Industrial EP(D)M Rubber Production Process: a First-Principle Data-Driven Modelling Approach

Abstract

Ethylene-Propylene-(Diene) rubber is one of the most worldwide used polymers, employed in a large variety of applications. Versalis EP(D)M is produced by suspension copolymerization, without the use of solvent, initiated by metallorganic component (Ziegler-Natta catalysts). In this work, a first-principle model to evaluate the performance of an industrial EP(D)M plant production is proposed. The polymerization mechanism was implemented in an Aspen Polymers v12.1 simulation of the industrial plant. Model parameters were tuned based on process data for selected product grades; model performance was cross-validated with further product grades and production line data. By leveraging external code (e.g., MATLAB/Python), process engineers can automate the tuning procedure, and accelerate sensitivity/optimization analysis. Both steady-state and dynamic simulations can successfully be employed to improve process understanding, analyse past conditions, monitor plant performance, and explore future process/product developments.