Fault Detection and Identification Using Modified Bayesian Classification on PCA Subspace

Abstract

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

Statistical process monitoring of a multiphase flow facility

“…Particularly, the Tennessee Eastman process simulator (Ricker, 2014) has been widely used for the assessment and comparison of the performance of different algorithms due to its realistic level of complexity and the challenges attached to the fact that it is a highly non-linear system. The popularity of this particular benchmark case is demonstrated by the high number of researchers who have used it in the last Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/conengprac years to prove the validity of a large variety of approaches based on PCA (Fan, Qin & Wang, 2014;Lau, Ghosh, Hussain & Che Hassan, 2013;Wang & Li, 2012;Rato & Reis, 2013), PLS (Ma, Hu, Yan & Shi, 2012;Yi, Hehe & Hongbo, 2013), CVA (Stubbs, Zhang & Morris, 2012;Odiowei & Cao, 2010), Fisher Discriminant Analysis (FDA) (He, Yang & Yang, 2008), Latent Subspace Projection (LSP) (Mori & Yu, 2014), Sensitive Fault Analysis (SFA) (Jiang & Yan, 2012) or Bayesian Networks (Liu & Chen, 2009;Verron, Li & Tiplica, 2010) among others.…”

Monitoring of carbon dioxide capture processes

“…The state variable resemblance among data points can be based on different principles [9] [10]. In this work k-means clustering is used to group data points based on their Euclidean distance to the cluster centroid [11].…”

Nonstationary fault detection and diagnosis for multimode processes

“…The PCA subspace has to be adapted when monitoring a time‐varying process. The detailed strategies of monitoring a time‐varying process with multiple operating states can be found in the previous work 29. In this example, the subspace was individually updated by using the two case data for fair comparisons.…”

“…The pairwise Fisher discriminant analysis (FDA) was then applied to normal data and each class of fault data to find fault directions that were used to generate contribution plots for isolating faulty variables. Liu and Chen29 used Bayesian classification to extract multiple operating regions from historical data. A fault identification index was derived based on the differences between normal and abnormal cluster centers and covariances.…”