网络环境下基于滑模观测器的故障重构与容错设计方法研究

Fault reconstruction is one of the most important research problems in system control and fault detection. Different from fault detection approach based on the residual error, Fault reconstruction estimates the fault directly, and improves the detecting ability for slow-varying or initial faults. Traditional fault detection theory requires the assumption that the communication between plant and fault detection filter is real-time. However, this assumption does not hold in network-based aerospace and complex industrial processes. Although the introduction of network holds a lot of advantages, at the same time it results in communication delay, packet loss and quantization error etc, which brings new challenging problems for the traditional fault detection theory. The aim of this proposal is to investigate the network-based fault reconstruction and fault-tolerant control problems. We would like to develop new theory and approach to describe the non-uniform sampling, communication delay, packet loss and signal quantization. Based on the presented mathematical model, it is desirable to solve the network-based fault reconstruction and fault-tolerant design by the theories of sampling control, impulsive control, stochastic control and variable structure control. Some previous development has been achieved, and it is desirable to build a systemic theory and approach for the network-based fault reconstruction and fault-tolerant design.

故障重构是系统控制与故障诊断领域的重要问题之一。与基于残差的故障诊断方法不同，故障重构方法直接对故障估计，提高了对于慢变或初始故障的检测能力。传统的故障诊断理论建立在一个理想假设条件之上：物理系统与故障检测滤波器之间的通讯是实时通讯的。另一方面，在当代航空航天、复杂工业过程等领域，网络的引入为工程应用带来众多优点，同时也造成了数据传输延迟、数据丢包、量化误差等问题，给传统的故障诊断理论与技术提出了新的挑战。本项目旨在研究网络化控制系统的故障重构和容错问题。提出一种新的理论和方法，以统一的框架描述非周期采样、信号延时、数据包丢失、信号量化等诸多因素；利用采样系统、脉冲系统、随机控制理论和变结构控制理论分析与综合的方法解决网络化控制系统的故障重构问题；本项目前期研究已取得重要进展，其最终研究结果有望建立一套关于网络环境下故障重构和容错设计的较为完整的理论体系。

本项目主要研究复杂网络环境下的故障重构，容错控制等问题。主要针对网络环境诱导的非周期采样、信号传输迟滞、数据包丢失以及信号量化误差等诸多因素影响下的可靠性控制问题来展开相关研究。项目基本按照计划进行，在基础理论及应用方面都取得了一些成果，在国际期刊上共发表相关论文10余篇。具体内容表述如下：项目研究人员在前期开展的网络环境下传感器/执行器故障系统的建模问题的研究基础上，针对直升机姿态控制、汽车主动悬架系统等实际物理系统，研究了滑模观测器的设计及其故障重构问题；随后研究了基于事件触发的非线性网络化系统的滤波问题和复杂工业网络下的高价系统的模型降阶问题；最后研究了网络化控制系统的容错控制问题，并将相关研究成果应用于主动悬架不确定系统的可靠控制以及直升机容错编队控制问题中。