The mechanical properties and in situ stress conditions of coal reservoirs critically control the effectiveness of hydraulic fracturing, yet the continuous acquisition of relevant parameters at the well scale is often limited by logging data availability and quality. To address this, an integrated workflow combining machine learning-based parameter inversion with a fracturing suitability evaluation framework was proposed for coalbed methane (CBM) reservoirs. A supervised neural network model was developed to establish nonlinear relationships between conventional logs and key parameters, including Young’s modulus, Poisson’s ratio, and horizontal principal stresses. Based on these inverted parameters, a dimensionless Fracturing Index (FI) was constructed to comprehensively characterize coal fracturability by integrating brittleness, fracture toughness, and stress conditions, with a density-based constraint introduced to ensure mechanical consistency. Point-scale FI values within coal seams were upscaled to the well scale for inter-well comparison and regional evaluation. Results showed that FI varied relatively little within individual wells but markedly between wells, reflecting systematic inter-well variations in mechanical and stress conditions, consistent with spatial patterns revealed by cross-well profiles. Correlation analysis from over ten wells with both FI and treatment data demonstrated positive relationships between FI and breakdown pressure, injected fluid volume, and proppant volume, confirming its engineering relevance. Consequently, a four-level FI-based classification scheme was established to identify favorable zones across the study area. This FI framework provides a practical, interpretable tool for early-stage CBM development, offering quantitative guidance for well prioritization, stimulation design, and regional planning in unfractured areas.
Jian et al. (Fri,) studied this question.