Method for detecting poly-alpha-olefin content in sample

Abstract

The invention relates to the field of spectrum detection, and discloses a method for detecting the content of poly-alpha-olefin in a sample. The method comprises the following steps: (1) providing a plurality of standard poly-alpha-olefin samples; (2) collecting a near infrared spectrum, a middle infrared spectrum and a Raman spectrum; selecting a characteristic spectral band to obtain a corresponding matrix; (3) taking a corresponding matrix, and performing regression analysis on the corresponding matrix and the poly-alpha-olefin content of a standard poly-alpha-olefin sample to obtain a prediction model; (4) performing weighted fusion on the prediction model to obtain a fusion model; and (5) taking a to-be-detected sample with unknown poly-alpha-olefin content, processing according to the mode in the step (2), and substituting the absorbance matrix and the spectral intensity matrix corresponding to the obtained characteristic spectral band into the fusion model to obtain the poly-alpha-olefin content. The method is simple to operate, high in analysis speed, high in efficiency, accurate in prediction and good in repeatability.

Description

Technical Field

[0001] The invention relates to the field of spectrum detection, and in particular to a method for detecting the content of poly-alpha-olefin in a sample. Background Art

[0002] Polyalphaolefin synthetic oil (PAO) is a base oil for producing high-quality lubricating oil. Its molecular structure is regular and uniform and is comb-shaped or tree-shaped. It has excellent low-temperature fluidity, oxidation stability, and viscosity-temperature properties. PAO is an olefin oligomer obtained by polymerizing monomer alpha-olefin under the action of a catalyst and then hydrogenating and refining. The monomer commonly used to synthesize PAO base oil is C6-C12 straight-chain olefin. Among these olefins, the PAO base oil produced with 1-decene as raw material has the best performance. The content of polyalphaolefin will affect the thermal stability, thermal oxidation stability and color of the product. Generally speaking, the higher the content of polyalphaolefin, the better the performance of the product.

[0003] The commonly used analytical methods for polyalphaolefin content in laboratories include high performance liquid chromatography, nuclear magnetic resonance analysis and gel chromatography, which are relatively reliable. However, the complex sample processing process and slow analysis process make these methods unable to become the best analytical methods in industrial production. Therefore, in order to improve the analysis efficiency, it is urgent to develop a method that can quickly, simply and accurately determine the polyalphaolefin content in polyalphaolefin synthetic oil. Summary of the invention

[0004] The purpose of the present invention is to overcome the above problems existing in the prior art and to provide a method for detecting the poly-alpha-olefin content in a sample, which has the advantages of simple operation, fast analysis speed, high efficiency, accurate prediction and good repeatability.

[0005] In order to achieve the above object, the present invention provides a method for detecting the content of poly-alpha-olefin in a sample, the method comprising:

[0006] (1) providing a plurality of standard poly-alpha-olefin samples with known and different poly-alpha-olefin contents;

[0007] (2) For the above standard poly-α-olefin samples, near infrared spectra X NIR , mid-infrared spectroscopy X MIR and Raman spectroscopy Raman ;

[0008] Select characteristic spectral bands of the above spectra respectively, and obtain corresponding absorbance matrix or spectral intensity matrix;

[0009] (3) Take the near infrared spectrum X NIR The absorbance matrix corresponding to the characteristic spectral bands, mid-infrared spectrum X MIRThe absorbance matrix and Raman spectrum X corresponding to the characteristic spectral band Raman The spectral intensity matrix corresponding to the characteristic spectral bands of NIR , P MIR and P Raman ;

[0010] (4) The prediction model P NIR , P MIR and P Raman Perform weighted fusion to obtain a fusion model M;

[0011] (5) Taking a sample with unknown poly-α-olefin content, processing it in the manner of step (2), and substituting the absorbance matrix and spectral intensity matrix corresponding to the obtained characteristic spectral band into the fusion model M to obtain the poly-α-olefin content in the sample.

[0012] The solution provided by the present invention can avoid tedious processing and analysis processes. The method is simple to operate, fast in analysis speed, high in efficiency, accurate in prediction and good in repeatability, and is more suitable for practical industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is an error comparison diagram between the fusion model of Example 1 and the prediction model of the comparative example. DETAILED DESCRIPTION

[0014] The endpoints and any values ​​of the ranges disclosed in this article are not limited to the precise ranges or values, and these ranges or values ​​should be understood to include values ​​close to these ranges or values. For numerical ranges, the endpoint values ​​of each range, the endpoint values ​​of each range and the individual point values, and the individual point values ​​can be combined with each other to obtain one or more new numerical ranges, which should be considered as specifically disclosed in this article.

[0015] The present invention provides a method for detecting the content of poly-alpha-olefin in a sample, the method comprising:

[0016] (1) providing a plurality of standard poly-alpha-olefin samples with known and different poly-alpha-olefin contents;

[0017] (2) For the above standard poly-α-olefin samples, near infrared spectra X NIR , mid-infrared spectroscopy X MIR and Raman spectroscopy Raman ;

[0018] Select characteristic spectral bands of the above spectra respectively, and obtain corresponding absorbance matrix or spectral intensity matrix;

[0019] (3) Take the near infrared spectrum XNIR The absorbance matrix corresponding to the characteristic spectral bands, mid-infrared spectrum X MIR The absorbance matrix and Raman spectrum X corresponding to the characteristic spectral band Raman The spectral intensity matrix corresponding to the characteristic spectral bands of NIR , P MIR and P Raman ;

[0020] (4) The prediction model P NIR , P MIR and P Raman Perform weighted fusion to obtain a fusion model M;

[0021] (5) Taking a sample with unknown poly-α-olefin content, processing it in the manner of step (2), and substituting the absorbance matrix and spectral intensity matrix corresponding to the obtained characteristic spectral band into the fusion model M to obtain the poly-α-olefin content in the sample.

[0022] The inventors of the present invention have found in their research that if the near infrared spectrum, mid-infrared spectrum and Raman spectrum of the poly-alpha-olefin sample are collected and processed to obtain a fusion model in the above manner, then for the sample to be tested with unknown poly-alpha-olefin content, the poly-alpha-olefin content can be quickly and accurately obtained by collecting the corresponding spectrum and substituting the absorbance matrix or spectral intensity matrix corresponding to the characteristic spectral band into the fusion model. In particular, the inventors of the present invention have found that the method of collecting near infrared spectrum, mid-infrared spectrum and Raman spectrum according to the present invention, and fusing the prediction models obtained by these three spectra to obtain the fusion model, can obtain higher accuracy than the method of collecting one of them, or collecting other types of spectra and fusing them.

[0023] It is understood that in near-infrared and mid-infrared spectra, the ordinate is generally absorbance; in Raman spectra, the ordinate is generally spectral intensity. In step (3), taking the absorbance matrix corresponding to the characteristic spectral band of the near-infrared spectrum as an example, a plurality of standard poly-alpha-olefin samples form an "M×N absorbance matrix", where M is the number of standard poly-alpha-olefin samples and N is the wavelength.

[0024] The method provided by the present invention can avoid the complicated sample processing process and slow analysis process in traditional methods (such as high performance liquid chromatography, nuclear magnetic resonance analysis and gel chromatography), has high accuracy and good repeatability, is simple to operate, fast, and is more suitable for practical industrial applications.

[0025] It is understood that the mid-infrared spectrum generally has a wavelength of 400-4000cm-1 The near infrared spectrum generally has a wave number range of 4000-12500cm -1 wavenumber range.

[0026] According to the present invention, preferably, the number of the standard poly-α-olefin samples is ≥50, more preferably 60-200 (for example, it can be 60, 80, 100, 120, 140, 160, 180, 200, and a range formed by any two of the above values, and a value within the range). This can further ensure the accuracy of the results.

[0027] The present invention has no particular limitation on the poly-α-olefin content in the standard poly-α-olefin sample, which can be within a wide range. According to the present invention, in step (1), the poly-α-olefin content in the standard poly-α-olefin sample satisfies: the mass concentration is greater than 30wt%, more preferably 42wt%-99.5wt% (for example, it can be 42wt%, 45wt%, 50wt%, 60wt%, 70wt%, 80wt%, 90wt%, 95wt%, 99.5wt% and the range formed by any two of the above values ​​and the value within the range).

[0028] It can be understood that when collecting samples, it is generally required that the sample conditions are relatively evenly distributed within a possible range. Specifically for this application, the poly-α-olefin content in the standard poly-α-olefin sample can be evenly distributed within the above range.

[0029] The spectrum collection conditions can also be carried out under conventional conditions in the art. Preferably, the spectrum collection conditions include: the collection temperature is 15-40°C, more preferably 23-27°C.

[0030] Preferably, the resolution of the acquisition is 2-16cm -1 , more preferably 6-10cm -1 .

[0031] According to the present invention, preferably, the collection wave number interval of the near infrared spectrum is 4000-10000cm -1 The optical path of the near infrared spectrum measurement can be in the range of 0.5-2 mm, and the number of scans can be 16-128 times.

[0032] Preferably, the mid-infrared spectrum is collected in the wavenumber range of 650-4000cm -1 The optical path of the mid-infrared spectrum measurement can be in the range of 0.5-2 mm, and the number of scans can be 16-128 times.

[0033] Preferably, the Raman spectrum is collected in a wavenumber range of 100-2500 cm -1The measurement optical path of Raman spectroscopy can be in the range of 0.5-2 mm, and the number of scans can be 16-128 times.

[0034] According to the present invention, preferably, before selecting the characteristic spectral band, the method further includes: preprocessing the spectrum to eliminate redundancy and / or noise in the spectrum. It can be understood that by preprocessing, baseline drift, spectral overlap, noise, etc. in the spectrum can be eliminated to further improve the accuracy of subsequent model construction.

[0035] Preferably, the near infrared spectrum X NIR and mid-infrared spectroscopy MIR The preprocessing methods are independently selected from at least one of the second order differential, standard normal variable transformation and first order differential, and more preferably the second order differential. The window width of the second order differential (referring to the width of the spectrum smoothing) can be 5-31. The window width of the second order differential corresponding to the near infrared spectrum and the mid infrared spectrum can be different.

[0036] According to the present invention, preferably, the Raman spectrum X Raman The preprocessing method is selected from at least one of multivariate scatter correction, standard normal variate transformation and baseline correction, and is more preferably multivariate scatter correction (MSC).

[0037] According to the present invention, preferably, the near infrared spectrum X NIR The wave number range of the characteristic spectral band is 4500-7800cm -1 , more preferably 4227-4651cm -1 、5422-6001cm -1 and 7416-9002cm -1 .

[0038] According to the present invention, preferably, the wave number interval of the characteristic spectral band of the mid-infrared spectrum is 650-4000cm -1 , more preferably 1603-1663cm -1 and 650-980cm -1 .

[0039] According to the present invention, preferably, the wave number interval of the characteristic spectral band of the Raman spectrum is 100-2500cm -1 , more preferably 1018-1755cm -1 .

[0040] Selecting the above wavebands can more comprehensively reflect the information of carbon molecules related to poly-α-olefins in the spectrum, and can further ensure the accuracy of the results.

[0041] According to the present invention, preferably, the regression analysis method is selected from at least one of least squares-support vector machine, partial least squares method and artificial neural network, more preferably least squares-support vector machine. The inventors of the present invention further found that for near infrared spectrum, mid-infrared spectrum and Raman spectrum, the least squares-support vector machine (LS-SVM) method is used for regression analysis, which can obtain higher accuracy than other regression analysis methods.

[0042] According to the present invention, preferably, the model parameter smoothing parameter σ and the regularization parameter γ in the least squares-support vector machine are determined according to at least one of the leave-one-out method, Bayesian reasoning and leave-one-out cross validation, and more preferably determined according to the leave-one-out cross validation. For leave-one-out cross validation, see Wong T T. Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation [J]. Pattern Recognition, 2015, 48 (9): 2839-2846.

[0043] According to the present invention, preferably, before the regression analysis, step (2) further includes: normalizing the absorbance matrix or spectral intensity matrix corresponding to the characteristic spectral band. It can be understood that the purpose of normalization is to scale the spectral intensity (normalize the highest value, and other intensities are expressed as decimals less than 1) to avoid loss of details of spectral information due to differences in absorbance or spectral intensity.

[0044] According to the present invention, preferably, the weighted fusion method of step (4) includes: near infrared spectrum X NIR The absorbance matrix corresponding to the characteristic spectral band is substituted into the prediction model P NIR , and get the prediction result Y NIR ;

[0045] Mid-infrared spectroscopy MIR The absorbance matrix corresponding to the characteristic spectral band is substituted into the prediction model P MIR , and get the prediction result Y MIR ;

[0046] Raman spectroscopy Raman The spectral intensity matrix corresponding to the characteristic spectral band is substituted into the prediction model P Raman , and get the prediction result Y Raman ;

[0047] Assign weights a, b and c to the above prediction results respectively. NIR +bY MIR +cYRaman Adjust a, b and c to make the calculation results close to the actual poly-α-olefin content; where a, b, c are greater than 0 and less than 1 respectively, and a+b+c=1; obtain the fusion model M=aP NIR +bP MIR +cP Raman It can be understood that, taking the absorbance matrix corresponding to the characteristic spectral band of the near-infrared spectrum as an example, one sample forms a "1×N matrix", where N is the wavelength; multiple samples form an "M×N matrix", where M is the number of samples and N is the wavelength; the above matrix serves as the basis for subsequent processing.

[0048] For example, in step (5), a sample with unknown poly-α-olefin content is taken to form an absorbance matrix corresponding to the characteristic spectral band of the near-infrared spectrum, an absorbance matrix corresponding to the characteristic spectral band of the mid-infrared spectrum, and a spectral intensity matrix corresponding to the characteristic spectral band of the Raman spectrum; the above matrices are respectively a "1×N matrix", where N is the wavelength.

[0049] It is understood that in the process of adjusting a, b and c, there are many possible combinations that can be used, and the values ​​of a, b and c that can calculate the result closer to the actual poly-α-olefin content can be selected from these combinations. Generally, the difference between the two is within 1×10 -4 Within.

[0050] The inventors of the present invention also found that, according to the above method, the prediction models obtained from the near-infrared spectrum, the mid-infrared spectrum and the Raman spectrum are weightedly fused, which can further ensure the accuracy of the fusion model.

[0051] It can be understood that since the fusion model M is obtained by predicting the model P NIR , P MIR and P Raman The prediction model P NIR , P MIR and P Raman It is directly related to the absorbance matrix or the spectral matrix. When the absorbance matrix and the spectral matrix are substituted into the fusion model M, it is substituted in the same way as substituting the absorbance matrix and the spectral intensity into the prediction model.

[0052] According to the present invention, preferably, in step (1), the sample is a poly-α-olefin synthetic oil (i.e., PAO). After obtaining the poly-α-olefin synthetic oil, the poly-α-olefin content therein can be determined by chromatography, for example, by using the ASTM D6352 method, and the poly-α-olefin content is confirmed based on the boiling point. The poly-α-olefin has a relatively high boiling point, and components with a boiling point exceeding 300° C. are considered to be poly-α-olefins.

[0053] It is understood that, generally, the poly-alpha-olefin synthetic oil produced with 1-decene as raw material has the best performance. Therefore, the poly-alpha-olefin in the poly-alpha-olefin synthetic oil preferably refers to hydrogenated poly-1-decene. The poly-alpha-olefin synthetic oil generally also contains impurities such as 1-dodecene, some linear alpha olefins, beta olefins, and a small amount of aromatics and alkanes.

[0054] Generally, when the sample is poly-alpha-olefin synthetic oil and the fusion model is constructed by the method provided by the present invention, generally, 0.2≤a≤0.7, 0.2≤b≤0.7. After collecting the near infrared spectrum, mid-infrared spectrum and Raman spectrum, MATLAB software can be used to perform the method provided by the present invention.

[0055] The present invention will be described in detail below by way of examples. In step (1) of the following examples,

[0056] The content of poly-α-olefin is confirmed by boiling point according to the ASTM D6352 method. Poly-α-olefin has a relatively high boiling point. Components with a boiling point exceeding 300° C. are considered to be poly-α-olefin.

[0057] The operations after spectrum acquisition were performed in MATLAB.

[0058] Example 1

[0059] Used to illustrate the method for detecting the poly-α-olefin content in a sample provided by the present invention

[0060] (1) 80 poly-α-olefin synthetic oil (PAO) samples were obtained and the poly-α-olefin content thereof was determined to be in the range of 63.70-97.41 wt %).

[0061] (2) For the above PAO samples, near infrared spectrum, mid infrared spectrum and Raman spectrum were collected. The collection temperature was 24°C and the collection resolution was 8 cm -1 ; The acquisition wave number range of near infrared spectrum is 4000-10000cm -1 , measured with an optical path of 0.5 mm and 64 scans; the mid-infrared spectrum was collected in the wavenumber range of 650-4000 cm -1 , measured with an optical path of 0.1 mm and 64 scans; the Raman spectrum was collected in a wavenumber range of 100-2500 cm -1 , measured with an optical path of 0.5 mm and 64 scans.

[0062] For the near-infrared spectrum, a second-order differential with a window width of 21 points was performed, and then the wave number interval was taken as 4227-4651cm -1 、5422-6001cm -1 and 7416-9002cm-1 The region is taken as the characteristic spectral band, and the corresponding absorbance matrix is ​​obtained;

[0063] For the mid-infrared spectrum, a second-order differential process with a window width of 27 points was performed, and then the wave number interval was taken as 1603-1663cm -1 and 650-980cm -1 The region is taken as the characteristic spectral band, and the corresponding absorbance matrix is ​​obtained;

[0064] The Raman spectrum was corrected for multivariate scattering and the wave number interval was 1018-1755cm -1 The area is taken as the characteristic spectral band, and the corresponding spectral intensity matrix is ​​obtained;

[0065] The absorbance matrix and spectral intensity matrix of the above characteristic spectral bands are normalized respectively (the highest value of the absorbance is normalized to unity, and the other values ​​are correspondingly decimals less than 1; or the highest value of the spectral intensity is normalized to unity, and the other values ​​are correspondingly decimals less than 1).

[0066] (3) The absorbance matrix corresponding to the near-infrared spectrum and the poly-alpha-olefin content of the corresponding standard poly-alpha-olefin sample are used to construct a prediction model P by using the least squares-support vector machine method. NIR ;

[0067] The least squares-support vector machine method was applied to the absorbance matrix corresponding to the mid-infrared spectrum and the polyalphaolefin content of the corresponding standard polyalphaolefin sample to construct a prediction model P MIR ;

[0068] The spectral intensity matrix corresponding to the Raman spectrum and the poly-alpha-olefin content of the corresponding standard poly-alpha-olefin sample are used to construct a prediction model P by applying the least squares-support vector machine method. Raman ;

[0069] In the above least squares-support vector machine method, the model parameters smoothing parameter σ and regularization parameter γ are determined by leave-one-out cross-validation.

[0070] (4) Substitute the absorbance matrix corresponding to the near-infrared spectrum into the prediction model P NIR , and get the prediction result Y NIR ;

[0071] Substitute the absorbance matrix corresponding to the mid-infrared spectrum into the prediction model P MIR , and get the prediction result Y MIR ;

[0072] Substitute the spectral intensity matrix corresponding to the Raman spectrum into the prediction model P Raman , and get the prediction result Y Raman ;

[0073] Assign weights a, b, c to the above prediction results respectively, according to aY NIR +bY MIR +cY Raman Adjust a, b and c to aY NIR +bY MIR +cY Raman The calculated result is close to the actual poly-α-olefin content.

[0074] The fusion model is obtained as M = aP NIR +bP MIR +cP Raman .

[0075] Among them, a=0.2626,b=0.3088,c=0.4286(a is Y NIR The weight of Y MIR The weight of Y Raman weight).

[0076] (5) 23 PAOs with unknown poly-α-olefin contents were selected as validation sets, and the poly-α-olefin contents thereof were determined by gas chromatography in step (1) (i.e., the measured values ​​in Table 1, which were distributed in the range of 79.17-96.98 wt %);

[0077] For each sample to be tested, the following processing is performed:

[0078] According to step (2), the characteristic spectral bands corresponding to the near infrared spectrum, mid-infrared spectrum and Raman spectrum, as well as the corresponding absorbance matrix and spectral intensity matrix (also normalized in the same way) are obtained. Substitute the above absorbance matrix and spectral intensity matrix into the fusion model to obtain the poly-α-olefin content in the sample to be tested (i.e., the calculated value in Table 1).

[0079] SEP (standard deviation of prediction) and RPD (ratio of standard deviation to root mean square error) are used to evaluate the performance of the fusion model, where:

[0080]

[0081]

[0082] In the above formula, i represents the i-th sample, i = 1, 2, ... m, m is the total number of samples in the validation set 23, y i,actual is the gas chromatography measured value of the i-th sample, y i,predicted is the calculated value of the i-th sample obtained by the fusion model, SD V is the standard deviation of the polyalphaolefin content of all samples in the validation set, SD V The calculation formula is as follows:

[0083]

[0084] Among them, i = 1, 2, ..., N, N is the number of samples in the validation set 23, X i represents the poly-α-olefin content of the i-th sample (value measured by gas chromatography), is the average value of the poly-α-olefin content of all samples in the validation set (measured by gas chromatography).

[0085] The SEP is 0.4399 and the RPD is 9.0974, indicating that the fusion model has a high accuracy.

[0086] In addition, the gas chromatography measured values ​​and the calculated values ​​of the fusion model for the 23 samples in the validation set, as well as the errors between the two (the calculated values ​​of the fusion model minus the gas chromatography measured values), are shown in Table 1.

[0087] Table 1

[0088] Sample No. Measured value, wt% Calculated value, wt% Error, wt% 1 79.17 78.23 -0.94 2 82.41 82.03 -0.38 3 83.61 83.50 -0.11 4 83.86 83.26 -0.60 5 85.08 85.03 -0.05 6 85.10 84.86 -0.24 7 86.03 85.90 -0.13 8 86.57 86.51 -0.06 9 87.25 87.10 -0.15 10 87.58 87.33 -0.25 11 87.79 85.86 -1.93 12 88.29 88.17 -0.12 13 89.19 89.32 0.13 14 89.24 89.00 -0.24 15 89.36 89.04 -0.32 16 89.74 89.43 -0.31 17 89.95 89.67 -0.28 18 91.01 90.11 -0.90 19 91.24 90.37 -0.87 20 91.73 91.22 -0.51 21 93.34 92.78 -0.56 22 93.42 93.08 -0.34 23 96.98 96.17 -0.81

[0089] It can be seen that the method provided by the present invention has a smaller error and a higher accuracy.

[0090] Example 2

[0091] Used to illustrate the repeatability of the fusion model obtained in Example 1

[0092] Take two samples with unknown poly-α-olefin content and measure the poly-α-olefin content therein respectively according to the gas chromatography method in step (1) of Example 1. The two samples are treated as follows:

[0093] Six parallel acquisitions were performed. In each parallel acquisition, the near-infrared spectrum, mid-infrared spectrum and Raman spectrum were acquired according to step (2) in Example 1, and the characteristic spectral bands were selected according to step (2) to obtain the corresponding absorbance matrix and spectral matrix and normalize them.

[0094] Substitute the absorbance matrix and spectrum matrix obtained from each parallel collection into the fusion model M.

[0095] For each sample, six parallel results (calculated values) were obtained, and the error was obtained by subtracting the measured value by gas chromatography from the calculated value each time. The RSD was used to evaluate the repeatability of the model. The smaller the value, the better the repeatability. The calculation formula of RSD is as follows:

[0096]

[0097] Where S represents the standard deviation of the calculated values, represents the average value of the sample calculation, i = 1, 2, ..., n, n represents the number of samples, x i Represents the calculated value of the i-th sample.

[0098] The results are shown in Table 2.

[0099] Table 2

[0100]

[0101] It can be seen that the method provided by the present invention has good repeatability.

[0102] Comparative Example 1

[0103] Take the verification set of Example 1 and obtain the absorbance matrix corresponding to the characteristic spectral band of the near-infrared spectrum, the absorbance matrix corresponding to the characteristic spectral band of the mid-infrared spectrum, and the absorbance matrix corresponding to the characteristic spectral band of the Raman spectrum in the manner of step (5).

[0104] Substitute the absorbance matrix corresponding to the near-infrared spectrum into the prediction model P NIR , and get the result R NIR , R NIR Subtract the value measured by gas chromatography to get the corresponding error;

[0105] Substitute the absorbance matrix corresponding to the mid-infrared spectrum into the prediction model P MIR , and get the result R MIR , R MIR Subtract the value measured by gas chromatography to get the corresponding error;

[0106] Substitute the absorbance matrix corresponding to the Raman spectrum into the prediction model P Raman , and get the result R Raman , R Raman Subtract the value measured by gas chromatography to get the corresponding error.

[0107] The error and R corresponding to the fusion model in Example 1 NIR The corresponding error, R MIR The corresponding error, R Raman The corresponding error distribution can be found in Figure 1 .

[0108] Among them, the SEP and RPD of the above prediction models or fusion models are shown in Table 3, and the calculation formulas of SEP and RPD are as follows.

[0109]

[0110]

[0111] Where i represents the i-th sample, i = 1, 2, ... m, m is the total number of samples in the validation set, y i,actual is the gas chromatography measured value of the i-th sample, y i,predicted is the calculated value of the i-th sample obtained by the prediction model, SD V is the standard deviation of the polyalphaolefin content of all samples in the validation set, SD V The calculation formula is as follows:

[0112]

[0113] Among them, i = 1, 2, ..., N, N is the number of samples in the validation set 23, X i represents the poly-α-olefin content of the i-th sample (value measured by gas chromatography), is the average value of the poly-α-olefin content of all samples in the validation set (measured by gas chromatography).

[0114] Table 3

[0115] <![CDATA[P NIR ]]> <![CDATA[P MIR ]]> <![CDATA[P Raman ]]> Fusion Model M SEP 1.4539 0.5594 0.5686 0.4399 RPD 2.7525 7.1538 7.0377 9.0974

[0116] It can be understood that the smaller the SEP and the larger the RPD, the better the accuracy of the model. It can be seen that the error of the scheme provided by the present invention is significantly smaller than the error of the model directly constructed using near infrared spectroscopy, mid-infrared spectroscopy and Raman spectroscopy.

[0117] From the above, it can be seen that the scheme provided by the present invention can avoid tedious processing and analysis processes. The method is simple to operate, fast in analysis speed, high in efficiency, accurate in prediction and good in repeatability, and is more suitable for practical industrial applications.

[0118] The preferred embodiments of the present invention are described in detail above, but the present invention is not limited thereto. Within the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications, including the combination of various technical features in any other suitable manner, and these simple modifications and combinations should also be regarded as the contents disclosed by the present invention and belong to the protection scope of the present invention.

Claims

1. A method for detecting the content of poly-α-olefin in a sample, characterized in that: The method includes: (1) providing a plurality of standard poly-alpha-olefin samples with known and different poly-alpha-olefin contents; (2) For the above standard poly-α-olefin samples, near infrared spectra X NIR , mid-infrared spectroscopy X MIR and Raman spectroscopy Raman ; Select characteristic spectral bands of the above spectra respectively, and obtain corresponding absorbance matrix or spectral intensity matrix; (3) Take the near infrared spectrum X NIR The absorbance matrix corresponding to the characteristic spectral bands, mid-infrared spectrum X MIR The absorbance matrix and Raman spectrum X corresponding to the characteristic spectral band Raman The spectral intensity matrix corresponding to the characteristic spectral bands of NIR , P MIR and P Raman ; (4) The prediction model P NIR , P MIR and P Raman Perform weighted fusion to obtain a fusion model M; (5) Taking a sample with unknown poly-α-olefin content, processing it in the manner of step (2), and substituting the absorbance matrix and spectral intensity matrix corresponding to the obtained characteristic spectral band into the fusion model M to obtain the poly-α-olefin content in the sample.

2. The method according to claim 1, wherein: In step (1), the number of the standard poly-α-olefin samples is ≥50, preferably 60-200; And / or, in step (1), the poly-α-olefin content in the standard poly-α-olefin sample satisfies: the mass concentration is greater than 30 wt%, preferably 42-99.5 wt%.

3. The method according to claim 1 or 2, wherein: In step (2), the spectrum collection conditions include: the collection temperature is 15-40°C, preferably 23-27°C; and / or, the acquisition resolution is 2-16cm -1 , preferably 6-10cm -1 .

4. The method according to claim 1, wherein: In step (2), the near infrared spectrum is collected in the wave number range of 4000-10000cm -1 ; And / or, the mid-infrared spectrum is collected in the wavenumber range of 650-4000cm -1 ; And / or, the Raman spectrum is collected in the wave number range of 100-2500cm -1 .

5. The method according to claim 1 or 4, wherein: Before selecting the characteristic spectral band, the method further includes: preprocessing the spectrum to eliminate redundant information and / or noise in the spectrum; Preferably, the near infrared spectrum X NIR and mid-infrared spectroscopy MIR The preprocessing methods are independently selected from at least one of second-order differential, standard normal variable transformation and first-order differential, more preferably second-order differential; Preferably, the Raman spectrum X Raman The preprocessing method is selected from at least one of multivariate scatter correction, standard normal variable transformation and baseline correction, and is more preferably multivariate scatter correction.

6. The method according to claim 1 or 4, wherein: Near Infrared SpectroscopyX NIR The wave number range of the characteristic spectral band is 4500-7800cm -1 , preferably 4227-4651cm -1 5422-6001cm -1 and 7416-9002cm -1 ; and / or, the characteristic spectral band of the mid-infrared spectrum has a wave number interval of 650-4000cm -1 , preferably 1603-1663cm -1 and 650-980cm -1 ; And / or, the wave number interval of the characteristic spectral band of the Raman spectrum is 100-2500cm -1 , preferably 1018-1755cm -1 .

7. The method according to claim 1, wherein: The regression analysis method is selected from at least one of least squares-support vector machine, partial least squares method and artificial neural network, preferably least squares-support vector machine.

8. The method according to claim 7, wherein: The model parameters smoothing parameter σ and regularization parameter γ in the least squares support vector machine are determined according to at least one of leave-one-out method, Bayesian reasoning and leave-one-out cross-validation, preferably according to leave-one-out cross-validation.

9. The method according to claim 1 or 2, wherein: Step (2) also includes: normalizing the absorbance matrix or spectral intensity matrix corresponding to the characteristic spectral band.

10. The method according to claim 1, wherein: Step (4) weighted fusion includes: near infrared spectrum X NIR The absorbance matrix corresponding to the characteristic spectral band is substituted into the prediction model P NIR , and get the prediction result Y NIR ; Mid-infrared spectroscopy MIR The absorbance matrix corresponding to the characteristic spectral band is substituted into the prediction model P MIR , and get the prediction result Y MIR ; Raman spectroscopy Raman The spectral intensity matrix corresponding to the characteristic spectral band is substituted into the prediction model P Raman , and get the prediction result Y Raman ; Assign weights a, b and c to the above prediction results respectively. NIR +bY MIR +cY Raman Adjust a, b and c to aY NIR +bY MIR +cY Raman The calculated result is close to the real poly-α-olefin content; wherein a, b, c are respectively greater than 0 and less than 1, and a+b+c=1; Get the fusion model M = aP NIR +bP MIR +cP Raman .

11. The method according to claim 1 or 10, wherein: In step (1), the sample is poly-α-olefin synthetic oil.

Images