Entwicklung eines empirischen Modells zur Bestimmung von Ergebnisgrößen aus FEM-Rollfalzprozessen

Roller hemming is an operation used in the construction of vehicle body parts and follows on from deep drawing, trimming and folding the flange to produce a metal-formed joint between the outer skin part and the inner part of vehicle body parts. The roll-in and the appearance of the hem can be pre-examined and optimized if necessary by means of the finite element method (FEM). This enables process planning and the commissioning process to be supported at an early stage. Alongside the targeted simulation-based investigation of individual processes, it is possible to conduct virtual FEA tests according to the methods of the “design of experiments” (DoE). As a result of study, an empirical model can be derived which quantitatively describes the correlation between the influencing factors investigated and the target figures (e.g. roll-in). This article describes the development of a response surface model on the basis of a quadratic regression approach and a neural network between FEA simulation results of certain factor combinations from pre-strain, geometry, flange length, rolling direction and pre-hemming robot paths. With the aid of this model, roll-in should be determined by means of known boundary conditions of different components, as well as the outer radius of the hem. The results from the empirical models are tested for applicability with the results of experimental tests using the example of a door.