A method for detecting trace impurities in ethanol solutions based on near-infrared spectroscopy

Abstract

This invention belongs to the field of impurity detection technology for ethanol solutions, and particularly relates to a method for detecting trace impurities in ethanol solutions based on near-infrared spectroscopy, comprising the following steps: (1) examining the standard deviation (RSD) value and calculating the required sample size N; (2) acquiring samples; (3) outlier removal and spectral preprocessing; (4) dividing the dataset; (5) using the difference spectrum method to screen feature signals; (6) establishing a "background silence" discrimination criterion; and (7) using an external prediction set for discrimination. The detection method model described in this invention has a wide range of applications. It establishes a model based on the background system, effectively identifying normal samples. For impurity samples or abnormal situations that are inconsistent with the near-infrared absorption characteristics of the background system, as long as they cause changes in the background system, they can be effectively detected. Compared with traditional PLS, PLSDA and other modeling methods, background silence discrimination only requires calculation using relevant statistical methods, resulting in a smaller overall computational load and more convenient data processing.

Description

Technical Field

[0001] This invention belongs to the field of impurity detection technology for ethanol solutions, and particularly relates to a method for detecting trace impurities in ethanol solutions based on near-infrared spectroscopy. Background Technology

[0002] Ethanol, as an important organic solvent, chemical raw material, and chemical fuel, has a wide range of applications in the medical, food, chemical, and pharmaceutical industries. In the pharmaceutical field, one important application of ethanol is as a precipitation medium for the extraction and separation of substances such as proteins, namely, the low-temperature ethanol precipitation method for protein extraction and separation. The low-temperature ethanol precipitation method is a classic process for plasma protein separation and purification. Its main steps involve adding low-temperature ethanol to the protein solution in a certain proportion, and then adjusting the pH to precipitate the target protein.

[0003] The low-temperature ethanol required for protein precipitation is typically between -15°C and -20°C. Before being added to the extraction tank, it needs to be cooled by a refrigerant (mainly ethylene glycol). This cooling process involves counter-current flow of ethanol and refrigerant in a plate heat exchanger to exchange heat and lower the ethanol temperature. Due to corrosion from the solution within the pipes and routine oxidative corrosion, subtle pipe damage may occur, posing a risk of refrigerant leakage into the ethanol solution. This can contaminate the subsequent protein precipitation process, affecting the purity of the final product and causing significant economic losses.

[0004] The main method for detecting refrigerant impurities in ethanol solutions is gas chromatography. Although this method has high sensitivity and accuracy, it cannot be used for online detection due to the large size of the instrument, high requirements for the working environment, and slow detection speed. Offline detection by sampling has a certain lag, and abnormalities often cannot be detected in time, which can still cause contamination in subsequent processes.

[0005] Therefore, we need a process analysis method to ensure that leaks can be detected and production stopped in a timely manner through real-time monitoring, preventing contaminated ethanol from entering the point of use and avoiding protein contamination and loss in the alcohol precipitation tank. Near-infrared spectroscopy (NIRS) is a PAT technique that can qualitatively or quantitatively identify impurities in solution and perform real-time online monitoring. It is sensitive to the combination and overtone absorption of hydrogen-containing group vibrations, making it particularly suitable for detecting hydrogen-containing groups. However, the use of near-infrared spectroscopy is limited by its relatively poor sensitivity and severe interference from background information.

[0006] Therefore, we propose a near-infrared spectroscopy background silencing method to eliminate background information interference as much as possible, thereby improving its sensitivity and accuracy in detecting trace impurities in ethanol solutions and providing a new solution for impurity detection in actual production. Summary of the Invention

[0007] To solve the problems existing in the above-mentioned prior art, the present invention provides a new method for discriminating impurities in near-infrared spectroscopy - background silence method, which can accurately discriminate trace refrigerant impurities in ethanol solution.

[0008] To achieve the above object, the present invention adopts the following technical solutions: A method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy, comprising the following steps: 1) Examine the sample size N required for calculating the relative standard deviation (RSD). Use MATLAB software to randomly generate 10 groups of data between 99 and 100, with 50 in each group. Calculate the RSD values between each group of data in turn, and count the amount of data when the RSD of each group reaches a stable state to determine the N value.

[0009] 2) Take a total of 55 ethanol solution samples in actual production, and use them as solvents to prepare a series of 46 groups of ethanol solutions with refrigerant impurities added at a concentration of one ten-thousandth to one percent. Collect their near-infrared spectra respectively, and obtain 55 ethanol sample spectra x and * , , * 46 impurity sample spectra.

[0010] 3) Eliminate abnormal samples, and obtain 52 ethanol sample spectra x and * 45 impurity sample spectra. Pretreat the normal sample spectra to eliminate baseline drift of the spectra and interference of other irrelevant information.

[0011] 4) Take the average value of the ethanol spectra x to obtain the average spectrum nx.

[0012] 5) Subtract nx from x and x respectively * to obtain the difference spectra of the two, select the negative peak of the difference spectrum as the characteristic peak, and record its wave number.

[0013] 6) Screen out the absorption values corresponding to the above wave numbers in the original spectra x and x * respectively, then sort the data points of the ethanol spectra using the KS algorithm, and divide the data set according to a certain ratio, that is, select the first 34 representative data points as the main body to calculate the RSD value, and calculate the RSD value of the remaining 18 data points with the main body data points respectively to obtain the maximum RSD value RSD max of the ethanol sample.

[0014] 7) Use the data points of the impurity spectra and 20 data points of the ethanol spectra as external validation set samples, and calculate the RSD value with the main body data points respectively. If RSD < RSDmax, it is determined as a normal sample; if RSD > RSDmax, it is determined as an abnormal sample.

[0015] Beneficial effects This invention discloses a method for detecting trace impurities in ethanol solutions based on near-infrared spectroscopy. Compared with existing technologies, this invention has the following advantages: 1. The model has a wide range of applications; it is built based on the background system, effectively identifying normal samples, while for impurity samples or abnormal situations that are inconsistent with the near-infrared absorption characteristics of the background system, as long as they cause changes in the background system, they can be effectively detected. In contrast, traditional modeling methods require a separate discrimination model for each type of impurity.

[0016] 2. Data processing is simple and computationally intensive; compared with traditional PLS, PLSDA and other modeling methods, background silence discrimination only requires the calculation of relevant statistical methods, the overall computational workload is small and data processing is more convenient. Attached Figure Description

[0017] Figure 1 RSD deviation statistics chart; Figure 2 Outlier removal chart for ethanol solution samples; Figure 3 Outlier removal chart for ethanol-coolant solution samples; Figure 4 : Spectrum image after SNV preprocessing; Figure 5 Differential spectra of ethanol and ethanol-coolant solution; Figure 6 : Result of the discrimination. Detailed Implementation

[0018] The present invention will now be described in detail. Before proceeding with the description, it should be understood that the terminology used in this specification and the appended claims should not be construed as limited to its general or dictionary meaning, but rather should be interpreted according to the meaning and concept corresponding to the technical aspects of the invention, based on the principle that the inventors are allowed to appropriately define the terms for the best interpretation. Therefore, the description presented herein is merely a preferred example for illustrative purposes and is not intended to limit the scope of the invention. It should be understood that other equivalents or modifications can be obtained from it without departing from the spirit and scope of the invention.

[0019] The following embodiments are merely examples illustrating implementations of the present invention and do not constitute any limitation on the present invention. Those skilled in the art will understand that modifications made without departing from the spirit and concept of the present invention fall within the protection scope of the present invention. Unless otherwise specified, the reagents and instruments used in the following embodiments are commercially available products. Example

[0020] The near-infrared spectroscopy "background silencing" method is used for the detection of refrigerant impurities in ethanol solutions, and includes the following steps: 1. Examination of the sample size required for RSD calculation: Ten sets of data between 99 and 100 were randomly generated using MATLAB software, with 50 data points in each set. The RSD value was calculated sequentially between each set. The previous value was subtracted from each subsequent RSD value to obtain the deviation between any two RSD values. A deviation within ±0.01 was considered a stable state. The results are shown in the attached figure. Figure 1 As shown, the number of data points for each group of RSD values ​​to reach a steady state is counted, and the sample size N required to apply this method is determined to be approximately 30.

[0021] Table 1. Statistical table of the number of calculations in which RSD values ​​reached a steady state. 2. Sample Acquisition: Fifty-five ethanol solution samples were taken from actual production, each sample containing 15 ml. An additional 20 ethanol solution samples were collected as an external prediction set. Two ml of each ethanol sample was reserved for spectral acquisition. The remaining ethanol was used as a solvent to prepare 23 ethanol-coolant solutions of different concentrations. Two sets of solutions were prepared for each concentration, for a total of 46 solutions. The preparation methods are shown in Table 2.

[0022] After preparation, the near-infrared spectrum of the sample was acquired using a Matrix-F Bruker near-infrared spectrometer, with a wavenumber range of 12000 cm⁻¹. -1 -4000 cm -1 The resolution is 4 cm. -1 The scan was performed 64 times, with each sample measured 3 times and the average taken as the final spectrum. This resulted in 55 spectra of ethanol samples and 55 spectra of impurity samples. * 46 images.

[0023] Table 2 Solution Preparation Details PS: Concentrated storage a: Accurately measure 1.0 ml of ethylene glycol and 19.0 ml of the production ethanol solution using a pipette, mix them evenly in a beaker to obtain 20 ml of 5% ethanol-ethylene glycol aqueous solution; Concentrated solution b: Accurately measure 1.0 ml of concentrated solution a and 9.0 ml of the production ethanol solution using a pipette, mix them evenly in a beaker to obtain 10 ml of 0.5% ethanol-diol aqueous solution.

[0024] 3. Outlier removal and spectral preprocessing: The spectra of the ethanol sample x and the spectra of the impurity sample x were analyzed respectively. * PCA analysis was performed to remove abnormal spectra, resulting in 52 spectra of ethanol samples and 45 spectra of impurity samples. Normal spectra were then preprocessed using SNV (Spectral Navier-Layered Spectra) to eliminate baseline drift and the effects of scattering. The results are attached. Figure 4 As shown.

[0025] 4. Split the dataset: The KS algorithm was used to sort the ethanol sample spectra x according to representativeness. The first 34 samples were selected as the main body for RSD calculation. The remaining 18 samples were used to calculate RSD with the main body to establish the normal range of RSD. The external prediction set was used to verify the reliability of the results.

[0026] 5. Differential spectral method for screening feature signals: The average spectrum nx is obtained by averaging the ethanol sample spectrum x; the average spectrum nx is obtained by averaging the ethanol sample spectrum x and the impurity sample spectrum x respectively. * Subtracting the average spectrum nx yields the difference spectrum between the two, as shown in the attached figure. Figure 5 As shown, the negative peak of the ethanol sample spectrum was taken as the research object, that is, the absorption of the background system was reduced due to the addition of impurities. A total of 6 characteristic wavenumbers corresponding to the peak were recorded.

[0027] 6. Establish criteria for judging the "background silence" method: The absorbance values ​​corresponding to characteristic wavenumbers were selected from the preprocessed spectra of both samples, and the RSD values ​​were calculated. The RSD values ​​of 18 ethanol samples were calculated against 34 main sample samples, and the maximum RSD value of the ethanol samples was obtained. max RSD max That is, it serves as the criterion for distinguishing between impurities and normal samples.

[0028] Table 3. RSD Calculation Results and Maximum Value Statistics 7. Use external prediction sets for discrimination: After the above processing was performed on 20 ethanol solution samples and 45 ethanol-refrigerant solution samples, RSD was calculated and compared with 34 ethanol samples to make a judgment.

[0029] 8. Judgment of the discrimination result: An impurity discrimination system is established based on the discrimination results at six wavenumber points. That is, the discrimination is performed in the form of probability according to the calculation results of multiple wavenumber points, so as to adapt to the needs of different application scenarios.

[0030] If there are 5-6 wavenumber points where RSD > RSD max This indicates that there is a very high probability that impurities have been introduced. If there are 3-4 wavenumber points where RSD > RSD max This indicates a high probability that impurities have been introduced. If there are 1-2 wavenumber points where RSD > RSD max This indicates that the possibility of impurities being mixed in is relatively small; If all are less than RSD max This indicates that no impurities have been added.

[0031] Due to RSD max The values ​​are determined based on only a small number of normal samples, which cannot eliminate all random errors and may misclassify some normal values ​​as outliers. Therefore, in practical work, to avoid a large number of "false positives," the discrimination criteria can be appropriately adjusted. In this experiment, we used RSD... max The RSD discrimination criteria for the six wavenumber points were adjusted to 0.30%, 0.32%, 0.32%, 1.40%, 2.20%, and 0.8%, respectively. Under these criteria, the method achieved a 100% recognition rate for normal ethanol samples and a 100% rejection rate for abnormal samples with impurities (minimum volume fraction of 0.01%). The results are attached. Figure 6 As shown.

[0032] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions claimed by the present invention.

Claims

1. A method for detecting trace impurities in ethanol solutions based on near-infrared spectroscopy, characterized in that, Includes the following steps: (1) Examine the standard deviation (RSD) value and calculate the required sample size N: Generate multiple sets of data between 99 and 100, calculate the standard deviation (RSD) value between each set of data, count the number of data sets whose RSD values ​​reach a steady state, and determine the required sample size N. (2) Obtaining samples: Ethanol solution samples were taken from actual production. A portion of each ethanol sample was reserved for spectral acquisition, resulting in the ethanol sample spectrum x. The remaining ethanol was used as a solvent to prepare ethanol-coolant solutions of different concentrations as impurity samples. Their near-infrared spectra were acquired, resulting in the impurity sample spectrum x. In addition, multiple ethanol solution samples were collected as external prediction samples, and their near-infrared spectra were acquired to obtain the external prediction sample spectra. (3) Outlier removal and spectral preprocessing: Outlier samples are removed, and the spectra of normal samples are preprocessed to eliminate baseline drift and other irrelevant information interference. (4) Divide the dataset: Use the KS algorithm to sort the ethanol sample spectra x after preprocessing in step (3) according to representativeness, divide the verification set according to proportion, select a representative part of the samples as the main sample for calculation; the remaining samples are used as comparison samples. (5) Differential spectral method for screening feature signals: Take the average value of the ethanol sample spectrum x to obtain the average spectrum nx; use the ethanol sample spectrum x and the impurity sample spectrum x respectively. Subtract nx to obtain the difference spectrum of the two, select the negative peak of the difference spectrum as the characteristic peak, and record its wavenumber as the characteristic wavenumber; (6) Establish the "background silencing" method discrimination criteria: screen out the spectra x of the pretreated ethanol sample and the spectra x of the impurity sample. The RSD value is calculated based on the absorption value corresponding to the characteristic wavenumber described in step (5). The calculation subject sample described in step (4) is used to calculate the RSD value. The comparison sample is compared with the calculation subject sample to calculate the RSD value, and the maximum RSD value of the ethanol sample is obtained. max RSD max That is, it serves as the criterion for distinguishing between impurities and normal samples; (7) Discrimination using external prediction set: The spectra of impurity samples x The data points and the data points of the external predicted sample spectrum in step (2) are used as the external validation set, and RSD values ​​are calculated with the data points of the calculation subject sample respectively. If the RSD <RSD max If RSD > RSD, then it is considered a normal sample. max If so, it is determined to be an abnormal sample; In step (7), an impurity discrimination system is established based on the discrimination results at the wavenumber points, that is, discrimination is performed in the form of probabilities according to the calculation results of multiple wavenumber points: If there are RSD > RSD at 5 / 6 to 6 / 6 wavenumber points max This indicates that there is a very high probability that impurities have been introduced. If there are RSD > RSD at 3 / 6 to 4 / 6 of the wavenumber points max This indicates a high probability that impurities have been introduced. If there are RSD > RSD at 1 / 6 to 2 / 6 of the wavenumber points max This indicates that the possibility of impurities being mixed in is relatively small; If all calculated RSD values ​​are less than RSD max This indicates that no impurities have been added.

2. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 1, characterized in that, In step (1), multiple sets of data between 99 and 100 are randomly generated using MATLAB software, with 50 data in each set. The standard deviation (RSD) value is calculated sequentially between each set of data.

3. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 2, characterized in that, In step (1), the obtained RSD value is subtracted from its previous value in turn to obtain the deviation between every two RSD values. The deviation is considered to be in a stable state when it is between ±0.

01.

4. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 1, characterized in that, In step (2), take 15 ml of the actual ethanol solution sample for production, reserve 2 ml of each ethanol sample for spectral acquisition, and use the remaining ethanol as a solvent to prepare ethanol-coolant solutions of different concentrations. Prepare two sets of solutions for each concentration.

5. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 4, characterized in that, In step (2), the volume fraction of refrigerant impurities in the ethanol-refrigerant solution is one ten-thousandth to one hundredth.

6. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 1, characterized in that, In step (2), the near-infrared spectrum of the sample was acquired using a Matrix-F Bruker near-infrared spectrometer with a wavenumber range of 12000 cm⁻¹. -1 ~4000 cm -1 The resolution is 4 cm. -1 The number of scans was 64, and each sample was measured 3 times and the average was taken as the final spectrum.

7. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 1, characterized in that, In step (3), the spectra x of the ethanol sample and the spectra x of the impurity sample are analyzed respectively. Perform PCA analysis to eliminate abnormal spectra; The normal spectrum was preprocessed using SNV to eliminate baseline drift and the effects of scattering on the spectrum.

8. The method for detecting trace impurities in ethanol solution based on near-infrared spectroscopy according to claim 1, characterized in that, In practical work, to avoid the occurrence of random errors, RSD can be used. max Adjust the judgment criteria appropriately.

Images