Think Ahead: Where Cognitive Predictive Maintenance is Leading EAM
One day in the not-too-distant future, data analytics platforms will give plant supervisors and maintenance technicians unprecedented powers of prescience over their industrial mission-critical assets.
Predictive maintenance, you mean? Yeah, we already employ PdM and have done so for some time. Tell us something we don't know.
Ah, but has your reliability maintenance program gone cognitive yet? If not, it soon might, as cognitive predictive maintenance is the next logical step in the evolution of enterprise asset management. Facilities across the world have already made room in their asset management protocols for this new norm.
What is cognitive predictive maintenance?
In one sense, cognitive predictive maintenance is synonymous with prescriptive maintenance, where computer programs recommend responses to emergent operational deficiencies in order to prevent them from snowballing into full-blown failures. These recommendations are based on a combination of historical failure data, real-time readings and related proprietary resources.
Specifically, cognitive PdM focuses on the lattermost element: digitized manuals, warranties or technicians' notes that can help guide operators, supervisors and maintenance professionals on how to address an impending disruption to production.
What cognitive predictive maintenance means for EAM
Dan Bigos, industrial manager for IBM's Watson IoT Predictive Maintenance, Quality and Warranty solutions, wrote on the IBM blog that the most advanced multivariate PdM capabilities on the market today used in certain applications can provide as much as 72 hours of lead time before an imminent failure event.
Whether that's enough time to preclude or mitigate the damage failures cause is, oddly, beside the point for now. What matters is that the utilization of data is widening the window between present operations and future challenges, and shrinking cost-pilfering threats that the industrial sector has long considered insurmountable.
As cognitive predictive maintenance becomes part of the standard package for EAM solutions, what practical values does it accord asset-intensive industrial businesses?
"If EAM had a mantra, it'd be this: 'Be proactive, not reactive.'"
Fast, intelligent response to danger
If EAM had a mantra, it'd be this: Be proactive, not reactive. Better to respond in a timely manner to minor repairs and inexpensive replacements than picking up the pieces after a hazardous, expensive, reputation-crushing disaster.
But while ancillary technology catches up to advances in EAM and maintenance analytics, cognitive PdM can act as an a valuable emergency response and recovery solution. Consider that between 2010 and 2016, advanced leak detection systems connected to pipelines for crude oil and refined products caught problems ahead of time in only about one-fifth of cases, according to research and analysis from the U.S. Pipeline and Hazardous Materials Safety Administration and Reuters. As such, O&G companies have a vested interest in accelerating their response times to leaks as detection technology improves.
Cognitive PdM software will act as the platform by which those augmented solutions with integrate with EAM schema - some already include their own sensors. But until the standard for detection technology rises, cognitive PdM grants industrial asset managers a leg up against failures that exacerbate quickly by assigning digital resources to specific failure modes or troublesome assets.
An automatically curated asset management library
A technician who can access the right data at the right time is invaluable. In the past, however, the only way to know for sure whether a particular work order or repair guideline was relevant to a particular task would be to have a technician tell you it is, possibly even dig it up for you. Not efficient in the slightest, but a necessary evil to uncover the truth or the path forward when threats of failure loom.
Deep reinforcement learning, a core component of cutting-edge cognitive PdM solutions, employs advanced algorithms to determine why historical digital documents and incoming data are relevant, organize them intuitively and incorporate feedback from technicians who used them in prescribed applications. Although manual input means a lot to deep learning networks, the real worth comes from their autonomy. Asset management teams have the sum of their collective knowledge at their fingertips, and no one had to sacrifice their important schedules to play librarian.