Computational fluid dynamics and artificial neural network based modeling of microplastics seperation using hydrocyclone

Microplastics represent a considerable environmental hazard, particularly to aquatic ecosystems. These particles can adversely impact marine organisms and infiltrate the food chain, potentially posing risks to human health. Due to their small size and ubiquitous distribution, the removal of microplastics from water presents a significant technical challenge. This study focuses on the separation of microplastics using hydrocyclones. A three-dimensional Eulerian–Eulerian computational fluid dynamics (CFD) model was used to investigate this process. The model showed strong agreement with experimental data in terms of pressure drop, water split, and axial velocity, confirming the reliability of the simulation results. In this work, design parameters such as the overflow diameter, underflow diameter, and the height of the cone and body parts of the hydrocyclone were investigated. Moreover, effect of the density of microplastics and the inlet velocity were studied as operating parameters. A statistical approach, analysis of variance (ANOVA), was utilized to quantify the influence of each parameter on the separation performance. To complement the statistical findings from ANOVA and to develop construct the nonlinear relationships between design variables and system performance, an artificial neural network (ANN) model was developed. This predictive model aimed to optimize two critical performance indicators: microplastic recovery percentage and water split percentage, based on geometrical inputs derived from CFD simulations. Through this integrated modeling approach, the study seeks to identify optimal geometric configurations that maximize microplastic recovery while minimizing water loss, thereby contributing to the development of a more efficient and environmentally sustainable separation system.