System and computer-implemented method for extrapolating calibration spectra
The present disclosure relates to a computer-implemented method for forecasting calibration spectra including a step of providing a machine learning model trained using historical calibration data corresponding to different gas species at different pressures. The computer-implemented method also includes steps of performing a calibration scan of one gas species at one pressure using an analyzer and generating calibration curves for the analyzer corresponding to one or multiple gas species at multiple pressures using the machine learning model and the calibration scan. Thereafter, a spectrum is obtained using the analyzer, and a concentration measurement is generated using the spectrum and at least one of the calibration curves.
1. A computer-implemented method for forecasting calibration spectra, comprising:
providing a machine learning model trained using historical calibration data corresponding to different gas species at different pressures;
performing a calibration scan of one gas species at one pressure using an analyzer;
generating calibration spectra for the analyzer corresponding to one or multiple gas species at multiple pressures using the machine learning model and the calibration scan;
obtaining a spectrum using the analyzer; and
generating a concentration measurement using the spectrum and at least one of the generated calibration spectra.
2. The computer-implemented method of claim 1 , further including generating calibration spectra for the analyzer corresponding to multiple gas species at multiple pressures.
3. The computer-implemented method of claim 1 , further including performing multiple calibration scans at one pressure, wherein each of the calibration scans corresponds to a different gas species at the one pressure using the analyzer.
4. The computer-implemented method of claim 1 , wherein the one gas is hydrogen sulfide, acetylene, ammonia, carbon dioxide, or water.
5. The computer-implemented method of claim 1 , wherein the one gas is oxygen, hydrogen chloride, methane, or carbon monoxide.
6. The computer-implemented method of claim 1 , wherein the one gas is methanol, ethane, ethylene, methyl acetylene, propadiene, nitrogen oxides, or sulfur oxides.
7. The computer-implemented method of claim 1 , wherein the machine learning model includes at least one of a neural network model, a partial least squares model, an inverse least square model, a classic least square model and a principal component regression model.
8. A system for forecasting calibration spectra, comprising:
computer hardware comprising:
at least one programmable processor; and
machine-readable medium storing instructions that, when executed by the at least one programmable processor, cause the computer hardware to perform operations comprising:
providing a machine learning model trained using historical calibration data corresponding to different gas species at different pressures;
performing a calibration scan of one gas species at one pressure using an analyzer;
generating calibration spectra for the analyzer corresponding to one or multiple gas species at multiple pressures using the machine learning model and the calibration scan;
obtaining a spectrum using the analyzer; and
generating a concentration measurement using the spectrum and at least one of the generated calibration spectra.
9. The system of claim 8 , further including generating calibration spectra for the analyzer corresponding to multiple gas species at multiple pressures.
10. The system of claim 8 , further including performing multiple calibration scans at one pressure, wherein each of the calibration scans corresponds to a different gas species at the one pressure using the analyzer.
11. The system of claim 8 , wherein the one gas is hydrogen sulfide, acetylene, ammonia, carbon dioxide, or water.
12. The system of claim 8 , wherein the one gas is oxygen, hydrochloric acid, methane, or carbon monoxide.
13. The system of claim 8 , wherein the one gas is methanol, ethane, ethylene, methyl acetylene, propadiene, nitrogen oxide, or sulfur oxide.
14. The system of claim 8 , further including providing a neural network model trained using historical calibration data corresponding to different gas species at different pressures and temperatures.
15. The system of claim 8 , wherein the machine learning model includes at least one of a neural network model, a partial least squares model, an inverse least square model, a classic least square model and a principal component regression model.
16. A computer program product comprising a machine-readable storage medium encoding instructions that, when executed by one or more programmable processors, cause the one or more programmable processors to perform operations comprising:
providing a machine learning model trained using historical calibration data corresponding to different gas species at different pressures;
performing a calibration scan of one gas species at one pressure using an analyzer;
generating calibration spectra for the analyzer corresponding to one or multiple gas species at multiple pressures using the machine learning model and the calibration scan;
obtaining a spectrum using the analyzer; and
generating a concentration measurement using the spectrum and at least one of the generated calibration spectra.
17. The computer program product of claim 16 , further including generating calibration spectra for the analyzer corresponding to multiple gas species at multiple pressures.
18. The computer program product of claim 16 , further including performing multiple calibration scans at one pressure, wherein each of the calibration scans corresponds to a different gas species at the one pressure using the analyzer.
19. The computer program product of claim 16 , further including providing a neural network model trained using historical calibration data corresponding to different gas species at different pressures and temperatures.
20. The computer program product of claim 16 , wherein the machine learning model includes at least one of a neural network model, a partial least squares model, an inverse least squares model, a classic least squares model and a principal component regression model.