Recent News
UNM receives $1.5 million to support computational workforce development
May 8, 2023
Tapia elected to Computing Research Association Board of Directors
March 3, 2023
UNM computer science students take part in HPC competition
March 3, 2023
Computer science professor, student part of AI panel on March 8
February 24, 2023
News Archives
Large Scale Optimization Algorithms for Homeland Security Applications
March 10, 2005
- Date: Thursday, March 10, 2005
- Time: 11:00 a.m.
- Place: Woodward 149
Dr. Bart G. Van Bloemen Waanders <bartv@sandia.gov>
Optimization and Uncertainty Estimation Department Sandia National Labs.
Large scale optimization algorithms are demonstrated on a range of important Homeland security applications. In the event of a chemical/biological contamination event, optimization algorithms can be utilized to help reconstruct initial conditions from a sparse set of sensors located throughout the domain. Given correct initial conditions, accurate forward simulations can then be used to help with the mitigation procedures. Inversion problems of this type are ill-conditioned, underdetermined and therefore difficult to solve, especially if inversion parameters exist at every discretized point in the domain. Intrusive all-at-once approaches are utilized to take advantage of the internal linear algebra of the forward simulation and thereby providing the most computationally efficient solution technique. Numerical results for hypothetical contamination events in internal facilities, external flows and water distributions are presented. In addition, optimization algorithms are applied to a decontamination process for a simple internal facility. A brief overview of fundamental optimization methods and implementations will be presented in addition to numerical results that compare and contrast these different optimization methods and implementations.