Rapid method for simulating gas spectra using reversed PCR temperature calibration models based on Hitran data

J. Bak

    Research output: Contribution to journalJournal articleResearch

    Abstract

    A computer program was produced to make rapid simulations of CO gas spectra at a spectral resolution of 1 cm(-1) and at temperatures ranging from 295 to 845 K and concentrations from 5 to 400 mg/m(3). The program is based on loadings and scores from three principal component regression (PCR) temperature calibration models. Three sets of 12 Hitran-simulated high-density spectra, each set spanning the entire temperature range at constant concentrations (50, 150, and 300 mg/m(3)), were used as calibration spectra in the PCR temperature models. All the spectra were convoluted with a sine-squared instrumental line-shape function and reduced in the number of data points prior to PCR modeling. The simulated spectra, calculated on the basis of the PCR model parameters, were next scaled by using the areas of the spectra to represent the input concentration. The program simulates spectra very rapidly, similar to 7 ms, which includes saving the simulated spectrum counting 962 data points to the hard disk. This is an improvement in computational speed of a factor of more than 10(3) compared to performing all the calculations from scratch. The actual program can be used to simulate CO spectra representing a uniform slab of gas at various temperatures, concentrations, and pathlengths. The gain in speed of the calculations of the spectra is based on the fact that the PCR models include mathematical pretreatments and compress the data effectively.
    Original languageEnglish
    JournalApplied Spectroscopy
    Volume53
    Issue number11
    Pages (from-to)1375-1381
    ISSN0003-7028
    DOIs
    Publication statusPublished - 1999

    Fingerprint

    Dive into the research topics of 'Rapid method for simulating gas spectra using reversed PCR temperature calibration models based on Hitran data'. Together they form a unique fingerprint.

    Cite this