Obtaining the mechanical parameters of SiC_f/SiC composites quickly and accurately is crucial for the performance evaluation and optimal design of novel turbine disc structures. A representative volume element (RVE) model of 2D woven SiC_f/SiC composites was developed using CT scanning and machine learning-driven image reconstruction techniques. The stress-strain curve was obtained by uniaxial tensile test, and the anisotropic mechanical parameters were obtained by inverse analysis using a non-dominated sorting genetic algorithm (NSGA-II). Subsequently, the uniaxial tension simulation was carried out based on the RVE model and mechanical parameters. The results show that the simulation curve is in good agreement with the test, and the errors of initial modulus and peak stress were 3.98% and 2.75%, respectively. Finally, the finite element models of the turbine disc with two braiding schemes were established to simulate the damage of the turbine disc. And the simulation results were verified by a centrifugal test. The failure modes of the two kinds of turbine discs are similar to the centrifugal test results, and the maximum rotating speed was close to the test results. The findings of this study provide a novel solution for obtaining the anisotropic mechanical parameters of SiC_f/SiC composites with different woven schemes.