Comparative Research On The Methods Of Estimation Of Aero Engine Gas Path Performance

Gas path health performance is an important part of engine health management. To improve the safety and reliability of aero-engines, this dissertation focuses on the kernel part of the health management system, which is the comparative study of methodologies for gas path performance estimation. Such an investigation is particularly important with the advent of on-condition maintenance capability of health management system.Firstly, intelligent algorithm is used to transform the performance estimation problem into an optimization problem. The genetic algorithm simulation experiments are carried out based on C++ and Matlab. The simulation results show that, although the traditional genetic algorithm is suitable in dealing with the nonlinear problems, there still exist low precision and local optimum problems. Aiming at these kinds of problems, an improvement scheme is proposed in this paper, by introducing information fusion and firefly algorithm. From the simulation, problems of traditional genetic algorithm can be solved through combined information fusion and genetic algorithm, and the estimation accuracy can be achieved within 5%. Although the firefly algorithm has high precision, it is very sensitive to the setting parameters, and a huge number of trial calculations should be done to obtain the setting parameters, so theoretical research is needed for further research.Secondly, the application of support vector machine in the gas path parameter estimation is studied. Parameter estimation problem is transformed into the modeling problem of measurement parameters and performance parameters. The support vector machine classifies the faults, then regresses the parameters. The simulation results show that when the degeneration rate is very small, the result is not that precise, which limits the using of support vector machine in performance estimation. This paper puts forward the improvement and correction of different measures, which introduces information fusion theory, regression correction strategy, the combination algorithm of genetic algorithm and support vector machine. Support vector regression with fusing correction strategy can control the estimation accuracy within 4%. Support vector machine fusing with genetic algorithm can overcome the defects of them, which can not only solve the local optimal problem during the process of genetic algorithm, but also reduce the dependence of the training samples, and control the estimation accuracy within 4%.Finally, comparative analysis of the estimation methods is carried out to summarize the limitations and application prospect of the methods, providing guidance and suggestions for further research.