Calibration of Elastoplastic Constitutive Model Parameters from Full-field Data with Automatic Differentiation-based Sensitivities.

摘要: We present a framework for calibration of parameters in elastoplastic constitutive models that is based on the use automatic differentiation. The model problem posed as partial differential equation-constrained optimization where finite element (FE) coupled equilibrium equation and evolution equations serves constraint. objective function quantifies mismatch between displacement predicted by FE full-field digital image correlation data, solved using gradient-based algorithms. Forward adjoint sensitivities are used to compute gradient at considerably less cost than its calculation from difference approximations. Through differentiation (AD), we need only write constraints terms AD objects, all derivatives required forward inverse problems obtained appropriately seeding evaluating these quantities. three numerical examples verify correctness gradient, demonstrate approach's parallel computation capabilities via application large-scale model, highlight formulation's ease extensibility other classes models.