Optimization strategy for the indirect isothermal extrusion of high-strength aluminum alloy based on temporal fusion transformer

Isothermal extrusion is an important method for ensuring the quality of extruded profiles. On the basis of the actual indirect extrusion process of 7075 aluminum alloy with high strength and toughness, this study used the combination of a finite element model and deep neural network to construct a prediction model for the temperature series of the indirect extrusion outlet. The extrusion velocity–stroke control curves of the isothermal extrusion of aluminum profiles were obtained on the premise of ensuring maximum production efficiency and the absence of defects. Results showed that the constructed model had high robustness and performed well on the validation set with an average error of 1.8% and that the temperature difference at the outlet of the profile obtained through isothermal extrusion would not exceed 5 °C. By using machine learning, finite element simulation or experiments can be omitted from process optimization, and the indirect isothermal extrusion process can be rapidly optimized. Then, consistency along the extrusion direction was proven by microstructure evolution simulation.