关键词: 复杂网络, 故障诊断, 聚类算法, 人工免疫系统

Abstract: A new immune network clustering method with the antibody mutation strategy and the idiotypic network regulation strategy is proposed to improve the data compression ratio and the classification accuracy of the immune network clustering algorithm. Firstly, a novel idiotypic fuzzy immune recognition hypersphere is constructed to update the antibody network based on the Timmis’s Artificial Recognition Ball and the fuzzy theory. In addition, according to the modularity concept which can describe the community structure of the complex network effectively, a modularity-based clustering criterion function is designed, and a new algorithm based on the modularity merging for extracting the memory antibody is proposed to realize the self-adaptive compression for the antibody network. Secondly, a new fuzzy immune network clustering algorithm is developed, which combines the artificial immune network clustering strategy with the modularity-based criterion function. The validity of the method is verified with the UCI data sets. Results show that the proposed method can obtain the reasonable memory antibody network and improve the data compression ratio and the classification accuracy. Finally, the method is applied to the fault diagnosis of a four-stage reciprocal compressor. Compared with the aiNet and fuzzy artificial immune network classification method (FAINC), the proposed method is superior in the classification accuracy and the compression ratio.

Key words: Artificial immune system, Clustering algorithms, Complex network, Fault diagnosis