A data driven methodology for supporting the design phase of forming tools

An important source of uncertainty in the design phase of forming tools for new body elements is owed to the springback phenomenon occurring in the formed parts when external forces are released. Process and tool shape designers should account for springback in advance and introduce geometric compensations to avoid systematic errors in the end product. However, the springback phenomenon is owed to the residual stresses within the sheet-metal, which – in their turn – are affected by several factors such as the material constitutive law, the forming method (i.e. die, addendum, and blank-holder shapes), holding and friction forces between part and tools, etc., eventually affecting the topological distribution of strain trajectories and residual stresses in the part. To foresee the influence of every factor - and their combined effect - on the springback is very hard, asking for long trial-and-error loops including explicit simulation of the deep drawing process, as well as the implicit simulation to solve for springback deformation. In such a way, designing steps for geometric compensation of systematic errors are time consuming, and strongly related to the experience of involved experts. In order to obtain an optimal compensation of die geometry rapidly, and unaffected by experts’ experience, this paper proposes a data driven methodology, able to formalize a large part of this knowledge implicitly, thus minimizing the needs for experience and the number of iterations required to solve the problem. This is obtained by combining ANN and FEM techniques to predict the springback deformation in a metal forming process. Starting from this result, a searching algorithm can finally be adopted to find the optimal set of parametric values, which compensate the predictable springback effect.