SCIENCE@home

Wisdom - Well Integrity System for Design, Operation and Maintenance

M. Santana,D. Colombo,A. Abrego,E. De Assis,F. Delesposte,R. Dias,L. E. G. Pulcinell,A. M. Souza,G. Martins,J. Papa

Published 2025 in OTC Brasil

ABSTRACT

This study introduces WISDOM, a platform designed to integrate offshore reliability data APIs into machine learning and statistical reliability models. The goal is to enhance decision-making for reliability and safety engineers by providing real-time analysis as a function of the well conditions. The platform consists of data aggregator pipelines for data collection and an extensible Model Repository ranging from classical statistics and machine learning models up to structural reliability and failure physics analysis, which provide the basis for reliability indicators and risk forecasting. WISDOM integrates offshore well data with machine learning and statistical models to predict reliability and related indicators. Failure time data is collected from historical well maintenance unstructured records and processed through pipelines using large language model (LLM) agents, with a pre-processing pipeline ensuring that proper data is used for model training. The Model Repository supports diverse model instances, managing versions and types to allow flexibility in predictive approaches. A scalable cloud system provides high-performance prediction APIs, enabling real-time analysis. The system dynamically updates, accessing the latest well conditions and model responses to ensure continuous adaptation for precise reliability and maintenance forecasting. Accordingly, the WISDOM design incorporates components and operational conditions as factors influencing failure rates and reliability curves, rather than relying solely on constant failure rates derived from historical averages, which is the current industry standard. These features ensure flexibility for risk assessment, while the deployment of WISDOM in cloud environments guarantees scalable real-time monitoring. As a result, WISDOM can enhance offshore well reliability analysis by using real-time data alongside predictive models. For instance, it can significantly improve the analysis of devices such as the DHSV by incorporating real-time pressure, temperature, and water column depth data, enabling precise failure probability assessments and highlighting the impact of covariates on failure probabilities across various scenarios. The proposed structure for the Model Repository allows flexibility in managing and updating diverse models, including statistical models, failure physics, structural reliability, and machine learning, ensuring the system's evolution and extensibility. Additionally, LLM agents facilitate the extraction of more insights, particularly from unstructured data. This broader approach empowers engineers to make informed, proactive maintenance decisions, optimizing the safety and efficiency of offshore assets.

PUBLICATION RECORD

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-20 of 20 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1