Skip to content

3W Dataset 2.0.0: a realistic and public dataset with rare undesirable real events in oil wells

Home / Papers / 3W Dataset 2.0.0: a realistic and public dataset with rare undesirable real events in oil wells

3W Dataset 2.0.0: a realistic and public dataset with rare undesirable real events in oil wells

Authors: Ricardo Emanuel Vaz Vargas, Afrânio José de Melo Junior, Celso J. Munaro, Cláudio Benevenuto de Campos Lima, Eduardo Toledo de Lima Júnior, Felipe Muntzberg Barrocas, Flávio Miguel Varejão, Guilherme Fidelis Peixer, Patrick Marques Ciarelli, Jader R. Barbosa, Jaime A. Lozano, Jean Carlos Dias de Araújo, J. N. E. Carneiro, Lucas Gouveia Omena Lopes, Lucas Pereira de Gouveia, Mateus de Araujo Fernandes, Matheus Lima Scramignon, Patrick Marques Ciarelli, Romualdo Branco, Rogério Leite Alves Pinto, Jaime A. Lozano Date: 2026-04-24 Paper ID: openalex:2507.01048

Summary

The 3W Dataset 2.0.0 is an updated, expert-labeled multivariate time series benchmark designed for detecting rare undesirable events in oil wells. This version expands upon the original 2019 release by incorporating more instances, additional monitored variables, and a refined data structure for improved accessibility. It serves as a critical resource for developing predictive maintenance methodologies capable of enabling early mitigating actions in the oil industry.

Key Contributions

  • Presents the updated version 2.0.0 of the 3W Dataset, featuring an increased number of instances and monitored variables.
  • Introduces a new labeling category and a refined data structure for improved efficiency and robustness in accessing multivariate oil well time series.
  • Provides a foundational benchmark for detecting and anticipating rare undesirable events in industrial oil production environments.

Open Questions & Future Work

Key Concepts

  • 3w-dataset: A public, expert-labeled multivariate time series dataset capturing rare undesirable events in oil wells.

Archivist Review

The 3W Dataset is a mature, field-defining benchmark that warrants a permanent conceptual and dataset entry. The identified open question regarding multilingual standardization highlights a critical barrier in industrial research datasets that extends beyond this specific instance. No other candidates were provided, and the approved items meet the strict criteria for long-term utility in the vault.

Approved Concepts

  • 3W Dataset: The dataset is a critical, industry-standard benchmark for multivariate time series anomaly detection in oil and gas production.

Approved Open Questions

  • Standardizing Multilingual Variable Names: Standardizing terminology in a global research dataset is critical for reproducibility and lowering the barrier to entry for international researchers.

Datasets

Metadata & Links

url
https://arxiv.org/abs/2507.01048
paper_id
2507.01048
paper_source
openalex
domain
time-series
tags
architectures
datasets
3W Dataset
concept_slugs
3w-dataset
dataset_slugs
3w-dataset
skill
TimeSeriesSkill
created_at
2026-04-27T06:02:13Z