An adaptive principal component analysis applied to sets of data provided by global analytical methods (UV-visible spectra, buffer capacity curves, respirometric tests) is proposed as a generic procedure for on-line and fast characterization of wastewater. The data-mining procedure is able to deal with a large amount of information, takes into account the normal variations of wastewater composition related to human activity, and enables a rapid detection of abnormal situations such as the presence of toxic substances by comparison of the actual wastewater state with a continuously updated reference. The procedure has been validated on municipal wastewater.
- principal components analysis
- © IWA Publishing 2003