Design And Implementation Of Network Fault Diagnosis Expert System Knowledge Base

Network is becoming a much important role in the economic, social activities of human being. Once a fault occurred in the network, there is always a demand for the network manager to correctly and rapidly locate the fault source, in order to make a speedy recovery. However, with the fast scale expanding and new services'appearance, the modern communication network became so complicated that many alarms were triggered by only one fault. Therefore the alarm event correlation analysis and intelligent fault diagnosis is needed.The expert system (ES) is a crucial application field of artificial intelligence, and its theories and technologies is becoming mature. To make a solution to the bottleneck problem of the traditional ES of knowledge acquiring, our project is going to construct an ES based on data mining technologies. In this background, the knowledge base (KB), a key part of the ES, is facing a big challenge. The traditional knowledge representation methods become incompetent while the ES complexity and knowledge amount increases.This thesis introduces the object-oriented technology into the construction of the knowledge base in our project to overcome the difficulties of traditional knowledge base faces. Object-oriented technology has a series of merits, such as natural expression, data abstract and code reuse supported, and easy to maintain and expand. The major content of this thesis include:a) Construct network management information model using object-oriented method. The information of managed network's topology, network resources, and network traffics and network events are all fact knowledges in the knowledge base. Using object-oriented technology, this thesis makes the representation of these fact knowledges via defining several fragments'Entity-Relationship, on which a representation of the association rules is designed based.b) The inference mechanism of object-oriented knowledge base is designed in this thesis. Differing from traditional ES, the inference machine is no longer an independent module, but being encapsulated in the association rule object as a member function. Thus the rule and inference mechanism compose an object-oriented knowledge base system, and the process of inference is accomplished via message passage.c) Based on the object-oriented encapsulation of inference mechanism and considering the specialty of the executive of association rules, this thesis implements the forward and backward inference, and the correctness of the implement of the backward inference is briefly proved.