Uncertainty quantification Of performance and stability of high-speed axial compressors

Abstract

Geometrical uncertainties in a compressor (due to manufacturing tolerance and/or in-service degradation) often result in flow asymmetry around the annulus of a compressor that jeopardises compressor stability and performance. Usually, sensitivity of compressor stability and performance for any parametric variation is arrived at by considering all blades to have same dimension. In reality, an inherent blade-to-blade variation causes the blades to have a probability distribution. These blades can be redistributed circumferentially resulting in adjacent passage areas between different blades to be completely random and hence the performance variation. Surrogate model is preferred for quantifying the effects of parametric variation on compressor stability and performance given its quick turnaround time vis-a-vis CFD and experiments. In this thesis, uncertainties for three test cases were considered: each representative of fans on military aircraft engines, fans on civil aircraft engines and a 1-stage transonic compressor used in industrial gas turbine. This research establishes a rule of thumb to arrange blades of differing dimensions around the compressor to eke out maximum performance and stability margin. The parameters tip gap and stagger angle represent manufacturing tolerance while in-service degradation was represented by leading edge damage. For both random tip gap variation (0.15% to 0.94% span) and random leading edge damage (4% to 18% chord), the compressor performance and stability boundaries were found to be best with a zigzag pattern of blade arrangement and worst with a sinusoidal pattern of arrangement. The converse was found to be true for blades having random stagger angle variation (± 2.25% change in nominal stagger angle). The best/worst arrangement of blades with differing dimensions was ascertained using a mix of CFD and travelling salesman (TSP) analogy. The TSP analogy is handy for determining the best arrangement when two or more parameters vary simultaneously. Generalised surrogate model was developed to accurately predict the performance of compressors undergoing random tip gap and stagger angle variation. Due to its robustness, the surrogate model was combined with Monte Carlo technique to gauge the impact of parametric variation on quantities of interest (QoI). The mean absolute percentage error between CFD and surrogate models of stagger angle and tip gap (for different QoI) were found to be less than 0.14% and 1.5% respectively. This de novo analysis considers only the aerodynamic effect from geometric variations while neglecting the associated aeroelastic effects. Detailed analyses based on past experience and physical reasoning were used to validate the numerical simulations.