Study of the flocculation kinetics of fish processing wastewater using a phenomenological model and machine learning techniques

• Autor(a) principal: Castamann , Gabriel
• Data de Publicação: 2022
• Outros Autores: Veit, Márcia Teresinha, Colombo, William Luis Reginatto, Palácio, Soraya Moreno, Gonçalves, Gilberto da Cunha, Barbieri, Jéssica Caroline Zanette
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• DOI: 10.33448/rsd-v11i11.33976
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/33976
• topic: Gradiente de velocidade; Cinética de floculação; Modelagem; Redes neurais artificiais.; Gradiente de velocidad; Cinética de floculación; Modelado; Redes neuronales artificiales.; Velocity gradient; Flocculation kinetics; Modeling; Artificial neural networks.

The coagulation/flocculation process is a widely used technique typically applied to solid-liquid separation for wastewater treatment, based on the principle of destabilization of colloidal particles in suspension, followed by the aggregation of these particles into structured flocs. In this process, the flocculation kinetics (velocity and time) plays a key role in the treatment performance, as it interferes with the flocs rupture and formation. Therefore, for the treatment of fish-processing wastewater, two coagulants (natural: Tanfloc SH®; inorganic: Ferric Chloride) were evaluated in the flocculation kinetics and the experimental data modeling was performed using a phenomenological and artificial neural networks (ANNs) model. For this purpose, different velocity gradients and slow-mixing times were used in jar test experiments for each coagulant,  and the aggregation (KA) and rupture (KB) coefficients of the formed flocs were determined. The most effective slow-mixing conditions (velocity and time) obtained for the effluent flocculation step were 16 s-1 and 20 min for the Tanfloc SH® coagulant and 24 s-1 and 30 min for the Ferric Chloride coagulant. The flocculation kinetic data were submitted to programming in ANNs using Python Software and to computational numerical iteration procedures using the Solver tool of the Microsoft Excel® program. Both models were able to adequately represent the flocculation kinetic experimental data, highlighting the ANNs as an alternative modeling tool to the mathematical models conventionally used.