A neuro-fuzzy approach to generating mold/die polishing sequences

A neuro-fuzzy approach to generating polishing sequences is presented. The measured initial surface roughness of the EDMed mold or die and the desired final polished surface roughness are the major inputs fed to the system. After receiving this input, the system consults its database for the polishing efficiency curves of the abrasive stone grain sizes available to it. With the outputs of this system being optimized for minimum polishing time, there is a selected sequence of grain sizes from among an available set, with each grain size used for each polishing step. There was a series of initial polishing experiments which were conducted for the different available grain sizes and workpiece roughness at different pressures and RPMʹs, with a change in roughness measured for polishing duration. From this, the database is constructed for designing the fuzzy logic rules. For constructing the exact membership function of the fuzzy interface, the neuro-fuzzy technique is combined with learning ability of neural network and the inferring ability of fuzzy logic system. Finally, while considering the stone-changing time, the actual experimental results from suggested polishing sequences are compared with the predicted value in order to establish the proposed approach.