Robustness and sensitivity analysis of a virtual process chain using the S-rail specimen applying random fields

An important part in robustness evaluation of production processes is the identification of shape deviations. A systematic approach is typically based on the numerical evaluation of a DoE and the application of metamodels. They provide knowledge on solver noise and sensitivities of individual model parameters. This article presents the sensitivity analysis workflow of a linked deep drawing and joining process chain. LS-DYNA®, optiSLang and SoS is used. The challenge is to separate simulative from process and material parameters of AA 6014. Spatial quantities like variations in geometry, thinning and strain have to be considered in the next process steps. At the same time the number of required virtual CAE model evaluations must be limited. The solution is based on nonlinear metamodels and random fields.