ML in Formation Evaluation: From LAS Files to Lithology Predictions

Formation evaluation has traditionally relied on manual interpretation of well logs by experienced petrophysicists. This process is time-consuming, subjective, and prone to human error. However, the convergence of machine learning and petroleum engineering is revolutionizing this field.

Our latest research demonstrates how neural networks trained on real-world well log data can identify geological formations with remarkable accuracy. By feeding LAS files into our proprietary model, we can predict lithology types, estimate porosity and permeability, and flag anomalies, all in under 30 seconds.

The key insight is that geological patterns are learnable. Rock formations leave distinctive "signatures" in gamma ray, neutron density, and resistivity logs. Machine learning excels at pattern recognition at scales humans cannot manually process. Our models have been trained on thousands of wells across different basins, capturing the variability of natural geological systems.

In this post, we walk through the technical pipeline: from LAS file parsing and normalization, through feature engineering, to final predictions with confidence intervals. We'll share real case studies from the Permian Basin and the North Sea, demonstrating how our Formation Evaluation legacy app reduced interpretation time from 2 weeks to 2 days while improving consistency.

Whether you're in exploration, appraisal, or development, this workflow can accelerate your petrophysical analysis and unlock deeper insights from your well data.