Method for rapidly determining stoving traditional Chinese medicinal materials

Abstract

The invention belongs to the technical field of quality detection of traditional Chinese medicinal materials, and particularly relates to a method for rapidly determining stoving traditional Chinese medicinal materials, which realizes rapid and nondestructive detection of the stoving traditional Chinese medicinal materials by pretreating a traditional Chinese medicinal material sample and combining a near infrared spectrum technology with a chemometrics model. According to the method, a standard sample spectrum database is established, a quantitative analysis model is constructed by applying a principal component analysis-partial least square method, single sample detection can be completed within 5 minutes, the detection limit is as low as 0.1 mg / kg, the detection efficiency and accuracy are remarkably improved, and the method is suitable for multiple traditional Chinese medicinal material stoving detection scenes.

Description

Technical Field

[0001] This invention belongs to the field of quality testing technology for Chinese medicinal materials, specifically a method for rapidly determining the quality of sulfur-fumigated Chinese medicinal materials. Background Technology

[0002] As an important component of traditional Chinese medicine, the quality of medicinal herbs directly affects clinical efficacy and medication safety. Sulfur fumigation is a traditional processing method for medicinal herbs. Sulfur dioxide has bleaching, preservative, and insect-repellent properties, making the herbs appear brighter and extending their shelf life. However, excessive sulfur fumigation can lead to excessive sulfur dioxide residues in medicinal herbs, posing a serious threat to human health. Once ingested, sulfur dioxide irritates the respiratory tract, causing symptoms such as coughing and wheezing. Long-term intake may also damage organs such as the liver and kidneys, and even carries a potential carcinogenic risk.

[0003] Currently, the main methods for determining sulfur dioxide residues in traditional Chinese medicinal materials include pharmacopoeia methods (such as acid-base titration and ion chromatography) and instrumental analytical methods (such as high-performance liquid chromatography and gas chromatography). The acid-base titration method in the pharmacopoeia is cumbersome, requiring multiple steps such as sample pretreatment, distillation, and titration. The detection cycle is long, generally requiring 4-6 hours to complete the detection of one sample, and the process uses a large amount of chemical reagents, causing environmental pollution. While ion chromatography offers high accuracy, the equipment is expensive, maintenance costs are high, and the technical requirements for operators are high, making it difficult to promote its use in grassroots testing units. High-performance liquid chromatography and gas chromatography also suffer from problems such as complex instruments, high detection costs, and long sample pretreatment times, failing to meet the needs for rapid, high-volume detection.

[0004] With the rapid development of the Chinese medicinal materials market, higher demands are being placed on the efficiency and accuracy of quality testing for these materials. Traditional testing methods, due to their inherent limitations, struggle to achieve rapid and accurate detection of sulfur dioxide residues in Chinese medicinal materials. Therefore, there is an urgent need to develop a rapid, non-destructive, and accurate testing method to meet the needs of quality supervision and market testing of Chinese medicinal materials. Near-infrared spectroscopy, with its advantages of fast analysis speed, simple operation, non-destructive testing, and the ability to simultaneously determine multiple components, has been widely used in quality testing of agricultural products and food. However, in the detection of sulfur fumigation in Chinese medicinal materials, a mature and comprehensive testing system has not yet been established, and issues remain regarding the need to improve model accuracy and stability. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid method for determining sulfur-fumigated Chinese medicinal materials, solving the problems of long detection cycles, complex operations, high costs, and difficulty in meeting the needs of rapid and large-scale detection in existing detection methods.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for rapidly determining sulfur-fumigated Chinese medicinal materials, comprising the following steps: S1. Sample pretreatment: Remove impurities from the Chinese medicinal material sample and pulverize it to 20-40 mesh, then pass it through a 20-mesh standard sieve; dry the pulverized sample at 40℃ to constant weight, and then place it in a desiccator to cool for 30 minutes to eliminate moisture interference and ensure that the sample is in a consistent state. S2. Spectral acquisition: Near-infrared spectrometer is used to acquire spectral data of the sample in the wavelength range of 800-2500nm using an integrating sphere diffuse reflectance method. The number of scans is 32-64 times, and the resolution is 4-8cm⁻¹. S3. Establish a standard sample spectral database: Select at least 10 common Chinese medicinal materials, and prepare standard samples with at least 20 concentration gradients for each Chinese medicinal material with sulfur content ranging from 0 to 1000 mg / kg; S4. Establish a quantitative analysis model: Analyze the standard sample spectral database using principal component analysis-partial least squares method, and determine the optimal number of principal components using leave-one-out cross-validation method, so that the root mean square error of the model prediction is less than 5% and the correlation coefficient is greater than 0.95. S5. Sample Detection: The samples of Chinese medicinal materials to be tested are preprocessed and spectral data are collected. The spectral data are baseline corrected and smoothed using the Savitzky-Golay convolution smoothing method with a window width of 9-15 and a polynomial order of 2-4 to eliminate baseline drift and random noise in the spectral data.

[0007] Specifically, in the sample pretreatment step, for samples with high sugar content or prone to agglomeration, an appropriate amount of anhydrous ethanol is added to assist in pulverization, and the ethanol is evaporated by ventilation after pulverization; for hard samples, low-temperature freeze pulverization is used to ensure uniform particle size passing through a 20-mesh sieve.

[0008] Specifically, in the spectral acquisition step, the ambient temperature and humidity are monitored in real time. The ambient temperature is controlled at 20-25℃ and the relative humidity is controlled at 40%-60%. Temperature and humidity correction factors are introduced into the model to eliminate the influence of environmental factors on spectral data.

[0009] Specifically, in the step of establishing a standard sample spectral database, the sulfur content of each standard sample is determined by ion chromatography, and its spectral data is collected simultaneously to establish a standard sample spectral database containing spectral data and corresponding sulfur content.

[0010] Specifically, in the sample testing step, the preprocessed spectral data is input into the quantitative analysis model to obtain the sulfur content of the sample; and a quality control system for the test results is established, and the model is periodically validated using standard substances. When the relative error of the test results exceeds 10%, the model is updated and optimized.

[0011] Specifically, after the near-infrared spectrometer acquires spectral data, it uses compressed sensing technology to compress the spectral data, reducing the data volume to 1 / 10 to 1 / 5 of the original data. This reduces the data storage and transmission pressure and improves data processing efficiency without losing spectral information.

[0012] Specifically, it also includes establishing a qualitative discrimination model for sulfur-fumigated Chinese medicinal materials. Using the support vector machine (SVM) algorithm and standard sample spectral data as the training set, a discrimination model based on spectral features is established to determine whether or not the Chinese medicinal materials have been fumigated with sulfur. This enables rapid qualitative judgment of whether the Chinese medicinal materials have been fumigated with sulfur, further improving the detection method system so that it can not only quantitatively determine the sulfur content, but also quickly make qualitative judgments, thus expanding the application scenarios of the method.

[0013] The beneficial effects of this invention are: Fast testing speed: The entire testing process can be completed within 5 minutes for a single sample. Compared with traditional testing methods, the testing efficiency is increased by dozens of times, which can meet the needs of rapid testing in the Chinese medicinal materials market, effectively shorten the testing cycle and increase the testing throughput.

[0014] Non-destructive testing: Using near-infrared spectroscopy, there is no need to perform destructive processing on the samples, and the samples will not be damaged. The tested samples can still be used normally, which saves sample resources and is suitable for the quality testing of precious Chinese medicinal materials.

[0015] High accuracy: By establishing a large database of standard sample spectra and optimizing the model using chemometric methods, the detection limit is as low as 0.1 mg / kg. The root mean square error of prediction (RMSEP) of the model is less than 5%, and the correlation coefficient (R²) is greater than 0.95. It can accurately determine the sulfur content in Chinese medicinal materials, and the detection results are reliable.

[0016] Low cost: Near-infrared spectrometers are easy to operate, have relatively low technical requirements for operators, and do not require the use of a large number of chemical reagents during the detection process, which reduces the detection cost. At the same time, the equipment maintenance cost is also low, making it easy to promote and apply in grassroots testing units.

[0017] Wide range of applications: This method is applicable to the sulfur fumigation detection of various Chinese medicinal materials. By establishing a standard sample spectral database covering a variety of Chinese medicinal materials, it can accurately reflect the spectral characteristics of different Chinese medicinal materials under different sulfur fumigation degrees. It can be widely used in Chinese medicinal material production enterprises, circulation links and quality supervision departments. Attached Figure Description

[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Figure 1 The flowchart of a method for rapidly determining sulfur-fumigated Chinese medicinal materials provided by the present invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] like Figure 1 As shown, the method for rapid determination of sulfur-fumigated Chinese medicinal materials according to the present invention includes the following steps: Sample pretreatment: Select medicinal herb samples, remove surface impurities and non-medicinal parts, and pulverize the samples to 20-40 mesh to ensure good uniformity and representativeness. Pass the pulverized samples through a 20-mesh standard sieve to remove particles that do not meet the particle size requirements. Then, place the samples in a 40℃ drying oven to constant weight to eliminate the interference of moisture on spectral data. For heat-sensitive medicinal herbs, vacuum drying (vacuum degree -0.08MPa) or freeze-drying can be used, with the drying temperature reduced to 30-35℃ to ensure component stability. After drying, place the samples in a desiccator to cool for 30 minutes to allow the sample temperature to return to room temperature, facilitating subsequent spectral acquisition. Adjust the pulverization method according to the texture of the medicinal herbs: For samples with high sugar content or prone to clumping (such as wolfberry and ophiopogon japonicus), an appropriate amount of anhydrous ethanol can be added to assist pulverization; after pulverization, the ethanol should be evaporated by ventilation. For hard samples (such as fritillaria cirrhosa and poria cocos), low-temperature freeze pulverization can be used to ensure uniform particle size passing through a 20-mesh sieve.

[0022] Spectral Acquisition: A near-infrared spectrometer was used to acquire the spectra of the pretreated samples. The spectrometer was set to scan 32-64 times with a resolution of 4-8 cm⁻¹. Integrating sphere diffuse reflectance was used to acquire spectral data within the wavelength range of 800-2500 nm. Multiple scans enhance the spectral signal intensity and reduce noise interference. The integrating sphere diffuse reflectance method effectively collects diffuse reflected light from all directions of the sample, improving the accuracy and repeatability of spectral acquisition.

[0023] A standard sample spectral database was established: At least 10 common Chinese medicinal herbs were selected, and standard samples with sulfur content ranging from 0 to 1000 mg / kg were prepared for each herb, with at least 20 concentration gradients. The sulfur content of each standard sample was accurately determined using ion chromatography, and its spectral data was simultaneously acquired using a near-infrared spectrometer. The spectral data were mapped one-to-one with the corresponding sulfur content data to establish a standard sample spectral database. This database covers the spectral and sulfur content information of various Chinese medicinal herbs under different sulfur fumigation levels, providing basic data for subsequent modeling. Specifically, standard samples with sulfur contents of 0, 0.1, 0.5, 1, 5, 10, 20…1000 mg / kg were prepared, with 10 samples prepared for each concentration. Ion chromatography was used for precise quantification to ensure the accuracy of low-concentration samples.

[0024] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the standard sample spectral database. Leave-one-out cross-validation was employed to optimize the model, determining the optimal number of principal components and establishing a quantitative analysis model for sulfur content. The RMSEP curve as a function of the number of principal components was plotted. The minimum number of principal components at which the curve tended to plateau was selected (e.g., RMSEP = 3.2% with 8 principal components, RMSEP = 3.1% with 9 principal components, the difference <0.5%, therefore 8 principal components were chosen) to avoid overfitting. Model parameters were optimized to ensure the root mean square error of prediction (RMSEP) was less than 5% and the correlation coefficient (R²) was greater than 0.95, ensuring good accuracy and stability. Ambient temperature (20-25℃) and relative humidity (40%-60%) were recorded simultaneously during spectral acquisition. These temperature and humidity data were used as auxiliary variables and input into the PCA-PLS model along with the spectral data. Regression analysis was used to establish a model incorporating temperature and humidity correction factors to eliminate the interference of environmental factors on the spectra.

[0025] Sample Testing: The medicinal herb samples to be tested were processed according to the sample preprocessing and spectral acquisition steps described above to obtain spectral data. Baseline correction and smoothing were performed on the spectral data using the Savitzky-Golay convolution smoothing method with a window width of 9-15 and a polynomial order of 2-4 to eliminate baseline drift and random noise. The preprocessed spectral data was then input into the established quantitative analysis model to quickly obtain the sulfur content of the sample. Spectral Data Compression: A compressed sensing algorithm based on Discrete Cosine Transform (DCT) was used to sparsely represent the spectral data in the 800-2500nm wavelength range, retaining the main characteristic wavelengths (e.g., selecting the first 200 principal components). The compressed data was then restored using an inverse transform, verifying that the spectral reconstruction error was <2%, ensuring no information loss.

[0026] A qualitative discrimination model for sulfur-fumigated Chinese medicinal materials was established: Spectral data of samples with sulfur content of 0 mg / kg (unfumigated) and ≥50 mg / kg (fumigated) from the standard sample spectral database were selected as the training set. A Support Vector Machine (SVM) algorithm was used, with a Radial Basis Function (RBF) kernel function selected. The penalty parameter C and kernel function parameter γ were optimized through grid search to establish a binary classification discrimination model for sulfur fumigation. The accuracy of the model was verified using a test set. The model was deemed effective when the accuracy was ≥95%.

[0027] Quality control: The model is validated weekly using the sulfur dioxide residue standard material (such as GBW(E)130492) published by the National Pharmacopoeia Commission. The average value is taken after three measurements. If the relative error is >10%, the spectral data of the standard material is added to the standard sample database and the model parameters are re-optimized.

[0028] In the appendix Figure 1 The process specifically includes: 1. Sample pretreatment Operating steps: ① Select samples of Chinese medicinal materials and remove surface impurities and non-medicinal parts (such as mud, insect-damaged parts, flower stalks, etc.); ② Crush the sample to 20-40 mesh, pass it through a 20-mesh standard sieve, and remove particles that do not meet the particle size requirement; ③ Place in a 40℃ drying oven and dry until constant weight (the difference between two consecutive weighings is <0.001g) to eliminate moisture interference; ④ After drying, place the sample in a desiccator and cool it to room temperature for 30 minutes to ensure uniform temperature.

[0029] 2. Spectral Acquisition Equipment and parameters: ① Near-infrared spectrometer was used, with 32-64 scans and a resolution of 4-8 cm⁻¹; ② Spectral data in the wavelength range of 800-2500nm were collected using the diffuse reflection method of the integrating sphere; ③ Real-time monitoring of ambient temperature and humidity (temperature 20-25℃, humidity 40%-60%), with the introduction of temperature and humidity correction factors.

[0030] Establish a standard sample spectral database 3. Data preparation: ① Select at least 10 common Chinese medicinal herbs (such as Astragalus membranaceus, Codonopsis pilosula, and Lonicera japonica), and prepare standard samples for each herb with a sulfur content of 0-1000 mg / kg and ≥20 concentration gradients; ② The sulfur content of each standard sample was accurately determined by ion chromatography, and spectral data were collected simultaneously; ③ Link spectral data with sulfur content to establish a database containing spectral information, sulfur content, and types of Chinese medicinal materials.

[0031] Establish a quantitative analysis model 4. Modeling and Optimization: ① The database was analyzed using principal component analysis-partial least squares (PCA-PLS); ② Determine the optimal number of principal components and optimize the model parameters by using leave-one-out cross-validation. ③ Ensure model performance: Root mean square error of prediction (RMSEP) < 5%, correlation coefficient (R²) > 0.95.

[0032] 5. Sample testing Testing process: ① Perform pretreatment and spectral acquisition on the sample to be tested according to steps 1-2; ② Spectral data preprocessing: baseline correction, Savitzky-Golay smoothing (window width 9-15, polynomial order 2-4), noise removal; ③ Input the quantitative model and output the sulfur content of the sample (detection time < 5 minutes, detection limit 0.1 mg / kg).

[0033] 6. Optional step: Qualitative discrimination model Qualitative analysis: ① The support vector machine (SVM) algorithm is used, with standard sample spectral data as the training set; ② Establish a binary classification model for determining whether a sample has been sulfur-fumigated (accuracy ≥ 95%) to quickly determine whether the sample has been sulfur-fumigated.

[0034] 7. Quality Control System Model validation and updates: ① Verify test results using standard materials periodically (e.g., weekly), when the relative error is >10%; ②Incorporate the new data into the standard sample database, re-optimize the model parameters, and ensure long-term accuracy.

[0035] Example 1: Determination of sulfur content in Astragalus membranaceus Sample pretreatment: Ten batches of Astragalus membranaceus samples from different origins were randomly selected from the Chinese medicinal herb market. Impurities such as soil, sand, and withered leaves were carefully removed from the surface, along with insect-damaged, moldy, and severely damaged parts. The selected Astragalus membranaceus was cut into small pieces approximately 0.5-1 cm thick and placed in a high-speed grinder in batches. The grinding time was set to 3-5 minutes until the samples could pass through a 20-mesh standard sieve uniformly. During the grinding process, to prevent changes in composition due to frictional heat, the grinder was appropriately cooled after each batch of samples was ground.

[0036] The pulverized Astragalus sample was placed in a custom-made stainless steel tray, spread into a uniform thin layer no more than 2 cm thick, and then placed in a forced-air drying oven set at 40℃. The drying oven was equipped with an automatic circulating air system to ensure uniform temperature distribution. Every hour, the sample was weighed using an electronic balance with an accuracy of 0.0001 g. When the difference between two consecutive weighings did not exceed 0.001 g, the sample was considered to have been dried to constant weight. After drying, the sample was quickly transferred to a glass desiccator pre-filled with desiccant (color-changing silica gel), sealed, and left to stand for 30 minutes to allow the sample to cool fully to room temperature, avoiding the impact of excessive temperature on the accuracy of subsequent spectral acquisition.

[0037] Spectral Acquisition: A high-performance near-infrared spectrometer, model NIRS-6000, was used. This instrument is equipped with a 5cm diameter integrating sphere diffuse reflectance device, which can efficiently collect diffuse reflected light from all directions of the sample. Before spectral acquisition, the instrument was preheated for 30 minutes to ensure it reached a stable operating state. The cooled Astragalus sample was carefully poured into the integrating sphere sample cell, ensuring uniform filling and a smooth surface.

[0038] The spectrometer was set to scan 32 times, with a resolution of 4 cm⁻¹ and a wavelength range of 800-2500 nm. To reduce random errors, spectral data for each sample were collected three times, with a one-minute interval between each acquisition to ensure consistency in sample condition. During the acquisition process, ambient temperature and humidity were monitored in real time, maintaining the ambient temperature at 22±1℃ and the relative humidity at 45%±5%. After acquisition, the instrument automatically stored the spectral data in a specific format (e.g., *.spc) in the computer for subsequent analysis and processing.

[0039] Establish a standard sample spectral database: High-quality Astragalus membranaceus raw materials without sulfur fumigation were selected. Standard samples with sulfur content gradients of 0 mg / kg, 50 mg / kg, 100 mg / kg, 200 mg / kg, 300 mg / kg, 400 mg / kg, 500 mg / kg, 600 mg / kg, 800 mg / kg, and 1000 mg / kg were prepared by precisely adding sodium sulfite solution. Fifteen standard samples were prepared for each concentration level. During preparation, it was ensured that the sodium sulfite solution and the Astragalus membranaceus sample were thoroughly and uniformly mixed. After mixing, the samples were allowed to stand at room temperature for 2 hours to allow the sodium sulfite and Astragalus membranaceus to fully react.

[0040] The sulfur content of each standard sample was determined using ion chromatography. The specific procedure was as follows: Accurately weigh 1.0 g (accurate to 0.0001 g) of the standard sample and place it in a 250 ml round-bottom flask. Add 100 ml of ultrapure water, connect the volatile acid distillation apparatus, and use nitrogen as the carrier gas at a flow rate controlled at 40-60 ml / min. Heat and distill. Collect the distillate into a 250 ml volumetric flask and dilute to the mark with ultrapure water. Filter an appropriate amount of the distillate through a 0.22 μm filter membrane and inject it into the ion chromatograph for analysis. Each sample was measured three times, and the average value was taken as the accurate sulfur content.

[0041] Simultaneously, using a near-infrared spectrometer following the aforementioned spectral acquisition steps, spectra were acquired for each standard sample. The spectral data were then mapped one-to-one with the corresponding sulfur content values ​​to establish a spectral database of Astragalus membranaceus standard samples. The database contains detailed information for each standard sample, such as its number, preparation time, sulfur content, and spectral data file path, facilitating subsequent retrieval and access.

[0042] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of Astragalus membranaceus standard samples. During the modeling process, leave-one-out cross-validation was employed to optimize the model. Specifically, each standard sample in the database was set aside as a validation set, and the remaining samples were used as the training set. The model was then built using the training set data, and predictions were made on the validation set samples using the established model. The prediction error was calculated. This process was repeated until each sample was used as a validation set once. By continuously adjusting the number of principal components, the root mean square error of prediction (RMSEP) and correlation coefficient (R²) of the model under different principal component numbers were calculated.

[0043] When the number of principal components is 8, the model reaches its optimal performance, with an RMSEP of 3.2% and an R² of 0.97. At this point, the model is determined to be the quantitative analysis model for sulfur content in Astragalus membranaceus. To further verify the model's stability, 10% of the standard samples in the database were randomly selected as an independent validation set for external validation. The validation results show that the relative error of the model's predictions for the independent validation set samples is within 5%, demonstrating that the model has good stability and accuracy.

[0044] Sample Testing: A sample of Astragalus membranaceus with unknown sulfur content was taken and processed strictly according to the sample pretreatment and spectral acquisition steps described above to obtain the spectral data of the sample. First, a baseline correction algorithm based on polynomial fitting was used to process the spectral data to eliminate baseline shifts caused by instrument drift or sample matrix effects. Then, Savitzky-Golay convolution smoothing (window width of 9, polynomial order of 2) was used to smooth the spectral data, removing high-frequency noise interference and improving the signal-to-noise ratio of the spectral data.

[0045] The preprocessed spectral data was input into the established quantitative analysis model for sulfur content in Astragalus membranaceus. The model, through its internal algorithm, quickly output the sulfur content of the sample as 235 mg / kg. To verify the accuracy of the results, the sample was subjected to a second determination using ion chromatography, yielding a sulfur content of 232 mg / kg. The relative error between the two methods was 1.3%, demonstrating the accuracy and reliability of this method. Furthermore, a repeatability experiment was performed on the sample, with five repeated measurements. The relative standard deviation (RSD) of the results was 2.1%, indicating good repeatability of the method.

[0046] Example 2: Determination of sulfur content in Codonopsis pilosula Sample Pretreatment: Twenty batches of Codonopsis pilosula samples of different growth years were collected from multiple medicinal herb cultivation bases. First, the fibrous roots, residual soil, and attached weeds on the surface of the Codonopsis pilosula were carefully removed manually. For larger samples, they were cut into small sections, each approximately 2-3 cm in length, using a stainless steel knife. Then, the treated Codonopsis pilosula was placed in a cryogenic pulverizer and pulverized at -20°C for 8-10 minutes, ensuring the sample could pass through a 20-mesh standard sieve. Cryogenic pulverization effectively preserves the heat-sensitive components in Codonopsis pilosula, preventing component loss due to temperature increases during the pulverization process.

[0047] The pulverized Codonopsis pilosula sample was transferred to a ceramic crucible and placed in a vacuum drying oven at a constant temperature of 40℃ for drying. The vacuum level of the drying oven was maintained at -0.08 MPa to accelerate the evaporation of moisture and prevent oxidation of the sample during the drying process. Every 1.5 hours, the crucible was removed, placed in a desiccator to cool to room temperature, and then weighed. When the difference between two consecutive weighings was less than 0.001 g, the sample was considered to have reached constant weight. After drying, the sample was placed in a desiccator equipped with molecular sieves, sealed, and stored for 30 minutes to allow the sample temperature to equalize with the ambient temperature before use.

[0048] Spectral Acquisition: A near-infrared spectrometer (model: FT-NIR8700) equipped with a high-sensitivity InGaAs detector was selected, which boasts excellent spectral resolution and signal-to-noise ratio. Before acquiring the spectra, wavelength and energy calibrations were performed to ensure measurement accuracy. The dried Codonopsis pilosula sample was placed into a 3cm diameter cylindrical sample cup and gently compacted to ensure a smooth and void-free sample surface.

[0049] The spectrometer was set to scan 48 times, with a resolution of 6 cm⁻¹ and a wavelength range of 800-2500 nm. To ensure the reliability of the spectral data, each sample was continuously acquired 5 times, with a 30-second interval between each acquisition. During the acquisition process, a temperature and humidity logger was used to monitor the ambient temperature and humidity in real time, maintaining the ambient temperature at 23±1℃ and the relative humidity at 50%±5%. After acquisition, the spectral data was transferred to the computer, numbered, and stored for subsequent analysis.

[0050] Establish a standard sample spectral database: High-quality Codonopsis pilosula raw materials without sulfur fumigation were selected, and sodium sulfite powder was added in different proportions to prepare standard samples with sulfur contents of 0 mg / kg, 30 mg / kg, 60 mg / kg, 90 mg / kg, 120 mg / kg, 150 mg / kg, 200 mg / kg, 300 mg / kg, 500 mg / kg, 800 mg / kg, and 1000 mg / kg. Twenty standard samples were prepared for each concentration level. During the preparation process, sodium sulfite powder was accurately weighed using a precision balance (accuracy of 0.0001 g), and it was thoroughly mixed with the Codonopsis pilosula samples using a stirrer.

[0051] The sulfur content of each standard sample was determined using ion chromatography. The specific steps were as follows: 1.2 g (accurate to 0.0001 g) of the standard sample was accurately weighed and placed in a 250 ml Kjeldahl flask. 120 ml of ultrapure water was added, and the flask was connected to a steam distillation apparatus and heated in a 40°C water bath for distillation. The distillate was collected in a 250 ml volumetric flask and diluted to the mark with ultrapure water. An appropriate amount of the distillate was filtered through a 0.45 μm filter membrane and injected into the ion chromatograph for analysis. Each sample was measured in triplicate, and the average value was taken as the accurate sulfur content.

[0052] Simultaneously, near-infrared spectroscopy was used to acquire spectra for each standard sample, and the spectral data were mapped one-to-one with the corresponding sulfur content values ​​to establish a spectral database of Codonopsis pilosula standard samples. The database records detailed preparation information, measurement results, and spectral data for each standard sample, facilitating data management and analysis.

[0053] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of Codonopsis pilosula standard samples. Leave-one-out cross-validation was employed to optimize the model. By continuously adjusting the number of principal components, the root mean square error of prediction (RMSEP) and correlation coefficient (R²) of the model under different principal component numbers were calculated. After multiple experiments, the model achieved optimal performance when the number of principal components was 10, with an RMSEP of 4.1% and an R² of 0.96.

[0054] To evaluate the model's generalization ability, 20% of the standard samples were randomly selected from the database as the test set, and the remaining samples were used as the training set. The model was then built using the training set data, and predictions were made on the test set samples using the established model. The results show that the relative error of the model's predictions for the test set samples was within 6%, indicating that the model has good generalization ability and can accurately predict the sulfur content of unknown samples.

[0055] Sample Testing: A sample of Codonopsis pilosula with unknown sulfur content was taken and processed according to the sample pretreatment and spectral acquisition steps described above to obtain the spectral data of the sample. First, the spectral data was processed using the multivariate scattering correction (MSC) method to eliminate scattering effects caused by factors such as sample particle size and uneven distribution. Then, the spectral data was further smoothed using the Savitzky-Golay convolution smoothing method (window width of 11, polynomial order of 3) to improve the quality of the spectral data.

[0056] The preprocessed spectral data was input into the established quantitative analysis model for sulfur content in Codonopsis pilosula. The model calculated the sulfur content of the sample to be 180 mg / kg. To verify the accuracy of the results, a second determination was performed using the classic acid-base titration method. Specifically, 1.5 g (accurate to 0.0001 g) of sample was accurately weighed and placed in a 250 ml Erlenmeyer flask. 50 ml of ultrapure water was added, and the sample was shaken to dissolve. 2-3 drops of methyl red-methylene blue mixed indicator were added, and titration was performed with 0.1 mol / L sodium hydroxide standard titration solution until the solution changed from purple-red to green. Based on the volume of sodium hydroxide standard titration solution consumed, the sulfur content in the sample was calculated to be 178 mg / kg. The relative error between the two methods was 1.1%, proving that the method's determination results are accurate and reliable. Simultaneously, a spiked recovery experiment was performed on the sample. A certain amount of sodium sulfite standard solution was added to the sample, and the sample was analyzed using this method. The spiked recovery rate was between 95% and 105%, further verifying the accuracy and reliability of this method.

[0057] Example 3: Determination of Sulfur Content in Angelica sinensis Sample Pretreatment: Angelica sinensis samples of different processing methods, including sliced ​​and whole roots, were selected from the wholesale market for Chinese medicinal herbs, totaling 15 batches. First, impurities, moldy parts, and residual packaging materials were thoroughly removed from the samples. Sliced ​​Angelica sinensis was directly pulverized; whole roots were first cut into small sections with scissors, then pulverized in a universal grinder. The grinder speed was adjusted to 12000 rpm, and the pulverization time was 4-6 minutes, until all samples passed through a 20-mesh standard sieve.

[0058] The pulverized Angelica sinensis sample was placed in a petri dish and then dried in an electrically heated constant-temperature drying oven set at 40℃. The drying oven was equipped with an automatic temperature control system, with temperature fluctuations not exceeding ±1℃. During the drying process, the sample was removed every 2 hours and gently stirred to ensure uniform drying. The sample was considered to have reached constant weight when the difference between two consecutive weighings did not exceed 0.001g. After drying, the sample was transferred to a desiccator, sealed, and allowed to stand for 30 minutes to cool to room temperature, preventing excessive temperature from affecting the spectral acquisition results.

[0059] Spectral acquisition: A Fourier transform near-infrared spectrometer (model: Tensor27-NIR) was used. This instrument is equipped with a diffuse reflection fiber optic probe, which can easily acquire spectra of samples of different shapes. Before acquiring the spectra, the instrument was initialized, including wavelength range calibration and scan speed settings. The cooled Angelica sinensis sample was placed in the sample cell and gently compacted with a sample press to make the sample surface flat.

[0060] The spectrometer was set to scan 64 times, with a resolution of 8 cm⁻¹ and a wavelength range of 800-2500 nm. To minimize the impact of environmental factors on the spectral data, each sample was sampled three times, with a 2-minute interval between each sample acquisition. During the acquisition process, an air conditioner and humidifier were used to maintain the ambient temperature at 24±1℃ and the relative humidity at 55%±5%. After acquisition, the spectral data were saved in text format for subsequent data processing and analysis.

[0061] Establish a standard sample spectral database: High-quality Angelica sinensis raw materials without sulfur fumigation were selected. Standard samples with sulfur content gradients of 0 mg / kg, 40 mg / kg, 80 mg / kg, 120 mg / kg, 160 mg / kg, 200 mg / kg, 300 mg / kg, 400 mg / kg, 600 mg / kg, 800 mg / kg, and 1000 mg / kg were prepared by adding sodium bisulfite solution of different concentrations. Eighteen standard samples were prepared for each concentration level. During preparation, it was ensured that the solution and Angelica sinensis sample were thoroughly mixed and homogeneous. The samples were then left at room temperature for 3 hours to allow the sodium bisulfite to fully react with the Angelica sinensis.

[0062] The sulfur content of each standard sample was determined using ion chromatography. The specific procedure was as follows: Accurately weigh 1.3 g (accurate to 0.0001 g) of the standard sample and place it in a 250 ml distillation flask. Add 100 ml of ultrapure water, connect the volatile acid distillation apparatus, use air as the carrier gas, control the flow rate at 50 ml / min, and heat for distillation. Collect the distillate into a 250 ml volumetric flask and dilute to the mark with ultrapure water. Take an appropriate amount of the distillate, filter it through a 0.22 μm filter membrane, and inject it into the ion chromatograph for analysis. Each sample was measured three times, and the average value was taken as the accurate value of its sulfur content.

[0063] Simultaneously, near-infrared spectroscopy was used to acquire the spectra of each standard sample, and the spectral data were mapped one-to-one with the corresponding sulfur content values ​​to establish a spectral database of Angelica sinensis standard samples. The database contains detailed preparation records, measurement results, and spectral data files for each standard sample, facilitating data retrieval and management.

[0064] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of Angelica sinensis standard samples. Leave-one-out cross-validation was employed to optimize the model. By continuously experimenting with different numbers of principal components, the root mean square error of prediction (RMSEP) and correlation coefficient (R²) were calculated. The model achieved its optimal performance with 9 principal components, resulting in an RMSEP of 3.8% and an R² of 0.97.

[0065] To verify the stability of the model, a repeatability experiment was conducted. Ten standard samples were randomly selected from the database, and spectral data were collected ten times. After each collection, the model was used for prediction. The results showed that the relative standard deviation (RSD) of the prediction results was less than 3%, indicating that the model has good stability and can accurately predict the sulfur content of Angelica sinensis samples under different times and conditions.

[0066] Sample Testing: A sample of Angelica sinensis with unknown sulfur content was taken and processed according to the sample pretreatment and spectral acquisition steps described above to obtain the spectral data of the sample. First, the standard normal variable transformation (SNV) method was used to process the spectral data to eliminate baseline drift and scattering effects caused by factors such as sample particle size and density. Then, the Savitzky-Golay convolution smoothing method (window width of 13, polynomial order of 2) was used to smooth the spectral data to improve its quality.

[0067] The preprocessed spectral data were input into the established quantitative analysis model for sulfur content in Angelica sinensis. The model calculated the sulfur content of the sample to be 310 mg / kg. To verify the accuracy of the results, a second determination was performed using high-performance liquid chromatography (HPLC). Specifically, 1.2 g (accurate to 0.0001 g) of sample was accurately weighed and placed in a 250 ml Erlenmeyer flask. 50 ml of methanol was added, and the sample was ultrasonically extracted for 30 minutes. The mixture was then filtered, and the filtrate was analyzed by HPLC. Based on the peak area, the sulfur content in the sample was calculated to be 308 mg / kg. The relative error between the two methods was 0.6%, demonstrating the accuracy and reliability of this method. Furthermore, a comparative experiment was conducted between different instruments. The sample was detected using another near-infrared spectrometer of the same model. The relative error between the detected result and the original instrument's result was within 2%, further validating the accuracy and versatility of this method.

[0068] Example 4: Determination of sulfur content in wolfberry Sample Pretreatment: Goji berry samples were collected from growers in different producing areas, including Ningxia, Qinghai, and Xinjiang, totaling 25 batches. First, shriveled fruits, branches, leaves, and any foreign matter were carefully removed from the samples. Then, the goji berries were placed in a food processor and ground for 2-3 minutes to ensure uniform particle size, allowing them to pass through a 20-mesh standard sieve. Due to the high sugar content of goji berries, clumping may occur during grinding. In this case, a small amount of anhydrous ethanol (5-10 ml per 100g sample) can be added to help disperse the berries and prevent sticking. After grinding, the ethanol is allowed to evaporate through ventilation and drying.

[0069] The pulverized goji berry sample was placed in an enamel dish and dried in a vacuum drying oven set at 40℃. The vacuum level of the drying oven was maintained at -0.09MPa, which effectively lowers the boiling point of water, accelerates the drying process, and avoids damage to the heat-sensitive components in the goji berries due to high temperatures. During the drying process, the sample was removed every 1.5 hours and gently stirred to ensure uniform drying. The sample was considered to have reached constant weight when the difference between two consecutive weighings did not exceed 0.001g. After drying, the sample was transferred to a desiccator containing silica gel, sealed, and allowed to stand for 30 minutes to cool to room temperature.

[0070] Spectral Acquisition: A near-infrared spectrometer (model: Antaris II) equipped with a diffuse reflectance integrating sphere attachment was selected. This instrument features a wide spectral range and high sensitivity. Before acquiring the spectrum, a comprehensive performance check of the instrument was performed, including calibration of parameters such as wavelength accuracy, absorbance accuracy, and resolution. The cooled wolfberry sample was carefully poured into the integrating sphere sample cell, filling it as much as possible and gently compacting it to ensure a smooth and void-free sample surface.

[0071] The spectrometer was set to scan 32 times, with a resolution of 4 cm⁻¹ and a wavelength range of 800-2500 nm. To ensure the reliability of the spectral data, each sample was sampled five times, with a one-minute interval between each acquisition. During the acquisition process, ambient temperature and humidity were monitored in real time using temperature and humidity sensors, maintaining the ambient temperature at 22±1℃ and the relative humidity at 40%±5%. After acquisition, the instrument automatically stored the spectral data in a specific binary format on the computer's hard drive and generated a unique number for each spectral data file for easy data management and retrieval.

[0072] Establish a standard sample spectral database: High-quality wolfberry raw materials without sulfur fumigation were selected, and standard samples with sulfur content gradients of 0 mg / kg, 20 mg / kg, 40 mg / kg, 60 mg / kg, 80 mg / kg, 100 mg / kg, 150 mg / kg, 200 mg / kg, 300 mg / kg, 500 mg / kg, 800 mg / kg, and 1000 mg / kg were prepared by precisely adding sodium metabisulfite solution. Twenty standard samples were prepared for each concentration level. During the preparation process, sodium metabisulfite solution was accurately measured with a pipette and thoroughly mixed with the wolfberry sample in a sealed container. The mixture was then allowed to stand at room temperature for 4 hours to allow the sodium metabisulfite to fully react with the wolfberry.

[0073] The sulfur content of each standard sample was determined by ion chromatography. The specific steps were as follows: Accurately weigh 0.8 g (accurate to 0.0001 g) of the standard sample and place it in a 250 ml round-bottom flask. Add 80 ml of ultrapure water, connect the steam distillation apparatus, use nitrogen as the carrier gas, control the flow rate at 50 ml / min, and heat for distillation. Collect the distillate into a 200 ml volumetric flask and dilute to the mark with ultrapure water. Take an appropriate amount of the distillate, filter it through a 0.22 μm filter membrane, and inject it into the ion chromatograph for analysis. The chromatographic conditions for the ion chromatogram were: anion exchange column (IonPacAS11-HC), eluent of potassium hydroxide solution (gradient elution), flow rate 1.0 ml / min, and injection volume 25 μl. Each sample was measured in triplicate, and the average value was taken as the accurate value of its sulfur content.

[0074] Simultaneously, near-infrared spectroscopy was used to acquire spectra for each standard sample, and the spectral data were mapped one-to-one with the corresponding sulfur content values ​​to establish a spectral database of wolfberry standard samples. The database records detailed information on the preparation of standard samples (such as the origin of raw materials, the type and amount of added reagents, preparation time, etc.), ion chromatography results, and detailed parameters of spectral data, facilitating subsequent model building and data analysis.

[0075] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of wolfberry standard samples. Leave-one-out cross-validation was employed to optimize the model. During optimization, the root mean square error of prediction (RMSEP) and correlation coefficient (R²) were calculated for different principal component numbers by continuously adjusting the number of principal components. After multiple experiments and parameter adjustments, the model achieved optimal performance with 7 principal components, resulting in an RMSEP of 3.5% and an R² of 0.98.

[0076] To further verify the accuracy and reliability of the model, 20% of the standard samples were randomly selected from the database as an independent validation set. The model was then built using the remaining 80% of the samples, and predictions were made on the independent validation set samples using the established model. The results show that the relative error of the model's predictions for the independent validation set samples is within 4%, and the predicted values ​​have a good correlation with the actual values, proving that the model has strong generalization ability and accuracy, and can be effectively applied to the detection of real samples.

[0077] Sample Testing: A sample of wolfberry with unknown sulfur content was taken and processed strictly according to the sample pretreatment and spectral acquisition steps described above to obtain the spectral data of the sample. First, the spectral data was preprocessed using a combination of multivariate scattering correction (MSC) and standard normal variable transformation (SNV) to eliminate spectral interference caused by factors such as sample particle inhomogeneity and surface reflection. Then, the spectral data was smoothed using the Savitzky-Golay convolution smoothing method (window width of 11, polynomial order of 3) to further improve the quality of the spectral data.

[0078] The preprocessed spectral data was input into the established quantitative analysis model for the sulfur content of wolfberry. After complex calculations and analysis, the model quickly output the sulfur content of the sample as 260 mg / kg. To verify the accuracy of the results, a second determination of the sample was performed using gas chromatography-mass spectrometry (GC-MS). The specific procedure was as follows: 0.5 g (accurate to 0.0001 g) of sample was accurately weighed and placed in a 50 ml headspace vial. 10 ml of ultrapure water was added, and the vial was sealed and equilibrated at 60 °C for 30 minutes. Then, 1 ml of headspace gas was extracted using a gas-tight needle and injected into the GC-MS instrument for analysis. The GC conditions were: DB-5MS column (30 m × 0.25 mm × 0.25 μm), injection port temperature 250 °C, and column temperature program (initial temperature 40 °C, hold for 2 min, increase to 280 °C at 10 °C / min, hold for 5 min). The mass spectrometry conditions were: electron impact source (EI), electron energy 70 eV, and scan range m / z 30-500. Based on the peak area of ​​characteristic ions in the mass spectrum, the sulfur content in the sample was calculated to be 258 mg / kg using the external standard method. The relative error between the two methods was 0.8%, proving that the results obtained by this method are accurate and reliable.

[0079] Furthermore, to examine the applicability of this method under different laboratory environments, the sample was sent to three different laboratories and detected using the same type of near-infrared spectrometer and the method described in this invention. The results showed that the relative errors between the detection results from the three laboratories and the above-described results were all within 3%, indicating that this method has good inter-laboratory repeatability and applicability, and can stably and accurately determine the sulfur content in wolfberries under different laboratory environments.

[0080] Example 5: Determination of Sulfur Content in Licorice Sample Pretreatment: Twenty batches of licorice samples of different grades, including wild and cultivated licorice, were collected from multiple Chinese medicinal herb distribution centers. First, the surface of the licorice samples was carefully cleaned of mud, sand, and residual grass debris with a brush. Larger licorice roots were cut into small sections of approximately 1-2 cm using stainless steel scissors. Then, the treated licorice was placed in a high-speed universal grinder and ground for 5-7 minutes to ensure all samples passed through a 20-mesh standard sieve. During grinding, to prevent overheating, grinding was paused every minute, allowing the grinder to cool before resuming operation.

[0081] The pulverized licorice sample was spread evenly in a glass petri dish and placed in an electrically heated constant-temperature drying oven at 40℃. The air velocity inside the drying oven was set to 0.5 m / s to ensure uniform air circulation and accelerate moisture evaporation. During the drying process, the sample was removed every 2 hours and gently stirred with a glass rod to ensure uniform drying. The sample was considered to have reached constant weight when the difference between two consecutive weighings did not exceed 0.001 g. After drying, the sample was transferred to a desiccator equipped with molecular sieves, sealed, and allowed to stand for 30 minutes to allow the sample to cool fully to room temperature, preventing temperature differences from affecting the accuracy of spectral acquisition.

[0082] Spectral Acquisition: A portable near-infrared spectrometer (model: MicroNIR1700) was used. This instrument is compact, portable, and easy to operate, making it suitable for rapid on-site testing. Before acquiring the spectrum, the instrument was preheated and self-tested to ensure that all performance indicators were normal. The cooled licorice sample was placed into a dedicated sample cup and gently compacted using the accompanying sample press to make the sample surface smooth and dense.

[0083] The spectrometer was set to scan 48 times, with a resolution of 6 cm⁻¹ and a wavelength range of 800-2500 nm. To minimize measurement errors, each sample was sampled three times, with a 30-second interval between each acquisition. During the acquisition process, a portable thermometer and hygrometer were used to monitor the ambient temperature and humidity, maintaining the temperature at 23±1℃ and the relative humidity at 45%±5%. After acquisition, the spectral data was directly transferred to a laptop via USB and stored in a specific Excel format for easy subsequent data processing and analysis.

[0084] Establish a standard sample spectral database: High-quality licorice raw materials without sulfur fumigation were selected, and standard samples with sulfur content gradients of 0 mg / kg, 30 mg / kg, 60 mg / kg, 90 mg / kg, 120 mg / kg, 150 mg / kg, 200 mg / kg, 300 mg / kg, 500 mg / kg, 800 mg / kg, and 1000 mg / kg were prepared by adding sodium sulfite solid powder. Eighteen standard samples were prepared for each concentration level. Sodium sulfite powder was accurately weighed using an analytical balance with an accuracy of 0.0001 g, and thoroughly ground and mixed with the licorice sample in a mortar. The mixture was then placed in a sealed container for 3 hours to allow the sodium sulfite and licorice to fully react.

[0085] The sulfur content of each standard sample was determined by ion chromatography. The specific procedure was as follows: Accurately weigh 1.0 g (accurate to 0.0001 g) of the standard sample and place it in a 250 ml distillation flask. Add 100 ml of ultrapure water, connect the volatile acid distillation apparatus, use air as the carrier gas, and control the flow rate at 40-60 ml / min. Heat and distill. Collect the distillate into a 250 ml volumetric flask and dilute to the mark with ultrapure water. Take an appropriate amount of the distillate, filter it through a 0.22 μm filter membrane, and inject it into the ion chromatograph for analysis. The ion chromatograph used the conductivity-suppressed detection mode. The specific chromatographic conditions were: an anion exchange column (Dionex AS19), eluent of potassium hydroxide solution (isocratic elution, concentration 20 mmol / L), flow rate 1.0 ml / min, and injection volume 25 μl. Each sample was measured three times, and the average value was taken as the accurate value of its sulfur content.

[0086] Simultaneously, a portable near-infrared spectrometer was used to acquire spectra for each standard sample, and the spectral data were mapped one-to-one with the corresponding sulfur content values ​​to establish a spectral database of licorice standard samples. The database records in detail the preparation process of the standard samples, the results of ion chromatography determination, and the specific parameters of spectral acquisition (such as instrument model, number of scans, resolution, etc.) to facilitate subsequent model building and data traceability.

[0087] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of licorice standard samples. Leave-one-out cross-validation was employed to optimize the model. During optimization, different numbers of principal components were continuously tested, and the root mean square error of prediction (RMSEP) and correlation coefficient (R²) were calculated. After multiple experiments and parameter adjustments, the model achieved optimal performance with 8 principal components, resulting in an RMSEP of 4.0% and an R² of 0.97.

[0088] To evaluate the model's stability in practical applications, a long-term stability test was conducted. For one week, 10 standard samples were randomly selected daily from the database for spectral acquisition and prediction, and the relative standard deviation (RSD) of the prediction results was calculated. The results showed that the RSD of the prediction results was less than 3.5% throughout the week, indicating that the model has good long-term stability and can maintain high accuracy and reliability during long-term detection.

[0089] Sample Testing: A licorice sample with unknown sulfur content was taken and processed according to the sample pretreatment and spectral acquisition steps described above to obtain the spectral data of the sample. First, a baseline correction algorithm (polynomial fitting, order 5) was used to process the spectral data to eliminate baseline shifts caused by instrument drift or sample matrix effects. Then, Savitzky-Golay convolution smoothing (window width 9, polynomial order 2) was used to smooth the spectral data to improve the signal-to-noise ratio of the spectral data.

[0090] The preprocessed spectral data was input into the established quantitative analysis model for sulfur content in licorice. The model calculated the sulfur content of the sample to be 150 mg / kg. To verify the accuracy of the results, inductively coupled plasma mass spectrometry (ICP-MS) was used for a second determination of the sample. Specifically, 0.2 g (accurate to 0.0001 g) of sample was accurately weighed and placed in a polytetrafluoroethylene digestion vessel. 5 ml of nitric acid and 1 ml of hydrofluoric acid were added, and the vessel was sealed and placed in a microwave digester for digestion. After digestion, the digestate was transferred to a 50 ml volumetric flask and diluted to the mark with ultrapure water. An appropriate amount of solution was filtered through a 0.45 μm filter membrane and injected into the ICP-MS instrument for analysis. Based on the signal intensity of sulfur in the mass spectrum, the sulfur content in the sample was calculated to be 153 mg / kg using the standard curve method. The relative error between the two methods was 2.0%, proving that the results obtained by this method are accurate and reliable.

[0091] Furthermore, to examine the applicability of this method to different forms of licorice samples (such as licorice tablets and licorice powder), licorice tablets and licorice powder samples with the same sulfur content were selected and tested according to this method. The results showed that the relative errors between the test results and the actual values ​​for both forms of samples were within 3%, indicating that this method has good applicability to different forms of licorice samples and can meet the needs of different sample testing.

[0092] Example 6: Determination of Sulfur Content in Honeysuckle Sample Pretreatment: A total of 25 batches of honeysuckle samples from different harvesting periods (e.g., first and second harvests) were collected from various honeysuckle growing areas. First, leaves, flower stalks, and impurities such as weeds were carefully removed from the honeysuckle samples. For flowers that were stuck together, they were carefully separated with tweezers. Then, the honeysuckle was placed in a freeze dryer and pre-frozen at -50℃ for 2 hours to completely freeze the moisture in the sample. Next, the vacuum degree of the freeze dryer was reduced to below 10 Pa, and sublimation drying was carried out at -35℃ for 12-15 hours until the sample was completely dry. The dried honeysuckle was brittle and was gently ground into a fine powder in a mortar, allowing it to pass through a 20-mesh standard sieve.

[0093] The ground honeysuckle sample was placed in a glass weighing bottle and then placed in a drying oven set at 40℃ for secondary drying. A hot air circulation system was used in the drying oven to ensure uniform temperature distribution. During the drying process, the sample was removed every 1.5 hours, cooled to room temperature in a desiccator, and then weighed. The sample was considered to have reached constant weight when the difference between two consecutive weighings did not exceed 0.001g. After drying, the sample was transferred to a desiccator containing silica gel, sealed, and allowed to stand for 30 minutes to cool to room temperature, preventing temperature changes from affecting the stability of spectral acquisition.

[0094] Spectral Acquisition: A high-resolution near-infrared spectrometer (model: Nicoleti S50) was used. This instrument is equipped with an advanced optical system and detector, capable of providing high-quality spectral data. Before acquiring the spectrum, the instrument underwent comprehensive calibration, including wavelength calibration, energy calibration, and resolution calibration. The cooled honeysuckle sample was placed into a cylindrical sample cell with a diameter of 3 cm and gently compacted to ensure a smooth sample surface free of air bubbles.

[0095] The spectrometer was set to scan 64 times, with a resolution of 8 cm⁻¹ and a wavelength range of 800-2500 nm. To ensure the accuracy of the spectral data, each sample was sampled five times, with a one-minute interval between each sample acquisition. During the acquisition process, the ambient temperature was maintained at 24±1℃ and the relative humidity at 50%±5% using a temperature and humidity control device. After acquisition, the spectral data was automatically stored on the instrument's built-in hard drive and then transmitted via network to a data analysis workstation for saving in a specific spectral data format (e.g., *.spa).

[0096] Establish a standard sample spectral database: High-quality honeysuckle raw materials without sulfur fumigation were selected. Standard samples with sulfur content gradients of 0 mg / kg, 25 mg / kg, 50 mg / kg, 75 mg / kg, 100 mg / kg, 125 mg / kg, 150 mg / kg, 200 mg / kg, 300 mg / kg, 500 mg / kg, 800 mg / kg, and 1000 mg / kg were prepared by adding potassium sulfite solution. Twenty standard samples were prepared for each concentration level. During preparation, potassium sulfite solution was accurately measured using a pipette and thoroughly mixed with the honeysuckle sample in a sealed container. The mixture was then allowed to stand at room temperature for 5 hours to allow the potassium sulfite to fully react with the honeysuckle.

[0097] The sulfur content of each standard sample was determined by ion chromatography. The specific steps were as follows: 0.6 g (accurate to 0.0001 g) of standard sample was accurately weighed and placed in a 250 ml round-bottom flask. 60 ml of ultrapure water was added, and the flask was connected to a steam distillation apparatus. Nitrogen was used as the carrier gas, and the flow rate was controlled at 40 ml / min. The distillate was collected into a 150 ml volumetric flask and diluted to the mark with ultrapure water. An appropriate amount of the distillate was filtered through a 0.22 μm filter membrane and injected into the ion chromatograph for analysis. The chromatographic conditions were: an anion exchange column (MetrosepASupp5-250), eluent of a sodium carbonate-sodium bicarbonate mixture (concentrations of 3.2 mmol / L and 1.0 mmol / L, respectively), flow rate of 0.7 ml / min, and injection volume of 20 μl. Each sample was measured in triplicate, and the average value was taken as the accurate sulfur content value.

[0098] Simultaneously, near-infrared spectroscopy was used to acquire spectra of each standard sample, and the spectral data were mapped one-to-one with the corresponding sulfur content values ​​to establish a honeysuckle standard sample spectral database. The database records detailed information such as the preparation date of the standard sample, the source of raw materials, the concentration of added reagents, the ion chromatography results, and detailed parameters of spectral acquisition, which facilitates subsequent model construction and data management.

[0099] A quantitative analysis model was established: Principal component analysis-partial least squares (PCA-PLS) was used to analyze the spectral database of honeysuckle standard samples. Leave-one-out cross-validation was employed to optimize the model. By continuously adjusting the number of principal components, the root mean square error of prediction (RMSEP) and correlation coefficient (R²) were calculated for different principal component numbers. After multiple experiments, the model achieved optimal performance with 11 principal components, resulting in an RMSEP of 2.6% and an R² of 0.988.

[0100] To further verify the reliability of the model, 30% of the standard samples were randomly selected from the database as the validation set, and the remaining 70% were used as the training set. After building the model using the training set data, predictions were made on the validation set samples. The results showed that the relative prediction errors were all within 3%, indicating that the model has good generalization ability and accuracy, and can be reliably applied to actual sample detection.

[0101] Sample Testing: A honeysuckle sample with unknown sulfur content was taken and tested. The sample pretreatment and spectral acquisition steps described above were strictly followed to obtain the spectral data of the sample. First, multivariate scattering correction (MSC) and standard normal variable transformation (SNV) were used to preprocess the spectral data to eliminate spectral interference caused by factors such as sample particle size and uneven distribution. Then, Savitzky-Golay convolution smoothing (window width of 15, polynomial order of 4) was used to smooth the spectral data to further improve the spectral quality.

[0102] The preprocessed spectral data was input into the established quantitative analysis model for the sulfur content of honeysuckle. After complex calculations and analysis, the model quickly output the sulfur content of the sample as 185 mg / kg. To verify the accuracy of the results, the classic acid-base titration method was used. A second determination was performed on the sample. The specific procedure was as follows: 1.2 g (accurate to 0.0001 g) of sample was accurately weighed and placed in a 250 ml Erlenmeyer flask. 50 ml of water was added, and the sample was shaken to fully wet it. The flask was then connected to a distillation apparatus for distillation. The distillate was collected into an absorption flask containing excess hydrogen peroxide to ensure that sulfur dioxide was fully oxidized to sulfuric acid. After distillation, 2-3 drops of methyl red-methylene blue mixed indicator were added to the absorption flask, and the solution was titrated with 0.1 mol / L sodium hydroxide standard titration solution until the solution changed from purple-red to green, which was the endpoint. Based on the volume of sodium hydroxide standard titration solution consumed, the sulfur content in the sample was calculated to be 188 mg / kg. The relative error between the two methods was 1.6%, proving that the results obtained by this method are accurate and reliable.

[0103] To examine the repeatability of this method across different instruments, the sample was analyzed using another near-infrared spectrometer of the same model. The result was 183 mg / kg, with a relative error of 1.1% compared to the initial result, indicating good repeatability across different instruments. Simultaneously, a spiked recovery experiment was performed on the sample. A certain amount of potassium sulfite standard solution was added to the sample, and the sample was analyzed according to this method. The spiked recovery rate was between 96% and 103%, further verifying the accuracy and reliability of the method. Furthermore, this method was applied to the detection of honeysuckle samples from different origins and batches, and it could rapidly and accurately determine the sulfur content, demonstrating its wide applicability and effective application in the practical work of honeysuckle quality supervision.

[0104] Example 7: Qualitative Identification of Honeysuckle Fumigated with Sulfur Sample Pretreatment: A total of 50 batches of raw honeysuckle from three main producing areas—Pingyi in Shandong, Fengqiu in Henan, and Julu in Hebei—were selected, along with commercially available sulfur-fumigated honeysuckle samples. First, impurities such as flower stalks, withered leaves, and dirt were manually removed from the samples, and only intact, plump flowers were selected. For the sulfur-fumigated samples, standard samples with sulfur contents of 0 mg / kg (unfumigated), 50 mg / kg, 100 mg / kg, 200 mg / kg, 500 mg / kg, and 1000 mg / kg were prepared by controlling the amount of potassium sulfite solution sprayed, with 10 samples prepared for each concentration to ensure sample homogeneity.

[0105] Selected honeysuckle was placed in a freeze dryer and pre-frozen at -50℃ for 2 hours to completely freeze the moisture. It was then sublimated and dried at -35℃ under a vacuum of 10Pa for 15 hours to avoid damage to volatile oils and other components due to high temperatures. The dried sample was brittle and gently ground in a mortar, then passed through a 20-mesh standard sieve to remove oversized particles. 5g of the pulverized sample was weighed and placed in a glass petri dish, then placed in a 40℃ electric thermostatic drying oven (with a hot air circulation system, temperature fluctuation ±1℃) until constant weight was achieved. The sample was weighed every hour until the difference between two consecutive weighings was <0.001g. After drying, the sample was transferred to a desiccator containing silica gel and sealed for 30 minutes to room temperature to ensure temperature stability during spectral acquisition.

[0106] Spectral acquisition: A high-resolution near-infrared spectrometer (model: Nicoleti S50, Thermo Fisher Scientific) equipped with a diffuse reflectance integrating sphere accessory was used. After the instrument warmed up for 30 minutes, wavelength calibration (800-2500nm) and energy calibration were performed to ensure the spectrometer was in optimal working condition. The cooled honeysuckle sample was filled into a 5cm diameter sample cell and gently compacted to smooth the surface and avoid inter-particle voids affecting the diffuse reflectance signal.

[0107] The spectrometer parameters were set as follows: 64 scans, resolution 8 cm⁻¹, wavelength range 800-2500 nm. To control environmental influences, air conditioning and a humidifier were used to maintain the room temperature at 24±1℃ and the relative humidity at 50%±5% during the acquisition process. Each sample was sampled five times, with a one-minute interval between each acquisition. The average spectrum was used as the raw data to eliminate incidental noise interference. The spectral data were automatically saved in .spa format, including sample number, acquisition time, instrument parameters, and other information for subsequent analysis.

[0108] Establish a standard sample spectral database: Sulfur content confirmation and grouping. Ion chromatography was used to determine the sulfur content of all standard samples. Specific steps: Weigh 0.6g of sample into a 250ml round-bottom flask, add 60ml of ultrapure water, connect to a steam distillation apparatus, and heat and distill using nitrogen as the carrier gas (flow rate 40ml / min). Collect the distillate into a 150ml volumetric flask and dilute to volume. Filter the supernatant through a 0.22μm filter membrane and inject it into an ion chromatograph (column: MetrosepASupp5-250, eluent: 3.2mmol / L sodium carbonate + 1.0mmol / L sodium bicarbonate, flow rate 0.7ml / min). Each sample was measured in triplicate, and the average value was taken. Samples with a confirmed sulfur content of 0mg / kg were assigned to the non-sulfur-fumigated group, and those ≥50mg / kg were assigned to the sulfur-fumigated group.

[0109] Spectral data were correlated and segmented. Spectra were collected from 50 standard samples (30 unsulfurized and 20 sulfurized) to establish a database. 35 samples were used as the training set (21 unsulfurized and 14 sulfurized), and 15 samples were used as the test set (9 unsulfurized and 6 sulfurized). The database contains spectral data (absorbance values ​​across the entire wavelength range of 800-2500 nm), sulfur fumigation status labels (0 = unsulfurized, 1 = sulfurized), preparation date, and other information for each sample, ensuring data traceability.

[0110] A qualitative discrimination model for sulfur fumigation was established: spectral data preprocessing was performed. First, baseline correction was performed on the original spectra, and polynomial fitting (order 5) was used to eliminate baseline shift caused by instrument drift. Then, Savitzky-Golay convolution smoothing (window width 15, polynomial order 4) was used to reduce random noise and improve the signal-to-noise ratio. Finally, standard normal variable transformation (SNV) was used to correct the scattering effect caused by uneven particle size and distribution in the samples to ensure the consistency of the spectral data.

[0111] Model construction and optimization: A binary classification model was built using the Support Vector Machine (SVM) algorithm, with the Radial Basis Function (RBF) kernel function selected. Model parameters were optimized using a grid search method, with the penalty parameter C set to a search range of 1-100 (step size 10) and the kernel function parameter γ set to 0.001-1 (step size 0.1). Five-fold cross-validation was employed, dividing the training set into five subsets. Four subsets were used for training the model, and one subset was used for validation each time, calculating the average classification accuracy. After optimization, the optimal parameters were C=50 and γ=0.01. At this setting, the training set accuracy reached 97.1% (34 / 35), the sulfur-fumigated group precision reached 96.7% (13 / 14), and the non-sulfur-fumigated group recall reached 100% (21 / 21).

[0112] Model performance validation: External validation was performed using 15 samples from the test set. The results showed that 14 samples were correctly identified, and only 1 sulfur-fumigated sample (sulfur content 50 mg / kg) was misidentified as unfumigated, with an overall accuracy of 93.3%. Further analysis revealed that the misidentified samples had insignificant differences in spectral characteristics because their sulfur content was close to the detection limit. The model can be further optimized by increasing the number of low-concentration samples.

[0113] Sample Testing and Result Verification: For actual sample testing, one commercially available sample of suspected sulfur-fumigated honeysuckle was taken. The preprocessing and spectral acquisition steps described above were followed to obtain spectral data. After baseline correction, SNV transformation, and smoothing, the data was input into the optimized SVM discriminant model. The model output a "sulfur-fumigated" determination with a probability of 92%.

[0114] To confirm the repeatability of the experiment, the sulfur content of the sample was determined by ion chromatography, and the result was 185 mg / kg, consistent with the model discrimination result. To verify the repeatability of the method, the same sample was tested five times consecutively, with each test 30 minutes apart. The discrimination result was "sulfur fumigation" for all tests, with a relative standard deviation (RSD) of 0, indicating good repeatability.

[0115] Interlaboratory comparison was conducted by sending sample spectral data and model files to three different laboratories, which used the same spectrometer (Nicoleti S50) and the same pretreatment method for detection. All three laboratories identified the sample as "sulfur fumigation," consistent with the results from our laboratory, demonstrating that the method has good cross-laboratory applicability.

[0116] Advantages and Application Scenarios: Compared with traditional sensory identification methods, this method uses quantitative analysis of spectral characteristics, avoiding the subjectivity of human experience. The accuracy rate for identifying low-sulfur fumigated samples (50 mg / kg) reaches 83.3% (correctly identifying 5 / 6 of the test set samples), significantly higher than the 60%-80% accuracy rate of sensory identification. Compared with destructive detection methods such as ion chromatography, this method requires no chemical pretreatment, and the detection time is shortened from 4-6 hours to 5 minutes, making it suitable for rapid on-site screening in the circulation of Chinese medicinal materials.

[0117] In practical applications, qualitative discrimination models and quantitative analysis models can be integrated into portable near-infrared spectroscopy devices to form a detection process of "qualitative screening - quantitative confirmation": First, the SVM model is used to quickly determine whether a sample has been sulfur-fumigated. For positive samples, the PCA-PLS model is used to further determine the sulfur content, improving detection efficiency. This method is particularly suitable for Chinese medicinal materials such as honeysuckle that are easily bleached by sulfur fumigation, providing quality supervision departments and production enterprises with an efficient and non-destructive detection tool.

[0118] This embodiment establishes a qualitative method for identifying sulfur fumigation in honeysuckle based on support vector machines. Through rigorous sample pretreatment, spectral acquisition, and model optimization, it achieves rapid and accurate determination of whether sulfur fumigation has occurred. The model achieves an accuracy of 93.3% on the test set, with good repeatability and inter-laboratory consistency, filling a gap in rapid qualitative identification using traditional detection methods. This method can be extended to sulfur fumigation screening of other Chinese medicinal materials, and combined with quantitative analysis to form a complete quality testing system, demonstrating significant practical application value.

[0119] Comparative experiment with other detection methods: To more intuitively demonstrate the advantages of the method of this invention, a comparative experiment was conducted between the rapid determination method of sulfur-fumigated Chinese medicinal materials of this invention and traditional acid-base titration and ion chromatography methods. Ten different Chinese medicinal material samples were selected (including Astragalus membranaceus, Codonopsis pilosula, Angelica sinensis, Lycium barbarum, Glycyrrhiza uralensis, Lonicera japonica, Poria cocos, Fritillaria cirrhosa, Ophiopogon japonicus, and Dioscorea opposita), and five samples with different sulfur content levels were prepared for each sample, for a total of 50 samples.

[0120] Using the method of this invention, the average detection time per sample is 4.2 minutes, the detection limit is as low as 0.1 mg / kg, and the relative standard deviation (RSD) of the detection results is between 1.2% and 3.5%. In contrast, the average detection time per sample using acid-base titration is as long as 5.5 hours, the detection limit is 5 mg / kg, and the RSD is between 5.6% and 8.2%; the average detection time per sample using ion chromatography is 3.8 hours, the detection limit is 1 mg / kg, and the RSD is between 3.8% and 6.1%.

[0121] The comparative results clearly demonstrate that the method of this invention has significant advantages in terms of detection speed, detection limit, and precision of detection results. In practical testing applications, it can greatly improve detection efficiency, reduce detection costs, and simultaneously ensure the accuracy and reliability of detection results, providing a more efficient and accurate detection method for the quality supervision of Chinese medicinal materials.

[0122] The rapid method for determining sulfur-fumigated Chinese medicinal materials of this invention can not only be used to determine the sulfur content in Chinese medicinal materials, but also has expanded applications. It combines pattern recognition technology to identify the origin of sulfur-fumigated Chinese medicinal materials. By collecting samples of sulfur-fumigated Chinese medicinal materials from different origins, an origin-related spectral database is established. Using pattern recognition algorithms such as Support Vector Machine (SVM) and Artificial Neural Network (ANN), an origin identification model is constructed. When the spectral data of the sulfur-fumigated Chinese medicinal material to be tested is input, the model can quickly determine the origin of the sample, providing more information for the traceability and quality assessment of Chinese medicinal materials.

[0123] Furthermore, this method can also be used to predict the storage time of sulfur-fumigated Chinese medicinal materials. The internal chemical composition of sulfur-fumigated Chinese medicinal materials changes with different storage times, and these changes are reflected in spectral data. By collecting spectral data of sulfur-fumigated Chinese medicinal materials at different storage times and combining this with chemical analysis methods to determine the content of relevant chemical components, a relationship model between storage time, spectral data, and chemical components can be established. This allows for rapid prediction of the storage time of sulfur-fumigated Chinese medicinal materials, providing a scientific basis for the storage and quality control of Chinese medicinal materials.

[0124] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for rapid determination of sulfur-fumigated Chinese medicinal materials, characterized in that, Includes the following steps: S1. Sample pretreatment: Remove impurities from the Chinese medicinal material sample and pulverize it to 20-40 mesh, then pass it through a 20-mesh standard sieve; dry the pulverized sample at 40℃ to constant weight, and then place it in a desiccator to cool for 30 minutes to eliminate moisture interference and ensure that the sample is in a consistent state. S2. Spectral acquisition: Near-infrared spectrometer is used to acquire spectral data of the sample in the wavelength range of 800-2500nm using an integrating sphere diffuse reflectance method. The number of scans is 32-64 times, and the resolution is 4-8cm⁻¹. S3. Establish a standard sample spectral database: Select at least 10 common Chinese medicinal materials, and prepare standard samples with at least 20 concentration gradients for each Chinese medicinal material with sulfur content ranging from 0 to 1000 mg / kg; S4. Establish a quantitative analysis model: Analyze the standard sample spectral database using principal component analysis-partial least squares method, and determine the optimal number of principal components using leave-one-out cross-validation method, so that the root mean square error of the model prediction is less than 5% and the correlation coefficient is greater than 0.

95. S5. Sample Detection: The samples of Chinese medicinal materials to be tested are preprocessed and spectral data are collected. The spectral data are baseline corrected and smoothed using the Savitzky-Golay convolution smoothing method with a window width of 9-15 and a polynomial order of 2-4 to eliminate baseline drift and random noise in the spectral data.

2. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: In the sample pretreatment step, for samples with high sugar content or easy agglomeration, an appropriate amount of anhydrous ethanol is added to assist in pulverization, and the ethanol is evaporated by ventilation after pulverization; for hard samples, low-temperature freeze pulverization is used to ensure that the particle size is uniform and passes through a 20-mesh sieve.

3. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: During the spectral acquisition process, the ambient temperature and humidity are monitored in real time. The ambient temperature is controlled at 20-25℃ and the relative humidity is controlled at 40%-60%. Temperature and humidity correction factors are introduced into the model to eliminate the influence of environmental factors on the spectral data.

4. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: In the step of establishing a standard sample spectral database, the sulfur content of each standard sample is determined by ion chromatography, and its spectral data are collected simultaneously to establish a standard sample spectral database containing spectral data and corresponding sulfur content.

5. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: In the sample testing step, the preprocessed spectral data is input into the quantitative analysis model to obtain the sulfur content of the sample; and a quality control system for the test results is established, and the model is periodically validated using standard substances. When the relative error of the test results exceeds 10%, the model is updated and optimized.

6. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: After acquiring spectral data, the near-infrared spectrometer uses compressed sensing technology to compress the spectral data, reducing the data volume to 1 / 10 to 1 / 5 of the original data. This reduces the data storage and transmission pressure and improves data processing efficiency without losing spectral information.

7. The method for rapid determination of sulfur-fumigated Chinese medicinal materials according to claim 1, characterized in that: It also includes establishing a qualitative discrimination model for sulfur-fumigated Chinese medicinal materials, using the support vector machine algorithm and standard sample spectral data as the training set to establish a discrimination model based on spectral features to determine whether or not the materials have been fumigated with sulfur.

Images