Neural network representation of mechanical fasteners in large-scale analyses
This paper presents an artificial neural network (NN) modeling approach for representing mechanical fasteners in large-scale finite element crash simulations for explicit analysis using LS-DYNA version R9.3.1. The NN-model is established to describe the local force-deformation response of point-connectors in automotive applications like self-piercing-rivets and flow-drill-screws. The behaviour from initial loading until failure or unloading is covered. Various architectures and complexities of feedforward NNs were evaluated and trained based on synthetic experiments generated from the constraint model proposed by Hanssen et al. [1]. The constraint model is available as *CONSTRAINED_SPR2 but was used in form of a cohesive element (8-noded, 4-point cohesive element with offsets for use with shells).
https://www.dynalook.com/conferences/13th-european-ls-dyna-conference-2021/simulation-methods/andre_ntnu.pdf/view
https://www.dynalook.com/@@site-logo/DYNAlook-Logo480x80.png
Neural network representation of mechanical fasteners in large-scale analyses
This paper presents an artificial neural network (NN) modeling approach for representing mechanical fasteners in large-scale finite element crash simulations for explicit analysis using LS-DYNA version R9.3.1. The NN-model is established to describe the local force-deformation response of point-connectors in automotive applications like self-piercing-rivets and flow-drill-screws. The behaviour from initial loading until failure or unloading is covered. Various architectures and complexities of feedforward NNs were evaluated and trained based on synthetic experiments generated from the constraint model proposed by Hanssen et al. [1]. The constraint model is available as *CONSTRAINED_SPR2 but was used in form of a cohesive element (8-noded, 4-point cohesive element with offsets for use with shells).
Andre_NTNU.pdf
— 1.3 MB