Abstract
The problem of fluoride contamination in natural water sources has become increasingly relevant, particularly in rural areas and regions with naturally elevated fluoride levels in groundwater. The lack or military destruction of centralized water treatment infrastructure in many such areas exacerbates the health risks associated with excessive fluoride intake. These challenges highlight the need for affordable and locally adaptable solutions for decentralized water purification. The study focuses on the neural network (NN) simulation and optimization of sorption-based water treatment processes aimed at the removal of fluoride compounds using iron-containing sorbents derived from water treatment waste. For the NN approximation of the sorption capacity of the sorbent and fluoride ion removal efficiency as well as for the subsequent search for a compromise optimum (simultaneous multigoal maximization), a Python-based program was developed.
To address the multi-objective optimization problem, Harrington’s desirability functions were applied. As the global optima search method, the “differential evolution” algorithm was selected. The calculated optimal point corresponds to the findings of the authors' previous studies on the influence of pH and sorbent dose on the sorption process.
Approaches presented are essential for the cost-effective adaptation of sorption technologies to local needs. The integration of statistical and AI-driven models facilitates the development of intelligent decision-support systems for water treatment, contributing to both scientific advancement and practical environmental protection.
The application of computational modeling forms a cornerstone of modern research on sorption technologies. It enhances scientific understanding, enables data-driven design, and promotes the scalable implementation of innovative, sustainable water purification solutions.