New multiobjective optimisation algorithms for assembly lines design
Résumé
Multiobjective optimisation methods for an assembly line design problem are presented in this paper. The studied problem consists of two parts. Among a set of candidate machines, the first part of the problem aims to assign a single machine to each workstation. The goal of the second part is to size the intermediate buffers. Two objectives are taken in consideration for our design problem: the minimisation of the cost of the line and the maximisation of the throughput rate. Different new multiobjective methods are developed to solve the problem. First, a multiobjective ant colony optimisation algorithm is proposed. Then, in order to get better results, the first algorithm is coupled with a guided local search. The third method proposed for the first time to solve our problem, called L-ant, is based on a multiobjective ant colony algorithm but using the Lorenz dominance. The fourth algorithm is another new method based on genetic algorithms and the Lorenz dominance and called Lorenz-archive. In order to compare the different methods to each others and to assess their efficiency, different measuring criteria are applied on the best fronts with the non-dominated solutions.