【作者】 王健康；

【导师】 黄向华；

【作者基本信息】 南京航空航天大学 ， 航空宇航推进理论与工程， 2013， 博士

【摘要】 随着对航空发动机性能要求的不断提高，航空发动机控制系统已由传统独立于飞机控制系统的设计方法，向飞行/推进系统综合控制方向发展。航空发动机模型基优化控制作为飞行/推进系统综合控制中的重要内容，可在基本不增加机载硬件的条件下，仅通过控制算法充分挖掘发动机某项或综合性能潜力，提升飞机的飞行品质。本文以航空发动机模型基优化控制为主线，重点对固定翼飞机/发动机综合控制与直升机/涡轴发动机综合控制开展研究，具体包括机载发动机模型研究、综合优化控制、高稳定性控制、综合抗扰控制等内容。论文首先对固定翼飞机/发动机综合优化控制进行研究。提出了基于相似理论建立线性与非线性相结合的发动机机载复合稳态模型，并将发动机部件蜕化因素纳入考虑，结合改进Kalman滤波器反映发动机性能蜕化时的真实状态。进而，提出了机载复合模型与FSQP算法相结合的发动机多模式性能优化控制，包括最大推力、最小油耗、最低涡轮温度等控制模式，并分别应用于飞机平飞加速、爬升、巡航、大马赫数飞行等飞行任务中，通过数字仿真验证了优化方案的有效性。其次，论文研究了超机动飞行时发动机的高稳定性控制。建立了基于特征选择及在线攻角预测的发动机喘振裕度估计模型，并基于该模型设计了发动机直接喘振裕度控制与发动机进气畸变预测补偿控制两种高稳定性控制方法，通过实时调整发动机工作点保证发动机的稳定性。其中，前者改变了传统基于传感器的控制形态，将喘振裕度作为被控量直接加入控制回路，取得了更为高效的控制效果，后者则以预测补偿为中心，在发动机原控制回路不变的基础上进行压比损失补偿控制，使得控制体系的实现更为简单。再者，对于直升机/涡轴发动机综合优化控制，论文采用数据驱动的方法建立了涡轴发动机简化模型与蜕化估计模块作为涡轴发动机机载模型，前者是基于包线划分设计小包线的神经网络映射模块，后者则通过设计先进的MRR-LSSVR算法离线训练，并以多输出特征选择算法筛选最具价值的模型输入来提升估计精度；进而基于上述机载模型设计了最小油耗、串行优化最小油耗、最大功率、最低涡轮温度等直/发综合优化控制模式，在涡轴发动机额定/非额定状态下均实现了优化精度与优化实时性兼优的良好效果。最后，针对直升机/涡轴发动机综合抗扰控制，以旋翼扭矩为关联量，引入预测控制技术，通过设计滚动优化机制克服由旋翼扭矩测量滞后、发动机动态响应、数据传输等引起的时滞效应，改善系统的抗扰性能。论文先后提出了基于离线预测模型的涡轴发动机非线性模型预测控制与基于在线预测模型的串级+NMPC混合预测控制两种方法，前者的预测模型通过MRR-LSSVR离线算法设计，具有较好的泛化能力，后者的预测模型通过OSP-LSSVR在线算法设计，可根据发动机的实时信息实现模型在线自我更新，两种方法较串级控制均实现了更优的抗扰控制效果，尤其后者具有更好的动静态控制品质、响应速度及鲁棒性。更多还原

【Abstract】 Traditionally, the aero-engine control system is designed to operate independently of aircraftflight control system. This design philosophy often produces a system where performance is compro-mised greatly to ensure operability and simplicity. Now, by the aid of advances in control, estimationand system modeling techniques, integrated flight/propulsion control (IFPC) has become an inevitabletrend, which can improve overall aircraft performance. As the heart of IFPC, model-based optimalcontrol enables an aero-engine to achieve its full potential using only control algorithms, withoutadding onboard hardware. This paper takes model-based optimal control as the main line, focusing onboth integrated control for fixed-wing aircraft/turbofan engine system and integrated control for hel-icopter/turbo-shaft engine system. In detail, some techniques of onboard engine modeling, perfor-mance seeking control (PSC), high stability engine control (HISTEC) and disturbance rejection con-trol (DRC) are designed respectively, and moreover, some satisfactory results are obtained.In order to improve real-time capability of PSC for fixed-wing aircraft/turbofan engine system, asteady-state hybrid engine model is proposed based on similarity theory in the whole flight envelope,which takes engine’s working under non-nominal condition into account. An estimation module ofturbofan engine performance deterioration, meanwhile, is developed using improved Kalman filter.After combining the hybrid engine model with the estimation module, an onboard simplified enginemodel is built for PSC. With this model and feasible sequential quadratic programming (FSQP) algo-rithm, PSC for fixed-wing aircraft/turbofan engine system is designed and performed, including themaximum thrust mode, the minimum fuel-consumption mode and the minimum turbine temperaturemode. At last, these PSC modes are applied to various flying missions of aircraft, such as climbing,speeding up, cruising and so on. And simulation results demonstrate that flight performance can beimproved obviously with good real-time ability of optimizing process.As to HISTEC for super-maneuvering flight mission, two control schemes are put forward basedon engine surge margin (Sm) estimation model, viz. direct Smcontrol and engine inlet distortion com-pensation control. The modeling process consists of two parts: Smbenchmark value model under rou-tine flight condition and Smloss value model under super-maneuvering flight condition. The former isdeveloped using nonlinear fitting method, and its input is fixed by Smfeature selection algorithmbased on least square support vector regression (LSSVR). The latter is obtained utilizing an attackangel predictive model which is established with online sliding parsimonious LSSVR (OSP-LSSVR).Direct Smcontrol for engine, which is different from conventional control scheme, puts Sminto engineclosed-loop and takes Smas controlled variable directly. In this way, direct Smcontrol can exploit po-tential performance of engine more effectively with the decrease of engine stability margin. As wellas direct Smcontrol, engine inlet distortion compensation control also can achieve a good control level based on the prediction and compensation of engine inlet distortion. To be more important, engineinlet distortion compensation control is able to correct turbine expansion ratio command on engineoriginal closed-loop, which can be realized easier than direct Smcontrol.PSC for helicopter/turbo-shaft engine system is another important research topic. As the same asPSC for fixed-wing aircraft/turbofan engine system, it is necessary to design an onboard simplifiedturbo-shaft engine model with adaptive ability. In this paper, both the simplified engine model and theestimation module of engine performance deterioration are put forward with data-based method. Thesimplified engine model is performed using block partition of flight envelop, and engine mappingmodule in every block is trained by BP Neural Network; the estimation module of engine perfor-mance deterioration is trained offline using muti-input muti-output recursive reduced LSSVR(MRR-LSSVR), and the input of model is determined by feature selection algorithm. Considering thecomprehensive actions of multi-output variables to select support vectors, MRR-LSSVR is proposedto solve multi-output problems, and holds better sparseness and generalization performance thancommon algorithms due to combining reduced technique with iterative strategy. On the basis of thesimplified engine model and the estimation module, PSC for helicopter/turbo-shaft engine system isdesigned with FSQP algorithm, including the minimum fuel-consumption mode, the cascaded mini-mum fuel-consumption mode, the maximum power mode and the minimum turbine temperature mode.At last, the above control modes are carried out respectively based on UH-60A helicopter/T700tur-bo-shaft engine simulation platform, and simulation results show that the proposed optimizationscheme is feasible and valid.Finally, constrained nonlinear model predictive control (NMPC) is applied to the engine controlsystem to improve dynamic disturbance rejection ability of integrated aircraft/turbo-shaft engine sys-tem. Taking the torque of helicopter rotor as the related variable, NMPC utilizes a predictive modeland a rolling optimizer to solve the time-lag effect caused by measuring lag of rotor torque, dynamicresponse of engine, data transmission and so on. In this essay, two kinds of NMPC schemes for tur-bo-shaft engine are put forward, i.e. NMPC based on offline predictive model and cascade+NMPChybrid predictive control based on online predictive model. As for the former, the predictive enginemodel is trained by adopting MRR-LSSVR algorithm, and the rolling optimizer is realized using SQPalgorithm libray. In comparison with cascade controller, NMPC based on offline predictive model candecrease the droop or overshoot of rotor speed remarkably during helicopter’s maneuver flight. Butfor the latter, OSP-LSSVR algorithm is used for the predictive engine model, and gives the model aself-renewal capacity and higher accuracy. Cascade+NMPC hybrid predictive control, therefore, hasbetter dynamic disturbance rejection ability and robustness in a large flight envelop, and enables thehelicopter to exhibit much greater maneuverability than the conventional control method.更多还原