Multivariate Data Mining for Estimating the Rate of Discoloration Material Accumulation in Drinking Water Systems

It has been demonstrated that particulate material accumulates as cohesive layers on pipeline surfaces in drinking water distribution systems (WDS) and that when mobilized this material can cause discoloration and other water quality issues. This paper investigates the factors involved in this accumulation rate from real world field data. A data-driven modelling approach is adopted, whereby two machine-learning methods are applied for multivariate data mining based on the observed phenomena. The results highlight bulk water iron concentration, pipe material and looped network areas as key descriptive parameters. Such understanding and expressions are important for pro-active network management.