基于网络的时滞系统分布式鲁棒故障检测方法研究

The project devotes in the modeling, analysis and design for network-based distributed fault detection of time-delay systems. For discrete linear time-delay systems and discrete nonlinear time-delay systems, considering network-induced delay and pocket dropouts, the more suitable mathematics model for distributed robust fault detection systems with Markov stochastic performance will be constructed. The consensus strategy will be applied in distributed estimation for sensor networks, and the coupling and coupled data between one sensor and its neighbors can be collaborated effectively. Then, the novel network-based distributed fault detection filter will be designed, and less conservative sufficient conditions for the obtained distributed robust fault detection dynamics be proposed. Simultaneously, in order to improve that the filter is as robust as possible to parameters uncertainties and disturbances, and as sensitive as possible to faults, the more appropriate H-infinity average consensus performance index and adaptive fuzzy threshold will be proposed. A systematically new distributed robust fault detection methodologies will be presented, and our developed theories will be employed to the fault detection problems of robot arms. The research result will perfect theoretical basis, and provide a theoretical basis and effective method for network-based distributed fault detection filter to detect the fault timely and accurately.

本课题研究基于网络的时滞系统分布式鲁棒故障检测的建模、分析与设计。分别针对离散线性时滞系统与离散非线性时滞系统，在网络诱导时滞和数据包丢失情况下，建立更恰当的马尔可夫随机特性的分布式鲁棒故障检测系统数学模型。将一致性策略应用到传感器网络的分布式估计中，有效协调传感器和其相邻传感器间的耦合信息，设计新的基于网络的分布式鲁棒故障检测滤波器，并给出保守性更小的分布式鲁棒故障检测系统存在的稳定性条件。提出一类更适合分布式鲁棒故障检测的H无穷平均一致性指标和模糊自适应阈值决策方法，提高分布式故障检测滤波器对外部扰动与不确定因素的鲁棒性及对故障的敏感性，同时降低漏报和误报率。形成一整套基于网络的时滞系统分布式鲁棒故障检测方法，并在机械臂系统中进行验证及应用。研究成果将进一步完善分布式鲁棒故障检测的理论基础，为基于网络的时滞系统分布式故障检测滤波器及时而准确的检测出故障提供了理论依据和行之有效的方法。

本课题针对网络环境下时滞系统的分布式控制、分布式滤波、鲁棒故障检测进行了深入的理论研究和实验分析。针对带有诱导时滞及数据包丢失的分布式传感器网络，将一致性策略应用于分布式传感器网络的估计中，研究基于一致性的时滞系统分布式鲁棒故障检测问题。针对一类带有量化和多包丢失的网络非线性控制系统，提出一种新的闭环故障检测策略，同时设计了控制器和故障检测滤波器，提高了检测性能和设计的自由度。有损传感器网络下，分别针对一类带有时变时滞的线性离散系统和不确定性的T-S模糊系统，分别提出一种新的基于事件触发机制的分布式H∞一致性滤波和分布式H∞故障检测策略。提高故障检测滤波器对外部扰动与不确定因素的鲁棒性及对故障的敏感性，同时降低漏报和误报率。针对传感器网络环境下，对时滞系统的研究结果进行了大量的仿真研究，仿真结果验证了所提出控制方法的有效性。初步形成一整套基于网络的时滞系统分布式鲁棒故障检测方法，完善了分布式鲁棒故障诊断的理论基础。