Application of Neural Network in Precision Prediction of Hat-Section Profiles in Rotary Draw Bending

It is difficult to control the accuracy of hat-section profile rotary-draw bent products because their shapes are complex and many factors influence the forming process. In this paper, a multi-layer feed-forward neural network based on improved BP (back propagation) algorithm is adopted to study the influence law of the main material properties and forming parameters on the accuracy of hat-section profile forming process. Rotary draw-bending precision prediction is realized and the technology parameters in the forming process are controlled according to the given precision.