Uncertainty quantification of leakage characteristics of stepped labyrinth seals based on sparse PC-Kriging model

Abstract Based on the sparse PC-Kriging method, a high-fidelity surrogate model was constructed to map the geometry of a stepped labyrinth seal to its leakage characteristics. Based on this surrogate model, Monte Carlo sampling was employed to analyze the impact of geometric errors on the uncertainty of the seal’s leakage characteristics and conduct sensitivity analysis. The results indicate that within a 99.7 % confidence interval, the maximum discharge coefficient can reach 0.584; compared to the baseline value of 0.52118, the worst-case scenario could lead to an approximate 12.1 % increase in the discharge coefficient, significantly increasing leakage. Sensitivity analysis reveals that the tip clearance of the fifth tooth exhibits the highest sensitivity to the discharge coefficient, approaching 80 % and dominating the response, while the remaining parameters show lower sensitivity, with 47 of them having sensitivities close to zero.