Artificial intelligence is taking predictive maintenance to the next level – picking up on subtle clues to optimize maintenance schedules, identify faults and predict breakdowns before they cause costly downtime or damage.
Predictive maintenance works by monitoring the sounds, vibrations and other characteristics of machinery in order to glean insight into how it’s performing. Modern machinery often has these sensors built-in, while external sensors can be retrofitted to older equipment to offer continuous condition monitoring and on-demand diagnostics.
While “predictive” maintenance helps predict an incident, making the most of this insight requires combining it with “prescriptive” maintenance. This advises on the best course of action to address or avoid the predicted incident, says Nicolas Layus – director of ADI OtoSense BU at Analog Devices.
An AI-driven sensing interpretation platform, ADI OtoSense is an agnostic solution designed to work with a wide range of machinery and sensors. This extends to monitoring ambient noise to detect issues perhaps overlooked by onboard sensors. For example, it can analyze the sound as parts fit together on an assembly line to check for faults, or listen for the faint scraping sounds which indicate a robotic arm has brushed against something.
ADI OtoSense does not just analyze all of this sensory data, it also takes advantage of human expertise to make sense of it all, Layus says.
“You can train our solution on any specific piece of equipment, in your exact operating environment, but that training doesn’t need to be performed by a data scientist or other specialist. You might have someone who has worked on a machine for decades, they listen to it, they feel it and they know exactly what it is doing at any given moment – they can be the one to train the system, to ensure that their domain expertise is not lost over time.”
Along with monitoring equipment already in service, the technology can also improve reliability and reduce downtime by improving quality assurance testing before machinery leaves the factory. This can reduce testing time significantly while improving overall accuracy.
Once the machinery is in the field, manufacturers can offer remote performance monitoring as a service to customers to improve performance and reduce downtime. A mutually beneficial arrangement could also see manufacturers take advantage of that real-world performance data in order to improve product designs.
Looking to the future, this kind of AI-driven continuous monitoring will empower more manufacturers to provide a full “as-a-service” model, already common in the software space with services such as Microsoft’s Office 365.
Just as Microsoft moved from selling one-off copies off Word and Excel to selling annual subscriptions, motor manufacturers could evolve beyond simply selling hardware to offering a torque-as-a-service model.
“This kind of as-a-service transformation is happening in many manufacturing sectors, but if you want to reliably offer this kind of service then you need to know exactly what’s happening with your assets such as motors in the field. Once manufacturers are armed with this insight, they have the power to completely reimagine their business model to be ready for the future.”