Artificial neural networks towards average properties targets in styrene ARGET-ATRP

Abstract

The application of artificial neural networks (ANN) in the synthesis of polystyrene via ARGET-ATRP was presented for the first time. In this research, it was utilized a deterministic modeling to train ANN operating in the direct and inverse way, that is, with the possibility of identifying reaction conditions from target polymer average properties and vice versa. Prediction deviations by ANN were less than 20% in all cases, and for monomer conversion and dispersity, these values did not exceed 10%. This approach provides an alternative possibility for intelligent control of the dispersity and degree of polymerization. It was exposed that the control strategies learned are robust and can be transferred to similar ARGET-ATRP reaction configurations. Moreover, it was demonstrated that the inverse ANN remains an outstanding alternative to overcome the limitations of traditional deterministic modeling, in which direct and rapid prediction of reaction conditions from the polymer properties as input parameters is difficult. Hence, we believe this work represents a bottom line for the use of modern techniques of artificial intelligence in the controlled synthesis of polymers.