Fault diagnosis method based on the EWMA dynamic kernel principal component analysis

As widely used method for multivariate statistical process monitoring and fault diagnosis, the conventional principal component analysis (PCA) method is limited to the application of linear and time-invariant systems, and it can’t handle the sequence related question of the data. To handle the nonlinear and time-varying characteristics of the real processes, and the sequence related question of the data, a new monitoring and fault diagnosis method based on the EWMA dynamic kernel PCA (EKPCA) for nonlinear process is proposed in this paper. The simulation results for monitoring and fault diagnosis of three water tank system show the effectiveness of this method.