Aeroengine Modeling Research Based On System Identification

Aeroengine modeling is an important part of the circuit simulation and semi-physical test of modern aeroengine control systems,which is of great significance for reducing the risk of aeroengine design and decreasing the test cost.The methods of aeroengine modeling can be divided into the component level modeling(CLM)based on physical method and the process mathematical model(PMM)based on system identification from test data.Although the component level modeling based on the working principle of the engine is widely applied in aeroengine modeling,it's necessary to consider the real-time and accuracy of the model.It's commonly happened that the aeroengine model established by CLM is slightly different from the test data of actual aeroengines when there are differences in the component characteristics.Therefore,it is necessary to apply the method of system identification into modifying the CLM.The process mathematical model of aeroengine can be identified based on the test data or simulation data from the CLM.In this thesis,two kinds of engine modeling methods were studied as follows:firstly,taking the turbojet engine as an example,the principle of the CLM was analyzed and methods of improving the real-time performance of the CLM were proposed.With the test data of the turbojet,the component characteristics were identified and the CLM was modified by the method of the particle swarm optimization(PSO)algorithm.Secondly,for the process mathematical model,the identification method of the linearization of the CLM and the identification method of the state variable model based on the test data were studied respectively.The pseudo-random excitation signal and the least squares method were used to identify the parameters of the state variable model,and the dynamic characteristics of the identified model were analyzed.According to the test data of the turboshaft engine,the Kalman filter algorithm was applied to identify the state variable model,and the influence of the excitation signal on the identification model was analyzed.The main research content is as follows:Firstly,the basic principle of the CLM based on the turbojet engine was studied.The primary goal of improving the real-time performance of the model was to improve the iteration speed of the internal component module and the calculation efficiency of external nonlinear equations.Correspondingly,the method of fitting and solving the temperature based on the known enthalpy or entropy value of gas and the nonlinear equilibrium equation by the Broyden method were promoted.And the interface method of the low-level C language combined with dynamic link library was used to ensure the stability of the model transplant in modeling.Secondly,for the error of the CLM and the experimental data above,the identification and correction method of the turbojet engine CLM was put forward.First of all,the principle of coupling calculation and interpolation of the characteristics of the compressor were taken into account.Then the particle swarm optimization algorithm was used to identify the coupling coefficient of the compressor and the new characteristics of the compressor were modified.The simulation results turned out that the model error was significantly reduced with the above method.Thirdly,the identification of the linearized model of the CLM was considered and the dynamic characteristics of the identification model was analyzed.Taking the CLM of turbofan engine JT9D as an example,the least square method was adopted to identify the linearized state variable model of the turbofan engine excited by the Pseudo Random Binary Signals(PRBS).And analyzing the influence of the input signal amplitude revealed that the input signal amplitude should not exceed 3%.In addition,the dynamic characteristics of the identified state variable model were discussed.It's shown that the low-frequency characteristics of the state variable model and the influence of the zero-pole distribution of the model.Furthermore,the identification of state variable model from the actual aeroengine test data was deliberated.Taking the test data of a turboshaft engine as an example,a new method of identifying the state variable model-Kalman filter method was proposed.It's evinced that the speed error of the identified speed response was less than 1%.The CLM of JT9D was used to simulate and analyze the influence of different excitation signals on the engine model identified by this method.An average method and another method of pseudo-random signal excitation were used to identify the state variable model with Kalman filter.The simulation results showed that the method above significantly mitigated the unilateral effect,and the positive and negative step response errors of the identified model were significantly reduced.