State estimation and fixed-horizon prediction for an industrial polypropylene polymerization reactor

Details

Supervisors

Mignon, Denis

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en chimie et science des matériaux, à finalité spécialisée

Abstract

enThis thesis was conducted in collaboration with TotalEnergies and focused on one of its polypropylene production lines at the Feluy polymer production plant. This unit, named PP3, produces a wide range of polypropylene grades and makes frequent transitions from one grade to another. Each grade has specifications notably in terms of mechanical and thermal properties. These properties are directly related to the operating conditions in the reactor which need to be precisely controlled. But operating the reactor is a complex job, because of the number of parameters involved, the interaction between them and the limited number of reliable measurements within the reactor. The product properties, for example, are only measured a few times a day from samples taken on a fixed schedule. Especially during grade transitions, in transient, dynamic regime, operation of the reactor is mainly heuristic, relying on operator experience and with little feedback from the system. For the profitability of the process, it is important that the grade transitions are carried out fast, while minimizing the amount of off-grade polymer produced during the adjustment of the reaction conditions. In this context, the goal of this work was to support reactor operation by developing a computational model capable of predicting and simulating the effects of input parameter variations on product properties.