Uncertainty quantification of effects of manufacturing errors on aerodynamic performance and flow of supersonic cascade

Abstract

To investigate the effect of normal manufacturing errors on the aerodynamic performance of supersonic cascade, a five-dimensional geometric variability model of the profile caused by normal manufacturing errors has been constructed based on Gaussian process and principal component analysis. A surrogate model has been proposed using the non-intrusive chaos polynomial method based on sparse grid technology, to predict the effect of manufacturing errors on the aerodynamic performance and flow of supersonic cascade under different conditions. Finally, based on a loss source model that is proposed to quantified the losses of various parts of the flow field, the corresponding flow mechanism has been explained. The results show that under the influence of random manufacturing errors: (1) In terms of aerodynamic performance, the total pressure loss coefficient of the supersonic cascade is approximately normally distributed, with no significant difference between the mean value and the nominal value but a huge deviation (16%∼25%) under different working conditions. For each working condition, the total pressure loss coefficient is most sensitive to the manufacturing errors of the leading edge, and is more sensitive to the manufacturing error of the suction surface than to the manufacturing error of the pressure surface, with the sensitivity gradually decreasing from the leading edge to the trailing edge. (2) In terms of flow field distribution, the mean flow field distribution differs significantly from the nominal flow field distribution at the area of shock, accompanied by huge fluctuations. And there are differences in the flow fluctuations under different working conditions. (3) The mechanism is that the manufacturing error of the leading edge would affect the bow shock strength and the acceleration process at leading edge, which will cause different degrees of impact on the wave pattern under different conditions, ultimately leading to cascade losses with different degrees of deviation.