Date of Award
Master of Science
Department of Electrical and Computer Engineering
Eugene Santos, PhD
This research presents an autonomous and computationally tractable method for scientific process analysis, combining an iterative algorithmic search and a recognition technique to discover multivariate linear and non-linear relations within experimental data series. These resultant data-driven relations provide researchers with a potentially real-time insight into experimental process phenomena and behavior. This method enables the efficient search of a potentially infinite space of relations within large data series to identify relations that accurately represent process phenomena. Proposed is a time series transformation that encodes and compresses real-valued data into a well-defined, discrete-space of 13 primitive elements where comparative evaluation between variables is both plausible and heuristically efficient. Additionally, this research develops and demonstrates binary discrete-space operations which accurately parallel their numeric-space equivalents. These operations extend the method's utility into trivariate relational analysis, and experimental evidence is offered supporting the existence of traceable multivariate signatures of incremental order within the discrete-space that can be exploited for higher dimensional analysis by means of an iterative best-n first search.
DTIC Accession Number
Conrad, David M., "Data-Driven Process Discovery: A Discrete Time Algebra for Relational Signal Analysis" (1996). Theses and Dissertations. 5856.