| Aeroengine performance-seeking control (PSC) is a technology of obtaining additional gains by improving the engine potential. With the application of PSC, the integrated fighting capability of modern fighter aircraft can be developed. The nonlinear model of aeroengine is the base of all related models used in PSC. Accurate on-board engine model is the key technology of PSC, and the model's accuracy is implemented by parameter estimation. Algorithms for solving the aeroengine mathematical model and parameter estimation are studied, which are two important technologies of PSC. In order to account for some practical problems of the two aspects, new theories and methods are introduced, and a series of research results are achieved.The component level mathematical model of aeroengine is built up and analyzed. (1) The metamodels of compressor characteristics are established using the metamodeling ideal. Then the compressor characteristics are reconstructed under the unknown conversion speeds to expand the initial characteristics data tables. (2) The component level mathematic model of turbofan engine is established. Problems are analyzed when carrying out the calculations of compressor characteristics interpolation and local nonlinear equation(s) solution, and methods of characteristics changing and using nondimensional variables are suggested to solve the problems. (3) The balance equations of turbofan are established, and the method to choose independent variables is given. (4) Problems, such as needing accurate iteration initial value, iteration breaking because of interpolation over range or calculation singular, and the divergence caused by bad Jacobian matrix condition number, are analyzed, when using the traditional Newton-Raphson method to solve the model.The traditional optimization algorithms are studied to solve the aerogine mathematical model. (1) The solution of balance equations is changed to a problem of nonlinear least square, using the least square criterion. The ideal using the optimization theory to solve the problem is proposed. (2) Two kinds of method based on Gauss-Newton (G-N) method, which are G-N method with step factor and Levenberg-Marqurdt (L-M) method, are studied. (3) The calculation of matrix inversion is changed to the problem of solving linear equations, which is the key computation in the iteration formula. The matrix orthogonal decomposition technology is applied to solve the problem of calculation convergence caused by the bad matrix condition number. (4) A hybrid algorithm is proposed, which combines both improved methods'advantages. According to the simulation, the solution with satisfied precision is acquired, and is better than the two basic algorithms.The modern optimization algorithms are studied to solve the aerogine mathematical model. (1) In order to avoid computing break during optimization, the nonlinear least square model with variables boundaries is developed. (2) Particle Swarm Optimization (PSO) algorithm is studied, and the appropriate algorithm parameters are selected and designed for solving the engine model. (3) Quantum-behaved Particle Swarm Optimization (QPSO) algorithm is studied, and the appropriate algorithm parameter is selected and designed for solving the engine model. According to the simulation, both algorithms with appropriate parameters can get satisfied solutions, and the feasibility using the two methods to solve the aeroengine model is verified.Parameter linear estimation algorithm of PSC is studied. (1) Adaptive discrete Kalman filter is designed to solve the problem of state model errors uncertain. (2) An indirect method to estimate measurable parameter biases is presented, which is based on CA model and Kalman filter with the nonlinear engine model joining the estimation process. (3) Federated filter is designed to solve the problems, which arise when using centralized Kalman filter to estimate component deviation parameters (CDP). Local filter with distributed processing mode and main filter with information fusion and distribution functions are design, which makes good use of the available measurements.The application of the federated filter can decrease the calculation consumption, improve the estimation precision and improve the tolerance of system. Parameter nonlinear estimation algorithm of PSC is studied. (1) On the base of analyzing reasons of component degradation, the nonlinear degradation model is established by coupling the basic balance equations with the expanded measurement balance equations, and the estimation of CDP is changed to a problem of solving nonlinear equations. (2) The testing data are preprocessed using robust Kalman filter. (3) The modern optimization algorithms are introduced to solve the degraded nonlinear model. In order to decrease the probability of get local minimum, the mutation operation is introduced into PSO and QPSO, and the CDPs are estimated using the improved algorithms with mutation. The magnitude of objective function value to get the satisfied estimation results is given.The optimization algorithms are introduced to the field of aeroengine model solution, and some problems using the traditional numeric method are solved in some degree. Two kinds of linear estimation algorithm overcome some disadvantages of the centralized Kalman filter by using the distributed computing mode, and accord with the trend of distributed control system development. The nonlinear estimation technology is studied preliminary, which uses the modern optimization algorithms. The results of the studies have some theoretical significance and practical value for the development of PSC, and can be referred for the field of aeroengine simulation and fault diagnosis.
|
PDF Full Text Request