A digital traceability assessment method and system for mulberry processed products
By constructing a database of mulberry variety characteristics and a multi-dimensional quantitative correlation model, the systematic management problem of mulberry variety characteristic traceability was solved, and the accurate traceability assessment and processing suitability analysis of mulberry processed products were realized, improving the accuracy and operability of the assessment.
Patent Information
- Application Number
- CN202611577313.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-13
- Publication Date
- 2026-11-17
AI Technical Summary
Existing fruit traceability methods lack systematic data management and rapid retrieval capabilities for mulberry variety characteristics, cannot fully characterize variety characteristics, and have not established a quantitative correlation between variety characteristics and processing technology, making it difficult to assess the suitability of processing and product consistency.
A database of mulberry variety characteristics was constructed, a standardized index collection system was defined, a variety characteristic matching algorithm was designed, the evaluation process was integrated, and a multi-dimensional quantitative correlation model of sugar content, acidity, anthocyanin content, juice yield, and seed ratio was established. A comprehensive score was generated using a matching algorithm based on Euclidean distance and cosine similarity.
It enables precise variety traceability and processing suitability assessment of mulberry processed products, improves the accuracy and operability of traceability assessment, and provides data support for process optimization.
Smart Images

Figure CN123118128A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of production traceability, specifically relating to a digital traceability assessment method and system for mulberry fruit processing products. Background Technology
[0002] With the rapid development of the food industry, food traceability technology has become an important means of ensuring product quality. The quality of processed mulberry products is closely related to the characteristics of the raw material varieties, but existing fruit traceability methods are mainly designed for common varieties, only collecting conventional indicators such as sugar content and acidity, and have not yet established a special traceability evaluation system for the characteristics of mulberry varieties.
[0003] The main shortcomings of existing technologies are: the lack of systematic data management and rapid retrieval capabilities for mulberry variety characteristics, making it impossible to fully characterize the varietal characteristics of mulberries; at the same time, the lack of a quantitative correlation between varietal characteristics and processing technology makes it difficult to assess the differences in adaptability of different varieties to the processing process, and also makes it impossible to accurately judge the traceability consistency between processed products and raw material varieties.
[0004] Therefore, there is an urgent need to establish a digital traceability assessment method for mulberry varieties. This method can be developed by constructing a variety characteristic database, establishing a standardized indicator collection system, designing a variety characteristic matching degree algorithm, and integrating the assessment process to achieve accurate variety traceability and processing suitability assessment of mulberry processed products. Summary of the Invention
[0005] The purpose of this invention is to provide a digital traceability evaluation method and system for mulberry fruit processing products, which can effectively solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Firstly, a digital traceability assessment method for mulberry fruit processed products includes the following specific steps: A database of mulberry variety characteristics was constructed, and the core fields of the database include sugar content, acidity, anthocyanin content, juice yield, and seed ratio. Establish a key mulberry variety indicator collection system, and define standardized collection methods and testing procedures for the sugar content, acidity, anthocyanin content, juice yield and seed ratio, respectively; A correlation model between varietal characteristics and processing adaptability is constructed, which includes a sugar content fermentation performance adaptation model, an acidity taste coordination model, an anthocyanin color stability model, and a juice yield process efficiency model. Design a variety feature matching algorithm to quantitatively compare the detection indicators of the product to be traced with the standard variety features in the database; Integrate the digital traceability assessment process for mulberries to generate a comprehensive traceability assessment report.
[0007] Secondly, a digital traceability and evaluation system for mulberry processed products includes: The database construction module is used to construct a mulberry variety characteristic database. The core fields of the mulberry variety characteristic database include sugar content, acidity, anthocyanin content, juice yield, and seed ratio. The indicator acquisition module is used to establish an indicator acquisition system for key mulberry varieties, and to define standardized acquisition methods and detection procedures for the sugar content, acidity, anthocyanin content, juice yield and seed ratio, respectively. The association model construction module is used to construct an association model between varietal characteristics and processing adaptability. The association model between varietal characteristics and processing adaptability includes a sugar content fermentation performance adaptability model, an acidity taste coordination model, an anthocyanin color stability model, and a juice yield process efficiency model. The matching degree algorithm module is used to design a variety feature matching degree algorithm to quantitatively compare the detection indicators of the product to be traced with the standard variety features in the mulberry variety feature database. The assessment integration module is used to integrate the digital traceability assessment process for mulberry fruit and generate a comprehensive traceability assessment report.
[0008] In summary, this application includes at least one of the following beneficial technical effects: 1. This invention establishes a complete varietal characteristic index system and standardized collection process specifically for mulberries, covering five core indicators: sugar content, acidity, anthocyanin content, juice yield, and seed ratio. It also designs targeted detection methods for the unique attributes of mulberries, such as high sugar content, easy browning, and small seeds, to make up for the lack of indicators in general fruit traceability methods and make the description of varietal characteristics more comprehensive and accurate.
[0009] 2. This invention constructs a multi-dimensional quantitative correlation model between varietal characteristics and processing suitability, including four sub-models: sugar content fermentation performance, acidity and taste coordination, anthocyanin color stability, and juice yield and process efficiency. It establishes a mathematical mapping between raw material varietal indicators and processing effects, so that traceability assessment can not only determine varietal consistency, but also predict the suitability of different varieties for the target processing technology, providing data support for process optimization.
[0010] 3. This invention designs a variety feature matching algorithm that integrates Euclidean distance and cosine similarity, and combines it with a weighted summation model to generate comprehensive scores and grading conclusions. This realizes a complete closed-loop process from sample information reception, index collection, feature comparison, suitability assessment to assessment report generation, which significantly improves the accuracy, operability and interpretability of mulberry fruit processing product traceability assessment. Attached Figure Description
[0011] Figure 1 A schematic diagram of the overall technical solution for the digital traceability assessment method for mulberry fruit processed products.
[0012] Figure 2 This is a schematic diagram illustrating the principle framework of the core evaluation algorithm.
[0013] Figure 3 This is a logical flowchart of the data acquisition and preprocessing stage.
[0014] Figure 4 A schematic diagram of the multi-level interaction relationships and data flow for tracing information association and evaluation analysis.
[0015] Figure 5 A diagram illustrating the comparison between the digital traceability assessment method for mulberry processed products and existing assessment methods. Detailed Implementation
[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the following description is provided in conjunction with the appendix. Figure 1 To be continued Figure 5 The present invention will be further described in detail below with reference to specific embodiments.
[0017] Firstly, this invention provides a digital traceability assessment method for mulberry processed products, encompassing the following core steps: Step S1, constructing a mulberry variety characteristic database; Step S2, establishing a key mulberry variety indicator collection system; Step S3, constructing a correlation model between variety characteristics and processing suitability; Step S4, designing a variety characteristic matching degree algorithm; and Step S5, integrating the mulberry digital traceability assessment process. Through the systematic integration and coordinated operation of the above steps, the entire chain of mulberry processed products, from raw materials to finished products, can be traced for variety characteristics and accurately assessed for traceability.
[0018] Firstly, regarding step S1 in the method of this embodiment, namely, constructing a mulberry variety characteristic database, which is used to establish a database system containing core indicator fields of mulberry varieties to provide a data foundation for subsequent variety indicator comparison and traceability evaluation, this is specifically achieved through the following sub-steps: Step S101 establishes the multi-dimensional data organization structure of the database. The core fields of this data structure include five varietal indicators: sugar content, acidity, anthocyanin content, juice yield, and seed percentage. The sugar content index is set to a range of 18% to 24%. This range was determined through statistical analysis of sugar content measurements of at least 20 main mulberry varieties, including the Yuezhen series, Da Shi, and Hongguo series, in major mulberry producing areas in China. This range covers the main sugar content distribution range of mulberry raw materials and can effectively characterize the sugar accumulation characteristics of different varieties.
[0019] Acidity was characterized by titratable acid content, determined by acid-base titration of the total organic acids in the sample, with the result expressed as grams of malic acid per 100 grams of sample. Anthocyanin content was characterized by its concentration per unit mass of fruit pulp, providing crucial reference for judging the quality of processed mulberry products.
[0020] Juice yield is calculated as the percentage of juice mass to raw material mass after juicing, and is used to measure the juicing performance of mulberry raw materials. Seed percentage is calculated as the percentage of seed mass to total fruit mass, and characterizes the distribution characteristics of seed material in mulberry fruit.
[0021] Step S102: Design a composite index structure for the database, which adopts a hybrid design pattern of hash index and B-tree index. The hash index is used to support exact matching queries for variety identifiers, enabling fast location within constant time complexity.
[0022] B-tree indexes are used to support numerical range queries and feature vector similarity comparisons, maintaining a stable logarithmic time complexity when performing operations such as sugar content range retrieval, acidity threshold filtering, and anthocyanin content sorting. Through this composite index structure, the database can automatically select the optimal execution path when processing mixed requests containing both exact matches and range queries, ensuring efficient operation of the fast retrieval of variety characteristics and feature vector similarity comparison functions.
[0023] Step S103: Configure the distributed storage architecture of the database. This architecture adopts a distributed key-value storage design, with the underlying layer based on a consistent hashing algorithm to achieve partitioned storage and load balancing of data. Variety characteristic records are stored serially in JSON format. Each record includes a variety identifier field, a collection timestamp field, a sugar content scalar field, an acidity scalar field, an anthocyanin content scalar field, a juice yield scalar field, a seed percentage scalar field, an origin attribute field, and a processing type field.
[0024] The variety identifier field is generated using UUID format to ensure the global uniqueness of each record. The data collection timestamp field supports historical data backtracking and querying. The origin attribute field is stored using administrative division codes, enabling origin tracing and statistical analysis according to provincial, municipal, and county-level administrative regions. The processing type field enumerates process paths such as fresh consumption, pulping, fermentation, concentration, and drying, providing process context information for subsequent processing suitability analysis.
[0025] Step S104: Set the database capacity and performance indicators. The database storage capacity is designed to be no less than 100,000 varietal characteristic records, covering common varietal types and their multiple batches of test data from major mulberry producing areas in China. The target response time for a single retrieval is less than 500 milliseconds. This indicator is ensured by the horizontal scalability of the distributed architecture and the query optimization strategy of the composite index structure.
[0026] The database cluster adopts a three-master, six-slave topology and improves the overall concurrent processing capability of the system through a read-write separation mechanism.
[0027] Step S105: Deploy the data acquisition and preprocessing module. This module is responsible for aggregating raw data on variety characteristics from multiple data sources. These data sources include variety identification experimental data from agricultural research institutions, testing report data from agricultural product quality testing centers, raw material warehousing testing data from mulberry processing enterprises, and real-time data collected by the traceability assessment system.
[0028] Data from different data sources is processed by a unified adaptation layer and converted into standardized JSON format records. Then, it undergoes preprocessing steps such as data cleaning, quality verification, and format validation before being written to the corresponding shard nodes of the distributed storage cluster.
[0029] The data cleaning stage identifies and handles missing values, outliers, and duplicate records. For missing values, if any field among the five core variety indicators is missing, the record is marked as an outlier and not written to the database. For outlier identification, the system uses the interquartile range (ICM) method for statistical outlier detection. For a specific variety indicator field currently being processed, let its sample dataset be... Calculate the first quartile of this dataset. and the third and fourth quartiles Let the interquartile range be: When a certain data point A data point is considered an outlier if the following inequality is satisfied: in, This represents the sample dataset corresponding to a certain indicator field of a variety currently being processed, containing the measurement values of multiple historical detection records in that indicator dimension; This represents the first quartile of the dataset, meaning that 25% of the sample values in the dataset are less than this value. The specific calculation method is as follows: The data in the dataset is sorted in ascending order, and the value at the 25th percentile is taken. This represents the third quartile of the dataset, meaning that 75% of the sample values in the dataset are less than this value. The specific calculation method is as follows: The data is sorted in ascending order, and the value at the 75th percentile is taken.
[0030] Represents the interquartile range, equal to and The difference is used to characterize the degree of dispersion of the dataset; This indicates the specific measured value of the currently tested sample in a certain variety indicator field.
[0031] Records identified as outliers are forwarded for manual review and are not automatically written to the database. For duplicate records, the system uses the combination of the variety identifier field and the collection timestamp field as the uniqueness criterion; if both are the same, the record is considered a duplicate and is discarded.
[0032] During system initialization, manually verified standard variety data is pre-entered as an initial sample set for calculating the initial values of each indicator field. and Thresholds. As the number of records entering the database accumulates, the system periodically recalculates the quartile thresholds of each indicator using recent historical data, thereby dynamically updating the judgment criteria.
[0033] The quality verification process performs rule-based validation based on the value constraints of each indicator field to ensure the accuracy and reasonableness of the data entering the database. Specifically, the sugar content indicator is verified to ensure it falls within the range of 18% to 24%. The format verification process checks the validity of the JSON structure and the correctness of the field types to prevent incorrectly formatted data from entering the database system.
[0034] The database provides a data write interface, supporting the submission of test data by third-party testing devices and data acquisition terminals. This interface supports both single-record writing and batch record writing modes.
[0035] To more clearly illustrate the above data acquisition and preprocessing process, the following description uses a specific example: A mulberry sample was tested and obtained five index values: sugar content 21.35%, acidity 1.26 g / 100g, anthocyanin content 186.5 mg / 100g, juice yield 71.0%, and seed content 4.3%. After the test data was entered into the system, the five core fields were first checked to ensure they were complete and not empty.
[0036] Then, outlier detection is performed: taking sugar content as an example, the system retrieves historical sugar content data for this variety to form a sample dataset. ,set up , ,but The lower bound threshold is The upper threshold is The current sugar content of 21.35% is within the normal range, and the other four indicators also passed outlier detection. The system then checked the combination of variety identifier and collection timestamp and found no duplicate records. After all quality and format verifications passed, the record was submitted to the database cluster via the data writing interface to complete the data entry.
[0037] The verification data of the above preprocessing process in actual operation is as follows: 500 mulberry variety detection records were processed continuously, the missing value detection rate was 100%, the outlier false positive rate was 8.3%, and the total processing time was about 4.2 seconds, which can meet the data entry efficiency requirements in actual business scenarios.
[0038] Through the above step S1, a mulberry variety characteristic database containing multi-dimensional variety indicator fields is constructed. This database has distributed storage, fast retrieval and data preprocessing capabilities, providing data support for subsequent variety characteristic matching and traceability evaluation.
[0039] For step S2, a key mulberry variety indicator collection system was established, defining standardized collection methods and testing procedures for the five key variety indicators: sugar content, acidity, anthocyanin content, juice yield, and seed ratio. This ensures that the indicator data obtained in the subsequent traceability assessment are comparable and repeatable, and is carried out according to the following steps.
[0040] Step S201: Standardized collection of sugar content index was performed using a handheld digital refractometer. Before the measurement, the mulberry sample was juiced, then centrifuged at 3000 rpm for 10 minutes using a low-speed centrifuge. The supernatant was collected and filtered through a 0.45-micron filter membrane to remove impurities, obtaining a clear juice solution for testing.
[0041] The ambient temperature for measurement should be controlled within the range of 20 to 25 degrees Celsius. If the ambient temperature exceeds this range, a constant temperature water bath should be used to adjust the sample temperature. Before use, the handheld digital refractometer should be zero-calibrated with distilled water and then calibrated and verified with a 20% sucrose standard solution.
[0042] Each measurement involved three readings, with the average value taken as the result. Each sample underwent three parallel measurements, and the arithmetic mean of the three parallel measurements was used as the sugar content value. All measurement results were formatted as a percentage to two decimal places. The measurement record included the sample number, collection time, personnel involved, equipment number, ambient temperature, and original measurement data.
[0043] Step S202: Standardize the collection of acidity index and determine the titratable acid content using acid-base titration. Take 20g of mulberry fruit sample and place it in a tissue homogenizer. Add an equal amount of distilled water and homogenize for 2 minutes. Transfer the homogenate to an Erlenmeyer flask and dilute with distilled water to 100ml.
[0044] Add 2 to 3 drops of phenolphthalein indicator to the conical flask and titrate with a 0.1 mol / L sodium hydroxide standard solution. Shake the conical flask continuously during titration until the solution turns a faint pink color that does not fade within 30 seconds; this is the titration endpoint. Record the volume of sodium hydroxide standard solution consumed.
[0045] The calculation of titratable acid content is based on the stoichiometric relationships in acid-base titration. The organic acids in mulberries are mainly represented by malic acid, which is a dicarboxylic acid; each mole of malic acid can neutralize 2 moles of sodium hydroxide. This formula originates from the basic principles of quantitative analysis using acid-base titration in analytical chemistry, and this step directly applies this classic calculation formula. The titratable acid content is calculated according to the formula: in It indicates the titratable acid content, with the unit being grams per 100 grams of sample. The value is the total amount of organic acids, calculated as malic acid, in every 100 grams of sample. This indicates the volume of sodium hydroxide standard solution consumed in the titration, in milliliters. This value is obtained directly during the titration process. The exact concentration of the sodium hydroxide standard solution is expressed in moles per liter, which is 0.1 moles per liter in this step. The conversion factor for malic acid is 0.067, and the unit is grams per millimole. This factor is derived from the conversion of the molar mass of malic acid (134.09 grams per mole) by dividing it by 2000 millimoles per mole and then by 1000 milligrams per gram. Its physical meaning is the mass of malic acid corresponding to one millimole of sodium hydroxide.
[0046] This indicates the dilution factor. In this step, 20 grams of sample is diluted to 100 ml and then used entirely for titration, so the dilution factor is 1, which is dimensionless. The sample mass is expressed in grams; in this step, it is 20 grams. Each sample is measured in triplicate, with the relative standard deviation controlled within 2%.
[0047] Step S203 involves standardized data collection for anthocyanin content, followed by spectrophotometric determination using the pH differential method. This method leverages the principle that anthocyanins exhibit different absorbance characteristics under different pH conditions to achieve quantitative analysis. This is a classic method in analytical chemistry for determining anthocyanin content, based on the Lambert-Beer law. Two measurement media are prepared: a pH 1.0 buffer solution is prepared from a 0.2 mol / L hydrochloric acid-potassium chloride solution, and a pH 4.5 buffer solution is prepared from a 0.2 mol / L acetate-sodium acetate solution.
[0048] Take 5 grams of mulberry fruit sample, add 20 ml of methanol solution containing 1% hydrochloric acid, and extract by shaking at 25 degrees Celsius in the dark for 1 hour. Then centrifuge at 4000 rpm for 10 minutes and take the supernatant to make up to 25 ml.
[0049] Take 2 ml of the extract and dilute to 25 ml with pH 1.0 buffer. Take another 2 ml of the extract and dilute to 25 ml with pH 4.5 buffer. Measure the absorbance of the sample extract at 520 nm in both buffer solutions.
[0050] The anthocyanin content is calculated using the following formula, which is directly derived from the classical analytical chemistry principle of the pH differential method: in The value indicates the anthocyanin content, expressed in milligrams per 100 grams of pulp. It represents the anthocyanin mass equivalent to cyanidin-3-glucoside in the sample. This represents the difference in absorbance values measured at pH 1.0 and pH 4.5, i.e. , which is a dimensionless absorbance difference that eliminates the interference of non-anthocyanin substances on absorbance; The value represents the average molecular weight of cyanidin-3-glucoside, which is 449.2 grams per mole. This value is the molar mass constant of cyanidin-3-glucoside reported in the literature. The dilution factor, in this step, is the ratio of the extraction liquid volume of 25 ml to the extraction liquid volume used for color development of 2 ml. , dimensionless.
[0051] The value of 26,900 represents the molar absorptivity of cyanidin-3-glucoside, expressed as liters per mole per centimeter. This value is the molar absorptivity constant of cyanidin-3-glucoside at a wavelength of 520 nanometers as reported in the literature. The optical path length of the cuvette is expressed in centimeters; the standard cuvette used in this step has an optical path length of 1 cm. The calculated results are characterized as milligrams of anthocyanins per 100 grams of fruit pulp. Each sample was measured in triplicate, with the relative standard deviation controlled within 5%.
[0052] Step S204: Standardize the collection of juice yield data using the pressing method. Take 500g of mulberry fruit sample, juice it for 3 minutes at rated speed using a juicer, filter the juice through double-layered gauze, and transfer it to a graduated cylinder to read the juice volume. Juice mass is calculated by multiplying the juice volume by the density of mulberry juice, which is taken as 1.05g / ml at room temperature.
[0053] Juice yield was calculated as a percentage by dividing the weight of the juice after pressing by the weight of the raw sample and then multiplying by 100%. Each sample was measured five times independently, with the coefficient of variation controlled within 3%. During the measurements, process parameters such as pressing pressure, pressing time, raw material temperature, and raw material maturity grade were recorded simultaneously.
[0054] Step S205: Standardize the collection of the seed percentage index using a gravimetric method. Take 200g of mulberry fruit sample and homogenize it using a tissue homogenizer, adding distilled water to maintain good fluidity. Wet-sieve the homogenized sample through a 60-mesh standard sieve (0.25mm aperture). Collect the seed material remaining on the sieve, rinse with distilled water to remove residual pulp, and transfer to pre-dried glass petri dishes. Place the petri dishes in a vacuum drying oven and dry at 70°C and -0.08 MPa for at least 6 hours until constant weight.
[0055] After the dried grain material was cooled to room temperature, it was weighed using an analytical balance, accurate to 0.001 g. The grain percentage was calculated by dividing the weight of the dried grain material by the total weight of the sample and then multiplying by 100%. Each sample was measured five times independently, with the coefficient of variation controlled within 5%.
[0056] Through step S2 above, a standardized data collection system for five key mulberry variety indicators is established. This system covers five dimensions: sugar content, acidity, anthocyanin content, juice yield, and seed percentage. For each indicator, a complete process is defined, from sample pretreatment and measurement operations to data recording and calculation formulas. This data collection system provides a standardized data source for the construction of the variety characteristic-processing suitability correlation model in subsequent step S3, and also provides reliable input for the comparative analysis of the variety characteristic matching degree algorithm in step S4.
[0057] The next step is step S3, which involves constructing a correlation model between varietal characteristics and processing suitability. This step uses data mining and machine learning techniques to establish a quantitative mapping relationship between mulberry varietal characteristics and processing effects, providing processing suitability analysis capabilities for traceability assessment. This includes the following implementation steps.
[0058] Step S301: Construct a sugar content-fermentation performance adaptation model using a multiple linear regression algorithm. The input variable of the model is the sugar content index, and the output variables include three fermentation performance indicators: fermentation efficiency index, alcohol conversion rate, and residual sugar content.
[0059] The fermentation efficiency index is defined as the ratio of actual alcohol yield to theoretical alcohol yield, expressed as a percentage, reflecting the overall efficiency of yeast in converting sugar into alcohol. The alcohol conversion rate is defined as the ratio of the mass of alcohol actually produced to the mass of sugar consumed, reflecting the degree of sugar-to-alcohol conversion. The residual sugar content is defined as the concentration of reducing sugars in the final fermentation sample, expressed in grams per liter, reflecting the thoroughness of the fermentation process.
[0060] The model training samples were derived from fermentation experiments of mulberries with no fewer than 500 different sugar content ranges, covering multiple concentration gradients within the sugar content range of 18% to 24%. The experiments used standardized fermentation process parameters: fermentation temperature was controlled at 22 degrees Celsius, fermentation time was set at 7 days, yeast inoculum amount was set at 0.5 grams per liter, and initial sulfur dioxide addition amount was set at 60 milligrams per liter.
[0061] The data collection criteria for training are as follows: at least three parallel fermenters are set up for each sugar concentration gradient. After fermentation, samples are taken three times from each parallel fermenter for index measurement. The average of the three measurements is taken as the measured value of the fermentation performance index of that parallel fermenter. The average of the measured values from the three parallel fermenters is then taken as the final training sample value of the output variable under that sugar concentration condition, in order to eliminate the influence of batch-to-batch differences on model training. By controlling the single variable of sugar concentration while fixing other process parameters, the quantitative response relationship between sugar concentration and various fermentation performance indicators is obtained. The least squares method is used to estimate the regression coefficients for model fitting.
[0062] The sugar content-fermentation performance matching model adopts a multiple linear regression equation. For the j-th output variable, the regression equation is expressed as: in Indicates the first The model predicted values of the output variables, when The predicted value of the fermentation efficiency index at that time is expressed as a percentage; when The predicted value of alcohol conversion rate at that time, in percentage; when The predicted value of residual sugar content at that time is expressed in grams per liter.
[0063] This indicates the input sugar content value, expressed as a percentage. Indicates the first The regression intercepts corresponding to each output variable; Indicates the first The regression coefficients corresponding to each output variable reflect the average change in the output variable for every 1 percentage point change in sugar content. The regression coefficients are estimated by fitting the training samples using the least squares method, and the model's coefficient of determination is greater than or equal to 0.85.
[0064] When applying the model, after obtaining the sugar content index value of the sample to be tested, it is used as... Substitute the values into the regression equation and calculate the predicted values for the three output variables. The predicted value of the fermentation efficiency index is directly used as the fermentation efficiency fit score. The predicted value of the alcohol conversion rate is multiplied by 100 and used as the alcohol conversion rate fit score. The predicted value of the residual sugar content is reverse normalized and used as the residual sugar content fit score; the lower the residual sugar content, the higher the score. The range for all three scores is set to 0 to 100.
[0065] Step S302: Construct an acidity-mouth balance model using partial least squares regression. This model establishes a mapping relationship between acidity indices and sensory evaluation scores, encompassing three dimensions: sweet-sour ratio balance, flavor equilibrium, and mouthfeel comfort. The model training data comes from the evaluation results of 20 trained sensory evaluators on 50 samples with different acidity levels. Each sample was evaluated three times to eliminate the influence of individual evaluator differences.
[0066] The basic principle of partial least squares regression is to extract principal components from the predictor variable matrix and the response variable matrix, and maximize the covariance between the extracted principal components to establish a regression relationship between the predictor variables and the response variable. In this step, the predictor variables are a matrix composed of acidity index values, and the response variables are a matrix composed of scores from three sensory evaluation dimensions.
[0067] During model training, a 10-fold cross-validation method is used to determine the optimal number of principal components: the training data is randomly divided into 10 equal parts, and one part is used as the validation set and the remaining 9 parts are used as the training set for modeling and validation in turn. The root mean square error of cross-validation under different numbers of principal components is calculated, and the number of principal components that minimizes the root mean square error of cross-validation is selected as the optimal number of principal components, thereby avoiding overfitting or underfitting of the model.
[0068] When applying the model, after obtaining the acidity index value of the sample to be tested, it is input into the trained partial least squares regression model. The model outputs predicted values for the three sensory dimensions. Acidity fit score. Calculated using a weighted average across three dimensions, i.e. ,in This indicates the predicted score for the sweet and sour ratio. This indicates the predicted value of the flavor balance score. The three predicted values represent the taste comfort score, with each predicted value ranging from 0 to 100. The score range is also from 0 to 100.
[0069] Step S303: Construct an anthocyanin-color stability model using first-order reaction kinetics. This model describes the kinetic law of anthocyanin content decaying with increasing temperature and time during processing. The expression for the first-order reaction kinetics equation is: in Indicates time The anthocyanin content at any given time, expressed in milligrams per 100 grams; This indicates the initial anthocyanin content, expressed in milligrams per 100 grams. This represents the decay rate constant, expressed in units of per minute. Indicates time, in minutes.
[0070] Model parameters include decay rate constant and half-life Half-life is determined by the formula The calculations are in minutes. A longer half-life indicates greater stability of anthocyanins.
[0071] The model training data came from experimental data on anthocyanin content changes under different temperature conditions during mulberry processing. The temperature settings covered four levels: 4°C, 20°C, 37°C, and 55°C, with an observation time span from 0 hours to 720 hours. Under different temperature conditions, the first-order reaction kinetic equation was fitted using the nonlinear least squares method to determine the decay rate constant corresponding to each temperature. .
[0072] When applying the model, obtain the initial anthocyanin content of the sample to be tested. Then, the corresponding decay rate constant is determined based on the target processing or storage temperature. Storage time Substitute into the above equation to calculate the predicted anthocyanin content. Color stability rating The ratio of predicted anthocyanin content to initial anthocyanin content is multiplied by 100 to calculate the result. The rating ranges from 0 to 100. A higher rating indicates a higher anthocyanin retention rate and better color stability.
[0073] Step S304: Construct a juice yield-processing efficiency model using a random forest regression algorithm. The input variables of this model include three features: seed percentage, fruit maturity, and juicing pressure.
[0074] Fruit maturity was quantitatively characterized by measuring the puncture force of the fruit using a hardness tester. The seed percentage was derived from the standardized data collected in step S2. The juicing pressure parameter was the set operating pressure of the juicing equipment. The model output variables were the juice yield index and turbidity index. The juice yield index reflects the proportion of raw material converted into usable juice, and the turbidity index reflects the suspended matter content in the juice.
[0075] The basic principle of random forest regression is to construct multiple decision trees and combine the prediction results of each decision tree to obtain the final prediction value. In this step, the random forest model integrates no fewer than 100 decision trees. The construction process of each decision tree is as follows: using the Bootstrap sampling method, a subset of samples of the same size as the original dataset is randomly drawn with replacement from the original training dataset as the training samples for that decision tree.
[0076] At each node split, some features are randomly selected from all input features as split candidate features, and the optimal split feature and split position are selected from them; the decision tree grows recursively in the above manner until the stopping condition is met.
[0077] The model training samples were derived from 300 sets of juicing experiment data with different combinations of process parameters. Each experiment was repeated three times to ensure statistical reliability. The data cleaning criteria were as follows: outlier samples with juice extraction efficiency index exceeding the mean plus or minus three standard deviations were removed, as were samples with negative or unreasonable turbidity values.
[0078] When applying the model, the seed percentage, fruit maturity, and preset juicing pressure parameters of the mulberry raw material are obtained to form a 3D feature vector input to the random forest model. The model inputs the feature vector into each decision tree, and each decision tree makes independent predictions. Finally, the arithmetic mean of the prediction results of all decision trees is taken as the final output value of the model. The predicted values of the juice extraction efficiency index and turbidity index are directly used as the corresponding fitness scores, with a score range of 0 to 100.
[0079] The model outputs a ranking of the importance of each input feature. Feature importance is calculated using out-of-bag error: for each decision tree, the prediction error is calculated using out-of-bag data; then, the values of a particular feature are randomly shuffled, and the out-of-bag prediction error is calculated again; the larger the difference between the two errors, the greater the influence of that feature on the prediction result. This importance ranking is used to guide the optimization of mulberry juice extraction process parameters and the evaluation of processing efficiency.
[0080] To more clearly illustrate the execution flow of the above four models, the following description uses specific examples: A batch of mulberry raw materials, after testing in step S2, yielded the following varietal characteristic indicators: sugar content 21.35%, acidity 1.26 g / 100g, anthocyanin content 186.5 mg / 100g, juice yield 71.0%, and seed content 4.3%. This batch of raw materials is planned for use in the processing of fermented mulberry wine products, and its suitability for various processing techniques needs to be evaluated.
[0081] Application of the sugar content-fermentation performance adaptation model: A sugar content of 21.35% was used as... Substituting into the regression equation, the predicted fermentation efficiency index was 82.3%, and the fermentation efficiency fit score was 82 points; the predicted alcohol conversion rate was 73.1%, and the alcohol conversion rate fit score was 73 points; the predicted residual sugar content was 5.0 g / L, and after reverse normalization, the residual sugar content fit score was 20 points, indicating that the residual sugar at the end of fermentation was low under this sugar content condition.
[0082] Application of the acidity-mouth balance model: When the acidity value of 1.26 grams per 100 grams is input into the model, the model predicts a score of 88 for the sweet-sour ratio, 85 for the flavor balance, 82 for the mouthfeel, and 85 for the acidity compatibility.
[0083] Anthocyanin-color stability model application: The target product is mulberry wine, the storage temperature is set at 20 degrees Celsius, and the target storage time is 180 days. The decay rate constant at this temperature... Substituting the values into the equation, the predicted anthocyanin content after 180 days was 81.3 mg per 100g, and the color stability score was 44 points, indicating that the anthocyanin degradation of this mulberry variety is relatively significant during long-term storage.
[0084] Juice yield-process efficiency model application: The proportion of seeds in this batch of raw materials is 4.3%, the fruit maturity index is 2.8 kg / cm², the juicing pressure is set to 0.3 MPa, and after inputting into the random forest model, the output juice yield efficiency index score is 86 points and the turbidity index score is 78 points.
[0085] The fitness scores of the above four models are summarized and used in step S5 to generate a comprehensive traceability evaluation report.
[0086] Through step S3, this embodiment establishes a model system for the correlation between varietal characteristics and processing adaptability, quantifies the mapping relationship between varietal indicators and processing effects from four dimensions: fermentation conversion, taste coordination, color stability, and juice extraction efficiency, and provides a source of adaptability scores for each dimension for the comprehensive traceability assessment in step S5.
[0087] The next step, S4, involves designing a variety characteristic matching algorithm. This algorithm quantitatively compares the detection indicators of the product to be traced with the standard variety characteristics in the database, and outputs a matching score as a quantitative basis for determining traceability consistency. This is implemented through the following sub-steps: Step S401: Define the query vector and reference vector set. Define the variety indicator vector obtained by standardizing and collecting the product to be traced in step S2 as the query vector. Query vector This is a 5-dimensional real vector, whose components are, in order, sugar content, acidity, anthocyanin content, juice yield, and grain percentage. .in This indicates the sugar content value. This indicates the acidity index value. This indicates the anthocyanin content index value. This indicates the juice yield index value. This indicates the percentage of grains in the index.
[0088] Define the standard variety feature vectors stored in the database constructed in step S1 as the reference vector set. Reference vector set Includes A standard variety feature vector, namely Each of these vectors is also a 5-dimensional real vector, denoted as the first... The reference vectors are .
[0089] Indicates the first The standard sugar content value for each standard variety Indicates the first Acidity standard values for each standard variety Indicates the first Standard values for anthocyanin content of each standard variety. Indicates the first Standard values for juice yield of each standard variety. Indicates the first The standard value for the percentage of grains in each standard variety.
[0090] Step S402: Perform Euclidean distance calculation, using the Euclidean distance calculation method as the primary matching strategy. The query vector... With reference vector set Each reference vector in Perform distance measurements one by one. The Euclidean distance calculation formula is as follows: in Represents the query vector With the reference vectors The Euclidean distance between them. Represents the query vector The Each component. Represents the reference vector The Each component. The value ranges from 1 to 5, corresponding to the five product indicator dimensions. The calculation involves first squaring the differences in each dimension, then summing them and taking the square root.
[0091] It should be noted that the five varietal indicators have different physical units and dimensions: sugar content is expressed as a percentage, acidity as grams per 100 grams, anthocyanin content as milligrams per 100 grams, juice yield as a percentage, and seed percentage as a percentage. The Euclidean distance value in this step is used only as a dimensionless measure of the overall degree of difference and is not assigned a specific physical unit. Its function is to compare the relative proximity between different vectors within the same dimensional system; the smaller the distance value, the closer the varietal characteristics of the product to be traced are to the characteristics of the standard varietal in the database.
[0092] Step S403: Perform a matching score normalization calculation, converting the Euclidean distance calculated in step S402 into a matching score. Determine the normalization coefficients. The value is taken from the reference vector set. The normalization coefficient is set to 1.2 times the maximum Euclidean distance between all pairs of standard variety vectors, so that the numerical coverage of the normalization coefficient is slightly larger than the maximum difference distance between varieties in the database, ensuring that the normalized score distribution is reasonable.
[0093] The 1.2 factor is set based on the following: In feature similarity comparison in the fields of machine learning and pattern recognition, the normalization factor usually needs to be slightly larger than the maximum value of the normalized data to avoid saturation compression of the score at the high end. 1.2 is a commonly used empirical value in this type of scenario, which can maintain the score discrimination while making the normalized score evenly distributed in the whole range.
[0094] Calculated from step S402 Select the minimum value from the Euclidean distances. ,Right now The matching score is calculated using the following formula: in This represents the matching score, with a value ranging from 0 to 100. Represents the query vector With reference vector set The minimum Euclidean distance between all reference vectors in the vector. This represents the normalization coefficient. When... At that time, the score is set to the maximum value of 100, indicating that the product to be traced is completely consistent with the characteristics of a certain standard variety. The higher the score, the closer the variety characteristics of the product to be traced are to the characteristics of the standard variety in the database.
[0095] Step S404: Perform cosine similarity calculation as an alternative matching strategy. Besides using Euclidean distance as the primary matching strategy, the variety feature matching algorithm can also use cosine similarity calculation as an alternative matching strategy. The query vector of the product to be traced... With reference vectors in the database The cosine of the angle between the two points is used as a similarity measure. The formula for calculating cosine similarity is as follows: in Represents the query vector With the reference vectors The cosine of the angle between them. Represents the query vector The Each component. Represents the reference vector The Each component. The value ranges from 1 to 5, corresponding to the five indicator dimensions of each variety.
[0096] In the calculation, the numerator is the dot product of the two vectors, and the denominator is the product of the magnitudes of the two vectors. Since the index values for each variety are all positive, the actual range of the cosine of the angle is 0 to 1. The closer the value is to 1, the more consistent the directions of the two vectors are, meaning the more similar the variety's characteristic patterns are. The cosine similarity matching score is converted to a 0 to 100-point score using a linear transformation formula. .
[0097] The cosine similarity method has advantages in handling directional differences in indicator vectors. When two vectors have consistent proportional relationships across various indicator dimensions but different overall amplitudes—that is, when all variety indicators are simultaneously high or low—the cosine value of the included angle remains close to 1, and the cosine similarity matching score remains high. This method can effectively identify situations where variety characteristic patterns are consistent but there are systematic deviations in absolute values, and can serve as an effective supplement to the Euclidean distance method.
[0098] Step S405: Perform threshold filtering and sorting output. The variety feature matching degree algorithm supports threshold filtering and sorting output functions. When the user sets a minimum matching degree threshold... hour, The value ranges from 0 to 100, and the algorithm only returns a matching score greater than or equal to 100. The results are recorded.
[0099] When a user requests the complete sorting results, the algorithm returns all reference varieties and their corresponding matching scores in descending order of matching score. If the user also sets a threshold... The algorithm then filters out those that satisfy the condition. After selecting the reference varieties, the screening results are then sorted and output in descending order of scores.
[0100] The algorithm accelerates the similarity comparison process through the database composite index structure designed in step S102, and achieves fast response with parallel computing strategy, with an average response time of less than 200 milliseconds for a single query.
[0101] To more clearly illustrate the complete execution flow of the above variety feature matching algorithm, the following description uses a specific example: A batch of mulberry raw materials labeled as the "Da Shi" variety was tested in step S2, yielding five varietal indicators: sugar content 21.35%, acidity 1.26 g / 100g, anthocyanin content 186.5 mg / 100g pulp, juice yield 71.0%, and seed percentage 4.3%. These five indicators were then organized into a query vector. .
[0102] The mulberry variety feature database constructed in step S1 stores the standard feature vectors of the Da Shi variety. The standard feature vector of the Hongguo No. 2 variety is The standard feature vector of the White Jade King variety is A total of 50 standard variety characteristics records were compiled.
[0103] Step S402 Euclidean distance calculation: Calculate the query vector sequentially. The Euclidean distance to the feature vector of each standard variety in the database. Taking the DaShi variety as an example: Minimum Euclidean distance after traversing all 50 reference vectors , corresponding to the ten major varieties.
[0104] Step S403: Normalization coefficient and score calculation: First, scan all 50 reference vectors in the database. After calculating the Euclidean distance between each pair of vectors, determine the maximum distance value as 15.0. Then, calculate the normalization coefficient. Match score: Step S404 Cosine Similarity Calculation: Taking the top ten varieties as an example, calculate the query vector. The cosine of the angle between the vector and the larger ten standard vector. The numerator is 41024.737, the query vector magnitude is 200.746, and the larger ten standard vector magnitude is 204.361. Cosine similarity matching score This indicates that the proportional relationships of the various indicators of the query vector and the Big Ten standard vector are highly consistent, with only minor differences in the overall amplitude.
[0105] Step S405 Threshold Filtering and Sorting Output: Assume the user sets a minimum matching threshold. Among all 50 reference varieties, the Da Shi variety scored 79.7 points, Hongguo No. 2 scored 68.2 points, and Baiyuwang scored 63.4 points. Only the Da Shi variety met the threshold condition, and the algorithm outputs the Da Shi variety and its matching score of 79.7 points. If the user requests a complete sort without setting a threshold, the algorithm outputs a list of scores for all reference varieties from highest to lowest.
[0106] The above examples demonstrate that the matching algorithm can effectively identify the standard varieties that are closest to the test sample, and evaluate the degree of matching of variety characteristics from different perspectives using two strategies: Euclidean distance and cosine similarity.
[0107] Step S5 completes the design of the variety feature matching algorithm in this embodiment. The algorithm uses Euclidean distance as the primary matching strategy, calculating the minimum distance by iterating through the query vector and the reference vector set, and then normalizing the result to obtain a matching score. Simultaneously, cosine similarity is used as an alternative matching strategy, measuring the similarity of variety features using the cosine value of the angle between the query vector and the reference vector set. The algorithm supports two result presentation methods: threshold filtering and sorted output. This matching algorithm provides a quantitative scoring basis for the variety matching conclusion section of the comprehensive traceability assessment report in step S5.
[0108] Finally, step S5 integrates the digital traceability and evaluation process for mulberry fruit. This step systematically integrates data collection, feature extraction, model matching, and result output to form a full-chain traceability and evaluation capability for variety characteristics from raw materials to finished products. This is achieved through the following sub-steps: Step S501: Receive and parse the sample information of the mulberry processed products to be traced. When receiving the sample information of the mulberry processed products to be traced, parse the key fields in the sample registration request, including sample number, sample type, source batch, processing technology description and customer identification information.
[0109] The sample type enumeration field distinguishes between different forms of mulberry samples, such as fresh mulberry fruit samples, mulberry pulp samples, fermented mulberry product samples, and concentrated mulberry juice samples. The source batch field links raw material supplier information and harvesting time information, providing data support for origin traceability. The processing technology description field records the processing path of the sample from raw materials to finished product, used for process context matching in subsequent processing suitability analysis.
[0110] Step S502: The key varietal indicator values of the sample to be tested are obtained by calling the varietal indicator collection system, and the appropriate indicator collection process is selected according to the sample type. For fresh mulberry samples, the standard collection method defined in step S2 is directly executed to obtain the values of the five core varietal indicators. For processed mulberry product samples, the raw material characteristics are inferred based on the processing technology type.
[0111] The processing method for mulberry pulp samples is as follows: Sugar content, acidity, anthocyanin content, and other indicators in the pulp are measured, and the varietal characteristics of the raw materials are inferred based on these measured values. The processing method for fermented mulberry product samples is as follows: Residual sugar content and alcohol content at the end of fermentation are measured, and the raw material characteristics are comprehensively estimated by combining these with the indicators of the pulp before fermentation. The processing method for concentrated mulberry juice samples is as follows: Soluble solids content in the concentrated juice is measured, and the varietal characteristics of the raw materials are inferred based on the concentration ratio.
[0112] Step S503: Call the variety feature matching degree algorithm for quantitative comparison. Input the indicator vector of the variety to be tested into the variety feature matching degree algorithm to perform the comparison operation. First, retrieve reference records from the database that are from the same production area and of the same variety type as the batch of the sample to be tested to narrow the matching search range and improve the comparison efficiency.
[0113] Then, the indicator vector of the variety to be tested is compared with the candidate reference records one by one. The distance is calculated one by one according to the Euclidean distance calculation formula defined in step S402, and the minimum Euclidean distance is converted into a matching score according to the matching score scoring formula defined in step S403. Finally, the reference variety with the highest matching score and its matching score are output. It also provides score distribution statistics and anomaly detection results.
[0114] Step S504: The variety characteristics and processing suitability association model is called to calculate the suitability score, and the four association models defined in step S3 are used for evaluation. For fresh mulberry fruit samples, the five variety index values measured in step S502 are directly input into each model.
[0115] For mulberry processed product samples, the raw material level variety index values obtained by back-deriving based on the processing technology type in step S502 are input into each model. These back-derived values have been converted into raw material level variety indicators and can be directly substituted into each associated model to calculate the suitability score without additional conversion.
[0116] The sugar content-fermentation performance matching model outputs a score on the compatibility of the tested sample with the fermentation process. This score reflects the degree to which sugar content affects yeast fermentation efficiency. The acidity-mouth-taste compatibility model outputs an acidity score for the test sample's suitability for the target taste. This score reflects the degree to which acidity level contributes to the flavor balance of the product.
[0117] The anthocyanin-color stability model outputs predicted scores for the color stability of the sample under different storage conditions. This score reflects the influence of anthocyanin content and processing and storage conditions on the product's appearance and quality. The juice yield-process efficiency model outputs a predicted score for the juice yield efficiency of the tested sample relative to the pressing process. This score reflects the combined influence of grain ratio and process parameters on juice extraction.
[0118] Step S505: Integrate the matching score and the suitability score to generate a comprehensive traceability evaluation report, and calculate the comprehensive score using a weighted summation model. The matching score weight is set to 0.4, and the weights of the four processing suitability scores are each set to 0.15, for a total weight of 0.6. The comprehensive score calculation formula is as follows: in This represents the overall score, with a value ranging from 0 to 100. The matching score output by step S503 reflects the degree of consistency between the sample under test and the characteristics of standard varieties in the database, and the value ranges from 0 to 100. This represents the fermentation efficiency fit score output by the sugar content-fermentation performance fit model in step S504, with a value range of 0 to 100. This represents the taste harmony score output by the acidity-taste harmony model in step S504, with a value range of 0 to 100. This represents the color stability fit score output by the anthocyanin-color stability model in step S504, with a value range of 0 to 100.
[0119] This represents the juice yield-process efficiency and turbidity compatibility score output by the juice yield-process efficiency model in step S504, with a value range of 0 to 100. All five scores are dimensionless values from 0 to 100, with uniform dimensions, allowing for weighted summation.
[0120] The comprehensive traceability assessment report includes three core sections: variety matching conclusions, processing suitability assessment, and traceability consistency level determination. The variety matching conclusions section lists the name of the standard variety that best matches the sample under test, along with its matching score. The document includes matching confidence intervals and a list of candidate varieties. The processing suitability assessment section lists the suitability scores of the tested samples for various processing techniques. , , , Analysis of compatibility advantages and suggestions for process optimization.
[0121] The section on determining the consistency level of traceability is based on a comprehensive score. The samples to be tested were divided into four traceability consistency levels. The excellent level corresponds to... A score between 85 and 100 indicates that the tested sample is highly consistent with the database record and has excellent processing suitability. A "good" grade corresponds to... A score between 70 and 84 indicates that the tested sample is basically consistent with the database record and has good processing compatibility. The corresponding pass / fail level is... A score between 55 and 69 indicates that the tested sample deviates somewhat from the database record, but is still within an acceptable range. The corresponding non-compliance level is... A score below 55 indicates a significant discrepancy between the tested sample and the database record, requiring further verification.
[0122] The comprehensive traceability assessment report is generated in structured JSON format and can also be converted to PDF document format for archiving and viewing. The report includes metadata information such as the version identifier of this embodiment, assessment timestamp, assessor signature, and digital signature verification, ensuring the legal validity and traceability of the report.
[0123] Through the above steps, this embodiment completes the full integration of the digital traceability assessment process for mulberry fruit. This process forms a complete closed loop, from sample information reception and variety indicator acquisition to matching degree and suitability score calculation and comprehensive assessment report generation, covering the entire chain of variety characteristic traceability and assessment from raw materials to finished products. This integrated process achieves the coordinated use of the indicator collection system in step S2, the correlation model system in step S3, and the matching degree algorithm in step S4, enabling the traceability assessment results to not only quantitatively reflect the consistency of variety origin but also comprehensively assess the suitability of different mulberry varieties for the target processing technology.
[0124] To facilitate understanding, the following process is illustrated using the example of a mulberry processing company receiving a batch of mulberry raw materials labeled as the "Da Shi" variety.
[0125] The company's quality control department selects representative samples from the raw material batches and sends them to the digital traceability assessment laboratory. After receiving the samples, the laboratory assigns a unique identifier and establishes a registration record, corresponding to step S501.
[0126] Step S502: Index collection: Five varietal index values were obtained through testing: sugar content 21.35%, acidity 1.26 g / 100g, anthocyanin content 186.5 mg / 100g pulp, juice yield 71.0%, and seed ratio 4.3%.
[0127] Step S503 Variety Feature Matching: Combine the above 5 indicator values into a query vector. Retrieve the standard feature vectors of the top ten varieties from the database. Calculate the Euclidean distance: The maximum Euclidean distance in the database is 15.0, and the normalization coefficient is... Match score: Step S504 Processing Suitability Assessment: Four correlation models are invoked, and the sugar content-fermentation performance suitability model outputs... Acidity-mouth balance model output The anthocyanin-color stability model predicts that after 30 days of storage at 20 degrees Celsius, [the following information is missing]. Juice yield - process efficiency model output point.
[0128] Step S505: Generating the Comprehensive Evaluation Report: Substitute each score into the weighted summation formula: The calculation yields: The traceability consistency level is good.
[0129] The report shows that the raw materials match the standard characteristics of the Da Shi variety and have good processing suitability, suggesting their use in high-end mulberry pulp or fermented product processing. The report also lists similar candidate varieties, including Hongguo No. 2 (71 points matching) and Baiyuwang (68 points matching).
[0130] On the other hand, the digital traceability and evaluation system for mulberry processed products disclosed in this invention includes: The database construction module is used to build a database of mulberry variety characteristics. The core fields of the database include sugar content, acidity, anthocyanin content, juice yield, and seed ratio. The indicator acquisition module is used to establish an indicator acquisition system for key mulberry varieties, defining standardized acquisition methods and testing procedures for sugar content, acidity, anthocyanin content, juice yield, and seed ratio. The association model construction module is used to construct association models between varietal characteristics and processing adaptability. These association models include a sugar content fermentation performance adaptation model, an acidity taste coordination model, an anthocyanin color stability model, and a juice yield process efficiency model. The matching degree algorithm module is used to design the variety feature matching degree algorithm, which quantitatively compares the detection indicators of the product to be traced with the standard variety features in the mulberry variety feature database. The assessment integration module is used to integrate the digital traceability assessment process for mulberries and generate a comprehensive traceability assessment report.
[0131] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects.
[0132] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.
Claims
1. A digital traceability assessment method for mulberry fruit processed products, characterized in that: Includes the following steps: A database of mulberry variety characteristics was constructed, and the core fields of the database include sugar content, acidity, anthocyanin content, juice yield, and seed ratio. Establish a key mulberry variety indicator collection system, and define standardized collection methods and testing procedures for the sugar content, acidity, anthocyanin content, juice yield and seed ratio, respectively; A correlation model between varietal characteristics and processing adaptability is constructed, which includes a sugar content fermentation performance adaptation model, an acidity taste coordination model, an anthocyanin color stability model, and a juice yield process efficiency model. Design a variety feature matching algorithm to quantitatively compare the detection indicators of the product to be traced with the standard variety features in the database; Integrate the digital traceability assessment process for mulberries to generate a comprehensive traceability assessment report.
2. The method according to claim 1, characterized in that: The sugar content fermentation performance adaptation model is constructed using a multiple linear regression algorithm. The input variable of the sugar content fermentation performance adaptation model is the sugar content index value, and the output variables of the sugar content fermentation performance adaptation model include the fermentation efficiency index, alcohol conversion rate, and residual sugar content. The acidity-taste coordination model is constructed using a partial least squares regression algorithm. The input variable of the acidity-taste coordination model is the acidity index value, and the output variables of the acidity-taste coordination model include scores in three dimensions: the coordination of sweet and sour ratio, the balance of flavor, and the comfort of taste. The anthocyanin color stability model is constructed using first-order reaction kinetics equations, and the expression for the anthocyanin color stability model is as follows: ,in For time Anthocyanin content at any given time This represents the initial anthocyanin content. The decay rate constant; The juice yield process efficiency model is constructed using a random forest regression algorithm. The input variables of the juice yield process efficiency model include the seed ratio index, fruit maturity index, and juicing pressure parameter.
3. The method according to claim 1, characterized in that: The variety feature matching algorithm defines the variety indicator vector of the product to be traced as the query vector. ,in This refers to the sugar content index. This refers to the acidity index value. This refers to the anthocyanin content index value. This refers to the juice yield index value. This refers to the percentage of grains in the index. The variety feature matching algorithm defines the standard variety feature vectors stored in the database as a reference vector set. ,in The total number of standard variety feature vectors stored in the database; The variety feature matching algorithm uses Euclidean distance to measure distances one by one. The Euclidean distance calculation formula is as follows: in For the query vector With the reference vectors The Euclidean distance between them For the first reference vectors The One portion, The values range from 1 to 5, corresponding to the sugar content index, acidity index, anthocyanin content index, juice yield index, and seed ratio index, respectively. The variety feature matching degree algorithm calculates the minimum Euclidean distance. Convert to a matching score using the following formula: ,in Score the matching degree. The minimum Euclidean distance between the query vector and all reference vectors in the reference vector set. This is the normalization coefficient.
4. The method according to claim 3, characterized in that: The variety feature matching algorithm also employs cosine similarity calculation as an alternative matching strategy. The cosine similarity calculation formula is as follows: ,in For the query vector With the reference vectors The cosine of the angle between them; The variety feature matching algorithm converts the cosine value of the included angle into a cosine similarity matching score using the following formula: ,in The cosine similarity matching score is given.
5. The method according to claim 1, characterized in that: The standardized collection of the sugar content index was carried out by measuring the supernatant of mulberry juice using a handheld digital refractometer. Before the measurement, the mulberry samples were juiced and centrifuged at 3000 rpm for 10 minutes. The supernatant was then filtered through a 0.45-micron filter membrane to obtain the test liquid. The measurement temperature was controlled within the range of 20 to 25 degrees Celsius. Three readings were taken for each measurement and the average value was taken. Three parallel measurements were performed for each sample. The standardized collection of the acidity index was performed using acid-base titration to determine the titratable acid content. 20 grams of mulberry fruit sample was added to an equal volume of distilled water and crushed for 2 minutes. After diluting to 100 ml, the solution was titrated with sodium hydroxide standard solution to the phenolphthalein indicator endpoint. The titratable acid content was calculated using the following formula: in The content of the titratable acid. The volume of sodium hydroxide standard solution consumed in the titration. To determine the precise concentration of the sodium hydroxide standard solution, This is the conversion factor for malic acid. This is the dilution factor. For sample quality.
6. The method according to claim 1, characterized in that: The anthocyanin content was standardized by spectrophotometric determination using the pH differential method. Absorbance values were measured at 520 nm in two buffer solutions, pH 1.0 and pH 4.5, and the anthocyanin content was calculated using the following formula: in The anthocyanin content is... This represents the difference in absorbance values measured at pH 1.0 and pH 4.
5. This refers to the average molecular weight of cyanidin 3-glucoside. Diluting factor The molar absorptivity of cyanidin 3-glucoside is given. The optical path length of the cuvette; The standardized collection of the juice yield index was determined by pressing method. 500 grams of mulberry fruit sample was juiced for 3 minutes, filtered through double gauze, and the juice volume was measured and converted into juice mass. The juice yield was obtained by dividing the juice mass by the raw material sample mass and multiplying by 100%. The standardized collection of the seed proportion index was determined by gravimetric method. 200 grams of mulberry fruit sample was homogenized and wet-sieved through a 60-mesh standard sieve. The seed material on the sieve was collected, dried to constant weight, and weighed. The seed proportion was obtained by dividing the weight of the dried seed material by the total weight of the sample and multiplying by 100%.
7. The method according to claim 1, characterized in that: The composite index structure of the mulberry variety feature database adopts a hybrid design pattern of hash index and B-tree index. The hash index is used to support exact matching queries of variety identifiers, and the B-tree index is used to support numerical range queries and feature vector similarity comparisons. The mulberry variety characteristic database adopts a distributed key-value storage architecture. The underlying layer of the distributed key-value storage architecture is based on the consistent hashing algorithm to realize the partitioned storage and load balancing of data. The variety characteristic records are serialized and stored in JSON format. Each variety characteristic record includes a variety identifier field, a collection timestamp field, a sugar content scalar field, an acidity scalar field, an anthocyanin content scalar field, a juice yield scalar field, a seed percentage scalar field, an origin attribute field, and a processing type field.
8. The method according to claim 1, characterized in that: The construction of the mulberry variety characteristic database also includes deploying a data acquisition and preprocessing module, which is responsible for performing the following steps: Data from different data sources is processed by a unified adaptation layer and converted into JSON format records, which are then cleaned, quality checked, and format verified in sequence. The data cleaning process employs the interquartile range method for statistical outlier detection. For a specific index field of a particular variety being processed, let the sample dataset be... Calculate the sample dataset First quartile and the third and fourth quartiles Let the interquartile range When data points satisfy or If an element is identified as an outlier, it is removed. The data acquisition and preprocessing module periodically recalculates the quartile thresholds of each indicator using recent historical data, thereby dynamically updating the judgment criteria.
9. The method according to claim 1, characterized in that: The integrated digital traceability and evaluation process for mulberry fruit includes: Receive and parse sample information of mulberry fruit processed products to be traced; The varietal index values of the sample to be tested are obtained by calling the key mulberry variety index collection system. The variety feature matching algorithm is invoked to perform quantitative comparison and output a matching score. The compatibility score is calculated by calling the correlation model between the variety characteristics and processing compatibility. The weighted summation model is used to calculate the overall score. The formula for calculating the overall score is as follows: in The overall score is as follows. Score the matching degree. The fermentation efficiency fit score output by the sugar content fermentation performance fit model. The acidity-taste harmony matching model outputs a taste harmony fit score. The color stability fit score output by the anthocyanin color stability model is... The juice yield process efficiency model outputs a score for the juice yield efficiency and turbidity fit. Based on the comprehensive score, the samples to be tested are divided into four traceability consistency levels: excellent, good, qualified, and unqualified.
10. A digital traceability and evaluation system for mulberry processed products, applied to the method described in any one of claims 1-9, characterized in that: include: The database construction module is used to construct a mulberry variety characteristic database. The core fields of the mulberry variety characteristic database include sugar content, acidity, anthocyanin content, juice yield, and seed ratio. The indicator acquisition module is used to establish an indicator acquisition system for key mulberry varieties, and to define standardized acquisition methods and detection procedures for the sugar content, acidity, anthocyanin content, juice yield and seed ratio, respectively. The association model construction module is used to construct an association model between varietal characteristics and processing adaptability. The association model between varietal characteristics and processing adaptability includes a sugar content fermentation performance adaptability model, an acidity taste coordination model, an anthocyanin color stability model, and a juice yield process efficiency model. The matching degree algorithm module is used to design a variety feature matching degree algorithm to quantitatively compare the detection indicators of the product to be traced with the standard variety features in the mulberry variety feature database. The assessment integration module is used to integrate the digital traceability assessment process for mulberry fruit and generate a comprehensive traceability assessment report.
