New data mining and calibration approaches to the assessment of water treatment efficiency

  • Authors:

  • M. Bieroza;A. Baker;J. Bridgeman

  • Affiliations:

  • Centre for Sustainable Water Management, Lancaster Environment Centre, Lancaster, LA1 4YQ, United Kingdom;School of Civil and Environmental Engineering and School of Biology, Earth and Environmental Sciences, The University of New South Wales, 110 King St., Manly Vale, NSW 2093, Australia;School of Civil Engineering, University of Birmingham, Edgbaston, Birmingham, B15 2TT, United Kingdom

  • Venue:
  • Advances in Engineering Software
  • Year:
  • 2012

Quantified Score

Hi-index 0.01

Visualization

Abstract

For the first time, the application of different robust data mining techniques to the assessment of water treatment performance is considered. Principal components analysis (PCA), parallel factor analysis (PARAFAC), and a self-organizing map (SOM) were used in the analysis of multivariate data characterising organic matter (OM) removal at 16 water treatment works. Decomposed fluorescence data from PCA, PARAFAC and SOM were used as input to calibrate fluorescence data with OM concentrations using stepwise regression (SR), partial least squares (PLS), multiple linear regression (MLR), and neural network with back-propagation algorithm (BPNN). The best results were obtained with combined PARAFAC/PLS and SOM/BPNN. Both the numerical accuracy and feasibility of the adopted solutions were compared and recommendations on the use of the above techniques for fluorescence data analysis are presented.