Improving full-field identification using progressive model enrichments

Abstract

Full-field identification methods such as finite element model updating or integrated digital image correlation minimize the gap between an experiment and a simulation by iterative schemes. Within the algorithms residual fields and sensitivity fields are used to achieve identification. This paper discusses how these same fields can be used to assess the quality of the identification and guide toward successive enrichment of the constitutive model to progressively reduce the experiment-model gap. A cyclic experiment on a dog-bone sample made of aluminum alloy is used as an example to identify the parameters of an elasto-plastic model with exponential hardening and anisotropic yielding.

Simultaneous full-field multi-experiment identification

“…In FEMU, the gap between the experiment and simulation of the same experiment is minimized by optimizing (i.e., updating) the unknown model parameters. Within this paper a similar method is applied, which is referred to as Integrated-DIC [14][15][16][17][18][19][20]. Integrated-DIC optimizes the gap between simulation and experiment directly on the captured images by integrating the identification step within the DIC algorithm.…”

Optimal procedure for the identification of constitutive parameters from experimentally measured displacement fields

“…Such cases of model error are much more demanding as they call for an appropriate metric in a space of constitutive laws, a question that is dicult to address on objective grounds. The simpler issue of asking whether the behavior that is known to belong to a large class of laws may be restricted to a subclass (that of specic symmetries or properties), thereby leading to strategies of gradual enrichments of considered models [39] is even considered beyond the scope of the following analyses. Similarly, it will be assumed that discretization errors are not relevant.…”

Classic and inverse compositional Gauss‐Newton in global DIC

“…Given f ( x ) and , DIC thus consists in computing the transformation ϕ ( x , u ) as the solution of the gray‐level conservation equation: where a digital image f maps any sampling point x ∈ Ω to a quantized gray‐level value f ( x ), and ∘ refers to the composition operator between two functions. The residual map r ( x , u ), quantifying the noncompliance with the gray‐level conservation, can be used to validate or improve the kinematic model . The unknown transformation ϕ ( x , u ) is of the form where u ( x ) is the unknown displacement field.…”

“…The residual map r(x, u), quantifying the noncompliance with the gray-level conservation, can be used to validate or improve the kinematic model. 20,21 The unknown transformation (x, u) is of the form…”