The Application Of Immune Theory In The Network Multiple Faults Diagnosis

The rapid development of communication technology and computer network technology makes the fault more complex, diverse, however, traditional diagnostic methods or intelligent diagnostic technology usually only diagnose a single type of fault and device, and they can’t meet the demand of present network. In order to solve network multiple faults, we should further study relevant intelligence technology on basis of theory and technology which has been researched to be mature in the field of single fault diagnosis to come true the main target of an efficient and accurate diagnostic system in network fault diagnosis fields.On the basis of understanding immune mechanisms of the biological immune system, we further study negative selection algorithm in the artificial immune system, coming up with a network fault diagnosis framework based on multi-agent technology for the current stage of network development. The framework consists of two agents, the central immune diagnostic agent and local diagnostic agent. The role of the central immune agent is to view, manage and make decisions for fault directives. Local diagnosis agent is composed of four different function modules. When the local diagnostic agents perceive a request from the outside world, it starts the appropriate diagnostic services to schedule, manage and collaborate. After handshaking, network information acquisition and processing module collects and processes local data to be perceived as detected data. The control and management module gives itself and detected data to the fault diagnosis module, and sends results to the fault response module. Fault response module records and transmits results to central immune agents by scheduling and handshaking.In this article, multiple fault diagnosis function is achieved by BP neural network and the combination rule of evidence theory in the fault diagnosis module. Gaussian artificial immune system is used to solve weights and bias of the BP neural network to ensure antibody diversity and improve the recognition accuracy of the BP neural network. The results of BP neural network output node are standardized, and used as the basic probability assignment of each type of fault. Then the basic reliability allocation or belief function is calculated based the combinational rule after combining fault types. In the end, the ultimate judgment of fault is achieved by the determining condition.The dynamic turned polling method is applied in the process of this experiment to collect network state information, in this way it can collect more accurate network status, and does not take up too much network bandwidth. The threshold of negative selection algorithm and evidence theory is analyzed and researched. After experimental verification, the framework of this paper for the diagnosis of multiple network faults is feasible.