Date of Award

2017

Document Type

Thesis

Degree Name

Bachelors

Department

Natural Sciences

First Advisor

Shipman, Steven

Area of Concentration

Natural Sciences

Abstract

There is a bottleneck in the field of spectroscopy between data collection and analysis, which can be addressed by the fast, automated assignment of spectral data. The AUTOFIT program[1] was designed to serve that need by quickly matching rotational constants to spectra with little user input and supervision, but can be improved by incorporating an optimization algorithm in the search for a solution. The Particle Swarm Optimization Algorithm (PSO)[2] was chosen as a robust algorithm to implement. PSO is part of a family of optimization algorithms called heuristic algorithms, which seeks approximately the best answer[3]. This is ideal for spectroscopic data, where an exact, precise match may not be found. PSO was tested for robustness against five standard fitness functions[4][5] used to test optimization algorithms (Ackley’s, a generalized Schaffers’s F6, Schaffers’s F7, Rastrigin’s, and Schwefel’s fitness functions) and then applied to a custom fitness function created for spectroscopic data. This thesis will explain the Particle Swarm Optimization Algorithm and how it works, how Autofit was modified to use PSO, the fitness function developed to work with spectroscopic data, and benchmark results

Download

Share

COinS