Development of an ANN-based predictive model for intelligent control of water hardness and TDS in industrial wastewater treatment using ion-exchange resins

Abstract. Water scarcity and environmental pollution remain among the most pressing global challenges of the 21st century, particularly in industrial regions. This study proposes an artificial neural network (ANN)-based predictive model for the intelligent control of water hardness (H) and total dissolved solids (TDS) mass concentration in the industrial wastewater treatment process using ion-exchange resins. Experimental data obtained from a pilot-scale purification system treating wastewater from the Kungrad Soda Plant in Uzbekistan were used to train and validate the model. The ANN was developed in MATLAB using a feedforward backpropagation algorithm, with H (in milligrams of calcium carbonate per litre, mg/L) and TDS (in mg/L) as input quantities and the servo valve opening degree (SerK) as the output quantity. The predictive model was trained on 80 experimental datasets and achieved high accuracy, with a mean squared error (MSE) of 9.72 × 10⁻⁴ and a regression coefficient R = 0.987, indicating a strong correlation between predicted and measured values. The trained ANN accurately modelled the nonlinear interdependence between influent water quality parameters and process control actions. For example, at input values of H = 2.0 mg/L and TDS = 20.0 mg/L, the model predicted a valve opening degree of 12.5%, which closely matched the empirical value. Similarly, when H = 3.43 mg/L and TDS = 35.5 mg/L, the model correctly predicted a minimised valve opening of 4.16%, confirming its predictive reliability across a broad operational range. These results demonstrate that the proposed ANN-based model can serve as an effective and reliable tool for real-time control and optimisation of wastewater treatment processes. Its ability to generalise from experimental data makes it particularly well-suited for dynamic and uncertain industrial environments, supporting smarter, data-driven decision-making in water resource management.