BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20230831T095746Z
LOCATION:Hall
DTSTART;TZID=Europe/Stockholm:20230627T193000
DTEND;TZID=Europe/Stockholm:20230627T213000
UID:submissions.pasc-conference.org_PASC23_sess116_pos115@linklings.com
SUMMARY:P55 - Novel Geometric Deep Learning Surrogate Framework for Non-Li
near Finite Element Simulations
DESCRIPTION:Poster\n\nSaurabh Deshpande (University of Luxembourg); Jakub
Lengiewicz (University of Luxembourg, IPPT PAN); and Stéphane Bordas (Univ
ersity of Luxembourg)\n\nConventional numerical methods are computationall
y expensive in simulating non-linear phenomena arising in mechanics. In th
is aspect, deep learning (DL) techniques are being increasingly used for a
ccelerating simulations in mechanics. However, existing DL methods perform
inefficiently as the size and complexity of the problem increases. In thi
s work we propose a novel geometric deep learning surrogate framework, whi
ch can efficiently find non-linear mappings between mesh-based datasets. I
n particular, we propose two novel neural network layers, Multichannel Agg
regation (MAg) layer, and the graph pooling layer, which are combined to c
onstitute a robust graph U-Net architecture. Our framework can efficiently
tackle problems involving complex fine meshes and scales efficiently to l
arge dimensional inputs. We validate the performance of our framework by l
earning on numerically generated non-linear finite element datasets and by
comparing the performance to state-of-the-art convolutional neural networ
k frameworks. In particular, we show that the proposed GDL framework is ab
le to accurately predict the nonlinear deformations of irregular soft bodi
es in real-time.
END:VEVENT
END:VCALENDAR