Files and links (1)
Free to read and download, Open Access
Near-infrared (NIR) spectroscopy is a widely used technique in quantitative analytical applications. Near-infrared spectroscopy is commonly used in clinical, environmental and industrial applications because of its compatibility with aqueous samples and with relatively thick samples. However, NIR spectra typically contain weak and highly overlapped spectral features which require multivariate data analysis techniques (chemometrics) to yield meaningful and chemically relevant information.
This dissertation consists of two main themes which include applications of NIR spectroscopy combined with chemometrics to (1) model temperatures in clinically relevant aqueous-based samples and (2) model temperature and moisture content in nylon-6,6 polymers. This research employed overtone and combination bands of C-H, O-H and N-H bonds situated in the 4000 - 5000 cm-1 region to develop partial least-squares (PLS) regression models to predict analyte properties such as temperature, concentration and moisture content.
The research described in the first part of this dissertation includes the development of a spectral preprocessing strategy based on the standard variate transform (SNV) and discrete wavelet transform (DWT) to isolate the low-frequency baseline information which carries the spectral features due to temperature fluctuations in aqueous-based samples. This approach was used to develop calibration models to determine the temperature of aqueous-based samples directly from their NIR spectra. This is an important development due to the fact that extreme temperature sensitivity of the underlying water bands can lead to poor quantitative analyte prediction results. These temperature models were developed using pH 7.4 phosphate buffer solutions spanning the range of 20 to 40.5 °C. Following the temperature models, a temperature-correction strategy based on the systematic pattern of concentration residuals was successfully developed to improve quantitative analyte predictions in aqueous-based samples. These analyte prediction models included glucose solutions and glucose-lactate mixture solutions prepared in pH 7.4 phosphate buffer. The computed temperature models gave excellent long-term prediction results. The temperature correction strategy gave promising results with the glucose solutions as well as the glucose-lactate mixture solutions.
The research presented in the second part of this dissertation includes the development of calibration models to determine the temperature and moisture content of a piece of nylon-6,6 polymer directly from its NIR spectrum combining SNV and DWT procedures followed by PLS regression. Both models gave good long-term prediction results and predicted well across different nylon-6,6 sheets. Computed moisture model provides a reliable and fast method to determine the moisture content of a nylon polymer when compared to existing techniques. Extended research towards polymer characterization including preliminary investigations of inhomogeneous nature of nylon polymers using infrared microscopy is documented in the latter part of this dissertation.