Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
5132258 | Chemometrics and Intelligent Laboratory Systems | 2017 | 11 Pages |
â¢New procedure to design a fault diagnosis system in a complex plant.â¢Using minimum number of classifiers in multiblock techniques.â¢The performance of diagnostic system is maintained.â¢The procedure can be completely automated.
This paper presents a procedure proposed for the multiblock-based fault diagnosis in complex plants using fewer classifiers, while keeping the best performance indexes. Such proposal can be completely automated so algorithms to this aim are also included. In order to prove its feasibility, this procedure has been applied to the Tennessee Eastman Process test problem using classifiers based on the Maximum a Posteriori Probability (MAP), k-Nearest Neighbors (kNN), Artificial Neural Networks (ANN) and Support Vector Machines (SVM).