<p>The Cerchar abrasivity index (CAI) serves as a critical parameter for predicting tool wear and optimizing equipment selection in geotechnical applications. Traditional CAI estimation methods rely heavily on equivalent quartz content (EQC) approaches, which oversimplify complex mineralogical compositions and often yield inadequate estimation accuracy. This study presents a comprehensive methodology for CAI estimation through detailed mineralogical analysis of 42 diverse rock samples collected from various geological formations across Korea, including sedimentary (21.4 %), igneous (54.8 %), and metamorphic (23.8 %) lithologies. X-Ray diffraction analysis identified 16 distinct mineral phases, revealing substantial compositional heterogeneity with quartz content ranging from 0 to 69.2 %. The research systematically evaluated multiple regression models, demonstrating that conventional EQC-based models exhibit limited accuracy (<i>R</i>² = 0.773, RMSE = 0.35), particularly for carbonate-dominated sedimentary rocks where estimation errors exceeded ±40%. In contrast, direct integration of individual mineral weight percentages significantly improved estimation performance (<i>R</i>² = 0.863, RMSE = 0.26), representing an 11.6 % enhancement over traditional methods and reducing errors to approximately 5% for problematic carbonate-rich samples. The optimal model, incorporating mineral compositions with porosity and Shore hardness parameters, achieved superior accuracy (<i>R</i>² = 0.896, RMSE = 0.18), representing a 49 % reduction in estimation error compared to EQC-only approaches. This study establishes estimation equations using mineral compositions and defines specific coefficients for each mineral phase.</p>