Fault Analysis And Detection Of Complex Dynamical Networks

With the rapid development of science and technology such as automation and intelligence, the functions of modern systems become more and more powerful; meanwhile, size and complexity of these systems increase exponentially. Moreover, major accidents may occur once these systems break down. For example, a glitch may trigger cascade failure in power grid, leading to blackout and even the collapse of the whole network. Therefore, constructing efficient fault propagation model and fault detection method is the premise and guarantee of these large-scale complex systems, as well as the hot research area of complex network. Based on the research products, this paper investigates the condition of cascade failure occurring and relative epidemic progress in-depth analysis, specially focuses on fault detection and fault-tolerant control methods. In summary, the main contributions of this paper are as follows:For cascade failure caused by random attack in flow network, modeling and analysis methods of cascade failure is discussed. By analyzing the effect of initial attack on damage, probabilistic cascade failure model is established, and the critical attack size ensuring no collapse of network is derived. Finally, the influence of initial load and tolerance parameter on the resilience of network is investigated by simulation, and the optimal distributions are found.For multi-agent network suffered from actuator or sensor fault, fault detection of multi-agent network is studied. By constructing a consensus-based estimator and residual generator, each agent equipped with a virtual subsystem is derived, as well the distributed fault detection scheme. This scheme can not only detect all the faulty agents but also ensure the normal agents complete consensus task. Finally, numerical example is simulated to demonstrate the correctness and effectiveness of the proposed scheme.For consensus of multi-agent network with actuator fault, fault-tolerant control of multi-agent network is investigated. By structuring distributed state observer, relative fault detection algorithm is derived. Combining linear-parameterized adaptive method, the distributed fault-tolerant consensus protocol is presented, which ensures multi-agent network achieves consensus even under faulty environment.