Principal component analysis and quantitative image analysis to predict effects...

• Author(s): Costa, J. C. ; Alves, M. M. ; Ferreira, E. C.
• Date: 2009
• Persistent ID: http://hdl.handle.net/1822/9137
• Origin: RepositóriUM - Universidade do Minho
• Subject(s): Detergent; Principal component analysis; Quantitative image analysis; Solvent; Toxic shock load
• Document Type: Article
• Language: English

Description

Principal component analysis (PCA) was applied to datasets gathering morphological, physiological and reactor performance information, from three toxic shock loads (SL1 – 1.6 mgdetergent/L; SL2 – 3.1 mgdetergent/L; SL3 – 40 mgsolvent/L) applied in an expanded granular sludge bed (EGSB) reactor. The PCA allowed the visualization of the main effects caused by the toxics, by clustering the samples according to its operational phase, exposure or recovery. The aim was to investigate the variables or group of variables that mostly contribute for the early detection of operational problems. The morphological parameters showed to be sensitive enough to detect the operational problems even before the COD removal efficiency decreased. As observed by the high loadings in the plane defined by the first and second principal components. PCA defined a new latent variable t[1], gathering the most relevant variability in dataset, that showed an immediate variation after the toxics were fed to the reactors. t[1] varied 262%, 254% and 80%, respectively, in SL1, SL2 and SL3. The high loadings/weights of the morphological parameters associated with this new variable express its influence in shock load monitoring and control, and consequently in operational problems recognition.