Unlocking Large Amount of Untapped Legacy Log Data Through High Performance AI-Driven Processing

Electrical log data from wellbores, acquired over nearly a century, are critical for subsurface characterization, providing insights into lithology, porosity, and fluid content. However, the vast volume, diverse formats, and variable quality of these datasets pose significant challenges for manual processing. To address this, an automated workflow integrating machine learning (ML) and artificial intelligence (AI) has been developed to process large-scale well log datasets efficiently and accurately.

The proposed workflow revisits conventional log processing steps, including data recognition, unit harmonization, error detection and correction, and log splicing, with two primary objectives: ensuring consistency and accuracy across multiple wells and enabling automation. The methodology incorporates advanced techniques such as Bayesian analysis, forward modeling, and ML-based reconstruction to assess log quality, identify and correct errors, and reconstruct missing or invalid intervals. A multi-well approach enhances accuracy by leveraging inter-well comparisons to detect anomalies and ensure geological consistency.

The workflow generates high-quality composite logs with accompanying quality indicators, streamlining the review process for subsurface specialists. Successfully applied to tens of thousands of wells, the methodology drastically reduces processing time from years to hours. This approach unlocks the value of extensive legacy datasets, improving subsurface modeling, operational efficiency, and decision-making.