key infrastructure component for any prognostic system. Data will be gathered from onboard sensors over a fairly long time frame.
"If you start gathering, say, vibration data on an axle, you want to capture data over years so you can see the trends that occur before a failure occurs," said Hall.
One of the keys for this type of early warning system is to avoid bringing equipment into shops when a failure is not imminent. That is one reason why prognostics must be based on real-world conditions, not data drawn from CAD files or other sources.
"Only when you gather historical data can you see what your limits should be. You'll have too many false alarms if you predict without good data," said Hall.