This work uses a double-coolant-temperature model to evaluate the cooling performance of a double-row film holes scheme employed in a high-pressure turbine nozzle guide vane, subjecting to the variability of film hole geometric parameters set by manufacturing deviation range. The numerical predictions are performed at engine-representative Maexit =0.85, inlet turbulence intensity of 16%, and Reexit, Cx =1.5%, design blowing ratio (BR=2.5) and typical high density ratio of DR=1.95. A dual non-deterministic analysis is conducted in this paper to analyze the cooling performance robustness under uncertain conditions of consistent and inconsistent film hole characteristic parameters. First, a fundamental uncertainty quantification framework, which combines a conical nozzle parameterized model, non-intrusion Polynomial Chaos UQ methodology, and k-Nearest Neighbor clustering algorithm, is built to quantify the cooling performance while geometric parameters of all film holes are consistent. With the same film hole imperfection characteristic parameters, the likelihood of all holes being on the edge of their tolerance range increases dramatically, which further leads to more extreme MFR fluctuations being observed, ranging from 0.51% to 2.31% (μ ± 3σ). Second, a flow parameter dimensionality reduction UQ approach (FPDR), which transforms the deviations of multiple geometric parameters into the fluctuations of several key flow parameters, is proposed to address the curse of dimensionality phenomenon in predicting the cooling performance under uncertain conditions of inconsistent film hole imperfection characteristic parameters.