An improved constrained multi-objective optimization evolutionary algorithm for carbon fibre drawing process

Abstract

In this paper, an improved ϵCMOEA/D-DE is proposed to improve the performance of CMOP algorithm and achieve parameter optimization in carbon fibre drawing process. In order to avoid overusing the infeasible solutions, two repair operators are introduced in the population evolution model. More specially, during the evolutionary process, when the constraint violation of infeasible solution exceeds a tolerance threshold, the proposed two repair operators are used to find a better solution to repair the infeasible solution. On the other hand, in order to enhance the convergence rate of the Differential Evolution (DE), a modified DE is proposed. Then, an ϵCMOEA/D-mDE-RO is proposed by incorporating these two improved strategies into the ϵCMOEA/D-DE. Subsequently, the performance of the proposed ϵCMOEA/D-mDE-RO is evaluated on the constrained test problem series and the experiment shows that the proposed algorithm outperforms the existing ϵCMOEA/D-DE. Finally, in order to further illustrate the application potential, the proposed algorithm is successfully applied in optimizing carbon fibre drawing process and the optimal draw ratio vector is obtained.