ホブ盤診断システムのための教師データ作成とその妥当性

抄録

There are various factors in hobbing processes that affect the profile deviation and helix deviation of gears. Accordingly, it is difficult to analyze the problems on the hobbing machine from the profile and helix deviations of the hobbed gears. Therefore, we have developed a system with artificial intelligence that estimates the motion errors in a hobbing machine from the profile and helix deviation of the hobbed gears. First, hobbing simulations provided the data on profile and helix deviations of hobbed gears, and correlation coefficients expressed the similarities of the simulated micro-geometries between tooth flanks. The correlation coefficients constituted the images used for training artificial intelligence. In the present paper, we compared the images created from the hobbing simulations with those obtained by actual hobbing processes and examined the possibility of the images constructed with the hobbing simulations in training artificial intelligence. As a result, the images showed the decent possibility of the training.