Modeling, analysis and multi-objective optimization of twist extrusion process using predictive models and meta-heuristic approaches, based on finite element results

Recently, twist extrusion has found extensive applications as a novel method of severe plastic deformation for grain refining of materials. In this paper, two prominent predictive models, response surface method and artificial neural network (ANN) are employed together with results of finite element simulation to model twist extrusion process. Twist angle, friction factor and ram speed are selected as input variables and imposed effective plastic strain, strain homogeneity and maximum punch force are considered as output parameters. Comparison between results shows that ANN outperforms response surface method in modeling twist extrusion process. In addition, statistical analysis of response surface shows that twist extrusion and friction factor have the most and ram speed has the least effect on output parameters at room temperature. Also, optimization of twist extrusion process was carried out by a combination of neural network model and multi-objective meta-heuristic optimization algorithms. For this reason, three prominent multi-objective algorithms, non-dominated sorting genetic algorithm, strength pareto evolutionary algorithm and multi-objective particle swarm optimization (MOPSO) were utilized. Results showed that MOPSO algorithm has relative superiority over other algorithms to find the optimal points.