zahr2017icme-affiliates-poster.bib

@comment{{This file has been generated by bib2bib 1.98}}

@comment{{Command line: bib2bib -ob html/content/bib/zahr2017icme-affiliates-poster.bib -c $key="zahr2017icme-affiliates-poster" /Users/mzahr/mjzdb/bib/mjz.bib}}

@conference{zahr2017icme-affiliates-poster,
  abstract = {The design and control of engineering and scientific systems often calls for the optimization of a system that depends on the interaction between multiple physical phenomena. For example, en- ergetically optimal flapping motions, which have applications in the design of energy harvesting mechanisms, micro aerial vehicles, and the understanding of biological systems, amounts to an optimization problem that critically depends on the interaction between highly separated fluid flow and large structural deformations.
In this work, we develop an adjoint-based optimization framework for multiphysics problems governed by coupled partial differential equations. High-order spatio-temporal discretizations are used for the underlying PDE and quantity of interest to obtain highly accurate simulations at a reasonable cost. A specialized high-order implicit-explicit Runge-Kutta scheme is used for the temporal discretization to ensure single-physics solvers can be leveraged without having to design a complex monolithic discretization. The fully discrete adjoint equations corresponding to this primal solver are derived and implemented to yield very precise gradients of quantities of interest, a necessity to obtain fast convergence from gradient-based optimization solvers. This framework has been applied to a number of applications including the design of energetically optimal flapping motions, the design of energy harvesting mechanisms, and data assimilation to dramatically enhance the resolution of magnetic resonance imaging devices.
This framework has also been extended to address the challenges posed by stochastic optimiza- tion where the input data for the PDE is not known with certainty. Such problems require an ensemble of primal and adjoint solves at each optimization iteration and dramatically increases the computational cost. To address this issue, a globally convergent multifidelity optimization framework has been developed that is capable of reducing the cost of solving stochastic opti- mization problems by several orders of magnitude. The method will also be demonstrated on various applications in fluid control.},
  address = {Stanford, California},
  author = {Zahr, Matthew J.},
  booktitle = {2016 Stanford Computational Mathematics and Engineering Affiliates Meeting},
  conftype = {poster},
  date-added = {2017-12-16 18:40:20 +0000},
  date-modified = {2017-12-16 18:41:51 +0000},
  poster = {content/posters/zahr2017aa-affiliates-poster.pdf},
  project = {trammo, trammo:stochromopt},
  title = {Efficient {PDE}-constrained optimization using adaptive model reduction},
  year = {5/1/2016}
}

This file was generated by bibtex2html 1.98.