Monitoring beyond the critical few

By bringing condition monitoring to the remaining rotating machinery, plants could multiply efficiencies, ROI, and cost savings.

>CLICK HERE to learn how Machine Health could revolutionize your reliability program with our all-new eBook<

The financial risks of ignoring balance‑of‑plant assets

Reflect on this: a typical power generation facility may house hundreds of rotating machines. Traditionally, only the “few” critical machines are labeled as “critical” due to their impact on production and the substantial costs associated with their potential failures. Determining criticality usually means the machine:

• Is un-spared (or not 100% spared)
• Has a substantial direct impact on process stream
• Operates continuously with planned outage intervals (measured in years rather than months)
• Has safety implications if it fails
• Has high downtime costs
• Or is very expensive to repair

Your facility may have many or fewer assets, but these ratios generally hold true in terms of the “critical few” and the “less-critical many.” However, your balance-of-plant (BOP) assets can also dramatically impact critical assets.

Critical assets are individually impactful because if they fail, production halts instantly. Understandably, most asset health management efforts and budgets are focused on these critical assets—they are usually the best instrumented machines in the enterprise.

BOP assets—often around 99%—are collectively impactful. For example, efficiency losses that manifest in condensers actually start in BOP systems such as vacuum pumps, circulating water pumps, and cooling towers. However, they are under-addressed because the cost of monitoring them has traditionally been too high.

In place of condition monitoring, a laissez-faire approach to balance-of-plant maintenance is more common, often consisting of one or more of these actions:

• Run to failure, which means putting off maintenance until the machine fails
• Manual data collection and human-intensive review of alarms and data
  • Only feasible if collection frequencies are less than 1x per month
  • Not practical if failure modes progress faster than 1x month
  • Asset coverage is limited based on available resources
• Online monitoring with manual review of alarms
  • Often impractical due to the large number of monitored machines—alarm fatigue becomes common
  • Overwhelmed by data, organizations are unable to turn data into actionable information
• Some form of general AI
  • Known as “generic AI,” where data from various sensors is fed into a general-purpose AI model
  • Generally this approach will yield suboptimal results for quite some time because of the long timelines required to clean your own data as well as build and train the model

In general, reliability teams tend to deprioritize BOP monitoring simply because there are too many assets to monitor and too many alarms to make the data meaningful.