Search

Various factors in hobbing processes affect profile and helix deviations of hobbed gears, which prevent problems on the hobbing machine from being analyzed using them. Identifying problems requires a system that allows itself to evaluate motion errors in a hobbing machine from them. The proposed one in this study has artificial intelligence for image analysis. Thus, the developed hobbing simulation provided the data on profile and helix deviations of hobbed gears derived from the motion errors in the hobbing process, which determined correlation coefficients to express the similarities of the simulated micro-geometries between tooth flanks of a hobbed gear. These correlation coefficients are color-coded according to their magnitude and converted into images to constitute training data for artificial intelligence. However, different motion errors could lead to similar ones in tests with helical gears. The present paper describes the comparisons between images obtained for spur and helical gears computed by hobbing simulation and the comparisons between ones produced from the simulated and actual micro-geometries. As a result, the motion error that causes similar images depends on gear geometries, i.e., whether spur or helical. Furthermore, actual micro-geometries depend on hob accuracies, which require their compensation. That yields equivalent images to those produced from simulated data.

A hobbed gear has pitch, profile, and helix deviations caused by various factors. However it is hard for even experts to specify which factors cause individual deviations from the measured ones. Therefore, a hobbing-machine-diagnosis system could play a key role in identifying the factors from hobbed-gear deviations. Artificial intelligence for image recognition, which is comparatively easy to use, could be appropriate for the system. Generally, a larger number of training data for artificial intelligence could improve the system accuracy, however, it must take a longer time for learning. Besides, efficient learning requires small image data. Typical gear-deviation diagrams are large images that include unnecessary information for the diagnosis system. Therefore, the image data for training have only information on the correlation coefficients between the deviations of two teeth selected at a time. The present paper describes a method for providing image data for learning. The image is the arrangement of correlation coefficients in rows and columns as a grayscale. This paper also describes the relationship between the image and the hobbing problems; e.g., radial or face runouts. As a result, the image data showed clear characteristics for the difference in the periodicity of the hobbing problem, however, did not show clear characteristics for the difference in the directionality.

In a hobbing process, many factors bring pitch, profile, and helix deviations to hobbed gears, and it is generally difficult to identify which factors cause the deviations. Therefore, a system that allows causal factors to be determined is required especially in the after-sales service of the hobbing machine manufactures. In the present study, a hobbing-machine-diagnosis system is being developed. Artificial intelligence will be employed in the system to easily determine the cause of deviations that occur in hobbed gears. The development of the system requires a lot of learning data for artificial intelligence. The data should be obtained in various conditions. However, collecting data from real cutting is almost impossible because of cost and time, so that hobbing simulation was carried out. The hobbing simulation has already been developed by many researchers. But those simulations are not enough for manufacturers of hobbing machines to evaluate problems in the machines. In addition, the conventional tooth profile/helix deviations diagrams are not appropriate for images used in artificial-intelligence systems, because they include redundant data. In the present paper, therefore, a network graph representing the correlation between every two teeth were discussed to examine whether the graph can be used as data for artificial-intelligence systems. As a result, an image of the network graph could be suitable for the system and could enable its data size to be reduced.

The high productivity of gear skiving has caused the process to attract a lot of attention in recent years and taper shaped skiving cutters are generally used in current gear skiving. A cutter axis in gear skiving is inclined to a gear axis and a taper shaped skiving cutter could cause tooth profile deviations of skived gears after re-sharpening of tool faces. Adjusting cutter position and/or re-profiling of cutter flanks are applied to enable tooth profile deviations of skived gears to remain within acceptable ranges when the adjustment is necessary. Adjusting cutter position is preferred to re-profiling of cutter flanks due to desire for low production cost. However, the universality of the current methods of calculating cutter positions is still not confirmed. The present study explains that the current methods for obtaining cutter positions could require individual calculating formulas according to cutter specifications for each particular gear. A universal method which avoids such complexity of making various formulas could contribute to permit gear skiving to become more useful in the gear machining field. Then such a desirable method is proposed and the reasonability of the method is confirmed by experiments. In addition, this study discusses the existence of scope in which the method can be applied and a method for checking collisions between cutter flanks and gear tooth flanks being cut after adjustment of cutter position.

Taper shaped skiving cutters are commonly used in gear skiving because they provide clearance angles under a simple gear arrangement and they could have lower cutting resistance than cylindrical skiving cutters. Their tooth flanks are generally ground by the same method of generation grinding with a trapezoidal wheel as ones for pinion cutter tooth flanks for gear shaping. In gear skiving, however, the process of tool-face re-grinding could increase profile deviations of cut gears due to the lack of an appropriate change in cutter profiles along the facewidth. In the present paper, computer programs were developed to simulate the generation grinding and the skiving with the cutting edges ground by the method. Skiving with a cutter ground by the method were carried out and skived tooth forms were compared with those calculated by the developed program. As a result, the comparisons could guarantee the reliability of the program. Furthermore, the program has revealed that the increase in a helix angle of a cutter yields the increase in tooth profile deviations of skived gears after regrinding the tool faces. Therefore, taper shaped cutters should be applied to internal gear skiving with extra caution.