Abstract: In the actual processing of the cascade blade, the leading edge is prone to exceeding the tolerance, while other parts of the blade are not. The discrepancy in processing tolerance results in a deterioration of the aerodynamic performance consistency of compressor cascades. This paper proposes a mathematical model that incorporates random deliations in blade processing across the entire arc length of the blade profile coupled with the leading-edge over tolerance. This study examines the impact of processing deviation uncertainties on the aerodynamic performance consistency of a high subsonic diffuser cascade. A non-intrusive polynomial chaos expansion method based on sparse grids was employed to achieve this. The results show that the consistency of the blade loss coefficient and static pressure ratio is optimal at the design angle of attack, while it is lowest at large positive angles of attack. The poor performance consistency can be attributsd to the fact that the leading edge deviations result in an increase in the mean and root-mean-square values of flow loss within the boundary layers around the blade leading edge. When using ±5% deviation from the design blade as the criterion to measure the blade loss coefficient, the probability of blade profile performance consistency is 99% at the design angle of attack and 79.4% at large positive angles of attack.