Process structure change detection by eigenvalue-based method

This paper proposes a novel eigenvalue-based method for process structure change detection. Based on this method, the number of components can be automatically determined, and the noise variance can also be estimated under limited data samples. Different from traditional methods, eigenvalues of the sampled data covariance matrix are used for structure change detection. Due to the difficulty in modeling the distribution of the calculated eigenvalues, the well-known one-class classification approach: support vector data description (SVDD) method is employed. To test the performance of the proposed method, two case studies are provided.