【作者】 刘敏；

【导师】 汪锐；

【作者基本信息】 大连理工大学 ， 航空宇航科学与技术， 2019， 硕士

【摘要】 作为涡扇发动机的核心系统,控制系统保证了发动机安全、可靠地工作,为飞机飞行提供推力。若控制系统发生故障,则很难达到期望的控制目标,甚至造成飞行事故。因此,为了减少维修费用,确保飞行的安全性,开展针对涡扇发动机控制系统的故障诊断与容错控制研究具有重要意义。本文以某型涡扇发动机为研究对象,采用拟合法建立小偏差模型,并通过伪逆法建立仿射参数依赖型的线性参数变化(Linear-Parameter-Varying,LPV)模型。当涡扇发动机控制系统发生故障时,需要快速、准确地进行诊断,并利用诊断出的故障信息进行容错控制,确保控制系统仍能正常工作,并满足一定的性能指标。为此,本文提出了一种基于线性分式变换(Linear Fractional Transformation,LFT)参数依赖的LPV故障估计器设计方法。首先,针对同时存在扰动、执行机构及传感器故障的涡扇发动机LPV模型,将LPV故障估计问题转化为基于LFT参数依赖的鲁棒H_?控制问题。然后,基于S-过程,给出了低保守性的故障估计器存在的充分条件。此外,提出了具有自适应调度特性的LPV故障估计器的设计算法,设计出的估计器参数能够随着LPV模型参数自适应变化,并且迅速检测和准确重构故障信号。针对存在扰动、乘性执行机构故障和加性传感器故障的涡扇发动机控制系统,本文设计了虚拟执行器的主动容错控制策略,对故障系统进行重构。该方法无需改变涡扇发动机的原控制器形式,通过重构系统可以保持控制系统稳定,并得到与无故障系统相近的控制效果,实现容错控制。此外,考虑到涡扇发动机由于气路部件性能退化而导致故障发生,本文基于机器学习算法建立了气路性能参数预测模型,对气路性能参数的变化趋势进行预测与分析,实现涡扇发动机气路部件的性能监视。首先,针对某型涡扇发动机启动时的地面试车数据,通过相关性分析选择预测模型的输入参数;然后,基于相空间重构技术,构建了输入-输出数据;最后,通过AdaBoost.RT_ELM算法,建立了气路性能参数预测模型,该预测模型能够实现对气路性能参数变化趋势的预测与分析,预测精度高,符合工程需求。更多还原

【Abstract】 As the core system of the turbofan aero-engine,the control system keeps the engine working safely and reliably which provides thrust for flight.If the fault of the control system occurs,it is difficult to obtain the anticipant control purpose,and even leads to flight accidents.Therefore,in order to reduce maintenance costs and improve flight safety,it is significant to study fault diagnosis and fault-tolerant control of turbofan aero-engine control system.In this paper,a turbofan aero-engine is taken as the research object.The small perturbation model is established by the linear fitting method,and the affine parameter-dependent linear-parameter-varying(LPV)model is derived by the matrix pseudo-inverse method.When the fault of the turbofan aero-engine control system occurs,the fault should be diagnosed quickly and accurately.Based on the fault information,the fault-tolerant control should be used to ensure the safety operation of control system and satisfy some specified performance.Therefore,the design algorithm of a LPV estimator which has linear fractional transformation(LFT)parameter dependency is proposed in this paper.Firstly,for the turbofan aero-engine LPV model with both actuator and sensor fault under disturbances,the LPV fault estimation problem is transformed into a robustH_?control problem with LFT parameter dependency.By using S-procedure,the sufficient condition for the existence of the fault estimator is proposed,which can lead to less conservative results.Moreover,the LPV fault estimator design algorithm with adaptive scheduling characteristics is presented,and the parameters of the designed estimator will change adaptively with the LPV model.The LPV fault estimator can detect fault rapidly and reconstruct fault accurately.For the turbofan aero-engine control system with actuator multiplicative fault and sensor additive fault under disturbances,an active fault-tolerant control strategy based on the virtual actuator is designed to reconstruct the fault system.By adopting this strategy,it is no need to redesign the original controller,then the stability of the control system is guaranteed,and the control effect is similar with the fault-free system.Therefore,the fault-tolerant control is achieved.Considering the turbofan aero-engine has fault due to the performance degradation of gas path components,a prediction model for gas path performance parameters based on machine learning algorithms is proposed in this paper to predict and analyze the variation trend of gas path parameters and realizes the performance monitoring of the gas path components.First,starting test data of the turbofan aero-engine,the correlation analysis is carried out to choose the input parameters of the prediction model.Then,the input-output samples are established with the phase space reconstruction.Finally,AdaBoost.RT_ELM algorithm is applied to establish the gas path performance parameters prediction model.The prediction model can predict and analyze the variation trend of gas path parameters,and has high accuracy which can meet the engineering requirements.更多还原