Hybrid algorithms for the twin–screw extrusion configuration problem

• For the TSCP hybridization of algorithms improves performance.
• The hybrid algorithms can be classified into two main approaches. The first is to use PLS as a post-processing of the non-dominated set of solutions generated by TPLS, MOACO, or MOEA.
• In this case, we obtained rather large improvements over the non-hybrid search methods used in the first phase.
• The second main approach to hybridization was to seed the initial set of solutions to PLS, MOACO and MOEA, respectively.

The twin-screw configuration problem (TSCP) arises in the context of polymer processing, where twin-screw extruders are used to prepare polymer blends, compounds or composites. The goal of the TSCP is to define the configuration of a screw from a given set of screw elements. The TSCP can be seen as a sequencing problem as the order of the screw elements on the screw axis has to be defined. It is also inherently a multi-objective problem since processing has to optimize various conflicting parameters related to the degree of mixing, shear rate, or mechanical energy input among others. In this article, we develop hybrid algorithms to tackle the bi-objective TSCP. The hybrid algorithms combine different local search procedures, including Pareto local search and two phase local search algorithms, with two different population-based algorithms, namely a multi-objective evolutionary algorithm and a multi-objective ant colony optimization algorithm. The experimental evaluation of these approaches shows that the best hybrid designs, combining Pareto local search with a multi-objective ant colony optimization approach, outperform the best algorithms that have been previously proposed for the TSCP.