ISSN:
1934-2659
Quelle:
Berkeley Electronic Press Academic Journals
Thema:
Chemie und Pharmazie
,
Werkstoffwissenschaften, Fertigungsverfahren, Fertigung
Notizen:
The outputs of statistical process control (SPC) tools developed for fault detection are comparatively examined while applied to actual data collected in an industrial plant. The influence of added information gathered from the plant operation under different strategies is analyzed. Particularly, standard principal component analysis (PCA), kernel PCA and the Hotelling's T2 charts are inspected for a reported problem. The effect of training the tools either with an extended historic databank obtained under standard operation, or including also non-conventional conditions, is studied. The ability of the tools to provide a specific alarm and identify the responsible variable is examined by analyzing the contributions per variable to the SPE and the T2 statistics. In addition, the capacity of the tested tools to adapt to a new operation strategy is compared.
Materialart:
Digitale Medien
URL:
http://www.bepress.com/cppm/vol3/iss1/4
