A method for detecting glyphosate products and their by-products based on near infrared spectroscopy

Abstract

The present invention provides a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy, comprising the following steps: pre-treating the glyphosate products and their byproducts to be detected, collecting near infrared spectra, and obtaining an original near infrared spectrum of the glyphosate products; inputting the original near infrared spectrum of the glyphosate products and their byproducts into an established quantitative model, and obtaining quantitative results of the glyphosate products and their byproducts. The present invention does not require the use of chemicals and is a high-precision, efficient, pollution-free, millisecond-level in-situ analysis technology, which greatly reduces the cost of the agent and provides a product production cycle.

Description

Technical Field

[0001] The invention relates to the technical field of drug analysis, and in particular to a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy. Background Art

[0002] Glyphosate, also known as N-(phosphonomethyl)glycine (abbreviated as PMG), is a low-toxic systemic broad-spectrum herbicide developed by Monsanto. With an output of 800,000 tons in 2021, it is the largest pesticide variety in the world in terms of production and sales. Currently, more than 50 crops are registered for use in my country, involving edible agricultural products such as fruits, corn and rice, as well as non-pesticide products such as rubber and mulberry. In recent years, due to the continued intensification of global extreme climate and the outbreak of the new coronavirus, the uncertainty and instability of the global food supply system have led to increasing attention from countries around the world to ensure food self-sufficiency. The area of ​​agricultural product cultivation will rebound, and the demand for agricultural products such as fertilizers and pesticides will also grow simultaneously. According to historical data, more than 80% of the glyphosate produced in China is exported, accounting for more than 50% of the total export of pesticides. The annual export volume is expected to be 400,000 tons. The main export countries are Brazil, Argentina, the United States and Australia.

[0003] Compared with domestic requirements, the requirements for glyphosate characteristic indicators and byproduct indicators are more stringent for export products. In particular, in 2020, the EU readjusted my country's glyphosate residue limit standards and product export standards. The adjustment of its pesticide management policy has increasingly higher requirements for future analysis, which will also have an impact on agricultural product exports. At present, there are two main glyphosate analysis methods commonly used internationally and domestically: high-performance liquid chromatography (arbitration method) and spectrophotometry (GB-T 12686-2017). The former requires sample pretreatment and has high requirements for instrumentation and equipment, large consumables, and is sensitive to temperature. The latter requires the use of toxic substances such as potassium bromide and sodium nitrite, and the process is complicated, the analysis cycle is long, and it takes two hours. In addition, the uniformity of the sample is required. Both methods will produce a large amount of acid, alkali and organic waste liquid, which needs to be outsourced for treatment, which also increases the production cost of glyphosate. At the same time, glyphosate is a low-toxic herbicide that has a certain irritating effect on human skin and mucous membranes, and long-term contact causes allergic reactions.

[0004] Near infrared spectroscopy is a detection technology that uses the material information contained in the near infrared spectrum to qualitatively and quantitatively analyze organic matter. It mainly reflects the fundamental frequency of 2000cm -1The above group information mainly includes hydrogen-containing groups (CH, OH, SH, NH), as well as other group information (C=C, C=O). This rich information determines that it can determine chemical composition and analyze physical properties, and can provide quality inspection and analysis solutions for the fields of petroleum, chemical, food, pharmaceutical and pesticide. In recent years, many scholars at home and abroad have successfully applied near-infrared spectroscopy technology to the identification of components in the fields of petroleum and agricultural products, but there has been no research on the application of near-infrared in the field of pesticides.

[0005] As the mainstream production route of glyphosate in my country, the alkyl ester method provides about 80% of the national demand, especially the glycine method. However, a series of side reactions are inevitably produced during the synthesis process due to formaldehyde and hydrolysis. Carbon spectrum, phosphorus spectrum and hydrogen spectrum detection mainly show glyphosate, N-methylglyphosate, etc., resulting in a complex composition of glyphosate products.

[0006] Therefore, developing an online in-situ analysis and detection method for glyphosate to achieve rapid and efficient analysis of product indicators has practical significance for the glyphosate and even pesticide industries. Summary of the invention

[0007] In view of this, the technical problem to be solved by the present invention is to provide a method for detecting glyphosate products and their by-products based on near-infrared spectroscopy. The detection method of the present invention can detect glyphosate products and their by-products quickly with good reproducibility and accuracy.

[0008] The present invention provides a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy, comprising the following steps:

[0009] A) pre-treating the glyphosate product to be tested and its by-products, and collecting near infrared spectra to obtain an original near infrared spectrum of the glyphosate product;

[0010] B) inputting the original near infrared spectra of the glyphosate product and its by-products into the established quantitative model to obtain the quantitative results of the glyphosate product and its by-products;

[0011] The quantitative model is obtained by the following method:

[0012] a) Determination of the content of glyphosate products and their by-products by high performance liquid chromatography;

[0013] b) performing near infrared spectroscopy scanning on the glyphosate product and its by-products to obtain original near infrared spectra of the glyphosate product and its by-products;

[0014] c) preprocessing the original near-infrared spectra of glyphosate products and their byproducts, and establishing a preliminary model by combining PLS fitting to build a spectral data matrix;

[0015] d) Evaluate and reoptimize the model based on cross-validation and / or external validation to obtain a quantitative model.

[0016] The invention provides a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy.

[0017] The glyphosate by-products of the present invention include glyphosate and N-methylglyphosate.

[0018] That is, the detection method of the present invention can perform qualitative and quantitative analysis on the contents of glyphosate, glyphosate and N-methylglyphosate in products containing the above substances, with good reproducibility and accuracy.

[0019] The first step is to establish a quantitative model.

[0020] The quantitative model of the present invention is first established by using high performance liquid chromatography to determine the content of glyphosate products and their by-products.

[0021] The present invention does not limit the determination method of the high performance liquid chromatography, which is well known to those skilled in the art. Preferably, the method of GB / T 12686-2017 can be used. Specifically, the components of glyphosate raw powder are determined by arbitration method-high performance liquid chromatography method (GB / T12686-2017), and the main analysis indicators include glyphosate, glyphosate and N-methylglyphosate. Each sample is measured twice as the final concentration value of the sample, and the difference between the parallel results is less than 0.5%, and quantification is performed by external standard method.

[0022] The sample can be dissolved in a mobile phase, and potassium dihydrogen phosphate and methanol aqueous solution with a pH of 1.9 are used as mobile phases, and a strong ion exchange column (SAX) and an ultraviolet detector (195nm) are used to measure the components in the sample. Measurement conditions: mobile phase potassium dihydrogen phosphate (mass concentration 0.027%) + methanol aqueous solution (methanol: water = 0.03), phosphoric acid is used to adjust the pH to 1.9, ultrasonic vibration for 10 to 20 minutes; flow rate: 1.5ml / min; column temperature: 30℃, injection volume 20μl. Each sample is measured twice as the final concentration value of the sample, and the difference between the parallel results is less than 0.5%, and quantification is performed by the external standard method.

[0023] The present invention preferably selects samples that can represent the glyphosate system. To ensure the robustness of the model and expand the content range, the amount of spectral scanning sample data is 150 to 600; more preferably 200 to 500.

[0024] The inventors found that within the above sample range, the established model has good accuracy.

[0025] If the sample groups are not controlled, the boundary content cannot be identified and the finished products in the middle concentration range will be overfitted, which will cause the model to fail.

[0026] The glyphosate product and its by-products are scanned by near infrared spectroscopy to obtain original near infrared spectra of the glyphosate product and its by-products.

[0027] According to the present invention, the near infrared spectrum acquisition parameters are specifically:

[0028] A near-infrared spectrometer was used to collect diffuse reflectance data of glyphosate samples. The main wavelength range was concentrated in the long-wave band of 1100-2526nm, with an average resolution of 2nm.

[0029] The specific scanning accessories for spectral data collection are sample rotating table + integrating sphere accessories, the sample and background scanning frequency is 64 times, as the final sample information, the number of single sample spectra is 2, the scanning speed is 10KHz; the sample temperature is 30℃.

[0030] The comprehensive effect of selecting the above-mentioned spectrum data acquisition parameters in the present invention makes the model data results finally established accurate and reliable.

[0031] Preprocessing the original near infrared spectra of glyphosate products and their byproducts. The preprocessing of the present invention includes spectrum preprocessing and model preprocessing;

[0032] According to the present invention, the spectral pretreatment indexes include: glyphosate 9090-6060cm -1 、5464-4608cm -1 ; Contains 445 wavelength points. Glyphosate is 9130-8090cm -1 、7504-5448cm -1 4600-4248cm -1 Contains 412 wavelength points; N-methylglyphosate is 8952-8100cm -1 、6000-4100cm -1 , contains 445 wavelength points, and the actual spectral data of each indicator is a matrix set of wavelength points.

[0033] The above-mentioned full scan of the present invention limits the frequency range of PLS ​​regression in order to reduce the influence of spectral noise and additional components in the sample, which may cause over-analysis and enhance the robustness of the model.

[0034] The model preprocessing of the present invention includes one or more of subtracting a straight line, minimum-maximum normalization, vector normalization (SNV), first-order derivative, second-order derivative, first-order derivative + vector normalization; first-order derivative + MSC;

[0035] The pretreatment index of the model of the present invention preferably includes: glyphosate is treated by first-order derivative + MSC; glyphosate is treated by first-order derivative + vector normalization; N-methylglyphosate is treated by first-order derivative method.

[0036] The detection method of the present invention adopts the above-mentioned specific preprocessing index to obtain the maximum orientation coefficient R 2 / R P 2 with the minimum root mean square error value.

[0037] After pretreatment, the spectral data matrix is ​​established in combination with PLS fitting to establish a preliminary model. In the PLS fitting process of the present invention, the matrix dimension is set to ≤10, which is specifically determined according to the following evaluation parameters and secondary optimization results.

[0038] On the one hand, a model with too high a dimension will affect the recognition of eigenvalues ​​and the performance of the model. On the other hand, it will complicate the calculation and storage. Therefore, it is necessary to reduce the dimension as much as possible during the modeling process.

[0039] The model is evaluated and re-optimized based on cross-validation and / or external validation to obtain a quantitative model.

[0040] The cross-check and / or external check of the present invention includes:

[0041] Glyphosate calibration uses external verification to obtain the best RMSEP and R 2 and RPD value; correction value: measurement value = 5:2, correction interval is 10, data length is 4;

[0042] Glyphosate verification uses cross-checking to obtain the best RMSECV, R 2 and RPD value;

[0043] N-methylglyphosate calibration uses external verification to obtain the best RMSEP, R 2 And RPD value; correction value: measurement value = 4:1, correction interval is 8, data length is 2.

[0044] Different chemical values ​​adopt specific evaluation methods and parameters to obtain the most robust effect. The selection of the above-mentioned test method of the present invention can obtain the best RMSEP, R 2 and RPD value.

[0045] The secondary optimization of the present invention includes:

[0046] The model was optimized secondary by removing outliers, selecting dimensions, adjusting Mahalanobis distance, and adjusting the relationship between leverage value and spectral residual.

[0047] The above optimization methods of the present invention are conventional optimization methods well known to those skilled in the art, and the present inventors can adopt one or more of them for optimization according to specific needs.

[0048] The model was optimized twice by removing outliers, selecting dimensions, adjusting Mahalanobis distance, and adjusting the leverage value and spectral residual relationship to ensure that the fitting value, true value, deviation and residual reached the optimal level. Finally, the optimal dimensions of the quantitative model were: glyphosate: 10, glyphosate and N-methylglyphosate were 7 and 8 respectively.

[0049] The present invention provides a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy. Firstly, the glyphosate product and its byproducts to be detected are pretreated, and then near infrared spectroscopy is collected to obtain an original near infrared spectrum of the glyphosate product.

[0050] The pretreatment here refers to sample pretreatment. The sample of the present invention needs to be pretreated by mobile phase dilution, ultrasonic vibration and the like.

[0051] In one embodiment of the present invention, the dilution can preferably be: 0.2 g to 100 mL, the mobile phase is diluted to the scale;

[0052] The preferred method for ultrasonic oscillation is: a normal temperature water bath for 10 minutes.

[0053] After preprocessing, the parameters for near infrared spectrum acquisition have been clearly described above. The specific parameters for near infrared spectrum acquisition are:

[0054] A near-infrared spectrometer was used to collect diffuse reflectance data of glyphosate samples. The main wavelength range was concentrated in the long-wave band of 1100-2526nm, with an average resolution of 2nm.

[0055] The specific scanning accessories for spectral data collection are sample rotating table + integrating sphere accessories, the sample and background scanning frequency is 64 times, as the final sample information, the number of single sample spectra is 2, the scanning speed is 10KHz; the sample temperature is 30°C. The above parameters can ensure the stability and uniformity of sample scanning and avoid the influence of instrument and ambient temperature.

[0056] The original near-infrared spectra of glyphosate products and their by-products are input into the established quantitative model to obtain the quantitative results of glyphosate products and their by-products.

[0057] The detection method of the present invention also includes using the quantitative results of the model detection in combination with whiteness or particle size indicators to classify the glyphosate product.

[0058] The whiteness evaluation of the present invention is preferably as follows: the whiteness of the sample is obtained by comparing the reflectance of the sample to blue light with the reflectance of the standard white board to blue light. The sample is loaded according to the sample box provided by the whiteness meter, and two measurements are performed simultaneously. The difference between the parallel measurement results is not greater than 0.2%, and the average value is taken as the measurement result;

[0059] The particle size is preferably evaluated by measuring the particle size using a Malvern 3000 particle size analyzer.

[0060] The chemical test indicators of glyphosate samples include glyphosate, glyphosate-enhanced and N-methylglyphosate. The threshold value is set for the standard product. The grade of the glyphosate product is determined by comparing the sample with the standard threshold value (see Table 10).

[0061] According to the main components and impurity content of glyphosate samples, including physical properties such as particle size and whiteness, glyphosate product grades A, B and C are determined (see Table 10). Grade A qualified products can be sold (exported) abroad, grade B qualified products can be sold domestically and abroad; grade C products are unqualified.

[0062] Through the above innovations, the present invention overcomes the problems of long analysis cycle, low precision and environmental pollution in the existing analysis technology, and realizes rapid and accurate evaluation of the quality of glyphosate products.

[0063] Compared with the traditional method, the present invention has the following advantages: 1. During the test, the finished glyphosate product is not damaged, and the test sample is retained to the greatest extent; 2. The analysis speed is fast and can be completed quickly within a dozen seconds; 3. During the test, no complicated pre-treatment is required, and no chemical reagents are used, so it is green and there is no professional requirement for analysts. 4. The accuracy and linearity are good. Whether it is online or offline analysis, the sample volume is large and each sample is repeatedly scanned 30 times or more, and human errors are eliminated; 5. It can realize the simultaneous detection of multiple chemical components, save material and manpower, and greatly improve the analysis efficiency. In short, the analysis method developed by the present invention can realize the rapid detection of glyphosate finished products, make timely adjustments according to the category classification regression process, ensure that the total qualified rate of the product is 100%, and greatly improve the utilization rate of elements. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is the near infrared spectrum scan of the sample;

[0065] Figure 2 is a normal distribution diagram of sample indicators; Figure 2(a) is a normal distribution test diagram of glyphosate model data, Figure 2(b) is a normal distribution test diagram of glyphosate model data, and Figure 2(c) is a normal distribution test diagram of N-methylglyphosate model data;

[0066] Figure 3 The results of model parameter testing for different spectral preprocessing methods are shown in the figure;

[0067] Figure 4 This is a diagram of the detection path method of glyphosate products and their byproducts using near infrared spectroscopy of the present invention. DETAILED DESCRIPTION

[0068] The present invention provides a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy. Those skilled in the art can refer to the content of this article and appropriately improve the process parameters to achieve it. It should be particularly noted that all similar replacements and modifications are obvious to those skilled in the art, and they all belong to the scope of protection of the present invention. The method and application of the present invention have been described through preferred embodiments, and relevant personnel can obviously modify or appropriately change and combine the method and application of this article without departing from the content, spirit and scope of the present invention to realize and apply the technology of the present invention.

[0069] In this application, the term "and / or" describes the association relationship of associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural.

[0070] In this application, "at least one" means one or more, "more than one" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items.

[0071] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution, some or all of the steps can be executed in parallel or sequentially, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0072] The numerical ranges and parameters involved in the present invention have presented the relevant numerical values ​​in the specific embodiments as accurately as possible. However, any numerical value essentially inevitably contains standard deviations due to individual testing methods. Therefore, unless otherwise explicitly stated, it should be understood that all numerical ranges or specific data used in this disclosure may have certain reasonable deviations within a certain range, for example: within 1% or 0.5%.

[0073] Some cases are recorded in the embodiments and comparative examples of the present invention, wherein the embodiments show certain implementation modes of the present invention. However, this does not mean that the effects of the present invention can only be achieved in these cases.

[0074] To further illustrate the present invention, a method for detecting glyphosate products and their byproducts based on near infrared spectroscopy provided by the present invention is described in detail below in conjunction with examples.

[0075] Control group: The components of glyphosate raw powder were determined by arbitration method-high performance liquid chromatography (GB / T 12686-2017), and the main analytical indicators included glyphosate, glyphosate and N-methylglyphosate. Each sample was measured twice as the final concentration value of the sample, and the difference between the parallel results was less than 0.5%, and quantification was performed by external standard method.

[0076] Table 1: Glyphosate product verification sample data (Arbitration Law)

[0077]

[0078] Comparative Example 1: (Experimental Group 1)

[0079] Ultraviolet spectrophotometry: weigh 0.1g of glyphosate product (accurate to 0.0001g). Dissolve the sample in 100mL, dilute with water to the scale, ultrasonically oscillate for 10min to dissolve the sample, cool to room temperature, and shake well. Use a pipette to transfer 2mL of the above solution into a 100mL volumetric flask, add 5mL of water, 0.5mL of sulfuric acid solution, 0.1mL of potassium bromide solution, and 0.5mL of sodium nitrite solution in sequence, plug the stopper tightly, shake well, leave it for 20min, dilute with water to the scale, and shake well. Open the stopper and leave it for 15min. Measure the absorbance at 242nm and measure it by the external standard method.

[0080] Table 2: Glyphosate product verification sample data (UV-spectrophotometry)

[0081]

[0082] By comparing Table 2 and Table 1, it can be seen that the results of the spectrophotometric method and the arbitration method are consistent, and the relative deviation is controlled within 0.2%, which meets the analysis requirements. However, because manual operation is inevitably required during the implementation of the method, the reproducibility of the method and the precision of the results are lower than the analysis level of the arbitration method.

[0083] Embodiment 2:

[0084] Near infrared in-situ spectroscopy analysis method: The following examples are strictly implemented according to the invention process, and the parameters are selected differently in different examples.

[0085] Example

[0086] A method for detecting glyphosate products and their byproducts based on near infrared spectroscopy, comprising: Figure 1 The following steps are displayed:

[0087] (1) Sample selection and concentration data matrix establishment

[0088] Collect samples and conduct chemical tests and analyses on sample components to ensure the reliability and accuracy of measurement results.

[0089] Sample source and storage: finished glyphosate products from a glyphosate manufacturer in Jiangsu. The collection period is 2 months. Samples that can represent the glyphosate system are selected. To ensure the robustness of the model and expand the content range, the spectral scanning sample data volume is 200-500. The finished glyphosate products sampled in the workshop are dried and stored in a vacuum drying oven.

[0090] In this step, high performance liquid chromatography is mainly used for analysis, and the main analytical indicators include glyphosate, glyphosate and N-methylglyphosate. The sample is dissolved in a mobile phase, and potassium dihydrogen phosphate and methanol aqueous solution with a pH of 1.9 are used as the mobile phase, and a strong ion exchange column (SAX) and an ultraviolet detector (195nm) are used to measure each component in the sample.

[0091] Measurement conditions: mobile phase: aqueous solution of potassium dihydrogen phosphate + methanol, adjusted to pH 1.9 with phosphoric acid, ultrasonic vibration for 10-20 minutes; flow rate: 1.5 ml / min; column temperature: 30°C, injection volume 20 ul.

[0092] Each sample was measured twice as the final concentration value of the sample, and the difference between the parallel results was less than 0.5%, and quantification was performed using the external standard method.

[0093] The selected modeling samples must conform to the normal distribution (Figure 2). The statistical analysis content is shown in Table 3:

[0094] Table 3 Chemical statistical analysis results of glyphosate product modeling samples

[0095] parameter Glyphosate Glyphosate N-methylglyphosate Maximum value (%) 97.26 2.344 0.65 Minimum value (%) 94 0.878 0.014 Mean(%) 95.743 1.5743 0.272 Standard Deviation 0.706 0.275 0.001 Median (%) 95.75 1.578 0.275

[0096] (2) Whiteness and particle size evaluation

[0097] The whiteness of the sample is obtained by comparing the reflectance of the sample to blue light with that of the standard white board. The sample is loaded into the sample box provided by the whiteness meter, and two consecutive measurements are performed at the same time. The difference between the parallel measurement results is no more than 0.2%, and the average value is taken as the measurement result;

[0098] The particle size was determined using a Malvern 3000 particle size analyzer;

[0099] (3) Near-infrared spectrum acquisition and spectrum preprocessing

[0100] A Bruker near-infrared spectrometer (MPA II) was used to collect diffuse reflectance data of glyphosate samples. The main wavelength range was concentrated in the long-wave band of 1100-2526 nm, with an average resolution of 2 nm.

[0101] In the above steps, the specific scanning accessories for spectral data collection are sample rotating table + integrating sphere accessories, the sample and background scanning frequency is 64 times, as the final sample information, the number of single sample spectra is 2, the scanning speed is 10KHz; the sample temperature is 30℃;

[0102] In order to ensure the quality of modeling and a good correlation between spectral data and content, the spectral data of glyphosate samples need to be preprocessed by smoothing, adjusting baseline drift, and flattening the background. The specific processing methods include: subtracting a straight line, minimum-maximum normalization, vector normalization (SNV), first-order derivative, second-order derivative, first-order derivative + vector normalization; one or more of the first-order derivative + MSC; the model parameters corresponding to different preprocessing methods are shown in Table 4.

[0103] Table 4 Glyphosate chemical index model parameters

[0104]

[0105]

[0106] Different spectra preprocessing methods for glyphosate indicators are different. The first-order derivative + MSC is used for glyphosate indicator; the first-order derivative + vector normalization is used for glyphosate; and the vector normalization method is used for N-methylglyphosate. The maximum orientation coefficient R can be obtained respectively. 2 / R P 2 with the minimum root mean square error value.

[0107] (4) Model establishment and verification

[0108] In order to reduce spectral overfitting, the frequency range of PLS ​​regression was limited. The optimal cross-selection ranges of near-infrared spectra corresponding to different chemical values ​​were: 9090-6060 cm for glyphosate -1 、5464-4608cm -1 ; Contains 445 wavelength points. Glyphosate is 9130-8090cm -1 、7504-5448cm -1 4600-4248cm -1 Contains 412 wavelength points; N-methylglyphosate is 8952-8100cm -1 、6000-4100cm -1 , contains 445 wavelength points, and the actual spectral data of each indicator is a matrix set of wavelength points.

[0109] The evaluation method of the present invention includes two methods: cross-check and external check. The model evaluation index includes the directional coefficient R 2 / R P 2, root mean square error RMSECV / RMSEP, relative analysis deviation RPD.

[0110] The evaluation methods and parameters of different indicators are as follows: Glyphosate: The external inspection method obtains the best RMSEP and R 2 and RPD value; correction value: measured value = 5:2, correction interval is 10, data length is 4; glyphosate calibration adopts cross-check method to obtain the best RMSECV, R 2 and RPD values; N-methylglyphosate calibration uses external verification to obtain the best RMSEP, R 2 and RPD value; correction value: measurement value = 4:1, correction interval is 8, data length is 2; model parameters are as follows:

[0111] Table 5 Model parameters corresponding to different calibration modes

[0112]

[0113]

[0114] (5) Model determination

[0115] The above-obtained model is optimized twice by removing outliers, selecting dimensions, adjusting Mahalanobis distance, and adjusting the relationship between leverage value and spectral residual to ensure that the fitting value, true value, deviation and residual reach the optimal level.

[0116] The optimal dimensions of the models in the above steps are: glyphosate: 10, glyphosate and N-methylglyphosate: 7 and 8, respectively;

[0117] Table 6 Optimal model parameters for different chemical components

[0118]

[0119] According to the above table, the three key output indicators of the model can all meet the high-precision quantitative measurement and analysis sample level.

[0120] The above process involves multiple parameter selections. According to different parameter setting series implementation examples, the difference analysis between models is carried out, and finally the optimal parameters are determined to determine the model.

[0121] Example 3: The difference from the above example is that the modeling sample data and concentration selection process are different. According to whether the sample size is controlled and whether the concentration interval setting range conforms to the normal distribution, a series of experimental groups are set, the full spectrum is collected, the first-order derivative processing is performed, and the model is optimized by cross-checking. The final verification and benchmarking results are as follows:

[0122] Table 7 Glyphosate product verification sample data (sample selection and spectrum processing)

[0123]

[0124]

[0125] According to the table above, we can find that the sample size control and concentration range screening group benchmark results are much better than other groups, especially for unqualified products and products with high glyphosate concentrations (>97%). The uncontrolled sample group cannot identify the boundary content, and the finished product in the middle concentration range is overfitted. For example, in sample E, the actual glyphosate concentration is 97.26%, and the model prediction result is only 96.38%, which means the model is invalid. Therefore, the modeling sample interval must be controlled to set a normal distribution within the average concentration range of the finished product.

[0126] Example 4

[0127] The modeling sample data and concentration data conform to the normal distribution. The difference from the above embodiment is that the spectral segment is screened and optimized according to the model evaluation parameters as in Example 2, not the full spectral segment, the spectrum is processed by the first-order derivative, and the model is optimized by cross-checking. The final verification and benchmarking results are shown in Table 7:

[0128] It can be concluded that by selecting specific spectral intervals, the accuracy of the model is further improved, especially for glyphosate and glyphosate, the relative deviation is controlled within 1%. The main reason is that not all spectral intervals are useful for specific analytical tasks. Some intervals may contain information that is irrelevant to the target analyte, or the quality of the information may be reduced due to factors such as spectral noise and baseline drift. Therefore, according to the fact that the glyphosate finished product is rich in main functional group information, selecting the appropriate spectral interval can remove this useless or interfering information, thereby improving the prediction accuracy and stability of the model.

[0129] Example 5

[0130] The modeling sample data and concentration data conform to the normal distribution. The spectral segment is screened and optimized according to the fitting model parameters. For non-full spectral segments, the spectra are processed in different preprocessing modes. The model is optimized by cross-checking. The final verification and benchmarking results are shown in Table 8:

[0131] Table 8 Glyphosate product verification sample data (spectrum preprocessing)

[0132]

[0133] Table 8 reflects the model verification and benchmarking results of different pretreatment methods. It can be seen that the best pretreatment methods corresponding to different indicators are different. Glyphosate is treated with first-order derivative + MSC; glyphosate is treated with first-order derivative + vector normalization; N-methylglyphosate is treated with vector normalization. The best benchmarking results can be obtained. The deviation between the predicted value and the true value is small. The robustness of the model is further improved compared with other pretreatment methods, and the sample content can be better predicted, which is consistent with the conclusion based on the evaluation and treatment methods of the model parameters.

[0134] Example 6: The modeling sample data and concentration data conform to the normal distribution. The spectral segment is screened and optimized according to the fitting model parameters. The spectrum is not a full spectral segment. The spectrum is processed in the mode of Example 2. The inspection method adopts external inspection or cross-inspection. Compared with group A, group B has a narrower measurement interval. The external verification parameters are debugged multiple times to verify that the concentration data is rich in the entire range as much as possible to ensure the robustness of the sample. The benchmarking results are shown in Table 9:

[0135] Table 9: Glyphosate product verification sample data (verification method)

[0136]

[0137] According to the above table, it can be found that the choice of test method is very important. Different indicators correspond to different test methods, and the predicted sample deviations are different. Because the structure of glyphosate is relatively complex, external verification must be attempted multiple times. Try to choose a larger sample size, such as Group B, to ensure full coverage and increase model robustness.

[0138] The above 2-6 comparison examples show that near-infrared spectroscopy is a complex analytical method, and the selection of sample data and model parameters is crucial to the performance of the model. In order to establish an accurate and reliable near-infrared model, it is necessary to fully consider factors such as the representativeness of the sample, the selection of spectral preprocessing methods, and the number of modeling samples. At the same time, it is also necessary to pay attention to the influence of the near-infrared detection limit and the influence of the accuracy of the reference method in order to select appropriate model parameters and evaluation models.

[0139] The above Example 2 comprehensively considers the above influencing factors and has a better prediction effect on the model. Therefore, the model constructed in Example 2 is used to perform unknown sample measurement and performance evaluation.

[0140] (6) Sample analysis and verification

[0141] First, the unknown samples to be tested are chemically determined by high performance liquid chromatography, and each sample is measured twice. Then the samples to be tested are spectrally scanned and input into the system. The main components and impurity content of the output glyphosate samples are compared with the standard threshold, and the glyphosate product grades A, B and C are determined based on physical properties such as particle size and whiteness. Grade A qualified products can be sold (exported) and Grade B qualified products can be sold domestically and abroad; Grade C products are unqualified and need to return to the production system for further improvement. The test results are as follows:

[0142] Table 10 Standard qualified product threshold

[0143] Classification Glyphosate % Glyphosate% N-methylglyphosate % A ≥95.5 ≤1.5 ≤0.35 B ≥95 - - C ≥95 - -

[0144] According to the model verification results, first, there are no abnormalities in the samples to be tested, the chemical values ​​are output normally, and the chemical analysis deviation is basically controlled within 3%, especially the indicator glyphosate, which is controlled within 0.5%; second, the model is stable, and the reproducibility is basically controlled within 3%, especially the glyphosate indicator, which is far better than the chemical analysis method and high-performance liquid chromatography. Third, see Table 11-3: 20 samples to be tested, chemical detection indicators, 2 samples are Class C, need to be returned to the process for further processing, 8 indicators are Class B, can be sold domestically, if they need to be exported abroad, they must be further washed and refined to further reduce by-products, and 10 samples meet Class A and export standards.

[0145] Table 11-1 Comparison between the true value of the sample to be tested and the model predicted value

[0146]

[0147]

[0148] Table 11-2 Results of the model verification of the samples to be tested

[0149]

[0150]

[0151] Table 11-3 Model prediction results and grade classification of samples to be tested

[0152]

[0153] Therefore, the process of the present invention can quickly realize the concentration analysis and classification of glyphosate and its byproducts. Compared with high-performance liquid chromatography and spectrophotometry, it has high accuracy and good reproducibility, plays a positive leading role in the glyphosate and even pesticide industries, and broadens the application fields of glyphosate analysis methods and near-infrared. The present invention has been industrially applied to help quickly identify the concentration content of glyphosate products leaving the factory.

[0154] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting glyphosate products and their byproducts based on near infrared spectroscopy, characterized in that: The steps include: A) pre-treating the glyphosate product and its by-products to be tested, and then collecting near infrared spectra to obtain an original near infrared spectrum of the glyphosate product; B) inputting the original near infrared spectra of glyphosate products and their by-products into the established quantitative model to obtain the quantitative results of glyphosate products and their by-products; The quantitative model is obtained by the following method: a) Determination of the content of glyphosate products and their by-products by high performance liquid chromatography; b) performing near infrared spectroscopy scanning on the glyphosate product and its by-products to obtain original near infrared spectra of the glyphosate product and its by-products; c) Preprocessing the original near-infrared spectra of glyphosate products and their byproducts, and establishing a preliminary model by combining PLS fitting to build a spectral data matrix; d) Evaluate and reoptimize the preliminary model based on cross-validation and / or external validation to obtain a quantitative model; The cross-check and / or external check includes: Glyphosate calibration uses external verification to obtain the best RMSEP and R 2 and RPD value; correction value: measured value = 5:2, correction interval is 10, data length is 4; glyphosate verification uses cross-checking to obtain the best RMSECV, R 2 and RPD values; N-methylglyphosate calibration uses external verification to obtain the best RMSEP, R 2 and RPD value; correction value: measurement value = 4:1, correction interval is 8, data length is 2.

2. The detection method according to claim 1, characterized in that: Step A) the glyphosate byproducts include glyphosate and N-methylglyphosate; The near infrared spectrum acquisition parameters are specifically: A near-infrared spectrometer was used to collect diffuse reflectance data of glyphosate samples. The main wavelength range was concentrated in the long-wave band of 1100~2526nm, with an average resolution of 2nm. The specific scanning accessories for spectral data collection are sample rotating table + integrating sphere accessories, the sample and background scanning frequency is 64 times, as the final sample information, the number of single sample spectra is 2, the scanning speed is 10KHz; the sample temperature is 30℃.

3. The detection method according to claim 1, characterized in that: The preprocessing in step c) includes spectrum preprocessing and model preprocessing; The model preprocessing includes one or more of subtracting a straight line, minimum-maximum normalization, vector normalization (SNV), first-order derivative, second-order derivative, first-order derivative+vector normalization; and first-order derivative+MSC.

4. The detection method according to claim 3, wherein the spectrum preprocessing index comprises: Glyphosate is 9090-6060 cm -1 、5464-4608 cm -1 ; Contains 445 wavelength points; Glyphosate is 9130-8090cm -1 、7504-5448cm -1 4600-4248cm -1 Contains 412 wavelength points; N-methylglyphosate is 8952-8100 cm -1 、6000-4100 cm -1 , contains 445 wavelength points, and the actual spectral data of each indicator is a matrix set of wavelength points.

5. The detection method according to claim 3, characterized in that: The model preprocessing indicators include: Glyphosate was treated with first-order derivative + MSC; glyphosate was treated with first-order derivative + vector normalization; N-methylglyphosate was treated with first-order derivative method.

6. The detection method according to claim 1, characterized in that: Step d) the secondary optimization comprises: The model was optimized secondary by removing outliers, selecting dimensions, adjusting Mahalanobis distance, and adjusting the relationship between leverage value and spectral residual.

7. The detection method according to claim 5, characterized in that: The optimal dimensions of the quantitative model are: glyphosate: 10, glyphosate and N-methylglyphosate: 7 and 8 respectively.

8. The detection method according to claim 1, characterized in that: The amount of sample data for performing near infrared spectroscopy scanning on the glyphosate product and its byproducts in step b) is 150 to 600.

9. The detection method according to claim 1, characterized in that: Step B) also includes classifying the glyphosate product using the quantitative results of the model test combined with whiteness or particle size indicators.

Images