PCA in a Bayesian framework for fault detection

In this paper, we give an original representation of Principal Component Analysis (PCA) for fault detection. PCA with its corresponding quadratic test statistics are integrated under a particular case of Bayesian Networks (BNs) named Conditional Gaussian Network (CGN). The proposed network maps a new observation to an orthogonal space and gives probabilities on the state of the system even when some data in the sample test are missing. An illustrative example is given on a simple process.