The present research employs a 3D scanning methodology to evaluate the roughness of rock joints and proposes the concepts of the embedment ratio (T) and 3D average inclination angle ( \({\overline\theta}_p\) ) to characterize the 3D roughness coefficient (JRC3D) of joints. Direct shear tests on joints with varying JRC3D revealed: (1) a positive correlation between T or \({\overline\theta}_p\) and concrete-rock interface (CRI) shear strength; (2) increased JRC3D enlarges damage zones but reduces peak shear displacement (sr); (3) an exponential JRC3D − peak dilation angle (ip) relationship implying asperity degradation. Leveraging these findings, established the NCRSM constitutive model via multivariate regression. Validation against experimental data and alternative models confirmed NCRSM’s accuracy in capturing CRI shear mechanisms, particularly strain-softening stage. Implemented in rock-socketed pile analysis, NCRSM reformulated shaft resistance transfer functions, achieved close agreement with FEM simulations. Parametric analysis demonstrated that \({\overline\theta}_p\) contributes more significantly than T to RSPs performance. When T ≥ 40%, its contribution to enhanced the load-bearing capacity of RSPs diminishes significantly. In contrast, the contribution of \({\overline\theta}_p\) remains unaffected and does not diminish despite its persistent increase. T − \({\overline\theta}_p\) synergy framework enables optimized RSPs design in complex rock formations.
