涡轴发动机控制系统传感器故障诊断与容错控制

发布时间：2018-04-05 14:33

本文选题：涡轴发动机　切入点：故障诊断　出处：《南京航空航天大学》2016年硕士论文

【摘要】：涡轴发动机FADEC系统中传感器和执行机构直接影响控制及整个发动机的安全稳定运行。为了保障涡轴发动机可靠性,本文开展了执行机构及其传感器故障诊断、气路传感器故障诊断以及容错控制相关研究。本文基于涡轴发动机非线性部件级模型开展上述研究工作,首先采用具有计算发散判定和校正机制的有限次通过改进Broyden拟牛顿法替代牛顿一次通过法,提高部件级模型动态过程计算实时性和精度。构建了基于执行机构模型和发动机逆模型的执行机构及其传感器故障诊断系统。采用传递函数结合拟合法建立执行机构数学模型;提出了一种自校正学习速率的在线BP神经网络算法,并应用于发动机逆模型的建立。通过数字仿真表明诊断系统对于执行机构故障以及LVDT传感器故障能够实现准确的定位和诊断,同时具有较好的适应能力。针对气路传感器故障诊断问题,分别采用基于自适应模型和基于在线极端学习机的诊断系统。提出一种基于方程组求解的自适应模型建立方法,具有较高的精度和自适应能力,并应用于传感器故障诊断;采用在线极端学习机算法对各气路传感器建立智能映射模块形成诊断系统,并引入“预学习”思路,使得该方法对于传感器单故障、双故障均具有出色的诊断能力和较高的重构精度,同时具有较好的实时性和鲁棒性。在实现传感器故障诊断后,继续对容错控制开展研究。设计了LVDT传感器以及气路传感器故障隔离方案;采用增广LQR方法设计多组控制器,并提出一种扭矩指令模型,在此基础上采用控制回路切换的方式实现主动容错控制,提高控制系统稳定性。最后,针对执行机及其传感器故障诊断系统以及基于在线极端学习机的气路传感器故障诊断系统设计半物理仿真试验,验证了算法的有效性。
[Abstract]:The sensors and actuators in the FADEC system of the swirl shaft engine directly affect the control and the safe and stable operation of the whole engine.In order to ensure the reliability of the vortex shaft engine, the related research on actuator and its sensor fault diagnosis, gas path sensor fault diagnosis and fault tolerant control are carried out in this paper.In this paper, based on the nonlinear component level model of the turbo-shaft engine, the finite order method with the mechanism of calculating divergence determination and correction is used to replace the Newtonian one-pass method by improving the Broyden quasi-Newton method.Improve the real-time and precision of dynamic process calculation of component-level model.The actuator and its sensor fault diagnosis system based on the actuator model and the engine inverse model are constructed.An online BP neural network algorithm for self-tuning learning rate is proposed, which is applied to the establishment of the inverse model of the engine, and the mathematical model of the actuator is established by using the transfer function combined with the fitting method, and an on-line BP neural network algorithm with self-tuning learning rate is proposed.The results of digital simulation show that the diagnosis system can accurately locate and diagnose the actuator faults and LVDT sensor faults, and has a good adaptability at the same time.To solve the problem of gas path sensor fault diagnosis, the diagnosis system based on adaptive model and online extreme learning machine is adopted respectively.An adaptive model building method based on solving equations is proposed, which has high accuracy and adaptive ability, and is applied to sensor fault diagnosis.The on-line extreme learning machine algorithm is used to establish an intelligent mapping module for each gas path sensor to form a diagnosis system, and the idea of "pre-learning" is introduced to make the method for the single fault of the sensor.Both faults have excellent diagnosis ability, high reconstruction accuracy, good real-time performance and robustness.After the sensor fault diagnosis is realized, the research on fault tolerant control is carried out.The fault isolation scheme of LVDT sensor and gas sensor is designed, and the method of augmented LQR is used to design multi-group controller, and a torque instruction model is proposed. On the basis of this, active fault-tolerant control is realized by switching control loop.Improve the stability of the control system.Finally, a semi-physical simulation test is designed for the fault diagnosis system of the actuator and its sensors and the gas sensor fault diagnosis system based on the on-line extreme learning machine, which verifies the effectiveness of the algorithm.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：V263.6

【参考文献】