On algorithms in FE-DIC
Résumé
Beyond simplifying the dialogue between experimentation and simulation, Finite Element Digital Image Correlation (FE-DIC) mainly allows to develop new hybrid approaches combining numerical models and experimental image data (integrated approaches, mechanical regularization, data-driven simulation…). These new approaches allow experimental solid mechanicians access to properties that could not be measured with conventional tools.
Since a decade, many efforts have been dedicated to improve the FE-DIC concept. Indeed, the method, initially developed in 2D and further extended in 3D (Digital Volume Correlation (DVC)), was also recently adapted to the context of Stereo-Digital Image Correlation (FE-SDIC).
In all these applications though, the FE-DIC method is, until now, essentially based on a Gauss-Newton algorithm. Most of the time, it is even a quasi-Gauss-Newton, since the gradient of the graylevel image is often kept constant during the iterations, for the construction of the left and right hand side operators.
In this presentation, we will review a series of other alternative algorithms to (quasi) Gauss-Newton to perform FE-DIC. For each considered algorithm, we will try to explain the theory, to illustrate on emblematic cases and to identify the advantages and drawbacks.