Brewing wheat variety identification method and device based on rheological characteristics
Through a neural network model based on rheological characteristic data, the problem of time-consuming and cost-effective identification of wheat varieties is solved, and rapid and accurate identification of brewed wheat varieties is achieved, which is suitable for industrial production.
Patent Information
- Application Number
- CN202510438879.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-09
- Publication Date
- 2025-07-11
AI Technical Summary
The prior art has problems in identifying wheat varieties that rely on artificial experience and are costly, complex molecular biological technology and are not suitable for industrial high-throughput screening.
Based on rheological characteristic data, a neural network is used to construct a wheat variety prediction model. By obtaining rheological characteristic data of brewed wheat and training, rapid identification of wheat varieties is achieved.
It improves the efficiency and accuracy of wheat variety detection, and is suitable for the rapid screening and identification of large-scale wine-making wheat samples.
Smart Images
Figure CN120296357A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wheat variety identification, and specifically to a method and device for identifying brewing wheat varieties based on rheological properties. Background Art
[0002] The identification of wheat varieties is of great significance in the brewing industry. Different wheat varieties have significant differences in starch structure, protein content, enzyme activity, etc., which affect their processing performance, fermentation behavior, and the flavor and quality of the final wine body. At present, the identification of wheat varieties mainly relies on methods such as manual experience judgment or molecular biology techniques (such as DNA molecular markers). The manual identification method depends on professional knowledge and experience, is easily affected by subjective factors, and takes a long time. Although molecular biology techniques are accurate, the detection process is complex, the cost is high, and the requirements for experimental conditions are high, making it difficult to meet the needs of high-throughput and rapid screening in industrial production. Summary of the Invention
[0003] In order to achieve industrial detection of wheat varieties, the present application provides a method and device for identifying brewing wheat varieties based on rheological properties.
[0004] The technical solution adopted by the present invention to solve the above problems is as follows:
[0005] A method for identifying brewing wheat varieties based on rheological properties, comprising:
[0006] Step 1: Obtain rheological property data in the brewing wheat sample data, label the wheat variety information, and construct a rheological property - wheat variety data set;
[0007] Step 2: Based on a neural network, use the rheological property data as input and the wheat variety as output to construct and train a wheat variety prediction model;
[0008] Step 3: Obtain the rheological property data of the wheat sample to be predicted, and input it into the wheat variety prediction model to obtain the prediction result.
[0009] Further, the wheat variety information includes the variety or origin.
[0010] Further, the rheological property data is rheological property time series data, sequence data of key indicators, and / or profile index data. Among them, the rheological property time series data refers to the torque data that changes with time during the rheological test; the key indicators refer to the torque peak value C1 during dough kneading, the protein weakening property index C2, the starch retrogradation property index C3, the starch hot paste gelatinization hot gel stability index C4, and the retrogradation property index C5 of gelatinized starch in the cooling stage; the profile index data includes: water absorption index, kneading index, gluten index, viscosity index, enzyme activity index, and retrogradation index.
[0011] Furthermore, the wheat variety prediction model is constructed based on MLP, LSTM, RNN, CNN, GRU, and / or Transformer.
[0012] Furthermore, the training process of the wheat variety prediction model is as follows:
[0013] The rheological property - wheat variety dataset is divided into a training set and a test set. The wheat variety prediction model is trained using the training set, and the training effect of the wheat variety prediction model is evaluated using the test set. When the error function of the test set is less than the error function threshold and the accuracy rate of the test set is higher than the accuracy rate threshold, the training is completed.
[0014] Furthermore, the error function is the cross - entropy loss function, and its expression is: In the formula, M is the number of samples; C is the number of wheat varieties; p ij is the true variety label of the i - th sample corresponding to the j - th wheat variety; is the probability that the i - th sample is predicted as the j - th wheat variety.
[0015] Furthermore, step 1 also includes: removing missing values and / or outliers in the rheological property - wheat variety dataset and performing normalization processing.
[0016] The device for identifying brewing wheat varieties based on rheological properties includes:
[0017] A data acquisition module, which is used to acquire rheological property data in the brewing wheat sample data, label the wheat variety information, and construct a rheological property - wheat variety dataset;
[0018] A model construction module; based on a neural network, using rheological property data as input and wheat variety as output, constructing and training a wheat variety prediction model;
[0019] A wheat variety prediction module: acquiring rheological property data of the wheat sample to be predicted and inputting it into the wheat variety prediction model to obtain a prediction result.
[0020] The beneficial effects of the present invention compared with the prior art are: a wheat variety prediction model is created based on rheological property data and its corresponding wheat varieties. Wheat variety identification is carried out based on the wheat variety prediction model, and the detection efficiency and accuracy are higher, which can be used for the rapid screening and identification of a large number of brewing wheat samples. Description of the Drawings
[0021] Figure 1 It is a flow chart of the method for identifying brewing wheat varieties based on rheological properties;
[0022] Figure 2Schematic diagram of the change of the loss function during the training process;
[0023] Figure 3 Schematic diagram of the result of predicting the wheat origin with the time series data of rheological properties;
[0024] Figure 4 Schematic diagram of the result of predicting the wheat variety with the time series data of rheological properties;
[0025] Figure 5 Schematic diagram of the result of predicting the wheat variety with the sequence data of key indicators;
[0026] Figure 6 Schematic diagram of the result of predicting the wheat variety with the sectional view index data;
[0027] Figure 7 Schematic diagram of the structure of the brewing wheat variety identification device based on rheological properties. Specific embodiments
[0028] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0029] Rheological properties are key parameters characterizing the physical properties of wheat dough, which can reflect the physical and chemical properties of wheat proteins and starches and have important reference value for the quality evaluation of brewing wheat. The rheological properties of wheat are measured using a mixing tester (by simulating the mixing and kneading process of dough to evaluate the rheological properties of dough). Through time series data analysis, key indicators such as the kneading and mixing properties of dough, the degree of protein weakening, and the starch retrogradation properties can be obtained. These indicators can effectively distinguish different varieties of wheat and provide a new method for the variety identification of brewing wheat.
[0030] As Figure 1 shown, the method for identifying brewing wheat varieties based on rheological properties includes:
[0031] Step 1: Obtain the rheological property data in the brewing wheat sample data, label the wheat variety information, and construct a rheological property - wheat variety data set.
[0032] The rheological property data are rheological property time series data, sequence data of key indicators, and / or profile index data. These three types of data can be used alone or in any combination. Among them, the rheological property time series data refer to the torque data that changes with time during the rheological test. The key indicators refer to the torque peak value C1 during dough kneading, the protein weakening property index C2, the starch retrogradation property index C3, the starch thermal gelatinization and hot paste stability index C4, and the retrogradation property index C5 of gelatinized starch during the cooling stage. The profile index data include: water absorption index, kneading index, gluten index, viscosity index, enzyme activity index, and retrogradation index. The wheat variety information includes the variety or origin.
[0033] To improve the data processing efficiency and accuracy, missing values and / or outliers in the rheological property - wheat variety dataset can also be removed and normalized.
[0034] Step 2: Based on the neural network, using the rheological property data as the input and the wheat variety as the output, construct and train a wheat variety prediction model.
[0035] The wheat variety prediction model is constructed based on MLP, LSTM, RNN, CNN, GRU, and / or Transformer. The wheat variety features can be represented by consecutive integers starting from 1 or one - hot vectors.
[0036] When training the wheat variety prediction model, divide the rheological property - wheat variety dataset into a training set and a test set. Use the training set to train the wheat variety prediction model, and use the test set to evaluate the training effect of the wheat variety prediction model. When the error function of the test set is less than the error function threshold and the accuracy of the test set is higher than the accuracy threshold, the training is completed.
[0037] In this embodiment, the error function uses the cross - entropy loss function, and the expression is: Loss = - In the formula, M is the number of samples; C is the number of wheat varieties; p ij is the true variety label of the i - th sample corresponding to the j - th wheat variety; is the probability that the i - th sample is predicted as the j - th wheat variety. Other loss functions can also be used, and there is no limitation here.
[0038] Step 3: Obtain the rheological property data of the wheat sample to be predicted, and input it into the wheat variety prediction model to obtain the prediction result.
[0039] Example 1
[0040] In this example, the rheological property time series data is used as the rheological property data to predict the origin of wheat.
[0041] Data acquisition was carried out using a hybrid tester, and time series sampling was performed at a sampling interval of 1 s. A total of 45 minutes of time series data was collected; a total of 100 sample data from 2 production areas were collected, 50 for each of the two production areas. The two production areas were recorded as production area 1 and production area 2, and were marked with integers 1 and 2 respectively.
[0042] The time series data is shown as follows:
[0043] Time / s Torque / N·m 1 0 2 0.012 3 0.017 4 0.024 5 0.031 …… ……
[0044] The dataset was randomly divided into a training set and a validation set at a ratio of 8:2. Therefore, the validation set included 20 sample data.
[0045] In the construction of the neural network in this embodiment, the data was resampled at a sampling frequency of 1 min, that is, a set of data included data at 45 time points of 1 min, 2 min,..., 45 min. A neural network model was constructed based on the LSTM network. During the training process, the loss function transformation diagrams of the training set and the test set are as Figure 2 shown, and the confusion matrix of the prediction results for different production areas is as Figure 3 shown. It can be seen that in this embodiment, a 100% accuracy rate for the test set can be achieved for discrimination.
[0046] Example 2
[0047] In this embodiment, the time series data of rheological properties was used as the rheological property data to predict the type of wheat.
[0048] In this embodiment, a total of 140 sample data from 2 varieties were collected. The two varieties were recorded as variety 1 and variety 2, and were marked with integers 1 and 2 respectively. Among them, variety 1 included 90 data, and variety 2 included 50 data.
[0049] Similar to Example 1, the dataset was randomly divided into a training set and a validation set at a ratio of 8:2. Therefore, the validation set included 28 sample data. In the construction of the neural network in this embodiment, the same as in Example 1, the data was resampled at a sampling frequency of 1 min, that is, a set of data included data at 45 time points of 1 min, 2 min,..., 45 min. A neural network model was constructed based on the LSTM network. During the training process, the confusion matrix of the prediction results for different varieties is as Figure 4 shown. It can be seen that in this embodiment, a 100% accuracy rate for the test set can be achieved for discrimination.
[0050] Example 3
[0051] In this embodiment, the sequence data of key indicators was used as the rheological property data to predict the type of wheat.
[0052] This embodiment uses five key indicators, C1, C2, C3, C4, and C5, to identify wheat varieties. A total of 1000 sample data of 2 varieties are collected. The two varieties are recorded as Variety 1 and Variety 2, and are marked with integers 1 and 2 respectively. Among them, Variety 1 includes 500 data, and Variety 2 includes 500 data.
[0053] Similar to Embodiment 1, the data set is randomly divided into a training set and a validation set according to 8:2. Therefore, the validation set includes 200 sample data. Based on the MLP network, a neural network model is constructed. During the training process, the confusion matrix of the prediction results of different varieties is as Figure 5 shown. It can be seen that in this embodiment, a 99% accuracy rate for the test set can be achieved for differentiation.
[0054] Embodiment 4
[0055] This embodiment uses the sectional profile index data as rheological property data to predict wheat varieties.
[0056] This embodiment uses six sectional profile index data, namely water absorption index, mixing index, gluten index, viscosity index, enzyme activity index, and retrogradation index, to identify wheat varieties. A total of 50 sample data of 2 varieties are collected. The two varieties are recorded as Variety 1 and Variety 2, and are marked with integers 1 and 2 respectively. Among them, Variety 1 includes 25 data, and Variety 2 includes 25 data.
[0057] Similar to Embodiment 1, the data set is randomly divided into a training set and a validation set according to 8:2. Therefore, the validation set includes 10 sample data. Based on the MLP network, a neural network model is constructed. During the training process, the confusion matrix of the prediction results of different varieties is as Figure 6 shown. It can be seen that in this embodiment, a 100% accuracy rate for the test set can be achieved for differentiation.
[0058] Correspondingly, this embodiment also provides a device for identifying brewing wheat varieties based on rheological properties, as Figure 7 shown, including:
[0059] A data acquisition module, configured to acquire rheological property data in brewing wheat sample data, label wheat variety information, and construct a rheological property - wheat variety data set;
[0060] A model construction module; based on a neural network, using rheological property data as input and wheat variety as output, constructing and training a wheat variety prediction model;
[0061] A wheat variety prediction module: acquiring rheological property data of a wheat sample to be predicted, and inputting it into the wheat variety prediction model to obtain a prediction result.
Claims
1. A method for identifying brewing wheat varieties based on rheological properties, characterized in that, Including: Step 1: Obtain the rheological property data in the brewing wheat sample data, label the wheat variety information, and construct a rheological property - wheat variety dataset; Step 2: Based on the neural network, use the rheological property data as the input and the wheat variety as the output to construct and train a wheat variety prediction model; Step 3: Obtain the rheological property data of the wheat sample to be predicted, and input it into the wheat variety prediction model to obtain the prediction result.
2. The method for identifying brewing wheat varieties based on rheological properties according to claim 1, characterized in that, The wheat variety information includes the variety or origin.
3. The method for identifying brewing wheat varieties based on rheological properties according to claim 1, characterized in that, The rheological property data is rheological property time - series data, sequence data of key indicators, and / or sectional profile index data. Among them, the rheological property time - series data refers to the torque data that changes with time during the rheological test; the key indicators refer to the torque peak value C1 when kneading dough, the protein weakening property index C2, the starch retrogradation property index C3, the starch thermal pasting and heat gel stability index C4, and the retrogradation property index C5 of gelatinized starch in the cooling stage; the sectional profile index data includes: water absorption index, kneading index, gluten index, viscosity index, enzyme activity index, and retrogradation index.
4. The method for identifying brewing wheat varieties based on rheological properties according to claim 1, characterized in that The wheat variety prediction model is constructed based on MLP, LSTM, RNN, CNN, GRU, and / or Transformer.
5. The method for identifying brewing wheat varieties based on rheological properties according to claim 1, characterized in that, The training process of the wheat variety prediction model is as follows: Divide the rheological property - wheat variety dataset into a training set and a test set. Use the training set to train the wheat variety prediction model, and use the test set to evaluate the training effect of the wheat variety prediction model. When the error function of the test set is less than the error function threshold and the accuracy of the test set is higher than the accuracy threshold, the training is completed.
6. The method for identifying brewing wheat varieties based on rheological properties according to claim 5, characterized in that, The error function is the cross-entropy loss function, and its expression is: In the formula, M is the number of samples; C is the number of wheat varieties; p ij is the true variety label of the i-th sample corresponding to the j-th wheat variety; is the probability that the i-th sample is predicted as the j-th wheat variety.
7. The method for identifying brewing wheat varieties based on rheological properties according to claim 1, characterized in that, Step 1 also includes: removing the missing values and / or outliers in the rheological property - wheat variety dataset and performing normalization processing.
8. A device for identifying brewing wheat varieties based on rheological properties, characterized in that, Including: A data acquisition module, which is used to obtain the rheological property data in the brewing wheat sample data, label the wheat variety information, and construct a rheological property - wheat variety dataset; A model construction module; based on the neural network, use the rheological property data as the input and the wheat variety as the output to construct and train a wheat variety prediction model; A wheat variety prediction module: obtain the rheological property data of the wheat sample to be predicted, and input it into the wheat variety prediction model to obtain the prediction result.