Statistical Process Monitoring with Machine Learning Kernel Methods: A Comprehensive Review

This paper explores the application of machine learning kernel methods in statistical process monitoring. Kernel methods are techniques that map data to high-dimensional spaces, enabling the linear separation of nonlinear relationships. In statistical process monitoring, these methods can be used to build monitoring models for anomaly detection and process control. The paper presents different types of kernel methods, including radial basis function kernel, polynomial kernel, linear kernel, and kernel principal component analysis, analyzing their performance and advantages across various applications. Additionally, the paper examines parameter selection and model optimization techniques for kernel methods, along with their comparison and integration with other statistical process monitoring methods. By investigating the strengths and limitations of kernel methods in statistical process monitoring, this paper offers new perspectives and approaches for industrial process monitoring.