petrobras/3W

Timely detections for more proactive and effective actions in offshore oil wells!

What it solves

It addresses the difficulty of detecting and classifying rare, undesirable events in offshore oil wells. Timely detection of these events is critical to prevent production losses, reduce high maintenance costs (such as the cost of maritime probes), and avoid environmental accidents or human casualties.

How it works

The project provides two primary resources:

  1. 3W Dataset: A public benchmark dataset containing real-world instances of undesirable events from three different sources, stored as Parquet files.
  2. 3W Toolkit: A Python-based software package designed to standardize the machine learning pipeline. It enables users to generate dataset overviews and perform comparative analysis of ML algorithms specifically for oil well event detection.

Who it’s for

It is designed for researchers, data scientists, startups, and oil operators who are developing machine learning models to monitor well integrity and subsea systems during the drilling, completion, and production phases.

Highlights

  • Real-world data: Provides one of the first public, realistic datasets of rare undesirable events in oil wells.
  • Standardized pipeline: The toolkit helps ensure reproducibility and fair comparative analysis of different ML approaches.
  • Industry-backed: Developed and maintained by Petrobras, specifically their Flow Assurance and Well Integrity departments.

Related

  • Project
    WenjieDu/PyPOTS

    A Python toolbox for machine learning on partially-observed time series, providing unified APIs for imputation, forecasting, and anomaly detection on data with missing values.

  • Project
    seisbench/seisbench

    An open-source Python toolbox that provides a unified API and benchmark datasets for applying and training machine learning models in seismology.

  • Project
    sintel-dev/Orion

    A machine learning library for unsupervised time series anomaly detection that provides verified ML pipelines to identify rare patterns in signals.

  • Project
    microsoft/responsible-ai-toolbox

    A suite of model and data exploration tools that enable developers to identify errors, assess fairness, and interpret AI model predictions for more responsible deployment.

  • Project
    AIStream-Peelout/flow-forecast

    An open-source deep learning framework for time series forecasting, classification, and anomaly detection featuring a wide array of state-of-the-art models.