Research On Intelligent Aeroengine Performance Deterioration Mitigating Control

Intelligent engine control technology has become an important development direction for future engine control system. As a branch of intelligent control technology, performance deterioration mitigating control(PDMC) has been paid widespread attention. The related key technonogies of PDMC which include engine modeling, robust control and health management are studied from the perspective of intelligent engine. To verify PDMC, simulations on turbofan engines are carried out.Modeling technologies for turbofan engines are studied firstly. To overcome the disadvantages caused by tranditional partial derivative method and fitting method in the state variable modeling process, an improved artificial bee colony(ABC) algorithm is presented and used to establish the aeroengine state variable model. The responses of the established state variable model are in accordance with that of the nonlinear model at the same steady point. Meanwhile, a simplified thrust model is established using artificial neural network and improved recursive reduced least square support vector regression(IRR-LSSVR). In this presented method, the simplified thrust model is divided into two sub-models. One uses flying condition to estimate the sensors signals, and the other uses the estimated sensors singnals to map the thrust, which overcomes the disadvantage of direct thrust estimation and increases the accuracy of simplified thrust model.The multivariable robust control method is also studied in this paper. A multi-objective optimization method called MOABC is proposed to solve the controller design problem in H2/H∞ based on ABC algorithm. The presented algorithm uses non-dominated solutions in the external archive to produce offspring. After that, food sources in the offspring and parent are sorted based on non-domination. Then, the food sources in the Pareto front are used to update external archive based on crowing distance method. Numerical simulations compared with NSGA-II and SPEA-II show that the MOABC method holds good convergence and distribution. Subsequently, H2/H∞ robust controller for aeroengine is designed based on MOABC optimization. H2 performance index and H∞ performance index are considered as two objectives of multi-objective optimization problem, and the controller parameters on the Pareto front are obtained. Simulation results on the nonlinear model demonstrate that the controller has advantages of strong robustness, anti-disturbance ability and good dynamic performance.The fault diagnosis for sensors and components of aero-engine is also studied in this paper. According to local learning and ensemble learning technologies, a method for sensors fault and abrupt components fault diagnosis of aero-engine is proposed based on support vector machine-extreme learning machine-Kalman filter(SVM-ELM-KF). The IRR-LSSVR is extended to classification machine to distinguish sensor faults and component faults. ELM is used for sensor fault location, while the improved Kalman filter is used for health parameters estimation of each component, and located the components fault consequently. Simulation results show that the proposed method for fault diagnosis can distinguish sensor faults and abrupt component faults accurately, and locate the faults effectively. Then, the model-based and data-driven-based(DDB) aero-engine component fault fusion diagnosis method is presented in this paper. The extreme learning machine(ELM) needs more nodes in hidden layer and has poor generalization ability, since ELM inputs weights and hidden layer biases are generated randomly. To overcome these drawbacks of ELM, an improved differential evolution(IDE) algrithm is presented to optimize the inputs weight and the hidden layer biases. Meantime, SVD-based Reduced-dimensional Kalman filters are used to estimate the health parameters which can solve the problem of limited number of sensors in the model based diagnosis method. To improve the accuracy of fault diagnosis for aero-engine component, the prediction results of healthy parameters based on both methods are fused by IRR-LSSVR. Simulation results show that the proposed fusion method improves the accuracy of fault diagnosis significantly. Subsequently, improved on-line sequential extreme learning machine(Im OS-ELM) is proposed. On one hand, the presented algorithm overcomes matrix singular and ill-posed problems by regularization, which can improve the predicted accuracy and achieve predictive ability at the start stage of training. On the other hand, the output layer weight vector was updated selectively based on generalization capability, and it largely reduced the mean training time of the algorithm. In order to verify the effectiveness of the proposed algorithm, simulation tests are carried out using time series data. The proposed algorithm achieves higher accuracy and faster speed. Finally, Im OS-ELM was applied to sensor fault detection and isolation of aero-engine. The simulation results show that the sensor faults diagnosis method using the proposed algorithm could detect and isolate faults of double-sensor failures and single-sensor drift, which also prove the validity and feasibility of the proposed algorithm.The thrust estimation and outer loop controller designed methods are studied. Considering aeroengine with perfoermance deterioration, thrust estimators based on ALQR and IRR-LSSVR are designed, and simulations verify the accuracy of the thrust estimators. Then, the simulator of aeroengine performance deterioration mitigating control is designed, and the integrated simuations are carried out. The integrated simulations with different degrees of engine performance degradation demonstrate that the PDMC can exploit the potential of the aeroengine and achieve thrust matching of different degradation level engines and the yaw of plane can be avoided. The simulations also show that the PDMC has a significant application prospects in engineering practice.