BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Inverse design of multiscale structures and meta-materials by topo
 logy optimization and dehomogenization - Prof. Ole Sigmund\, Technical Uni
 versity of Denmark
DTSTART:20240614T130000Z
DTEND:20240614T140000Z
UID:TALK213337@talks.cam.ac.uk
CONTACT:46601
DESCRIPTION:The large geometrical freedom allowed by developments in addit
 ive manufacturing (AM) allows for realization of extreme\, complex structu
 res and has spurred a surge in the use and manufacturing of architected la
 ttice structures and meta materials. Apart from purely structural applicat
 ions\, inverse design and multiscale optimization has applications in ther
 mal\, optical\, hydraulic and many others.\n\nTopology optimization (TO) i
 s a highly efficient inverse design tool for optimizing complex problems m
 odelled by partial differential equations and builds on voxel-based design
  parameterizations\, deterministic optimization approaches based on adjoin
 t methods for computing gradients. TO has been implemented as interactive 
 apps (see TopOpt and TopOpt3D apps on AppStore) and has been applied to fu
 ll-scale airplane wing design with discretizations by billions of finite e
 lements on super computers.\n\nDifferent paths may be considered for savin
 g CPU costs. Although popular\, and showing promise in many areas\, AI or 
 CNN-based approaches have so far not been proven efficient in solving TO p
 roblems directly. Main issues are high cost of training and data generatio
 n (high break-even costs) and low generality (change of boundary condition
 s requires retraining).\n\nA promising path is multiscale optimization tha
 t makes use of the knowledge of the homogenized properties of extreme micr
 ostructures to perform the optimization on coarse meshes and sub-sequently
  realizing the multiscale structure by so-called dehomogenization. Dehomog
 enization is based on computer graphics techniques and the combined proced
 ure thus has potential to be solved interactively on apps\, as for the sin
 gle scale apps mentioned above\, albeit with much higher resolutions.\n\nT
 he talk will give a brief introduction to TO and a discussion of the relev
 ance of various optimization techniques\, followed by a deeper dive into m
 ultiscale and dehomogenization techniques that take stiffness\, strength a
 s well as local and global buckling stability of multiscale lattice struct
 ures into account. Time permitting\, recent results on design of meta mate
 rials with extreme non-linear responses as well as non-perfect realization
 s will also be included.
LOCATION:Department of Engineering - LR4
END:VEVENT
END:VCALENDAR