Research On Hybrid Diagnosis And Self-Healing Based On Multi-Agent

At present, the automation operation of large scale complex systems, such as gridcomputing, cloud computing, aerospace and water-power engineering, has beenimproved constantly, the scale is widening and the related complexity is increasingrapidly.During the running processes of these large scale complex systems, it is inevitablethat the sensors, operation units and the internal devices of the system could fail. Oncethe fault cannot be detected and repaired in a timely manner, it may lead to the failureand paralysis of the whole system. It will not only result in the economic loss, evenendanger the personal safety. So, maintaining the reliability and stability of the systemis especially important for the favorable process of the large scale system. In fact, thecurrent status of the large scale system is just like that one person has just only onebrain that has to manage the whole complex body functions, including the heart beat,breathe, blood circulation, et al. However, human beings feel no burden actually. Afterresearch this bionics problem, it turns out that the key is that human beings have a set ofthorough autonomic nerves system which partakes the function of body selfmanagement.Based on the opinion of bionics, this thesis simulates the human beings’ autonomicnerves system and develops the framework of the implementation mechanism of theautonomous hybrid fault diagnosis and self-healing in the large scale complex system,integrating the characteristic of the large scale complex system. This thesis defines thedifferent functions of the neuron agent, the axon agent, the peripheral nerve agent andthe central nerve agent respectively in the bionic autonomic nervous system, proposesalgorithms for hybrid fault diagnosis and self-healing, and specifies the operatingmechanism of the system. Then, using the results oriented diagnosis and healingmechanism, the system decides the security level of the faults and infers the possibleimpact on the system based on the real time system symptoms. Furthermore, it decidesthe diagnosis process for the faults and heals the components which need repairsrespectively based on the healing strategies. Differing from the traditional processes, the proposed system has no need to test and locate the malfunction components and adoptsa novel cooperative mechanism. When the local\neuron agent cannot fulfill the faultsdiagnosis and healing tasks, the peripheral nerve agent will participate in thecooperative diagnose and the assist healing based on the collaboration andcollaboratively between different types of agents. These agents work togethercollaborate to maintain the stability and high efficiency of the running system. As aresult, the thesis provides a case which amplifies the whole procedures include thedecision of faults security level, the selection of the diagnosis process and the finalhealing implementation.At last, the availability and the high efficiency of the proposed hybrid faultdiagnosis and self-healing model based on multi-agent system are certified byexperiments.