The Application Of Fuzzy Reasoning In Network Fault Diagnose

With the development of Network currentlly, generally people’s lives and study area are increasingly dependent on computer network, as well as the social economy. However, due to the development of computer network, its scale constantly expanding rapidly and many hardware and software service does never stop increasing which is the reason the operation of the network is not always in the normal state. For such a huge complex network environment, once the network has the abnormal operation, a large amount of relevant alarms triggered by a variety of equipment produced by different manufacturers will appear. Besides, because of the correlation between these equipments, the alarm events might propagate through the network and give rise of the alarm storm. Above all, we should give enough emphasis on network fault management.Whether the network can run stably and safely completely depend on the network fault diagnosis performance. Besides, when the network is in an abnormal state, the ablity whether the network can quickly locate the fault and restore the normal operation is also an important criterion to verity the fault diagnosis performance. In the network fault diagnosis, by removing redundant alarms or get rid of the correlation between alarms, clear diagnosis information is presented to the managers instead of amorphous alarm data. Besides, traditional network fault diagnosis system does not consider the uncertainty of the relationship between the alarms and the fault. Instead, only use hard division method to solve that relationship.On account of the above problems, this article uses fuzzy association rule mining and fuzzy reasoning methods which are based on fuzzy theroy to complete the alarm correlation analysis and locate the root alarm.1. According to the characteristics of the network alarm, the paper puts forward a method to acquire the fuzzy membership degree of fuzzy reasoning based on fuzzy clustering. In this section, in one hand, it analyzes the process of acquiring effective alarm information from alarm information structure and the process to quantify each dimension of alarm information; In the other hand, after the quantization, it comprehensively consider the effect of the each alarm information factor using fuzzy clustering and improves the fuzzy clustering algorithm by minimum clustering volume ideology; At last, it gets the fuzzy membership degree distribution relative to the root alarm.2. After obtaining the fuzzy membership degree distribution, this section gives the fuzzy association rule mining algorithm based on traditional association rule mining algorithm and fuzzy theory. Then, it establishes the rule library for fuzzy reasoning.3. Based on the fuzzy implication operator and fuzzy synthetic decision, it gives a new control strategy of fuzzy reasoning process. In this part, we improve and construct the fuzzy implication operator with ‘xor’ relationship; Specific resolution of rule conflict and search strategy is given; weighting method is used to do the defuzzification of fuzzy conclusion. Besides, it verifies the efficiency and feasibility of the new reasoning algorithm from the aspects of theory and simulation analysis. The conclusion shows that the fuzzy reasoning algorithm based on fuzzy association rules can locate the network root alarm accurately and effectively.