ASSA Inversion Model Upgrades: Implementation of Raytrace AEL, and Bayesian Posterior Probability Density Distributions

PDF

Authors
  1. MacGillivray, A.
  2. Austin, M.
Corporate Authors
Defence R&D Canada - Atlantic, Dartmouth NS (CAN);Jasco Research Ltd, Victoria BC (CAN)
Abstract
The Adaptive Simplex Simulated Annealing (ASSA) Fortran code currently used by DRDC Atlantic to solve a series of geoacoustic inversion problems has been expanded to solve a new type of problem and to provide Bayesian statistical analysis of the model result uncertainties. The new problem type involves the localization of acoustic sources and receivers given relative arrival time data. A new forward model has been added to the program for solving the localization problem using a ray tracing approach. A new cost function, to be minimized in the optimization of the localization problem, has also been added to the program based on the root mean square (rms) time difference of the relative arrival times. The Bayesian analysis is performed using a Gibbs sampling approach to determine a posteriori probability density (PPD) distributions, providing estimates of model uncertainties for geoacoustic inversion problems. In order to preserve the general nature of the ASSA code and to allow future program enhancements, the new features have been added in a modular fashion and are pure extensions to the existing code that do not change the main functionality of the original ASSA program.

Il y a un résumé en français ici.

Keywords
ASSA;Adaptive simplex simulated annealing;Passive localization;Geo-acoustic inversion;Bayesian analysis;Inversion uncertainties
Report Number
DRDC-ATLANTIC-CR-2003-251 — Contractor Report
Date of publication
01 Jan 2005
Number of Pages
50
DSTKIM No
CA026435
CANDIS No
524298
Format(s):
CD ROM

Permanent link

Document 1 of 1

Date modified: