Experimental Study and Modelling Through Artificial Intelligence of the Separation by Azeotropic Distillation of Solvents Used in Industrial Recycling Processes

The use of aqueous acidic solvents plays a crucial role in industrial processes, in particular in the recycle operations. Very often, recycling and material recovery operations are essential for the circular economy of the products and materials present in the market. In order to make sustainable the solvent based recovery processes, it is essential to be able to recycle the solvents themselves, bringing them back to the required composition and purity. Aqueous acid solvents are largely applied in industry and their recovery by distillation can be challenging due to the strongly non ideal thermodynamic behaviour of water-acids mixtures. A possible solution is the addition of an entrainer to conduct the azeotropic distillation. As part of the funded project named RE-POLY.AI, conducted in collaboration with Radici InNova s.c. a r.l. (R&D company of RadiciGroup) and Téchnéos s.r.l., experimental tests were performed using process fluids from an industrial recycling. A three-meter continuous distillation column with 15 physical trays was used to evaluate different process conditions and identify the optimal configuration for the separation. This research was performed combining experimental test with a predictive artificial intelligence (AI) framework that integrates Bayesian optimization based on Gaussian Process modelling and ensemble Random Forest regression with uncertainty quantification capabilities. Trained on data from multiple test conditions, this AI-driven approach guided the selection of new experimental setups and provided insights into optimized separation configurations.