Papers by Author: Jia Zhou

Abstract: In the present study, the extrusion process for the AZ31B magnesium alloy was simulated using a DEFORM-3D software package to establish a database in order to provide input data for artificial neural networks (ANN). The network model was trained by taking extrusion ratio, ram speed, shape complexity and ram displacement as the input variables and the extrusion load and exit temperature as the output parameters. The data from FEM simulations were submitted for ANN as a training file and then ANN built were used to predict the target parameters. The ANN predicted results were found to be in agreement with the FEM simulated and experimental measured ones.

Abstract: The present case study addressed a practical problem of wall thickness attenuation during extrusion to produce a complex thin-walled hollow magnesium profile. A HyperWorks FEM software package was employed to aid in identifying the causes for the wall thickness attenuation. Recommendations were made to adjust the interspacing between the mandrels and the height of the welding chamber. The modified dies yielded much improved results in terms of velocity and hydrostatic pressure uniformity. The wall thickness of the extrudate predicted using FEM simulation was very close to experimental measurements. The case study demonstrated the feasibility of using FEM simulation as a useful tool to solve industrial problems encountered in the production of complex profiles.