NextPrevious

STREAM 4 | Analytics, big data, AI, machine and deep learning: An Integrated Approach towards Capturing Machinery Health

  • What is hybrid modelling and how can it give you deeper insights than Neural Nets, Regression and Principal Component Analyses
  • How do you inform node weights and network levels
  • How do you incorporate into traditional modelling and extract meaning
  • How integrated modelling can be used for high-precision failure prediction/state analysis

Predictive Maintenance is currently regarded as the gold standard for machinery maintenance where companies endeavor to identify emerging equipment failures and propose proactive corrective action.  The goal is to eliminate the cost and downtime associated with pure reactive maintenance while eliminating the unnecessary overhead brought on by Preventative Maintenance practices.  The incumbent approach, however, has potential limitations in sectors such as Industrial Gases where contractual requirements related to onstream time lead to a multi-year outage planning time horizon.  Furthermore, an equipment failure needs to be placed in the proper systems context to understand relative criticality.  With these challenges in mind, Air Products has been endeavoring to develop the next generation in Machinery Health which combines Predictive Maintenance approaches with RAM (Reliability-Availability-Maintainability) analysis to drive effective maintenance decisions over multiple time scales. We are leveraging diverse IIoT sensor technology and infrastructure to capture the current equipment state and then apply advanced analytics to project likelihood of failure over time. This approach integrates our maintenance decision-making with the customer requirements and production schedules to deliver highest level of safety, reliability and on-stream time.

 

NextPrevious