基于性能监控的故障诊断与容错控制一体化设计方法研究

This project is dedicated to performance supervised fault diagnosis and fault-tolerant control of complex automatic control systems, which can be linear or nonlinear and consist of numerous feedback control loops. With ever increasing demands for high asset reliability, availability as well as safety in industrial processes, fault diagnosis and fault-tolerant control issues have received considerable attention in research and industrial application domains. The overall goal of this project is to develop efficient fault diagnosis and fault-tolerant control schemes in an integrated manner and as a part of online life cycle management of automatic control systems. .The first objective of our project is to propose a performance supervised fault diagnosis strategy, which promises higher fault diagnosis performance and builds the basis for a reinforcement learning based fault-tolerant control strategy. Developing reinforcement learning based fault-tolerant control schemes is the second objective of our work, in which numerous online maintenance actions are considered. Integrating the performance supervised fault diagnosis and fault-tolerant control algorithms into an online life cycle management system builds the further project objective. The online life cycle management system coordinates and optimizes the implementation of real-time maintenance actions. .From the methodological viewpoint, our work contributes to integrated application of model-based control theoretical methods and data-driven machine learning techniques. Moreover, it is also a project objective to investigate real-time ability of the proposed fault diagnosis and fault–tolerant control algorithms and to demonstrate their integration into the online life cycle management on a benchmark demonstrator.

本项目致力于研究在系统性能监控意义下复杂多回路自动控制系统的故障诊断和容错控制问题。随着对设备稳定性以及工业过程安全性等要求的不断提高，故障诊断和容错控制问题引起了学术和工业界的广泛关注。本项目的总体目标是提出高效的故障诊断和容错控制一体化设计方案，并将其作为自动控制系统在线生命周期管理的重要组成部分。.项目的首要目标是提出一种性能指标监督下的故障诊断策略，在保证高效故障诊断性能的基础上，为进一步提出基于强化学习的容错控制奠定基础。其次，多种在线维护措施将在基于强化学习的容错控制方法中被进一步研究。最终将提出的故障诊断和容错控制一体化设计方法结合到在线生命周期管理系统中，对实时维护措施进行在线协调和优化。.从方法论的角度来看，本项目致力于基于模型的控制理论和数据驱动的机器学习技术的综合应用。故障诊断与容错控制算法的实时计算能力，及其在生命管理系统中的应用和工业实现也是我们重要的研究目的。

本项目针对在系统性能监控意义下复杂多回路自动控制系统的故障诊断和容错控制问题，提出了高效的故障诊断和容错控制一体化设计方案，并将其作为自动控制系统在线生命周期管理的重要组成部分。首先，本项目提出了一种性能指标监督下的故障诊断策略和在线维护措施将在基于强化学习的容错控制方法，并将提出的故障诊断和容错控制一体化设计方法结合到在线生命周期管理系统中，对实时维护措施进行在线协调和优化。本项目着重研究了基于模型的控制理论和数据驱动的机器学习技术的综合应用。故障诊断与容错控制算法的实时计算能力，及其在生命管理系统中的应用和工业实现。