INVESTIGATIONS ON SPRINGBACK OF BENT TUBES USING DESIGN OF EXPERIMENT AND ARTIFICIAL NEURAL NETWORKS

The bending is one of the most used operations applied to the metallic tubes. In the least years the demands regarding the accuracy of bent tubes has dramatically increased. One of the problems which still affect the accuracy of bent tubes is the bend angle error caused by the springback phenomenon. In this paper is present a factorial experiment in order to determine the relative importance of four factors on the springback of bent tubes. In order to obtain a representative experimental data, it was used a factorial experiment of type L9 (34). The relative influence of each factor on the springback was determined using the ANOVA method.The obtained experimental data were used for training of a Back-Propagation Neural Network (BPNN), in order to predict the springback of bent tubes. The input layer on the artificial network, consist from 4 inputs. These inputs were: tube diameter, bending radius, bending angle and yield stress.