Seminar : Stéphane ROUX

Learning through images

Abstract

Digital Image Correlation (DIC) was originally conceived as a measurement of kinematic fields through a time series of images.  It is therefore a valuable element to be compared with the numerical simulation of these same kinematic fields, for the identification and validation (or even invalidation) of mechanical models and the measurement of the corresponding "material" parameters.  We will quickly present the foundations of this approach in the case of optical and tomographic imaging.
Taking a step back from the initial CID, the same approach can be used to match an image to its model, which requires a strong dialogue between these two components, and allows attention to be focused on a small number of parameters but chosen for their relevance.  The model itself can be more or less elaborate, ranging from a fine description to a "proto-model", such as a simple set of properties.   These notions will be illustrated by a heterogeneous set of examples, some of which are chosen for their shared roots between MSSMat and LMT.

Lien de connexion : Zoom, ID de réunion : 833 9379 3079