Prediction of particle transport an‎d dispersion in treatment units an‎d open channels, analysis of particle interactions with surfaces an‎d baffles, an‎d monitoring of processes affected by various environmental factors have always been key topics in environmental studies. Among these, the investigation of wastewater inflow into settling basins, determination of the optimal location of inlets an‎d outlets, an‎d the appropriate placement of bottom baffles to enhance solid particle sedimentation efficiency are of particular importance. In such cases, the advection-dispersion equation (ADE) is widely employed as the principal analytical framework for modeling these processes. In this research, with the aim of accurately simulating the sedimentation process under two-dimensional flow conditions in settling basins, the numerical solution of the advection–dispersion equation was carried out using deep learning an‎d the Physics-Informed Neural Network (PINN) approach. The modeling framework was implemented in Python using the powerful SciANN library. The influence of parameters such as inlet an‎d outlet positions, location an‎d height of the bottom baffle, flow discharge, an‎d inlet concentration was investigated using both experimental measurements an‎d analytical solutions in simplified (baffle-free) conditions. Experimental tests were conducted in several series on a pilot-scale settling basin. In these tests, the baseline condition (without baffle) was first eva‎luated, followed by scenarios with bottom baffles of varying heights an‎d distances from the inlet. The results demonstrated that the presence of a bottom baffle led to the breakup of the inlet jet, reduction of short-circuiting, an‎d an increase in particle residence time. Based on the analysis of different experimental series, for the given basin dimensions, the optimal baffle height was found to be about 0.09–0.12 times the flow depth, an‎d the optimal distance from the inlet was about 0.10–0.15 times the basin length. Comparison with the no-baffle condition showed a significant improvement in sedimentation efficiency, while the PINN model was able to accurately reproduce this trend an‎d predict the two-dimensional flow behavior an‎d suspended particle distribution in the basin. Moreover, comparison of the PINN outputs with the analytical solution of the ADE an‎d the pilot-scale experimental data confirmed the strong consistency of results an‎d the ability of the PINN to reliably represent the physical phenomena. Overall, the findings of this study highlight that the application of PINNs, especially when combined with experimental data, can serve as a novel an‎d powerful tool for accurate an‎d cost-effective simulation of complex environmental processes such as particle sedimentation in settling basins.

هستي هاشمي نژاد , محمدنويد مقيم

بشير موحديان عطار

ناصر طالب بيدختي , سعيد صرامي