WO2023180141A1 - Procédé d'étalonnage d'un ou de plusieurs spectromètres - Google Patents

Procédé d'étalonnage d'un ou de plusieurs spectromètres Download PDF

Info

Publication number
WO2023180141A1
WO2023180141A1 PCT/EP2023/056629 EP2023056629W WO2023180141A1 WO 2023180141 A1 WO2023180141 A1 WO 2023180141A1 EP 2023056629 W EP2023056629 W EP 2023056629W WO 2023180141 A1 WO2023180141 A1 WO 2023180141A1
Authority
WO
WIPO (PCT)
Prior art keywords
spectrometer
spectra
calibration
transform
calibrated
Prior art date
Application number
PCT/EP2023/056629
Other languages
English (en)
Inventor
Jonathon SPEED
Victoria GRIGSON
Kiran HAROON
Original Assignee
Keit Limited
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Keit Limited filed Critical Keit Limited
Publication of WO2023180141A1 publication Critical patent/WO2023180141A1/fr

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
    • G01J3/28—Investigating the spectrum
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/27—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands using photo-electric detection ; circuits for computing concentration
    • G01N21/274—Calibration, base line adjustment, drift correction
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
    • G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
    • G01N2021/3595—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using FTIR
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2201/00—Features of devices classified in G01N21/00
    • G01N2201/12—Circuits of general importance; Signal processing
    • G01N2201/129—Using chemometrical methods

Definitions

  • the present invention relates to methods of calibrating a spectrometer such as molecular vibrational spectroscopy spectrometer.
  • the calibration is produced more quickly than conventional methods.
  • the methods may be expanded and used to calibrate multiple spectrometers such as may be found at an industrial process plant.
  • Instruments range from those that are designed solely for laboratory use to those that are more robust and designed to be used in a manufacturing environment.
  • the Michelson interferometer comprises a partially reflecting mirror 10 and a pair of reflecting mirrors 21 , 22.
  • the partially reflecting mirror 10 divides the light into two components.
  • the mirrors 21 , 22 and partially reflecting mirror 10 are arranged so that light passes along two linear paths, namely legs 31 , 32. Each component travels a return path along a different one of the legs.
  • the components recombine through the partially reflecting mirror and arrive at the detector 30 to form interference fringes.
  • the optical path difference between the two legs can be varied. As the path difference is changed the separation between fringes becomes greater or smaller. If the path length difference is varied linearly with time, the signal from the detector will vary sinusoidally with a period determined by the wavelength and path difference.
  • the calibration of a spectrometers generally requires two steps.
  • the first step is that of wavelength registration, which could also be described as wavelength calibration.
  • a known wavelength standard is used to set the wavelength position and scale for measurements on the spectrometer.
  • the second step is that of intensity or absorption calibration, and uses chemicals with known absorption characteristics.
  • the wavelength registration or calibration of a Michelson interferometer commonly uses a reference light source such as a helium-neon laser.
  • the reference light source is different to the source used for the spectroscopy itself.
  • a broadband infrared light source may be used.
  • the helium-neon laser calibration source is shown schematically at 40.
  • a helium-neon laser has a wavelength of 632.8nm.
  • the Michelson interferometer relies on path difference between the two legs of the interferometer. For a single wavelength of 632.8nm the signal at the detector will peak each time the path difference crosses a multiple of this wavelength. The peaks will occur in a similar manner to if there were multiple wavelengths present in the source each separated by 632.8nm. This principle can be used in calibrating a Michelson interferometer over a wide wavelength range such as from 2-25pm.
  • the intensity or absorption calibration measures the absorption of a number of chemicals spread across the wavelength range of interest, which are then compared to known standards.
  • a single chemical such as a polystyrene film may be used which has lots of chemical markers in the single chemical.
  • the known relative intensity of different peaks in the spectrum of a chemical can be used to set the absorption measurements for the spectrometer being calibrated.
  • interferometers based on light, such as infrared, travelling around a loop have been developed.
  • light or infrared travelling around a loop.
  • the light or infrared is again split into two components which travel slightly different paths around a loop.
  • the paths are similar offset triangular paths.
  • the detector for this arrangement is different to that of the Michelson interferometer.
  • the Michelson interferometer simply requires a single pixel detector, whereas the arrangement of figure 1B requires multiple pixels in a linear or rectangular array.
  • the two components of light, or infrared, on reaching the detector 30’ have already travelled different path lengths and so the interference pattern is spread across the detector 30’.
  • the interference pattern is spread across the detector and so all wavelengths are measured simultaneously.
  • the technique of calibrating across a wide wavelength rage using a single, for example, helium-neon source is not possible.
  • spectrometers according to the arrangement of figure 1 B a number of factors provide slight differences in measurements from one spectrometer to another. These factors include the relative position and angle of the detector, the sensitivity of the detector, and differences in sensitivity from one pixel to another.
  • the arrangement of figure 1 B also includes mirrors 2T and 22’ which are curved and provide some focusing effect on the beams. Another variability between spectrometers is that the curvature of the mirrors from one spectrometer to another may differ slightly.
  • One area where it is particularly advantageous for spectrometers to provide calibrated outputs that are directly comparable, that is any difference between spectrometers is substantially eliminated, is where multiple spectrometers are used on a single industrial process.
  • Another area is where a process has, for example, multiple identical process lines, chamber or tanks and it is desirable that measurements from spectrometers are directly comparable.
  • the present invention provides a method of calibrating a spectrometer, comprising: acquiring for a first spectrometer, the first spectrometer being a calibrated reference spectrometer, first spectra of one or more reference chemicals; acquiring for a second spectrometer, the second spectrometer being a spectrometer to be calibrated, second spectra of the one or more reference chemicals; generating a calibration transform or calibration matrix from the first spectra and the second spectra, the calibration transform providing a calibration for the second spectrometer, wherein the calibration transform translates spectra from the second spectrometer to spectra that are calibrated based on the calibration of the first spectrometer; and uploading the calibration transform to the second spectrometer.
  • the spectra that are translated, which are generated on the second spectrometer may be considered to be raw spectra.
  • the spectra may have been processed from absorption spectra to absorbance spectra by eliminating loss or absorbance of the optics of an interferometer in the spectrometer.
  • the method of using a calibration transform provides a faster way to achieve a high quality calibration on a spectrometer than the conventional manner of making numerous detailed measurements of many chemicals. This conventional approach has been applied to the reference or master spectrometer and this detailed information is, at least partly, re-used in the calibration transform for the new or second spectrometer.
  • This re-use of calibration data from the reference or master instrument results in the calibration for the second or new spectrometer being produced much more rapidly than for the reference or master spectrometer.
  • the use of spectra of reference chemicals generated by the second or new spectrometer tailors the calibration from the master and allows it to be used for the second spectrometer.
  • the spectra from the slave spectrometer that have been translated based on the calibration transform may be known as a translated spectrum.
  • the translated spectra of a given sample will have features, in terms of wavenumber and absorbance that substantially correspond to the features of spectrum of the same sample measured on the master or reference spectrometer.
  • spectra are generated, for example, with respect to frequency or wavelength, or absorbance or intensity.
  • peaks will have substantially the same width or wavenumber ranges and relative height or absorption as if measured on the master spectrometer.
  • the second spectrometer that is the spectrometer to be calibrated, is described as a slave spectrometer. However, this does not mean that it requires the master or reference spectrometer to operate, but just that calibration transfer has occurred from the master to it.
  • Embodiments described herein indicate that the first spectra and second spectra, and other spectra mentioned, are generated by molecular vibrational spectroscopy. However, other spectroscopic techniques are also possible. Nevertheless, in preferred embodiments the first spectra, the second spectra and other spectra are generated using the same spectroscopic technique, such as infrared spectroscopy, near infrared spectroscopy, mid infrared spectroscopy, UV-visible spectroscopy or Raman spectroscopy. The first spectra and the second spectra may be generated by FTIR spectroscopy.
  • the first spectrometer which is a calibrated reference spectrometer
  • the second spectrometer which is a spectrometer to be calibrated
  • the first spectrometer which is the reference spectrometer, may be a virtual spectrometer and the first spectra may comprise spectra collected from a plurality of spectrometers.
  • the plurality of spectrometers forming the virtual spectrometer may also be of substantially the same design.
  • the first spectra and second spectra may be collected over the same wavelength, frequency or wavenumber range. If there is a difference between the first and second spectra then the spectra having the wider range maybe reduced so that the ranges match.
  • the spectra of the one or more reference chemicals may comprise at least two spectra of each of the one or more reference chemicals.
  • the method may further comprise verifying that the second spectra of the one or more reference chemicals generated by the spectrometer to be calibrated are within an acceptance range of the first spectra of the one or more reference chemicals for the reference spectrometer. This verification may be performed using principal component analysis and/or determining a confidence interval for the first spectra in to which the second spectra should fall.
  • the one or more reference chemicals may comprise one of more of: water, isopropanol, acetone, decane, decanol, hexane, hexanol, toluene, methyl benzoate, ethylene glycol and a sample collected from a process.
  • the step of generating the calibration transform may comprise generating the calibration transform using one of: a spectral space transform (SST); direct standardization (DS); and piecewise direct standardization (PDS).
  • SST spectral space transform
  • DS direct standardization
  • PDS piecewise direct standardization
  • the step of generating the calibration transform may comprise removing noise from the spectra, determining differences between the first spectra and second spectra, and based on the differences determining the calibration transform to transform spectra from the second spectrometer such that they are calibrated as if measured on the first spectrometer.
  • the step of generating the calibration transform using a spectral space transform may comprise: generating a first matrix of data of the first spectra for the one or more reference chemicals obtained for the reference spectrometer; generating a second matrix of data of the second spectra of the one or more reference chemicals generated by the spectrometer to be calibrated; combining the first matrix and second matrix to from a combined matrix, X CO mb; performing Single Variable Decomposition or other decomposition techniques to separate noise from the combined matrix; partitioning the resultant matrix into two submatrices corresponding to the spectral data of the first matrix and the second matrix; and uploading the two submatrices to the second spectrometer along with an algorithm to provide the calibration transform.
  • the step of generating the calibration transform may comprise using the spectral space transform and may further comprise selecting the number of factors to be used in the SVD based on the number of reference chemicals.
  • the method my further comprise augmenting the calibration transform with spectra of one or more samples collected on the second spectrometer. This augmentation increases the accuracy of the translated spectra.
  • the step of augmenting may comprise: obtaining, from the second spectrometer, spectra of one or more samples; translating the spectra of the one or more samples using the calibration transform; adding the spectra obtained from the second spectrometer and the translated spectra to the first and second spectra of reference chemicals; regenerating the calibration transform based on the added spectra and the first and second spectra to produce an augmented calibration transform; and uploading the augmented calibration transform to the second spectrometer. Accordingly, the augmentation requires only a small number of extra spectra to be collected and it is only required to be collected using the second spectrometer.
  • Additional corresponding spectra can be collected using the master or reference spectrometer but this amounts to an extension of the number of reference chemicals for which spectra are collected in preparing the original calibration transfer. Furthermore, it is not always possible to have the reference spectrometer present at this stage so use of additional spectra from only the spectrometer being calibrated provides convenience to this augmentation.
  • the one or more samples for the augmentation may be obtained from the process such as an industrial or biotechnological process.
  • the number of spectra of the one or more samples may be less than the number of first spectra and/or second spectra.
  • the method may further comprise applying a single calibration to a plurality of spectrometers.
  • the spectrometers to be calibrated may comprise the second spectrometer and a third spectrometer.
  • the third spectrometer may have a calibration transform generated according to the method set out above for second spectrometer.
  • the method may further comprise: obtaining, from the second spectrometer, spectra of one or more samples; translating the spectra of the one or more samples using the calibration transform of the second spectrometer; adding the spectra obtained from the second spectrometer and the translated spectra from the second spectrometer to the first and second spectra of reference chemicals; obtaining, from the third spectrometer, spectra of one or more samples; translating the spectra of the one or more samples using the calibration transform of the third spectrometer; adding the spectra obtained from the third spectrometer and the translated spectra from the third spectrometer to the first and second spectra of reference chemicals; regenerating the calibration transform based on the added spectra from the second spectrometer, third spectrometer and the first and second spectra to produce an augmented calibration transform; and uploading the augmented calibration transform to the second and third spectrometers.
  • the one or more samples of which spectra are obtained by the third spectrometer are preferably substantially the same chemicals.
  • they may be samples obtained from different but corresponding parts of a process.
  • a process may be split into multiple sub-processes where each sub-process performs the same function.
  • Each sub-process may start with same chemicals and aim to end with the same output.
  • each sub-process may include a fermentation tank and respective spectrometers measure the contents in the fermentation tanks.
  • the one or more samples may be obtained from the process.
  • the number of spectra of the one or more samples may be less than the number of first spectra and/or second spectra.
  • the present invention provides a method of providing a calibrated spectrum of a sample by a spectrometer, the method comprising: calibrating the spectrometer according to any of the methods set out above, wherein the spectrometer is the second spectrometer; generating a spectrum of a sample using the second spectrometer; translating the spectrum of the sample using the calibration transform; and outputting the translated spectrum to a display or computing device.
  • the method may further comprise: initializing the second spectrometer, wherein initializing comprises: loading the calibration transform into a memory of the second spectrometer; and generating an absorption spectrum using the second spectrometer, in the absence of a sample, to determine the absorbance of spectrometer components in a beam path within the second spectrometer, the method further comprising: generating, by the second spectrometer, an absorption spectrum representing a raw spectrum of the sample, the raw spectrum collected by a spectrometer using molecular vibrational spectroscopy; creating an absorbance spectrum based on the absorption spectrum in the absence of a sample and the absorption spectrum representing a raw spectrum; translating the absorbance spectrum using the calibration transform to generate a translated absorbance spectrum of the sample.
  • raw spectrum we may mean a spectrum that has undergone no or minimal processing such that the data is that received directly from the detector of the spectrometer.
  • the method may further comprise outputting the translated absorbance spectrum of the sample.
  • the method may further comprise comparing the translated absorbance spectrum with a chemometric model and outputting determined characteristics of the sample based on the comparison to the chemometric model.
  • the present invention provides a computer program that when executed cause a computer, analyser, spectrometer or processor to perform any of the methods set out herein.
  • the present invention provides computer readable media, such as non-transitory computer readable media, having stored thereon instructions, such as computer program code, that when executed cause a computer, analyser, spectrometer or processor to perform any of the methods set out herein.
  • the present invention provides a spectrometer comprising an interferometer, a memory and a processor, the spectrometer configured to perform any of the methods et out herein.
  • the present invention provides a spectrometer, the spectrometer calibrated according to any of the calibration methods set out above, the spectrometer may comprise an interferometer and broadband infrared source such as mid-infrared source, the spectrometer may be configured to direct light or radiation from the source through a sample and analyse the light or radiation that has passed through the sample using an interferometer, the spectrometer configured to: generate data relating to a spectrum of a sample; translate the data or spectrum of the sample using the calibration transform; and output the translated spectrum to a display or computing device.
  • the spectrometer may comprise an interferometer and broadband infrared source such as mid-infrared source
  • the spectrometer may be configured to direct light or radiation from the source through a sample and analyse the light or radiation that has passed through the sample using an interferometer, the spectrometer configured to: generate data relating to a spectrum of a sample; translate the data or spectrum of the sample using the calibration transform; and output the translated spectrum to a display
  • the present invention provides apparatus comprising: a spectrometer for performing molecular vibrational spectroscopy; and a computer comprising a processor and memory, wherein the apparatus is configured to store and process the computer program set out above.
  • the first or second spectrometer, or any other spectrometer referenced herein may comprise an interferometer based on a Sagnac interferometer in which an input beam is divided and directed around opposing paths, each comprising a loop, the paths having a path difference therebetween and being incident on a detector.
  • the detector may be arranged to monitor interference patterns produced thereon.
  • the interferometer may comprise comprising path differencing optics and a detector, wherein: the path differencing optics comprise at least two reflecting elements and a beamsplitter, the beam splitter arranged to divide an input beam into first and second beams respectively traversing first and second paths towards the detector; first and second beams respectively directed to first and second reflecting elements, the first reflecting element arranged to direct the first beam to the second reflecting element, the second reflecting element arranged to direct the first beam to the beamsplitter and on towards the detector; the second reflecting element arranged to direct the second beam to the first reflecting element, the first reflecting element arranged to direct the second beam to the beamsplitter and on along a second path towards the detector, the first and second paths having an optical path difference; and the detector is arranged to detect a pattern produced by interference of the first and second beams, wherein the first and second reflecting elements are concavely curved to focus the first and second beams onto the detector.
  • figure 1A is schematic diagram of a Michelson interferometer
  • figure 1B is a schematic diagram of a Sagnac based interferometer
  • figure 2 is a flow diagram of a method of calibrating a spectrometer using a calibration transform and using the spectrometer to provide calibrated output data
  • figure 3 are infrared absorption spectra of three reference chemicals: water, acetone and isopropanol
  • figure 4 is a plot of principal component, PC1 and PC2, values of the reference spectra from a master instrument and instrument to be calibrated
  • figure 5 is a flow diagram of a method of generating a calibration transform
  • figure 6A is a bar chart showing the weighting of SVD factors
  • figure 6B is plot showing the first SVD factor generated from the slave and master data
  • figure 7 is a graph showing how the raw data from the instrument to be calibrated is translated to the master, and is compared to the
  • figure 8 is a flow diagram of a method of calibrating a spectrometer with an augmented calibration transform and using the spectrometer to provide calibrated output data
  • figures 9A and 9B schematically show the respective spectra that may be used in producing a calibration transform with and without augmentation
  • figures 10A and 10B are graphical plots demonstrating the fit of translated and augmented data in comparison to a master spectrometer
  • figure 11 shows six graphs indicating how calibration augmentation improves the accuracy of calibration transfer
  • figure 12 is a flow diagram of a method of calibrating a plurality of spectrometers using a calibration transform and augmentation across the multiple spectrometers
  • figure 13 is a schematic diagram of a spectrometer coupled to a process
  • figure 14 is a flow diagram of an example of initializing a spectrometer through to providing a chemometric output.
  • Figure 2 is a flow diagram of a method of calibrating a spectrometer.
  • the master instrument is known to have a high quality calibration which may have been built by taking many spectra of a broad range of chemicals. Such calibrations are time consuming to build.
  • the present invention aims to build a corresponding calibration for another instrument with good accuracy but over much shorter timescales.
  • the master spectrometer and the new spectrometer are of the same design or type, although the present methods are not limited thereto.
  • the new spectrometer that is to be calibrated may be considered to be the “slave” spectrometer.
  • the reference spectra collected on the master spectrometer preferably comprise one or more spectra of each of one or more chemicals.
  • the chemicals we will refer to as reference chemicals.
  • the one or more chemicals have features such as absorption peaks, in the spectra over at least the desired range of wavenumber over which the spectra are taken, and more preferably have features across the full wavenumber or frequency range of operation of the spectrometer.
  • the spectrometer to be calibrated may be a newly manufactured spectrometer which is ready to undergo acceptance tests to verify that it is operating correctly.
  • the acceptance tests may include taking spectra of a number of chemicals and then performing some checks on the resulting spectra.
  • the checks may include performing a peak-fit on the spectra and analysing the signal-to-noise at different points on the spectra, the peak centre and the peak width (such as measured by full width half maximum). Since in the method of figure 2 it is required to obtain spectra of the same reference chemicals acquired for the master instrument it is preferable that the chemicals used in the acceptance tests are the reference chemicals.
  • the reference chemicals may be selected based on a number of different factors. It is preferable that the chemicals used are readily available in relatively pure form. Using higher purity reduces the impurities such that for higher purity chemicals there is greater confidence that the chemical being measured is the same when calibrating one device compared to another. Furthermore, the chemicals preferably should be inexpensive, readily available and non-toxic. In one embodiment, it is preferable to use the following chemicals:
  • spectra will be available that have been collected on the master spectrometer.
  • spectra are generated of the reference chemicals.
  • at least two spectra are collected of each chemical.
  • Figure 3 shows spectra collected on an instrument to be calibrated. Three spectra have been collected for each of three chemicals.
  • the three chemicals are water, isopropanol and acetone. As can be seen between the three chemicals they have features or peaks that are spread across the frequency range of operation of the spectrometer.
  • the spectra of water shows the wide peak at around 3300cm' 1 for the OH bond stretching in water.
  • the isopropanol spectra show OH stretching at around 3300cm' 1 , C-H stretching at around 3000cm' 1 and various other peaks.
  • Having spectra of the same chemicals from the master spectrometer and the spectrometer to be calibrated allows the two instruments to be compared. That is, any difference seen between the spectra of the same chemical from instrument to the other is very likely to be due to differences between the instruments themselves. This difference is used to build a calibration transform so that we can make the spectrometer to be calibrated look like the master spectrometer.
  • the chemicals are not limited to those listed above, but may also be samples taken from a process such as an industrial process or biotechnological process. Alternatively, references samples that look similar to process samples may be made up in the laboratory.
  • the chemicals in the second list above have application in specific chemical sectors.
  • decane has application in the petrochemical industry as do other alkanes.
  • Methyl benzoate has a similar chemical structure to polystyrene and polystyrene film is widely used by spectrometer manufacturers for calibrations.
  • Methyl benzoate is a liquid and so can be used in arrangements where the chemical is required to be a liquid. In one example, it is considered that decane and hexane could be used together as chemicals for calibrating a spectrometer for the petrochemical industry.
  • the next step in figure 2 is to build the calibration transform.
  • PCA principal component analysis
  • MD multivariate decomposition
  • MCR multivariate curve resolution
  • the model may be built using data from a number of instruments and a number of chemicals. This would produce a model of a “virtual” master.
  • the master spectral data may all be from a single master spectrometer.
  • the PCA model determines principal components to fit the master data.
  • the first and second principal components describe the main variation across the data.
  • Each set of data can be described by a score for each principal component and a residual.
  • the score is a measure of how well or how much of the principal component is used to fit the spectra to the model.
  • the residuals relate to any difference remaining when the fit to the principal components has been taken into account.
  • Figure 4 shows results of scores for reference spectra of the reference chemicals obtained from a number of instruments.
  • Figure 4 shows the scores resulting for three spectra from each of three reference or master instruments.
  • the scores in relation to PC1 and PC2 for nine reference spectra for each of water, isopropanol and acetone are shown.
  • the PCA model has been built based on these nine reference spectra, which thereby provide a “virtual” master instrument.
  • Spectra obtained from the instrument being calibrated from three measurements of each of the three reference chemical have been projected on to the model and their PC1 and PC2 scores obtained and added to the plot of figure 4.
  • Figure 4 shows an ellipse around the PC data relating to each reference chemical.
  • the ellipse represents two standard deviations of the range of master values centred about the mean. Two standard deviations equates to a 95% confidence interval of the data points. Hence, the ellipse is a 95% confidence interval. If the PC data from the instrument being calibrated lies within this ellipse the spectra provided by that instrument may be considered to be acceptable. If the spectra lie outside of this range something is wrong and new reference spectra from the instrument being calibrated will need to be obtained. For example, the reference chemicals may have been contaminated. If the new spectra also show problems then there may be a problem with the instrument.
  • Figure 4 also indicates why it is desirable to have more than one spectrum for the reference chemicals. With just a single measurement it is not possible to determine a standard deviation and confidence ellipse around the PC data. Hence, for a data point from the instrument being calibrated falling a distance away from the master data points it is not possible to determine if the distance away indicates a problem with the calibration data. Hence, it is preferable that two or more reference spectra for the reference chemicals measured by the master spectrometer are provided.
  • the next step 130 in the flow diagram is to build a calibration transform.
  • the figure indicates an SST matrix, Spectral Space Transform, as the type of calibration transform and we will describe this in more detail below.
  • Other types of transform are available such as Direct Standardization (DS) and Piecewise Direct Standardization (PDS).
  • the SST transform is used to make one instrument or set of test conditions look like another.
  • this approach is used to make a new, uncalibrated instrument look like a master instrument that has a proven high standard calibration.
  • the approach avoids having to go through a lengthy calibration process for the new instrument by using the data that has been built up on the master instrument.
  • Xi and X 2 are matrices of spectra for a number of samples respectively measured under calibration and test/sampling conditions
  • C is a concentration matrix representing concentrations of spectroscopically active components in the samples
  • Si and S 2 are matrices of the spectra of pure chemical components
  • Ei and E 2 are residuals matrices.
  • Xtrans Xtest (S2 T ) + Sl T + Xtest - Xtest (S2 T ) + S2 T where the superscript “+” represents the Moore-Penrose generalized inverse.
  • Figure 5 shows in a flow diagram how this works in practice.
  • steps 210 and 220 spectra of reference chemicals obtained from the instrument to be calibrated and spectra of reference chemicals obtained from the master instrument are input to the SST generator 240.
  • the spectra from the master will have been calibrated using the calibration of the master instrument.
  • To determine the SST matrix we need to provide to the SST generator the number of variables to use in the Single Variable Decomposition. This is very likely to correspond to the number of chemicals being used as the reference chemicals. Hence, for our example of using water, isopropanol and acetone, this would be three.
  • the number of factors may be less than the number of chemicals and if, in chemometric space, the reference chemicals are orthogonal then the number of factors may be more. When evaluating the SVD more factors are often found than the number of reference chemicals. However, these additional factors represent noise. By limiting the number of factors used noise can be eliminated to avoid it being mapped from one spectrometer to another. Typically, it would be expected to use a number of factors of between 1 and 3 but could be as many as 10 depending on the number of, and which, reference chemicals are used. In general the number of factors used may be limited to those factors which are greater than a threshold value, such as 0.2 or 0.1 .
  • Figure 6A shows the values of factors in an SVD analysis for three reference chemicals.
  • the values of the first three factors are significant whereas the values of subsequent factors can be seen to be very small.
  • the fourth factor can be seen to be nonzero but is not used in the subsequent analysis because it relates to noise.
  • Figure 6B is a graph showing how one of the factors varies with wavenumber (in cm -1 ).
  • the factor selected is the first factor.
  • the plot shows the value for that factor for the instrument being calibrated, which is marked as “slave” in the figure, along with the value for the master instrument. It is possible to show the master and slave values since the SVD analysis is performed on the master and slave data together.
  • the values for the first factor, shown in the figure, are the biggest changes for these spectra.
  • Figure 6B shows that there is a significant difference at around 3300cm' 1 .
  • the slave has a higher output and so needs to be downweighted, resulting in the larger negative value at this wavenumber. A similar but smaller difference can be seen at 1250cm' 1 .
  • figure 6B shows that the SVD analysis identifies the differences. These differences are then corrected by the SST transform.
  • Figure 7 is a graph showing how a spectrum from the instrument to be calibrated (shown as “slave”) is translated towards the master spectrometer.
  • the narrow line shows the raw spectrum of one of the reference chemicals, in this case isopropanol, as measured by the spectrometer to be calibrated (marked as slave on graph).
  • the thick line is the spectrum of the same reference chemical measured on the master instrument.
  • the dashed line is the spectrum from the instrument to be calibrated after translation using the SST algorithm to look like it was measured on the master instrument.
  • the dashed line for the translated data largely overlies the thick line for the master instrument data. Hence, there is a very good match between the master and translated data and the differences between the untranslated spectrum and the master have been eliminated. The match remains strong across the whole range of wavenumbers of the spectrum.
  • the step of outputting the SST transform is shown as optional.
  • the step of generating the calibration transform may be performed on a separate computer and then output from the computer and uploaded to the spectrometer being calibrated.
  • the reference data from the master spectrometer may be loaded into the spectrometer being calibrated and the spectrometer being calibrated itself generates the SST transform.
  • Figure 2 which we referred to earlier and is the flow diagram of how to calibrate and use a spectrometer once calibrated, also has a step 140 of uploading or installing the calibration transform to the spectrometer.
  • the step of loading or storing the calibration transform may comprise storing the transform in the firmware of the instrument . This means that every spectrum that is acquired on the spectrometer is made to look like the master instrument and this happens continuously and in real time. By configuring the spectrometer in this way there is no need to have to take the data away and perform a calibration off-line. The calibration happens in the instrument immediately after or even during data acquisition. In this way it is possible that only a translated spectrum is ever made available to the customer.
  • the calibration transform stored in the spectrometer, there is no possibility of using the wrong calibration transform as is possible if the calibration is performed off-line at a separate computer.
  • the spectrometer is turned on, as the spectrometer software initializes, the calibration transform is automatically accessed and loaded to the accessible memory and/or processor ready for use.
  • the final two steps in figure 2 are to acquire data 150 of a sample on the newly calibrated spectrometer and to output 160 a translated spectrum of the sample.
  • the sample may be a sample collected from an industrial or biotechnological process line.
  • the method of calibration transfer described above may further include augmenting the calibration derived from the master instrument with data from the instrument being calibrated.
  • the spectrometer being calibrated we refer to the spectrometer being calibrated as the slave spectrometer. However, this should not be taken to mean that the slave spectrometer is run and controlled from the master. It is simply used as a convenient way of describing the instrument being calibrated.
  • Steps 710 and 720 correspond to the steps 110 and 120 of figure 2. That is, they comprise at step 710 acquiring spectra of the reference chemicals on the calibrated master or reference instrument and at step 720 acquiring spectra of the reference chemicals on the instrument to be calibrated, that is, the slave instrument.
  • Step 730 comprises augmenting the data for the calibration with further data collected on the slave instrument. This may comprise data from, or representative of, the process and chemicals to which the spectrometer will be deployed. For example, the slave spectrometer may collect spectra of ethanol if the slave is to be deployed to a process where ethanol is to be measured.
  • spectra of reference chemicals were collected on a master spectrometer. In one example, we described that this might be nine spectra for each of the three chemicals: water, isopropanol and acetone. As these spectra have been collected on the master instrument they will be calibrated. Additionally, spectra of the same three reference chemicals are generated by the instrument to be calibrated. These two groups of spectra are shown schematically in figure 9A. In embodiment of figures 2 to 7, the calibration transform is built on this data.
  • additional spectra are collected on the instrument to be calibrated. These spectra are collected after having first generated the calibration transform of the embodiment of figures 2 to 7. Hence, the additional spectra generated by the instrument to be calibrated have been processed by the calibration transform such that the spectra appear as they have been generated by the master spectrometer.
  • This translated data is indicated at 910 in figure 9.
  • the raw data measured on the spectrometer to be calibrated is indicated at 920 in figure 9.
  • the translated data along with the raw data are added to the master and slave data as shown in figure 9 such that the translated spectra are added to the master data Xi and the raw spectra are added to the slave data X2.
  • the calibration transform generator is then operated again to generate an augmented calibration to include this data.
  • the additional spectra collected for the augmentation may be from a sample from the process that is going to be monitored by the new instrument, for example, ethanol.
  • the spectra may be collected on a laboratory version of the sample or on another chemical that is different but relevant to the process.
  • step 750 the steps of acquiring data on the calibrated instrument, step 760, and outputting a translated/calibrated spectrum, step 770, correspond to steps 150 and 160 in figure 2.
  • Figure 10A is a graph comparing three types of calibration.
  • the graph uses spectral data that has been processed by principal component analysis (PCA).
  • PCA principal component analysis
  • the spectra of ethanol have been used to build a principal component analysis (PCA) model.
  • PCA principal component analysis
  • one of the principal components will map to ethanol in chemometric space.
  • the spectra collected on the master spectrometer are projected on to the model. Scores of the spectra against the principal component corresponding to ethanol will provide an indication of the amount of ethanol present.
  • these data points are plotted as diamonds and make up the majority of the data points.
  • the vertical axis represents a determined amount of ethanol based on the PCA analysis and scores.
  • the horizontal axis corresponds
  • FIG. 10A further data is added that has been collected using a slave spectrometer.
  • the square data points (largely in the range 5 to 10 of the horizontal axis) are based on spectra of process samples that have been measured on a slave spectrometer, which has been calibrated based on the calibration of the master spectrometer with augmentation.
  • the augmentation comprises additional raw data from the slave that is added to the slave reference data (X2) and corresponding spectra taken on the master instrument and added to the master reference data (X-i).
  • the triangular data points have been produced in a similar way using augmentation.
  • the samples are laboratory samples, for example in ranges of ethanol concentration that it wasn’t possible to obtain from the process.
  • triangular shaped data points are for laboratory samples and also use a calibration augmented with slave and master data.
  • the data points in figure 10A have undergone partial least squares (PLS) analysis to fit a line to the data.
  • Figure 10B shows how well the data points fit the line.
  • the dashed ellipse is a 95% confidence interval and both groups of translated and augmented data, namely the square and triangular data points fall within the ellipse.
  • calibration transfer and augmentation is effective in making slave data look like master data.
  • Figure 11 shows more clearly how augmentation can be used to improve the accuracy of calibration transfer.
  • Figure 11 shows six graphs, respectively for (a) glucose, (b) xylose, (c) brix (a solution of aqueous dissolved sugars used as a standard), (d) lignin, (e) acetic acid and (f) furfural.
  • the data from a purely translated measurement is shown on the left hand plot and data from a translated and augmented calibration with 99% master data and 1% slave data is shown on the right.
  • the magnitude of the error is dramatically decreased showing how much improved the calibration is. For example, for glucose the magnitude of the error is reduced by around a factor of eight.
  • the amount of augmentation is small.
  • the amount of slave data added to the master data for the calibration may range from 1% to 25% but is more typically 1% to 10% and preferably as little as 1% to 5%.
  • FIG 12 is a flow diagram of another method according to the present invention.
  • multiple instruments are used to build a calibration.
  • the approach is particularly beneficial if a single calibration is used for multiple instruments, because when two or more instruments are required it decreases the need to maintain multiple calibrations. For example, consider a customer with eight fermenters and a spectrometer for each one. If a single calibration is built it can be built with equal weighting for all eight fermenters. That single calibration can be used on all eight instruments.
  • the flow chart of this embodiment sets out the method for achieving this. The method starts with 810 and 820 which correspond to the first two steps in the flow charts of figures 2 and 8.
  • Spectra for reference chemicals are obtained on a master instrument, which may be a virtual instrument. Spectra for the same reference chemicals are obtained for the instrument to be calibrated. Here this instrument is designated as instrument x. Similar to figures 2 and 8 a calibration transform is built and uploaded to the instrument x, as shown at step 830. The same process may be performed with other instruments, such as instruments y and z. For these other instruments the same spectra from the master spectrometer may be used, but the spectrometers y and z are required to produce their own spectra for the reference chemicals, in a similar way to step 820 for instrument x. A separate calibration transform is built for instrument y and for instrument z and uploaded to the respective instrument. All three instruments, x, y and z, will now have calibration transforms translating them to the master spectrometer.
  • step 840 we perform a step similar to the calibration augmentation of figures 8 and 9.
  • One or more spectra are collected on instrument x and the calibration transform for instrument x is applied at step 850 such that the spectra appear as though they were generated by the master spectrometer.
  • the spectra may be of chemicals from or relevant to the process.
  • the translated spectra along with the raw spectra are added to the data set for generating an augmented calibration.
  • Corresponding translated spectra and raw spectra from instruments y and z are also added to the data set for generating the augmented calibration, as shown at step 865.
  • a single calibration transform is built using the master data and data from the three spectrometers.
  • This single calibration is then uploaded to each of instruments x, y and z, as shown at steps 880 and 885.
  • This calibration may be known as a global calibration.
  • This approach may be readily expanded and applied to the example of eight spectrometers operating on eight fermenters. This approach provides a rapid method for calibrating multiple spectrometers quickly and may also eliminate variability between spectrometers.
  • all of the spectrometers could be translated to the master spectrometer, such as a virtual master spectrometer. This would mean stopping the method at step 830 of figure 12. Each spectrometer would have a different calibration transform for translating it to the master.
  • Figure 13 shows an example spectrometer 2 coupled to a process 1 for measuring a vibrational spectrum of the process.
  • This may for example be an infrared absorption spectrum, such as an FTIR spectrum, or a Raman spectrum.
  • the spectrometer may be coupled to any part of a process in which there is product, batch, other material of the process or by-product of the process.
  • the part of the process may include a pipe or tank.
  • Spectrometer 2 may be coupled to the process by a probe 5.
  • Light or electromagnetic radiation which may be infra-red, is emitted from a source which may be part of interferometer 3.
  • the infra-red radiation may preferably be mid infra-red such as in the range 600-4000cm -1 or 800-1800cm -1 .
  • the probe 3 directs the light to the process and couples it back to the interferometer for measurement.
  • the end of the probe may comprise a window that allows the light to interact with the process material, such as to allow the process material to absorb the light at particular wavelengths.
  • the interferometer 3 generates a pattern, such as an interference pattern on a detector.
  • the interferometer may be the interferometer of figure 1B.
  • the signal from the detector representing the interference pattern is sent to electronics 4 for processing.
  • Display 6 may provide an output spectrum, such as an FTIR spectrum, showing characteristics of the process.
  • Electronics may comprises a memory and processor, such as forming a computing device, that processes that data.
  • 6 may be a computing device including a display.
  • the spectrometer and display may be incorporated together as one. Aspects of the present invention relate to the processing and calibration of the spectrometer 2.
  • the spectrometer comprises an interferometer such as the one described in WO 2011/086357 A1 or US 9046412 B, which are hereby incorporated by reference.
  • the spectrometer may be different but preferably is a form of molecular spectroscopy such as vibrational spectroscopy.
  • the spectrometer may be located away from the process to perform off-line analysis. In both cases the spectrometer may be an FTIR spectrometer such as the IRmadilloTM spectrometer by the applicant or any other suitable process spectrometer.
  • Various aspects of the analysis may be performed by a computer which forms part of the spectrometer or may be coupled to the spectrometer.
  • the computer may receive the spectra such as the FTIR spectra and perform the analyses described below.
  • computer programs may be provided for performing the analyses.
  • FIG. 14 is a flow diagram showing the example operation of a calibrated spectrometer.
  • the calibration transform may be stored in the firmware of the instrument.
  • the spectrometer takes a spectrum of its optics. That is, it takes a spectrum without a sample being present. This may be done by diverting a some of the spectrometer’s broadband source output around the internal optics of the device. This also takes into account any variation in intensity of the source across the frequency range of operation of the spectrometer. This spectrum is indicated in figure 14 as the generation of the reference spectrum.
  • the spectrometer is ready for operation and may collect data or spectra of the sample to be analysed. This produces a raw spectrum of the sample. Next the reference spectrum of the internal optics and source is compared against the raw spectrum to determine an “absorbance spectrum”.
  • Absorbance , A is given by the equation
  • A logio (Io / 1) where Io is the recorded intensity at a given frequency for the reference spectrum and I is the intensity at the same frequency for the sample being analysed.
  • the calibration transform is applied to the absorbance spectrum in the next step to create the translated absorbance spectrum.
  • a chemometric model may be applied to the translated absorbance spectrum.
  • the chemometric model may be used to interpret the resulting spectrum, for example, by providing an indication of the chemicals present and/or the amounts thereof.
  • the chemometric model is specific to the process and may be provided on a computing device connected to the spectrometer. Since calibration transfer allows spectra from one spectrometer to look like they are from another spectrometer, instruments can be swapped out and replaced without having to rebuild the chemometric model. This could be done quickly into a process and the chemometric model analysed output would start running without having to change the settings in the model.

Landscapes

  • Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • General Health & Medical Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)

Abstract

L'invention concerne un procédé d'étalonnage d'un spectromètre, comprenant : l'acquisition pour un premier spectromètre, le premier spectromètre étant un spectromètre de référence étalonné, de premiers spectres d'un ou plusieurs produits chimiques de référence; l'acquisition pour un second spectromètre, le second spectromètre étant un spectromètre à étalonner, de seconds spectres du ou des produits chimiques de référence, les premiers spectres et les seconds spectres étant générés à l'aide d'une spectroscopie vibrationnelle moléculaire; la génération d'une transformée d'étalonnage à partir des premiers spectres et des seconds spectres, la transformée d'étalonnage fournissant un étalonnage pour le second spectromètre, la transformée d'étalonnage transformant des spectres du second spectromètre en spectres qui sont étalonnés sur la base de l'étalonnage du premier spectromètre; et le téléchargement en amont de la transformée d'étalonnage vers le second spectromètre. Le procédé peut comprendre l'augmentation de la transformée d'étalonnage avec des spectres d'échantillons collectés sur le second spectromètre. L'invention concerne également des procédés d'étalonnage d'une pluralité de spectromètres.
PCT/EP2023/056629 2022-03-22 2023-03-15 Procédé d'étalonnage d'un ou de plusieurs spectromètres WO2023180141A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GBGB2203991.1A GB202203991D0 (en) 2022-03-22 2022-03-22 Method of calibrating one or more spectrometers
GB2203991.1 2022-03-22

Publications (1)

Publication Number Publication Date
WO2023180141A1 true WO2023180141A1 (fr) 2023-09-28

Family

ID=81344798

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/056629 WO2023180141A1 (fr) 2022-03-22 2023-03-15 Procédé d'étalonnage d'un ou de plusieurs spectromètres

Country Status (2)

Country Link
GB (1) GB202203991D0 (fr)
WO (1) WO2023180141A1 (fr)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011086357A1 (fr) 2010-01-18 2011-07-21 The Science And Technology Facilities Council Interféromètre spectromètre compact

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011086357A1 (fr) 2010-01-18 2011-07-21 The Science And Technology Facilities Council Interféromètre spectromètre compact
US9046412B2 (en) 2010-01-18 2015-06-02 The Science And Technology Facilities Council Compact interferometer spectrometer

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
CHEN ET AL.: "Systematic prediction error correction: A novel strategy for maintaining the predictive abilities of multivariate calibration models", ANALYST, vol. 13, 2011, pages 98 - 106, XP055314419, DOI: 10.1039/C0AN00171F
MISHRA PUNEET ET AL: "Are standard sample measurements still needed to transfer multivariate calibration models between near-infrared spectrometers? The answer is not always", TRAC TRENDS IN ANALYTICAL CHEMISTRY, ELSEVIER, AMSTERDAM, NL, vol. 143, 15 May 2021 (2021-05-15), XP086800985, ISSN: 0165-9936, [retrieved on 20210515], DOI: 10.1016/J.TRAC.2021.116331 *
SULUB ET AL: "Near-infrared multivariate calibration updating using placebo: A content uniformity determination of pharmaceutical tablets", VIBRATIONAL SPECTROSCOPY, ELSEVIER SCIENCE, AMSTERDAM, NL, vol. 46, no. 2, 28 November 2007 (2007-11-28), pages 128 - 134, XP022493310, ISSN: 0924-2031, DOI: 10.1016/J.VIBSPEC.2007.11.004 *
WEN DU ET AL: "Maintaining the predictive abilities of multivariate calibration models by spectral space transformation", ANALYTICA CHIMICA ACTA, ELSEVIER, AMSTERDAM, NL, vol. 690, no. 1, 4 February 2011 (2011-02-04), pages 64 - 70, XP028366844, ISSN: 0003-2670, [retrieved on 20110212], DOI: 10.1016/J.ACA.2011.02.014 *

Also Published As

Publication number Publication date
GB202203991D0 (en) 2022-05-04

Similar Documents

Publication Publication Date Title
EP3332231B1 (fr) Détermination et correction d'écarts d'enregistrement de fréquence pour une spectroscopie quantitative
US8154731B2 (en) Method and a system for the assessment of samples
US11953427B2 (en) Reconstruction of frequency registration for quantitative spectroscopy
US6922645B2 (en) Augmented classical least squares multivariate spectral analysis
AU2005318683B2 (en) A method for standardising a spectrometer
EP3332230B1 (fr) Reconstruction d'écarts d'enregistrement de fréquence pour une spectroscopie quantitative
EP0800066A2 (fr) Etalonnage précis en longueur d'onde d'un spectromètre
EP3133380B1 (fr) Procédé de correction de sortie de photodétecteur utilisé pour analyseur spectroscopique ou spectroscope, analyseur spectroscopique ou spectroscope utilisant ce procédé et programme pour analyseur spectroscopique ou spectroscope instruisant ce procédé
CN102364328A (zh) 光谱分析仪以及光谱分析方法
RU2569052C1 (ru) Способ компенсации дрейфа частоты в интерферометре
RU2571185C2 (ru) Способ компенсации дрейфа амплитуды в спектрометре и спектрометр, осуществляющий указанный способ
Elsayed et al. On the environmental gas sensing using MEMS FTIR spectrometer in the near-infrared region
WO2023180141A1 (fr) Procédé d'étalonnage d'un ou de plusieurs spectromètres
KR101972883B1 (ko) 정적 변환 방식의 푸리에 변환 분광 시스템
JPH0414298B2 (fr)
JP2022527850A (ja) 分光測定装置を構成するための方法
US20230304860A1 (en) Generalized artificial intelligence modeler for ultra-wide-scale deployment of spectral devices
CN114993941B (zh) 一种免标定抗振动的吸收光谱测量方法与系统
WO2023183499A1 (fr) Modélisateur d'intelligence artificielle généralisé pour un déploiement à 'échelle ultra-large de dispositifs spectraux
KR20220123465A (ko) 큐벳을 통한 광학 경로 길이 결정 방법
Noble FT-IR SPECTROSCOPY
WO2019080451A1 (fr) Procédé et appareil de commande pour analyseur de spectre infrarouge à transformée de fourier et support d'enregistrement et dispositif informatique
CN117664883A (zh) 变压器故障检测方法、装置、存储介质及电子设备
US20140197315A1 (en) Spectral variance compressive detection system, device, and process

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23712198

Country of ref document: EP

Kind code of ref document: A1