Machine learning based approach to predict ductile damage model parameters for polycrystalline metals

Damage models for ductile materials typically need to be parameterized, often with the appropriate parameters changing a given material depending on loading conditions. This can make parameterizing these computationally expensive, since an inverse problem must solved each condition. Using standard modeling techniques requires hundreds or thousands of high-fidelity computer simulations estimate optimal parameters. Additionally, time human expert is required set up model. Machine learning has recently emerged as alternative approach in settings, where machine model trained offline manner and new quickly generated fly, after training complete. work utilizes such workflow enable rapid parameterization damage called TEPLA The efficiently much faster, compared previously employed methods, Bayesian calibration. results demonstrate good accuracy synthetic test dataset validated against experimental data.