Method for identifying Maotai-flavor liquor based on surface enhanced Raman spectroscopy

Abstract

The invention relates to a Maotai-flavor liquor identification method based on surface enhanced Raman spectroscopy, and belongs to the technical field of liquor analysis and detection, the method comprises the following steps: S1, sample pretreatment: mixing a Maotai-flavor liquor sample with a gold nanostar solution, adding a potassium chloride coagulant reagent, and fully and uniformly mixing to obtain a to-be-detected sample solution; s2, Raman spectrum collection: placing the to-be-detected sample solution on a Raman spectrometer sample table for Raman spectrum collection; s3, preprocessing spectral data: carrying out background deduction and smoothing processing on the collected original spectrum, identifying a quantitative peak and calculating a peak height integral value, taking an average value measured for multiple times as final spectral data, and selecting a C-C-O stretching vibration peak of ethanol as an internal standard peak to carry out normalization processing on all spectrums; and S4, inputting the extracted Raman spectrum characteristic data into the established decision-making tree model to realize identification of the producing area, the source and the quality grade of the Maotai-flavor liquor.

Description

Technical Field

[0001] This invention relates to the field of liquor analysis and detection technology, and in particular to a method for identifying sauce-flavored liquor based on surface-enhanced Raman spectroscopy. Background Technology

[0002] The inconsistent quality of Maotai-flavor liquor products poses a significant challenge to its quality control. Therefore, identifying Maotai-flavor liquor from different origins is crucial for product quality control, origin traceability, and market supervision. Currently, the main methods for identifying liquor quality include sensory evaluation, gas / liquid chromatography-mass spectrometry (GC / MS), electronic nose / tongue, and spectrometry. Sensory evaluation, as the primary method for evaluating liquor quality, is fast, convenient, and real-time; however, it requires a large number of evaluators and is easily affected by the evaluation environment and subjective biases. GC / MS and other GC-MS methods can obtain specific component information of liquor samples, but these instruments are expensive, complex to operate, and often require sample pretreatment, which is time-consuming and labor-intensive, and the analytical results are delayed, limiting their ability to monitor liquor quality in real time. Electronic nose / tongue, infrared spectroscopy, and fluorescence spectroscopy are simple and fast to operate, but these methods have low detection resolution, making it difficult to effectively distinguish characteristic component information.

[0003] In recent years, Raman spectroscopy has become more suitable for analyzing aqueous solutions due to its advantages such as no need for complex pretreatment, fast detection speed, and absence of interference from water or CO2 gases, compared to the more commonly used infrared spectroscopy. For example, CN108896527A discloses a method for rapidly identifying genuine and counterfeit baijiu (Chinese liquor) using principal component analysis of Raman spectroscopy. This method includes the following steps: collecting Raman spectral data of genuine and counterfeit baijiu and dividing them into training and test sets; representing the raw Raman spectral data of the collected samples using a matrix, importing the raw data of each sample into each row of this matrix, where each column represents the Raman scattering intensity of each baijiu sample in the training set at a certain wavenumber; extracting multiple principal component data from the training set data using principal component analysis; performing normalization and then linear regression to fit a linear classifier; importing the obtained data into the linear classifier to obtain the predicted value Z for the baijiu sample in the test set; setting a classifier threshold θ, where the classifier classifies the sample as genuine when Z>θ and as counterfeit when Z<θ.

[0004] However, conventional Raman scattering signals are weak, especially in the presence of a fluorescence background, where the Raman signal is easily masked by the fluorescence signal. Figure 1 The average Raman spectrum of 61 types of Maotai-flavor liquor from 21 domestic manufacturers is shown. Figure 1As can be seen, the Raman spectral trends of Maotai-flavor liquors from different manufacturers are basically the same, with similar peak shapes and heights. The main components of liquor are ethanol and water, with trace components (acids, esters, alcohols, aldehydes, ketones, etc.) produced during brewing accounting for only 1% to 2%. Raman spectral intensity is directly proportional to the concentration of substances, therefore, trace components are difficult to observe directly from the spectral signal. The noise background of the Raman spectra of Maotai-flavor liquors from different manufacturers also varies. Noise interference can lead to inaccurate subsequent analyses; therefore, preprocessing the Raman spectral data, including smoothing and denoising, and normalization, to remove signal contributions irrelevant to the sample, can improve the accuracy of the experimental analysis. Figure 2 The image shows the Raman spectrum after preprocessing. The preprocessed Raman spectrum is more intuitive, but it is still impossible to directly distinguish between different manufacturers' Maotai-flavor liquors with the naked eye.

[0005] Therefore, there is currently a lack of a simple, fast, and effective method for visual online detection and identification of Maotai-flavor liquor from different production areas in China.

[0006] Furthermore, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the applicant studied a large number of documents and patents when making this invention, but due to space limitations, not all details and contents were listed in detail. However, this does not mean that the present invention does not possess the features of these prior art. On the contrary, the present invention already possesses all the features of the prior art, and the applicant reserves the right to add relevant prior art to the background art. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method for identifying Maotai-flavor liquor based on surface-enhanced Raman spectroscopy, which can quickly determine the origin and quality of Maotai-flavor liquor, thereby solving at least some of the above-mentioned technical problems.

[0008] This invention discloses a method for identifying Maotai-flavor liquor based on surface-enhanced Raman spectroscopy, which includes the following steps: S1. Sample pretreatment: Take a sample of Maotai-flavor liquor and mix it with gold nanostar solution, then add coagulation accelerator and mix thoroughly to obtain the sample solution to be tested. S2. Raman spectroscopy acquisition: Place the sample solution to be tested on the sample stage of the Raman spectrometer for Raman spectroscopy acquisition; S3. Spectral data preprocessing: Background subtraction and smoothing are performed on the acquired raw spectra, quantitative peaks are identified and peak height integral values ​​are calculated, the average value of multiple measurements is taken as the final spectral data, and the CCO stretching vibration peak of ethanol is selected as the internal standard peak to normalize all spectra. S4. Input the extracted Raman spectral feature data into the established decision tree model to identify the origin and quality grade of Maotai-flavor liquor.

[0009] This invention significantly enhances the Raman signal intensity of trace components in Maotai-flavor liquor by using gold nanostars to reinforce the substrate, effectively overcoming the technical bottleneck of conventional Raman spectroscopy, which struggles to distinguish liquor components due to weak signals. Specifically, the pointed structure on the surface of the gold nanostars generates a strong localized surface plasmon resonance effect, resulting in a superior SERS signal enhancement compared to gold nanospheres. Simultaneously, the addition of potassium chloride as a coagulation promoter adjusts the ionic strength of the solution, promoting the adsorption of analyte molecules on the surface of the gold nanostars and improving signal stability and reproducibility. Furthermore, the CCO stretching vibration peak of ethanol (880 cm⁻¹) is utilized... -1 The sample concentration difference and the fluctuation of detection conditions were used as internal standards for normalization, which eliminated the influence of sample concentration differences and detection conditions on spectral intensity. This provided high-precision feature data for subsequent decision tree models, thereby enabling accurate identification of the origin and quality grade of Maotai-flavor liquor.

[0010] According to a preferred embodiment, in step S4, the model is established through the following steps: S4.1 Establishing an Orthogonal Partial Least Squares-Discriminant Analysis (OPLS-DA) Model: Import the preprocessed Raman spectral data into chemometrics software, establish an OPLS-DA model based on different origins and quality classification standards, and screen key Raman shift points with VIP>1 by variable importance projection values; S4.2 Constructing a discrimination model: Randomly divide the selected key Raman displacement point data into training and test sets, establish a classification prediction model, optimize the model parameters through the training set, and verify the discrimination accuracy of the model using the test set.

[0011] This invention effectively solves the problems of high dimensionality and multicollinearity among variables in SERS spectral data of baijiu (Chinese liquor) by screening key Raman shift points with VIP>1 using the OPLS-DA model. This model quantifies and evaluates feature variables based on variable importance projection values, selecting the feature sites that contribute most to the identification of baijiu origin and quality, thus avoiding overfitting or underfitting caused by improper feature selection in traditional methods. The selected key shift points (e.g., 1750 cm⁻¹) are used to identify key Raman shift points. -1 C=O vibration in nearby esters or amides, 1600 cm -1 The chemical structure of components in baijiu (Chinese liquor) is closely related to aromatic C=C vibrations, making the model construction more consistent with the chemical essence of baijiu quality evaluation. This feature importance-based screening method significantly reduces model complexity while improving classification accuracy, providing scientific and reliable feature inputs for subsequent construction of decision tree identification models.

[0012] According to a preferred embodiment, in step S4.1, during the establishment of the OPLS-DA model, the preprocessed Raman spectral data are used to establish corresponding category labels according to geographical origin, river basin location, or market price level standards. The calculation of the variable importance projection value is based on the contribution weight of each variable in the model to the classification result, and the spectral shift points with variable importance projection values ​​greater than 1 are used as key feature variables.

[0013] This invention establishes category labels for preprocessed Raman spectral data according to actual standards for evaluating the quality of baijiu (Chinese liquor) (geographical origin, river basin location, and market price level), making the model construction more closely aligned with the actual evaluation system of the baijiu industry. By using Raman shift points with VIP>1 as key feature variables, this invention accurately captures the chemical essence of baijiu origin and quality differences, avoiding model bias caused by random feature selection. This feature selection method based on actual classification standards in the baijiu industry enables the established identification model to accurately reflect the regional characteristics and quality levels of baijiu, providing a scientific basis that conforms to industry standards for baijiu origin traceability and quality grading, and avoiding identification bias caused by mismatched classification standards in traditional methods.

[0014] According to a preferred embodiment, among the key Raman displacement points screened by the OPLS-DA model, 1600 cm -1 The nearby Raman shift is related to the C=C skeleton vibration in aromatic compounds, 1280 cm⁻¹ -1 The nearby Raman shift corresponds to the CO absorption peak of ester or ether molecules, at 2200 cm⁻¹. -1 The nearby Raman shifts are related to the stretching vibrations of C≡C and C≡N in alkynes; among the key Raman shift points screened by the OPLS-DA model, 1750 cm⁻¹ is... -1 The nearby Raman shifts are related to the C=O bonds in esters or amides; among the key Raman shift points screened by the OPLS-DA model, 1100 cm⁻¹ is... -1 The nearby Raman shift is a characteristic peak of the CO stretching vibration in alcohol molecules or the C-C bond stretching vibration in carbon chain compounds.

[0015] This invention, through the correspondence between key Raman shift points screened using the OPLS-DA model and specific chemical structures, provides a clear chemical explanation for the analysis of baijiu (Chinese liquor) components. (1600 cm) -1 The nearby Raman shifts are related to the C=C skeleton vibrations of aromatic compounds, reflecting the differences in the content of pyrazine aroma components in baijiu. Pyrazines are key compounds that distinguish soy sauce-flavored baijiu from Guizhou, Sichuan, and Hunan. (1280 cm) -1 The nearby Raman shifts correspond to CO absorption peaks of ester or ether molecules, indicating the content of ester or ether compounds; 2200 cm⁻¹ -1The nearby Raman shifts are correlated with the C≡C and C≡N stretching vibrations of alkynes, reflecting the content of specific nitrogen-containing compounds. The chemical correlation between these characteristic peaks and the components of baijiu (Chinese liquor) gives the model identification results clear chemical significance, providing a bridge from spectral characteristics to chemical composition for baijiu quality analysis, and enhancing the scientific rigor and interpretability of the identification results.

[0016] According to a preferred embodiment, the identification model constructed in step S4.2 is a decision tree model.

[0017] This invention employs a decision tree model as the identification model. Compared to SVM and KNN models, it demonstrates 100% accuracy in distinguishing between different origins and qualities of Maotai-flavor liquor, significantly improving identification precision. The decision tree model is characterized by its clear structure and ease of interpretation. Each internal node represents a judgment condition for a spectral feature variable, and the leaf nodes correspond to the final classification label, making the identification results of liquor origin and quality clearly interpretable. This model selection not only avoids the overfitting risk associated with complex models but also provides intuitive classification rules, making the identification process both efficient and transparent. It offers a precise and understandable solution for liquor quality identification, making it particularly suitable for rapid testing needs in actual production environments.

[0018] According to a preferred embodiment, in step S3, the acquired raw Raman spectral data undergoes background subtraction processing. Background subtraction methods include polynomial fitting, iterative approximation algorithms, and wavelet transform. Subsequently, the spectral data is smoothed using moving average, Savitzky-Golay filter, or multi-point smoothing algorithm. Then, characteristic peaks in the spectrum are identified and quantitatively analyzed, calculating the peak height, peak area, or peak intensity integral value of each peak. Finally, the spectral data is normalized, selecting the ethanol molecule at 880 cm⁻¹. -1 The nearby CCO stretching vibration peak is used as an internal standard peak. The intensity value of this peak in the spectrum of each sample is set as a unified standard value, thereby achieving the standardization of the intensity of the entire spectrum.

[0019] This invention employs polynomial fitting, iterative approximation algorithms, and wavelet transform for background subtraction, effectively removing fluorescence background and Rayleigh scattering interference from the SERS spectrum of baijiu (Chinese liquor) while preserving authentic Raman scattering information. Smoothing is achieved using a Savitzky-Golay filter, effectively suppressing high-frequency noise while retaining the shape and position information of spectral characteristic peaks. (At 880 cm⁻¹) -1The CCO stretching vibration peak of ethanol was normalized using an internal standard, eliminating the influence of sample concentration differences and fluctuations in detection conditions on spectral intensity and making the spectral data of different samples comparable. These data processing steps significantly improved the spectral signal-to-noise ratio and data quality, providing high-quality and reliable input data for subsequent feature extraction and model construction, and avoiding the negative impact of noise and background interference on the identification results.

[0020] According to a preferred embodiment, in step S2, the pretreated Maotai-flavor liquor sample is placed at the detection position of a Raman spectrometer. The sample is excited by laser to generate a Raman scattering signal, and the scattered light is collected and analyzed by the spectrometer's detection system to obtain the characteristic Raman spectral information of the sample.

[0021] This invention employs a 532 nm laser excitation source, effectively avoiding common fluorescence background interference in baijiu (Chinese liquor) while ensuring sufficient signal intensity. By setting a 10x objective lens, good spatial resolution and signal collection efficiency are achieved, ensuring the representativeness of the detection area. An integration time of 6 seconds and a cumulative count of 5 balance signal intensity and detection efficiency, avoiding reduced throughput due to excessively long integration times. These optimized parameters result in Raman spectra with high signal-to-noise ratio and good reproducibility, providing high-quality spectral data for the characteristic identification of baijiu components and solving the technical problem of signal masking caused by fluorescence interference in conventional Raman spectroscopy for baijiu detection.

[0022] According to a preferred embodiment, in step S1, the sample of Maotai-flavor liquor to be tested is mixed with the prepared gold nanostar-enhanced substrate. The two are brought into full contact by stirring or oscillation. The stirring intensity and time are controlled during the mixing process to ensure that the components to be tested in the liquor can be effectively adsorbed onto the surface of the gold nanostar, thereby obtaining a test sample with surface-enhanced Raman scattering effect. An electrolyte solution, which is a potassium chloride solution, is added to the mixing system as a coagulation accelerator.

[0023] This invention utilizes the synergistic effect of gold nanostars as an enhanced substrate and potassium chloride as a coagulation accelerator to solve the technical challenge of weak and unstable signals in SERS detection of baijiu (Chinese liquor). The needle-like structure on the surface of the gold nanostars generates a strong tip enhancement effect, significantly improving the Raman signal enhancement. Potassium chloride, as an electrolyte coagulation accelerator, promotes the adsorption of analyte molecules on the gold nanostar surface by adjusting the ionic strength of the solution, while simultaneously stabilizing the SERS signal. This synergistic effect avoids the destruction of the gold nanostar structure caused by vigorous stirring, ensuring the stability and reproducibility of the detection. By optimizing the mixing ratio and the amount of coagulation accelerator added, this invention achieves effective contact between the baijiu sample and the gold nanostars, providing a high-quality SERS signal basis for the identification of baijiu origin and quality.

[0024] According to a preferred embodiment, the following steps may be performed before step S1: S0. Preparation of gold nanostar-reinforced substrate: After heating the chloroauric acid solution to boiling, add trisodium citrate solution to react and generate gold nanospheres. The gold nanosphere precipitate is obtained by centrifugation. The prepared gold nanospheres are dispersed in an organic solvent and stirred to mix the components thoroughly to generate a gold nanostar solution. The gold nanostar-reinforced substrate is then obtained by centrifugation and washing with an alcohol solvent.

[0025] This invention successfully synthesizes gold nanostars with a tip-enhancing effect by reacting chloroauric acid solution with trisodium citrate to prepare gold nanospheres as a precursor, followed by anisotropic growth of these nanospheres in a DMF solution containing PVP and HAuCl4. This preparation method precisely controls the morphology and size of the gold nanostars by adjusting the reaction temperature, time, stirring conditions, and centrifugation parameters, resulting in a diameter of approximately 50 nm and multiple needle-like structures on the surface, effectively enhancing the SERS signal. The tip structure of the gold nanostars generates a strong local electric field enhancement, significantly improving the Raman signal enhancement effect for trace components in baijiu (Chinese liquor). This preparation method avoids the problem of insufficient signal enhancement in baijiu detection using traditional SERS substrates, providing a highly sensitive surface enhancement platform for baijiu SERS detection.

[0026] According to a preferred embodiment, the organic solvent is N,N-dimethylformamide, and a protective agent, polyvinylpyrrolidone, and a metal precursor, chloroauric acid, are added to react and generate a gold nanostar solution. After the reaction is completed, the product is collected by centrifugation, with the centrifugation speed being higher than the separation speed of the gold nanospheres, and the alcohol solvent is methanol.

[0027] This invention uses N,N-dimethylformamide (DMF) as an organic solvent, whose polar aprotic properties are beneficial for the anisotropic growth of gold nanostars. Polyvinylpyrrolidone (PVP), with a molecular weight of approximately 20,000, is added as a protective agent. Through selective adsorption on different crystal faces of the gold nanoparticles, the crystal growth direction is controlled, forming a star-shaped structure with a tip-reinforcement effect. Washing with methanol effectively removes unreacted reagents and byproducts, ensuring the purity and stability of the gold nanostar-reinforced substrate. The optimization of these specific parameters results in the prepared gold nanostars exhibiting optimal SERS performance. Their UV-Vis spectra show a significant plasmon absorption peak near 600 nm, indicating the formation of the tip structure. This provides a high-quality surface-reinforced substrate for baijiu (Chinese liquor) detection, effectively solving the technical bottleneck of insufficient substrate performance in baijiu SERS detection. Attached Figure Description

[0028] Figure 1The average Raman spectrum of 61 types of Maotai-flavor liquor from 21 domestic manufacturers; Figure 2 The average Raman spectra of 21 domestic manufacturers' Maotai-flavor liquor after pretreatment; Figure 3 The flowchart of the method for identifying Maotai-flavor liquor provided by the present invention; Figure 4 Transmission electron microscopy images of gold nanospheres and gold nanostars; Figure 5 The UV-Vis spectra of gold nanospheres and gold nanostar sols are shown. Figure 6 The following are Raman spectra of (A) gold nanospheres and gold nanostars sol, (B) SERS Raman spectra of GL-1, (C) average Raman spectra of Sichuan, Guizhou and Hunan sauce-flavored liquors, and (D) average SERS spectra of Sichuan, Guizhou and Hunan sauce-flavored liquors. Figure 7 SERS data plots for Maotai-flavor liquor from different geographical production areas, analyzed using the OPLS-DA model; Figure 8 SERS data plots for Maotai-flavor liquor from different river basins, analyzed using the OPLS-DA model; Figure 9 SERS data plots for Maotai-flavor liquor at different market price levels, analyzed using the OPLS-DA model. Detailed Implementation

[0029] The following is a detailed explanation with reference to the accompanying drawings.

[0030] like Figure 3 As shown, this invention discloses a method for identifying Maotai-flavor liquor based on surface-enhanced Raman spectroscopy, which includes the following steps: S0. Preparation of gold nanostar-reinforced substrate: After heating the chloroauric acid solution to boiling, add trisodium citrate solution to react and generate gold nanospheres. The gold nanosphere precipitate is obtained by centrifugation. The prepared gold nanospheres are dispersed in an organic solvent and stirred to mix the components thoroughly and evenly to react and generate gold nanostar solution. The gold nanostar-reinforced substrate is then obtained by centrifugation and washing with an alcohol solvent. S1. Sample pretreatment: Take a sample of Maotai-flavor liquor and mix it with gold nanostar solution, then add coagulation accelerator and mix thoroughly to obtain the sample solution to be tested. S2. Raman spectroscopy acquisition: Place the sample solution to be tested on the sample stage of the Raman spectrometer for Raman spectroscopy acquisition; S3. Spectral data preprocessing: Background subtraction and smoothing are performed on the acquired raw spectra, quantitative peaks are identified and peak height integral values ​​are calculated, the average value of multiple measurements is taken as the final spectral data, and the CCO stretching vibration peak of ethanol is selected as the internal standard peak to normalize all spectra. S4. Input the extracted Raman spectral feature data into the established decision tree model to identify the origin and quality grade of Maotai-flavor liquor.

[0031] Preferably, in step S0, gold nanospheres are first prepared as precursors for gold nanostars. This process involves reacting an aqueous solution containing chloroauric acid (HAuCl4) with a reducing agent, trisodium citrate, under heating conditions. By controlling the reaction temperature, reaction time, and stirring conditions, the gold ions in the chloroauric acid are reduced to metallic gold atoms, which then gradually aggregate to form spherical nanoparticles. In a specific embodiment, 100 mL of chloroauric acid solution (0.5 mM) is heated to boiling, and then 5 mL of a 1% (w / w) trisodium citrate solution is added as a reducing agent. Continuous stirring ensures thorough mixing of the reactants, and the reaction time is controlled at 10-20 minutes (preferably 15 minutes) until the solution color changes from pale yellow to wine red, indicating that the formation reaction of the gold nanospheres is complete. After the reaction is complete, the product can be separated and purified by high-speed centrifugation to separate the gold nanospheres from the reaction mixture. The centrifugation parameters can be adjusted according to the particle size characteristics of the gold nanospheres. For example, centrifugation at 4000 rpm for 60 to 120 minutes (preferably 90 minutes) is used. The precipitate obtained is the gold nanospheres (AuNPs).

[0032] Furthermore, based on the obtained gold nanospheres, a gold nanostar-reinforced substrate with a star-shaped structure is further prepared. This process requires dispersing the gold nanospheres as seed particles in an organic solvent system containing a protective agent and a metal precursor for an anisotropic growth reaction. Specifically, the separated gold nanospheres are redispersed in an organic solvent, which can be a polar aprotic solvent such as N,N-dimethylformamide (DMF), while adding the protective agent polyvinylpyrrolidone (PVP) and the metal precursor chloroauric acid (HAuCl4). The components are thoroughly mixed by stirring. In a specific embodiment, 200 μL of the previously prepared AuNPs can be added to a DMF solution (30 mL) containing PVP (5 mM) and HAuCl4 (0.3 mM). The protective agent polyvinylpyrrolidone typically has a molecular weight of around 20,000, and its function is to selectively adsorb onto different crystal faces of the gold nanoparticles to control the crystal growth direction, thereby forming a star-shaped structure with sharp protrusions. During the stirring reaction, the color of the reaction system gradually changes from wine red to deep blue, indicating the formation of gold nanostar structures. After the reaction is complete, the product needs to be collected by centrifugation at a speed higher than that of the gold nanospheres, for example, centrifuging at 8000 rpm for 5 to 15 minutes (preferably 10 minutes). After separation, the product is washed with an alcohol solvent such as methanol to remove unreacted reagents and byproducts, ultimately obtaining a gold nanostar-reinforced substrate with a tip-enhancing effect. The temperature control, stirring speed, reaction time, and other parameters in the above preparation process can be adjusted according to actual needs to obtain gold nanostar materials with optimal surface-enhanced Raman scattering properties.

[0033] The morphology of gold nanospheres and gold nanostars was observed using transmission electron microscopy, such as... Figure 4 As shown. From Figure 4 (A) and Figure 4 (B) It can be seen that the synthesized gold nanospheres are relatively uniform in size, with a diameter of about 20 nm. Figure 4 (C) and Figure 4 (D) is a gold nanostar with a diameter of approximately 50 nm and multiple needle-like structures on its surface. The small tip size of the spikes on the surface of the gold nanostar enables a good surface-enhanced Raman effect. The UV-Vis spectra of the gold nanospheres and gold nanostar sols show that the gold nanospheres and gold nanostars exhibit plasmon absorption peaks near wavelengths of 520 nm and 600 nm, respectively, indicating the presence of plasmon absorption peaks. Figure 5 As shown.

[0034] Preferably, in step S1, the sample of Maotai-flavor liquor to be tested can be thoroughly mixed and contacted with the prepared gold nanostar-enhanced substrate to ensure that the analyte in the liquor can effectively adsorb onto the surface of the gold nanostar, thereby obtaining a sample with surface-enhanced Raman scattering effect. This process first involves mixing the Maotai-flavor liquor sample with the gold nanostar solution in a predetermined ratio, ensuring thorough contact through gentle stirring or shaking. The stirring intensity and time should be controlled during mixing to avoid damage or aggregation of the gold nanostar structure due to vigorous stirring. To further improve the stability and reproducibility of the detection, an appropriate amount of electrolyte solution can be added to the mixture as a coagulation promoter. The mechanism of this coagulation promoter is to promote the adsorption of analyte molecules on the surface of the gold nanostar by adjusting the ionic strength of the solution, while also helping to enhance the stability of the Raman signal. In a specific embodiment, potassium chloride solution can be selected as the coagulation promoter, with a concentration range controlled between 0.01 and 1 mol / L, for example, a 0.1 mol / L potassium chloride solution.

[0035] Preferably, the volume ratio of each component during sample mixing has a significant impact on the final detection results and can be optimized based on the concentration of gold nanostars, the characteristics of the baijiu sample, and the expected detection sensitivity. The volume ratio of the baijiu sample to the gold nanostar solution can be controlled within the range of 1:1 to 1:5, for example, a 1:1 ratio. The amount of coagulation accelerator added relative to the total sample volume also needs to be appropriately controlled, accounting for 5% to 30% of the total volume, for example, 10%. The mixing operation can be carried out at room temperature, with the mixing time controlled within the range of 1 to 10 minutes. Uniform mixing can be achieved by gentle vortexing or manual shaking. Environmental conditions such as temperature and light during sample pretreatment should also be controlled to avoid changes in sample composition or degradation of gold nanostar performance due to external interference. In addition, the mixed sample solution should be subjected to Raman spectroscopy as soon as possible to avoid component volatilization or changes in adsorption equilibrium due to prolonged storage. In practice, it is recommended to complete the detection within 30 minutes after mixing to ensure the accuracy of the results. The entire sample pretreatment process should be carried out in a clean environment to avoid interference from foreign impurities. The containers and tools used should be cleaned and kept dry beforehand.

[0036] Preferably, in step S2, the pretreated Maotai-flavor liquor sample can be placed at the detection position of a Raman spectrometer. The sample is excited by a laser to generate a Raman scattering signal, and the scattered light is collected and analyzed using the spectrometer's detection system to obtain the sample's characteristic Raman spectral information. First, the laser excitation source and parameters can be selected. The wavelength selection of the laser plays a crucial role in avoiding interference from the sample's fluorescence background and obtaining a high-quality Raman signal. An appropriate excitation wavelength can be selected based on the sample's optical characteristics. In specific implementation, a visible laser can be used as the excitation source, with a wavelength range between 400 and 700 nm, such as a 532 nm green laser. The laser power setting needs to ensure sufficient signal strength while avoiding thermal damage to the sample. The power density is controlled at the milliwatt level, and precise control is achieved through a power adjustment device.

[0037] Preferably, during spectral acquisition, the laser beam can be focused onto a specific area of ​​the sample using a microscope system. The magnification of the objective lens directly affects the spatial resolution and signal collection efficiency. A range of objectives, from low to high magnification, can be selected; for example, a 10x objective lens provides good signal collection capability while ensuring sufficient detection area. The detection wavenumber range of the spectrometer needs to cover the characteristic Raman shift range of the analyte, and can be set between 400 and 2800 cm⁻¹. -1 To ensure the capture of characteristic vibrational information of various organic components in baijiu, the integration time needs to balance signal strength and detection efficiency. Too short an integration time may lead to insufficient signal-to-noise ratio, while too long an integration time will reduce detection throughput. Optimization can be performed within the range of several seconds to tens of seconds. To improve the reliability and reproducibility of spectral data, multiple cumulative scans can be used to repeatedly detect the same location. Signal averaging can be used to reduce noise. The number of cumulative scans is set to 3-10 times (preferably 5 times). Furthermore, to ensure the accuracy and consistency of the detection results, each sample can be repeatedly measured at different locations, and representative spectral data can be obtained through statistical analysis.

[0038] Preferably, in step S3, a series of mathematical processing and signal optimization operations can be performed on the acquired raw Raman spectral data to eliminate the influence of instrument noise, background interference, and baseline drift on the spectral quality, thereby obtaining accurate and reliable spectral feature information for subsequent qualitative and quantitative analysis. This process can begin with background subtraction. Since fluorescence background, Rayleigh scattering, and other stray light signals present in the sample can interfere with the identification and analysis of Raman characteristic peaks, an appropriate background subtraction algorithm is needed to separate these non-Raman signal components from the original spectrum. Background subtraction methods can include polynomial fitting, iterative approximation algorithms, and wavelet transform techniques. These methods can effectively identify and remove continuous background signals in the spectrum, preserving the true Raman scattering information.

[0039] Preferably, in addition to background subtraction, the spectral data also needs to be smoothed to reduce the impact of random noise and improve the signal-to-noise ratio and visual clarity of the spectrum. Smoothing can employ techniques such as moving averages, Savitzky-Golay filters, or multi-point smoothing algorithms. These methods can effectively suppress high-frequency noise while preserving the shape and position information of spectral characteristic peaks. In specific implementation, appropriate smoothing parameters, such as the smoothing window width and polynomial order, can be selected based on the noise level of the spectrum and the analysis requirements. Subsequently, characteristic peaks in the spectrum can be identified and quantitatively analyzed. Peak detection algorithms can automatically identify the main Raman characteristic peaks in the spectrum and calculate quantitative parameters such as peak height, peak area, or peak intensity integral value for each peak. The calculation of the peak intensity integral value can use methods such as trapezoidal integration or Gaussian fitting integration to obtain more accurate peak intensity information. To eliminate the influence of concentration differences between different samples and fluctuations in detection conditions on spectral intensity, the spectral data also needs to be normalized, that is, the intensity values ​​of all spectra are standardized relative to a selected internal standard peak.

[0040] Preferably, in the normalization process, the selection of the internal standard peak is crucial for ensuring the accuracy of comparisons between different samples. A characteristic peak with stable content and a strong Raman signal in the sample can be selected as the internal standard, such as the CCO stretching vibration peak of ethanol molecules at a specific wavenumber position. In a specific embodiment, 880 cm⁻¹ can be selected. -1 The nearby ethanol characteristic peak is used as an internal standard peak. By setting the intensity value of this peak in the spectra of each sample to a uniform standard value, the intensity of the entire spectrum is standardized. To improve the accuracy and reproducibility of data processing, each sample can be measured repeatedly, and the resulting multiple sets of spectral data can be statistically analyzed, such as calculating statistical parameters like the mean and standard deviation, to obtain more representative spectral characteristic data.

[0041] Figure 6(A) indicates that gold nanospheres and gold nanostar sols themselves do not exhibit Raman absorption, but the gold nanostar sols show Raman absorption at 1500 cm⁻¹. -1 Nearby, there are residual reagent peaks, which are due to the stretching vibrations of -CH3 in DMF and PVP. From... Figure 6 As can be seen in (B), the SERS spectrum of GL-1, a type of soy sauce-flavored liquor, is richer than that of GL-1 Raman spectrum, with many small Raman absorption peaks added. This is mainly due to the enhanced Raman absorption effect of some trace components under the SERS effect. Figure 6 (C) and Figure 6 (D) shows the average Raman spectrum and average Raman enhanced spectrum of Maotai-flavor liquor from the three provinces, respectively. Comparison reveals that the SERS spectra of Maotai-flavor liquor from the three provinces are concentrated in the range of 1500–2500 cm⁻¹. -1 The absorption peak intensity of the Raman shift is enhanced, especially at 1600 cm⁻¹. -1 2100~2300 cm -1 1600 cm -1 This location is related to the C=C skeleton vibration in aromatic compounds, while the stretching vibrations of C≡C and C≡N in alkynes occur at 2200 cm⁻¹. -1 Nearby. Although gold nanomaterials enhanced the Raman signal of Maotai-flavor liquor, it was still impossible to distinguish Maotai-flavor liquor from three provinces by the naked eye. Therefore, further analysis of the SERS spectra of Maotai-flavor liquor samples using chemometrics is needed.

[0042] Preferably, in step S4, for the Maotai-flavor liquor sample to be tested, its extracted Raman spectral characteristic data can be input into the established decision tree model to achieve rapid identification of the origin and quality grade of the Maotai-flavor liquor. Further, in step S4, the model can be established first through the following steps: S4.1 Establishing an Orthogonal Partial Least Squares-Discriminant Analysis (OPLS-DA) Model: Import the preprocessed Raman spectral data into chemometrics software, establish an OPLS-DA model based on different origins and quality classification standards, and screen key Raman shift points with VIP>1 by variable importance projection values; S4.2 Constructing a discrimination model: Randomly divide the selected key Raman displacement point data into training and test sets, establish a classification prediction model, optimize the model parameters through the training set, and verify the discrimination accuracy of the model using the test set.

[0043] Preferably, steps S4.1 and S4.2 can be used to establish a model using known samples of Maotai-flavor liquor.

[0044] Preferably, the preprocessed Raman spectral data of known Maotai-flavor liquor samples can be imported into specialized chemometric analysis software. A classification and identification model can then be constructed using a multivariate statistical analysis method combining orthogonal partial least squares (OPLS) and discriminant analysis (DA). This allows for accurate identification of the origin and quality grade of the Maotai-flavor liquor samples. The modeling process begins with organizing and classifying the spectral data, categorizing all samples' Raman spectral data according to different standards, such as geographical origin, river basin location, or market price level. OPLS-DA, as a supervised learning method, effectively addresses issues like multicollinearity and the curse of dimensionality in spectral data, constructing an optimal classification function by extracting the correlation information between spectral variables and category labels. Optionally, the OPLS-DA model can be built and run using SIMCA 14.1 software.

[0045] Preferably, during model construction, feature selection and importance assessment can be performed on a large number of Raman spectral variables to identify the key spectral shift points that contribute most to classification. This process can employ variable importance projection value analysis, which quantifies the contribution of each spectral variable in the model to screen out feature variables with significant discriminative power. The calculation of variable importance projection values ​​is based on the contribution weight of each variable in the model to the classification results. A certain threshold standard can be set to determine the selected variables; for example, spectral shift points with variable importance projection values ​​greater than a certain critical value can be used as key feature variables. In specific implementations, corresponding screening criteria can be set according to the requirements of different classification tasks. For example, hundreds of key sites can be screened for province classification tasks, while a relatively small number of target key sites may need to be screened for more refined quality grade classification. The selected key spectral shift points can not only significantly reduce model complexity and improve computational efficiency, but also enhance the interpretability and robustness of the model.

[0046] For example, this invention selected 61 samples of Maotai-flavor liquor as known Maotai-flavor liquor samples. These 61 samples can be divided into Sichuan, Guizhou, and Hunan provinces according to their place of origin; into the left bank and right bank of the Chishui River according to their location; and into 16 Maotai-flavor liquors priced between 100 and 500 yuan, 24 Maotai-flavor liquors priced between 500 and 1000 yuan, 10 Maotai-flavor liquors priced between 1000 and 1500 yuan, and 11 Maotai-flavor liquors priced above 1500 yuan. Figure 7 and Figure 8 It can be seen that combining the SERS data of 61 Maotai-flavor liquors with the OPLS-DA model can effectively distinguish Maotai-flavor liquors from different production areas and from the left and right banks of the Chishui River. From Figure 9It can be seen that the OPLS-DA model also has a certain ability to distinguish between different price levels of Maotai-flavor liquor. In the OPLS-DA model, the Raman shift points with VIP>1 are the key Raman shift points for distinguishing Maotai-flavor liquor from different origins and of different qualities. Therefore, Raman shift points with VIP>1 are selected as data for the subsequent establishment of the identification model.

[0047] In the OPLS-DA model of Maotai-flavor liquor from three provinces (i.e., production areas), there were 409 Raman shift points with VIP>1, of which 69 Raman shift points had VIP>1.5. (2100 cm) -1 The largest VIP values ​​were observed at nearby Raman displacement points, which may be related to the C≡C symmetric stretching vibration. The overall organic acid content in Guizhou-origin Maotai-flavor liquor is higher than that in Sichuan and other regions. Octanoic acid, hexanoic acid, and acetic acid are important compounds that distinguish Maotai-flavor liquor from Guizhou, Sichuan, and other regions. The remaining displacement points with VIP > 1.5 are mainly concentrated in the 2200–2300 cm⁻¹ range. -1 1650 cm -1 and 1280 cm -1 Nearby, of which 1600 cm -1 The nearby Raman shift is related to the C=C skeleton vibration in aromatic compounds. Pyrazine is an important aroma component in Maotai-flavor liquor, with the highest pyrazine content (5.36 mg / L) in Maotai-flavor liquor from Guizhou, followed by the central and northern regions; 1280 cm -1 The nearby Raman shifts correspond to the CO absorption peaks of ester or ether molecules; the stretching vibrations of C≡C and C≡N in alkynes appear at 2200 cm⁻¹. -1 Nearby. Geographically, Guizhou avoids the high-altitude air currents and has a unique warm and humid climate, while Sichuan has hot and rainy summers. The differences in geographical environment and climate can lead to differences in brewing raw materials and microbial species, thus causing differences in the composition of baijiu.

[0048] In the OPLS-DA model of soy sauce-flavored baijiu on both banks of the Chishui River, there are 598 Raman shift points with VIP>1, of which 138 Raman shift points have VIP>1.5. (1750 cm) -1 The VIP value is highest at the nearest displacement point, followed by 2100 cm. -1 and 1800cm -1 Nearby, of which 1750 cm -1 Nearby Raman shifts are related to C=O bonds in esters or amides. The levels of ethyl hexanoate, ethyl formate, ethyl 2-methylbutyrate, ethyl heptanoate, and ethyl furoate in wine samples from the right bank of the Chishui River were significantly higher than those from the Renhuai and Xishui production areas on the left bank of the Chishui River. (1800 cm) -1 The nearby Raman shift is related to the vibration of C=O in acid anhydrides or acyl chlorides.

[0049] exist Figure 9 In the model, baijiu priced between 500 and 1000 yuan mainly clustered on the lower half of the Y-axis, baijiu priced between 100 and 500 yuan mainly clustered on the upper half of the Y-axis, and baijiu priced between 1000 and 1500 yuan mainly clustered on the left half of the X-axis. In the OPLS-DA model of different grades of baijiu, there were 418 Raman shift points with VIP > 1, of which 81 Raman shift points had VIP > 1.5. (1100 cm) -1 The largest Raman shift VIP value is found near this point, representing the characteristic peak of CO stretching vibration in alcohol molecules or C / C bond stretching vibration in certain carbon chain compounds. The remaining Raman shift points with VIP > 1.5 are mainly concentrated around 1800 cm⁻¹. -1 1600 cm -1 nearby.

[0050] For example, 90% of the 61 SERS spectra were randomly selected as the training set, and the remaining 10% were used as the test set. The results are shown in Table 1. Comparing the discrimination results of SVM, KNN, and TREE models, it was found that the TREE model had the best effect in distinguishing between different origins and different qualities of Maotai liquor. Therefore, this invention selects the TREE model (i.e., the decision tree model) as the identification model.

[0051] Table 1. Identification results of three identification models for sauce-flavored liquors of different origins and qualities. Furthermore, in step S4.2, when constructing the decision tree identification model, the decision tree algorithm and parameters can be selected and configured first. Considering the complexity and diversity of the liquor sample classification problem, mature decision tree construction methods such as CART, ID3, or C4.5 algorithms can be used. These algorithms can decompose complex multi-classification problems into a series of simple binary classification judgments through recursive partitioning. The core advantage of the decision tree model lies in its intuitive tree structure and easy-to-understand classification rules. Each internal node represents a judgment condition for a spectral feature variable, each branch represents a judgment result, and the leaf nodes correspond to the final classification label.

[0052] In the specific construction process of a decision tree, evaluation criteria and stopping conditions for node splitting can be determined to ensure that the generated decision tree has both good classification performance and avoids overfitting. Node splitting evaluation criteria can include indicators such as information gain, information gain ratio, and Gini impurity. These criteria can quantify the contribution of different feature variables to sample classification, thereby guiding the algorithm to select the optimal splitting variable and split point. The splitting process starts from the root node, selects the optimal splitting variable by calculating the information gain value of all candidate variables, and divides the sample set into two subsets, corresponding to the left and right subtrees, based on the optimal split point of this variable. This process is recursively performed until a preset stopping condition is met, such as the number of samples in a node falling below a certain threshold, the purity of the node reaching a predetermined standard, or the tree depth exceeding a maximum limit. In a specific implementation, the minimum number of samples in a leaf node can be set to 1% to 10% of the total number of samples; for example, a 5% proportion can effectively balance model complexity and classification accuracy. The maximum tree depth can be set in the range of 5 to 20 layers, adjusted and optimized according to the complexity of the actual classification task.

[0053] It should be noted that the specific embodiments described above are exemplary. Those skilled in the art can devise various solutions inspired by the disclosure of this invention, and these solutions all fall within the scope of this invention and its protection. Those skilled in the art should understand that this specification and its accompanying drawings are illustrative and not intended to limit the scope of the claims. The scope of protection of this invention is defined by the claims and their equivalents. This specification contains multiple inventive concepts; terms such as "preferredly," "according to a preferred embodiment," or "optionally" indicate that the corresponding paragraph discloses an independent concept. The applicant reserves the right to file divisional applications based on each inventive concept.

Claims

1. A method for identifying Maotai-flavor liquor based on surface-enhanced Raman spectroscopy, characterized in that, It includes the following steps: S1. Sample pretreatment: Take a sample of Maotai-flavor liquor and mix it with gold nanostar solution, then add coagulation accelerator and mix thoroughly to obtain the sample solution to be tested. S2. Raman spectroscopy acquisition: Place the sample solution to be tested on the sample stage of the Raman spectrometer for Raman spectroscopy acquisition; S3. Spectral data preprocessing: Background subtraction and smoothing are performed on the acquired raw spectra, quantitative peaks are identified and peak height integral values ​​are calculated, the average value of multiple measurements is taken as the final spectral data, and the CCO stretching vibration peak of ethanol is selected as the internal standard peak to normalize all spectra. S4. Input the extracted Raman spectral feature data into the established decision tree model to identify the origin and quality grade of Maotai-flavor liquor.

2. The method according to claim 1, characterized in that, In step S4, the model is established through the following steps: S4.1 Establishing an Orthogonal Partial Least Squares-Discriminant Analysis (OPLS-DA) Model: Import the preprocessed Raman spectral data into chemometrics software, establish an OPLS-DA model based on different origins and quality classification standards, and screen key Raman shift points with VIP>1 by variable importance projection values; S4.2 Constructing a discrimination model: Randomly divide the selected key Raman displacement point data into training and test sets, establish a classification prediction model, optimize the model parameters through the training set, and verify the discrimination accuracy of the model using the test set.

3. The method according to claim 2, characterized in that, In step S4.1, during the establishment of the OPLS-DA model, the preprocessed Raman spectral data are labeled with corresponding categories according to geographical origin, river basin location, or market price level. The calculation of the variable importance projection value is based on the contribution weight of each variable in the model to the classification result, and the spectral shift points with variable importance projection values ​​greater than 1 are taken as key feature variables.

4. The method according to claim 3, characterized in that, Among the key Raman displacement points selected by the OPLS-DA model, 1600 cm -1 The nearby Raman shift is related to the C=C skeleton vibration in aromatic compounds, 1280 cm⁻¹ -1 The nearby Raman shift corresponds to the CO absorption peak of ester or ether molecules, at 2200 cm⁻¹. -1 The nearby Raman shifts are related to the stretching vibrations of C≡C and C≡N in alkynes; among the key Raman shift points screened by the OPLS-DA model, 1750 cm⁻¹ is... -1 The nearby Raman shifts are related to the C=O bonds in esters or amides; among the key Raman shift points screened by the OPLS-DA model, 1100 cm⁻¹ is... -1 The nearby Raman shift is a characteristic peak of the CO stretching vibration in alcohol molecules or the C-C bond stretching vibration in carbon chain compounds.

5. The method according to claim 2, characterized in that, The identification model constructed in step S4.2 is a decision tree model.

6. The method according to claim 1, characterized in that, In step S3, background subtraction is performed on the acquired raw Raman spectral data. Background subtraction methods include polynomial fitting, iterative approximation algorithms, and wavelet transform. Subsequently, the spectral data is smoothed using moving average, Savitzky-Golay filter, or multi-point smoothing algorithm. Then, characteristic peaks in the spectrum are identified and quantitatively analyzed, calculating the peak height, peak area, or peak intensity integral value for each peak. Finally, the spectral data is normalized, selecting the ethanol molecule at 880 cm⁻¹. -1 The nearby CCO stretching vibration peak is used as an internal standard peak. The intensity value of this peak in the spectrum of each sample is set as a unified standard value, thereby achieving the standardization of the intensity of the entire spectrum.

7. The method according to claim 1, characterized in that, In step S2, the pretreated Maotai-flavor liquor sample is placed at the detection position of the Raman spectrometer. The sample is excited by laser to generate a Raman scattering signal, and the scattered light is collected and analyzed by the spectrometer's detection system to obtain the characteristic Raman spectral information of the sample.

8. The method according to claim 1, characterized in that, In step S1, the sample of Maotai-flavor liquor to be tested is mixed with the prepared gold nanostar-enhanced substrate. The two are brought into full contact by stirring or shaking. The stirring intensity and time are controlled during the mixing process to ensure that the components to be tested in the liquor can be effectively adsorbed onto the surface of the gold nanostar, thereby obtaining a test sample with surface-enhanced Raman scattering effect. An electrolyte solution, which is potassium chloride solution, is added to the mixing system as a coagulation accelerator.

9. The method according to claim 1 or 8, characterized in that, Before step S1, the following steps may be performed: S0. Preparation of gold nanostar-reinforced substrate: After heating the chloroauric acid solution to boiling, add trisodium citrate solution to react and generate gold nanospheres. The gold nanosphere precipitate is obtained by centrifugation. The prepared gold nanospheres are dispersed in an organic solvent and stirred to mix the components thoroughly to generate a gold nanostar solution. The gold nanostar-reinforced substrate is then obtained by centrifugation and washing with an alcohol solvent.

10. The method according to claim 9, characterized in that, The organic solvent is N,N-dimethylformamide, and the protective agent polyvinylpyrrolidone and the metal precursor chloroauric acid are added to react and generate a gold nanostar solution. After the reaction is completed, the product is collected by centrifugation, and the centrifugation speed is higher than the separation speed of the gold nanospheres. The alcohol solvent is methanol.

Images