In our previous study, TiAl alloys with varied cross-sections were prepared via the electromagnetic cold crucible zone melting (ECCZM) method. However, due to the unstable temperature field, the equiaxed grain region appeared in the columnar grain, disrupting the microstructure continuity. Herein, the processing method was redesigned based on numerical calculation and machine learning (ML). Real-time adjusted processing parameters were established to stabilize the temperature field. Firstly, the equation of absorbed heating power about withdrawing distance was calculated based on the magnetic field distribution in the cold crucible, and the input power was adjusted in real-time accordingly. Subsequently, a combination of real-time adjusted input power and withdrawing rate was established based on the balance of absorbed heating power and heat loss in the mushy zone. The specimen with real-time adjusted input power and withdrawing rate was also prepared. In addition, the matched input power and withdrawing rate were optimized by ML, and the TiAl alloy without equiaxed grain was obtained via the new processing method. Microstructure and mechanical properties of the directionally solidified TiAl alloy with the conventional ECCZM method (CM) and the real-time adjusted (RA) method were compared. It was found that equiaxed grains would induce intergranular fracture. The ML-optimized RA method can eliminate equiaxed grain and obviate the occurrence of intergranular fracture during tensile deformation, thus preventing early fracture and improving the tensile strength.
Read full abstract