Fault Detection and Identification Using Modified Bayesian Classification on PCA Subspace

摘要: A novel process monitoring method based on modified Bayesian classification PCA subspace is proposed. Fault detection and identification are the major steps to diagnose root causes of a fault. However, before faulty variables from abnormal operations identified, different operating states need be clustered historical data. The proposed approach modifies cluster data into groups. Therefore, new fault index derived center covariance. An industrial compressor used demonstrate effectiveness approach. In example, process-insight-based were monitored along with measured variables. capability diagnosis has been improved, since indices directly related characteristics.