Success Story

From Labeling Restart Patterns in Oil Well Data to Scalable Tolerance Corridor Generation for Brazil’s National Oil & Gas Company

Download as pdf

Restarting a production process—whether an oil well, furnace, power plant, or any other complex asset —is a highly dynamic and risk-prone operation. In oil and gas, hydrate blockages during well restarts lead to significant delays for dissociation, and can cause significant costs and weeks of unplanned effort. Early detection of deviations is vital to prevent escalation and minimize losses.

Petrobras, Brazil’s national oil and gas company sought a way to define dynamic tolerance corridors for monitoring pressure and temperature measurements during restarts of more than 400 oil wells.

With Visplore’s capabilities to label well restarts and build tolerance corridors in a scalable way, we can improve the monitoring of our wells nationwide with truly accurate information.

PreviousNext

Challenges

  • Extensive sensor data history, but with sparsely labeled restart events.

  • Limited labeled data, resulting in statistically weak monitoring corridors.

  • Multiple restart patterns to distinguish (e.g., dieselvs. gas-filled restarts).

  • High variability between individual assets — no “one-size-fits-all” corridor.

  • No automation to update corridors as new data becomes available.

The Solution

From a small dataset to a complex data-powered monitoring solution that enables:

  • Pattern search: identify similar restarts using pressure and temperature history.

  • Graphical labeling: interactively define rules to classify restart patterns types.

  • Pattern expansion: automatically classify all historic restarts using labeled examples.

  • Tolerance corridor generation: create statistically based corridors for each asset.

  • Adaptive corridors: workflow for dynamic corridor adjustment as new restarts occur.

How it works

The Results

  • Hundreds to thousands of hours saved for manual labeling and corridor generation.

  • Statistically backed corridors for monitoring based on 100% more events labeled.

  • Up to 70% fewer potential hydrate plug incidents on restart that can be foreseen on time.

  • Avoiding 1 blockage event per year across 100+ wells saves from 30 to 300 days of
    downtime.

Why It Matters for Any Industry

Data Scientists can:

  • Turn pattern mining and labeling from days of scripting into minutes of graphical interaction.

  • Build statistically strong, adaptive models from just a few labeled samples.

  • Scale monitoring logic across hundreds of similar, yet non-identical, assets.

  • Enable the potential for fully automated corridor retraining and self-improving models.

Whether you’re managing oil wells, power plants, or process-intensive assets, accurate pattern monitoring can mean the difference between smooth production and unexpected downtime.

Master your data with Visplore. Start your free trial today!

BOOK LIVE DEMO

You may also be interested in…

How Luminus reduces downtimes by locating CCGT boiler leaks from shutdown events. [Read more]

How Hafslund Eco cut Francis turbine maintenance downtime by 50%, avoiding €1.1 M revenue loss.

How Hafslund cut turbine maintenance downtime by 50%, avoiding €1.1 M revenue loss. [Read more]