CN114897218A - Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters - Google Patents
Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters Download PDFInfo
- Publication number
- CN114897218A CN114897218A CN202210400897.1A CN202210400897A CN114897218A CN 114897218 A CN114897218 A CN 114897218A CN 202210400897 A CN202210400897 A CN 202210400897A CN 114897218 A CN114897218 A CN 114897218A
- Authority
- CN
- China
- Prior art keywords
- soy sauce
- days
- index
- natural oil
- fermentation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000855 fermentation Methods 0.000 title claims abstract description 76
- 230000004151 fermentation Effects 0.000 title claims abstract description 76
- 235000013555 soy sauce Nutrition 0.000 title claims abstract description 60
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 34
- 238000011156 evaluation Methods 0.000 claims abstract description 24
- 239000002994 raw material Substances 0.000 claims abstract description 15
- 230000015654 memory Effects 0.000 claims abstract description 14
- 238000010606 normalization Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims abstract description 8
- 235000010469 Glycine max Nutrition 0.000 claims abstract description 7
- 244000068988 Glycine max Species 0.000 claims abstract description 7
- 238000012545 processing Methods 0.000 claims abstract description 7
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 15
- 229910052757 nitrogen Inorganic materials 0.000 claims description 8
- 239000002253 acid Substances 0.000 claims description 5
- 150000001413 amino acids Chemical class 0.000 claims description 5
- 150000003839 salts Chemical class 0.000 claims description 5
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 abstract description 26
- 239000000047 product Substances 0.000 description 9
- 210000004027 cell Anatomy 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000012549 training Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 230000007547 defect Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 235000013305 food Nutrition 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000002354 daily effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Pure & Applied Mathematics (AREA)
- Quality & Reliability (AREA)
- Educational Administration (AREA)
- Mathematical Analysis (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Primary Health Care (AREA)
- Probability & Statistics with Applications (AREA)
Abstract
The invention discloses a method for predicting soy sauce yield according to Guangdong soy sauce fermentation process parameters, which comprises the following steps: s1, setting evaluation indexes of natural oil in the fermentation tank at 15 days, 45 days and 90 days after feeding and numerical values of temperature indexes from 15 days to 180 days after feeding; s2, respectively carrying out normalization processing on the numerical values of the natural oil evaluation index and the temperature index, and then converting the numerical values into two three-dimensional matrixes to obtain two index sequences; s3, respectively inputting the two obtained index sequences into two long-short term memory neural networks, and finally obtaining two one-dimensional feature vectors; s4, splicing and fusing the quality index of the fermentation tank raw material soybean with the two one-dimensional eigenvectors obtained in S3 to form a long eigenvector, reducing dimensions through a full-connection layer, and outputting the predicted value of the soy sauce yield of each fermentation tank. The invention solves the problems that the fermentation process of the prior soy sauce natural oil is not transparent and the yield is difficult to quantify according to the fermentation process, and has guiding significance for controlling the fermentation production process of the soy sauce.
Description
Technical Field
The invention relates to a method for predicting soy sauce yield according to Guangdong soy sauce fermentation process parameters.
Background
Soy sauce plays an extremely important role in both our daily life and in the food industry. With the improvement of the whole technical level, the management level and the living standard of people in the food industry in China, the yield and the output value of soy sauce in China are rapidly increased in recent years.
At present, the soy sauce products accounting for 70 percent of the whole country adopt a wide-range high-salt dilute state fermentation process. The Guangdong soy sauce adopts a production process of natural fermentation of sun-dried night dew, the fermentation process is influenced by multiple factors such as raw material proportion, strains, climate environment, equipment parameters and the like, and the quality fluctuation range is large.
The fermentation of the Cantonese soy sauce is a multi-parameter, long-period and dynamic change process, most of adopted equipment is of a traditional non-standard type, and the problem of difficulty in acquiring real-time data exists. Therefore, although the actual parameters of the material in the processing period, such as the temperature, are changed values, the retained data are only single-point data and the representativeness of the data is very limited because the data are collected manually all the time. The key quality indexes in the soy sauce fermentation process fluctuate along with seasonal changes, for example, in the aspect of fermentation, the formula of the soy sauce mash needs to be adjusted according to climate change, and the value of the collected single-point data is further reduced.
In a word, the Guangdong soy sauce fermentation process always adopts a mode of manual recording, so that the problems of very limited traceability data volume and high analysis difficulty exist, and the optimization development of the Guangdong soy sauce fermentation process is restricted. Even providing guidance for the optimization of the moromi formula for the same month of the next year presents difficulties. This also causes a problem that the soy sauce yield in the first and fourth seasons of the year is always remarkably low.
Disclosure of Invention
The prediction of the soy sauce yield plays an important role in exploring the defects in the production process, reducing the raw material conversion cost, optimizing the whole production flow and realizing sustainable high-quality production. The invention mainly aims to search a method for predicting the soy sauce yield according to the Guangdong soy sauce fermentation process parameters based on currently recorded relevant data in the Guangdong soy sauce fermentation process so as to help to find key parameters for improving the product yield and guide the regulation and control mode of the soy sauce fermentation production process.
In the current Cantonese soy sauce fermentation process, soy sauce fermentation needs to be dried in the sun and exposed at night for 180 days, fermentation tanks are respectively subjected to sampling detection on 15 th day, 45 th day and 90 th day after feeding, numerical values of natural oil fermentation indexes including total acid (g/100mL), amino acid nitrogen (g/100mL), salt (g/100mL), pH value and the like after the corresponding days of fermentation are obtained, and the temperature condition of the fermentation tanks every day is recorded from the feeding day.
In order to realize the purpose of the invention, the technical scheme is as follows: a method for predicting soy sauce yield according to Guangdong soy sauce fermentation process parameters comprises the following steps:
s1, setting evaluation indexes of natural oil in the fermentation tank at 15 days, 45 days and 90 days after feeding and numerical values of temperature indexes from 15 days to 180 days after feeding;
s2, respectively carrying out normalization processing on the numerical values of the natural oil evaluation index and the temperature index, eliminating dimensional influence, and then respectively converting the numerical values into two three-dimensional matrixes comprising fermentation tank sample number, time sequence and characteristic index according to the characteristics of the two kinds of data to obtain a natural oil evaluation index sequence and a temperature index sequence for input of a neural network;
s3, respectively inputting the two index sequences obtained in S2 into two long-short term memory neural networks to refine the eigenvectors, and finally obtaining two one-dimensional eigenvectors;
s4, splicing and fusing the quality index of the fermentation tank raw material soybean and the two one-dimensional eigenvectors obtained in S3 on the characteristic dimension to form a long eigenvector, reducing the dimension through a full-connection layer, and outputting a predicted value of the soy sauce yield of each fermentation tank;
the weight parameters of the two long-short term memory neural networks in step S3 are determined as follows:
s3-1, acquiring evaluation indexes of natural oil in the fermentation tank at 15 days, 45 days and 90 days after feeding and historical values of temperature indexes from 15 days to 180 days after feeding;
s3-2, respectively carrying out normalization processing on the numerical values of the natural oil evaluation index and the temperature index, eliminating dimensional influence, and then respectively converting the numerical values into two three-dimensional matrixes comprising fermentation tank sample number, time sequence and characteristic index according to the characteristics of the two kinds of data to obtain a natural oil evaluation index sequence and a temperature index sequence for input of a neural network;
s3-3, respectively inputting the two index sequences obtained in the step S3-2 into two long-short term memory neural networks to refine the eigenvectors, and finally obtaining two one-dimensional eigenvectors;
s3-4, splicing and fusing the quality index of the fermentation tank raw material soybean and the two one-dimensional eigenvectors obtained in the step S3-3 on the characteristic dimension to form a long eigenvector, reducing the dimension through a full connection layer, and outputting a predicted value of the soy sauce yield of each fermentation tank;
s3-5 measures the difference between the yield predicted value and the true value in S3-4 through a loss function, calculates the loss of the neural network through forward propagation, updates the weight parameters of the S3-3 neural network through a back propagation algorithm, and repeats S3-3 to S3-5 until the loss function converges.
After training through S3-5, it is demonstrated that the production process and the product yield can be associated through S3-1 to S3-4. According to the characteristic that the neural network can automatically extract the characteristics, the corresponding relation between the fermentation time, the temperature and the natural oil evaluation index is sought by utilizing the neural network, the initial state information of the raw materials is supplemented by splicing the index data of the raw materials of the fermentation tank, so that the constructed model can accurately describe the relation between the process parameters and the soy sauce yield, and the problems that the existing soy sauce natural oil fermentation process is opaque and the yield is difficult to quantify according to the fermentation process are solved.
Preferred embodiments of the present invention include:
the natural oil evaluation indices are total acid (g/100mL), amino acid nitrogen (g/100mL), salt (g/100mL), and pH, and the temperature indices are maximum temperature (deg.C) and minimum temperature (deg.C) per day.
The loss function adopted in S3-5 is a SmoothL1 loss function, the SmoothL1 loss function can avoid the defects of break points and non-smoothness in model training, and the expression is as follows:
and x represents the difference between the predicted value and the true value, the loss function is insensitive to outliers and abnormal values, the gradient change is relatively smaller, and the robustness of the model is improved.
The quality indexes are total nitrogen (g/100mL) and moisture (%).
In the above steps, MinMax is adopted for normalization, so that the value of the data is compressed to be between 0 and 1, and the specific expression is as follows:
xi represents the input vector, and in order to quantify the degree of variation of the index with the number of days, the values of max and min are selected among the indexes of the whole day sequence of all fermenters.
The long-short term memory neural network in the above steps is bidirectional, and can process the sequence obtained in S2 from two opposite directions, learn the time dependence of the index, and enable the extracted deep features to better retain the implicit information of the index.
Has the advantages that:
according to the characteristic that the neural network can automatically extract the characteristics, the corresponding relation between the fermentation time, the temperature and the natural oil evaluation index is sought by utilizing the neural network, the initial state information of the raw materials is supplemented by splicing the index data of the raw materials of the fermentation tank, so that the constructed model can accurately describe the relation between the process parameters and the soy sauce yield, and the problems that the existing soy sauce natural oil fermentation process is opaque and the yield is difficult to quantify according to the fermentation process are solved. The method can establish the association between the production process and the product yield, thereby predicting the product yield according to the range of the controllable parameters in the production process, and further reversely deducing the controllable production parameter interval capable of obtaining stable yield, thereby guiding stable production and increasing economic benefit.
Drawings
FIG. 1 is a block flow diagram of the present invention;
FIG. 2 is a block diagram of a neural network in accordance with a preferred embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be further described below with reference to specific embodiments and accompanying drawings. It is to be understood that the specific examples described below are merely preferred embodiments of the invention and are not to be construed as the only alternative embodiments. All other embodiments obtained by those skilled in the art without inventive step based on the technical idea of the present invention should fall within the protection scope of the present invention as long as the protection scope of the present invention is claimed in the claims.
Fermentation of Cantonese soy sauce requires exposure to the sun for 180 days. Generally, sampling tests are performed at the fermentation stages of 15 days, 45 days and 90 days after feeding the materials into the fermentation tank, and numerical values of fermentation indexes of the natural oil after fermentation corresponding to the days are obtained, wherein the numerical values comprise total acid (g/100mL), amino acid nitrogen (g/100mL), salt (g/100mL) and pH value so as to represent the variation trend of the fermentation degree. The four indexes are not independent, and interaction exists among the four indexes. Meanwhile, the temperature change of each fermenter per day was recorded from the day of feeding.
The specific process for predicting the soy sauce yield based on the related data obtained by the Cantonese soy sauce fermentation process is as follows:
step 1: historical values of natural oil evaluation indexes in a fermentation tank at 15 days, 45 days and 90 days after feeding and temperature indexes from 15 days to 180 days after feeding are obtained, wherein the natural oil evaluation indexes are total acid (g/100mL), amino acid nitrogen (g/100mL), salt (g/100mL) and pH value, and the temperature indexes are used as the highest temperature (DEG C) and the lowest temperature (DEG C) of each day for the sake of convenience.
Only temperature data after 15 days are obtained in this step, because the natural oil evaluation index in the invention is recorded from day 15, and the temperature data from day 15 of the batch fermentation has an influence on the index.
Step 2: the first step is as follows: and respectively carrying out normalization treatment on the numerical values of the natural oil evaluation index and the temperature index to eliminate dimensional influence. The normalization processing mode specifically comprises the following steps:
carrying out MinMax normalization, wherein the expression is as follows:
xi represents the input vector, and in order to quantify the degree of variation of the index with the number of days, the values of max and min are selected among the indexes of the whole day sequence of all fermenters.
The second step is that: and respectively converting the two data into two three-dimensional matrixes comprising a fermentation tank sample number N, a time sequence (the natural oil indexes have 3 acquisition times, namely days 15, 45 and 90, and the temperature data are acquired from day 15 to day 180 and 165 acquisition times) and a characteristic index (the natural oil indexes have 4, the time indexes have 2, namely the highest temperature and the lowest temperature of the day) according to the characteristics of the two data to obtain a natural oil evaluation index sequence and a temperature index sequence for inputting a neural network model. Finally, for natural oil fermentation index, it can be extended to a sequence of nx 3 x 4; for the temperature index, it can be extended to a sequence of N × 165 × 2.
And step 3: and (3) respectively inputting the two index sequences obtained in the step (2) into two long-short term memory neural networks to refine the feature vectors to obtain two one-dimensional feature vectors.
The long-short term memory neural network adopted here is bidirectional, and can process the sequence obtained in step 2 from two opposite directions, and learn the time dependence of the index, so that the extracted deep features can better retain the implicit information of the index. Of course, the present invention may also employ unidirectional long-short term memory neural networks.
The long and short term memory neural network adopted by the invention performs increment and abandon on input information through an internal gate structure comprising a forgetting gate, an updating gate and an output gate. Wherein, the forgetting gate is used for controlling whether the information of the last neuron needs to be discarded at the current layer. If the hidden state at the moment of t is h t Then the forget gate can be expressed as:
f t ＝σ(W f *[h t-1 ，x t ]+b f )，
firstly, connecting the hidden state at the last moment, and obtaining a number from 0 to 1 through a sigmoid activation function, wherein 0 represents all abandonments, 1 represents all reservations, and the numbers between the abandonments and the reservations represent the ratio of the reservations; the update gate, which incorporates the past cell state information, the hidden information inside the cell at the old time, and the new input data, can be expressed as:
i t ＝σ(W i *[h t-1 ，x t ]+b i )；
the output gate can obtain the output value of the current cell and the hidden state value passed to the next cell, where the hidden state includes the related information input previously, and the output value of the current cell can be expressed as:
o t ＝σ(W o *[h t-1 ，x t ]+b o )，
the hidden state of the next cell can be expressed as:
h t ＝o t *tanh(C t )，
wherein, C t Indicating the state of the cell at the current time.
The long-short term memory neural network is realized through a Pythrch framework, the Pythrch is a manifold development framework in deep learning, supports dynamic building of the neural network, and has strong expansibility.
And 4, step 4: and (4) splicing and fusing the two eigenvectors output by the neural network in the step (3) on the characteristic dimension to form one long eigenvector. In order to combine the quality information of raw material soybeans to form continuous prediction of a production line, the quality indexes of soybeans are spliced on the basis of splicing and fusing long characteristic vectors, and the quality indexes adopted by the method comprise total nitrogen (g/100mL) and water (%), which are shown in figure 2. And reducing the dimension of the spliced final vector through a full-connection layer, and outputting the predicted value of the soy sauce yield of each fermentation tank.
And 5: and 4, measuring the difference between the yield prediction value and the true value in the step 4 by using a SmoothL1 loss function, calculating the loss of the neural network through forward propagation, updating the weight parameter of the neural network in the step 3 by using a backward propagation algorithm, repeating the steps 3-5 until the loss function is converged, and then predicting the soy sauce yield of each fermentation tank according to the steps 2-4 by setting the relevant parameters in the step 1.
The reason why the SmoothL1 loss function is adopted to train the model is that the defects of break points and non-smoothness in model training can be avoided, and the expression is as follows:
and x represents the difference between the predicted value and the true value, the loss function is insensitive to outliers and abnormal values, the gradient change is relatively smaller, and the robustness of the model is improved.
After the training of the step 5, the production process and the product yield can be associated through the steps 1 to 4, so that the product yield can be predicted according to the range of the controllable parameters in the production process, and the controllable production parameter interval capable of obtaining the stable yield can be deduced reversely, so that stable production is guided, and the economic benefit is increased.
The method is based on 3600 production data from the Cantonese soy sauce fermentation production line, and a training set and a testing set of the neural network are divided according to the ratio of 7: 3. After the weight parameters of the network are trained through the data of the training set, accuracy verification is carried out on the test set. The Mean Absolute Error (MAE) of the yield predictions in the resulting test set was 0.109, while the mean absolute deviation of the actual yields was 0.176. Therefore, the prediction yield has reference value by the method.
The yield prediction is a key task in the soy sauce production process, and the soy sauce yield prediction plays an important role in exploring the defects in the production process, reducing the raw material conversion cost, optimizing the whole production flow and realizing sustainable high-quality production. The method utilizes the characteristic that the neural network can automatically extract the characteristics, seeks the corresponding relation among the fermentation time, the temperature and the natural oil evaluation index, supplements the initial state information of the raw materials by splicing the index data of the raw materials of the fermentation tank, enables the constructed model to describe the relation between the process parameters and the soy sauce yield more accurately, and solves the problems that the existing soy sauce natural oil fermentation process is opaque, and the yield is difficult to quantify according to the fermentation process. The method can establish the association between the production process and the product yield, thereby predicting the product yield according to the range of the controllable parameters in the production process, and further reversely deducing the controllable production parameter interval capable of obtaining stable yield, thereby guiding stable production and increasing economic benefit.
The foregoing are merely preferred embodiments of the invention. However, the scope of protection of the present patent is not limited thereto. Those skilled in the art should also realize that such equivalent alterations and modifications are possible in light of the above teachings and within the spirit and scope of the appended claims.
Claims (6)
1. A method for predicting soy sauce yield according to Guangdong soy sauce fermentation process parameters is characterized by comprising the following steps:
s1, setting evaluation indexes of natural oil in the fermentation tank at 15 days, 45 days and 90 days after feeding and values of temperature indexes from 15 days to 180 days after feeding;
s2, respectively carrying out normalization processing on the numerical values of the natural oil evaluation index and the temperature index, eliminating dimensional influence, and then respectively converting the numerical values into two three-dimensional matrixes comprising fermentation tank sample number, time sequence and characteristic index according to the characteristics of the two kinds of data to obtain a natural oil evaluation index sequence and a temperature index sequence for input of a neural network;
s3, respectively inputting the two index sequences obtained in S2 into two long-short term memory neural networks to refine the eigenvectors, and finally obtaining two one-dimensional eigenvectors;
s4, splicing and fusing the quality index of the fermentation tank raw material soybean and the two one-dimensional eigenvectors obtained in S3 on the characteristic dimension to form a long eigenvector, reducing the dimension through a full-connection layer, and outputting a predicted value of the soy sauce yield of each fermentation tank;
the weight parameters of the two long-short term memory neural networks in step S3 are determined as follows:
s3-1, acquiring evaluation indexes of natural oil in the fermentation tank at 15 days, 45 days and 90 days after feeding and historical values of temperature indexes from 15 days to 180 days after feeding;
s3-2, respectively carrying out normalization processing on the numerical values of the natural oil evaluation index and the temperature index, eliminating dimensional influence, and then respectively converting the numerical values into two three-dimensional matrixes comprising fermentation tank sample number, time sequence and characteristic index according to the characteristics of the two kinds of data to obtain a natural oil evaluation index sequence and a temperature index sequence which are used for inputting a neural network;
s3-3, respectively inputting the two index sequences obtained in the step S3-2 into two long-short term memory neural networks to refine the eigenvectors, and finally obtaining two one-dimensional eigenvectors;
s3-4, splicing and fusing the quality index of the fermentation tank raw material soybean and the two one-dimensional eigenvectors obtained in the step S3-3 on the characteristic dimension to form a long eigenvector, reducing the dimension through a full connection layer, and outputting a predicted value of the soy sauce yield of each fermentation tank;
s3-5 measures the difference between the yield predicted value and the true value in S3-4 through a loss function, calculates the loss of the neural network through forward propagation, updates the weight parameters of the S3-3 neural network through a back propagation algorithm, and repeats S3-3 to S3-5 until the loss function converges.
2. The method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters of claim 1, wherein the natural oil evaluation indexes are total acid (g/100mL), amino acid nitrogen (g/100mL), salt content (g/100mL), and pH value, and the temperature indexes are maximum temperature (deg.C) and minimum temperature (deg.C) per day.
3. The method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters of claim 1, wherein the loss function used in S3-5 is SmoothL1 loss function, and the expression is:
x represents the difference between the predicted value and the true value.
4. The method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters of claim 1, wherein the quality indicators are total nitrogen (g/100mL) and moisture (%).
5. The method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters of claim 1, wherein MinMax is used for normalization, so as to compress the data value to 0-1, and the specific expression is as follows:
xi represents the input vector, and the values of max and min are selected in the index of the whole day sequence of all fermenters.
6. The method for predicting soy sauce yield according to Guangdong soy sauce fermentation process parameters as claimed in claim 1, wherein the long-short term memory neural network is a bidirectional long-short term memory neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210400897.1A CN114897218A (en) | 2022-04-15 | 2022-04-15 | Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210400897.1A CN114897218A (en) | 2022-04-15 | 2022-04-15 | Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114897218A true CN114897218A (en) | 2022-08-12 |
Family
ID=82718178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210400897.1A Pending CN114897218A (en) | 2022-04-15 | 2022-04-15 | Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114897218A (en) |
-
2022
- 2022-04-15 CN CN202210400897.1A patent/CN114897218A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109060001B (en) | Multi-working-condition process soft measurement modeling method based on feature transfer learning | |
| CN108898215B (en) | Intelligent sludge bulking identification method based on two-type fuzzy neural network | |
| CN111461922B (en) | Real-time prediction method for hot spot temperature of transformer based on extreme learning machine | |
| CN108710974B (en) | Water ammonia nitrogen prediction method and device based on deep belief network | |
| CN114169638A (en) | Water quality prediction method and device | |
| CN111768000A (en) | Industrial process data modeling method for online adaptive fine-tuning deep learning | |
| CN101995891B (en) | Method for online analysis of water content of solid master batch recovery system in aromatic acid production | |
| CN113836783A (en) | Digital regression model modeling method for main beam temperature-induced deflection monitoring reference value of cable-stayed bridge | |
| CN114403486B (en) | Intelligent control method of airflow type cut-tobacco drier based on local peak value coding circulation network | |
| CN110908361B (en) | Fermentation process soft measurement method based on online twin support vector regression | |
| CN116975645A (en) | Industrial process soft measurement modeling method based on VAE-MRCNN | |
| CN114117852B (en) | Regional heat load rolling prediction method based on finite difference working domain division | |
| CN114239397A (en) | Soft measurement modeling method based on dynamic feature extraction and local weighted deep learning | |
| CN116484275B (en) | Aluminum electrolysis anode effect monitoring method based on self-adaptive space-time diagram convolutional neural network | |
| CN106599391B (en) | Association vector machine soft measurement modeling method based on dynamic weighting of triangle angle values | |
| CN114897218A (en) | Method for predicting soy sauce yield according to Cantonese soy sauce fermentation process parameters | |
| CN116483036B (en) | Transformer-based self-encoder soft measurement modeling method | |
| CN115879750B (en) | Aquatic seedling environment monitoring management system and method | |
| CN110909492A (en) | Sewage treatment process soft measurement method based on extreme gradient lifting algorithm | |
| CN110619176A (en) | Aviation kerosene flash point prediction method based on DBN-RLSSVM | |
| CN114417740B (en) | Deep sea breeding situation sensing method | |
| CN114997486A (en) | Effluent residual chlorine prediction method of water works based on width learning network | |
| CN113837487A (en) | Power system load prediction method based on combined model | |
| Zhang et al. | An intelligent Attentional-GRU-based model for dynamic blood glucose prediction | |
| CN108764337B (en) | Underwater robot propeller fault degree identification method based on relative gray correlation degree boundary constraint |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |