Application of BPANN in spinning deformation of thin-walled tubular parts with longitudinal inner ribs^①

来源期刊：中南大学学报(英文版)2004年第1期

论文作者：江树勇 李萍 薛克敏

文章页码：27 - 30

Key words：artificial neural network; back-propagation; ball spinning; power spinning

Abstract: Back-propagation artificial neural network (BPANN) is used in ball backward spinning in order to form thin-walled tubular parts with longitudinal inner ribs. By selecting the process parameters which have a great influence on the height of inner ribs as well as fish scale on the surface of the spun part, a BPANN of 3-8-1 structure is established for pre-dicting the height of inner rib and recognizing the fish scale defect. Experiments data have proved that the average relative error between the measured value and the predicted value of the height of inner rib is not more than 5%. It is evident that BPANN can not only predict the height of inner ribs of the spun part accurately, but recognize and prevent the occurrence of the quality defect of fish scale successfully, and combining BPANN with the ball backward spinning is essential to ob-tain the desired spun part.