Gasoline near infrared spectroscopy conversion method

Abstract

This disclosure relates to a method for near-infrared spectral conversion of gasoline. Hydrocarbon compounds are used as standard samples, and their spectra are acquired on two spectrometers. A fitting algorithm is then used to obtain a fitting coefficient vector between the standard spectral matrix obtained from the first spectrometer and the spectrum of a typical gasoline sample. This fitting coefficient vector is then multiplied by the standard spectral matrices obtained from both spectrometers to obtain the gasoline fitted spectral matrix. A mathematical transformation is then performed on this resulting matrix to obtain the gasoline near-infrared spectral conversion coefficients between the first and second spectrometers. This method eliminates minor spectral differences between near-infrared spectra of gasoline samples measured by different spectrometers. In practical applications, it eliminates the need to collect gasoline standard samples, saving manpower and resources, while improving the accuracy of spectral conversion. It is suitable for rapid analysis of gasoline sample composition and properties.

Description

Technical Field

[0001] This disclosure relates to a method for near-infrared spectral conversion of gasoline. Background Technology

[0002] Existing research on crude oil detection is based on various modern instrumental analytical techniques, including chromatography, mass spectrometry, nuclear magnetic resonance, and infrared / near-infrared spectroscopy. Among these, near-infrared spectroscopy offers advantages such as speed and non-destructive analysis. However, in practical applications, if a near-infrared spectral database established on one type of near-infrared spectrometer is used on another, the differences between the spectra can lead to incorrect identification results. For example, the same type of crude oil might be misidentified as two different types. Currently, common methods to address the inconsistency in spectra acquired by different instruments include Direct Standardization (DS) and Mean Spectra Subtraction Correction (MSSC). However, these methods all have drawbacks, such as the time-consuming and labor-intensive process of collecting actual standard samples, and the lack of sample representativeness, which can further contribute to erroneous identification results. Summary of the Invention

[0003] The purpose of this disclosure is to provide a method for converting near-infrared spectra of gasoline, which can eliminate the slight spectral differences between near-infrared spectra of gasoline samples measured by different spectrometers and improve the accuracy of spectral conversion.

[0004] To achieve the above objectives, this disclosure provides a method for near-infrared spectral conversion of gasoline, the method comprising the following steps:

[0005] Multiple standard substance samples are obtained, and a first spectrometer is used to acquire the spectra of the standard substance samples to obtain a first spectral matrix. A second spectrometer is then used to acquire the spectra of the standard substance samples to obtain a second spectral matrix. The standard substance samples are pure hydrocarbon compounds.

[0006] Obtain typical spectra of multiple typical gasoline samples on the first spectrometer;

[0007] By fitting each typical spectrum to the first spectral matrix, a set of fitting coefficient vectors is obtained;

[0008] The product of the fitting coefficient vector and the first spectral matrix is ​​calculated to obtain the first gasoline fitting spectral matrix;

[0009] The product of the fitting coefficient vector and the second spectral matrix is ​​calculated to obtain the second gasoline fitting spectral matrix;

[0010] Calculate the conversion coefficients of the first gasoline fitted spectral matrix and the second gasoline fitted spectral matrix, and use them as the gasoline near-infrared spectral conversion coefficients between the first spectrometer and the second spectrometer.

[0011] Optionally, the standard material samples include alkane samples, olefin samples, and aromatic samples.

[0012] Optionally, the alkane sample is a pure hydrocarbon compound of alkane monomers selected from C5 to C12, the olefin sample is a pure hydrocarbon compound of olefin monomers selected from C5 to C12, and the aromatic sample is a pure hydrocarbon compound of aromatic monomers selected from C6 to C8.

[0013] Optionally, the step of fitting each typical spectrum to obtain a set of fitting coefficient vectors by fitting the first spectral matrix includes:

[0014] The non-zero fitting coefficients are determined by using the non-negative constrained least squares method, and then the non-zero fitting coefficients are normalized to obtain the fitting coefficient vector.

[0015] Optionally, the step of fitting each typical spectrum to obtain a set of fitting coefficient vectors by fitting the first spectral matrix includes:

[0016] Using the first spectral matrix X A All spectra in the spectrum are fitted to the typical spectrum according to the following equation (1):

[0017]

[0018] In equation (1), y A Representing the typical spectrum, v i Let k be the spectrum of the i-th standard sample in the first spectral matrix, and k be the number of spectra in the first spectral matrix. i Let be the fitting coefficients corresponding to the spectrum of the i-th standard sample in the first spectral matrix, and satisfy the objective function shown in equation (2):

[0019]

[0020] Extract the non-zero fitting coefficients from all the obtained fitting coefficients and normalize them according to the following formula (3):

[0021]

[0022] In equation (3), b i is the normalized fitting coefficient, and g is the number of spectra in the first spectral matrix corresponding to the non-zero fitting coefficient;

[0023] The obtained normalized fitting coefficients are combined into the fitting coefficient vector.

[0024] Optionally, calculating the conversion coefficients between the first gasoline fitting spectral matrix and the second gasoline fitting spectral matrix includes:

[0025] The conversion coefficients are calculated according to the following formula (4):

[0026] E A ×F=E B Equation (4),

[0027] In equation (4), E A E represents the fitted spectral matrix of the first gasoline. B represents the second gasoline fitted spectral matrix, and F is the conversion coefficient.

[0028] Optionally, the first spectral matrix and the second spectral matrix each have a selected characteristic spectral region of 6500–8900 cm⁻¹. -1 The spectral matrix was obtained from the absorbance.

[0029] Optionally, the first spectrometer and the second spectrometer are different models of near-infrared spectrometers.

[0030] Optionally, the first spectrometer and the second spectrometer are both Fourier transform near-infrared spectrometers.

[0031] Optionally, the detection conditions include: a resolution of 2–16 cm. -1 The wavenumber range is 4000–10000 cm⁻¹ -1 The number of scans ranges from 16 to 128.

[0032] Through the above technical solution, this disclosure uses pure hydrocarbon compounds as standard material samples, acquiring their spectra on two spectrometers respectively. A fitting algorithm is then used to obtain the fitting coefficient vector between the standard material spectral matrix from the first spectrometer and the spectrum of a typical gasoline sample. This fitting coefficient vector is then multiplied by the standard material spectral matrices obtained from both spectrometers to obtain the gasoline fitted spectral matrix. A mathematical transformation is then performed on this resulting matrix to obtain the gasoline near-infrared spectral conversion coefficient between the first and second spectrometers. This method eliminates the slight spectral differences between the near-infrared spectra of gasoline samples measured by different spectrometers. In practical applications, it eliminates the need to collect gasoline standard samples, saving manpower and resources, while improving the accuracy of spectral conversion. It is suitable for the rapid analysis of gasoline sample composition and properties.

[0033] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0034] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings:

[0035] Figure 1 This is a schematic flowchart of a gasoline near-infrared spectral conversion method provided in one embodiment of the present disclosure. Detailed Implementation

[0036] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0037] This disclosure provides a method for near-infrared spectral conversion of gasoline, such as... Figure 1 As shown, the process includes the following steps S101 to S106:

[0038] S101. Obtain multiple standard substance samples, perform spectral acquisition on the standard substance samples using a first spectrometer to obtain a first spectral matrix, and perform spectral acquisition on the standard substance samples using a second spectrometer to obtain a second spectral matrix; wherein, the standard substance samples are pure hydrocarbon compounds;

[0039] S102. Obtain typical spectra of multiple typical gasoline samples on the first spectrometer;

[0040] S103. Fit each of the typical spectra using the first spectral matrix to obtain a set of fitting coefficient vectors;

[0041] S104. Calculate the product of the fitting coefficient vector and the first spectral matrix to obtain the first gasoline fitting spectral matrix;

[0042] S105. Calculate the product of the fitting coefficient vector and the second spectral matrix to obtain the second gasoline fitting spectral matrix;

[0043] S106. Calculate the conversion coefficients of the first gasoline fitting spectral matrix and the second gasoline fitting spectral matrix, and use them as the gasoline near-infrared spectral conversion coefficients between the first spectrometer and the second spectrometer.

[0044] This disclosure addresses the wavelength differences between near-infrared spectra of gasoline samples measured by different spectrometers by using pure hydrocarbon compounds as standard material samples for spectral fitting and model transfer. The method is simple and has high accuracy in spectral conversion.

[0045] In step S101, the first spectrometer is the source instrument (or host), and the second spectrometer is the target instrument (or slave). The method disclosed herein is applicable to two near-infrared spectrometers of different models, meaning the first spectrometer and the second spectrometer are different models of near-infrared spectrometers. Specifically, the first spectrometer and the second spectrometer can be different models of the same type of near-infrared spectrometer, or they can be different models of different types of near-infrared spectrometers. The type of near-infrared spectrometer includes, for example, a Fourier transform near-infrared spectrometer, a grating spectrometer, etc. Differences between different models of near-infrared spectrometers include, for example, differences in the structure of key components, wavelength, and absorbance.

[0046] In one specific embodiment, the first spectrometer and the second spectrometer are both Fourier transform near-infrared spectrometers. The conditions for detecting the sample using a Fourier transform near-infrared spectrometer are well known to those skilled in the art. Preferably, the detection of the standard material sample using the first spectrometer and the detection of the standard material sample using the second spectrometer are performed under the same detection conditions to improve the accuracy of the spectral conversion. Specifically, the detection can be transmission detection, using a cuvette injection method with a transmission path of 0.5 mm; the detection conditions may include a resolution of 2–16 cm⁻¹. -1 The wavenumber range is 4000–10000 cm⁻¹ -1 The number of scans ranges from 16 to 128.

[0047] In one embodiment, the standard substance sample is a hydrocarbon monomer pure hydrocarbon compound, including alkane samples, olefin samples, and aromatic hydrocarbon samples. Further, the alkane sample is a alkane monomer pure hydrocarbon compound selected from C5 to C12, including n-alkanes and / or isoalkanes; the olefin sample is an olefin monomer pure hydrocarbon compound selected from C5 to C12, including monoolefins and / or polyolefins; and the aromatic hydrocarbon sample is an aromatic monomer pure hydrocarbon compound selected from C6 to C8.

[0048] The first spectral matrix and the second spectral matrix are each spectral matrices obtained through optional spectral preprocessing. Specifically, the spectral preprocessing may include, but is not limited to, one or more of the following: differentiation processing, normalization processing, normalization processing, and wavelet transform processing.

[0049] In one embodiment, the first spectral matrix and the second spectral matrix each select a characteristic spectral region of 6500–8900 cm⁻¹. -1 The spectral matrix is ​​obtained from the absorbance. Constructing the spectral matrix by selecting absorbance within an appropriate characteristic spectral range is beneficial for further improving the accuracy of spectral conversion.

[0050] In step S102, the typical gasoline sample refers to a gasoline sample with known composition and / or properties. It can be a typical gasoline sample from the gasoline database established on the first spectrometer or an actual gasoline sample collected on-site. The number of typical gasoline samples can be specifically 10 to 30.

[0051] In step S103, the fitting coefficient vector is preferably a normalized non-zero fitting coefficient vector, for example, determined by the non-negative constrained least squares method. Specifically, the non-zero fitting coefficients are determined by the non-negative constrained least squares method, and the non-zero fitting coefficients are normalized to obtain the fitting coefficient vector.

[0052] For example, using the first spectral matrix X A All spectra in the spectrum are fitted to the typical spectrum according to the following equation (1):

[0053]

[0054] In equation (1), y A Representing the typical spectrum, v i Let k be the spectrum of the i-th standard sample in the first spectral matrix, and k be the number of spectra in the first spectral matrix. i Let be the fitting coefficients corresponding to the spectrum of the i-th standard sample in the first spectral matrix, and satisfy the objective function shown in equation (2):

[0055]

[0056] Extract the non-zero fitting coefficients from all the obtained fitting coefficients and normalize them according to the following formula (3):

[0057]

[0058] In equation (3), b i is the normalized fitting coefficient, and g is the number of spectra in the first spectral matrix corresponding to the non-zero fitting coefficient;

[0059] The obtained normalized fitting coefficients are combined into the fitting coefficient vector.

[0060] In steps S104 and S105, the gasoline fitting spectral matrix on the two spectrometers can be obtained by multiplying the fitting coefficient vector with the standard substance spectral matrix on the first spectrometer and the second spectrometer, respectively.

[0061] In step S106, the gasoline fitting spectral matrix on the first spectrometer and the second spectrometer is mathematically converted to obtain the data relationship between the gasoline fitting spectral matrices on the two spectrometers (i.e., the gasoline near-infrared spectral conversion coefficient). For example, it can be calculated and solved by matrix analysis, principal component regression algorithm, etc.

[0062] For example, the conversion coefficients are calculated according to the following formula (4):

[0063] E A ×F=E B Equation (4),

[0064] In equation (4), E A E represents the fitted spectral matrix of the first gasoline. B represents the second gasoline fitted spectral matrix, and F is the conversion coefficient.

[0065] The obtained gasoline near-infrared spectral conversion coefficient can realize the conversion of the spectral database from the source machine (first spectrometer) to the target machine (second spectrometer). That is, the gasoline spectral matrix A established on the first spectrometer is calculated and the gasoline near-infrared spectral conversion coefficient is used to obtain the gasoline spectral matrix B on the second spectrometer.

[0066] This disclosed method is applicable to near-infrared spectral conversion between different spectrometers in gasoline sample detection. There are no special limitations on the gasoline sample; for example, it may include one or more of straight-run gasoline, alkylated gasoline, coking gasoline, catalytic cracking gasoline, catalytic reformed gasoline, and S-Zorb gasoline. In practical applications, a spectral correction model can be further established based on the spectral conversion for rapid analysis of gasoline sample composition, physical properties, etc. This disclosure does not impose special limitations on subsequent steps of the spectral conversion (such as spectral correction, sample analysis, etc.), and conventional methods in the art can be used.

[0067] The method disclosed herein first simultaneously measures the near-infrared spectra of standard material samples on both the source and target spectrometers. Then, it fits the spectrum of a typical gasoline sample collected on the source spectrometer to obtain a fitting coefficient vector. This fitting coefficient vector is multiplied by the standard material spectral matrix collected on the target spectrometer to obtain a virtual typical gasoline spectral matrix on the target spectrometer. The virtual typical gasoline spectral matrices on the source and target spectrometers are then mathematically transformed to obtain conversion coefficients. The model established on the source spectrometer can then be used to obtain the model on the target spectrometer, thereby realizing the conversion between near-infrared spectra measured by different spectrometers.

[0068] The present disclosure is further illustrated below with examples, but it is not limited thereto.

[0069] In the example, instrument a, used to acquire the near-infrared spectrum of gasoline, was a Thermo Anastar II Fourier transform near-infrared spectrometer, and instrument b, used an ABB MB 3600 Fourier transform near-infrared spectrometer. The spectral acquisition conditions for both instruments a and b were: a resolution of 8 cm⁻¹. -1 Wavenumber range 4000–10000 cm⁻¹ -1 The cumulative number of scans was 64, using transmission measurement method.

[0070] Example 1

[0071] In this embodiment, a near-infrared spectral database of gasoline has been established on instrument a. The database contains 200 gasoline samples and their corresponding research octane numbers.

[0072] (1) Collect spectra of pure compound samples of normal C5, C6, C7, C8, C9, C10, C11, C12 alkanes, normal C5, C6, C7, C8, C9, C10, C11, C12 alkenes, and C6, C7, C8 aromatics on instrument a and instrument b, respectively, and select 6500-8900 cm⁻¹. -1 The absorbance in the spectral region is used to establish the near-infrared absorbance matrix of pure compound standard substances, i.e., the first spectral matrix A. pure Second spectral matrix B pure .

[0073] In an actual application at a petrochemical refinery, 25 gasoline samples processed daily were collected. Near-infrared spectra were acquired on instrument b, and their octane numbers were determined using the national standard method GB / T 5487. These samples were used as unknown samples to test the spectral conversion effect.

[0074] (2) Select 20 typical gasoline samples from the gasoline database established on instrument a, and use the first spectral matrix A mentioned above. pure For each typical spectrum, a fitting was performed, resulting in 20 sets of normalized non-zero fitting coefficient vectors. Specifically,

[0075] Using all the spectra in the first spectral matrix, the typical spectrum is fitted according to the following equation (1):

[0076]

[0077] In equation (1), y A Representing the typical spectrum, v i Let k be the spectrum of the i-th standard sample in the first spectral matrix, and k be the number of spectra in the first spectral matrix. i Let be the fitting coefficients corresponding to the spectrum of the i-th standard sample in the first spectral matrix, and satisfy the objective function shown in equation (2):

[0078]

[0079] Extract the non-zero fitting coefficients from all the obtained fitting coefficients and normalize them according to the following formula (3):

[0080]

[0081] In equation (3), b i is the normalized fitting coefficient, and g is the number of spectra in the first spectral matrix corresponding to the non-zero fitting coefficient;

[0082] The obtained normalized fitting coefficients are combined into the fitting coefficient vector.

[0083] (3) The standard substance spectral matrix A collected by instrument a and instrument b pure and B pure Multiply by the above 20 sets of fitting coefficient vectors respectively to obtain the gasoline fitting spectral matrices of instrument a and instrument b, namely the first gasoline fitting spectral matrix EA and the second gasoline fitting spectral matrix EB.

[0084] (4) Calculate the transformation coefficient F of EA×F=EB based on matrix analysis.

[0085] (5) The gasoline spectral matrix A established on instrument a is multiplied by the conversion coefficient F, i.e., A×F=B a Thus, matrix B was obtained. a This completes the spectral conversion from instrument a to instrument b.

[0086] (6) Using gasoline near-infrared database B a A gasoline octane number prediction model was established using partial least squares (PLS) to predict the octane number of 25 unknown gasoline samples collected by instrument b. The results are shown in Table 1.

[0087] (7) Establish a partial least squares (PLS) spectral correction model on instrument a, and determine the optimal principal factor f. best Under the given conditions, the model was used to predict the results of the 25 gasoline samples collected by instrument a. The comparison results are shown in Table 2.

[0088] The formulas for calculating the relevant statistical parameters are as follows:

[0089]

[0090] in, Let y be the predicted value for the i-th sample. i (i = 1, 2, ..., n) represents the true value of the i-th sample, and n is the number of samples in the test set.

[0091] Deviation = Measured value - Predicted value.

[0092] Table 1

[0093]

[0094] Table 2

[0095]

[0096] Comparing Tables 1 and 2, it can be seen that the prediction accuracy of the PLS model established on the second spectrometer after spectral conversion using the method of this disclosure is basically the same as that before spectral conversion (on the first spectrometer). This indicates that the method of this disclosure can eliminate the small spectral differences between the near-infrared spectra of gasoline samples measured by different spectrometers and realize spectral conversion between different spectrometers.

[0097] Comparative Example 1

[0098] The conventional DS algorithm is used to perform spectral conversion between instrument a and instrument b, specifically as follows:

[0099] (1) In the actual application of a petrochemical refinery, 20 typical gasoline sample spectra were collected on instrument a and instrument b respectively as standard samples for model transfer, and the 6500-8900 cm⁻¹ range was selected. -1 The absorbance in the spectral region is used to establish the near-infrared spectral absorbance matrix of the gasoline standard, i.e., the first spectral matrix A. s Second spectral matrix B s .

[0100] Twenty-five gasoline samples processed daily were collected, and near-infrared spectra were acquired on instrument b. Their octane numbers were determined using the national standard method GB / T5487. These samples were used as unknown samples to test the spectral conversion effect.

[0101] (2) The first spectral matrix A s Second spectral matrix B s Perform eigenvalue decomposition separately, A s =U1S1V1,B s =U2S2V2,A s Representing host B s The slave device undergoes the same spectral preprocessing as in Example 1.

[0102] (3) Let Z1 = U1S1, Z2 = U2S2.

[0103] (4) Solve for the regression coefficient F, and use matrix analysis to find the transformation coefficient F of Z1×F=Z2.

[0104] (5) Multiply the gasoline spectral library matrix A established on instrument a by the conversion coefficient F, i.e., A×F=B a 'This yields matrix B' aThat is, the spectral conversion from instrument a to instrument b is completed.

[0105] (6) Using gasoline near-infrared database B a A gasoline octane number prediction model was established using partial least squares (PLS) to predict the octane number of 25 unknown gasoline samples collected by instrument b. The results are shown in Table 3.

[0106] Table 3

[0107]

[0108]

[0109] As can be seen from the comparison of Tables 1 and 3, compared with the traditional method for spectral conversion, the PLS model established on the second spectrometer after spectral conversion using the method of this disclosure has better prediction accuracy than the transfer algorithm of the traditional DS model. Furthermore, in practical applications, it is not necessary to collect gasoline standard samples, saving manpower and resources.

[0110] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0111] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0112] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for converting near-infrared spectra of gasoline, characterized in that, The method includes the following steps: Multiple standard substance samples are obtained, and a first spectrometer is used to acquire the spectra of the standard substance samples to obtain a first spectral matrix. A second spectrometer is then used to acquire the spectra of the standard substance samples to obtain a second spectral matrix. The standard substance samples are pure hydrocarbon compounds. Obtain typical spectra of multiple typical gasoline samples on the first spectrometer; Using the first spectral matrix X A All spectra in the spectrum are fitted to the typical spectrum according to the following equation (1): y A= Formula (1), In equation (1), y A Representing the typical spectrum, i This represents the spectrum of the i-th standard substance sample in the first spectral matrix. k The number of spectra in the first spectral matrix. a i Let be the fitting coefficients corresponding to the spectrum of the i-th standard sample in the first spectral matrix, and satisfy the objective function shown in equation (2): min||y A - ||, s.t. a i ≥0, equation (2), Extract the non-zero fitting coefficients from all the obtained fitting coefficients and normalize them according to the following formula (3): Equation (3), In equation (3), b i These are the normalized fitting coefficients. g The number of spectra in the first spectral matrix corresponding to the non-zero fitting coefficients; All the obtained normalized fitting coefficients are combined into a set of fitting coefficient vectors; The product of the fitting coefficient vector and the first spectral matrix is ​​calculated to obtain the first gasoline fitting spectral matrix; The product of the fitting coefficient vector and the second spectral matrix is ​​calculated to obtain the second gasoline fitting spectral matrix; The conversion coefficients of the first gasoline fitting spectral matrix and the second gasoline fitting spectral matrix are calculated according to the following formula (4), and are used as the gasoline near-infrared spectral conversion coefficients between the first spectrometer and the second spectrometer; E A × F = E B Equation (4), In equation (4), E A E represents the fitted spectral matrix of the first gasoline. B represents the second gasoline fitted spectral matrix, and F is the conversion coefficient.

2. The method according to claim 1, wherein, The standard material samples include alkane samples, olefin samples, and aromatic samples.

3. The method according to claim 2, wherein, The alkane sample is a pure hydrocarbon compound selected from C5 to C12 alkane monomers, the olefin sample is a pure hydrocarbon compound selected from C5 to C12 olefin monomers, and the aromatic sample is a pure hydrocarbon compound selected from C6 to C8 aromatic monomers.

4. The method according to claim 1, wherein, The first spectral matrix and the second spectral matrix each have a selected characteristic spectral region of 6500~8900cm. -1 The spectral matrix was obtained from the absorbance.

5. The method according to claim 1, wherein, The first spectrometer and the second spectrometer are different models of near-infrared spectrometers.

6. The method according to claim 5, wherein, The first spectrometer and the second spectrometer are both Fourier transform near-infrared spectrometers.

7. The method according to claim 6, wherein, The detection conditions for the first spectrometer and the second spectrometer respectively include: a resolution of 2~16cm. -1 The wavenumber range is 4000~10000 cm⁻¹ -1 The number of scans is 16 to 128.

Images