ISSN print edition: 0366-6352 ISSN electronic edition: 1336-9075 Registr. No.: MK SR 9/7 Published monthly

Modeling and prediction of operating parameters in electrodialysis process for water desalination: Integration of machine learning and response surface methodology

Pankaj D. Indurkar, Prashant Upadhyay, and Vaibhav Kulshrestha

Membrane Science & Separation Technology Division, CSIR-Central Salt and Marine Chemicals Research Institute, Bhavnagar, India

E-mail: pankajdi@csmcri.res.in

Received: 5 July 2025  Accepted: 24 January 2026

Abstract:

This study evaluated water desalination efficiency focused on the influence of key process parameters viz. voltage, volume ratio, flowrate and concentration with their modeling and optimization using combination of Response Surface Methodology and Artificial Neural Network techniques by electrodialysis technology. Response Surface Methodology and Artificial Neural Network was used to develop models to predict salt removal, current efficiency and energy consumption. A total of thirty experimental data were fitted and ANOVA analysis was used to validate the accuracy of the models. Response Surface Methodology results showed that all the parameters significantly influenced removal efficiency. The regression model effectively predicted the optimal operating conditions of electrodialysis process with highest efficiency. Three-dimensional surface plots and two-dimensional counter plots were created to show the effect of operating parameters. The predictive performance of the Response Surface Methodology models was compared with that of the machine learning Artificial Neural Network model using Levenberg-Marquardth algorithm. At optimum operating conditions such as voltage of 1.5, volume ratio of 0.67, flowrate of 163 mL/min and concentration of 0.18 M, Response Surface Methodology predict salt removal of 92.52%, current efficiency of 50.05% and energy consumption of 1.81 kWh/kg, while Artificial Neural Network predict 93.85%, 50.29% and 1.99 kWh/kg, respectively. The study revealed that, Response Surface Methodology accurately predicted the interactions and significance of the process parameters than Artificial Neural Network model. Overall, this study suggests that Response Surface Methodology is an effective tool for optimization of water desalination by electrodialysis technology with minimal experiments.

Graphical abstract

The alternative text for this image may have been generated using AI.

Keywords: Desalination; Electrodialysis; Central composite design; Response surface methodology; Artificial neural network

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-026-04701-z

Chemical Papers 80 (4) 4409–4428 (2026)