Design of data acquisition and analysis systems in near infrared spectroscopy: A virtual instrument approach.

摘要

A dual-channel NIR dispersive spectrometer has been assembled from commercially available component parts. In addition, an original interface for the instrument was written in the LabVIEW programming language, which was customized for both absorption and reflectance experiments. The interface not only controls and acquires spectral data from the instrument, it also manages the data storage and spectral display operations of the spectrometer.; An evaluation of the performance parameters of the spectrometer was carried out using a trichloromethane standard. The wavelength scale of the spectrometer was calibrated with the following trichloromethane wavelengths acquired to the nearest 0.4 nm on a standardized FT-NIR spectrometer: 1151 nm, 1411 nm, 1691 nm, and 1858 nm. Stray light levels at the above wavelengths were also evaluated. The evaluation revealed that the stray light levels in the spectrometer were dependent on the first order bandwidth of the long pass filter. The optimum spectral bandpass (or full-width-at-half-maximum) of the spectrometer was determined to be 10 nm at a slit width of 0.4 mm. The photometric accuracy of the spectrometer was determined to be 0.48%, 5.1%, 0.4% and 0.26% at 1511 nm, 1411 nm, 1691 run, and 1858 nm, respectively. The optimum transmittance that yielded the lowest photometric (concentration) error was determined to be 37% at 1691 nm.; The analytical capability of the instrument was evaluated by using it to explore the feasibility of determining the capsaicinoid content of chili pepper extracts using Partial Least Squares (PLS) regression. Despite the limited number of samples, results of the analysis show that the capsaicinoid concentration of the extracts (serrano, caribe, piquin, cherry hot, jalapeno, habanero, and thai) were highly correlated to their corresponding NIR wavelengths (R2 = 0.999).; Finally, a series of LabVIEW programs for chemometric analyses (Principal Components Analysis (PCA), Principal Components Regression (PCR), and PLS) have been written with particular emphasis to the analysis of NIR process data. The routines were tested by using them to analyze NIR absorbance spectra of diesel fuel samples for cetane number, total aromatics, boiling point at 50% recovery, viscosity, density, and freezing temperature. Plots of the predicted versus actual values for all the diesel fuel properties gave correlation coefficients better than 0.94 for both PCR and PLS calibration.