Generation of Tool Path in Incremental Forming of Regular Polygonal Pyramid Shape Using Neural Network and Smoothing Formed Surface

Abstract

Neural network models corresponding to regular polygonal pyramids were generated, and tool paths for incremental forming were generated to improve forming accuracy. The learning shapes were regular tetragonal, pentagonal, and hexagonal pyramids. SUS304 sheets with dimensions of 150 mm x 150 mm x 0.3 mm were used as specimens. The Z-coordinates after positioning and the base circle radius lengths of each shape were used as shape date for learning. The base circle radius lengths were determined in three patterns with different height ranges, and three models were created. The Z-coordinate and the radius length of base circles were output as tool paths, and they were smoothed to improve surface accuracy. As a result, the forming accuracy of the formed parts excluding the flange, was improved, and the surface accuracy was improved by smoothing the generated tool paths.