Currently, there is a lack of in-situ or model test results for cone penetration tests (CPTs) conducted in deep, dense sand layers under high overburden stresses, restricting the development of empirical relationships between CPT results and the characteristics of such deep, dense sand layers. This study addresses this gap by proposing an empirical relationship to predict the relative density of dense silica sand based on stress level and cone tip resistance. The relationship was developed through CPTs performed in a calibration chamber using dense sand specimens (with relative densities of 74%-91%) subjected to high stresses (under overburden stresses of 0.5-2.0 MPa) and numerical simulations employing the large deformation finite element method. The Arbitrary Lagrangian Eulerian method was used to regularly regenerate the mesh to prevent soil element distortion around the cone tip. Additionally, the modified Mohr-Coulomb model was integrated to capture the stress-strain behavior of dense silica sand under high stresses. A reasonable agreement was achieved between the numerical and experimental penetration profiles, which verifies the reliability of the numerical model. A sufficient number of parametric analyses were carried out, and then an empirical equation was proposed to establish the relationship between the relative density of dense sand, stress level and cone resistance. The empirical equation provides predictions with acceptable accuracy, as the discrepancies between the predicted and measured relative density values fall within ±30%.