Fermented grain layer identification method and device based on lactic acid, acetic acid and hexanoic acid content
Through data sets and neural network models based on lactic acid, acetic acid and hexanoic acid content, the accuracy and automated identification of traditional lactic acid hierarchical methods are solved, and the high accuracy recognition of lactic acid level is achieved, which is suitable for the brewing process of strong aroma white wine.
Patent Information
- Application Number
- CN202510364898.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-26
- Publication Date
- 2025-07-11
AI Technical Summary
The traditional granule layering method is not very accurate, cannot be digitized and automated, and it is difficult to identify the granule levels that have been released from the cellar.
The three major acid-granule hierarchical data sets are constructed based on the contents of lactic acid, acetic acid and hexachloric acid, and the hierarchical prediction model is established and trained. The neural network is used to perform hierarchical prediction, including MLP, LSTM, RNN, CNN, GRU and/or Transformer models. The hierarchical levels are predicted by detecting the lactic acid, acetic acid and hexachloric acid content in the scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered scattered sca
It achieves high accuracy identification of the grease level, with a wide range of applications, and is suitable for accurate identification of the grease level in the brewing process of strong fragrance liquor.
Smart Images
Figure CN120299562A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of mash layer identification, and in particular to a mash layer identification method and device based on lactic acid, acetic acid and caproic acid content. Background Art
[0002] In the brewing process of Luzhou-flavor liquor, accurate identification of the mash layers has an important impact on the flavor of the wine, the control of the fermentation process and the quality of the final product. The traditional mash layering method is mainly determined based on the yellow water line or the position of the mash in the cellar. The above method is not accurate and has strong limitations. It is impossible to effectively identify the layer of the mash that has been taken out of the cellar, and it is difficult to form a digital and automated identification method. Summary of the invention
[0003] In order to facilitate the identification of mash layers, the present application provides a mash layer identification method and device based on the lactic acid, acetic acid and caproic acid content.
[0004] The technical solution adopted by the present invention to solve the above problems is:
[0005] The method for identifying the level of mash based on the content of lactic acid, acetic acid and caproic acid includes:
[0006] Step 1: Obtain the contents of lactic acid, acetic acid, and caproic acid in the fermented grains sample data, annotate the fermented grains hierarchical information, and construct the three major acids-fermented grains hierarchical data set;
[0007] Step 2: Based on the three acid-gravy level data sets, with lactic acid, acetic acid and caproic acid content as input and the gravy level as output, a gravy level prediction model is established and trained;
[0008] Step 3: Obtain the lactic acid, acetic acid and caproic acid contents of the mash to be predicted, and input them into the trained mash level prediction model to obtain the mash level prediction results.
[0009] Furthermore, data on the contents of lactic acid, acetic acid and caproic acid are collected in the form of mass concentration or molar concentration.
[0010] Furthermore, the mash layer is an upper mash layer, a middle mash layer or a bottom mash layer.
[0011] Furthermore, the mash hierarchical prediction model is constructed using MLP, LSTM, RNN, CNN, GRU and / or Transformer.
[0012] Furthermore, the steps of training the mash hierarchical prediction model are as follows:
[0013] Divide the data in the three major acids-pit mud layer dataset into a training set and a test set according to a preset ratio. Use the training set to train the pit mud layer prediction model, and use the test set to determine the prediction accuracy of the pit mud layer prediction model. 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 of the pit mud layer prediction model is completed.
[0014] Furthermore, the error function is the cross-entropy loss function, and the expression is as follows:
[0015]
[0016] In the formula, M is the number of samples; C is the number of pit mud layers; p ij is the true label of the i-th sample corresponding to the j-th pit mud layer; is the probability that the i-th sample is predicted as the j-th pit mud layer.
[0017] Furthermore, after obtaining the contents of lactic acid, acetic acid, and caproic acid, it also includes preprocessing them, including removing missing values and outliers and performing normalization processing.
[0018] Furthermore, it also includes:
[0019] Regularly update and / or supplement the three major acids-pit mud layer dataset, and synchronously perform fine-tuning training and updating on the pit mud layer prediction model.
[0020] The pit mud layer identification device based on the contents of lactic acid, acetic acid, and caproic acid includes:
[0021] Data acquisition module: used to acquire the contents of lactic acid, acetic acid, and caproic acid in the pit mud sample data and the corresponding pit mud layer information, and construct the three major acids-pit mud layer dataset;
[0022] Model creation and training module: Based on the three major acids-pit mud layer dataset, using the contents of lactic acid, acetic acid, and caproic acid as inputs and the pit mud layer as the output, establish and train the pit mud layer prediction model;
[0023] Pit mud layer prediction module: Acquire the contents of lactic acid, acetic acid, and caproic acid of the pit mud to be predicted, and input them into the trained pit mud layer prediction model to obtain the prediction result of the pit mud layer.
[0024] Furthermore, it also includes:
[0025] Abnormal alarm module: used to regularly acquire the prediction result of the pit mud layer. When the prediction result accuracy of the pit mud layer is lower than the preset threshold, trigger an alarm and update the three major acids-pit mud layer dataset and the pit mud layer prediction model.
[0026] The beneficial effects of the present invention compared with the prior art are as follows: Based on the contents of lactic acid, acetic acid and caproic acid, the present invention constructs a three-acid-pit mud hierarchical dataset, creates and trains a pit mud hierarchical prediction model, and performs hierarchical prediction based on the pit mud hierarchical prediction model, with higher accuracy. By detecting the contents of lactic acid, acetic acid and caproic acid in the pit mud, the pit mud hierarchy can be predicted, which is more convenient to use and has a wider application range. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 FIG. is a flowchart of a method for identifying the pit mud hierarchy based on the contents of lactic acid, acetic acid and caproic acid;
[0028] Figure 2 FIG. is a schematic diagram showing the change of the loss function value during the training process of the embodiment;
[0029] Figure 3 FIG. is the prediction results of three pit mud hierarchies of the test set;
[0030] Figure 4 FIG. is a schematic structural diagram of a device for identifying the pit mud hierarchy based on the contents of lactic acid, acetic acid and caproic acid. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0031] In order 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. 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.
[0032] As Figure 1 shown, the method for identifying the pit mud hierarchy based on the contents of lactic acid, acetic acid and caproic acid includes:
[0033] Step 1: Obtain the contents of lactic acid, acetic acid and caproic acid in the pit mud sample data, and label the pit mud hierarchy information to construct a three-acid-pit mud hierarchical dataset; wherein, the contents of lactic acid, acetic acid and caproic acid are collected in the form of mass concentration or molar concentration, and the pit mud hierarchy is the upper layer of the pit mud, the middle layer of the pit mud or the bottom layer of the pit mud.
[0034] In this embodiment, the units of lactic acid and acetic acid are g / 100 g of pit mud, and the unit of caproic acid is mg / 100 g of pit mud; the data example is as follows:
[0035] Sample 1: upper layer of the pit mud, lactic acid 1.86, acetic acid 0.11, caproic acid 51.69;
[0036] Sample...
[0037] Sample 38: middle layer of the pit mud, lactic acid 3.44, acetic acid 0.27, caproic acid 31.18;
[0038] Sample...
[0039] Sample 76: Bottom layer of fermented grains, lactic acid 4.23, acetic acid 0.35, caproic acid 71.81;
[0040] Sample…
[0041] Step 2: Based on the three - acid - fermented grains layer dataset, using the lactic acid, acetic acid, and caproic acid contents as inputs and the fermented grains layer as the output, establish and train a fermented grains layer prediction model.
[0042] The fermented grains layer prediction model is constructed based on a neural network. The neural network model includes one model or a combined model of multiple neural network models such as MLP, LSTM, RNN, CNN, GRU, and Transformer. In this embodiment, the fermented grains layer prediction model is constructed based on MLP. In the model, since the inputs are the lactic acid, acetic acid, and caproic acid content data and the outputs are the upper layer of fermented grains, the middle layer of fermented grains, and the bottom layer of fermented grains, both the input dimension and the output dimension are 3.
[0043] The steps for training the fermented grains layer prediction model are as follows: Divide the data in the three - acid - fermented grains layer dataset into a training set and a test set according to a preset ratio. Use the training set to train the fermented grains layer prediction model, and use the test set to determine the prediction accuracy of the fermented grains layer 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 of the fermented grains layer prediction model is completed.
[0044] In this embodiment, the error function uses the cross - entropy loss function, and the expression is as follows:
[0045]
[0046] In the formula, M is the number of samples; C is the number of fermented grains layers; p ij is the true label of the i - th sample corresponding to the j - th fermented grains layer; is the probability that the i - th sample is predicted as the j - th fermented grains layer. Other loss functions can also be used, such as mean squared error, mean absolute error, etc.
[0047] In this embodiment, a total of 119 groups of data of three fermented grains layers are collected. The dataset is randomly divided into a training set (95 groups of data) and a test set (24 groups of data) according to a ratio of 80% and 20%. The change of the loss function value during the training process is as Figure 2 shown. The accuracy of the predicted values and the true values of the test set is used as an index to evaluate the model effect. When the accuracy is greater than 90%, it is considered that the model is well - trained and the training can be considered completed. In this embodiment, the prediction results of the three fermented grains layers of the test set are as Figure 3 shown. For the three fermented grains layers, the accuracy reaches 92% and the training is completed.
[0048] In this embodiment, the upper layer and the bottom layer of the fermented grains can be predicted with 100% accuracy. Since the middle layer of the fermented grains connects the upper layer and the bottom layer, it is in line with the objective law that the test set samples are predicted as the upper layer or the middle layer of the fermented grains.
[0049] Step 3: Obtain the contents of lactic acid, acetic acid and caproic acid in the fermented grains to be predicted, and input them into the trained fermented grains layer prediction model to obtain the prediction result of the fermented grains layer.
[0050] To improve the prediction accuracy of the model, the data can also be preprocessed after obtaining the data, such as removing the missing values or outliers in the three major acids-fermented grains layer dataset; performing normalization processing on the lactic acid, acetic acid and caproic acid data, etc. The normalization can be implemented by using the sklearn.preprocessing.MinMaxScaler function, or other preprocessing means can also be used, which are not limited here.
[0051] To improve the real-time performance of the prediction model, the three major acids-fermented grains layer dataset can also be updated and / or supplemented regularly, and the fermented grains layer prediction model can be fine-tuned and trained and updated synchronously. Regularly can be a preset time, or when the prediction accuracy of the model is lower than the accuracy threshold, etc.
[0052] Correspondingly, this embodiment also provides a fermented grains layer recognition device based on the contents of lactic acid, acetic acid and caproic acid, as Figure 4 shown, including:
[0053] Data acquisition module: used to obtain the contents of lactic acid, acetic acid and caproic acid in the fermented grains sample data, and the corresponding fermented grains layer information, and construct a three major acids-fermented grains layer dataset;
[0054] Model creation and training module: Based on the three major acids-fermented grains layer dataset, with the contents of lactic acid, acetic acid and caproic acid as the input and the fermented grains layer as the output, establish and train a fermented grains layer prediction model;
[0055] Fermented grains layer prediction module: Obtain the contents of lactic acid, acetic acid and caproic acid in the fermented grains to be predicted, and input them into the trained fermented grains layer prediction model to obtain the prediction result of the fermented grains layer.
[0056] Furthermore, it further includes: used to regularly obtain the prediction result of the fermented grains layer. When the accuracy of the prediction result of the fermented grains layer is lower than the preset threshold, trigger an alarm, and update the three major acids-fermented grains layer dataset and the fermented grains layer prediction model.
Claims
1. A method for identifying the levels of fermented grains based on the contents of lactic acid, acetic acid and caproic acid, characterized in that Including: Step 1: Obtain the contents of lactic acid, acetic acid, and caproic acid in the fermented grains sample data, mark the fermented grains layer information, and construct a three-acid-fermented grains layer dataset; Step 2: Based on the three-acid-fermented grains layer dataset, using the contents of lactic acid, acetic acid, and caproic acid as inputs and the fermented grains layer as the output, establish and train a fermented grains layer prediction model; Step 3: Obtain the contents of lactic acid, acetic acid, and caproic acid in the fermented grains to be predicted, and input them into the trained fermented grains layer prediction model to obtain the prediction result of the fermented grains layer.
2. The method for identifying the distiller's grains layer based on the contents of lactic acid, acetic acid, and caproic acid according to claim 1, wherein The contents of lactic acid, acetic acid, and caproic acid are collected in the form of mass concentration or molar concentration.
3. The method for identifying the levels of fermented grains based on the contents of lactic acid, acetic acid, and caproic acid according to claim 1, wherein The fermented grains layer is the upper layer of fermented grains, the middle layer of fermented grains, or the bottom layer of fermented grains.
4. The method for identifying the levels of fermented grains based on the contents of lactic acid, acetic acid and caproic acid according to claim 1, wherein The fermented grains layer prediction model is constructed using MLP, LSTM, RNN, CNN, GRU, and / or Transformer.
5. The method for identifying the distiller's grains layer based on the contents of lactic acid, acetic acid, and caproic acid according to claim 1, wherein The steps for training the fermented grains layer prediction model are: Divide the data in the three-acid-fermented grains layer dataset into a training set and a test set according to a preset ratio, use the training set to train the fermented grains layer prediction model, use the test set to determine the prediction accuracy of the fermented grains layer prediction model. When the error function of the test set is less than the error function threshold and the test set accuracy rate is higher than the accuracy rate threshold, the training of the fermented grains layer prediction model is completed.
6. The method for identifying the levels of fermented grains based on the contents of lactic acid, acetic acid, and caproic acid according to claim 5, wherein, The error function is the cross-entropy loss function, and the expression is as follows: Where M is the number of samples; C is the number of fermented grains layers; p ij is the true label of the i-th sample corresponding to the j-th fermented grains layer; is the probability that the i-th sample is predicted as the j-th fermented grains layer.
7. The method for identifying the distiller's grains layer based on the contents of lactic acid, acetic acid, and caproic acid according to claim 1, wherein After obtaining the contents of lactic acid, acetic acid, and caproic acid, it also includes preprocessing them, including removing missing values and outliers, and performing normalization processing.
8. The method for identifying the distiller's grains layer based on the contents of lactic acid, acetic acid and caproic acid according to claim 1, wherein, Also including: Regularly update and / or supplement the three-acid-fermented grains layer dataset, and synchronously perform fine-tuning training and update on the fermented grains layer prediction model.
9. A device for identifying the layers of fermented grains based on the contents of lactic acid, acetic acid and caproic acid, characterized in that Including: Data acquisition module: Used to obtain the contents of lactic acid, acetic acid, and caproic acid in the fermented grains sample data, and the corresponding fermented grains layer information, and construct a three-acid-fermented grains layer dataset; Model creation and training module: Based on the three-acid-fermented grains layer dataset, using the contents of lactic acid, acetic acid, and caproic acid as inputs and the fermented grains layer as the output, establish and train a fermented grains layer prediction model; Fermented grains layer prediction module: Obtain the contents of lactic acid, acetic acid, and caproic acid in the fermented grains to be predicted, and input them into the trained fermented grains layer prediction model to obtain the prediction result of the fermented grains layer.
10. The apparatus for identifying the levels of fermented grains based on the contents of lactic acid, acetic acid and caproic acid according to claim 9, wherein, Also including: Abnormal alarm module: Used to regularly obtain the prediction result of the fermented grains layer. When the prediction result accuracy of the fermented grains layer is lower than the preset threshold, trigger an alarm, and update the three-acid-fermented grains layer dataset and the fermented grains layer prediction model.