To quantify the impact of overlying karst caves on the stability of large-section tunnels in karst strata, a study was conducted based on the Tongzi Tunnel, combining numerical simulation and orthogonal experiments. First, the reliability of the Midas GTS NX plane strain model was validated using monitoring data from the on-site karst-free section. Subsequently, a three-factor, three-level orthogonal experiment was designed to investigate the influence of filler type (sandy soil, clay, void), cave-tunnel distance (4 m, 5 m, 6 m), and cave diameter (8 m, 10 m, 12 m) on tunnel crown settlement and surrounding displacement. Through range analysis and variance analysis, the order of influence and statistical significance of each factor were clarified, and a multiple linear regression model for crown settlement prediction was established. The results indicate: (1) Filler type is the most significant factor affecting crown settlement (range Z = 3.97), with the order of influence being: filler type > cave-tunnel distance > cave diameter, and it exhibits high statistical significance ; (2) Cave-tunnel distance is the most critical parameter controlling the horizontal convergence of the tunnel; (3) The established crown settlement prediction model demonstrates high accuracy (R2 = 0.986), providing a quantitative theoretical tool for stability assessment and construction control in karst tunnel engineering.