Investigation and prediction of springback in rotary-draw tube bending process using finite element method

Abstract

Rotary-draw tube bending operation is one of the most universal methods used for the tube forming processes. Similar to the other forming methods, some problems such as wall thinning, cross-section distortion, wrinkling, and springback can also be seen on the tubes formed by rotary-draw bending operations. Springback is a very common problem and its prediction plays a crucial role in increasing the efficiency of the tube bending operations and also to overcome the difficulties in the assembly processes. Tube diameter, wall thickness, bend radius, bend angle, and coefficient of friction can be considered as the most effective parameters that cause the variation of springback magnitude. In this study, not a simple one-at-a-time sensitivity analysis, but a thorough investigation of the springback phenomena involving interactions between the geometrical and mechanical parameters is done and surrogate models are developed via the data obtained from finite element analysis using a multi-purpose explicit and implicit finite element software LS-DYNA to analyze the non-linear response of structures. The constructed surrogate models can be utilized to perform fast prediction of springback for a given combination of parameters. Three different surrogate modeling techniques are exploited and it is found that the linear polynomial response surface approximations can provide acceptable accuracy. Finally, experiments are conducted to validate the accuracy of surrogate models. It is observed that the cross-validation error predictions are close to the errors observed in the experiments.