A Comprehensive Benchmark Dataset for Sheet Metal Forming: Advancing Machine Learning and Surrogate Modelling in Pro-cess Simulations
Pascal Heinzelmann, Sebastian Baum, Kim Rouven Riedmüller, Mathias Liewald, Michael Weyrich
Abstract
FEM simulations are widely used for process development in the field of sheet metal forming to streamline tool design, shorten development cycles and minimize costly try-out processes. However, significant manual adjustments remain necessary due to deviations between simulation predictions and actual production outcomes, stemming from modelling simplifications. To further accelerate development cycles and try-out phases, it is essential to improve simulation accuracy and reduce computational demands. Here, surrogate models derived from simulation data by using machine learning provide a promising solution. In this context, the present paper introduces a comprehensive dataset designed to train surrogate models for optimizing sheet metal forming processes. The dataset includes extensive FE simulation data of deep-drawn sheet metal parts with an example geometry, considering diverse material and process parameters. It captures interactions among the tool geometry, material properties and process conditions, providing insights into stress distributions and strain paths. An example application demonstrates using the dataset to model the impact of material and process parameters on the forming limit diagram (FLD) of a deep-drawn part. The dataset, along with detailed documentation of the simulation setup, parameter scope and data formats, is available to the scientific community to facilitate further research in sheet metal forming optimization.