| Résumé | The prevalence of microplastics (MPs) pollution has cast a shadow over aquatic ecosystems and their inhabitants, including humans. Each year, water bodies transport millions of tons of plastic into the ocean, with a considerable portion of this plastic settling in aquatic environments, leading to complex deposition patterns that impact aquatic ecosystems. This study introduces a novel application of machine learning combined with dimensionless analysis to model MPs deposition. The model offers practical value for predicting MPs behavior under varied flow conditions, supporting global efforts in pollution monitoring and mitigation. In this regard, this study aimed to develop a novel machine learning model to predict the deposition patterns of spherical and cylindrical microplastics (MPs) with identical particle diameters (dₚ) and flow dynamic viscosities (v), utilizing laboratory-generated datasets. To achieve this, different generic test scenarios were conducted to collect data for various cases of the water depth in the channel (w), the flow velocity (uf), the water depth in the channel where it undergoes deepening (h), the slope applied to the channel’s bed (s), and the particle shape (including spherical and cylindrical). Eleven models including different dimensionless combinations (obtained using Buckingham theorem) of input variables were taken into account. Statistical evaluation metrics were used to determine the best model for predicting either spherical or cylindrical MPs. The model including all four dimensionless inputs was found to be the best model based on data for either spherical (R = 0.94, MAE = 0.06, RMSE = 0.09, and BIAS = 0.01) or cylindrical MPs (R = 0.90, MAE = 0.09, RMSE = 0.11, and BIAS = 0.02). The sensitivity analysis and confidence intervals revealed that the ratio of the water depth in the channel’s bed deepening to the particle diameter had the most significant influence on the deposition patterns of both spherical and cylindrical MPs. These findings underscore the potential of machine learning approaches in advancing our understanding and prediction of microplastic behavior in aquatic environments, contributing to improved environmental monitoring and management strategies |
|---|
