A method for identifying and warning abnormal temperature in liquor fermentation process
By rapidly binding the temperature measurement system and neural network model, and fitting the temperature curve with historical data, the problems of automation and accuracy of temperature monitoring during the fermentation process of baijiu were solved, enabling timely early warning of temperature anomalies and ensuring the smooth progress of the fermentation process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN PROVINCIAL ANALYSIS & TESTING SERVICE CENT
- Filing Date
- 2022-10-27
- Publication Date
- 2026-05-26
AI Technical Summary
In the current process of baijiu fermentation, the fermentation temperature monitoring mainly relies on manual judgment, which lacks automation and data integration, resulting in untimely and inappropriate judgments and failing to effectively guide the production process.
A rapid binding temperature measurement system is used to collect temperature data. Temperature curves are fitted by combining historical data. Abnormal temperatures are automatically identified and warned through clustering and neural network models, including the application of polynomial functions to fill missing data and BP neural network models.
It enables timely and appropriate identification of temperature anomalies during the fermentation process of baijiu, ensuring the normal progress of fermentation and improving the automation and accuracy of temperature monitoring.
Smart Images
Figure CN115600147B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a method for identifying and issuing early warnings of abnormal fermentation temperatures during the fermentation process of baijiu (Chinese liquor). Background Technology
[0002] Currently, in the fermentation process of baijiu production, fermentation temperature is the main indicator for judging the quality of fermentation.
[0003] In existing technologies, fermentation temperature monitoring can generally only collect temperature data. Whether the fermentation temperature is abnormal can only be determined manually. Furthermore, the determination of fermentation temperature usually relies on a simple judgment of whether the temperature is too high or too low. The data source is singular and cannot be automatically combined with the production fermentation process. As a result, the judgment is prone to being untimely and inappropriate, and it does not have any guiding significance for the production fermentation process. Summary of the Invention
[0004] In view of this, the present invention provides a method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor). By relying on historical data and combining it with fermentation time, abnormal temperatures during the fermentation process can be identified in real time.
[0005] To solve the above technical problems, the technical solution of the present invention is to adopt a method for identifying and warning of abnormal temperatures in the fermentation process of baijiu (Chinese liquor), including:
[0006] Temperature data from several temperature measurement points during the fermentation process in the target fermentation pit were collected at regular intervals, divided into several time points.
[0007] The temperature data from the same temperature measurement point during a single fermentation process are fitted into a temperature curve.
[0008] The temperature curve is scored based on the final fermentation results;
[0009] Filter out temperature curves with scores higher than the threshold;
[0010] The temperature curves of the same temperature measurement point are fitted to obtain the temperature range of each time point, and the temperature ranges of all time points are merged into the standard temperature band of the temperature measurement point.
[0011] The temperature data collected during the fermentation process in the target fermentation pit is analyzed using a standard curve, and an alarm is triggered if the temperature data is abnormal.
[0012] As an improvement, a rapid binding temperature measurement system is used to collect temperature data during the fermentation process of the target fermentation pit; the rapid binding temperature measurement system includes a server and several monitoring instruments; each of the several monitoring instruments is equipped with its own unique information tag representing identity information; it also includes a rapid binding device, which is connected to the server; the rapid binding device is equipped with a reading module for reading information tags.
[0013] As a further improvement, the method for periodically collecting temperature data from several temperature measurement points during the fermentation process in the target fermentation pit includes:
[0014] Before data collection begins, locate the target cellar where you intend to install the monitoring instrument in the app or mini-program for quick device binding;
[0015] The rapid binding device obtains the identity information on the information tag on the monitoring instrument through scanning, reading and other methods, and binds the monitoring instrument to the target pit.
[0016] After the binding is completed, the monitoring instrument is placed at the corresponding temperature measurement point in the target cellar, and the monitoring instrument will transmit the monitored temperature data back to the server.
[0017] As another further improvement, when fitting temperature data from the same measurement point to a temperature curve, if temperature gaps are found, they are filled in, including:
[0018] Define a polynomial function;
[0019] The existing temperature data is substituted into a polynomial function to solve for the function with the smallest squared error.
[0020] The function retrieves the temperature data for the missing time point by inputting the time point of the missing temperature data into the function.
[0021] As an improvement, the method of fitting the temperature curves of the same selected temperature measurement points to obtain the temperature range for each time point includes:
[0022] The temperature data on the highest-scoring temperature curve is used as the cluster center, and the temperature data with the closest Euclidean distance to the cluster center is selected.
[0023] Calculate the average value of the selected temperature data as the cluster center, perform clustering iteration, and re-select the temperature data with the closest Euclidean distance to the cluster center. Repeat this step until the iteration is stable.
[0024] The selected temperature data is fitted into a temperature range.
[0025] As an improvement, the temperature anomaly includes:
[0026] The temperature is too high; the temperature at a certain time point collected exceeds the temperature range of that time point in the standard temperature band.
[0027] The temperature is too low; the temperature at a certain time point is lower than the temperature range of that time point in the standard temperature band.
[0028] The temperature rises too quickly, and the temperature at N consecutive time points exceeds the temperature range of the corresponding time point in the standard temperature band.
[0029] The temperature rises too slowly, and the temperature collected at M consecutive time points is lower than the temperature range of the corresponding time point in the standard temperature band.
[0030] Temperature stagnation occurred, and the temperature remained unchanged for L consecutive time points.
[0031] As an improvement, the standard temperature band is optimized using the collected temperature data.
[0032] This invention also provides another method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor), including:
[0033] Temperature data from several temperature measurement points during the fermentation process in the target fermentation pit were collected at regular intervals, divided into several time points.
[0034] During the primary fermentation process, the temperature data of the same temperature measurement point are fitted into several temperature curves according to the time points. The temperature curve of each time point includes the temperature data of that time point and all previous time points.
[0035] Temperature anomalies for each curve are manually identified to form training and validation sets;
[0036] For each temperature measurement point, a neural network model corresponding to the number of time points is constructed. The model is trained by inputting the corresponding training set and validated using the corresponding validation set.
[0037] The collected temperature data is input into the corresponding neural network model for identification and the identification result is output. If the identification result shows that the temperature is abnormal, an alarm is triggered.
[0038] As an improvement, the output of the neural network model is an array containing 5 elements. The 5 elements represent 5 states: temperature too high, temperature too low, temperature rises too fast, temperature rises too slow, and temperature stagnation. Each element has two states: 0 and 1, where 1 represents yes and 0 represents no.
[0039] As an improvement, the neural network model is a BP neural network model; the ratio of the training set to the validation set formed by manual identification is 7:3.
[0040] The advantages of this invention are: by combining historical data from temperature measurement points in the fermentation pit, and through analysis and screening of the historical data, the optimal temperature range at that time point is obtained, providing a reference for the subsequent fermentation process, and timely and appropriate abnormal temperature identification and early warning, effectively ensuring the normal fermentation of baijiu. Attached Figure Description
[0041] Figure 1This is a flowchart of Example 1.
[0042] Figure 2 This is a flowchart of Example 2. Detailed Implementation
[0043] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to specific embodiments.
[0044] There are many types of baijiu (Chinese liquor) and their brewing processes vary greatly. Each manufacturer and even each fermentation pit has its own unique fermentation temperature requirements. Currently, the control of fermentation temperature generally relies on the experience of technicians, which lacks rigor and is not timely enough in response to abnormal temperatures.
[0045] To solve this problem, such as Figure 1 As shown, Embodiment 1 of the present invention provides a method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor), the specific steps of which include:
[0046] S11 collects temperature data from several temperature measurement points during the fermentation process of the target fermentation pit at regular intervals.
[0047] Distilleries typically have multiple fermentation pits, and each pit usually requires multiple temperature monitoring points, such as three layers (top, middle, and bottom), with each layer arranged in a nine-square grid, resulting in 27 temperature monitoring points. Therefore, each pit requires multiple monitoring instruments. Existing pit monitoring systems and instruments usually rely on manual data entry to bind the pits to the instruments, which is inconvenient. Operators are prone to mismatching the pits and temperature monitoring points with the instruments, leading to erroneous monitoring data.
[0048] In this embodiment, temperature data acquisition relies on a rapid binding temperature measurement system. The rapid binding temperature measurement system includes a server and several monitoring instruments; each monitoring instrument is equipped with its own unique information tag representing its identity; it also includes a rapid binding device connected to the server; the rapid binding device is equipped with a reading module for reading the information tags. The steps for acquiring temperature data using the rapid binding temperature measurement system include:
[0049] Before the S111 data collection begins, locate the target cellar and temperature measurement point where you intend to install the monitoring instrument in the APP or mini-program for quick device binding.
[0050] S112 uses a quick-binding device to obtain the identity information on the information tag on the monitoring instrument through scanning, reading, etc., and binds the monitoring instrument to the target pit. The information tag includes one or more of the following: barcode, QR code, and electronic tag (such as NFC tag, RFID tag). To obtain the information from the information tag, the quick-binding device can be one or more of the following: mobile phone, tablet computer, handheld terminal. The reading module on the quick-binding device can be one or more of the following: camera, NFC reading module, RFID reading module, etc., that works with the information tag. Of course, to achieve binding, the quick-binding device should also include compatible software, such as an APP or mini-program.
[0051] After the binding is completed, the S113 will place the monitoring instrument at the corresponding temperature measurement point in the target cellar, and the monitoring instrument will transmit the monitored temperature data back to the server.
[0052] Due to differences in fermentation processes, the fermentation time varies. Several time points can be set based on the fermentation time to collect temperature data at regular intervals. For example, if a type of baijiu (Chinese liquor) requires 80 days to ferment, the interval between these time points can be set to one day, meaning temperature data can be collected every 24 hours.
[0053] S12 fits the temperature data from the same temperature measurement point during a single fermentation process into a temperature curve.
[0054] During a single fermentation process, temperature data is collected simultaneously from multiple temperature measurement points, resulting in multiple fitted temperature curves. For example, if a wine cellar has 27 temperature measurement points, then there will be 27 fitted temperature curves for a single fermentation process. The horizontal axis of this curve represents time, and the vertical axis represents temperature.
[0055] During the temperature data collection process, data loss can occur due to equipment malfunctions, transmission errors, or other reasons. This is reflected in the temperature curve as the entire curve being invalidated due to missing data points. In the early days, due to data scarcity, each collected curve was extremely valuable. Therefore, to avoid wasting data, it was necessary to fill in missing data points. A relatively simple existing method is to fill in intermediate points using the average of two consecutive points, but this method is not feasible for cases of continuously missing data. The method for filling in missing temperature data in this invention is as follows:
[0056] S121 Defines a polynomial function y = a0 + a1x + a2x 2 +a3x 3 Where y is temperature, x is time point, and a0 to a3 are missing coefficients. 。
[0057] S122 substitutes the existing temperature data into the polynomial function y = a0 + a1x + a2x2 +a3x 3 In this process, the function that minimizes the sum of squared errors is calculated. The purpose is to correct for missing coefficients in the polynomial function, obtaining a function for predicting missing temperature data.
[0058] S123 inputs the time point of the missing temperature data into the function obtained in step S122 to obtain the temperature data at that time point, and then fills the temperature data of the missing time point into the temperature curve to obtain a complete temperature curve again.
[0059] S13 scores the temperature curve based on the final fermentation results.
[0060] Since the quality of fermentation results is relatively subjective, it needs to be manually scored by the winery's process personnel. For example, on a 100-point scale, an excellent fermentation result corresponds to a temperature curve of 90-100 points, a good temperature curve of 80-90 points, an average temperature curve of 70-80 points, and an unqualified temperature curve of below 70 points.
[0061] S14 filters out temperature curves with scores higher than the threshold.
[0062] Due to the subjectivity of fermentation results and the multi-factor influence of the fermentation process, even a temperature curve with a score of 100 may not be optimal at every time point. Therefore, it is necessary to obtain a certain number of temperature curves with high scores for subsequent identification and early warning work. In this invention, a threshold is set to filter temperature curves. Furthermore, the threshold setting can vary depending on the amount of data. When there is less data initially, the threshold can be set lower to retain more curves, providing more reference data for later temperature range fitting. Conversely, when the data becomes larger later, the threshold can be increased to make the later fitted range more accurate.
[0063] S15 fits the temperature curve of the same temperature measurement point selected from the screening to obtain the temperature range of each time point, and merges the temperature ranges of all time points into the standard temperature band of the temperature measurement point.
[0064] Before fitting a standard temperature band, the temperature data needs to be processed to fit a temperature range for each time point. Specifically, this includes:
[0065] S151 uses the temperature data on the highest-scoring temperature curve as the cluster center and selects the temperature data with the closest Euclidean distance to the cluster center. For example, if the highest score among the currently collected temperature curves is 98, then the temperature data at each time point on that curve will be used as the first cluster center for that time point. Then, the 50% of the temperature data at that time point from other temperature curves that have the closest Euclidean distance to the cluster center will be selected.
[0066] S152 calculates the average value of the selected temperature data as the cluster center and performs clustering iterations, then re-selects the temperature data with the closest Euclidean distance to the cluster center. This step is repeated until the iteration is stable. The average value of the selected 50% of temperature data is calculated and used as the center for re-clustering. Then, the average value of the 50% of temperature points with the closest Euclidean distance to the new cluster center is calculated again and used as the cluster center. This process is repeated until the iteration is stable. Iterative stability means that the selected temperature data remains unchanged.
[0067] S153 fits the selected temperature data into a temperature range. From the final 50% of temperature data, a maximum and a minimum point are selected to form the temperature range.
[0068] By merging the temperature ranges at each time point, a standard temperature band for that temperature measurement point can be obtained.
[0069] S6 uses a standard curve to identify the temperature data collected during the fermentation process in the target fermentation pit, and triggers an alarm if the temperature data is abnormal.
[0070] The so-called temperature anomalies in this invention are of five types:
[0071] The temperature is too high; the temperature at a certain time point collected exceeds the temperature range of that time point in the standard temperature band.
[0072] The temperature is too low; the temperature at a certain time point is lower than the temperature range of that time point in the standard temperature band.
[0073] The temperature rises too quickly, and the temperature at N consecutive time points exceeds the temperature range of the corresponding time point in the standard temperature band.
[0074] The temperature rises too slowly, and the temperature collected at M consecutive time points is lower than the temperature range of the corresponding time point in the standard temperature band.
[0075] Temperature stagnation occurred, and the temperature remained unchanged for L consecutive time points.
[0076] When the collected temperature data meets one of the above abnormal conditions, the system will issue an alarm and display the type of temperature abnormality.
[0077] S17 optimizes the standard temperature band using the collected temperature data.
[0078] A certain amount of data is needed to fit a standard temperature band. Before that, to ensure smooth operation, an initial standard curve can be manually established for identifying abnormal temperatures. Once a sufficient amount of temperature curves has been collected, the standard temperature band can be fitted according to steps S12-S15. Furthermore, the more high-resolution temperature curves collected subsequently, the more accurate the fitted standard temperature band will be.
[0079] In addition, when the process is adjusted or there is a significant deviation from the current standard temperature range, it is also necessary to refit and optimize.
[0080] like Figure 2 As shown in Embodiment 2 of the present invention, another method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) is also provided. This method uses an artificial intelligence neural network model to classify and judge temperature data, thereby identifying whether the temperature is abnormal. Since the artificial intelligence neural network model requires massive amounts of data for training to adjust parameter weights and output more accurate results, in the early stages of a project where data is insufficient but temperature anomaly identification is required, the method of Embodiment 1 can be used to collect data and identify temperatures simultaneously. Once sufficient data has been collected, the method of this embodiment can be used to identify abnormal temperatures. The specific steps are as follows:
[0081] S21 Collects temperature data from several temperature measurement points during the fermentation process of the target fermentation pit at several time points; this step is the same as in Example 1, and will not be repeated here.
[0082] S22 fits the temperature data of the same temperature measurement point during the primary fermentation process into several temperature curves according to time points. The temperature curve of each time point includes the temperature data of that time point and all previous time points.
[0083] Unlike Example 1, this example requires fitting a temperature curve for each time point of each temperature measurement point, and this temperature curve includes the temperature data of all previous time points. For example, the temperature curve at the 75th time point includes all the temperature data of the previous 75 time points, the temperature curve at the 76th time point includes all the temperature data of the previous 76 time points, and so on.
[0084] S23 manually identifies temperature anomalies in each curve to form training and validation sets.
[0085] Abnormalities in each temperature curve were manually identified. The identification results included: normal, excessively high temperature, excessively low temperature, excessively rapid temperature rise, excessively slow temperature rise, and stagnant temperature. The identification criteria were consistent with those in Example 1. 70% of the identified temperature curves were used as the training set, and 30% as the validation set, i.e., the ratio of training set to validation set was 7:3.
[0086] S24 constructs a neural network model corresponding to the number of time points for each temperature measurement point, inputs the corresponding training set for training, and uses the corresponding validation set for validation, in order to train the neural network model.
[0087] In this embodiment, a backpropagation (BP) neural network model is used, with 3 hidden layers containing 5, 4, and 3 neurons respectively; a learning rate of 0.01; and a training error of 0.01. After the model is built, the training set is input into the model, and its output is compared with the validation set. Then, the parameter weights are adjusted to complete the model training.
[0088] The model outputs a 5-tuple A[a1,a2,a3,a4,a5], where a1 represents excessively high temperature, a2 represents excessively low temperature, a3 represents excessively rapid temperature rise, a4 represents excessively slow temperature rise, and a5 represents stagnant temperature. Each element has only two states: 0 and 1, where 1 represents yes and 0 represents no. For example, [0,0,1,0,0] represents excessively rapid temperature rise, while [0,0,0,0,0] represents normal temperature rise.
[0089] S25 inputs the collected temperature data into the corresponding neural network model for identification and outputs the identification result. If the identification result shows that the temperature is abnormal, an alarm is triggered.
[0090] After the model is trained, the collected temperature curves are input into the model. The output array can be used to determine whether there is a temperature anomaly. If so, the system will issue an alarm and display the type of temperature anomaly.
[0091] The above are merely preferred embodiments of the present invention. It should be noted that the above preferred embodiments should not be considered as limitations on the present invention, and the scope of protection of the present invention should be determined by the scope defined in the claims. For those skilled in the art, several improvements and modifications can be made without departing from the spirit and scope of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor), characterized in that... include: Temperature data from several temperature measurement points during the fermentation process in the target fermentation pit were collected at regular intervals, divided into several time points. The temperature data from the same temperature measurement point during a single fermentation process are fitted into a temperature curve. The temperature curve is scored based on the final fermentation results; Filter out temperature curves with scores higher than the threshold; The temperature curves of the selected temperature measurement points are fitted to obtain the temperature range for each time point. The temperature ranges of all time points are then merged into a standard temperature band for the temperature measurement points, including: The temperature data on the highest-scoring temperature curve is used as the cluster center, and the temperature data with the closest Euclidean distance to the cluster center is selected. Calculate the average value of the selected temperature data as the cluster center, perform clustering iteration, and re-select the temperature data with the closest Euclidean distance to the cluster center. Repeat this step until the iteration is stable. Fit a temperature range to the selected portion of the temperature data; The temperature data collected during the fermentation process in the target fermentation pit is analyzed using a standard curve, and an alarm is triggered if the temperature data is abnormal.
2. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 1, characterized in that: A rapid binding temperature measurement system is used to collect temperature data during the fermentation process in the target fermentation pit. The rapid binding temperature measurement system includes a server and several monitoring instruments. Each of the monitoring instruments is equipped with a unique information tag that represents its identity. The system also includes a rapid binding device, which is connected to the server. The rapid binding device is equipped with a reading module for reading the information tags.
3. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 2, characterized in that... The method for collecting temperature data at several temperature measurement points during the fermentation process of the target fermentation pit includes: before the collection begins, finding the target fermentation pit where the monitoring instrument is to be installed in the APP or mini-program for quick device binding. The rapid binding device obtains the identity information on the information tag on the monitoring instrument through scanning, reading and other methods, and binds the monitoring instrument to the target pit. After the binding is completed, the monitoring instrument is placed at the corresponding temperature measurement point in the target cellar, and the monitoring instrument will transmit the monitored temperature data back to the server.
4. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 1, characterized in that... When fitting temperature data from the same temperature measurement point to a temperature curve, if missing temperature data is found, it is filled in, including: defining a polynomial function; The existing temperature data is substituted into a polynomial function to solve for the function with the smallest squared error. The function retrieves the temperature data for the missing time point by inputting the time point of the missing temperature data into the function.
5. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 1, characterized in that... The temperature anomalies include: excessively high temperatures, where the temperature at a certain point in time exceeds the temperature range of that point in the standard temperature band; The temperature is too low; the temperature at a certain time point is lower than the temperature range of that time point in the standard temperature band. The temperature rises too quickly, and the temperature at N consecutive time points exceeds the temperature range of the corresponding time point in the standard temperature band. The temperature rises too slowly, and the temperature collected at M consecutive time points is lower than the temperature range of the corresponding time point in the standard temperature band. Temperature stagnation occurred, and the temperature remained unchanged for L consecutive time points.
6. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 1, characterized in that... The standard temperature band is optimized using the collected temperature data.
7. A method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor), characterized in that... include: Temperature data from several temperature measurement points during the fermentation process in the target fermentation pit were collected at regular intervals, divided into several time points. During the primary fermentation process, the temperature data of the same temperature measurement point are fitted into several temperature curves according to the time points. The temperature curve of each time point includes the temperature data of that time point and all previous time points. Temperature anomalies for each curve are manually identified to form training and validation sets; For each temperature measurement point, a neural network model corresponding to the number of time points is constructed. The model is trained by inputting the corresponding training set and validated using the corresponding validation set. The collected temperature data is input into the corresponding neural network model for identification and the identification result is output. If the identification result shows that the temperature is abnormal, an alarm is triggered.
8. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 7, characterized in that: The neural network model outputs an array containing 5 elements. These 5 elements represent five states: excessively high temperature, excessively low temperature, excessively rapid temperature rise, excessively slow temperature rise, and stagnant temperature. Each element has two states: 0 and 1. 1 represents yes, and 0 represents no.
9. The method for identifying and warning of abnormal temperatures during the fermentation process of baijiu (Chinese liquor) according to claim 7, characterized in that: The neural network model is a BP neural network model; the ratio of the training set to the validation set formed by manual identification is 7:3.