The distribution and shape of abrasive grains on the grinding wheel surface influence the grinding characteristics. Therefore, we previously proposed a method to evaluate the diamond abrasive grain shape using image processing. Fourteen factors, including area and degree of damage, were evaluated numerically for regularly shaped diamond abrasive grains. In addition, the relationship between the shape and crushing strength was verified. However, ragged shaped abrasive grains or abrasive grains which have small wear flats could not be evaluated. In this study, a method for classifying such abrasive grains using deep learning is proposed. Finally, the effectiveness of the proposed method was verified by experiment.

Milling of tungsten carbide using two types of nano-polycrystalline diamond (NPD) ball end mills with different finishing conditions, as well as single-crystalline diamond and conventional polycrystalline diamond ball end mills, was performed to verify the differences in machining characteristics of each tool. Compared to other diamond tools, the surface machined by the polished NPD tool maintained high quality over a long machining distance. In both NPD tools with polished and unpolished edges, there was a difference in the machined surface roughness during long-distance machining. The finished state of the flank face was directly reflected on the machined surface. Furthermore, we studied an NPD tool fabrication method providing high surface quality using a femtosecond laser to enable greater productivity. We confirmed that the total machining time could be maintained within 1 hour and could be fully automated to achieve the final shape.

Generally, mirror finishing of brittle materials, such as quartz, is performed by free abrasive grain polishing. To reduce the processing time, we are researching other ways to obtain the same surface quality as polishing using a thermoplastic resin bonded diamond wheel. This study investigated the possibility of reuse through wheel regeneration. The grinding results with this recycled wheel show a roughness of 4.8 nmRa at the end of an optical fiber and a minimum connection loss of 0.027 dB. After grinding with the recycled wheel, the surface properties at the fiber end face were equivalent to the results with a new wheel.