Research On Uncertainty Quantification And Design Method Of Aero-Engine Compressors

Many uncertainties exist in the aero-engine compressor,resulting in significant performance dispersion and frequent unqualified problems.Therefore,it is necessary to conduct in-depth research on compressor uncertainty quantification and find out the critical uncertainty factors and the coupling mechanism affecting the performance dispersion of the compressor.Then the performance dispersion of the compressor can be effectively controlled,which is essential to improve the aero-engine performance.Firstly,the geometric uncertainty characteristics of compressors have been investigated based on actual measurement data.An automatic and accurate identification and statistical analysis program to extract the geometric uncertainty characteristics of the compressor is developed,and the geometric uncertainty characteristics are analyzed based on thousands of actual measurement data of the compressor blade has been completed.The research shows that there are many complex uncertainties in manufacturing,and the geometry shows significant systematic and random deviations,with local relative deviations of up to 50% order of magnitude.Next,the sensitivity and coupling effect of the uncertainty factors of the compressor is studied.Due to the strong nonlinearity of the compressor,the uncertainty mechanism is still unclear.To address this problem,the nonlinear neural network is used as the surrogate model.At the same time,a linear model based on game theory is constructed to explain the local prediction of the neural network,establishing an interpretable method to extract the physical mechanism from the "black-box" neural network.The study reveals that the impact of a large number of uncertainties on the performance of a compressor is not simple linear addition but a significant nonlinear coupling effect.This result points out the necessity of simultaneously considering high-dimensional uncertainties to assess the performance dispersion of a compressor.Then,research on the dimensionality reduction method for the high-dimensional uncertainty quantification problem of compressors is carried out.To address the challenge of the large sample size for high-dimensional uncertainty quantification,a new method of feature selection based on the bi-directional search is proposed,which couples critical parameter identification with the model construction process.Using this method,an order of magnitude reduction in the required sample size is achieved with comparable model accuracy,implementing the high-dimensional uncertainty quantification modeling study in a multi-stage compressor for the first time.Based on the uncertainty quantification result,the performance dispersion control and robust design optimization are carried on for a multi-stage compressor,achieving a significant effect of a 2.0% increase in the average efficiency and a 26% reduction in the efficiency dispersion.Finally,the establishment of the aero-engine uncertainty-based design system is discussed.By comparing with the deterministic design system of aero-engine currently adopted at home and abroad,the paper systematically proposes how to establish the aeroengine uncertainty design system from the theoretical level.The uncertainty-based design system aims to achieve successful research and development with a single iteration.The performance dispersion is quantitatively evaluated and controlled at each design stage to produce a comprehensive optimal design scheme in terms of performance,reliability,and cost throughout the life cycle.To realize the reform from the deterministic design system to the uncertainty-based design system for aero-engines,changes required are discussed in five basic elements,including process,method,platform,specification,and organization.The research results of the thesis have been applied to evaluate and control the performance dispersion of compressors for several types of aero-engines,including the second-generation,third-generation,and fourth-generation aero-engines.