一类模型不确定闭环非线性控制系统的故障诊断研究及应用验证

With the increasing of system complexity and uncertainty, more and more practical engineering plants are modeled as nonlinear closed-loop systems with model uncertainties. A fault in a closed-loop system may transfer from one point to another and the relationship between fault, input, and output are complex. Also, the modes of fault grow fast for high dimension systems. As a result, fault diagnosis for uncertain nonlinear closed-loop systems is becoming a challenging problem to be solved, which concerns with the issues of fault modeling, the robustness of fault detection, and the accuracy of fault estimation. The quantitative relationship between fault model and open-loop plant parameters will be established and, based on this, novel approaches to fault modeling and fault diagnosis will be developed for a class of uncertain nonlinear closed-loop systems. A strong tracking filter in Krein space will be developed to estimate fault of uncertain nonlinear closed-loop systems. To simultaneously deal with the problems of robustness of residual to uncertainty, the sensitivity to fault and the closed-loop system control performance, an integration of residual generator and feedback controller will be designed by applying model matching and multi-objective optimization methods. A new control performance based fault detection strategy will be explored to deal with the issues of incipient small fault detection. Computer simulations, physical platform experiments, as well as actual flight experiments will be carried out via autonomous aircrafts. Through this project, both theoretical breakthrough and technical improvement of fault diagnosis will be achieved and some of the results should provide strong support to guarantee the safety of practical closed-loop control engineering systems.

工程系统复杂性和不确定性快速增加，模型不确定性、非线性、多变量强耦合等显著呈现，闭环运行情况下控制系统故障诊断面临故障广泛传播与复杂演化以及故障模式快速增长等挑战性问题。本项目针对一类模型不确定闭环非线性控制系统故障诊断，系统地开展故障建模与故障诊断创新理论方法研究与应用验证。挖掘闭环运行情况下表征故障动态行为的特征向量，建立故障模型与开环被控过程参数之间映射关系；提出基于Krein空间的强跟踪故障估计方法，解决一类模型不确定性闭环非线性控制系统高精度故障估计问题；研究基于模型匹配的残差产生器与反馈控制器多目标优化设计，提出解决故障诊断鲁棒性以及鲁棒约束下故障检测正确性的方法；在基于闭环反馈控制器性能评估框架下，创立适用于闭环非线性控制系统微小故障检测的新方法。本项目将致力于取得国际领先的系统性故障诊断研究成果并在实际无人机实验平台上应用验证，为闭环工程系统安全运行提供理论依据和技术支撑。

本项目针对一类模型不确定闭环非线性控制系统故障诊断开展研究，力求给出故障建模与故障诊断的新理论、新技术和新方法，并面向无人机自主飞行控制系统开展仿真实验及应用验证。将机理分析、系统辨识、机器学习以及故障诊断理论相结合，挖掘闭环运行情况下表征故障动态行为的特征向量，建立了数据驱动的动态潜变量故障模型，给出了机理分析与频域辨识结合的无人机飞行控制系统模型，提出了机理建模与机器学习结合的故障建模新方法；给出了参数故障的多项式表征，提出了基于PWCS的闭环非线性控制系统故障可检测度分析方法。针对闭环非线性系统故障特征弱化、多源扰动影响以及非线性可行解难以求取的问题，将闭环非线性故障诊断问题归结为H∞故障估计问题，揭示了与Krein空间投影的对应关系，给了一类模型不确定性闭环非线性控制系统的高精度鲁棒故障估计新方法，并进一步应用于闭环控制系统故障估计与补偿的多目标集成设计。将故障自适应估计与贡献分析结合，基于模型匹配给出故障特征提取新方法，发展了不确定环境下闭环控制系统故障检测与分离理论。针对受范数有界未知输入影响的离散时变系统，将故障检测问题归结为未知输入最小能量估计问题，创立了适用于不确定环境闭环非线性控制系统鲁棒故障检测的新框架。提出了基于残差信息深度学习的无人机多传感器故障诊断方法，发展非线性系统微小故障快速诊断理论。本项目致力于取得国际领先的系统性的研究成果并在实际的无人机实验平台上进行验证，为提高我国无人机飞行的安全性提供理论支撑和技术保障。