Study On The Nonisothermal Stamping Of Magnesium Alloy Based On An Adaptive SVR-ELM Ensemble Surrogate Model

Magnesium alloy has great application prospect in industrial fields with excellent comprehensive properties.Because of the hexagonal close-packed(HCP)structure,magnesium alloy has a poor plastic deformation ability and the industry application of magnesium alloy is severely limited.While the magnesium alloy is heated to a proper temperature,its plasticity is obviously better than that of normal temperature and the plastic forming ability will be greatly improved.Therefore,the application of nonisothermal forming technology in the forming of magnesium alloy has been paid more and more attention.How to improve the accuracy of the finite element simulation prediction in nonisothermal stamping of magnesium alloy and how to improve the quality of magnesium alloy forming parts by the control of plasticity differentiation are focused on in this research.The main work is as follows:Based on the Support Vector Regression(SVR)and the Extreme Learning Machine(ELM),the ensemble surrogate model SVR-ELM was established by weighting method.An adaptive approach was proposed to update the sample spaces of the ensemble surrogate model and the weight coefficient of ensemble surrogate model was calculated by heuristic algorithm.The quantum genetic algorithm was modified by dynamic quantum rotation gate in order to improve the performance of the algorithm.The NUMISHEET 2011 AZ31 B cross-shaped cup part was taken as the research object.The SVR-ELM ensemble surrogate model was estabilished between parameters of Johnson-Cook constitutive model and temperature of parts.Based on adaptive method,sample spaces and surrogate model could be updated by local optimal solution obtained during optimization process.The minimum value of the error function between the simulation result and the test data was found by using improved quantum genetic algorithm and then the optimum constitutive parameters for nonisothermal stamping of magnesium alloy AZ31 B would be obtained.Based on the finite element model of NUMISHEET 2011 cross-shaped cup part,the key area of blank holder affecting forming quality was divided into subareas and used as input parameters.The SVR-ELM ensemble surrogate model was estabilished between subareas and the reduction rate uniformity of the side wall thickness.The ensemble surrogate model could be updated by an adaptive method,and the optimum subareas would be obtained by improved quantum genetic algorithm.So the plasticity difference of the blank flange was realized by controlling temperature on the blank flange,and material flow was controlled better.Shapeoptimization was carried out based on shape optimization theory in stress concentration areas of blank holder to improve the stress distribution.A discretized method was used to disperse the key area of blank holder affecting forming quality in nonisothermal stamping of magnesium alloy.The finite element analysis of nonisothermal stamping with the discrete structure of blank holder was carried out to obtain the reaction force of each discrete unit.The set of obtained reaction forces was analyzed and the topological optimization models based on reaction force subset was used to optimize the structure of blank holder.Compared with the previous method based on the surrogate model and optimization algorithm,the approch of using the reaction force in the forming process to optimize the blank holder can reduce design variables obviously and it has higher efficiency.The plasticity differentiation of the blank flange was realized by using the optimized blank holder to control the temperature on the blank flange.This approach effectively improved the quality of the forming parts in nonisothermal stamping of Magnesium alloy.