Using data mining to predict sludge and filamentous microorganism sedimentation

Abstract

This study attempted to develop statistical regression models for predicting the settleability of activated sludge based on the quality of incoming sewage and on the identified dominant filamentous species. As part of the analyses conducted for the purpose, classification models are presented that enable identification of the respective filamentous microorganisms, based on the working parameters of the bioreactor and the quality of the influent. The study calculations demonstrated that the modeling methods based on artificial neural networks, random forests, and boost trees can be applied for the identification of filamentous microorganisms Microthrix parvicella, Nostocoida sp., and Thiotrix sp. in activated sludge chambers in the STP located in Sitkówka-Nowiny. The best predictive capacity, covering identification of the above-mentioned filamentous bacterial species in activated sludge chambers, was observed for statistical models obtained by the random forest method.