Detection method for analyzing content of melamine in dairy product through nanogold luminosity method

Abstract

The invention discloses a detection method for analyzing the content of melamine in a dairy product by a nanogold luminosity method, which comprises the following steps: S1, collecting a dairy product sample to form a clear dairy product sample suitable for spectrum detection; s2, preparing gold nanoparticles with tunable specific surface areas; s3, acquiring original spectral data containing multiple absorption peaks and background signals; s4, obtaining corrected spectrum data after background elimination; s5, extracting target peak data representing the melamine concentration; and S6, carrying out peak matching on the target peak data and a pre-constructed melamine standard peak template library to form a quantitative detection result of melamine in the dairy product sample, and carrying out threshold comparison on the quantitative detection result to generate a final detection and judgment result. According to the invention, the interpretability of the detection result and the recheck controllability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dairy products, in particular to a method for detecting melamine content in dairy products by using a nano-gold photometric method. Background Art

[0002] At present, the detection of melamine in dairy products mainly adopts chromatography combined with mass spectrometry technology. Some studies also use ELISA immunoassay and Raman spectroscopy. Although they have certain advantages in sensitivity and stability, there are still several technical bottlenecks and practical challenges.

[0003] First, traditional chromatography-mass spectrometry detection methods have complicated procedures, long detection times, and strong dependence on instruments. They usually require professional technicians to operate and are difficult to be widely promoted in general dairy processing companies or circulation links. In addition, the high equipment costs and maintenance costs also limit its application scope in small and medium-sized testing institutions. Secondly, although the ELISA method based on immunoassay is simple and fast, it is easily affected by interferences in complex dairy matrices, and there is a risk of false positives and false negatives. The repeatability and comparability of the test results are not high.

[0004] In the field of spectral detection, the combination of nanomaterials and UV-visible spectrophotometry has received widespread attention. Utilizing the surface-enhanced properties of nanomaterials to improve the response signal to melamine has become a research hotspot. However, existing detection methods based on gold nanoparticles are mostly laboratory-based and lack systematic signal processing mechanisms and data fitting methods. This leads to problems such as difficulty in effectively eliminating background signals, inaccurate target peak extraction, and large detection errors. Especially when facing samples with complex ingredients and strong matrix interference such as dairy products, traditional spectral processing methods cannot strike a balance between sensitivity and stability, affecting the reliability of detection results.

[0005] In addition, most existing spectral identification methods rely on a single peak or fixed segment for analysis, failing to fully consider the offset and distortion of the melamine absorption peak in different matrix environments, nor introducing advanced intelligent algorithms for multi-peak fitting and template comparison. Therefore, their adaptability and fault tolerance in complex samples are limited. There is an urgent need to develop an efficient detection method that can accurately separate spectral signals, suppress background interference, and accurately identify the characteristic peaks of melamine, so as to achieve rapid, accurate, and quantifiable analysis of the melamine content in dairy products. Summary of the Invention

[0006] One purpose of the present invention is to provide a method for analyzing the melamine content in dairy products using nanogold photometry, which improves the interpretability and retest controllability of the test results.

[0007] A method for analyzing melamine content in dairy products using nanogold spectrophotometry according to an embodiment of the present invention comprises the following steps:

[0008] S1. Collecting and pre-processing a dairy sample to remove interfering components and obtain a clear dairy sample suitable for spectral detection;

[0009] S2. preparing gold nanoparticles having a tunable specific surface area and adding the gold nanoparticles to the pretreated clear dairy sample, so that the melamine in the clear dairy sample specifically binds to the gold nanoparticles;

[0010] S3. Spectral scanning and acquisition of the clear dairy sample after the reaction to obtain raw spectral data containing multiple absorption peaks and background signals;

[0011] S4. Using the blank control sample to construct a background spectrum model, and applying the background spectrum model to the original spectral data to eliminate the background signal, and obtain the corrected spectral data after background elimination;

[0012] S5. Using an ant colony multi-peak fitting algorithm module with an adaptive pheromone update mechanism, the corrected spectral data is subjected to multi-peak signal separation and fitting, extracting target peak data representing melamine concentration;

[0013] S6. Peak match the target peak data with the pre-constructed melamine standard peak template library to form a quantitative detection result of melamine in the dairy sample, perform threshold comparison on the quantitative detection result, and generate the final detection judgment result.

[0014] Optionally, the S1 includes the following steps:

[0015] S11. Collect dairy samples from multiple sources. The total number of sampling batches is N, and each dairy sample is recorded as S i , where i = 1, 2, ..., N, constituting the dairy sample set S;

[0016] S12. For each dairy sample S in the dairy sample set S i Performing preliminary impurity removal treatment, removing fat components by centrifugation and removing protein components by protein precipitation, to form a preliminary dairy sample collection;

[0017] S13. The pH of the preliminary dairy sample set was adjusted to the target detection range and further filtered through a 0.22 μm pore size filter membrane to remove particulate suspended impurities to obtain a clear dairy sample;

[0018] S14. Construct a clear dairy sample dataset S for all processed clear dairy samples clear , the clear dairy sample data set meets the optical transmittance threshold condition. The transmittance of the clear dairy sample is greater than or equal to the preset minimum transmittance standard.

[0019] Optionally, S2 includes the following steps:

[0020] S21. Synthesize gold nanoparticles and obtain a collection of gold nanoparticles of different sizes by regulating the reducing agent concentration, reaction temperature and reaction time G = {g1, g2, ..., g M}, where g j represents the jth type of gold nanoparticle size distribution, M is the total number of adjustable particle size types;

[0021] S22. For each type of gold nanoparticle g in the gold nanoparticle set G. j Determine its specific surface area A j , construct the gold nanoparticle specific surface area data set A={A1,A2,...,A M}, where A j Indicates that the particle size type is g j The specific surface area per unit mass of gold nanoparticles;

[0022] S23. Compare the gold nanoparticle surface area dataset A with the pre-built clear milk sample dataset S clear Adaptive analysis was performed based on initial estimates of melamine concentrations in clear dairy samples. Choose the most suitable type of gold nanoparticles Constructing an enhanced spectral response system;

[0023] S24. Select the most suitable type of gold nanoparticles Add gold nanoparticles at a preset concentration ρ g Add to the corresponding clear dairy sample S″ i ∈S clear , generate the reaction sample set R:

[0024] R={R1,R2,...,R N};

[0025] Among them, R i Indicates clear dairy sample S″ i With gold nanoparticles The sample after reaction meets the mass ratio control conditions:

[0026]

[0027] Among them, ρ g is the concentration of gold nanoparticles added, in mg / mL, m g is the mass of gold nanoparticles added, is the sample volume;

[0028] S25. In the reaction sample set R, melamine molecules bind to the surface of the gold nanoparticles through specific interactions to form a gold nanoparticle-melamine composite structure.

[0029] Optionally, S3 includes the following steps:

[0030] S31. Perform a spectral scanning operation on the reaction sample set R, using a UV-visible spectrophotometer to scan each clear dairy sample R within a preset wavelength range. i Perform full-band scanning to obtain the spectral absorption function A corresponding to the sample i (λ), where λ represents the scanning wavelength, A i (λ) represents the clear dairy sample R i Absorbance at scanning wavelength λ;

[0031] S32. All the spectral absorption functions A obtained by scanning are i (λ) is sorted into the original spectral dataset D according to the sample order raw ;

[0032] S33. For the original spectral dataset D raw Perform initial signal analysis to identify a multiple peak structure including melamine-induced absorption peaks, coexisting impurity absorption peaks, and matrix background interference signals;

[0033] S34. The original spectral dataset D raw Each clear dairy sample spectrum A i (λ) is expressed as a composite spectral structure according to the following composition:

[0034] A i (λ)=A i,target (λ)+A i,interf (λ)+A i,bg (λ);

[0035] Among them, A i,target (λ) represents the absorption peak induced by melamine, A i,interf (λ) represents the absorption peak of coexisting impurities, A i,bg (λ) represents the matrix background interference signal.

[0036] Optionally, the S4 includes the following steps:

[0037] S41. Collect clear dairy samples that do not contain melamine as blank control samples to construct a blank dairy sample set S blank ={B1,B2,...,B K}, where B k represents the kth sample, K represents the number of blank dairy samples;

[0038] S42. For blank dairy sample set S blank Each blank dairy sample B k Perform UV-visible spectrum scanning to obtain background spectrum function Where λ is the scanning wavelength, Blank dairy sample B k Absorbance at wavelength λ;

[0039] S43. Introducing the multi-component statistical perturbation factor of the dairy matrix, all blank background spectral functions The components are weighted according to the interference factors in dairy products, including protein residues, lactose and trace mineral ions, to construct the interference factor weight vector w bg :

[0040] w bg ={w1,w2,...,w K};

[0041] Among them, w k represents the comprehensive influence weight of the interference factor on the background signal in the kth sample, which is calculated based on its correlation with the melamine characteristic segment in the spectral response;

[0042] S44. Based on the perturbation factor weight vector w bg Perform weighted fusion on blank dairy sample spectra to construct a matrix-sensitive background spectrum model The matrix-sensitive background spectrum model reflects the composite background distribution characteristics close to the real dairy matrix in the actual detection environment:

[0043]

[0044] S45. The matrix-sensitive background spectrum model Applied to the original spectral dataset D raw For each clear dairy sample in , the corrected spectral function of the sample is obtained:

[0045]

[0046] in, In order to remove the complex background of dairy products and retain the corrected spectral data of the characteristic absorption peaks of nano-gold-melamine, a corrected spectral dataset D was constructed. corr .

[0047] Optionally, the S5 includes the following steps:

[0048] S51. Input the corrected spectral data set into the ant colony multi-peak fitting algorithm module with each corrected spectral function is the fitting objective function;

[0049] S52. Construct the theoretical fitting model of sample i as the superposition of multiple peak functions, fitting the spectral function Defined as:

[0050]

[0051] Among them, G(p ij ,λ) is the jth fitting peak function, p ij =(μ ij ,σ ij ,h ij ) represent the peak center, half-maximum width and peak height, respectively;

[0052] S53. Construct the spectral fitting residual as the objective function f by fitting the objective function and the spectral function i (P i ):

[0053]

[0054] Where Λ is the scanning wavelength range, P i is the peak parameter set to be optimized;

[0055] S54. In each round of iteration, the ant colony takes the objective function f i (P i ) is minimized, and the path selection mechanism is used to search in the peak parameter space. The path transition probability is defined as:

[0056]

[0057] in, is the current pheromone, represents the heuristic function value, α, β are adjustment factors, and ∈ is a small constant to prevent division by zero;

[0058] S55. Update according to the pheromone reinforcement optimal strategy, and set the current global optimal path to be The corresponding minimum objective function value The update rules are:

[0059]

[0060] Among them, ρ is the pheromone volatility factor, Q is the enhancement factor, and the peak points in the path will get the update increment

[0061] S56. During the iterative process of the fitting paths formed in steps S54 and S55, minimizing the residual between the corrected spectrum and the fitted spectrum is optimized. After the ant colony multi-peak fitting algorithm converges, the fitting path with the smallest residual is selected from all fitting paths as the optimal fitting path for the current sample. The entire set of peak parameters fitted in the optimal fitting path is extracted. Within the extracted set of peak parameters, the fitting peaks whose peak centers are within the wavelength range of the standard characteristic absorption band of melamine are identified to form the target melamine peak set for the sample.

[0062] S57. The melamine target peak sets of all samples are aggregated and integrated to form a unified melamine target peak data set, wherein the target peak data set is used to characterize the characteristic absorption response of melamine generated by each clear dairy sample after binding to the gold nanoparticles.

[0063] Optionally, the S6 includes the following steps:

[0064] S61. Construct a melamine standard peak template library. The melamine standard peak template library is constructed based on standard samples under different known concentration gradients. The target peak parameters are extracted using the same ant colony multi-peak fitting algorithm as S5 to form a standard template set T = {T1, T2, ..., T L}, each template T l Contains the corresponding concentration C l The characteristic peak parameter set of melamine under

[0065] S62. Calling the melamine target peak dataset P target , and perform parameter-level matching with each template in the standard peak template set T. The matching process constructs a similarity score based on the weighted differences in peak center, half-height width and peak height to calculate the melamine target peak data set With each template T l Similarity scores form a template similarity list S i ={s i1 ,s i2 ,...,s iL}, where s il Indicates the similarity between sample i and template l;

[0066] S63. In the template similarity list S i In the example, select the template T corresponding to the one with the largest similarity l As the best matching template for this sample, the corresponding concentration is And record the similarity score of the best matching template At the same time, call the minimum value of the fitting residual of S5 sample i Form a set of sample concentration estimates and matching credibility indicators: a set of estimated dairy sample concentration values: Minimum set of fitted residuals: The best matching template similarity score set:

[0067] S74. Estimated concentrations based on each milk sample Minimum fitting residual And the similarity score of the best matching template Three thresholds are set as the basis for detection and judgment, including the melamine concentration safety detection threshold C safe , fitting residual credible threshold f thres and similarity lower limit threshold s thres ;

[0068] S65. Output the final detection result as the melamine concentration analysis output result in the dairy product.

[0069] Optionally, the S74 detection and determination rules are as follows:

[0070] When the estimated concentration of the sample is less than or equal to the concentration safety threshold And the fitting residual does not exceed the credible threshold At the same time, the template similarity is not lower than the similarity threshold When , it is judged as a qualified sample;

[0071] When the estimated concentration of the sample is higher than the concentration safety threshold At the same time, both the fitting residual and the template similarity meet the credibility condition, that is, and Determine samples that exceed the standard;

[0072] When the sample fitting residual exceeds the credible threshold Or the template matching similarity is lower than the similarity threshold Regardless of the concentration estimation result, the sample is judged as suspicious and it is recommended to enter the re-inspection process to ensure the reliability of the test results.

[0073] Optionally, the final detection and judgment result includes a sample number, a fitting path number, an estimated concentration value, a best-matching template number, a fitting residual value, a template similarity score, and a detection and judgment label.

[0074] The beneficial effects of the present invention are:

[0075] (1) In the background elimination process, the present invention introduces a multi-component statistical perturbation factor weight vector based on the actual matrix composition of dairy products. Combined with the ultraviolet-visible spectrum data of blank samples, a dynamic weighted fusion background spectrum model is established, which can dynamically adjust the background spectrum contrast according to the changes in the proportion of protein residues, lactose, and trace ion components in the sample, significantly improving the ability to identify and eliminate non-characteristic interference peaks in complex samples.

[0076] (2) The present invention designs a multi-peak fitting model based on ant colony optimization and introduces a parallel strategy of pheromone reinforcement and local path penalty, which effectively improves the convergence speed and solution diversity of the global optimal path search. By dynamically updating the fitting path pheromone in the high-dimensional peak parameter space, the algorithm can prioritize exploring the area with the smallest residual in each round of iteration, while suppressing the risk of falling into the local minimum.

[0077] (3) The present invention establishes a peak parameter template matching model. By comparing the similarity between the three-dimensional characteristic parameters of the sample fitting peak, namely, “central wavelength-half-maximum width-peak height” and the standard template, a peak-level concentration mapping and credibility scoring system is formed. This not only achieves more fine-grained quantitative judgment, but also integrates the template similarity score and fitting residual as multiple bases for detection judgment, thereby improving the interpretability and retest controllability of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 The present invention provides a flow chart of a method for analyzing melamine content in dairy products using nano-gold photometry. DETAILED DESCRIPTION

[0080] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0081] refer to Figure 1 A method for analyzing the content of melamine in dairy products by nano-gold spectrophotometry comprises the following steps:

[0082] S1. Collecting and pre-processing a dairy sample to remove interfering components and obtain a clear dairy sample suitable for spectral detection;

[0083] S2. preparing gold nanoparticles having a tunable specific surface area and adding the gold nanoparticles to the pretreated clear dairy sample, so that the melamine in the clear dairy sample specifically binds to the gold nanoparticles;

[0084] S3. Spectral scanning and acquisition of the clear dairy sample after the reaction to obtain raw spectral data containing multiple absorption peaks and background signals;

[0085] S4. Using the blank control sample to construct a background spectrum model, and applying the background spectrum model to the original spectral data to eliminate the background signal, and obtain the corrected spectral data after background elimination;

[0086] S5. Using an ant colony multi-peak fitting algorithm module with an adaptive pheromone update mechanism, the corrected spectral data is subjected to multi-peak signal separation and fitting, extracting target peak data representing melamine concentration;

[0087] S6. Peak match the target peak data with the pre-constructed melamine standard peak template library to form a quantitative detection result of melamine in the dairy sample, perform threshold comparison on the quantitative detection result, and generate the final detection judgment result.

[0088] In this embodiment, S1 includes the following steps:

[0089] S11. Collect dairy samples from multiple sources. The total number of sampling batches is N, and each dairy sample is recorded as S i , where i = 1, 2, ..., N, constituting the dairy sample set S;

[0090] S12. For each dairy sample S in the dairy sample set S i Performing preliminary impurity removal treatment, removing fat components by centrifugation and removing protein components by protein precipitation, to form a preliminary dairy sample collection;

[0091] S13. The pH of the preliminary dairy sample set was adjusted to the target detection range and further filtered through a 0.22 μm pore size filter membrane to remove particulate suspended impurities to obtain a clear dairy sample;

[0092] S14. Construct a clear dairy sample dataset S for all processed clear dairy samples clear , the clear dairy sample data set meets the optical transmittance threshold condition. The transmittance of the clear dairy sample is greater than or equal to the preset minimum transmittance standard.

[0093] In this embodiment, S2 includes the following steps:

[0094] S21. Synthesize gold nanoparticles and obtain a collection of gold nanoparticles of different sizes by regulating the reducing agent concentration, reaction temperature and reaction time G = {g1, g2, ..., g M}, where g j represents the jth type of gold nanoparticle size distribution, M is the total number of adjustable particle size types;

[0095] S22. For each type of gold nanoparticle g in the gold nanoparticle set G. j Determine its specific surface area A j , construct the gold nanoparticle specific surface area data set A={A1,A2,...,A M}, where A j Indicates that the particle size type is g j The specific surface area per unit mass of gold nanoparticles;

[0096] S23. Compare the gold nanoparticle surface area dataset A with the pre-built clear milk sample dataset S clear Adaptive analysis was performed based on initial estimates of melamine concentrations in clear dairy samples. Choose the most suitable type of gold nanoparticles Constructing an enhanced spectral response system;

[0097] S24. Select the most suitable type of gold nanoparticles Add gold nanoparticles at a preset concentration ρ g Add to the corresponding clear dairy sample S″ i ∈S clear , generate the reaction sample set R:

[0098] R={R1,R2,...,R N};

[0099] Among them, R i Indicates clear dairy sample S″ i With gold nanoparticles The sample after reaction meets the mass ratio control conditions:

[0100]

[0101] Among them, ρ g is the concentration of gold nanoparticles added, in mg / mL, m g is the mass of gold nanoparticles added, is the sample volume;

[0102] S25. In the reaction sample set R, melamine molecules bind to the surface of the gold nanoparticles through specific interactions to form a gold nanoparticle-melamine composite structure.

[0103] In this embodiment, S3 includes the following steps:

[0104] S31. Perform a spectral scanning operation on the reaction sample set R, using a UV-visible spectrophotometer to scan each clear dairy sample R within a preset wavelength range. i Perform full-band scanning to obtain the spectral absorption function A corresponding to the samplei (λ), where λ represents the scanning wavelength, A i (λ) represents the clear dairy sample R i Absorbance at scanning wavelength λ;

[0105] S32. All the spectral absorption functions A obtained by scanning are i (λ) is sorted into the original spectral dataset D according to the sample order raw ;

[0106] S33. For the original spectral dataset D raw Perform initial signal analysis to identify a multiple peak structure including melamine-induced absorption peaks, coexisting impurity absorption peaks, and matrix background interference signals;

[0107] S34. The original spectral dataset D raw Each clear dairy sample spectrum A i (λ) is expressed as a composite spectral structure according to the following composition:

[0108] A i (λ)=A i,target (λ)+A i,interf (λ)+A i,bg (λ);

[0109] Among them, A i,target (λ) represents the absorption peak induced by melamine, A i,interf (λ) represents the absorption peak of coexisting impurities, A i,bg (λ) represents the matrix background interference signal.

[0110] In this embodiment, S4 includes the following steps:

[0111] S41. Collect clear dairy samples that do not contain melamine as blank control samples to construct a blank dairy sample set S blank ={B1,B2,...,B K}, where B k represents the kth sample, K represents the number of blank dairy samples;

[0112] S42. For blank dairy sample set S blank Each blank dairy sample B k Perform UV-visible spectrum scanning to obtain background spectrum function Where λ is the scanning wavelength, Blank dairy sample B k Absorbance at wavelength λ;

[0113] S43. Introducing the multi-component statistical perturbation factor of the dairy matrix, all blank background spectral functions The components are weighted according to the interference factors in dairy products, including protein residues, lactose and trace mineral ions, to construct the interference factor weight vector w bg :

[0114] w bg ={w1,w2,...,w K};

[0115] Among them, w k represents the comprehensive influence weight of the interference factor on the background signal in the kth sample, which is calculated based on its correlation with the melamine characteristic segment in the spectral response;

[0116] S44. Based on the perturbation factor weight vector w bg Perform weighted fusion on blank dairy sample spectra to construct a matrix-sensitive background spectrum model The matrix-sensitive background spectrum model reflects the composite background distribution characteristics close to the real dairy matrix in the actual detection environment:

[0117]

[0118] S45. Matrix-sensitive background spectrum model Applied to the original spectral dataset D raw For each clear dairy sample in , the corrected spectral function of the sample is obtained:

[0119]

[0120] in, In order to remove the complex background of dairy products and retain the corrected spectral data of the characteristic absorption peaks of nano-gold-melamine, a corrected spectral dataset D was constructed. corr .

[0121] In this embodiment, S5 includes the following steps:

[0122] S51. Input the corrected spectral data set into the ant colony multi-peak fitting algorithm module with each corrected spectral function is the fitting objective function;

[0123] S52. Construct the theoretical fitting model of sample i as the superposition of multiple peak functions, fitting the spectral function Defined as:

[0124]

[0125] Among them, G(p ij ,λ) is the jth fitting peak function, p ij =(μ ij ,σ ij ,h ij) represent the peak center, half-maximum width and peak height, respectively;

[0126] S53. Construct the spectral fitting residual as the objective function f by fitting the objective function and the spectral function i (P i ):

[0127]

[0128] Where Λ is the scanning wavelength range, P i is the peak parameter set to be optimized;

[0129] S54. In each round of iteration, the ant colony takes the objective function f i (P i ) is minimized, and the path selection mechanism is used to search in the peak parameter space. The path transition probability is defined as:

[0130]

[0131] in, is the current pheromone, represents the heuristic function value, α, β are adjustment factors, and ∈ is a small constant to prevent division by zero;

[0132] S55. Update according to the pheromone reinforcement optimal strategy, and set the current global optimal path to be The corresponding minimum objective function value The update rules are:

[0133]

[0134] Among them, ρ is the pheromone volatility factor, Q is the enhancement factor, and the peak points in the path will get the update increment

[0135] S56. During the iterative process of the fitting paths formed in steps S54 and S55, minimizing the residual between the corrected spectrum and the fitted spectrum is optimized. After the ant colony multi-peak fitting algorithm converges, the fitting path with the smallest residual is selected from all fitting paths as the optimal fitting path for the current sample. The entire set of peak parameters fitted in the optimal fitting path is extracted. Within the extracted set of peak parameters, the fitting peaks whose peak centers are within the wavelength range of the standard characteristic absorption band of melamine are identified to form the target melamine peak set for the sample.

[0136] S57. The melamine target peak sets of all samples are aggregated and integrated to form a unified melamine target peak data set. The target peak data set is used to characterize the characteristic absorption response of melamine produced by each clear dairy sample after binding to the gold nanoparticles.

[0137] In this embodiment, S6 includes the following steps:

[0138] S61. Construct a melamine standard peak template library. The melamine standard peak template library is constructed based on standard samples under different known concentration gradients. The target peak parameters are extracted using the same ant colony multi-peak fitting algorithm as S5 to form a standard template set T = {T1, T2, ..., T L}, each template T l Contains the corresponding concentration C l The characteristic peak parameter set of melamine under

[0139] S62. Calling the melamine target peak dataset P target , and perform parameter-level matching with each template in the standard peak template set T. The matching process constructs a similarity score based on the weighted differences in peak center, half-height width and peak height to calculate the melamine target peak data set With each template T l Similarity scores form a template similarity list S i ={s i1 ,s i2 ,...,s iL}, where s il Indicates the similarity between sample i and template l;

[0140] S63. In the template similarity list S i In the example, select the template T corresponding to the one with the largest similarity l As the best matching template for this sample, the corresponding concentration is And record the similarity score of the best matching template At the same time, call the minimum value of the fitting residual of S5 sample i Form a set of sample concentration estimates and matching credibility indicators: a set of estimated dairy sample concentration values: Minimum set of fitted residuals: The best matching template similarity score set:

[0141] S74. Estimated concentrations based on each milk sample Minimum fitting residual And the similarity score of the best matching template Three thresholds are set as the basis for detection and judgment, including the melamine concentration safety detection threshold C safe , fitting residual credible threshold f thres and similarity lower limit threshold s thres ;

[0142] S65. Output the final detection result as the melamine concentration analysis output result in the dairy product.

[0143] In this embodiment, the S74 detection and determination rules are as follows:

[0144] When the estimated concentration of the sample is less than or equal to the concentration safety threshold And the fitting residual does not exceed the credible threshold At the same time, the template similarity is not lower than the similarity threshold When , it is judged as a qualified sample;

[0145] When the estimated concentration of the sample is higher than the concentration safety threshold At the same time, both the fitting residual and the template similarity meet the credibility condition, that is, and Determine samples that exceed the standard;

[0146] When the sample fitting residual exceeds the credible threshold Or the template matching similarity is lower than the similarity threshold Regardless of the concentration estimation result, the sample is judged as suspicious and it is recommended to enter the re-inspection process to ensure the reliability of the test results.

[0147] In this embodiment, the final detection result includes a sample number, a fitting path number, an estimated concentration value, a best matching template number, a fitting residual value, a template similarity score, and a detection label.

[0148] Example 1:

[0149] On September 22, 2024, the Food Quality Supervision and Inspection Center in City A received a commission from a large dairy company (hereinafter referred to as "A Dairy"), requesting it to test the melamine content of a batch of ambient temperature pure milk products that had recently been shipped from the factory. There were a total of 3 production batches of this batch of products, namely batch numbers JNSY-0922A, JNSY-0922B and JNSY-0922C. Each batch provided 6 bottles of samples, for a total of 18 bottles of samples.

[0150] On the day of testing, laboratory inspectors began to number and register the samples for pre-treatment. The pre-treatment operations included using a 5000rpm high-speed centrifuge to remove the upper layer of cream (processing time 10 minutes), followed by adding 1mL of 10% trichloroacetic acid solution to each bottle for protein precipitation, and then adjusting the pH to 6.5 with a buffer solution. The samples were then filtered through a 0.22μm pore size filter membrane to remove particulate impurities. Finally, it was confirmed that the transmittance of all samples was ≥86%, meeting the test conditions.

[0151] The inspector numbers the pre-treated samples as A01 to A18 and begins the gold nanoparticle reaction phase. The system recommends adding gold nanoparticles of different sizes to each sample based on the initial protein residue. For example, samples A04 and A05 have high protein contents, so the system automatically recommends using 30nm particles (with a specific surface area of approximately 60.8m 2 / g), while A01 and A02 use 20nm particles (specific surface area of 45.2m 2 / g), the concentration of the gold nanoparticle solution was 20 μg / mL, the reaction temperature was controlled at 37°C, and the static reaction time was set to 25 minutes.

[0152] The inspector used a Shimadzu UV-2600 ultraviolet-visible spectrometer to perform spectral scanning on the reacted samples, setting the wavelength range to 190–400 nm with a step size of 1 nm, and obtained an absorbance sequence of 211 wavelength points. The scanning results of sample A09 showed an abnormal peak for the first time: a clear absorption peak appeared at a wavelength of 241 nm, with an intensity of 0.452 AU, which was significantly higher than the normal range.

[0153] During the background elimination stage, the system used 10 bottles of melamine-free samples collected by the laboratory the day before as blank controls, automatically generated a disturbance factor weight model, and dynamically eliminated background signals. It is worth noting that sample A09 still retained a peak value of 0.412AU after background elimination, with a signal-to-noise ratio of 15.4. The system initially judged it as a "high-risk sample."

[0154] The system automatically input the corrected spectrum of sample A09 into the ant colony multi-peak fitting module. After 18 rounds of iterative optimization, the objective function residual of the optimal path was reduced to 0.0067. The fitted main peak was located at 241.1nm, the half-width was 4.6nm, and the peak height was 0.415AU. The fitting residual and the standard template matching error were both lower than the system-set threshold. The system's final matching concentration was estimated to be 1.68mg / kg.

[0155] At the same time, the traditional detection process was also in progress. Experimenter Chen Na used the Waters Acquity UPLC system coupled with the XEVOT Q-S mass spectrometer for liquid chromatography-mass spectrometry analysis. After treatment, sample A09 showed a melamine concentration of 1.72 mg / kg. The concentration difference between the two methods was only 0.04 mg / kg. The system confidence score was 98.3%, and the fitting residual score was 99.1%. The result was judged to be an "exceeding the standard sample."

[0156] The detection system automatically generates an early warning report for the sample that exceeds the standard, which contains the following fields:

[0157] Sample number: A09; concentration estimate: 1.68 mg / kg; fitting residual: 0.0067; matching template number: STD_241nm_T3; template similarity score: 0.983; judgment label: exceeded the standard; report generation time: 2024-09-22 14:05; automatic notification: sent to the email address of the quality control person in charge of A Dairy and backed up to the supervision platform.

[0158] After receiving the report at 15:30 on the same day, Dairy A immediately suspended the shipment of this batch of products and initiated an internal traceability process. It ultimately confirmed that this batch of raw milk came from a certain ranch, and that the self-inspection records submitted by the ranch were missing, which constituted an illegal operation. This move prevented potential unqualified dairy products from entering the market.

[0159] To further validate the system's performance, the laboratory tested 72 dairy samples between September 23rd and 25th, 36 of which were compared using the proposed method and 36 using LC-MS / MS. The samples covered a concentration range of 0.0 to 4.5 mg / kg. Statistical analysis showed that the proposed method had an average error of ±0.036 mg / kg in the 0.1-2.0 mg / kg range, with an average detection time of 32 minutes, a 45% reduction in time compared to LC-MS / MS, and eliminated the need for expensive reagents and high-end mass spectrometry equipment.

[0160] In the background elimination process, the present invention introduces a multi-component statistical perturbation factor weight vector constructed based on the actual matrix composition of dairy products. Combined with the ultraviolet-visible spectrum data of blank samples, a dynamic weighted fusion background spectrum model is established, which can dynamically adjust the background spectrum contrast according to the changes in the proportion of protein residues, lactose, and trace ion components in the sample, significantly improving the ability to identify and eliminate non-characteristic interference peaks in complex samples.

[0161] The present invention designs a multi-peak fitting model based on ant colony optimization and introduces a parallel strategy of pheromone reinforcement and local path penalty, which effectively improves the convergence speed and solution diversity of the global optimal path search. By dynamically updating the fitting path pheromone in the high-dimensional peak parameter space, the algorithm can prioritize exploring the area with the smallest residual in each round of iteration while suppressing the risk of falling into local minima.

[0162] The present invention establishes a peak parameter template matching model. By comparing the similarity between the three-dimensional characteristic parameters of "central wavelength-half-maximum width-peak height" of the sample fitting peak and the standard template, a peak-level concentration mapping and credibility scoring system is formed. It not only achieves finer-grained quantitative judgment, but also integrates template similarity scoring and fitting residuals as multiple bases for detection judgment, thereby improving the interpretability and re-inspection controllability of the detection results.

[0163] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A method for analyzing the content of melamine in dairy products by nanogold spectrophotometry, characterized in that: The steps include: S1. Collecting and pre-processing a dairy sample to remove interfering components and obtain a clear dairy sample suitable for spectral detection; S2. preparing gold nanoparticles having a tunable specific surface area and adding the gold nanoparticles to the pretreated clear dairy sample, so that the melamine in the clear dairy sample specifically binds to the gold nanoparticles; S3. Spectral scanning and acquisition of the clear dairy sample after the reaction to obtain raw spectral data containing multiple absorption peaks and background signals; S4. Using the blank control sample to construct a background spectrum model, and applying the background spectrum model to the original spectral data to eliminate the background signal, and obtain the corrected spectral data after background elimination; S5. Using an ant colony multi-peak fitting algorithm module with an adaptive pheromone update mechanism, the corrected spectral data is subjected to multi-peak signal separation and fitting, extracting target peak data representing melamine concentration; S6. Peak match the target peak data with the pre-constructed melamine standard peak template library to form a quantitative detection result of melamine in the dairy sample, perform threshold comparison on the quantitative detection result, and generate the final detection judgment result.

2. The method for analyzing the melamine content in dairy products by nanogold photometry according to claim 1, characterized in that: Said S1 comprises the following steps: S11. Collect dairy samples from multiple sources. The total number of sampling batches is N, and each dairy sample is recorded as S i , where i = 1, 2, ..., N, constituting the dairy sample set S; S12. For each dairy sample S in the dairy sample set S i Performing preliminary impurity removal treatment, removing fat components by centrifugation and removing protein components by protein precipitation, to form a preliminary dairy sample collection; S13. The pH of the preliminary dairy sample set was adjusted to the target detection range and further filtered through a 0.22 μm pore size filter membrane to remove particulate suspended impurities to obtain a clear dairy sample; S14. Construct a clear dairy sample dataset S for all processed clear dairy samples clear , the clear dairy sample data set meets the optical transmittance threshold condition. The transmittance of the clear dairy sample is greater than or equal to the preset minimum transmittance standard.

3. The method for analyzing the melamine content in dairy products by nanogold photometry according to claim 1, characterized in that: The S2 comprises the following steps: S21. Synthesizing gold nanoparticles, obtaining a collection of gold nanoparticles G of varying sizes by controlling the reducing agent concentration, reaction temperature, and reaction time; S22. For each type of gold nanoparticle g in the gold nanoparticle set G. j Determine its specific surface area A j , construct nano-gold particle specific surface area dataset A; S23. Compare the gold nanoparticle surface area dataset A with the pre-built clear milk sample dataset S clear Adaptive analysis was performed based on initial estimates of melamine concentrations in clear dairy samples. Choose the most suitable type of gold nanoparticles Constructing an enhanced spectral response system; S24. Select the most suitable type of gold nanoparticles Add gold nanoparticles at a preset concentration ρ g Add to the corresponding clear dairy sample S″ i ∈S clear , generate a reaction sample set R, which includes the clear dairy sample S″ i With gold nanoparticles Sample after reaction; S25. In the reaction sample set R, melamine molecules bind to the surface of the gold nanoparticles through specific interactions to form a gold nanoparticle-melamine composite structure.

4. The method for analyzing the melamine content in dairy products by nanogold photometry according to claim 1, wherein: The S3 includes the following steps: S31. Perform a spectral scanning operation on the reaction sample set R, using a UV-visible spectrophotometer to scan each clear dairy sample R within a preset wavelength range. i Perform full-band scanning to obtain the spectral absorption function A corresponding to the sample i (λ), where λ represents the scanning wavelength, A i (λ) represents the clear dairy sample R i Absorbance at scanning wavelength λ; S32. All the spectral absorption functions A obtained by scanning are i (λ) is sorted into the original spectral dataset D according to the sample order raw ; S33. For the original spectral dataset D raw Perform initial signal analysis to identify a multiple peak structure including melamine-induced absorption peaks, coexisting impurity absorption peaks, and matrix background interference signals; S34. The original spectral dataset D raw Each clear dairy sample spectrum A i (λ) is expressed as a composite spectral structure according to the following composition: A i (λ)=A i,target (λ)+A i,interf (λ)+A i,bg (l); Among them, A i,target (λ) represents the absorption peak induced by melamine, A i,interf (λ) represents the absorption peak of coexisting impurities, A i,bg (λ) represents the matrix background interference signal.

5. The method for analyzing the melamine content in dairy products by nano-gold photometry according to claim 1, characterized in that: The S4 comprises the following steps: S41. Collect clear dairy samples that do not contain melamine as blank control samples to construct a blank dairy sample set S blank ; S42. For blank dairy sample set S blank Each blank dairy sample B k Perform UV-visible spectrum scanning to obtain background spectrum function Where λ is the scanning wavelength, Blank dairy sample B k Absorbance at wavelength λ; S43. Introducing the multi-component statistical perturbation factor of the dairy matrix, all blank background spectral functions The components are weighted according to the interference factors in dairy products, including protein residues, lactose and trace mineral ions, to construct the interference factor weight vector w bg ; S44. Based on the perturbation factor weight vector w bg Perform weighted fusion on blank dairy sample spectra to construct a matrix-sensitive background spectrum model The matrix-sensitive background spectrum model reflects the composite background distribution characteristics close to the real dairy matrix in the actual detection environment: Among them, w k Represents the perturbation factor weight vector w bg The comprehensive influence weight of the interference factor on the background signal in the kth sample is calculated based on its correlation with the melamine characteristic segment in the spectral response; S45. The matrix-sensitive background spectrum model Applied to the original spectral dataset D raw For each clear dairy sample in , the corrected spectral function of the sample is obtained: in, In order to remove the complex background of dairy products and retain the corrected spectral data of the characteristic absorption peaks of nano-gold-melamine, a corrected spectral dataset D was constructed. corr .

6. The method for analyzing the melamine content in dairy products by nano-gold photometry according to claim 1, characterized in that: The S5 comprises the following steps: S51. Input the corrected spectral data set into the ant colony multi-peak fitting algorithm module with each corrected spectral function is the fitting objective function; S52. Construct the theoretical fitting model of sample i as the superposition of multiple peak functions, fitting the spectral function Defined as: Among them, G(p ij ,λ) is the jth fitting peak function, p ij =(μ ij ,σ ij ,h ij ) represent the peak center, half-maximum width and peak height, respectively; S53. Construct the spectral fitting residual as the objective function f by fitting the objective function and the spectral function i (P i ); S54. In each round of iteration, the ant colony takes the objective function f i (P i ) is minimized and the peak parameter space is searched through the path selection mechanism; S55. Update according to the pheromone reinforcement optimal strategy, and set the current global optimal path to be The corresponding minimum objective function value The update rules are: Among them, ρ is the pheromone volatility factor, Q is the enhancement factor, and the peak points in the path will get the update increment S56. During the iterative process of the fitting paths formed in steps S54 and S55, minimizing the residual between the corrected spectrum and the fitted spectrum is optimized. After the ant colony multi-peak fitting algorithm converges, the fitting path with the smallest residual is selected from all fitting paths as the optimal fitting path for the current sample. The entire set of peak parameters fitted in the optimal fitting path is extracted. Within the extracted set of peak parameters, the fitting peaks whose peak centers are within the wavelength range of the standard characteristic absorption band of melamine are identified to form the target melamine peak set for the sample. S57. The melamine target peak sets of all samples are aggregated and integrated to form a unified melamine target peak data set, wherein the target peak data set is used to characterize the characteristic absorption response of melamine generated by each clear dairy sample after binding to the gold nanoparticles.

7. The method for analyzing the melamine content in dairy products by nano-gold photometry according to claim 1, characterized in that: The S6 comprises the following steps: S61. Construct a melamine standard peak template library. The melamine standard peak template library is constructed based on standard samples under different known concentration gradients. The target peak parameters are extracted using the same ant colony multi-peak fitting algorithm as S5 to form a standard template set T. Each template T in the standard template set T l Contains the corresponding concentration C l The characteristic peak parameter set of melamine under S62. Calling the melamine target peak dataset P target , and perform parameter-level matching with each template in the standard peak template set T. The matching process constructs a similarity score based on the weighted differences in peak center, half-height width and peak height to calculate the melamine target peak data set With each template T l Similarity scores form a template similarity list S i ={s i1 ,s i2 ,...,s iL }, where s il Indicates the similarity between sample i and template l; S63. In the template similarity list S i In the example, select the template T corresponding to the one with the largest similarity l As the best matching template for this sample, the corresponding concentration is And record the similarity score of the best matching template At the same time, call the minimum value of the fitting residual of S5 sample i Form a set of sample concentration estimates and matching credibility indicators: a set of estimated dairy sample concentration values: Minimum set of fitted residuals: The best matching template similarity score set: S74. Estimated concentrations based on each milk sample Minimum fitting residual And the similarity score of the best matching template Three thresholds are set as the basis for detection and judgment, including the melamine concentration safety detection threshold C safe , fitting residual credible threshold f thres and similarity lower limit threshold s thres ; S65. Output the final detection result as the melamine concentration analysis output result in the dairy product.

8. The method for analyzing the melamine content in dairy products by nano-gold photometry according to claim 7, characterized in that: The S74 detection and judgment rules are as follows: When the estimated concentration of the sample is less than or equal to the concentration safety threshold And the fitting residual does not exceed the credible threshold At the same time, the template similarity is not lower than the similarity threshold When , it is judged as a qualified sample; When the estimated concentration of the sample is higher than the concentration safety threshold At the same time, both the fitting residual and the template similarity meet the credibility condition, that is, and Determine samples that exceed the standard; When the sample fitting residual exceeds the credible threshold Or the template matching similarity is lower than the similarity threshold Regardless of the concentration estimation result, the sample is judged as suspicious and it is recommended to enter the re-inspection process to ensure the reliability of the test results.

9. The method for analyzing the melamine content in dairy products by nano-gold photometry according to claim 7, characterized in that: The final detection result includes the sample number, fitting path number, estimated concentration value, best matching template number, fitting residual value, template similarity score and detection determination label.

Images