BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:First steps of numerical simulation using Artificial Intelligence
DTSTART:20191009T153000Z
DTEND:20191009T160000Z
DTSTAMP:20231004T162200Z
UID:indico-contribution-4947@indico.esa.int
DESCRIPTION:Speakers: Pierre Alliez (INRIA Sophia Antipolis)\, François B
runetti (Dorea)\, Vincent Vadez (Dorea)\n\nAs for the thermal simulation o
f a satellite\, its design is crucial. It involves the integration in the
thermal and mechanical model of reduced parts of equipment or supports (ad
ditive manufacturing) from 3D models finite elements. The goal of the thes
is is the drastic reduction of 3D geometric models that minimizes the alte
ration of the physical properties (here the calculation of the view surfac
es) induced by the model reduction during the numerical simulation.\n\nWe
are currently working on **an approximation guided by view factors**. The
aim is to design a geometric approximation method where each atomic simpli
fication or optimization operation is guided by preserving the view factor
s of the reference model\, rather than preserving the geometry. The ultima
te aim being a radiative thermal computation on a "small" model made of a
few hundreds of facets instead of millions of facets. \nThe next step cons
ists in **a supervised learning of the geometric error metric** (using **d
eep learning**). Deep learning is a subset of machine learning (which is a
lso a subset of Artificial Intelligence). In both cases\, algorithms appea
r to learn by analyzing huge amounts of data. Deep learning processes data
using computing units\, called neurons\, arranged into ordered sections\,
called layers. A stack of those layers form what we call a neural network
\, which is the kind of model we will use.\nWhen working with supervised a
lgorithms\, the input data is labeled and has a specific expected result.
We train a model to predict the labels of the given training examples. As
training progresses\, the predictions or classifications become more accur
ate. Our idea is to use deep learning as follows : from a database of expe
rtly generated 3D models\, we will design a supervised machine learning me
thod using multiple layers to learn the geometric error metric able to gov
ern an automatic approximation algorithm so that the resulting thermal sim
ulation is as accurate as possible to a reference calculation. \nWe alread
y developed and are still improving a software relying on a new hierarchic
al geometric data structure for the efficient computation of view factors\
, which allows us to have a precise reference case.\n\nhttps://indico.esa.
int/event/308/contributions/4947/
LOCATION:Newton 2
RELATED-TO:indico-event-308@indico.esa.int
URL:https://indico.esa.int/event/308/contributions/4947/
END:VEVENT
END:VCALENDAR