Research Of Network Fault Diagnosis Technology Based On Fuzzy Reasoning

If a node in the network breaks down, it may lead to alarms generated by other nodes. In this way, the alarm spread continually, and formed an alarm storm eventually.It is difficult to wipe off the correlations among them and locate the root alarms. In actual network, the boundary of alarm caused by fault contains a certain ambiguity.Alarm attribute is neither true nor false, but often in a numerical interval value. If the points near partition threshold are not handled properly, the effect of these points on the partition threshold may be over-emphasized or ignored. This has a negative effect on the accuracy of network fault diagnosis. In the other hand, usually there is no one-to-one relationship between the cause of the fault and fault phenomenon, but a fuzzy relationship. In this thesis, fuzzy theory and fuzzy reasoning technology are applied in network fault diagnosis. In order to locate the root alarm more quickly and more accurately, the thesis mainly contains following aspects.1) The association rule extracted from the knowledge base indicates the association relationship among the alarms, however, it does not reveal the causal relationship among them. In this thesis, a kind of prediction model constructed by an improved BP neural network is used to predict the direction,which alarm propagates in. An association rule used for fuzzy reasoning is inputted to the prediction model. Then, either FMP or FMT is used to calculate the degree of membership according to the result generated by prediction model.2) During the process of fault diagnosis, it may take much time to match the fuzzy alarm with the fuzzy association rule. Besides, a reasonable inference path is needed to avoid the phenomenon such as reasoning loop. All the situations above have a negative effect on the efficiency and accuracy of fault diagnosis.So the thesis deeply researches the strategies in fuzzy reasoning, such as fuzzy matching strategy, conflict resolution strategy, search strategy, driving strategy of fuzzy reasoning.3) Attempted to enhance the accuracy of fault diagnosis, a new fuzzy reasoning algorithm, a kind of weighted synthesis reasoning algorithm based on fuzzy similarity, is put forward in this thesis. Also an appropriate defuzzification method is applied to explain the fuzzy result.According to the simulation results get by the network fault diagnosis system based on fuzzy reasoning, we get the following conclusions. The prediction model used to predict the direction can make the system select FMP or FMT dynamically. It leads to the whole process of fault diagnosis back towards the root alarm, and makes the process controllable. Besides, there is less redundant information produced during the process of fault diagnosis. So the efficiency of fault diagnosis is improved. The fuzzy reasoning control module can effectively improve the matching efficiency between alarms and rules, It is benefit for constructing a reasonable reasoning path, avoiding the appearance of reasoning loop and solving the problem of matching conflict. The new fuzzy reasoning algorithm put forward in this thesis makes the result more accurate. Thus the whole system can locate the root alarm effectively and quickly.