A machine learning algorithm walks into a bar…
Have you heard the joke about the machine learning algorithm that walks into the bar?
Bartender asks, “What’ll you have?”
Algorithm says, “What’s everyone else having?”
While the joke is not truly characteristic to the use of machine learning in asset performance management (APM) applications, it does highlight a key consideration for use of analytics to improve equipment performance. Industrial equipment has widely varied designs, functional expectations and operating contexts which makes broad application of any given algorithm difficult. If you are not careful in your analytic design, you may run the risk of getting “what everyone one else is having” for all your equipment.
Any successful machine learning or deterministic algorithm which predicts equipment performance, identifies anomalies, or recommends actions must be underpinned by an asset strategy. Asset strategies define how equipment can potentially fail, what to monitor/maintain to prevent failure and determine what should be done when a failure risk is detected. The asset strategy is a critical success factor for any algorithm to be effective as it brings the core business, engineering and operational context to whatever mathematical approach is utilized to automate and scale equipment failure management efforts.
To learn more about how Itus Digital has approached Asset Performance Management, take a look at the recent ARC view by Paula Hollywood at ARC Advisory group @ ARC Advisory Group – Itus Digital – ARC View