Many areas of science now generate huge volumes of data that present visualization, modeling, and interpretation challenges. Methods for effectively representing the original data in a reduced coordinate space are therefore receiving much attention. The purpose of this research is to test the hypothesis that molecular computing of vectors for transformation matrices enables spectra to be represented in any arbitrary coordinate system. New coordinate systems are selected to reduce the dimensionality of the spectral hyperspace and simplify the mechanical/electrical/computational construction of a spectrometer. A novel integrated sensing and processing system, termed Molecular Factor Computing (MFC) based near infrared (NIR) spectrometer, is proposed in this dissertation. In an MFC -based NIR spectrometer, spectral features are encoded by the transmission spectrum of MFC filters which effectively compute the calibration function or the discriminant functions by weighing the signals received from a broad wavelength band. Compared with the conventional spectrometers, the novel NIR analyzer proposed in this work is orders of magnitude faster and more rugged than traditional spectroscopy instruments without sacrificing the accuracy that makes it an ideal analytical tool for process analysis. Two different MFC filter-generating algorithms are developed and tested for searching a near-infrared spectral library to select molecular filters for MFC-based spectroscopy. One using genetic algorithms coupled with predictive modeling methods to select MFC filters from a spectral library for quantitative prediction is firstly described. The second filter-generating algorithm designed to select MFC filters for qualitative classification purpose is then presented. The concept of molecular factor computing (MFC)-based predictive spectroscopy is demonstrated with quantitative analysis of ethanol-in-water mixtures in a MFC-based prototype instrument.
| Identifer | oai:union.ndltd.org:uky.edu/oai:uknowledge.uky.edu:gradschool_diss-1486 |
| Date | 01 January 2007 |
| Creators | Dai, Bin |
| Publisher | UKnowledge |
| Source Sets | University of Kentucky |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | University of Kentucky Doctoral Dissertations |
Page generated in 0.002 seconds