The invention discloses a brake disc based on Bayesian inference and an optimization design method.The brake disc comprises a brake disc inner disc face, the brake disc inner disc face is of a scraggly gradually-changing curved surface structure in the radial direction, and the gradually-changing curved surface structure comprises a middle arc-shaped outer convex face protruding outwards; the method comprises the following steps: taking R1, R2, R3, L1 and L2 as design variables, selecting a satisfactory average Mises stress of the brake disc surface as an observation value, and calculating the average Mises stress of the brake disc surface according to the average Mises stress of the brake disc surface, the average Mises stress of the brake disc surface, the average Mises stress of the brake disc surface, the average Mises stress of the brake disc surface and the average Mises stress of the brake disc surface, and the average Mises stress of the brake disc surface and the average Mises stress of the brake disc surface. The design variables are reversely optimized based on the Bayesian inference theory, information updating of the design variables is achieved, and finally the optimal solution of the design variables is determined; the internal heat dissipation area is increased, the heat dissipation capacity of the brake disc is improved, the brake disc structure has good heat crack resistance, the thermal fatigue life of the brake disc can be prolonged, and the brake safety of a vehicle is improved.