Improving one-dimensional pollution dispersion modeling in rivers using ANFIS and ANN-based GA optimized models.

Area of Science:

• Environmental Engineering
• Water Resources Management
• Computational Fluid Dynamics

Background:

• Accurate simulation of pollutant transport is crucial for river engineering and water quality management.
• The one-dimensional advection-dispersion equation (1D-ADE) is a standard model, but its accuracy depends on precise estimation of the longitudinal dispersion coefficient (Dx).
• Traditional methods for Dx estimation often lack the accuracy required for complex river systems.

Purpose of the Study:

• To develop and evaluate a hybrid numerical-intelligence model for enhanced prediction of pollutant dispersion in open-channel flows.
• To improve the estimation of the longitudinal dispersion coefficient (Dx) using artificial intelligence techniques.
• To integrate optimized artificial intelligence models with a numerical 1D-ADE solver for superior water quality simulation.

Main Methods:

• Development of a hybrid model combining artificial intelligence (AI) modules with a numerical 1D-ADE solver.
• Optimization of AI models, including Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANNs), using the Genetic Algorithm (GA).
• Numerical solution of the 1D-ADE using the Physically Influenced Scheme (PIS) within the finite volume method.
• Comparison of AI-based Dx estimations (ANN, ANFIS, ANFIS-GA, ANN-GA) and empirical methods against observed data from 505 river sections.

Main Results:

• The ANFIS-GA model demonstrated the highest accuracy in estimating the longitudinal dispersion coefficient (Dx) compared to other AI and empirical methods.
• The hybrid PIS-ANFIS-GA model significantly outperformed classical methods (PIS-ANFIS, PIS-empirical) in simulating pollutant transport, as validated by R², RMSE, MAE, and NSE metrics.
• The developed hybrid model accurately predicts dispersion processes in rivers, showing improved performance against analytical solutions and measured concentration hydrographs.

Conclusions:

• The hybrid numerical-intelligence model offers a more accurate and reliable approach for sediment and pollutant dispersion prediction in open-channel flows.
• The integration of GA-optimized ANFIS and ANN models represents a novel and effective advancement in water quality modeling.
• This approach enhances the practical applicability of advanced AI techniques in environmental and river engineering studies.