Abstract: To investigate the sealing failure mechanism of the support shoulder under thermo-mechanical loading conditions, research was conducted across three scales: macro, meso, and micro. A macro-scale simulation model of the engine under service conditions was established, alongside a meso-scale wear model for the contact surface of the support shoulder and a micro-scale fluid leakage model. This integrated approach enabled precise calculation of the leakage rate under both service conditions and the influence of machining marks. Based on the multi-scale analysis model, the influence laws of key parameters on the leakage rate of the support shoulder were systematically investigated. The parameters included initial preload force, impact load amplitude, friction coefficient of the sealing surface, and elastic modulus of the engine block. Finally, Gaussian process regression (GPR) and particle swarm optimization (PSO) algorithms were employed to optimize bolt preload force and load amplitude. The research results indicate that the leakage rate exhibits a negative correlation with the initial preload force and the elastic modulus of the engine block. However, influenced by the contact width and contact stress, the leakage rate demonstrates a dynamic variation with increasing impact load amplitude and friction coefficient: it increases initially and then decreases. The optimization yields an optimal preload force of F=135.490 kN and an optimal load amplitude of p=28.944 MPa. At these optimal parameters, the leakage rate is 1.685×10⁻⁶ g/s, representing a significant decline of about 78.5%, compared to the leakage rateunder actual operating conditions.