An evaluation method for the processing characteristics of fresh Daqu (a type of Chinese koji) blanks based on texture mechanics analysis.

By using structural mechanics analysis and support vector machine model, the sensory quality of fresh Daqu (fermentation starter) blanks is quantitatively evaluated, solving the problems of wheat pulverization and forming quality in the koji-making process and improving the quality control of baijiu brewing.

CN122283048APending Publication Date: 2026-06-26CHINA NAT RES INST OF FOOD & FERMENTATION IND CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610354918.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-23
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing technologies make it difficult to objectively quantify and evaluate the sensory quality of fresh Daqu (fermentation starter) blanks, resulting in difficulties in accurately controlling the degree of wheat crushing, the suitability of ingredients, and the quality of molding during the Daqu-making process, which affects the quality of Baijiu brewing.

Method used

Using the textural mechanics analysis method, the hardness, elasticity, chewiness, and cohesiveness of fresh koji blanks were measured by a texture analyzer. A support vector machine (SVM) model was established to construct a koji blank discrimination model, thereby realizing the quantitative evaluation of sensory quality.

Benefits of technology

This method enables the quantitative characterization of wheat processing characteristics for koji making, providing a rapid, accurate evaluation method that is not influenced by the subjective experience of production personnel, thereby improving quality control in the koji making process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122283048A_ABST
    Figure CN122283048A_ABST
Patent Text Reader

Abstract

This invention discloses an evaluation method for the processing characteristics of fresh Daqu (fermented wheat koji) blanks based on textural mechanics analysis. This method quantifies the sensory analysis results of the wheat-flour mixture used in koji production, allowing for an objective evaluation of the sensory quality of the koji blanks based on specific data. The method comprises the following steps: preparing multiple samples of fresh koji blanks as test samples; using a texture analyzer to measure the physical properties of each sample as evaluation indicators of the sensory characteristics of the fresh koji blanks. The measured physical properties include hardness, elasticity, chewiness, cohesion, and adhesiveness; determining the stability of the measured physical properties of different test samples and establishing a discrimination model for different wheat varieties of koji blanks; and determining the correlation between the measured physical property parameters and the sensory evaluation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of mechanical and sensory index testing technology for fresh Daqu (fermented koji) blanks in the koji-making process of Baijiu brewing, and particularly to an evaluation method based on the analysis of the processing characteristics of fresh Daqu blanks using texture mechanics. Background Technology

[0002] High-quality koji (fermented starter culture) should be yellow in color, have a firm, tortoise-shell shape, be elastic to the touch, and have a suitable density. Strict inspection is required to ensure the wheat powder used for koji production is appropriate, the koji ingredients are suitable, and the height and density of the formed koji are within the process standards. In the mixing stage of koji production, the grain-to-water ratio is crucial; the water content is generally controlled between 36% and 40%. Too much water leads to spoilage and mold growth, while too little water makes it difficult to bind and is detrimental to the growth of beneficial microorganisms. After forming, the koji should have full, neat corners, uniform density, and smooth, flat surfaces on all six sides. The surface slurry is then removed. This slurry promotes the formation of nutritional advantages on the surface of the koji and enhances its water retention, facilitating the "coating" of the koji. TPA (Total Texture Analysis) technology, also known as total texture analysis, simulates human touch to test samples, quantifying human sensory perception of food. It is now widely used to test the sensory quality of various foods.

[0003] For the reasons mentioned above, this invention aims to develop a key physical index analysis method using TPA analysis technology combined with quantitative texture technology, for the research and optimization of the production process of Baijiu koji. The data results can be applied to the selection of wheat varieties for koji making, the optimization of the grain and water ratio in the mixing process, and the objective quantification and characterization of sensory quality indicators such as looseness, compactness, agglomeration, viscosity, hardness, and elasticity in the koji forming process. Summary of the Invention

[0004] In order to solve the above-mentioned technical problems, the present invention provides an evaluation method for the processing characteristics of fresh Daqu (fermented starter culture) blanks based on the analysis of texture mechanics. This method can express the sensory analysis results of the wheat grain flour mixture for koji making in a quantitative form and objectively evaluate the sensory quality of the koji blanks based on specific data.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An evaluation method for the processing characteristics of fresh Daqu (a type of Chinese koji) blanks based on textural mechanics analysis includes the following steps: S1. Prepare multiple fresh koji blanks as samples to be tested; S2. The physical properties of each sample were measured using a texture analyzer to serve as an evaluation index for the sensory characteristics of fresh koji. The measured physical properties included hardness, elasticity, chewiness, cohesion, and adhesion. S3. Determine the stability of the determination of various physical property parameters of different test samples, and establish a discrimination model for different varieties of wheat koji. The steps for determining the stability of the measured physical properties of the sample are as follows: S3.1.1 Measure the physical properties of the sample to be tested multiple times and obtain the average value of each physical property. ; S3.1.2, Based on the average value of physical property parameters The standard deviations of the corresponding physical property parameters are obtained using the following formulas. and coefficient of variation value , The lower the value, the smaller the dispersion of the data. In the formula: Each data point refers to the result of each specific measurement of a mechanical parameter; This refers to the average value of all measurements of this mechanical parameter; This refers to the number of data points, i.e., the number of parallel tests performed on the same sample. The standard deviation (SD) of multiple parallel tests for each mechanical parameter of the same sample should be less than 10% of the average value of that parameter, and the coefficient of variation (CV) should generally be ≤10%, indicating stable testing conditions and good repeatability. Specifically, CV, as a normalized statistical indicator measuring the dispersion of observed values, can eliminate the influence of differences in dimensions and mean levels on the result determination, objectively reflecting the precision of the testing process. A fluctuation range within 10% proves that the method has good repeatability and data stability during operation, ensuring the accuracy and reliability of subsequent analysis of differences in the mechanical quality of different molded blanks.

[0006] The steps for constructing the blank discrimination model include: S3.2.1 Data Preparation and Preprocessing: The collected mechanical parameters were organized into a feature matrix representing the sensory quality of the sample to be tested. Min-Max normalization was then applied to transform the original feature x into a standardized feature x. To eliminate the impact of differences in the units and magnitudes of different parameters on the model: in: It is the first For all values ​​of a feature column, the normalization operation ensures that all feature values ​​fall within the range of _____. Within the range, this improves the training stability of subsequent models; S3.2.2, SVM Model Construction and Training: By selecting the RBF kernel function and setting a penalty coefficient, data transformation and adjustment are performed on the mechanical parameters of different pre-processed blanks to enhance the model's generalization ability. The RBF kernel function is defined as follows: in: , This represents two sample feature vectors; This represents the square of the Euclidean distance between two vectors; γ It is a key parameter of the RBF kernel, controlling the complexity of the function and the degree of fit. Specifically, γ The value of γ directly affects the smoothness of the decision boundary: too small a value will lead to underfitting of the model and an inability to distinguish the differences in mechanical properties between different types of sprues; too large a value will cause excessive distortion of the decision boundary, incorporating noise into the model and reducing generalization ability. A commonly used empirical formula for determining the value of γ is: Where k is the feature dimension (5 here), and median is the median of the squared Euclidean distances among all samples. This formula makes the kernel function take a moderate value at typical sample distances.

[0007] S3.2.3 Model Evaluation and Validation: k-fold cross-validation is used to evaluate the model's generalization ability on unknown data, avoiding overfitting and ensuring the reliability of the results. Specifically, the dataset is divided into k mutually exclusive subsets. Each time, k-1 subsets are used to train the model, and the remaining subset is used as the validation set. This process is repeated k times. The performance metrics (such as accuracy) of the blank discrimination model are calculated using the following formula: in, To determine the average performance metric of the model; It is the first The accuracy of the classification is defined as the ratio of the number of correctly classified samples to the total number of samples, and the expression is as follows: in: For accuracy; For true examples, it refers to the number of samples where the model correctly predicts the true variety A. A false counterexample refers to a real variety A, but the model incorrectly predicted the number of samples of real variety A. False positives refer to the number of samples where the actual variety is not A, but the model predicts that the actual variety is A. A true counterexample is a sample number of cases where the actual variety is not A, and the model correctly predicts that the actual variety is not A. S3.2.4 Result Analysis and Interpretation: The overall accuracy of the discrimination model is obtained, the confusion situation is analyzed in detail, the model performance is quantified, its advantages and possible misjudgment boundaries are clarified, and a basis is provided for application. S4. Determination of the correlation between physical property measurement parameters and sensory evaluation: S4.1, the procedure for determining physical property parameters is the same as in S2: transfer the sample to be tested from the sealed bag to the sample cup in several batches using a spatula until the sample to be tested is level with the rim of the cup. Then place the sample cup containing the sample to be tested directly below the center of the test probe. After aligning the probe and the rim of the sample cup, select the full texture analysis mode in the texture analyzer control software, start the test program, and collect the hardness, elasticity, chewiness, cohesiveness, and adhesiveness texture parameters of the shaped blank sample.

[0008] S4.2 Sensory evaluation includes: Elasticity evaluation: Press the surface of the blank lightly with your finger and release it quickly, then observe the degree to which it returns to its original shape. High-quality blanks should show obvious rebound. Hardness evaluation: Apply moderate pressure with your palm or fingers to feel its ability to resist deformation. The hardness should be moderate, neither too hard nor too soft. Hand feel evaluation: Take an appropriate amount of koji material, squeeze it into a ball, and observe its shape retention after releasing it. High-quality koji should not easily fall apart after being formed into a ball, and should have a moderate tightness and uniform internal structure.

[0009] The preparation steps for the above-mentioned test samples are as follows: different varieties of wheat are collected, crushed, and then water and yeast are added to prepare different varieties of test samples. Each variety of test sample is collected separately in a sealed bag for later use.

[0010] An auxiliary testing mechanism is also provided between the detection platform and the vertical lifting module of the texture analyzer. The auxiliary testing mechanism includes a bracket placed on the detection platform, a sample cup fixing plate on the bracket, a sample cup for loading fresh koji blanks on the sample cup fixing plate, and the port of the sample cup is located directly below the detection platform. The detection platform has a test probe, and the area of ​​the test end of the test probe is equal to the area of ​​the bottom of the sample cup.

[0011] In step S2, the sample to be tested in the sealed bag is transferred to the sample cup in multiple batches using a spatula until the sample to be tested in the sample cup is level with the rim of the cup. Then, the sample cup containing the sample to be tested is placed directly below the center of the test probe to determine the physical property parameters.

[0012] When using a texture analyzer to determine physical properties, the test procedure and parameter selection are as follows: Test type: Compression mode; Test method: TPA full texture test; Test target type: Distance compression distance; Test target: 10mm; Test conditions: Trigger load force 5N; speed before test 2mm / s; speed during test 1mm / s; speed after test 2mm / s.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention transforms the vague textual descriptions of hardness, elasticity, viscosity, compactness, and looseness in traditional koji-making wheat processing characteristics into quantifiable mechanical indicators. It then establishes objective and scientific sensory characteristics of koji-making wheat using these quantifiable indicators. Through a constructed koji-making discriminant model, the sensory analysis results of the koji-making wheat-flour mixture are expressed in quantitative form. Based on specific data, the sensory quality of the koji-making material is objectively evaluated, providing a rapid, accurate, and independent method for judging the processing characteristics of small-scale koji-making wheat in the koji-making process of baijiu brewing. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the assembly structure of the texture analyzer and auxiliary testing components in this invention; Figure 2 These are the sensor response diagrams of the texture analyzer probes for fresh wheat koji blanks of four varieties: A, B, C, and D. Figure 3 This is a scatter plot of the elastic mechanical parameters of fresh wheat koji from four varieties: A, B, C, and D. Figure 4 This is a scatter plot of the chewiness mechanical parameters of fresh wheat koji from four varieties: A, B, C, and D. Figure 5 This is a scatter plot of the cohesive mechanical parameters of fresh wheat koji from four varieties: A, B, C, and D. Figure 6 This is a scatter plot of the adhesive mechanical parameters of fresh wheat koji from four varieties: A, B, C, and D. Figure 7 This is a scatter plot of the mechanical parameters of hardness for fresh wheat koji (fermentation cakes) of four varieties: A, B, C, and D. Figure 8 These are the SVM algorithm results for the fresh wheat koji of four varieties: A, B, C, and D.

[0015] Attached diagram labels: 1. Texture analyzer, 11. Detection platform, 12. Vertical lifting module, 2. Support, 3. Sample cup fixing plate, 4. Sample cup, 13. Test probe. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0017] Example 1 An evaluation method for the processing characteristics of fresh Daqu (a type of Chinese koji) blanks based on textural mechanics analysis includes the following steps: S1. Prepare multiple fresh koji blanks as test samples; specifically, collect different varieties of wheat, crush them separately, add water and koji to make different varieties of test samples, and collect each variety of test sample separately in a sealed bag for later use. S2, such as Figure 1 As shown, a texture analyzer 1 is used to measure the physical properties of each sample to be tested, serving as an evaluation index for the sensory characteristics of fresh koji. The measured physical properties include hardness, elasticity, chewiness, cohesion, and adhesion. Specifically, an auxiliary testing mechanism is provided between the detection platform 11 and the vertical lifting module 12 of the texture analyzer 1. The auxiliary testing mechanism includes a support 2 placed on the detection platform 11, a sample cup fixing plate 3 on the support 2, and a sample cup 4 for loading fresh koji on the sample cup fixing plate 3. The port of the sample cup 4 is located directly below the detection platform 11. The detection platform 11 has a test probe 13, and the area of ​​the test end of the test probe 13 is equal to the area of ​​the bottom of the sample cup 4. Specifically, before measuring the physical properties of TPA, the sample to be tested in the sealed bag is transferred to the sample cup 4 in multiple transfers using a spatula until the sample to be tested in the sample cup 4 is level with the rim of the cup. Then, the sample cup 4 containing the sample to be tested is placed directly below the center of the test probe 13 for measuring the physical properties. When using texture analyzer 1 to measure physical property parameters, the test procedure and parameter selection are as follows: Test type: Compression mode; Test method: TPA full texture test; Test target type: Distance compression distance; Test target: 10mm; Test conditions: Trigger load 5N; initial velocity 2mm / s; velocity during test 1mm / s; velocity after test 2mm / s S3. Determine the stability of the determination of various physical property parameters of different test samples, and establish a discrimination model for different varieties of wheat koji. The steps for determining the stability of the measured physical properties of the sample are as follows: S3.1.1 Measure the physical properties of the sample to be tested multiple times and obtain the average value of each physical property. ; S3.1.2, Based on the average value of physical property parameters The standard deviations of the corresponding physical property parameters are obtained using the following formulas. and coefficient of variation value , The lower the value, the smaller the dispersion of the data. In the formula: Each data point refers to the result of each specific measurement of a mechanical parameter; This refers to the average value of all measurements of this mechanical parameter; This refers to the number of data points, i.e., the number of parallel tests performed on the same sample. The standard deviation (SD) of multiple parallel tests for each mechanical parameter of the same sample should be less than 10% of the average value of that parameter, and the coefficient of variation (CV) should generally be ≤10%, indicating stable testing conditions and good repeatability. Specifically, CV, as a normalized statistical indicator measuring the dispersion of observed values, can eliminate the influence of differences in dimensions and mean levels on the result determination, objectively reflecting the precision of the testing process. A fluctuation range within 10% proves that the method has good repeatability and data stability during operation, ensuring the accuracy and reliability of subsequent analysis of differences in the mechanical quality of different molded blanks.

[0018] The steps for constructing the blank discrimination model include: S3.2.1 Data Preparation and Preprocessing: The collected mechanical parameters are organized into the feature matrix of the sensory quality of the sample to be tested, and Min-Max normalization is used to convert the original feature x into standardized feature x. To eliminate the impact of differences in the units and magnitudes of different parameters on the model: Where X j These are all the values ​​of the j-th feature column. The normalization operation ensures that all feature values ​​fall within the [0,1] interval, which improves the training stability of the subsequent model. S3.2.2 SVM Model Construction and Training: The RBF kernel function is selected and a penalty coefficient is set. Data transformation and adjustment are performed on the mechanical parameters of different pre-processed blanks to enhance the model's generalization ability. The RBF kernel function is defined as follows: in, , This represents two sample feature vectors; This represents the square of the Euclidean distance between two vectors; γ It is a key parameter of the RBF kernel, controlling the complexity of the function and the degree of fit. Specifically, γ The value of γ directly affects the smoothness of the decision boundary. Too small a value leads to underfitting of the model, making it unable to distinguish the differences in mechanical properties between different types of sprue; too large a value causes excessive distortion of the decision boundary, incorporating noise into the model and reducing generalization ability. Therefore, in this embodiment, γ The numerical values ​​are determined using the following empirical formula: in, k is the dimension of the feature; median is the median of the squared Euclidean distances among all samples. This formula makes the kernel function take a moderate value at typical sample distances.

[0019] S3.2.3 Model Evaluation and Validation: k-fold cross-validation is used to evaluate the model's generalization ability on unknown data, avoiding overfitting and ensuring the reliability of the results. Specifically, the dataset is divided into k mutually exclusive subsets. Each time, k-1 subsets are used to train the model, and the remaining subset is used as the validation set, repeated k times. Model performance metrics (such as accuracy) are measured using... Average calculation: in, To determine the average performance metric of the model; It is the first The accuracy of the classification is defined as the ratio of the number of correctly classified samples to the total number of samples, and the expression is as follows: in: For accuracy; For true examples, it refers to the number of samples where the model correctly predicts the true variety A. A false counterexample refers to a real variety A, but the model incorrectly predicted the number of samples of real variety A. False positives refer to the number of samples where the actual variety is not A, but the model predicts that the actual variety is A. A true counterexample is a sample number of cases where the actual variety is not A, and the model correctly predicts that the actual variety is not A. S3.2.4 Result Analysis and Interpretation: Obtain the accuracy of the discriminant model at each fold. and average accuracy An average accuracy rate >85% indicates that the characteristic differences in the mechanical properties of different koji blanks can be effectively distinguished by SVM, demonstrating the model's excellent discriminative ability. In practical operation, the model's confusion is specifically analyzed based on the mechanical data feedback from different koji blank samples. The model's performance is quantified, its advantages and potential misjudgment boundaries are clarified, providing a basis for the application of texture analysis equipment in the mechanical quality judgment of koji production.

[0020] S4. Determination of the correlation between TPA physical property measurement parameters and sensory evaluation S4.1 The TPA physical property parameter determination procedure is the same as S2. The sample to be tested in the sealed bag is transferred to the sample cup 4 in multiple times with a spatula until the sample to be tested in the sample cup 4 is level with the cup mouth. Then, the sample cup 4 containing the sample to be tested is placed directly below the center of the test probe 13. After aligning the probe 13 and the cup mouth of the sample cup 4, the Total Texture Analysis (TPA) mode is selected in the texture analyzer control software, the test program is started, and the hardness, elasticity, chewiness, cohesiveness and adhesion texture parameters of the shaped blank sample are collected.

[0021] S4.2 Sensory evaluation includes: Elasticity evaluation: Press the surface of the blank lightly with your finger and release it quickly, then observe the degree to which it returns to its original shape. High-quality blanks should show obvious rebound. Hardness evaluation: Apply moderate pressure with your palm or fingers to feel its ability to resist deformation. The hardness should be moderate, neither too hard nor too soft. Hand feel evaluation: Take an appropriate amount of koji material, squeeze it into a ball, and observe its shape retention after releasing it. High-quality koji should not easily fall apart after being formed into a ball, and should have a moderate tightness and uniform internal structure.

[0022] Verification Example Every half hour, samples of wheat flour mixtures of various varieties were collected from the outlet of the mechanical mixing hopper as three parallel samples. At each sampling point, more than 500g of wheat flour mixture was placed in a sealed bag for later use.

[0023] Use a spatula to fill each portion of wheat flour mixture into the sample cup in multiple portions, 130g each time, keeping the height level with the mouth of the sample cup, and then place it directly below the center of the test probe, ready for testing.

[0024] The selection of parameters for measuring physical properties: The five parameters of elasticity, hardness, resilience, adhesion, and cohesion are selected based on the following reasons: The hardness is the maximum force value of the first compression test; the measured spectrum has a clear correspondence with the standard spectrum, which can be used to characterize the sensory indicators of the wheat mixture processing process for koji making; Resilience is calculated as recoverable work / compression work, yielding a value between 0 and 1. The larger the ratio, the better the resilience. It can be used to evaluate the looseness of wheat flour mixtures used in koji making. Therefore, resilience is also used as one of the indicators for evaluating the sensory characteristics of materials. Cohesion is the ratio of the compression work of the second cycle to the hardness work of the first cycle in the TPA experiment, which characterizes the internal bond strength and solidity of the sample structure; therefore, it is also used as one of the indicators for evaluating the sensory characteristics of fresh koji blanks. The elasticity, hardness, resilience, and cohesiveness of fresh koji blanks were measured; the measured spectra were compared to determine the parameters required for the test; and based on the actual sample state and sensory characteristics, as well as the sensory definitions of each index, five indicators—elasticity, hardness, resilience, cohesiveness, and adhesion—were selected as evaluation indicators for the sensory characteristics of fresh koji blanks. Repeat the test multiple times on the same mixed grain powder sample and record the data each time. Calculate the average value, which is the physical property data of this fresh koji sample. To investigate the stability of the measured physical properties of fresh koji samples: calculate the coefficient of variation (CV) based on the data of each physical property, and determine the stability by the coefficient of variation. The correlation between physical property parameters and sensory evaluation was determined by repeatedly testing various physical property parameters of different varieties of wheat grain flour mixtures (four varieties of wheat for koji making, A, B, C, and D). Based on the mechanical parameters of the blank, a sensory quality discrimination model for the blank was established using the support vector machine analysis method. The detailed steps include ① data preparation and preprocessing ② SVM model construction and training ③ model evaluation and verification ④ result analysis and interpretation.

[0025] In the implementation case, the correlation between physical property measurement parameters and sensory evaluation was determined.

[0026] Sensory evaluation of fresh koji blanks was conducted, and the determined test parameters were used to analyze fresh koji blank samples from different wheat varieties.

[0027] Figure 2 The sensor response results of four different varieties of wheat flour mixtures (A, B, C, and D) were tested, and the relationship between the applied force and the sample displacement was recorded. From the sensor response graph, it can be preliminarily determined that there are differences in the data fed back from the two extrusion processes of different varieties of shaped blanks under the TPA detection mode. Based on the probe data, the mechanical parameters of the specific shaped blanks were further calculated, as shown in Table 1.

[0028] In Table 1, Average represents the mean, and SD represents the standard deviation. The coefficient of variation (CV) was calculated based on the data for each physical property parameter. The CV value was used to determine the stability of the texture analyzer and the auxiliary analysis platform specified in this patent in detecting the processing characteristics of wheat flour. The data shows significant differences in the mechanical properties of fresh koji (fermented wheat cakes) corresponding to different wheat varieties: varieties A and C exhibit higher cohesion than varieties B and D. The sensory characteristic measurements of different wheat varieties show that the standard deviations for cohesion, chewiness, elasticity, and hardness are all less than 10% of the mean, and the coefficients of variation are all less than 10%. This indicates that the method used to determine these four mechanical properties of koji has good repeatability and stability, and can objectively and accurately reflect the physical properties. However, the coefficient of variation for adhesion is relatively large, possibly due to the low surface viscosity of the material itself.

[0029] Table 1. Parameters measured by the texture analyzer for processing characteristics of four wheat varieties in the examples. The five mechanical parameters used in Table 1 to monitor the koji-making characteristics of four wheat varieties (A, B, C, and D) are visualized using a scatter plot, allowing for a more intuitive assessment of the koji-forming characteristics of different wheat varieties. Figure 3-7 It can be seen that there are significant differences in elasticity, hardness, cohesion, and chewiness among the four wheat varieties A, B, C, and D between groups, while the parallelism within each group is relatively good. Due to the material characteristics of the koji block itself, the surface viscosity is low, and the adhesion values ​​of variety D fluctuate greatly within the three tests.

[0030] Sensory evaluations were conducted on four varieties of wheat flour and the processed koji blanks, and the results are shown in Table 2. By comparing the data in Tables 1 and 2, it was found that the hardness, elasticity, resilience, and cohesion collected by the texture analyzer were consistent with the sensory evaluations of different wheat varieties of koji blanks reported by the koji-making workshop technicians.

[0031] Table 2 Sensory evaluation of the processing characteristics of four varieties of koji blanks in the examples. Figure 8 The results of the discriminant model constructed using the support vector machine analysis method are shown, based on five mechanical parameters of the wheat koji blanks from four wheat varieties in the example. Detailed data processing steps are as follows: Data preparation and preprocessing: The five mechanical parameters were organized into a blank feature matrix and normalized. SVM Model Construction and Training: The RBF kernel function was selected, and the penalty coefficient C=10 was set. The mechanical parameters of the four varieties of pre-processed koji blanks were converted and adjusted to enhance the generalization ability of the model. Model evaluation and validation: Four-fold cross-validation was used, dividing the data into four parts, and using three parts for training and one part for testing in turn, for a total of four cycles; Results Analysis and Interpretation: The accuracy of the obtained discrimination model reached 91.67%. Specifically, in this embodiment, two varieties C were misclassified as varieties B.

[0032] The above results indicate that the rapid sensory quality discrimination model for wheat koji based on the mechanical parameters obtained from texture analysis has a high level of accuracy and can be effectively used for the rapid identification of the mechanical quality of koji.

[0033] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. An evaluation method for processing characteristics of fresh Daqu Qu Bai based on texture mechanics analysis, characterized in that, Includes the following steps: S1. Prepare multiple fresh koji blanks as samples to be tested; S2. The physical properties of each sample to be tested were measured using a texture analyzer (1) as an evaluation index of the sensory characteristics of fresh koji blanks. The measured physical properties included hardness, elasticity, chewiness, cohesion, and adhesion. S3. Determine the stability of the determination of various physical property parameters of different test samples, and establish a discrimination model for different varieties of wheat koji. The stability determination steps for the measurement of each physical property parameter of the sample to be tested are as follows: S3.1.1, multiple times of measuring each physical property parameter of the sample to be measured, and obtaining the average value of each physical property parameter ; S3.1.2, average value of the physical property parameter and the standard deviation of the corresponding physical property parameter is obtained by the following formula and the coefficient of variation value , The lower the data dispersion is, the smaller the data dispersion is: ; ; In the formula: is each data point, indicating the result of each specific measurement of the mechanical parameter; indicates the average value of all measurement results of the mechanical parameter; is the number of data points, i.e. the number of parallel detections of the same sample; S4. Determination of the correlation between physical property measurement parameters and sensory evaluation; S4.1, the physical property parameter determination steps are the same as in S2: the sample to be tested in the sealed bag is transferred to the sample cup (4) in multiple batches using a spatula until the sample to be tested in the sample cup (4) is level with the mouth of the cup. Then the sample cup (4) containing the sample to be tested is placed directly below the center of the test probe (13). After aligning the probe (13) and the mouth of the sample cup (4), the full texture analysis mode is selected in the texture analyzer control software, the test program is started, and the hardness, elasticity, chewiness, cohesiveness and adhesion texture parameters of the shaped blank sample are collected. S4.2 Sensory evaluation includes: Elasticity evaluation: Press the surface of the blank lightly with your finger and release it quickly, then observe the degree to which it returns to its original shape. High-quality blanks should show obvious rebound. Hardness evaluation: Apply moderate pressure with your palm or fingers to feel its ability to resist deformation. The hardness should be moderate, neither too hard nor too soft. Hand feel evaluation: Take an appropriate amount of koji material, squeeze it into a ball, and observe its shape retention after releasing it. High-quality koji should not easily fall apart after being formed into a ball, and should have a moderate tightness and uniform internal structure.

2. The method according to claim 1, wherein the method is characterized by, The preparation steps for the sample to be tested are as follows: Different varieties of wheat were collected, crushed, and then mixed with water and yeast to prepare different test samples. Each test sample was collected separately in a sealed bag for later use.

3. The method according to claim 2, wherein the method is characterized by, An auxiliary testing mechanism is provided between the testing platform (11) and the vertical lifting module (12) of the texture analyzer (1). The auxiliary testing mechanism includes a bracket (2) placed on the testing platform (11). A sample cup fixing plate (3) is provided on the bracket (2). A sample cup (4) for loading fresh koji is provided on the sample cup fixing plate (3). The port of the sample cup (4) is located directly below the testing platform (11). A test probe (13) is provided on the testing platform (11). The area of ​​the test end of the test probe (13) is equal to the area of ​​the bottom of the sample cup (4).

4. The method according to claim 3, wherein the method is characterized by, In step S2, the sample to be tested in the sealed bag is transferred to the sample cup (4) in multiple steps using a spatula until the sample to be tested in the sample cup (4) is level with the mouth of the cup. Then the sample cup (4) containing the sample to be tested is placed directly below the center of the test probe (13) to determine the physical parameters of TPA.

5. The method according to claim 1, wherein the method is characterized by, When the texture analyzer (1) performs TPA determination, the TPA test procedure and test parameters are selected as follows: Test type: Compression mode; Test method: TPA full texture test; Test target type: Distance compression distance; Test target: 10mm; Test conditions: Trigger load force 5N; speed before test 2mm / s; speed during test 1mm / s; speed after test 2mm / s.

6. The method according to claim 1, wherein the method is characterized by, The steps for constructing the blank discrimination model include: S3.2.1 Data Preparation and Preprocessing: The collected mechanical parameters are arranged as the characteristic matrix of the sensory quality of the sample to be tested, and Min-Max normalization is used to convert the original characteristics x into standardized characteristics x , so as to eliminate the influence of different parameter units and magnitude differences on the model: ; where: X j is all values of the jth feature column, and the normalization operation can ensure that all feature values fall within the [0, 1] interval; S3.2.2, SVM Model Construction and Training: Select the RBF kernel function and set the penalty coefficient. The RBF kernel function is defined as follows: ; where: , denotes two sample feature vectors; denotes the squared Euclidean distance between two vectors; γ is a key parameter of the RBF kernel and is determined by the following empirical formula: ; wherein: k is the dimension of the feature; median is the median of the squared Euclidean distances between all samples; S3.2.3 Model Evaluation and Validation: k-fold cross-validation is used to evaluate the model's generalization ability on unknown data to ensure the reliability of the results. Specifically, the dataset is divided into k mutually exclusive subsets. Each time, k-1 subsets are used to train the model, and the remaining subset is used as the validation set. This process is repeated k times. The performance metrics of the blank discrimination model (such as accuracy) are calculated using the following formula: ; wherein: is the average performance indicator of the discriminant model; is the accuracy of the i-th fold, accuracy being defined as the ratio of the number of correctly classified samples to the total number of samples, expressed as follows: is the accuracy of the i-th fold, accuracy being defined as the ratio of the number of correctly classified samples to the total number of samples, expressed as follows: ; wherein: is the accuracy rate; is the number of true positive samples, which means the real variety A and the model correctly predicts the real variety A; is the number of false negative samples, which means the real variety A but the model incorrectly predicts the real variety A; is the number of false positive samples, which means the real variety is not A but the model predicts the real variety not A as the real variety A; is the number of true negative samples, which means the real variety is not A and the model correctly predicts the real variety not A; S3.2.4 Result Analysis and Interpretation: An average accuracy of >85% indicates that the characteristic differences in the mechanical properties of different blanks can be effectively distinguished by SVM.