A Classification Method for Abnormal Patterns of Complex Electromechanical System for Discriminant Analysis Nuclear Entropy

摘要

In order to solve the complexity and dynamic of massive data of complex electromechanical system in process industry, and to put up with fault data precisely and classify the fault types accurately and efficiently, this paper proposes a kernel entropy discriminant analysis method based on information entropy. Firstly, the concept of information entropy is introduced. Because of the redundancy of information, the redundancy degree is related to the uncertainty of information. The average information amount after excluding redundancy can eliminate the effective classification of abnormal patterns. Secondly, the kernel entropy component analysis is used to nonlinearly map and reduce the dimension of the data. The entropy-based parameter selection method is made to calculate steps and KEDA algorithm steps, so as to classify the space after dimension reduction. Finally, the effectiveness of the algorithm is verified by combining with the TE process dataset, and the effectiveness, superiority and rationality of the algorithm are verified by simulation experiments. The results show that KEDA method proposed in this paper has certain application value compared with other methods.