Study of Membranes with Nanotubes to Enhance Osmosis Desalination Efficiency by Using Machine Learning towards Sustainable Water Management

Abstract

Water resources management is one of the most important issues nowadays. The necessity of sustainable management of water resources, as well as finding a solution to the water shortage crisis, is a question of our survival on our planet. One of the most important ways to solve this problem is to use water purification systems for wastewater resources, and one of the most necessary reasons for the research of water desalination systems and their development is the problem related to water scarcity and the crisis in the world that has arisen because of it. The present study employs a carbon nanotube-containing nanocomposite to enhance membrane performance. Additionally, the rise in flow brought on by a reduction in the membrane’s clogging surface was investigated. The filtration of brackish water using synthetic polyamide reverse osmosis nanocomposite membrane, which has an electroconductivity of 4000 Ds/cm, helped the study achieve its goal. In order to improve porosity and hydrophilicity, the modified raw, multi-walled carbon nanotube membrane was implanted using the polymerization process. Every 30 min, the rates of water flow and rejection were evaluated. The study’s findings demonstrated that the membranes have soft hydrophilic surfaces, and by varying concentrations of nanocomposite materials in a prescribed way, the water flux increased up to 30.8 L/m2h, which was notable when compared to the water flux of the straightforward polyamide membranes. Our findings revealed that nanocomposite membranes significantly decreased fouling and clogging, and that the rejection rate was greater than 97 percent for all pyrrole-based membranes. Finally, an artificial neural network is utilized to propose a predictive model for predicting flux through membranes. The model benefits hyperparameter tuning, so it has the best performance among all the studied models. The model has a mean absolute error of 1.36% and an R2 of 0.98.