基于区间估计的不确定动态系统故障诊断与重构方法研究

Fault diagnosis and reconstruction are of great significance for ensuring safe operation of systems. Owing to the restricted conditions of designing traditional observers, such as the matching condition and the nonlinear styles, there exists too much conservatism by using traditional observers. Due to the neglect of the influence of the disturbance information, point estimation often lead to inexact diagnosis results. The precondition of interval estimation is relatively loose, and interval estimation can reflect the influence of disturbance on estimated results. Therefore, by using interval estimation, this project will carry out the research on fault diagnosis and reconstruction of a kind of uncertain dynamic systems: Firstly, a residual system which is sensitive to faults will be constructed to diagnose faults. The factors of the width of interval estimation will be clear to get low rates of false positives and false negatives. Then, the methods of decoupling actuator fault will be proposed, and the actuator fault algebraic interval reconstruction can be given. Class-proportional integral interval observer will be proposed to estimate system states and actuator fault simultaneously. Finally, with the help of the extended system states, the system with sensor fault will be transformed into a descriptor system to fulfill the sensor fault interval reconstruction. The key scientific problems will be solved: the building of a correlation model between system transformation and the width of interval estimation; the construction of a class-proportional integral interval observer. Some innovative results will be achieved, and some high level research papers will be published. The results of this project will overcome some weaknesses of the traditional observer to some extent, it is hoped to improve the performance of fault diagnosis.

故障诊断与重构对于保障系统安全运行具有重要意义。传统观测器设计因受匹配条件和非线性等因素影响，使得故障诊断与重构需较强的保守条件；据此得到的点估计结果无法如实反映不确定性的影响，常导致故障诊断结果出现偏差。区间估计方法设计前提相对宽松，可以较全面包含不确定性的信息。为此，本项目拟针对一类不确定动态系统，借助于区间估计开展故障诊断与重构研究：构造对故障敏感的残差系统，分析影响估计区间的因素，降低故障诊断漏报率和误报率；提出解耦执行器故障的方法，建立执行器故障代数区间重构，构造对系统状态和执行器故障同时估计的区间观测器；借助于系统状态扩展，将系统转化为广义系统，实现传感器故障区间重构。重点解决所遇到的关键科学问题：建立系统变换影响区间估计的关联模型，构造类-比例积分区间观测器。取得一些创新性研究成果，发表一些高水平论文，以弥补现有传统观测器设计方法存在的一些不足，有望进一步改善故障诊断效果。

项目主要针对不确定动态系统进行了状态区间估计方法研究，利用状态区间估计结果实现故障诊断，并研究故障的区间重构方法。相关研究考虑了多种不确定性情况，成果集中在状态估计、故障检测、故障重构等故障诊断领域，同时对切换系统、多智能体系统对象进行了初步研究。具体实现了：（1）利用区间估计结果构造了系统输出的合理区间范围，用实际输出是否处于该区间进行故障检测，对区间观测器和可达集分析得到的区间进行模型关联对比研究，初步分析了影响故障漏报率和误报率的因素；（2）分别基于区间观测器和可达集分析方法，对执行器故障进行了区间重构，基于比例积分点估计观测器和可达集分析方法，设计了可以直接进行执行器故障区间估计的类-PIO观测器；（3）基于状态扩展和Zonotopic Kalman滤波器方法，构造了广义系统模型，实现了对传感器故障的估计。项目执行过程中，研究人员围绕设定的研究内容，瞄准研究目标，提出了一系列的创新理论和方法，共撰写了研究论文29篇，已发表学术论文22篇（SCI检索论文8篇，EI检索论文7篇），取得的创新性研究成果，提升了故障诊断效果，扩大了故障诊断范围，完善了故障诊断理论，实现了故障诊断技术储备。