Cranford, Scott (2014) 2-D inversions of gravity data for multi-sided polygons using particle swarm optimization. Masters thesis, Memorial University of Newfoundland.
PDF (Thesis with accompanying files in PDF portfolio.)
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.
Since its introduction in 1995 particle swarm optimization (PSO) has been an area of great interest for many optimization problems, including geophysical inversions, as it does not use gradient information to find solutions. PSO is a global searching method, which, in principle, should avoid local minima, and find the global minima. PSO operates by moving a group of particles around the search space according to a simple mathematical formula which is dependent on particle positions and velocities. The particles movements are influenced by their own best known positions, pbest, and the overall best known position, gbest, which is updated each iteration. This thesis presents a method for inverting 2-D geophysical data using multisided polygons and PSO which has never been investigated in the literature. The first implementation of the algorithm presented below was designed to minimize several standard benchmark functions. The second implementation augmented the existing algorithm by including a forward modeling method, and model parameterization to invert 2-D gravity data. Investigations were then carried out to discover more robust methods of parameterization and constraint handling methods because unconstrained inversions tend to turn themselves inside-out, i.e., become infeasible or geologically implausible. A penalty function was found to be the best solution for constraint handling, with an appropriate penalty value. The algorithm was tested with ten-sided models and noisy data, and returned excellent inversion results for all obstacles presented to it.
|Item Type:||Thesis (Masters)|
|Additional Information:||Includes bibliographical references (pages 165-166).|
|Department(s):||Science, Faculty of > Earth Sciences|
|Library of Congress Subject Heading:||Mathematical optimization; Particles (Nuclear physics)--Mathematical models; Inverse problems (Differential equations)—Numerical solutions; Swarm intelligence|
Actions (login required)