【作者】 林鹏；

【导师】 王聪；

【作者基本信息】 华南理工大学 ， 控制理论与控制工程， 2017， 博士

【摘要】 涡扇航空发动机是目前世界上军用和大型民用飞机常用的动力装置。轴流压气机作为涡扇航空发动机的核心部件之一,其工作负荷能力和稳定性对整个系统的工作效率和可靠性有着重要作用。对其失稳的机理分析、预测和控制是长期困扰涡扇航空发动机性能提升的技术瓶颈。随着现代涡扇航空发动机朝着更高单级压比和更少级数的发展,必然要求压气机叶片负荷越来越高,导致发动机工作裕度和抗干扰能力的下降。进气道出口流场对发动机正常运行的影响逐渐明显,压气机进口气流非均匀(进口畸变)已是诱发其内部气流流动失稳的重要因素之一,极大地限制了轴流压气机的性能及稳定运行范围。为了获取涡扇航空发动机压缩系统更高的可靠性和效率,借助于轴流压气机失稳预测技术的研究,在出现失稳现象前,能够采取有效的预警方案,实现在轴流压气机出现微小失稳征兆时就及时诊断,确保在满足稳定性要求的同时最大限度提升涡扇航空发动机的性能。此外,失稳预警的研究还可以为主动控制提供充足的反应时间,调整涡扇航空发动机进口畸变条件下的工作状态。因此,有必要在轴流压气机失稳的建模与预测方面进行深入研究,对进口畸变条件下轴流压气机内部气流流动失稳触发机理的认识,并对其进行有效地预警,对于提高压缩系统的抗畸变能力和拓宽轴流压气机的稳定工作范围,具有十分重要的理论价值与实际意义。为了深入理解进口畸变下轴流压气机的失稳机理及有效地对其预警,考虑压气机进口周向总压畸变情况,本文开展了以下几个方面的工作:1.进口畸变下轴流压气机失速机理的分析。考虑一类压气机前缘放置畸变片的周向总压畸变情况,提出了一个高阶畸变模型用于描述轴流压气机进口畸变下的旋转失速现象,特别是失速初始扰动的发展过程。首先,选取经典的局部流量二次型函数来定量地描述均匀气流通过畸变片的压力升。通过联立稳态气流下的压气机特性函数,构造一个特定畸变下压气机稳态特征函数来表达压气机内气流与畸变扰动气流间的耦合并表达整个系统的动态。其次,基于各组件的流动模型,采用Galerkin截取和周向空间离散化方法,推导了一个以状态空间形式的高阶畸变模型。该模型不仅能够描述已有模型或实验得到的进口畸变下旋转失速基本特性(轴向速度周向不一致,压气机性能下降以及失速提前发生),而且能够描述畸变下旋转失速初始扰动的发展过程。最后,基于Mansoux-C2压气机相应的参数,对18阶的畸变模型(相对应N=8阶失速模态)进行数值仿真来验证该高阶畸变模型的有效性。2.基于确定学习理论轴流压气机旋转失速预警的研究。针对周向总压畸变,提出了基于确定学习理论轴流压气机进口畸变条件下的失速预警方案,实现了在压气机出现微小失速征兆时给出预警信号。该失速预警方案主要是对复杂的非线性失速先兆动态进行辨识和识别的过程,通过对被测系统的失速先兆模式的识别给出失速预警信号。首先,采用确定学习理论,在未知建模动态存在的条件下,分别对系统失速前正常模式和失速先兆模式的内部动态进行局部准确建模,所获得的建模结果以常数权值形式的径向基函数(RBF)神经网络存储在失速模式库中。其次,基于失速模式库,采用动态模式识别方法构造动态估计器,获取待测系统的动态误差,并在L1范数度量下通过最小残差原理判断待测系统是否进入初始扰动状态,以此对待测系统提出预警信号。在失速先兆动态的辨识阶段,基于确定学习理论实现了对结构不确定复杂非线性系统的动态建模,使得该失速预警算法具有一定的鲁棒性。最后,通过数值仿真和低速轴流压气机试验台的畸变实验,对所提失速预警方案的有效性进行了理论和实验验证。3.短尺度突尖型失速预警的研究。基于确定学习理论,进一步地研究了轴流压气机畸变条件下突尖型失速预警。在多级高速发动机实际的运行中,短尺度的突尖型失速是较为普遍存在的一种流动失稳现象,特别是压气机进口畸变条件下能够诱发突尖型失速的发生。针对进口畸变诱发的突尖型失速时态数据,提出了基于数据的失速预警方案。该失速预警方案通过对包含外部扰动的微小且快速变化的失速动态进行RBF神经网络建模和动态模式快速识别,实现了在不同运行条件下对短尺度突尖型失速的提前预警,为多级高速压气机试验台的失速预警研究提供理论基础。最后,采用高阶畸变模型产生的突尖型失速数据,数值仿真验证了该基于数据的突尖型失速预警方法的有效性。更多还原

【Abstract】 Turbofan engine is the most common power device for military and civil aircraft in the world.As one of the core components of turbofan engine,axial flow compressor(axial compressor)has an important role in the working efficiency and reliability of the whole system.The demand for higher pressure ratios and lower weight in modern aircraft engine compressors tends to move the aerodynamic load of each stage to its limits.Unfortunately,this leads to decreased operating range and an unfavorable response to inlet flow nonuniformity(inlet distortion).It is well known in practice that inlet distortion is one of the most important causes resulting in rotating stall and has a serious impact on compressor performance.Compared with uniform inlet flow,which is an assumption in normal compressor design and experiment,inlet distortion increases the blade loading in the low mass flow zone,which,in turn,increases the incidence angle leading to stall precursors.That is,inlet distortion triggers the initial disturbances and induces the occurrence of rotating stall in advance.There is quite a lot differences in the transient process of stall inception with the presence of inlet distortion.Given the unavailability of a reliable stall warning system,modern highly loaded compressors are still designed to have a wide and conservative stall margin.Therefore,an efficient stall warning scheme for compressors with inlet distortion is of great significance to the reduction of stall margin.In additional,the stall warning could have significant practical benefit if the warning is sufficiently in advance of the stall so as to permit time for control system response.Rotating stall has to be avoided in any situation during engine operation.For this reason a detailed knowledge of the flow phenomena of the compressor in normal conditions as well as near the stability limit is essential.The purpose of this thesis is to investigate the detection of these different stall inception patterns based on Deterministic Learning(DL)algorithm.It is necessary to analyze in detail whether the DL algorithm is applicable to the modeling and early detection of rotating stall with inlet distortion,so that the proposed stall warning scheme can be applied more widely to different compressor operating conditions.The contributions of this thesis are as follows:1.Considering the practical limitation of the engine operation,a circumferential total pressure distortion at the compressor inlet is discussed.A high-order distortion model is proposed for analyzing the rotating stall inception process induced by inlet distortion in axial compressors.A distortion-generating screen in the compressor inlet is considered.By assuming a quadratic function for the local flow total pressure-drop,a new steady compressor characteristic under inlet distortion is constructed to describe the coupling between the compressor and the distortion screen,and then dominates the compression system dynamics.The high-order distortion model in a real-valued state-space form is derived based on a generalized harmonics expansion of the perturbation flow and a spatial discretization in the circumferential angle θ.Considering the high-order stall modes,the wave propagating around the compressor annulus has a rich harmonic structure for a compression system with inlet distortion,and then the high-order distortion model can more accurately depict the dynamics of rotating stall,especially the transient behavior of stall inception.2.Based on the high-order distortion model developed,a stall warning scheme employing DL algorithm is proposed for aircraft engines with inlet distortion.The high-order distortion model can capture the transient behavior of stall inception for a test rig and be suitable for the analysis of stall warning.In the identification phase,locally accurate approximations of the unknown dynamics along the known state trajectories are achieved from the high-order distortion model in the presence of unstructured uncertainty.The obtained knowledge of system dynamics is stored in constant radial basis function(RBF)networks.In the recognition phase,a bank of estimators is constructed using the stored constant RBF networks to represent the learning normal and stall inception patterns.By comparing each estimator with a test system,the average L1 norms of the residuals are taken as the measure of the dynamical differences between the test system and the learning patterns.The occurrence of stall inception as an early warning signal can be rapidly detected according to the smallest residual principle.Finally,numerical simulation and experiments results are given to show the effectiveness of stall warning approach.3.In general,spike-type stall inception is a prevalent type of fluid dynamic instability leading to rotating stall in high speed compressors.An effective spike-type stall warning approach for axial compressors with inlet distortion is proposed by using DL algorithm.Based on the temporal data sequences,locally-accurate approximation of the underlying the normal and spike-type stall inception dynamics is achieved.The occurrence of spike-type stall inception as an early warning signal can be rapidly detected according to the smallest residual principle.The spike-type stall warning approach can provide early detection of the short-time changes in stall inception and be applicable in different operational conditions.更多还原