Machine-learning based multi-objective optimization of helically coiled tube flocculators for water treatment

Abstract

The limitations of water resources and their importance for human life necessitates special attention to water purification processes. Here, the flocculation process, in which unstable particles aggregate, have been considered. In particular, we focused on helically coiled tube flocculators (HCTFs), which have certain advantages over conventional hydraulic flocculators. The simulation of the flocculation process in HCFTs has been done using the computational fluid dynamics (CFD) to study the effect of different geometrical and hydrodynamic variables on the performance of HCTFs. Small variations of velocity gradient, Gˆp95, and particle’s residence time, MI, are considered as the performance parameters. CFD results show that the design variables can change the performance parameters up to 60% and 40%, respectively. Using CFD for the training of machine-learning regression models, multi-objective optimization of HCTFs is performed. Our results show that a helical pipe with the diameter of 8.35 mm, curvature radius of 42.5 mm, helical pitch of 62.1 mm, and flow rate of 1.08 L min−1 provides the best performance among the range of input variables, where both performance parameters are minimum. Moreover, it was shown that the optimum results can be generalized to other scales with less than 2% difference in outputs by using equal non-dimensional groups.