Clay Formation Characterization Using Supervised Machine Learning and Stochastic Multimineral Modeli

Abstract

A recent study supporting the formation evaluation of argillaceous units considered for deep geological storage of radioactive waste will be presented in this session. Conducted within the TBO project, this work integrates over 5 km of core data from nine boreholes with a comprehensive logging suite — including Spectral GR, Density–Neutron–PEF–Sigma, and Elemental Spectroscopy — as well as high-resolution core measurements (XRF, MSCL).

The workflow highlights:

• Rigorous depth matching, bias checking, and QC across datasets.

• Application of supervised machine learning to enhance quantification of key accessory minerals such as smectite, chlorite, and TOC.

• Integration into a stochastic Multimin inversion, calibrated to core porosity, grain density, and XRD mineralogy.

• Monte Carlo uncertainty analysis with 800 iterations ensuring robustness of results.

The resulting model delivers a continuous, uncertainty-qualified description of mineralogy and porosity — critical for the safety assessment of radioactive waste repositories.

Speaker : Serge Marnat

Senior Petrophysicist & Technical Director – Ad Terra Consultancy

With over 30 years of industry experience, Serge Marnat is a distinguished geologist engineer from IFP School. His career spans TotalEnergies, Addax Petroleum, and Ad Terra, covering diverse geological settings from nickel mining to complex subsurface reservoirs.
He is recognized for pioneering advances in deterministic and stochastic multimineral workflows, machine learning applications, and uncertainty evaluation within formation characterization.

Join Us

Don’t miss this opportunity to gain insights into the next generation of petrophysical workflows that blend machine learning and stochastic modeling for high-stakes geological evaluations.

🎟 Free attendance – registration required. Send by email to Federico Games: federicogames@gmail.com.