MI-NiDIA: A scalable framework for modeling flocculation kinetics and floc evolution in water treatment[Formula presented]

Abstract

This paper presents a scalable framework for modeling floc evolution and flocculation kinetics in water treatment. Unlike the existing methods that subjects Non-intrusive Dynamic Image Analysis (NiDIA) data to complex mathematical concepts, the proposed software devised a scaling concept for NiDIA data and designed an effective algorithm with the capability to predict varying floc lengths and the underlying kinetics under a broad flocculation conditions (Gf and Tf). Technically, the designed machine-intelligence framework (MI-NiDIA) involves data preprocessing, automatic parameter selection, validation and prediction of floc length evolution with metrics. For instance, MI-NiDIA-MLP recorded R2 of 0.95–1.0 for varying floc length at Gf60s−1.