Research On Turbo-shaft Engine Control System Sensor Fault Diagnosis And Fault Tolerant Control

In FADEC system of turboshaft engine, sensors and actuators are crucial elements which directly affect the security and stability of the whole engine. In order to improve the reliability of turboshaft engine, research on actuator, gas path sensor fault diagnosis and fault tolerant control methods has been studied in this paper.All of the interrelated reserach is based on the nonlinear component level model of turboshaft engine. First of all, an improved Broyden quasi-Newton method with divergence determination and correction mechanisms was used to replace the Newton single iteration per time step method, which really improved the real-time property and precision of dynamic calculation process of the model.Fault diagnosis system of actuator and its sensor has been built based on an actuator model and an inverse engine model. Transfer function combined with fitting method was used to establish the actuator model. An inproved online BP network with self-correcting learning rate was proposed to establish the inverse engine model. The digital simulation shows that the diagnosis system not only can achieve accurate diagnosis and location the faults of actuator and its sensor but also have good adaptability.For the problem of gas path sensor fault, diagnosis systems based on adaptive model and online extreme learning machine respectively were proposed in this paper. The adaptive model which has high precision and adaptability was bulit through the new method of equations solving, then it was applied to the sensor fault diagnosis; Intelligent mapping module of each gas path sensor was establish by online extreme learning machine added wtih advance learning thinking. The simulation results show that the diagnosis system can isolate and diagnose single even double sensor faults and predict reconstructed signal accurately. Real-time performance and robustnesst of the system is also satisfied.After the realization of sensor fault diagnosis, research on the fault tolerant control was carried out. LVDT sensor and gas path sensor fault isolation scheme was designed. Multiple sets of controllers was built by augmented LQR method. A torque command modeling method was also proposed. On this basis, a fault tolerant control system using switching of control circuit was achieved to improve the control system stability.Finally, semi-physical simulation experiments of actuator, LVDT sensor fault diagnosis system and gas path sensor fault diagnosis system based on online extreme learning machine were carried out to verify the validity and reliability of the algorithm.