Unsupervised Machine Learning for Reservoir Rock Type Classification in the Ngimbang Formation, Northeast Java Basin, Indonesia

Abstract

Reservoir characterisation is a fundamental step in hydrocarbon exploration and production, requiring accurate identification of rock types and productive zones. Conventional petrophysical analysis alone is often insufficient to capture the full complexity of heterogeneous carbonate formations. This study applies unsupervised machine learning, specifically K-means clustering, to classify reservoir rock types within the Lower Ngimbang Formation of the AXEL Field, Northeast Java Basin, Indonesia. Wireline log data from two wells  AXEL-1 and AXEL-2 comprising Gamma Ray (GR), Deep Resistivity (ResD), Neutron Porosity (NPHI), Bulk Density (RHOB), and Sonic (DT) logs were used as input. Quantitative petrophysical analysis was first conducted to estimate shale volume (Vsh), effective porosity (PHIE), and water saturation (Sw) using  Indonesian Equation. K-means clustering was subsequently applied, starting from 20 initial micro-clusters that were progressively consolidated into five geologically meaningful rock types: Hydrocarbon, Water-Bearing, Tight Formations, Coal, and Shale. The selection of five final groups was guided by a two-step validation approach using the Cluster Groups Randomness variance ratio generated within Interactive Petrophysics, which confirmed K=5 as the most geologically defensible configuration. In AXEL-1 Zone 1 yielded a net pay of 46.82 m (avg PHIE = 0.089, avg Sw = 0.352), confirmed by DST gas flow rates of up to 7.599 MMCFGPD, while AXEL-2 Zone 1 returned a net pay of 12.50 m (avg PHIE = 0.095, avg Sw = 0.057) reflecting lower hydrocarbon potential. The Hydrocarbon Potential facies (Group 3) identified by K-means showed strong spatial agreement with conventionally defined productive zones. These results demonstrate that unsupervised machine learning provides a practical and reproducible complementary framework for reservoir characterisation and rock type classification in complex carbonate reservoirs.