Open AccessResearch on Kernel-Distance-Based AEWMA-t Control Method
Author(s) -
Yatang Yang,
Fumin Chen,
Yiyu Tian
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1550/6/062014
Subject(s) - hypersphere , kernel (algebra) , mathematics , variable kernel density estimation , kernel principal component analysis , multivariate statistics , statistics , process (computing) , multivariate normal distribution , kernel method , computer science , artificial intelligence , support vector machine , combinatorics , operating system
Aimed at the complex data correlation among multivariate quality characteristics of products, inability to obey the assumptions of traditional control methods, low data volume in small and medium batch production, and inaccurate parameter estimation in process control, an AEWMA-t control method based on kernel distance data is proposed. Firstly, based on the support vector data description algorithm, the hypersphere is trained by normal samples. Then the kernel distance from the sample to the center of the hypersphere is calculated and the kernel distance is normally converted. Finally, the process is controlled with AEWMA-t control method. Case analysis shows that, compared with the traditional multivariate control methods, this method has good ability to detect anomalies in the mean deviation interval of each process, and is not restricted by the distribution and size of the data samples.