Rapid detection method for raffinate concentration in hydrogen peroxide production process

Abstract

The invention relates to the field of spectrum detection, and discloses a rapid detection method for raffinate concentration in a hydrogen peroxide production process. The method comprises the following steps: (1) obtaining a plurality of standard raffinate; (2) obtaining a characteristic spectrum wave band of the standard raffinate; (3) obtaining a global principal component score of each characteristic spectral band; segmenting each characteristic spectrum wave band, and obtaining a block score of each segment of wave band after segmentation; (4) using the global principal component score and the score of each block as input variables, using the raffinate concentration as an output variable, and adopting a random forest algorithm to construct a prediction model; and (5) obtaining the raffinate to be detected and the corresponding global principal component score and block score, and substituting the scores into the prediction model to obtain the raffinate concentration. The method is particularly suitable for detection of the raffinate concentration of the raffinate, and has the advantages of rapidness, simplicity, high accuracy, good repeatability, and better operation and accuracy advantages.

Description

Technical Field

[0001] The present invention relates to the field of spectral detection, and in particular to a method for quickly detecting raffinate concentration in a hydrogen peroxide production process. Background Art

[0002] The production of hydrogen peroxide (Hydrogen Peroxide) primarily utilizes the anthraquinone process. However, the raw materials, intermediates, and products used in this process are often flammable, explosive, or contain combustion-supporting substances. Therefore, accurate and timely monitoring of the quality indicators of the hydrogen peroxide production process is crucial to guide production.

[0003] The raffinate is the liquid remaining after hydrogen peroxide is extracted from the oxidation reaction during the anthraquinone process for hydrogen peroxide production. Its primary components are anthraquinone and solvent, along with some catalysts and organic impurities such as organic acids. It also contains a small amount of hydrogen peroxide. The concentration of hydrogen peroxide in the raffinate is the raffinate concentration. Raffinate concentration is a critical analytical parameter in the anthraquinone process, reflecting the quality of hydrogen peroxide production and closely related to production safety. The raffinate concentration is generally determined by taking a sample, adding potassium permanganate titrant, and calculating the raffinate concentration (the actual hydrogen peroxide content in the sample) based on the amount of titrant consumed at the titration endpoint. To ensure hydrogen peroxide production quality and meet safety requirements, raffinate concentration analysis is performed very frequently during production, typically every 2-4 hours. However, the currently prevalent manual titration method is cumbersome, slow, and requires high operator input. Frequent titration analysis is labor-intensive and prone to human error.

[0004] Therefore, it is necessary to develop a method for accurately, simply and quickly obtaining the raffinate concentration of the raffinate, so as to quickly and accurately determine the raffinate concentration in the raffinate during the production process of hydrogen peroxide. Summary of the Invention

[0005] The purpose of the present invention is to overcome the above-mentioned problems existing in the prior art and provide a method for rapid detection of raffinate concentration in the hydrogen peroxide production process. The method is particularly suitable for detecting raffinate concentration and is fast, simple, highly accurate, and reproducible, and has advantages in operation and accuracy.

[0006] In order to achieve the above object, the present invention provides a method for rapidly detecting the raffinate concentration in the hydrogen peroxide production process, the method comprising:

[0007] (1) obtaining a plurality of standard raffinates with known and different raffinate concentrations;

[0008] (2) obtaining a spectrum of the standard raffinate and selecting a characteristic spectral band;

[0009] (3) Using multi-block principal component analysis to obtain the global principal component score of each characteristic spectral band; each characteristic spectral band is segmented, and using multi-block principal component analysis to obtain the principal component score of each segmented band, which is the block score of each segmented band;

[0010] (4) The global principal component scores and the scores of each block corresponding to each characteristic spectral band are used as input variables, and the corresponding residual concentration obtained in step (1) is used as the output variable. A random forest algorithm is used to construct a prediction model between the input variables and the output variables;

[0011] (5) Obtain a raffinate to be tested with an unknown raffinate concentration, obtain the corresponding global principal component score and block score according to the method in steps (2)-(3), and substitute them into the prediction model obtained in step (4) to obtain the raffinate concentration of the raffinate to be tested.

[0012] The above technical solution can quickly, simply, and accurately determine the raffinate concentration of the raffinate, and the method has good repeatability and stability. Compared with the existing technology, the method provided by the present invention can avoid complex manual processing, is more efficient, and greatly improves the analytical operating environment, which has obvious advantages. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 3 is a correlation diagram of the measured values ​​and the calculated values ​​obtained in Example 1 of the present invention. DETAILED DESCRIPTION

[0014] The endpoints of the ranges and any values ​​disclosed herein 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 endpoints of each range, the endpoints of each range and 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 to be specifically disclosed herein.

[0015] The present invention provides a method for quickly detecting the raffinate concentration in a hydrogen peroxide production process, the method comprising:

[0016] (1) obtaining a plurality of standard raffinates with known and different raffinate concentrations;

[0017] (2) obtaining a spectrum of the standard raffinate and selecting a characteristic spectral band;

[0018] (3) Using multi-block principal component analysis to obtain the global principal component score of each characteristic spectral band; each characteristic spectral band is segmented, and using multi-block principal component analysis to obtain the principal component score of each segmented band, which is the block score of each segmented band;

[0019] (4) The global principal component scores and the scores of each block corresponding to each characteristic spectral band are used as input variables, and the corresponding residual concentration obtained in step (1) is used as the output variable. A random forest algorithm is used to construct a prediction model between the input variables and the output variables;

[0020] (5) Obtain a raffinate to be tested with an unknown raffinate concentration, obtain the corresponding global principal component score and block score according to the method in steps (2)-(3), and substitute them into the prediction model obtained in step (4) to obtain the raffinate concentration of the raffinate to be tested.

[0021] It is understood that hydrogen peroxide is generally produced using the anthraquinone process, and the main reaction liquids involved in the production process generally include hydrogenation liquid, oxidation liquid, and raffinate. The main process of the anthraquinone process includes: anthraquinone (generally 2-ethylanthraquinone) is mixed with an organic solvent (such as a mixed solvent of C9-C10 heavy aromatic hydrocarbons, trioctyl phosphate, and tetrabutyl urea) to form a working liquid. The hydrogenation stage: Under conditions of pressure of 0.3MPa or above, temperature of 40-80°C, and catalyst (such as Pd catalyst), the anthraquinone in the working liquid is hydrogenated and reduced with H2 to produce anthrahydroquinone or tetrahydroanthrahydroquinone. The oxidation stage: The material after the hydrogenation stage is further oxidized with O2 at 30-60°C under slight compression, so that the anthrahydroquinone and tetrahydroanthrahydroquinone are oxidized to produce H2O2 and anthraquinone. The material after the oxidation stage is then extracted (the raffinate is the solution remaining after the H2O2 extraction), regenerated, refined, and concentrated to produce a 20wt%-50wt% H2O2 aqueous solution.

[0022] The raffinate concentration refers to the concentration of hydrogen peroxide in the remaining liquid after hydrogen peroxide extraction, and this concentration is generally low. The inventors of the present invention have discovered that, for raffinates from anthraquinone processes, processing them according to the above method, particularly extracting global scores and block scores through multi-block principal component analysis, can extract as much spectral information related to raffinate concentration as possible while largely eliminating the spectral influence of free water in the raffinate, resulting in highly accurate analysis results. A prediction model is established using a random forest algorithm with nonlinear regression, which exhibits good stability and ensures high accuracy when used to calculate unknown raffinate concentrations. In particular, raffinate concentrations are generally low, making traditional manual detection methods more prone to error risks. Compared to traditional manual methods or prediction models constructed using other methods (such as partial least squares (PLS)), the method provided by the present invention produces more accurate results for raffinates with low raffinate concentrations, which are more prone to error. The method of the present invention is particularly suitable for detecting raffinate concentration in raffinates.

[0023] According to the present invention, preferably, in step (1), the number of standard raffinates is not less than 100, preferably 150-250 (for example, it can be 150, 160, 170, 180, 200, 220, 250, and a range formed by any two of the above values, and values ​​within the range). It is understood that in order to obtain higher accuracy, a slightly larger sample size is generally required when collecting samples. The inventors of the present invention have found that within the above range, the accuracy and stability of the results obtained by the prediction model can be further guaranteed.

[0024] According to the present invention, preferably, in step (1), the raffinate concentration of the standard raffinate is 0.03-1 g / L, preferably 0.05-0.6 g / L (for example, it can be 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, and any range formed by any two of the above values, and values ​​within the range). It is understood that when selecting samples, in order to make the resulting model more accurate, it is generally necessary to uniformly distribute the sample conditions within the possible range. Therefore, the raffinate concentration of the standard raffinate is preferably also uniformly distributed within the above range. For example, the difference between the raffinate concentrations of two adjacent raffinate concentration values ​​is preferably within the range of 0.008-0.08 g / L.

[0025] According to the present invention, preferably, in step (2), the spectrum is a near-infrared spectrum.

[0026] The present invention has no particular limitation on the specific method of obtaining the near-infrared spectrum, and the method can be performed according to conventional methods in the art. However, preferably, the conditions for obtaining the near-infrared spectrum include: a temperature of 12-62°C, preferably 20-30°C (for example, 20, 22, 25, 28, 30, and a range formed by any two of the above values, and a value within the range); a wave number range of 3500-12000 cm -1 ; Resolution is 2-16cm -1 , preferably 6-12 (for example, it can be 6, 7, 8, 9, 10, 11, 12 and the range formed by any two of the above values ​​and the value within the range) cm -1 . Can scan 64-128 times.

[0027] When collecting near-infrared spectra, the sample can be injected into the cuvette to about two-thirds of its volume, and the cuvette containing the sample can be sealed with a sealing film. The sealed cuvette can be placed in a temperature-controlled sample cell holder for spectrum collection. The optical path of the cuvette can be 1-2 mm. The instrument used when collecting the near-infrared spectra is not particularly limited and can be a conventional choice in the art, for example, an Antaris II near-infrared spectrometer (Thermo Fisher Corporation) can be used.

[0028] According to the present invention, preferably, the wave number interval of the characteristic spectral band corresponding to the near infrared spectrum is 4556-7596 cm -1 The inventors of the present invention have further discovered that the selection of the above characteristic spectral bands can further fully reflect the information related to the raffinate concentration.

[0029] According to the present invention, preferably, before performing step (3), the method further comprises: preprocessing the characteristic spectral band to reduce redundant information and / or noise in the spectrum.

[0030] According to the present invention, the preprocessing method is preferably selected from the second-order differential or the first-order differential, and more preferably the first-order differential with a window width of 15-27 (e.g., 15, 17, 19, 21, 23, 25, 27, and any range formed by any two of the above values, and values ​​within the range). The inventors of the present invention further discovered that the use of this preprocessing method can better cooperate with the subsequent multi-block principal component analysis method, and even if the raffinate concentration in the raffinate is relatively low, it can further ensure the acquisition of sufficient information and further guarantee the accuracy of the results.

[0031] According to the present invention, preferably, in step (3), the number of segments is 2-4, preferably 3.

[0032] More preferably, in step (3), the number of segments is 3, and the wave number range of the first segment is 4556-5476 cm -1 The second wave number range is 5476-6396cm -1 The third wave number range is 6396-7596cm -1 .

[0033] It is understood that the multi-block principal component analysis method (MBPCA method) is a commonly used data processing method by dimensionality reduction, but the inventors of the present invention have found that for the raffinate, the above method, combined with the characteristic spectral bands, segmentation and random forest algorithm as described above, can accurately determine the raffinate concentration result in the raffinate. The MBPCA method is a common method, and those skilled in the art are familiar with its principles. For example, see Westerhuis JA, Kourti T, MacGregor J F. Analysis of multiblock and hierarchical PCA and PLS models [J]. Journal of Chemometrics: A Journal of the Chemometrics Society, 1998, 12 (5): 301-321. The general principle of MBPCA is as follows:

[0034] For a B-block spectral matrix, block X 1 , Block X 2 ,…,block X B , the matrix X, X=[X 1 X 2 …X B ], MBPCA is decomposed in two steps according to the following formula:

[0035] Full matrix decomposition: X = TP T +E

[0036] Block matrix factorization: X b =T b P bT +E b ,b=1,2,…,B

[0037] Where T and P are the global score matrix and load matrix respectively, T b 、P b They are block matrices X b The score matrix and loading matrix of .

[0038] The MBPCA algorithm steps are as follows:

[0039] (1) Initialize the global score vector t

[0040] (2) Calculate the block score vector t b and the block load vector p b ,b=1,2,…,B

[0041] p b =X bT / t T t

[0042] p b =p b / ||p b ||

[0043] t b =X b p b

[0044] (3) Calculate the block global vector t and weight w, b = 1, 2, ..., B

[0045] T=[t 1 t 2 … t B ]

[0046] w=T T t / t T t

[0047] w=w / ||w||

[0048] t=Tw

[0049] (4) Return to (2) and loop to calculate t until convergence, and obtain t1, w1, and t1 of the first principal component. b and p1 b

[0050] (5) Calculate the residual matrix X2 of each block b , b=1,2,…,B,X2 b =X b -(tt T / t T t)X b

[0051] (6) Use X2 b Replace X b , return to steps (1) to (4) and calculate t2, w2, and t2 of the second principal component b and p2 b

[0052] (7) According to the above cycle, calculate the t of all A principal components in turn. i 、w i , t i b and p i b , i=1,2,…,A.

[0053] Generally, the number of global principal component scores of each characteristic spectral band and the number of block scores of each segmented band can be independently greater than 6, for example, can be independently 8-12.

[0054] Among them, the Random Forest (RF) algorithm is a fusion classification algorithm that includes many decision trees and voting strategies. It belongs to an integrated algorithm and is also a natural nonlinear modeling tool that can be used for classification or regression analysis. The inventors of the present invention have found in their research that, compared with other algorithms, such as the classic partial least squares (PLS) linear modeling algorithm, the use of the Random Forest algorithm to construct the prediction model of the present invention has high accuracy, good tolerance to outliers and noise, and is not prone to overfitting; in particular, when combined with the above-mentioned multi-block principal component analysis method, the obtained model can also ensure high accuracy for the raffinate with generally low raffinate concentration. As for the Random Forest algorithm itself, it is a common algorithm. Those skilled in the art are also familiar with the principle. The above algorithm has been open sourced and applied. For its specific content, please refer to the description in Fang Kuangnan; Wu Jianbin; Zhu Jianping; Xie Bangchang. A review of the research on random forest method [J]. Statistics and Information Forum, 2011, 26(03): 32-38. The algorithm can be implemented in MATLAB.

[0055] Among them, in the random forest (RF) algorithm, the ntree (generally referring to the number of sampling training) and mtry (generally referring to the number of features selected for each sampling training) parameters can also be controlled. In the present invention, ntree can be set to 100-500, and mtry can be set to 9-18, especially 12-15.

[0056] In the present invention, the multi-block principal component analysis method and the random forest method are organically combined. That is, the global principal component score and each block score obtained by the multi-block principal component analysis method are used as input variables of the random forest method, and the actual value of the raffinate concentration is used as the output variable. This ensures that the obtained prediction model has good accuracy and stability when determining the raffinate concentration in the raffinate to be measured.

[0057] It can be understood that in step (5), the raffinate to be tested with unknown raffinate concentration is obtained, and the spectrum and characteristic spectral band are obtained (corresponding preprocessing and segmentation are performed) in accordance with the method in steps (2)-(3), and the global principal component score of the characteristic spectral band and the block score of each band after segmentation are obtained accordingly, and they are substituted into the prediction model to obtain the result of the raffinate concentration.

[0058] The present invention will be described in detail below through examples.

[0059] In the following examples, the raffinate was obtained during the production of hydrogen peroxide using the anthraquinone process.

[0060] In step (1) of the following examples, the raffinate concentration of the standard raffinate is determined by potassium permanganate titration, and the raffinate concentration value obtained in this manner is the actual measured value.

[0061] The instrument used to collect near-infrared spectra was an Antaris II near-infrared spectrometer (Thermo Fisher Scientific).

[0062] The operations after spectrum acquisition were performed in MATLAB.

[0063] Example 1

[0064] To illustrate the method provided by the present invention

[0065] (1) 200 raffinates were obtained, and their corresponding measured values ​​were obtained. The measured values ​​were distributed in the range of 0.06-0.51 g / L, and the difference between every two adjacent raffinate concentration values ​​was in the range of 0.01-0.05 g / L.

[0066] (2) The samples were injected into cuvettes (optical path length 2 mm) until the volume of the cuvettes reached two-thirds. The cuvettes containing the samples were sealed with sealing film and placed in a temperature-controlled sample cell rack for spectrum acquisition. The conditions for acquiring near-infrared spectra included: temperature of 25°C, and the acquisition wavenumber range of 3500-10000 cm -1 , with a resolution of 8cm -1 , scanned 128 times. For the obtained near-infrared spectrum, the characteristic spectrum band 4556-7596cm was selected -1 .

[0067] (3) Perform first-order differentiation of the characteristic spectral bands obtained above with a window width of 21.

[0068] The multi-block principal component analysis method is used for each characteristic spectral band after the above processing. Among them, the number of global principal component scores of each characteristic spectral band is 9.

[0069] Then, each characteristic spectral band is divided into three sections according to the following method. The wave number range of the first section is 4556-5476cm -1 The second wave number range is 5476-6396cm -1 The third wave number range is 6396-7596cm -1 Using multi-block principal component analysis, we obtain nine block scores for each segmented band, for a total of 27 block scores for the three segments. Therefore, for each characteristic spectral band, there are nine global principal component scores and 27 block scores, for a total of 36 scores.

[0070] (4) Using the random forest algorithm, the 36 scores of each sample are taken as input variables, and the measured value of the residual concentration corresponding to each sample is taken as the output variable. ntree is set to 400 and mtry is set to 12 to construct a prediction model M between the input variables and the output variables.

[0071] (5) Take 39 raffinates with unknown raffinate concentrations as the validation set and process them according to the method in steps (2)-(3). Then substitute the obtained global principal component scores and each block score into the model obtained in step (4) to obtain the corresponding raffinate concentration (calculated value).

[0072] In order to verify the accuracy of the prediction model, the measured values ​​corresponding to each of the 39 samples in the validation set were also obtained.

[0073] The measured and calculated values ​​of the 39 samples in the validation set, as well as the deviation between the two (calculated value minus measured value) are shown in Table 1. In addition, the root mean square error of prediction (RMSEP) and correlation coefficient (R) are used to evaluate the performance of the model, where

[0074]

[0075] Among them, n is the total number of samples in the validation set, y i is the measured value of the i-th sample, is the calculated value of the i-th sample.

[0076]

[0077] Among them, n is the total number of samples in the validation set, y i is the measured value of the i-th sample, is the calculated value of the i-th sample, is the average value of the actual values ​​of the validation set samples. By calculation, we can get R = 0.988.

[0078] Table 1

[0079]

[0080]

[0081] In addition, a correlation diagram is drawn using the measured values ​​(x-axis) and calculated values ​​(y-axis) of the 39 samples in the validation set, see Figure 1 (where R is the correlation coefficient).

[0082] It can be seen that the method provided by the present invention can quickly, simply and accurately determine the raffinate concentration of the raffinate in the process of producing hydrogen peroxide by the anthraquinone method, and the prediction results at various concentrations are relatively accurate and have high accuracy.

[0083] Example 2

[0084] Used to evaluate the repeatability of the method provided by the present invention

[0085] Take two raffinates to be tested, and repeat the near-infrared spectrum measurement three times for each sample according to the method of step (2) in Example 1. The three near-infrared spectra are processed as follows: obtain the characteristic spectral band according to the method in Example 1, and then obtain the global principal component score and block score according to step (3) in Example 1. Substitute the predicted model obtained in Example 1 to obtain the raffinate concentration (calculated value). In this way, three results for each sample in three parallel calculations are obtained, as shown in Table 2. Among them, the relative standard deviation is calculated as follows:

[0086]

[0087] 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 iRepresents the calculated value of the i-th sample.

[0088] Table 2

[0089]

[0090] As shown in Table 2, the relative standard deviations of the three replicate calculations of the raffinate concentrations for different samples using the method of the present invention were all below 3.0%. Therefore, the method provided by the present invention exhibits good repeatability when determining the raffinate concentrations of the raffinate samples to be tested. As demonstrated above, the method provided by the present invention exhibits high repeatability and good stability.

[0091] Example 3

[0092] The method of Example 1 is followed, except that ntree is set to 60 and mtry is set to 6. The results are shown in Table 3.

[0093] Table 3

[0094]

[0095]

[0096] It can be seen that compared with Example 1, the RMSEP is slightly increased, and Example 1 has better accuracy.

[0097] Comparative Example 1

[0098] The method of Example 1 was followed, except that the random forest algorithm was replaced with the partial least squares algorithm to establish the model. The results are shown in Table 4.

[0099] Table 4

[0100]

[0101] It can be seen that the accuracy of the model obtained in this way is poorer than that of the solution provided by the present invention.

[0102] The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited thereto. Within the technical concept of the present invention, various simple variations of the technical solution of the present invention may be made, including combining the various technical features in any other appropriate manner. These simple variations and combinations should also be regarded as disclosed in the present invention and fall within the scope of protection of the present invention.

Claims

1. A rapid detection method for raffinate concentration in a hydrogen peroxide production process, characterized in that: The method includes: (1) obtaining a plurality of standard raffinates with known and different raffinate concentrations; (2) obtaining a spectrum of the standard raffinate and selecting a characteristic spectral band; (3) Using multi-block principal component analysis to obtain the global principal component score of each characteristic spectral band; each characteristic spectral band is segmented, and using multi-block principal component analysis to obtain the principal component score of each segmented band, which is the block score of each segmented band; (4) The global principal component scores and the scores of each block corresponding to each characteristic spectral band are used as input variables, and the corresponding residual concentration obtained in step (1) is used as the output variable. A random forest algorithm is used to construct a prediction model between the input variables and the output variables; (5) Obtain a raffinate to be tested with an unknown raffinate concentration, obtain the corresponding global principal component score and block score according to the method in steps (2)-(3), and substitute them into the prediction model obtained in step (4) to obtain the raffinate concentration of the raffinate to be tested.

2. The method according to claim 1, wherein In step (1), the number of standard raffinates is not less than 100, preferably 150-250.

3. The method according to claim 1 or 2, wherein: In step (1), the raffinate concentration of the standard raffinate is 0.03-1 g / L, preferably 0.05-0.6 g / L.

4. The method according to claim 1, wherein In step (2), the spectrum is a near-infrared spectrum.

5. The method according to claim 4, wherein The conditions for obtaining near-infrared spectra include: temperature of 12-62°C, preferably 20-30°C; wave number range of 3500-12000cm -1 ; Resolution is 2-16cm -1 , preferably 6-12cm -1 .

6. The method according to claim 4, wherein: The wave number range of the characteristic spectral band corresponding to the near-infrared spectrum is 4556-7596 cm -1 .

7. The method according to claim 4 or 6, wherein: Before performing step (3), the method further includes: preprocessing the characteristic spectrum band to reduce redundant information and / or noise in the spectrum.

8. The method according to claim 7, wherein: The preprocessing method is selected from the second order differential or the first order differential, and more preferably the first order differential with a window width of 15-27.

9. The method according to claim 6, wherein: In step (3), the number of segments is 2-4, preferably 3.

10. The method according to claim 9, wherein: In step (3), the number of segments is 3, and the wave number range of the first segment is 4556-5476cm -1 The second wave number range is 5476-6396cm -1 The third wave number range is 6396-7596cm -1 .

Images