BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Engineering Department Mechanics Colloquia Researc
h Seminars
SUMMARY:Model-Free Data-Driven Science: Cutting out the Mi
ddleman - Prof Michael Ortiz\, California Institut
e of Technology
DTSTART;TZID=Europe/London:20210219T160000
DTEND;TZID=Europe/London:20210219T170000
UID:TALK156577AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/156577
DESCRIPTION:We have developed a new computing paradigm\, which
we refer to as Data-Driven Computing\, according
to which calculations are carried out directly fro
m experimental material data and pertinent kinemat
ic constraints and conservation laws\, such as com
patibility and equilibrium\, thus bypassing entire
ly the empirical material modeling step of convent
ional computing altogether. Data-driven solvers se
ek to assign to each material point the state from
a prespecified data set that is closest to satisf
ying the conservation laws. Equivalently\, data-dr
iven solvers aim to find the state satisfying the
conservation laws that is closest to the data set.
The resulting data-driven problem thus consists o
f the minimization of a distance function to the d
ata set in phase space subject to constraints intr
oduced by the conservation laws. We demonstrate th
e data-driven paradigm and investigate the perform
ance of data-driven solvers by means of several ex
amples of application\, including statics and dyna
mics of nonlinear three-dimensional trusses\, line
ar and nonlinear elasticity\, dynamics and plastic
ity\, including scattered data and stochastic beha
vior. In these tests\, the data-driven solvers exh
ibit good convergence properties both with respect
to the number of data points and with regard to l
ocal data assignment\, including noisy material da
ta sets containing outliers. The variational struc
ture of the data-driven problem also renders it am
enable to analysis. We find that the classical sol
utions are recovered as a special case of Data-Dri
ven solutions. We identify conditions for converge
nce of Data-Driven solutions corresponding to sequ
ences of approximating material data sets. Special
ization to constant material data set sequences in
turn establishes an appropriate notion of relaxat
ion. We find that relaxation within the Data-Drive
n framework is fundamentally different from the cl
assical relaxation of energy functions. For instan
ce\, we show that in the Data-Driven framework the
relaxation of a bistable material leads to effect
ive material data sets that are not graphs. I will
finish my presentation with highlights on work in
progress\, including experimental material data m
ining and identification\, material data generatio
n through multiscale analysis and fast search and
data structure algorithms as a form of ansatz-free
learning.\n\n
LOCATION:Zoom Meeting ID: 819 1682 8857
CONTACT:Hilde Hambro
END:VEVENT
END:VCALENDAR