Study On The Intelligent Modeling Of The Blade Aerodynamic Force In Compressor

The blade is one of the key parts of aviation compressors,which plays a role in energy conversion in work engineering.Its quality directly affects the overall performance and operation safety of the engine.With the continuous pursuit of energy and power,the compressor continues to develop in the direction of high load,high efficiency,and large flow.The internal environment of the compressor is very bad when it is working.The three-dimensional strong nonlinear and unsteady characteristics make the blade surface of the compressor bear a high centrifugal load,and the problem of blade flutter becomes more and more prominent under the action of complex unsteady aerodynamic load.Blade flutter will lead to structural vibration divergence and damage of blades in a short time.Under high rotating speed,the centrifugal force and aerodynamic load on the blade are very large,resulting in high cycle fatigue failure of the blade easily,which causes the blade to be damaged in a short time,and the harm is very great.Therefore,it is of great significance to accurately predict the flow field of vibrating blades and the unsteady aerodynamic force generated during the compressor design stage.Due to the fluid-structure coupling characteristics,three-dimensional strong nonlinearity,and inconstancy characteristics of the blade flow field,the analysis of blade vibration under airflow excitation force is still one of the technical difficulties.Due to the complexity of the flow in the impeller,the direct CFD method is used to analyze the blade flutter of the compressor.Therefore,the establishment of an efficient and accurate unsteady aerodynamic model is the basis of the aeroelastic analysis of the compressor.In recent years,the unsteady aerodynamic force model based on CFD technology has been developed.Through the learning of a small number of CFD samples,ROM can achieve the accuracy close to that of CFD simulation,which improves the computational efficiency of unsteady aerodynamic force by one or two orders of magnitude.This advantage makes ROM easy to be coupled with other disciplinary models,and used for multidisciplinary analysis,control,and optimization design research.In this project,high-fidelity sample data are obtained based on CFD technology.The mathematical relationship between structural displacement and aerodynamic force is established by using the achievements in the field of machine learning.The physical characteristics of the system are analyzed and the complex fluid-structure coupling phenomenon is explained by using mathematical principles.The dynamic mode decomposition method was used to decompose the three-dimensional aerodynamic snapshot of the compressor blade flow field,and the complex high-dimensional flow field was decomposed into low-dimensional decomposition,to extract the coherent structure of the complex system and the frequency and growth information of each model.According to the extracted coherent structure,the reconstruction and prediction model of flow field evolution was established.The prediction results of the machine learning method and dynamic mode decomposition method are compared with the aerodynamic values obtained from the CFD simulation.The determination coefficient R2 and root mean square error（RMSE）were used as evaluation indexes.The results show that the prediction results based on the machine learning model are in good agreement with the CFD results,and the computational efficiency is significantly improved.It is indicated that the aerodynamic intelligent model based on the machine learning method and the dynamic modal decomposition method is worthy of further study in the evaluation of blade vibration stability.