Multi-objective optimization of single screw polymer extrusion based on artificial intelligence

Abstract

The performance of the single screw polymer extrusion process depends on the definition of the best set of design variables, including operating conditions and/or geometrical parameters, which can be seen as a multi-objective optimization problem where several objectives and constraints must be satisfied simultaneously. The most efficient way to solve this problem consists in linking a modelling routine with optimization algorithms able to deal with multi-objectives, for example, those based on a population of solutions. This implies that the modelling routine must be run several times, and, thus, the computation time can become expensive, since they are based on the use of sophisticated numerical methods due to the need to obtain reliable results [1]. The aim of this work is to present an alternative based on the use of Artificial Intelligence (AI) techniques in order to reduce the number of modelling evaluations required during the optimization process. This analysis will be based on the use of a data analysis technique, named DAMICORE, able to define important interrelations between all variables related to extrusion and, then, optimize the process [2,3,4]. For that purpose, the results obtained for three practical examples will be presented and discussed. These case studies include the optimization of screw geometrical parameters, barrel grooves section and rotational barrel segment. It will be shown that the results obtained, taking into consideration the design variables, the objectives and the constraints defined, are in agreement with the expected thermomechanical behaviour of the process.