Process alarm log mining and modeling based on complex networks

To analyze the correlation between alarm logs generated during complex chemical processes, a mining and modeling method of process alarm logs is proposed in this paper. The proposed method mines the correlation between alarm logs based on the Word2Vec and Pearson correlation coefficient, and establishes a quantitative modeling based on the complex networks. Firstly, the alarm variables are mapped to real-valued vectors using the Word2Vec. Then, the association relationships among the alarm logs are analyzed using the Pearson correlation analysis, and the Pearson correlation coefficient matrix and its Boolean matrix are obtained. The Boolean matrix is set as the adjacency matrix, and complex networks can be quantitatively modelled. Finally, the effectiveness of the proposed method is verified by an actual case.