Machine learning for the improvement of springback modelling

Abstract

New demands in the automotive industry have led to an increase in the use of Advanced High-Strength sheet metal materials. However, higher values of strength are usually achieved at the expense of reduced formability and increased sensitivity of the springback. Today, springback is one of the more important factors that influence the quality of sheet metal forming products. During the forming process, sheet metal undergoes a complicated deformation history, which is why the accurate prediction of the springback level can be very difficult. Today, a good compromise between the finite element method (FEM) simulation and the real stamping process can be achieved, but there is still limited reliability of the FEM springback prediction. In this paper, the machine learning (ML) approach was used to update the FEM for springback modelling. Combined models are tuned to better reflect the measured experimental data. Springback of sheet metal products is a very complex problem. It is the result of the stress state in the material following the forming process and means a change of shape in the sheet metal forming product after the withdrawal of the forming forces. Steel sheets with high strength and aluminium alloys are more sensitive to the springback effect due to a greater degree of elastic deformation than conventional mild steels. A comprehensive examination of the current estimation of the springback of sheet metal after forming is shown in the work [1]. Analytical, experimental and numerical approaches are introduced in detail. Analytical solutions in their completeness are only valid for simple ideal cases, but it also provides advanced understanding of the relations between some of the material and the process parameters that increase or decrease the level of springback [2]. Numerical simulations are a kind of approximation for the real behaviour of sheet metal during processing. By modelling the FEM (finite element method) the physical structure of the sheet metal and tools is converted into a mathematical model for the solution to use the numerical procedure, in particular, the modelling material properties of the sheet and the contact and friction conditions of the tool depend on the reliability of the forming process with computer simulations. For the conversion of sheet metal material to a numerical model a few material models were developed. The advantages and disadvantages of the different material models and the standard experiments to determine their parameters are presented in the work [3]. It is necessary to know and understand the influence of the numerical parameters on the simulation results for the successful application of the FEM to predict springback. Full reviews of the numerical simulation of sheet metal forming are presented by many authors [4, 5]. These are guidelines for the determination of material models and numerical simulation parameters for different materials, followed by a comparison with the results of