We are pleased to announce the seminar of 

Dr. Mathieu AUCEJO, Associate professor (MCF-HDR) at the Laboratoire de Mécanique des Structures et des Systèmes Couplés - Cnam Paris, on "Parsimonious Bayesian Filters : Application to effort identification". 

On 15/12/2022 at 14h30 via Teams, and face-to-face in Amphi 3 at ISAE-Supméca.

Abstract :  

Bayesian filters of the Kalman type have been used for about fifteen years in structural dynamics to estimate jointly or sequentially the state of a system, as well as the forces applied to it. When we are only interested in the problem of force identification, the filters used in the literature have two major defects. The first concerns the smoothing of the reconstructed force fields when the density of identification points becomes important. The second defect is related to the drift phenomenon that appears when only acceleration measurements are used for the reconstruction.
To overcome these problems, we have developed a parsimonious Bayesian filter that introduces a constraint on the spatial distribution of the force field to be reconstructed, as part of Julian GHIBAUDO’s thesis. For this purpose, we used the Bayesian paradigm to obtain a unified formulation of the problem of estimating the state and the control of a system, allowing to introduce in a rigorous way the hypothesis of spatial parsimony of the effort field to be identified.
A comparison of the Bayesian parsimonious filter developed with the "state of the art" filters will be proposed on a common example.