Detailed preform design procedure considering the effect of heat treatment in IN718 disk forging

Abstract

We investigated the impact of heat treatment on preform deduction during the hot forging of In718 turbine disks. Proposed preform design method leverages a Convolutional Neural Network (CNN) to aim for minimized damage and uniform grain size in the final forged products. For damage calculation, we employ the Lemaitre damage model, while the Johnson-Mehl-Avrami-Kolmogorov (JMAK) model is utilized for grain size calculation. The CNN, trained on datasets of NURBS-generated preform shapes and final forged disks from them, showed superior performance in terms of damage and grain size of forged disk compared with other existing approaches. Moreover, the results indicate that preforms deduced with heat treatment considerations can lead to significant improvements in forging results, including up to a 17% reduction in average grain size and a 16% decrease in standard deviation compared to preforms without heat treatment consideration. These results underscore the importance of considering heat treatment in preform design, offering valuable insights for industries where the integrity of forged products, like aerospace, is critical.

Keywords

• Inconel 718
• Disk forging
• Convolutional neural network
• Lemaitre damage model
• JMAK model
• Heat treatment