Development of an artificial neural network to predict springback in air vee bending

Springback is a serious problem in the air vee bending process because of i ts inconsistency. An on-line tool to control springback is more reliable th an an analytical model which might not be able to control the stroke of the machine in real-time. Therefore, one might resort to adaptive control or u se an artificial neural network (ANN) trainer, either using experimental da ta or analytical predictions (or both), and use it for real-time control of the machine tool. The inconsistency in springback is then reduced to withi n acceptable limits. Adaptive control would need several strokes to complet e the job, but it is envisaged that the job could be completed in a single stroke with the ANN. The present paper discusses the development of an ANN which can be used to train and later to predict the springback, as well as the punch travel, to achieve the desired angle in a single stroke in an air vee bending process.