Modelling Contact of Curved Surfaces in FE Models

Engineered products sometimes feature interfaces which are curved in nature which when represented by a faceted FEA model can lead to various issues and challenges to solve and work robustly. Traditionally such interfaces are modelled using contact algorithms such as the penalty methods or the augmented lagrangian methods. These techniques work well with matched meshes with little or no relative movement, but non-matched meshes introduce mesh faceting problems, problems which can also occur when a matched mesh is subject to large scale sliding. Faceting can lead to modelling challenges including fringes on stress distributions with artificially high stresses or non-robust solutions. Various solutions to this problem exist such as penetration removal, offset algorithms and geometry based contact but these all have limitations such as accuracy, stability or validity beyond small scale deformation. A new curved contact algorithm was developed in collaboration with Sandia National Laboratories in order to resolve these issues. the new algorithm using node-based reconstructions of the surface from the academic literature and incorporates them into SNL's custom engineering code. Both the Nagata and Gregory reconstruction were tested for speed and robustness. The developed algorithm showed great promise in test problems not only with significantly better accuracy but also the potential for faster convergence for implicit non-linear contact problems. This presentation will detail how the new algorithm was developed and the details behind the final implementation. Additionally testing of the algorithm will be presented including test examples and the results of real world application.