Abstract:

In the entire production cycle, which consists of material development, shape design, and process optimization, data-driven methods offer new ways to infer the complex relationships between input and output variables at each step and apply the inferences for the efficient optimization. In this seminar, we will present a well-organized data-driven design strategy that takes into account design space size, data reliability, and data set size for a variety of design problems, as well as multiple research outcomes for material, structure, and process design. In particular, we will introduce three obstacles associated with the application of data-driven design: extrapolation, data-fidelity, and small datasets, and discuss how to overcome each one. Specific applications include digital composite optimization for high strength and toughness, phononic crystal optimization for maximum wave-focusing, optimization of bio-inspired composites for combined strength and toughness, injection molding process optimization with respect to production speed and defect rate, and a combined theoretical and data-driven approach to efficient particulate-reinforced composite design.

ZOOM

Join Zoom Meeting
https://oist.zoom.us/j/91014769125?pwd=dUk5RmVUQzllWkc1c1pmQ0YwRFo3dz09

Meeting ID: 910 1476 9125

Passcode: 481561