Enhancement of microstructure homogeneity and hardness after heat treatment of forged flex splines with CNN-based preform design

This study presents the application of a convolutional neural network (CNN)-based preform design aimed at improving the deformation homogeneity of flex splines, a crucial component in strain wave gears. Flex splines were forged using an existing preform and a CNN-designed preform, followed by isothermal normalizing heat treatments. The resulting microstructures and Rockwell hardness distributions were systematically compared. After forging, the flex spline fabricated with the CNN-designed preform exhibited a reduction in grain size and an increase in grain orientation spread (GOS) and kernel average misorientation (KAM) relative to the existing preform. The standard deviations of grain size, GOS, and KAM were reduced by 59 %, 22 %, and 55 %, respectively, indicating enhanced microstructural homogeneity. Although the average hardness remained unchanged, its standard deviation decreased by 74 %, reflecting improved mechanical uniformity. After isothermal normalizing, the CNN-designed preform demonstrated higher overall hardness levels and a pronounced suppression of grain growth, with grain size reduced by up to 88.8 %. Furthermore, KAM and GOS showed strong positive correlations with hardness, suggesting that improved deformation homogeneity contributes to the stability of the microstructure and mechanical property uniformity after heat treatment. These findings demonstrate that CNN-based preform design, targeting effective strain homogenization, offers a highly effective approach to refining the microstructure and enhancing mechanical performance in the manufacturing process of flex splines.

Abstract licence: CC BY-NC-ND

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.