基于记忆调度优化的鲁棒故障估计与闭环补偿容错控制方法研究

The model mismatch factors deeply affect the practicability of active fault tolerant control. Therefore, the study of reliable fault tolerant control for mismatch models has important theoretical significance and high engineering value. This project is centered on memory scheduling strategy to explore the theory and method of using finite data to improve model-based fault estimation and closed-loop compensation control performances. The main research contents include: polyhedron modeling and vertex discrimination techniques for mismatch models; unbalance sequential data processing techniques for memory scheduling applications; fault estimation and closed-loop compensation control methods based on switching memory scheduling. The salient features of the study are summarized as three aspects. First, a novel active fault tolerant control strategy is proposed based on memory scheduling method. Such memory scheduling manner can overcome the defect of traditional fault-tolerant design method, which is that the system dynamic knowledge and data information are not comprehensively utilized. From this respect, the proposed memory scheduling fault tolerant control method can enhance the accuracy of fault estimation and the robustness of closed-loop compensation control. Second, the methods of polyhedron modeling and data completeness processing for memory scheduling are further proposed to solve the problem that model mismatch and unbalanced data influence the effectiveness of memory scheduling. And these designs are helpful to improve the practical reliability of fault-tolerant control. Third, several mechanisms of switching memory scheduling based on performance trigger are proposed to realize the proper and conditional use of data, which can improve the data utilization benefit and finally enhance the regulation capability of active fault tolerant control.

模型不匹配因素深刻影响着主动容错控制的实用性，因此面向不匹配模型的可靠容错控制研究具有重要的理论意义和很高的工程价值。本项目以记忆调度思想为核心，探究利用有限数据改善基于模型故障估计以及闭环补偿控制性能的理论与方法。主要研究内容包括：面向不匹配模型的多面体建模以及顶点判别技术；面向记忆调度的不平衡时序数据处理技术；基于切换记忆调度的故障估计与闭环补偿控制方法。研究的显著特点是：提出基于记忆调度方法的新颖主动容错控制策略，解决了单一容错设计方法未综合利用系统动态知识和数据信息的缺陷，有助于提高故障估计的准确性和闭环补偿控制的鲁棒性；提出面向记忆调度的多面体建模和数据完备性处理方法，分别解决了模型不匹配和数据不平衡影响记忆调度有效性的问题，有助于提高容错控制的实用可靠性；提出基于性能触发的切换记忆调度机制，实现了数据的适时且视情调用，有助于提高数据利用效益以及提升主动容错控制的调控能力。

本项目针对复杂工业系统和关键设备在容错控制设计时面临的模型不匹配这一普遍存在的问题，构建了基于记忆调度的多面体依赖故障估计和闭环补偿控制框架，研究了相应的多面体区间动态建模、时序数据表征与重构、记忆调度、性能触发、广义估计参数化、双模预测调节、快速插值优化等一系列方法，综合提高了故障估计的准确性、补偿控制的可靠性以及闭环优化的可行性。主要成果包括：基于参数时变模型建立了不匹配多面体系统的一般表征形式，并引入未知输入估计理论与预见控制策略实现了快速故障估计以及有限预测数据调度容错控制；基于马尔科夫参数序列辨识技术实现了面向记忆调度的多面体建模和数据完备性处理，解决了模型不匹配和数据不平衡影响记忆调度效率的问题，提高了故障估计与容错控制的实用性；基于不变集理论构造了不匹配模型的最小容错容许域以及最大容错容许域，容许域面积的大小直观地刻画了容错控制的容错能力；引入集不变原理、双模预测、插值分解等技术并构造了基于事件触发的切换约束容错优化机制，实现了预测数据的适时且视情调用，提高了闭环容错控制的鲁棒性以及可行性；搭建了用于验证容错控制方法有效性的电路实验平台，实现了面向算法的硬件在环快速原型设计。总体而言，研究成果从理论意义与工程价值角度分别针对不匹配模型的可靠容错控制问题给出了具体解决方法，为生产过程的安全运行提供了一些基础理论和关键技术的参考。