Premium
Novel approaches for predicting efficiency in helically coiled tube flocculators using regression models and artificial neural networks
Author(s) -
Oliveira D. S.,
Teixeira E. C.,
Donadel C. B.
Publication year - 2020
Publication title -
water and environment journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.437
H-Index - 37
eISSN - 1747-6593
pISSN - 1747-6585
DOI - 10.1111/wej.12484
Subject(s) - mean squared error , artificial neural network , linear regression , coefficient of determination , mathematics , regression , regression analysis , computer science , statistics , artificial intelligence
In this paper, prediction models for turbidity removal efficiency (TRE) in helically coiled tube flocculators (HCTFs) are presented. The TRE was determined by physically modelling a compact, high‐performance and low detention time clarification system composed of a HCTF coupled to a decantation system. The values of hydrodynamic representative parameters of the flow were determined by CFD modelling. Eighty‐four different configurations of HCTFs were evaluated. Multiple linear/non‐linear regression and artificial neural network analyses were performed. A determination coefficient ( R 2 ) of 0.81 was obtained using multiple linear regression with the geometric and hydraulic parameters. In this model, the root mean squared error (RMSE) was 3.29%. Adding hydrodynamic parameters and using the artificial neural networks, R 2 reaches 0.96 and RMSE decay to 1.58%. These results indicate that the use of effective efficiency prediction models can be helpful in the design of new flocculation units and for the improvement of existing ones.