Multivariate optical elements for nonlinear calibration
AssigneeHalliburton Energy Services, Inc.
InventorsMichael L. Myrick, Luisa Profeta
Life of this patent
Estimating the term…
Abstract
The present subject matter is direct to methodologies for calibrating data obtained from an optical analysis system. An initial calibration matrix of sampled analyte concentrations is modified using mean-centering techniques and selection of low and high analyte concentration spectra to produce a two-point calibration. A modified calibration matrix is produced by generating a non-linear calibration matrix by multiplying the initial calibration matrix by the two-point calibration. In an alternate embodiment, an initial multivariate optical element design is modified by iteratively adjusting the design based on standard error of calibration determination based on non-linerly fitted functions.
Claims12
A method for calibrating data produced by an optical analysis system, comprising:
- providing an initial calibration matrix of sample concentrations including a first predetermined number of analyte concentration levels and a second predetermined number of levels of an interferent species;
- mean-centering the samples at each analyte concentration level to produce vectors representing interferences for the individual levels of analyte concentration;
- calculating the mean spectrum of the first predetermined number of levels of low analyte concentration spectra;
- calculating the mean spectrum of the first predetermined number of levels of high analyte concentration spectra;
- establishing the calculated low analyte mean spectrum and the calculated high analyte mean spectrum as end-member spectra of a two-point calibration;
- generating a non-linear calibration matrix by multiplying the initial calibration matrix by the two-point calibration;
- and
- evaluating optical analysis system data based on the generated non-linear calibration matrix.
A method for calibrating data produced by a multivariate optical element analysis system, comprising:
- selecting an initial multivariate optical element (MOE) design;
- calculating the spectrum for the initial MOE design to determine the MOE vector;
- providing an initial calibration matrix of sample concentrations of a plurality of analyte concentrations and a plurality of levels of an interferent;
- calculating calibration spectra based on analyte concentration levels;
- calculating the dot product of the MOE vector with each calibration spectra;
- obtaining the best non-linear fitting function between the calibration concentrations and the dot product;
- determining the standard error of the calibration;
- and
- optimizing the standard error of the calibration by modifying the design of the MOE.
Dependencies are inferred from the claim text.