Case studyOil & gasReduce downtime

Compressor train uptime +12% over twelve months.

Multivariate health model catches degradation hours before trip on a critical gas compressor.

Talk to an expertDownload 2-page PDF
Anomaly detectionForecastingPredictive maintenance
RESIDUAL · CMP_042anomaly-v3.1
Top contributors: vib +3.2σ · tmp +1.1σ
At a glancetimeframes and caveats below
+12%
uptime
12 months
Audited compressor train
+41%
MTBF
Same period
vs prior 12-month baseline
3
early-catch events
Same period
Confirmed in post-mortem
8 weeks
to first early catch
Pilot
Single asset
Challenge

What was happening before.

Trips happened with little warning. Vibration alarms tripped only after damage had already started, and bearing temps were a late indicator.

Constraints
  • Asset is uninstrumented for some health proxies; we use what's in the Historian.
  • Cannot recommend manual operator action without an MOC sign-off.
  • Maintenance windows are scarce; the model must be confident enough to justify pulling forward an outage.
Solution

What we built, on what data.

Data
  • › Historian tags: vibration RMS, bearing temp, suction/discharge P&T, lube oil ΔT, motor current
  • › Maintenance log: work orders, parts, downtime hours
  • › Trip log
  1. 01
    Define a healthy envelope per operating mode
    Used Data Explorer to label clean steady-state runs across the past three years.
  2. 02
    Train a multivariate residual model
    Anomaly score rises hours before trip; validated on three historical events.
  3. 03
    Forecast time-to-event
    Project residual forward to recommend the next safe maintenance window.
  4. 04
    Hand off to maintenance planning
    Recommendations land directly in the planner's queue with severity and confidence.
Adoption

Who uses it, and where it fits in the day.

Used by the rotating-equipment engineer daily. Maintenance planner sees recommendations in the queue. Operations notified only when severity warrants.

Modules in use
Anomaly detectionForecastingPredictive maintenance
Results

Quantified, with timeframe and caveats.

  • R1Audited uptime +12% on the critical compressor train.
  • R2MTBF +41% vs prior 12-month baseline.
  • R3Three confirmed early-catch events with post-mortem sign-off.
RESIDUAL · CMP_042anomaly-v3.1
Top contributors: vib +3.2σ · tmp +1.1σ

Operational outcome (proof rule): Audited uptime on the critical compressor train improved 12% over 12 months; MTBF up 41% with three confirmed early-catch events.

Voice from the floor
"We schedule the intervention. We don't get woken up at 2am for it anymore. That's the difference."
Rotating Equipment Engineer · Reliability · Midstream Operator
Replication

Relevant for teams trying to…

Oil & gas teams responsible for compressors, pumps, or turbines with scarce maintenance windows.

Solution: Reduce downtime →Industry: Oil & gas →
Lessons
  • One healthy envelope per operating mode beats a single global model on rotating equipment.
  • Wire recommendations into the planner's queue, not a separate dashboard, or they don't get acted on.
  • Three confirmed early catches is enough to fund the next two years. Document them carefully.
Get started

Talk to an expert about a similar use case.

Tell us your unit and your KPI. We'll come back with what data you'd need and a realistic timeline based on Midstream Operator's rollout.

Book a demoDownload 2-page PDF