A Consistency Evaluation Method for Beverage Products Based on Volatile Substance Fingerprint Spectra

Abstract

This invention discloses a method for evaluating the consistency of beverage products based on volatile substance fingerprinting, belonging to the field of beverage product quality testing technology. The technical solution includes performing GC-MS full scan analysis on standard samples and test samples respectively to obtain component peak areas. After screening, the top 10 components with the highest peak areas in both the standard sample and the test sample are selected. Similarity analyses (existence similarity, ordinal similarity, and response similarity) are performed on these two samples to obtain a comprehensive similarity score. If the comprehensive similarity score is greater than 80%, the test sample can be considered similar to the standard sample, i.e., a qualified sample. This invention is applied to the consistency evaluation of beverage products, solving the problem that existing beverage products (such as juice, carbonated drinks, and tea drinks) have complex volatile flavor components, making it difficult for traditional methods to comprehensively evaluate batch-to-batch consistency.

Description

Technical Field

[0001] This invention belongs to the field of beverage product quality testing technology, and in particular relates to a method for evaluating the consistency of beverage products based on the fingerprint spectrum of volatile substances. Background Technology

[0002] Batch-to-batch consistency evaluation of beverage products is crucial for quality control. Existing methods include sensory evaluation, quantitative analysis, model-based discrimination, electronic noses, and visualization technologies. However, these methods suffer from drawbacks such as the high subjectivity of sensory evaluation, the high cost of quantitative analysis (requiring numerous standards to maintain calibration curves), the abstract and difficult-to-understand output of mathematical models (e.g., PLS-DA), and the inability of electronic noses to distinguish between structurally similar substances. Specifically: 1. Sensory evaluation is highly subjective: human evaluation is affected by individual differences and has poor repeatability. The complexity of aromas in beverage, alcoholic, and spice samples is unimaginable; a single sample can yield at least a hundred volatile components through a low-resolution GCMS full scan. Because sensory analysis is highly subjective and difficult to standardize, the reliability and repeatability of the tests cannot be guaranteed.

[0003] 2. High cost of quantitative analysis: Requires a large number of standards to establish calibration curves, and maintenance is time-consuming. Traditional instrumental analysis methods primarily utilize GC-MS (gas chromatography-mass spectrometry) to quantitatively analyze key flavor components in samples, thereby evaluating their flavor. However, the possibility of quantifying over a hundred compounds in complex flavor samples is very small. Quantitative analysis requires a large variety of standard substances, resulting in high procurement costs, and some flavor substances lack high-purity standards. The standard curves used for quantification of each compound require regular maintenance; maintaining hundreds of calibration curves would be a massive and time-consuming task. Furthermore, due to the volatility of flavor substances, frequent calibration curve maintenance may actually cause fluctuations in the detection data.

[0004] 3. Model-based discrimination methods have low practicality, produce abstract results, require specialized software, and are difficult to understand. In recent years, artificial intelligence and machine learning models, represented by partial least squares discriminant analysis (PLS-DA), have been used to analyze large amounts of aroma data to discover new flavor compounds in food samples and predict their aroma quality. However, model-based discriminant analysis requires expensive specialized software, and its complex mathematical principles are difficult for researchers in the food and flavor fields to understand. Ultimately, while the output can be visualized, it lacks intuitive and quantifiable indicators, making it difficult to compare samples.

[0005] 4. The electronic nose cannot distinguish between compounds with similar structures (CN110412010A). The bottleneck in electronic noses distinguishing structurally similar objects lies in the insufficient resolution of sensors and the analytical power of algorithms. Although optimizations have been achieved in some scenarios through material innovation (such as MOF), spatiotemporal feature fusion, and multimodal technologies, auxiliary separation techniques are still needed in high-precision applications. Ultimately, breakthroughs in biomimetic receptor design and quantum-level sensing technology are required for a definitive solution.

[0006] 5. Visualization techniques have not solved the problem of quantitative comparison (CN202011415170.8) Overall, the methods for evaluating batch consistency of flavored beverage products, whether sensory evaluation, key ingredient rating, or model evaluation, cannot adequately meet the current needs. Summary of the Invention

[0007] In view of the shortcomings of existing technologies, the technical problem to be solved by this invention is that the volatile flavor components in existing beverage products (such as juice, carbonated drinks, tea drinks, etc.) are complex and traditional methods are difficult to comprehensively evaluate batch-to-batch consistency. This invention proposes a consistency evaluation method for beverage products based on volatile substance fingerprinting, which is objective, low-cost, and universally applicable.

[0008] To solve the aforementioned technical problem, the technical solution adopted by the present invention is as follows: This invention provides a consistency evaluation method for beverage products based on volatile substance fingerprinting. GC-MS full scan analysis is performed on standard samples and test samples to obtain component peak areas. After screening, the top 10 components with the highest peak areas in both the standard and test samples are selected. Existence similarity, ordinal similarity, and response similarity analyses are performed on these top 10 components to obtain a comprehensive similarity score. The comprehensive similarity score is calculated using the following method: Overall similarity = Existence similarity × 0.7 + Ranking similarity × 0.2 + Response similarity × 0.1; If the overall similarity is greater than 80%, the sample to be tested can be identified as a sample similar to the standard sample, that is, a qualified sample.

[0009] In some embodiments, GC-MS full scan analysis of standard samples and test samples includes: using a weakly polar capillary column; using helium as the carrier gas at a flow rate of 1.5 mL / min; the column temperature program is: holding at 40°C for 3 min, then increasing to 250°C at 3°C / min and holding for 10 min; the injection port temperature is 230°C; the electron bombardment energy is 70 eV; and the ion source temperature is 230°C.

[0010] In some embodiments, solid-phase microextraction is also used to extract the standard sample and the sample to be tested separately.

[0011] In some embodiments, the extraction conditions are: SPME fiber; extraction temperature 35°C; pre-heating time 1 min; extraction time 5 min; and desorption time 5 min.

[0012] In some embodiments, a modified Jaccard similarity coefficient is used to compare the similarity and difference between the top 10 components in peak area of ​​the standard sample and the top 10 components in peak area of ​​the sample to be tested, and the existence of similarity is obtained.

[0013] In some embodiments, similarity is defined as the number of intersections between the top 10 components in peak area of ​​the sample to be tested and the top 10 components in peak area of ​​the standard sample, divided by 10.

[0014] In some embodiments, the Spearman rank correlation coefficient is used to calculate the order correlation between the top 10 components of the peak area of ​​the test sample and the perfect matching items among the top 10 components of the peak area of ​​the standard sample, and the ranking similarity is obtained.

[0015] In some embodiments, ranking similarity is calculated using the following method:

[0016] In the above formula, n is the number of grades, and d is the grade difference between the two paired variables.

[0017] In some embodiments, response similarity is evaluated by calculating the cosine of the angle between two vectors. The response similarity is calculated using the following method:

[0018] In the above formula, I exp To detect the sample response, I ref The response is that of the control sample.

[0019] In some embodiments, the standard sample is a sample that has been unanimously recognized by a panel of experts as having flavor characteristics representative of such products.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a consistency evaluation method for beverage products based on volatile substance fingerprinting. By comparing the volatile substance GC-MS full scan spectrum of each batch of products with that of a standard sample, the method evaluates whether the product quality meets the requirements of the product quality standard. This method is applicable to production quality control, product R&D optimization, and supply chain management, and improves the objectivity, accuracy, and efficiency of beverage product flavor evaluation. Attached Figure Description

[0021] Figure 1This is a GC-MS full scan (SCAN) image of volatile substances in a barley tea standard sample provided in an embodiment of the present invention. Figure 2 This is a GC-MS full scan (SCAN) image of volatile substances in barley tea sample #1 provided in an embodiment of the present invention. Figure 3 This is a GC-MS full scan (SCAN) image of volatile substances in barley tea sample #2 provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions in specific embodiments of the present invention will be described in detail and completely below. Obviously, the described embodiments are only some specific implementations of the overall technical solution of the present invention, and not all implementations. Based on the overall concept of the present invention, all other embodiments obtained by those skilled in the art fall within the protection scope of the present invention.

[0023] This invention provides a consistency evaluation method for beverage products based on volatile substance fingerprinting. GC-MS full scan analysis is performed on standard samples and test samples to obtain component peak areas. After screening, the top 10 components with the highest peak areas in both the standard and test samples are selected. Existence similarity, ordinal similarity, and response similarity analyses are performed on these top 10 components to obtain a comprehensive similarity score. The comprehensive similarity score is calculated using the following method: Overall similarity = Existence similarity × 0.7 + Ranking similarity × 0.2 + Response similarity × 0.1; If the overall similarity is greater than 80%, the sample to be tested can be identified as a sample similar to the standard sample, that is, a qualified sample.

[0024] Fingerprint spectroscopy evaluation method has wide applications in various fields (CN115639288B, CN119310227A, etc.), and its advantages are mainly reflected in the following aspects: 1. Holistic and comprehensive Multi-component simultaneous analysis: Fingerprinting avoids the limitations of traditional methods that rely on a single key substance, and can simultaneously analyze dozens to hundreds of chemical components, which can comprehensively reflect the intrinsic quality of products in quality control. Revealing synergistic effects of components: In the process of food and drug analysis and research and development, it can reflect the interaction between components (such as synergistic or antagonistic effects), which is closer to the actual flavor activity and reveals that multiple components synergistically affect the overall flavor of the sample. 2. Stability and anti-interference performance Batch consistency control: During the production process of flavored products such as beverages, the similarity of fingerprint patterns (≥0.90) can be calculated using the cosine similarity method to effectively monitor the flavor consistency of different batches of products. Environment: Fingerprint patterns have a high tolerance for differences in sample pretreatment, making them more suitable for industrial quality control; 3. High-precision identification capability Microscopic difference identification: Isomer differentiation, fingerprint spectrum can identify the difference in peak retention time of isomers; adulteration detection, fingerprint spectrum can accurately identify adulteration by the absence of characteristic peaks or abnormal peak heights. Product raw material origin traceability: Due to differences in soil trace elements, the peak area ratio of relevant components in the fingerprint spectrum of raw materials from different origins is significantly different; 4. High efficiency and high throughput Rapid batch detection: GC-MS technology can complete multi-component scanning of samples within minutes; Automation and digitalization: Fingerprint data can obtain a large amount of information about related components, significantly reducing information transmission costs; 5. Comprehensive quality control throughout the entire process It can realize closed-loop management from raw material identification and process monitoring to finished product evaluation. In summary, fingerprinting evaluation methods provide powerful analytical tools in multiple fields, helping to improve product quality and the efficiency of new product development.

[0025] The above-mentioned consistency evaluation method for beverage products based on volatile substance fingerprint spectrum compares the volatile substance GC-MS full scan spectrum of each batch of products with that of standard samples to evaluate whether the product quality meets the requirements of product quality standards. It is applicable to production quality control, product R&D optimization and supply chain management, and improves the objectivity, accuracy and efficiency of beverage product flavor evaluation.

[0026] It should be noted that traditional fingerprint spectral evaluation is based on the marker compounds of the control sample as a reference, comparing the test sample with the control sample, which is a targeted test. However, if the test sample contains key components not present in the control sample, this method cannot provide an objective similarity score. This invention, however, employs a similarity comparison step: comparing the top ten components with the highest peak area response in both the control and test samples, which is a non-targeted test. If an unexpectedly high-response component appears in the test sample, it can be quickly identified, and the similarity score will be significantly reduced, thus objectively evaluating the differences between the two samples.

[0027] The above technical solution comprehensively determines the similarity by considering presence similarity, order similarity, and response similarity. It evaluates the similarity between two samples from a macroscopic to a microscopic scale: First, it judges from a macroscopic perspective that the key components are present in both samples (presence similarity); second, it compares whether the order of the responses of the key components is consistent at the local level (order similarity); finally, it compares the responses of each key component at the microscopic level (response similarity). This comprehensive and multi-dimensional comparison mode is significantly better than the comparison of a single factor and largely avoids false negatives that occur in traditional fingerprint spectral identification.

[0028] Meanwhile, the above technical solution also provides coefficients for existence similarity, ranking similarity, and response similarity. It should be noted that, logically, existence similarity has the highest weight in determining these coefficients. First, it must be ensured that the key component is present in both samples; its absence constitutes a significant dissimilarity factor. Second, even when the key component is present, the ranking of its response values ​​(ranking similarity) is also important, but its weight is lower than existence similarity. Finally, comparing the absolute values ​​of the key component's responses in the two samples is relatively less important. The specific weight values ​​(0.7, 0.2, 0.1) are inferred from empirical values ​​and cannot be changed once determined, otherwise the measurement scale would change.

[0029] In some embodiments, GC-MS full scan analysis of standard samples and test samples includes: using a weakly polar capillary column; using helium as the carrier gas at a flow rate of 1.5 mL / min; the column temperature program is: holding at 40°C for 3 min, then increasing to 250°C at 3°C / min and holding for 10 min; the injection port temperature is 230°C; the electron bombardment energy is 70 eV; and the ion source temperature is 230°C.

[0030] The qualitative analysis of the above technical solutions was performed using the National Institute of Standards and Technology (NIST) standard mass spectrometry library.

[0031] GCMS Total Ion Chromatography (TIC) Peak Area Screening: GC-MS full scan analysis is used to obtain the peak areas of the components; these peak areas are then screened to remove noise signals such as column bleed; the peak areas are then sorted, and the top 10 components by peak area are selected for similarity analysis.

[0032] In some embodiments, solid-phase microextraction is also used to extract the standard sample and the sample to be tested separately. In some embodiments, the extraction conditions are: SPME fiber; extraction temperature 35°C; pre-heating time 1 min; extraction time 5 min; elution time 5 min.

[0033] In some embodiments, a modified Jaccard similarity coefficient is used to compare the similarity and differences between the top 10 components in peak area of ​​the standard sample and the top 10 components in peak area of ​​the test sample, to obtain the existence similarity. In some embodiments, the existence similarity = the number of intersections between the top 10 components in peak area of ​​the test sample and the top 10 components in peak area of ​​the standard sample / 10.

[0034] The modified Jaccard similarity coefficient is an indicator used to measure the difference between two sets, comparing the similarity and differences between finite sample sets. The higher the Jaccard coefficient value, the higher the sample similarity.

[0035] In some embodiments, the Spearman rank correlation coefficient is used to calculate the order correlation between the top 10 components of the peak area of ​​the test sample and the perfect match items among the top 10 components of the peak area of ​​the standard sample, thus obtaining the order similarity. In some embodiments, the order similarity is calculated by the following method:

[0036] In the above formula, n is the number of grades, and d is the grade difference between the two paired variables.

[0037] The Spearman's Rank Correlation is used to calculate ordinal correlation only for exact matches, making it suitable for measuring the monotonic relationship between two variables. To calculate the Spearman's coefficient, each matching element is assigned a rank in both lists, and then the correlation coefficient between these ranks is calculated. A coefficient of 1 indicates identical ranks, -1 indicates opposite ranks, and 0 indicates no correlation. This provides a quantified value for ranking similarity.

[0038] In some embodiments, response similarity is evaluated by calculating the cosine of the angle between two vectors. The response similarity is calculated using the following method:

[0039] In the above formula, I exp To detect the sample response, I ref The response is that of the control sample.

[0040] In some embodiments, the standard sample is a sample that has been unanimously recognized by a panel of experts as having flavor characteristics representative of such products.

[0041] In the production of flavored products such as beverages and alcoholic beverages, how to achieve rapid, objective, and low-cost evaluation of batch-to-batch flavor consistency, and overcome the shortcomings of traditional methods such as the subjectivity of sensory evaluation, the high cost of quantitative analysis (requiring a large number of standards to maintain calibration curves), the abstract and difficult-to-understand output results of mathematical models (such as PLS-DA), and the inability of electronic noses to distinguish structurally similar substances, has always been a problem in this field. The evaluation method provided by this invention is objective when used for evaluating the quality consistency of beverage flavored products: it uses instrument data to replace manual sensory evaluation; it is low-cost: it does not require quantitative analysis of standards, reducing maintenance costs; it is interpretable: the flavor index is easier for production personnel to understand than mathematical models; and it is universally applicable: it is suitable for various flavored products such as fruit juices, tea drinks, and beer. The above technical solution uses solid-phase microextraction combined with gas chromatography-mass spectrometry to perform qualitative analysis of complex sample flavors and screen key volatile components and design weighted (threshold) flavor indices, achieving rapid, accurate, efficient, and low-cost quantitative evaluation of sample flavor and its stability. It can be applied to the flavor design and optimization of products such as alcoholic beverages, beverages, and spices.

[0042] To provide a clearer and more detailed description of the beverage product consistency evaluation method based on volatile substance fingerprinting provided in this invention, specific embodiments will be described below.

[0043] Example 1 Original Barley Tea #1 I. Collection of Standard (Sample) Spectra (1) Determination of standard sample: Original barley tea with a production date of July 13, 2024, which was unanimously recognized by the six judges of the evaluation expert group.

[0044] (2) Preparation of standard samples: Take 5 mL of sample and place it in a 20 mL headspace vial, tighten the cap, and place the sample vial in the sample tray of the autosampler for analysis.

[0045] (3) Solid phase microextraction conditions: SPME fiber 65µm DVB / CAR / PDMS; extraction temperature 35℃; preheating time 1min; extraction time 5min; desorption time 5min.

[0046] (4) Gas chromatography-mass spectrometry conditions: The analytical capillary column used was a weakly polar column (30 m long, 0.32 mm inner diameter, 0.25 μm film thickness). The carrier gas was helium, with a flow rate of 1.5 mL / min. The column temperature program was: 40 °C for 3 min, then increased to 250 °C at 3 °C / min and held for 10 min. The injection port temperature was 230 °C. The electron impact energy was 70 eV, and the ion source temperature was 230 °C.

[0047] (5) Qualitative analysis: National Institute of Standards and Technology (NIST) Standard Mass Spectrometry Library.

[0048] (6) GCMS Total Ion Chromatography (TIC) Peak Area Screening: The peak areas of the components are obtained by GC-MS full scan analysis, as shown in the figure. Figure 1 As shown in Table 1, the samples were screened to remove noise signals such as column loss. The component with the largest peak area was selected as the base peak, with a response value of 999. The other components were then sorted according to their relative peak areas to the base peak, and the top 10 components with the largest peak areas were selected for similarity analysis.

[0049] Table 1. Top 10 components by relative peak area in standard samples

[0050] II. Collection of Spectra of Test Samples (1)-(6) Same as above, GC-MS full scan analysis yields the component peak areas as follows Figure 2 As shown.

[0051] (7) The top 10 components of the relative peak area of ​​the test sample and the top 10 components of the relative peak area of ​​the standard sample are compared using the Jaccard similarity coefficient, as shown in Table 2.

[0052] Table 2 Comparison of the top 10 components with relative peak area in test sample 1# and standard sample.

[0053] Similarity exists = intersection / 10 = 9 / 10 = 0.9 (*Since the number of components in the two datasets involved in this invention is fixed at 10, the standard Jaccard similarity coefficient = intersection / union is transformed here into similarity = intersection / 10) Hexanal was found only in the benchmark sample, while nonanal was found only in test sample #1. Therefore, these two compounds will be eliminated in the next round of comparison.

[0054] (8) The ranking similarity ρ was calculated using Spearman's Rank Correlation, as shown in Table 3.

[0055] Table 3 Comparison of the top 10 components by relative peak area in the test samples and standard samples.

[0056]

[0057]

[0058] The similarity of this sample order is 0.55, indicating that there is a moderate positive correlation between the order of the two columns.

[0059] (9) The response similarity of the overlapping components in the top 10 components of the relative peak area of ​​the two samples is calculated based on the cosine similarity, as shown in Table 4.

[0060] Table 4 Comparison of peak areas of overlapping components among the top 10 components in relative peak area of ​​test samples and standard samples.

[0061] Dot product calculation

[0062] (999×920)+(291×460)+(199×809)+(189×999)+(160×308)+(141×261)+(123×553)+(83×460)+(70×329)=1618052 Length calculation

[0063] Test sample vector magnitude =

[0064] Standard sample vector magnitude =

[0065] Response similarity =(1618052) / (1870.6×1109.2)≈0.78 (10) Calculation of comprehensive similarity The overall similarity between the fingerprint spectrum of sample #1 (Maixiang) and the standard barley tea is calculated as follows: Existence Similarity × 0.7 + Ranking Similarity × 0.2 + Response Similarity × 0.1 = 0.9 × 0.7 + 0.55 × 0.2 + 0.78 × 0.1 ≈ 81.8%, which meets the requirements for qualified samples.

[0066] (11) Verification Using Agilent's MS Similarity Calculation Software, the original barley tea #1 was compared with the benchmark sample, and the similarity was 88.1%, which is similar to the result in this case.

[0067] Example 2 Original Barley Tea #2 I. Collection of Standard (Sample) Spectra Same as original barley tea #1 II. Collection of Spectra of Test Samples (1)-(6) are the same as the original barley tea 1#(1)-(6). The peak areas of the components were obtained by GC-MS full scan analysis as follows: Figure 3 As shown.

[0068] (7) The Jaccard similarity coefficient was used to compare the top 10 components of the relative peak area of ​​the test sample with the top 10 components of the relative peak area of ​​the standard sample, as shown in Table 5. Hexanal appeared only in the benchmark sample; while 2-ethyl-5-methylpyrazine appeared only in test sample 2#, so these two compounds will be eliminated in the next round of comparison.

[0069] Table 5 Comparison of the top 10 components with relative peak area in test sample #2 and standard sample.

[0070] Similarity exists = intersection / 10 = 9 / 10 = 0.9 (*Since the number of components in the two datasets involved in this invention is fixed at 10, the standard Jaccard similarity coefficient = intersection / union is transformed here into similarity = intersection / 10) (8) The similarity of the ranking was calculated using Spearman's Rank Correlation, and the result was ρ=0.477.

[0071] (9) The response similarity MF = 0.78 of the overlapping components in the top 10 components of the relative peak area of ​​the two samples is calculated according to the cosine similarity, as shown in Table 6.

[0072] Table 6 Comparison of peak areas of overlapping components among the top 10 components with relative peak areas in test sample 2# and standard sample.

[0073] Dot product calculation =1572336 Standard sample vector magnitude =1109.235322 Test sample vector magnitude =1864.985255 Response similarity =0.782156001 (10) Calculation of comprehensive similarity The overall similarity of the fingerprint spectrum of sample 2# with that of the standard sample is approximately 80% (0.9×0.7+0.477×0.2+0.78×0.1), which meets the requirements for qualified samples.

[0074] (11) Verification Using Agilent's MS Similarity Calculation Software, the original barley tea #2 was compared with the benchmark sample, and the similarity was 86.2%, which is similar to the result in this case.

[0075] Comparative Example 1 Chinese patent CN 115639288 B, entitled "A method for determining the GC fingerprint spectrum of Ganlu disinfectant pills and its application", is used as comparative example 1.

[0076] Comparative Example 1 mainly relates to a gas chromatography (GC) fingerprinting method for the traditional Chinese medicine preparation Ganlu Xiaodu Wan and its application in quality control. Ganlu Xiaodu Wan is a traditional Chinese medicine compound preparation composed of 11 medicinal materials (such as agastache, forsythia, and scutellaria), used for clearing heat and detoxifying, and aromatically resolving dampness. The core innovation of the patent lies in establishing a highly efficient and reliable gas chromatography fingerprinting analysis method, which solves the problem of insufficient qualitative analysis of volatile components in existing technologies, and realizes comprehensive quality control from raw materials to finished products. The following is a structured summary from four aspects: method, application, validation, and advantages.

[0077] The biggest difference between this invention and Comparative Example 1 lies in their technical objectives and methodologies: this invention targets flavor consistency in the food and beverage industry, employing a comprehensive similarity algorithm that does not require standard substances; Comparative Example 1 targets traditional Chinese medicine preparations, establishing a fingerprint spectral system that includes quantification and attribution. This difference makes this invention more suitable for industrialized flavor product quality inspection in terms of cost, efficiency, and interpretability, while Comparative Example 1 is more suitable for the entire process of pharmaceutical quality control.

[0078] Comparative Example 2 Chinese patent with publication number CN 119310227 A, entitled "A vapor phase fingerprint spectrum of a transverse diaphragm and its construction method and application", is used as comparative example 2.

[0079] Comparative Example 2 addresses the challenge of identifying *Calophyllum membranaceum*, a plant in the Clusiaceae family, by innovatively establishing a fingerprinting method based on gas chromatography (GC). *Calophyllum membranaceum*, a traditional Chinese medicine with properties of dispelling wind and dampness and strengthening bones and muscles, is often difficult to accurately identify due to its similar appearance to other plants in the same genus (such as *Calophyllum rubrum* and *Calophyllum rubrum*). Traditional thin-layer chromatography methods suffer from limitations such as limited separation capacity and susceptibility to environmental interference. This patent achieves rapid and accurate identification through five specific chromatographic peaks.

[0080] Comparative Example 2 focuses more on the static identification of the authenticity of Chinese medicinal materials, while this invention solves the beverage industry's need for rapid, objective, and low-cost consistency evaluation through a non-quantitative multi-dimensional similarity algorithm. The two are fundamentally different in terms of target field, method design, and practicality.

[0081] Comparative Example 3 The Chinese patent with publication number CN 119355152 A, entitled "Construction method and application of fingerprint spectrum of Chushi Weiling Decoction", is used as comparative example 3.

[0082] Comparative Example 3 addresses the quality control challenges of the traditional Chinese medicine compound preparation, Chushi Weiling Decoction, by innovatively establishing a fingerprinting method based on high-performance liquid chromatography (HPLC). Chushi Weiling Decoction is composed of 14 traditional Chinese medicines (such as tangerine peel, gardenia, and saposhnikovia root), possessing the effects of clearing heat and dampness, strengthening the spleen, and promoting diuresis. Traditional methods struggle to comprehensively monitor its complex components. This invention, through optimizing pretreatment, chromatographic conditions, and peak assignment strategies, constructs a standard fingerprint spectrum containing 21 characteristic peaks, achieving end-to-end quality monitoring from raw materials to finished product.

[0083] Comparative Example 3 describes a component stability control method for the traditional Chinese medicine industry, relying on fixed characteristic peak matching and reference standard verification, focusing on the overall quality control of repeated formulations. In contrast, this invention is a rapid flavor consistency detection method for the food and beverage industry, achieving low-cost, highly interpretable batch monitoring through a non-quantitative, multi-dimensional similarity algorithm; the two differ fundamentally in their technical approaches and application scenarios.

[0084] Comparative Example 4 Chinese patent with publication number CN 119510630 A and title "Method for establishing and application of HPLC fingerprint of tanshinone polysaccharide" was used as comparative example 4.

[0085] Comparative Example 4 addresses the issue that the polysaccharide composition of Salvia miltiorrhiza is complex due to differences in origin, making it difficult for traditional detection methods to comprehensively control its quality. An innovative method for fingerprinting Salvia miltiorrhiza polysaccharides based on high performance liquid chromatography (HPLC) was established. Combined with the spectral-activity relationship analysis of antioxidant and anticoagulant activities, this provides a scientific basis for the quality evaluation of Salvia miltiorrhiza.

[0086] Comparative Example 4 is an in-depth analysis method for research and quality control of traditional Chinese medicine, focusing on the discovery of component-efficacy correlations. In contrast, this invention is a rapid flavor detection tool for food and beverage production lines, with low cost and high interpretability as its core. The two present a stark contrast in their technical approaches: the former pursues efficiency and practicality, while the latter pursues academic precision.

[0087] Comparative Example 5 The Chinese patent with publication number CN 119534734 A, entitled "Construction method and application of fingerprint spectrum of Angelica sinensis", is used as comparative example 5.

[0088] Comparative Example 5 solved the quality control problem of multiple components of Angelica sinensis decoction by using dual-wavelength HPLC fingerprinting and multi-dimensional verification. It is both scientific (rigorous methodology) and practical (suitable for industrial quality control), providing a model for the quality evaluation of traditional Chinese medicine compound preparations.

[0089] Comparative Example 5 presents a component-pharmacodynamic correlation analysis method for traditional Chinese medicine research, emphasizing academic depth. In contrast, this invention offers a low-cost, rapid flavor detection solution for food and beverage production lines, prioritizing efficiency and interpretability. The two complement each other in their target areas (food vs. traditional Chinese medicine), technology selection (GC-MS vs. HPLC), and result application (batch consistency vs. pharmacodynamic traceability).

[0090] Comparative Example 6 Chinese patent with publication number CN 113295789 A, entitled "A method for establishing and detecting UPLC fingerprint spectrum of Wanglaoji herbal tea concentrate", is used as comparative example 6.

[0091] Although both prior art document 6 (CN113295789A) and this invention involve fingerprinting technology, they differ significantly in their application areas and technical positioning: Prior art document 6 focuses on the quality control of Wanglaoji herbal tea concentrate, belonging to the field of plant-based beverage quality control, and applies a component-oriented quality control method. It establishes fingerprints of 14 characteristic components using UPLC technology, representing traditional fingerprinting technology and emphasizing the qualitative and quantitative analysis of specific components. This invention, on the other hand, targets batch consistency evaluation for flavored products such as beverages and alcoholic beverages, with a wider range of applications (fruit juice, carbonated drinks, tea drinks, etc.). It belongs to a flavor-oriented consistency evaluation method, analyzes the overall similarity of volatile flavor components using GC-MS, and proposes an innovative evaluation algorithm to achieve flavor consistency evaluation through triple similarity calculation.

[0092] Comparative Example 7 Chinese patent with publication number CN 111487343 A and title "A method for establishing the fingerprint spectrum of Baoyuantang preparation" is used as comparative example 7.

[0093] Comparative Example 7 demonstrates the rigor and scientific nature of quality control in traditional Chinese medicine, while this invention highlights the need for rapid, low-cost solutions in industrial production. The two present a stark contrast in their technical approaches and application scenarios, collectively enriching the application scope and methodological system of fingerprinting technology.

[0094] Existing patents primarily based on fingerprinting mainly focus on quality monitoring of traditional Chinese medicine (TCM). Their common characteristics include: multi-component holistic quality control, overcoming the limitations of single indicators; patents generally emphasize that traditional single-component detection (such as content determination) is insufficient to comprehensively reflect the quality of complex systems (such as TCM compound formulas), while fingerprinting achieves holistic evaluation by simultaneously analyzing multiple components (10–30 common peaks); chemical component identification and attribution: most patents explicitly identify the chemical components corresponding to key peaks (such as reference standard verification) and analyze their source medicinal materials; standardized methods: chromatographic conditions are highly uniform, with liquid chromatography being the most common method; patent applications are mainly used for authenticity verification and quality evaluation, process optimization and standard setting; multiple chemometric applications: 20% of patents introduce chemical pattern recognition (such as cluster analysis and principal component analysis) to assist in quality evaluation.

[0095] The main innovation of this invention lies in its core algorithm, which employs comprehensive similarity calculation and uses a triple ( This invention uses a weighted evaluation of flavor consistency based on similarity (existence similarity, ranking similarity, and response similarity). It overcomes the shortcomings of subjective sensory evaluation, high cost of quantitative analysis, low practicality of model discrimination, and abstract and difficult-to-understand output. This invention is mainly applied to flavor products such as beverages and alcoholic drinks (e.g., juice, tea, beer), evaluating batch-to-batch flavor consistency through volatile fingerprinting. The core objective is to solve the quality control problem of flavor stability in industrial production, emphasizing rapid and low-cost application in supply chain management.

[0096] The innovative advantages of this invention are: algorithm universality: the comprehensive similarity method is applicable to a variety of flavor products (juice, tea, etc.) without the need to adjust core parameters; low cost and high efficiency: it eliminates the need for standard product procurement and calibration curve maintenance, making it suitable for high-throughput industrial quality inspection; strong interpretability: the numerical results (comprehensive similarity %) are easier for non-professionals to understand than mathematical models.

Claims

1. A method for evaluating the consistency of beverage products based on volatile substance fingerprinting, characterized in that, GC-MS full scan analysis was performed on the standard sample and the test sample to obtain the peak area of ​​each component. After screening, the top 10 components with the largest peak areas in both the standard sample and the test sample were selected. Existence similarity, ordinal similarity, and response similarity analyses were performed on these top 10 components to obtain the comprehensive similarity score. The comprehensive similarity score was calculated using the following method: Overall similarity = Existence similarity × 0.7 + Ranking similarity × 0.2 + Response similarity × 0.1; If the overall similarity is greater than 80%, the sample to be tested can be identified as a sample similar to the standard sample, that is, a qualified sample; GC-MS full scan analysis of standard samples and test samples was performed separately, including: using a weakly polar capillary column; using helium as the carrier gas at a flow rate of 1.5 mL / min; the column temperature program was: 40℃ for 3 min, then increasing to 250℃ at 3℃ / min and holding for 10 min; the injection port temperature was 230℃; the electron bombardment energy was 70 eV; and the ion source temperature was 230℃. It also includes using solid-phase microextraction to extract standard samples and test samples separately; Extraction conditions were: SPME fiber; extraction temperature 35℃; pre-heating time 1 min; extraction time 5 min; desorption time 5 min. The beverage product is barley tea.

2. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 1, characterized in that, A modified Jaccard similarity coefficient was used to compare the similarity and difference between the top 10 components in peak area of ​​the standard sample and the top 10 components in peak area of ​​the sample to be tested, and the similarity was obtained.

3. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 2, characterized in that, Similarity = Number of intersections between the top 10 components in peak area of ​​the sample to be tested and the top 10 components in peak area of ​​the standard sample / 10.

4. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 1, characterized in that, The Spearman rank correlation coefficient was used to calculate the order correlation between the top 10 components of the peak area of ​​the test sample and the top 10 components of the peak area of ​​the standard sample that were completely matched, and the ranking similarity was obtained.

5. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 4, characterized in that, Ranking similarity is calculated using the following method: In the above formula, n is the number of grades, and d is the grade difference between the two paired variables.

6. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 1, characterized in that, Response similarity is evaluated by calculating the cosine of the angle between two vectors. The response similarity is calculated using the following method: In the above formula, I exp To detect the sample response, I ref The response is that of the control sample.

7. The method for evaluating the consistency of beverage products based on volatile substance fingerprinting according to claim 1, characterized in that, The standard sample is a sample whose flavor is unanimously recognized by the expert evaluation panel as representative of the characteristics of this type of product.

Images