Real-time monitoring method for fermentation process of high-degree rice wine based on intelligent sensor

Abstract

The invention relates to the technical field of fermentation process monitoring, and discloses a high-degree rice wine fermentation process real-time monitoring method based on an intelligent sensor. The method comprises the following steps: firstly, constructing a fermentation monitoring model and updating based on real-time sensor data, wherein the real-time sensor data comprises parameters such as a temperature fluctuation value; key parameters and the like in the model are identified through a data fusion algorithm; generating an initial adjustment strategy set based on the key parameters and the sensitive nodes; after receiving a user interaction instruction, dynamically correcting the initial adjustment strategy set; and finally, executing the corrected strategy and updating the model based on result feedback. According to the method, the fermentation process can be monitored in real time, key parameters and abnormal conditions can be accurately identified, an adjustment strategy is dynamically generated and corrected, and the stability and controllability of rice wine fermentation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fermentation process monitoring, and in particular to a real-time monitoring method for the fermentation process of high-proof rice wine based on an intelligent sensor. Background Art

[0002] Rice wine fermentation is a process that combines tradition and technology. It involves the metabolic activities of multiple microorganisms and is significantly influenced by environmental factors and operating conditions. Traditional rice wine fermentation production relies heavily on operator experience for process control, with regular sampling and testing used to assess fermentation status. This approach has significant limitations.

[0003] Sampling tests often have long intervals between tests, making it difficult to reflect subtle changes in the fermentation process in real time, and it is easy to miss the optimal opportunity for adjustment. For example, a sudden temperature fluctuation can affect microbial activity in a short period of time, but due to the sampling interval, operators may not be able to detect it in time, causing the fermentation direction to deviate from expectations. Furthermore, the accuracy of manual testing is significantly affected by factors such as the operator's skill level and sense of responsibility. Test results from different operators may vary, which in turn affects the judgment of the fermentation status.

[0004] With the advancement of industrialized production, the scale of rice wine fermentation continues to expand. Traditional empirical control methods are no longer able to meet the needs of large-scale, standardized production. Currently, some manufacturers have begun to introduce sensors to monitor the fermentation process, but existing monitoring methods still have many problems. For one thing, the data types collected by sensors are relatively simple, mostly focusing on a few parameters such as temperature and humidity, which makes it difficult to fully reflect the complex state of the fermentation system. For another, the collected data is often not effectively integrated and analyzed, existing only as independent numerical values. It is impossible to establish correlations between parameters, making it difficult to accurately identify key parameters and abnormal parameters.

[0005] Furthermore, existing monitoring systems lack dynamic adjustment capabilities. When abnormal parameters are detected, they are unable to generate tailored adjustment strategies based on the specific situation, and decision-making still relies on manual experience. Furthermore, sensors may drift over long periods of use, resulting in decreased data accuracy. Existing systems often lack comprehensive calibration mechanisms, further impacting monitoring reliability.

[0006] During rice wine fermentation, complex interactions exist between various parameters. Changes in one parameter can trigger chain reactions in multiple related parameters. For example, a shift in pH can affect the rate of alcohol growth, while changes in temperature can simultaneously affect pH and microbial metabolic efficiency. Existing technologies are unable to effectively capture these interdependencies, resulting in a limited understanding of the fermentation process and difficulties in achieving precise control. Furthermore, the lack of identification of sensitive nodes makes it difficult to quickly locate the root cause of fermentation anomalies, increasing the difficulty and time cost of troubleshooting.

[0007] Furthermore, different rice wine production companies differ in terms of fermentation equipment and raw material formulations. Existing monitoring methods often lack universality and are difficult to adapt to different production scenarios. When companies change raw materials or adjust production scale, the monitoring system requires extensive reconfiguration, which is cumbersome and inefficient. These issues are hindering the high-quality development of the rice wine fermentation industry, and a technical solution that can achieve real-time, accurate, and dynamic monitoring and control is urgently needed. Summary of the Invention

[0008] The purpose of the present invention is to provide a real-time monitoring method for the fermentation process of high-proof rice wine based on an intelligent sensor to solve the problems raised in the above background technology.

[0009] To achieve the above object, the present invention provides a method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor, the method comprising:

[0010] S100, constructing a fermentation monitoring model and updating the fermentation monitoring model based on real-time sensor data; wherein the fermentation monitoring model includes multiple parameter nodes, associations between parameter nodes, and status information of each parameter node, and the real-time sensor data includes temperature fluctuation value, humidity change rate, pH value offset, and alcohol content growth rate;

[0011] S200, identifying key parameters and abnormal parameters in the fermentation monitoring model and confirming sensitive nodes through a data fusion algorithm;

[0012] S300, generating an initial adjustment strategy set based on the key parameters and sensitive nodes; the initial adjustment strategy set includes a sensor calibration solution, an environment control solution, and a priority sorting solution;

[0013] S400: Receive a user interaction instruction, determine an adjustment operation type according to the user interaction instruction, and dynamically modify the initial adjustment strategy set based on the adjustment operation type;

[0014] S500: Execute the dynamically revised initial adjustment strategy set, and update the fermentation monitoring model based on the execution result feedback.

[0015] Preferably, the specific implementation of constructing the fermentation monitoring model in S100 includes:

[0016] S110, collecting historical fermentation data, wherein the historical fermentation data includes raw material batch records, fermentation tank temperature and humidity curves, pH value change logs, and alcohol content test reports;

[0017] S120: Perform multi-dimensional feature extraction on the historical fermentation data to generate status information of parameter nodes; the status information includes parameter type, sensitivity, control priority, parameter stability, historical matching accuracy, and dynamic weight; wherein:

[0018] The control priority is dynamically assigned according to the fermentation stage requirements, raw material characteristics and process standards to obtain the corresponding control priority coefficient;

[0019] The dynamic weight is composed of the weighted sum of three parts, specifically including:

[0020] Divide the historical matching accuracy of the parameter node by the preset highest matching benchmark value, and then multiply it by the weight adjustment factor to obtain the historical matching ratio as the first part;

[0021] Divide the current control priority coefficient by the average priority coefficient, and then multiply it by the weight adjustment factor to get the current priority ratio as the second part;

[0022] Divide the parameter stability by the total monitoring range, and then multiply it by the weight adjustment factor to get the stability ratio as the third part;

[0023] Adding the first part, the second part and the third part to obtain the dynamic weight;

[0024] S130, constructing a topological structure of the fermentation monitoring model based on a database; in the topological structure:

[0025] Parameter nodes are connected by directed edges, which represent the direction of parameter influence and the degree of association;

[0026] Each parameter node is bound to the state information to form a complete node description including parameter type label, sensitivity value, control priority coefficient, parameter stability, historical matching accuracy and dynamic weight parameters.

[0027] Preferably, the degree of association in S130 is specifically:

[0028] S130.1. Based on the historical fermentation data collected in S110, calculate the parameter linkage frequency and linkage amplitude between adjacent parameter nodes;

[0029] S130.2. Calculate an association closeness value based on the parameter linkage frequency, the parameter linkage amplitude, and the node status information generated in S120. The association closeness value is calculated as follows:

[0030] Divide the parameter linkage frequency by the set maximum linkage frequency reference value to obtain the normalized linkage frequency ratio;

[0031] Divide the parameter linkage amplitude by the set maximum linkage amplitude reference value to obtain the normalized linkage amplitude ratio;

[0032] The average value of the sensitivity of the source node and the target node is taken as the contribution value of the sensitivity to the association closeness;

[0033] Multiply the above three parts of the results by the preset weight coefficients respectively, and satisfy the condition that the sum of the weight coefficients is 1. The correlation closeness value is the sum of the three parts;

[0034] S130.3. Assign labels to directed edges based on the association density value:

[0035] If the association density value is greater than or equal to the first critical value, a high association density label is assigned to the directed edge;

[0036] If the association density value is between the second critical value and the first critical value, a medium association density label is assigned to the directed edge;

[0037] If the association density value is less than the second critical value, a low association density label is assigned to the directed edge;

[0038] S130.4. Based on the dynamic weight adjustment threshold in S120, mark the core link:

[0039] Re-adjusting the first critical value according to the dynamic weight of the parameter node calculated in S120 so that the first critical value increases as the importance of the node increases;

[0040] Only when the association closeness value reaches or exceeds the adjusted first critical value, the association relationship is marked as a core link.

[0041] Preferably, the step S200 specifically includes the following sub-steps:

[0042] S210, using a breadth-first search algorithm to traverse the fermentation monitoring model and extract all complete parameter chains from initial fermentation to terminal ripening;

[0043] S220, calculating the parameter stability of each parameter chain;

[0044] S230, marking key parameters and abnormal parameters based on parameter stability; specifically including the following sub-steps:

[0045] Comparing the parameter stability value with a preset stability threshold interval, where the preset stability threshold interval includes a stability threshold and an abnormality threshold;

[0046] If the parameter stability value is greater than or equal to the stability threshold, the corresponding parameter chain is marked as a key parameter;

[0047] If the parameter stability value is less than the abnormal threshold, the corresponding parameter chain is marked as an abnormal parameter;

[0048] S240: Perform differential analysis on key parameters and abnormal parameters and identify sensitive nodes. This specifically includes the following sub-steps:

[0049] Extract the correlation edge with the highest correlation closeness value among the key parameters, mark it as the core link, and allocate additional monitoring frequency to the core link to ensure its accuracy;

[0050] For each abnormal parameter chain, the dynamic weights of all nodes in the parameter chain are extracted, and the node with the lowest dynamic weight is screened out. If the dynamic weights of multiple nodes are the same and all are the lowest value, their historical matching accuracy rates are further compared, and the node with the lowest matching rate is identified as a sensitive node.

[0051] The core link information of key parameters and the sensitive node information of abnormal parameters are associated and stored.

[0052] Preferably, the calculation of parameter stability in S220 includes the following sub-steps:

[0053] Extract the dynamic weights of all nodes in the parameter chain and calculate the geometric mean of the dynamic weights of the nodes;

[0054] Extract the labels of all associated edges in the parameter chain and calculate the arithmetic mean of the association closeness values;

[0055] Statistical parameter chain time span, where the time span is defined as the sum of monitoring time intervals between adjacent nodes in the parameter chain;

[0056] The parameter stability value is calculated using the following comprehensive evaluation logic:

[0057] Multiply the geometric mean of the node dynamic weights by the arithmetic mean of the association closeness values ​​to obtain a comprehensive evaluation value;

[0058] Adding the parameter chain time span to a preset zero-proof constant, where the zero-proof constant is a very small positive number to avoid calculation anomalies;

[0059] The comprehensive evaluation value is divided by the time adjustment value to obtain the parameter stability value; the higher the parameter stability value, the better the monitoring reliability and continuity of the parameter chain.

[0060] Preferably, the specific implementation of generating the initial adjustment strategy set in S300 includes:

[0061] S310: Generate a sensor calibration plan for abnormal parameters; specifically, the following sub-steps are included:

[0062] Extracting the abnormal parameters marked in S230 and S240 and their corresponding sensitive node information;

[0063] Traversing the adjacent nodes of the sensitive node and screening out adjacent nodes whose redundant monitoring capabilities meet preset conditions; the screening process is specifically that the current monitoring load of the adjacent node is lower than the preset load threshold and its sensitivity is greater than the current demand;

[0064] Calculate the monitoring deviation of the sensitive node, that is, the difference between the current monitoring value and the standard value;

[0065] Based on the redundant monitoring capabilities and correlation closeness values ​​of adjacent nodes, monitoring requirements are allocated according to the following rules:

[0066] Prioritize adjacent nodes with high association closeness values;

[0067] The allocated amount does not exceed a set percentage threshold of the redundant monitoring capacity of adjacent nodes;

[0068] generating a sensor calibration instruction, and adding the sensor calibration instruction to an initial adjustment strategy set;

[0069] S320: Generate an environment control plan for the core link; specifically, the plan includes the following sub-steps:

[0070] Extracting the key parameters and core link information marked in S230;

[0071] Obtain the current environmental parameter values ​​and external interference data of the core link;

[0072] If the current environmental parameter value exceeds the preset parameter threshold or the external interference intensity is higher than the preset interference threshold, perform the following adjustments:

[0073] Execute parameter optimization strategy, specifically:

[0074] Through the topological structure of S130, a link in the replacement parameter chain is searched for a link whose correlation closeness value is not lower than that of the original parameter chain and whose time span is shorter; if no replacement parameter chain is found, the original parameter chain is retained, but an early warning signal is triggered and a control recommendation report is generated, wherein the control recommendation report includes adding monitoring equipment, temporarily activating backup sensors, or adjusting the sampling period;

[0075] Implement interference avoidance strategies, specifically:

[0076] Calculate the interference coefficient of the node in the parameter chain based on the historical interference times and total monitoring times recorded in the raw material batch records and fermentation tank data collected by S110;

[0077] Nodes with interference coefficients lower than a preset threshold and adjacent monitoring periods are selected as backup nodes;

[0078] Replace the high-interference node with a backup node and recalculate the correlation density value and time span of the parameter chain after replacement;

[0079] An optimized environment control instruction is generated, and the environment control instruction is added to the initial adjustment strategy set.

[0080] Preferably, the control priority in S120 is dynamically assigned according to the fermentation stage requirements, raw material characteristics and process standards, and the corresponding control priority coefficient is obtained as follows:

[0081] The fermentation stage is divided into three stages: early, middle and late, and the stage demand weights are quantified based on process standards;

[0082] Classify raw material batches into superior and ordinary categories, and assign weights to raw material characteristics based on raw material freshness, starch content, and impurity ratio;

[0083] The control priority coefficient is calculated using the following priority coefficient generation rules:

[0084] Assign basic weights to fermentation stage requirements and raw material characteristics respectively;

[0085] The priority coefficient is generated by weighting and summing the basic weight and the process violation cost ratio;

[0086] The priority coefficient is adjusted in conjunction with the node dynamic weight to ensure that high-priority parameter nodes are preferentially matched to high-priority control requirements;

[0087] The specific implementation of generating the initial adjustment strategy set in S300 further includes:

[0088] S330: Generate a priority sorting scheme for regulating priority conflicts; specifically, the scheme includes the following sub-steps:

[0089] Receive the fermentation data updated by S110 in real time and analyze the fermentation stage requirements and raw material characteristics;

[0090] Recalculate the control priority coefficient of the conflicting node based on the priority coefficient generation rule defined in S120;

[0091] If multiple control requirements conflict on the same parameter node, the following rules apply:

[0092] Priority is given to ensuring the regulation of higher demand classification during the fermentation stage;

[0093] If the stage demand classification is the same, the priority is determined according to the raw material characteristic weight;

[0094] If the stage demand classification and the raw material characteristic are the same, the process violation cost proportion is sorted from high to low;

[0095] Then, according to the updated priority coefficient, the regulation queue of the parameter node is dynamically adjusted:

[0096] The regulation demand with the highest priority coefficient is placed at the top for processing, and a reserved monitoring resource is allocated thereto;

[0097] The regulation demand with the second highest priority coefficient generates a delay processing suggestion, including the expected processing time and the recommended alternative parameter node;

[0098] Finally, the adjusted priority instruction is added to the initial adjustment strategy set.

[0099] Preferably, the adjustment operation type in the S400 includes parameter node adjustment, associated relationship correction, and weight dynamic update;

[0100] The S400 generates a target adjustment strategy set, specifically including the following steps:

[0101] S410, if the adjustment operation type is parameter node adjustment, the dynamic weight of the affected node is recalculated, and the sensor calibration scheme is updated;

[0102] S420, if the adjustment operation type is associated relationship correction, the associated closeness value is adjusted and the core link is re-evaluated to generate a new environment regulation scheme;

[0103] S430, if the adjustment operation type is weight dynamic update, the original dynamic weight is covered based on the direct weight value input by the user, and the priority sorting scheme is simultaneously corrected.

[0104] Preferably, the S500 specifically includes the following sub-steps:

[0105] S510, execute the initial adjustment strategy set after dynamic correction, including the sensor calibration scheme, the environment regulation scheme, and the priority sorting scheme;

[0106] S520, real-time collection of feedback data in the execution process, the feedback data including sensor calibration completion rate, actual regulation value of environmental parameters, regulation demand processing time limit, and node monitoring load change value;

[0107] S530, adjust the dynamic weight of the parameter node in the fermentation monitoring model;

[0108] S540, updating the topology structure and the correlation closeness value of the fermentation monitoring model based on the updated dynamic weight;

[0109] The adjusting the dynamic weight of the parameter node in the fermentation monitoring model comprises:

[0110] Based on the feedback data collected in S520, the historical matching accuracy, the current control priority coefficient and the parameter stability of the parameter node are recalculated, and the dynamic weight of each parameter node is updated:

[0111] If the sensor calibration completion rate is lower than the set threshold, the value of the weight adjustment factor is increased, and the influence of the historical matching accuracy on the dynamic weight is strengthened;

[0112] If the environmental parameter control value deviation exceeds the tolerance range, the value of the weight adjustment factor is increased, and the weight proportion of the current control priority coefficient is improved;

[0113] If the monitoring load change value does not reach the target, the value of the weight adjustment factor is increased, and the contribution of the parameter stability in the dynamic weight is enhanced.

[0114] Preferably, the specific implementation of collecting the feedback data in the execution process in S520 comprises:

[0115] Triggering a data collection instruction at a preset time interval, and the time interval is dynamically adjusted according to the fermentation stage requirement;

[0116] For the sensor calibration scheme, the working state of the calibration equipment, the deviation of the monitoring value before and after calibration, and the time consumption of the calibration operation are collected;

[0117] For the environmental control scheme, the running parameters of the control equipment, the real-time fluctuation curve of the environmental parameters and the control response delay time are collected;

[0118] For the priority sorting scheme, the actual processing order of the control requirement, the waiting time of the delayed processing requirement and the monitoring data of the alternative parameter node are collected.

[0119] Compared with the prior art, the beneficial effects of the present application are:

[0120] The high meter wine fermentation process real-time monitoring method based on intelligent sensors can comprehensively and dynamically reflect various states in the fermentation process by constructing a fermentation monitoring model and continuously updating based on real-time sensor data. The model includes multiple parameter nodes, the correlation between the parameter nodes and the state information of each parameter node, combined with real-time sensor data such as temperature fluctuation value, humidity change rate, pH value offset and alcohol content growth rate, which makes the description of the fermentation process more detailed and comprehensive, breaking the limitations of isolated and static data in traditional monitoring.

[0121] The application of the data fusion algorithm realizes accurate identification of key parameters and abnormal parameters in the fermentation monitoring model and effective confirmation of sensitive nodes. This process no longer relies on simple judgment of a single parameter, but through correlation analysis of multiple parameters, the key parameters that play a leading role in the fermentation process are mined, abnormal parameters that may affect the fermentation quality are discovered in time, and sensitive nodes that are easily disturbed by external factors are locked, so that the operator can clearly grasp the core link and potential risk point of the fermentation process, and change the previous situation of passive response to fermentation abnormalities.

[0122] The initial adjustment strategy set generated based on the key parameters and sensitive nodes covers sensor calibration schemes, environment control schemes and priority sorting schemes, providing diversified choices for the adjustment of the fermentation process. The sensor calibration scheme can solve the problem of sensor data drift in long-term use, the environment control scheme can improve the external conditions affecting fermentation, and the priority sorting scheme can clarify the order of operation when multiple adjustment requirements exist, avoiding the blindness and chaos of adjustment operation.

[0123] Receiving user interaction instructions and dynamically modifying the initial adjustment strategy set according to the instructions enhances the flexibility and practicality of the method. Different production scenarios and different fermentation stages may require different adjustment methods, and users can issue interaction instructions based on actual experience and specific circumstances to make the adjustment strategy more suitable for actual needs, overcoming the lack of adaptability of fixed adjustment mode in complex production environment.

[0124] Executing the dynamically modified adjustment strategy set and updating the fermentation monitoring model based on the execution result feedback forms a closed-loop monitoring and control system. The result of each adjustment operation is fed back to the model, enabling the model to continuously learn and optimize, improving the prediction and judgment ability of the fermentation process. With the continuous updating of the model, the identification accuracy of key parameters, abnormal parameters and sensitive nodes will be higher and higher, and the generated adjustment strategy will be more and more accurate, thus promoting the rice wine fermentation process to a more stable and efficient direction, reducing product quality fluctuations and resource waste caused by fermentation abnormalities. BRIEF DESCRIPTION OF DRAWINGS

[0125] Figure 1 The working principle diagram of the intelligent sensor-based high-quality rice wine fermentation process real-time monitoring method described in the present application;

[0126] Figure 2 The flowchart for constructing the fermentation monitoring model;

[0127] Figure 3 The flowchart for identifying key parameters, abnormal parameters and confirming sensitive nodes;

[0128] Figure 4Flowchart for execution result feedback and model update. DETAILED DESCRIPTION

[0129] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0130] See also Figures 1-4 The present invention provides a real-time monitoring method for the fermentation process of high-proof rice wine based on an intelligent sensor, and the specific steps are as follows:

[0131] A fermentation monitoring model is constructed and updated based on real-time sensor data; wherein the fermentation monitoring model includes multiple parameter nodes, the relationship between the parameter nodes, and the status information of each parameter node, and the real-time sensor data includes temperature fluctuation value, humidity change rate, pH value offset and alcohol content growth rate.

[0132] The data fusion algorithm is used to identify key parameters and abnormal parameters in the fermentation monitoring model and confirm sensitive nodes.

[0133] An initial adjustment strategy set is generated based on the key parameters and sensitive nodes; the initial adjustment strategy set includes a sensor calibration scheme, an environment control scheme, and a priority sorting scheme.

[0134] A user interaction instruction is received, and an adjustment operation type is determined according to the user interaction instruction, and the initial adjustment strategy set is dynamically modified based on the adjustment operation type.

[0135] The dynamically revised initial adjustment strategy set is executed, and the fermentation monitoring model is updated based on the execution result feedback.

[0136] Example 1: In the real-time monitoring method for the fermentation process of high-proof rice wine based on intelligent sensors, building a fermentation monitoring model and updating the model based on real-time sensor data are important basic steps. When building the fermentation monitoring model, the first step is to collect historical fermentation data. This historical fermentation data covers multiple aspects, including raw material batch records, which record in detail the relevant information of different batches of raw materials, such as the type, source, batch number, etc. of the raw materials; the fermentation tank temperature and humidity curve, which records the changes in temperature and humidity in the fermentation tank over time during the fermentation process; the pH value change log, which clearly presents the dynamic changes in pH value during the fermentation process; and the alcohol content test report, which contains the test results of alcohol content at different stages.

[0137] After collecting historical fermentation data, it is necessary to perform multi-dimensional feature extraction on these data to generate the status information of the parameter nodes. The status information of the parameter nodes contains multiple key parts, including parameter type, which is used to distinguish different types of parameters; sensitivity, which reflects the sensitivity of the parameter to the fermentation process; control priority, which reflects the importance of the parameter in the control; parameter stability, which indicates the stability of the parameter; historical matching accuracy, which is the degree of match between the historical monitoring data of the parameter and the actual situation; and dynamic weight. Among them, the determination of the control priority needs to be dynamically assigned according to the requirements of the fermentation stage, the characteristics of the raw materials and the process standards, and then the corresponding control priority coefficient is obtained. The fermentation stage may be divided into different periods, and each period has different requirements for various parameters. The characteristics of the raw materials also vary depending on the raw materials. The process standards stipulate the corresponding parameter ranges and requirements. These factors are combined to determine the control priority coefficient.

[0138] The calculation of dynamic weight consists of a weighted sum of three parts. The first part is to divide the parameter node's historical matching accuracy by the preset highest matching benchmark value, then multiply it by the weight adjustment factor to obtain the historical matching percentage. The preset highest matching benchmark value is a standard value set based on historical data and experience. This calculation reflects the impact of historical matching on the dynamic weight. The second part is to divide the current control priority coefficient by the average priority coefficient, then multiply it by the weight adjustment factor to obtain the current priority percentage. The average priority coefficient is the average of the control priority coefficients of all parameter nodes, which reflects the relative importance of the current control priority within the overall system. The third part is to divide the parameter stability by the total monitoring range, then multiply it by the weight adjustment factor to obtain the stability percentage. The total monitoring range is the range of possible parameter monitoring values. This calculation measures the contribution of parameter stability to the dynamic weight. Finally, these three parts are added together to obtain the dynamic weight.

[0139] After processing and extracting features from historical fermentation data, the topological structure of the fermentation monitoring model is constructed based on the database. Within this topological structure, parameter nodes are connected by directed edges that represent the direction of influence and the degree of correlation between parameters. Each parameter node is bound to previously generated state information, forming a complete node description that includes parameter type labels, sensitivity values, control priority coefficients, parameter stability, historical matching accuracy, and dynamic weight parameters. This approach allows the constructed fermentation monitoring model to comprehensively reflect the various parameters and their interrelationships during the fermentation of high-proof rice wine, providing a foundation for subsequent model updates based on real-time sensor data, as well as for real-time monitoring and adjustment of the entire fermentation process.

[0140] In actual applications, real-time sensors will collect data such as temperature fluctuations, humidity change rates, pH offsets, and alcohol growth rates. These real-time data will be used to update the constructed fermentation monitoring model so that the model can reflect the actual situation of the fermentation process in real time, thereby more accurately monitoring and analyzing the fermentation process.

[0141] Example 2: In a real-time, intelligent sensor-based method for monitoring the fermentation process of high-proof rice wine, determining the correlation between parameter nodes is a key step in constructing the topological structure of the fermentation monitoring model. This correlation requires a series of statistical and computational analyses based on historical fermentation data to accurately reflect the degree of mutual influence and correlation strength between parameters.

[0142] Based on previously collected historical fermentation data, the frequency and magnitude of parameter linkage between adjacent parameter nodes are statistically analyzed. This historical fermentation data includes real-time monitoring data for each parameter during the fermentation process. By analyzing this data, we can determine the linkage between parameter changes between adjacent parameter nodes within a certain timeframe. For example, when the temperature parameter changes, whether the adjacent humidity parameter also changes accordingly, and the number of such changes is the linkage frequency, while the magnitude of the humidity parameter change is the linkage magnitude.

[0143] After obtaining the parameter linkage frequency and linkage amplitude, the correlation density value is calculated based on this data and the previously generated node status information. The first step is to divide the parameter linkage frequency by the set maximum linkage frequency reference value to obtain a normalized linkage frequency ratio. The maximum linkage frequency reference value set here is determined based on the maximum value of the parameter linkage frequency in the historical data. Through this normalization process, linkage frequencies of different ranges can be converted into comparable ratio values.

[0144] The second step is to divide the parameter linkage amplitude by the set maximum linkage amplitude benchmark value to obtain the normalized linkage amplitude ratio. Again, the maximum linkage amplitude benchmark value is set based on the maximum parameter linkage amplitude in the historical data. Normalization makes the linkage amplitude comparable across different parameters.

[0145] The third step is to take the average of the sensitivity of the source and target nodes and use this as the contribution to the closeness of association. The node sensitivity is determined in the previously generated state information and reflects the sensitivity of the node parameters to the fermentation process. By taking the average of the sensitivity of the source and target nodes, we can comprehensively consider the impact of the sensitivity characteristics of both nodes on the closeness of association.

[0146] After completing the calculations for the three components above, each result is multiplied by a preset weighting factor, and the sum of these weighting factors must be 1. This weighted calculation ultimately yields the correlation density value. The preset weighting factors are pre-set based on the characteristics and experience of the fermentation process, and are used to reflect the different weights of linkage frequency, linkage amplitude, and sensitivity in the correlation density calculation.

[0147] After obtaining the association density value, it is necessary to assign labels to the directed edges based on this value. The specific assignment rules are as follows: if the association density value is greater than or equal to the first critical value, the directed edge is assigned a high association density label; if the association density value is between the second critical value and the first critical value, the directed edge is assigned a medium association density label; if the association density value is less than the second critical value, the directed edge is assigned a low association density label. The first and second critical values ​​are thresholds set based on the actual conditions and requirements of the fermentation process and are used to categorize association density.

[0148] The critical value is adjusted based on the previously calculated dynamic weight of the parameter node, and the core link is marked. The specific operation is to readjust the first critical value based on the dynamic weight of the parameter node, so that the first critical value increases as the importance of the node increases. The higher the dynamic weight of the node, the more important the node is in the fermentation process. Correspondingly, the standard for its association closeness to become a core link will also increase. Only when the association closeness value reaches or exceeds the adjusted first critical value will the association relationship be marked as a core link.

[0149] By determining the degree of correlation and marking core links through this series of steps, the topology of the fermentation monitoring model can accurately reflect the degree of correlation and importance between parameter nodes. Directed edges with high correlation indicate strong mutual influence between parameters, while core links represent parameter relationships that require special attention during the fermentation process. In actual fermentation monitoring, this information can help the monitoring system more specifically monitor and analyze key parameters and core links, thereby better understanding the dynamic changes in the fermentation process. For example, for parameter nodes connected by directed edges marked with high correlation, the monitoring system can appropriately increase the monitoring frequency to obtain more timely information on changes in these parameters. Core links, on the other hand, require greater attention, as their changes may have a significant impact on the entire fermentation process.

[0150] By fully utilizing historical data and using scientific calculation methods, we ensure that the determination of the closeness of association can accurately reflect the actual correlation between parameters during the fermentation process, thus guaranteeing the accuracy and reliability of the fermentation monitoring model. At the same time, by adjusting the critical value based on the dynamic weight of the node to mark the core link, the model can more flexibly adapt to the changes in the importance of parameters in different fermentation stages and different raw material characteristics, thereby improving the adaptability and practicality of the model. In the subsequent fermentation monitoring process, with the continuous input of real-time data and the update of the model, the closeness of association value and the marking of the core link will also change accordingly, so as to continuously and accurately reflect the correlation between the parameters in the fermentation process, providing a more accurate basis for the optimization and regulation of the fermentation process.

[0151] Example 3: Identifying key parameters and abnormal parameters in the fermentation monitoring model and confirming sensitive nodes are important steps to ensure monitoring accuracy. This process uses multi-step data processing and logical analysis to achieve a systematic evaluation of the parameter chain. The specific implementation method is as follows:

[0152] A breadth-first search algorithm is used to traverse the fermentation monitoring model, extracting the complete chain of parameters from initial fermentation to final maturity. The fermentation monitoring model stores parameter nodes and their relationships in a topological structure. Starting from the initial node, the breadth-first search algorithm traverses adjacent nodes layer by layer, ensuring that all possible parameter transmission paths are covered, thereby obtaining a complete chain of parameter evolution, including temperature, humidity, pH value, alcohol content, and so on.

[0153] Calculating the parameter stability of each parameter chain involves comprehensive processing of multi-dimensional data. The specific steps are: first, extract the dynamic weights of all nodes in the parameter chain and calculate the geometric mean of these dynamic weights; then, extract the labels of all associated edges in the parameter chain and calculate the arithmetic mean of the association closeness values; and simultaneously calculate the time span of the parameter chain, which is defined as the sum of the monitoring time intervals between adjacent nodes in the parameter chain.

[0154] The parameter stability value is calculated using the following comprehensive evaluation logic: the geometric mean of the node dynamic weight is multiplied by the arithmetic mean of the association closeness value to obtain a comprehensive evaluation value; then the parameter chain time span is added to the preset anti-zero constant, which is a very small positive number (such as 10^-6) to avoid calculation anomalies when the time span is zero; finally, the comprehensive evaluation value is divided by the time adjustment value to obtain the parameter stability value. It can be expressed as:

[0155] Among them, w i represents the dynamic weight of the i-th node in the parameter chain, n is the number of nodes in the parameter chain; c jThe association density value of the jth association edge in the parameter chain, m is the number of association edges in the parameter chain, t is the time span of the parameter chain, and k is a preset anti-zero constant. The higher the parameter stability value, the better the monitoring reliability and continuity of the parameter chain.

[0156] Based on the parameter stability value, the key parameters and abnormal parameters are marked. The parameter stability value is compared with a preset stability threshold interval, which includes a stability threshold and an abnormal threshold. If the parameter stability value is greater than or equal to the stability threshold, the corresponding parameter chain is marked as a key parameter, which shows higher stability and reliability in the fermentation process and has a more significant impact on the fermentation process; if the parameter stability value is less than the abnormal threshold, the corresponding parameter chain is marked as an abnormal parameter, indicating that the monitoring data of the parameter chain may fluctuate or be abnormal, which needs to be focused on.

[0157] After completing the parameter marking, differential analysis is performed on the key parameters and abnormal parameters, and sensitive nodes are confirmed. For key parameters, the association edge with the highest association density value is extracted and marked as a core link. The core link represents the strongest association relationship between parameters, and additional monitoring frequency is allocated to the core link to ensure its accuracy, such as shortening the data collection time by 50% based on the regular monitoring interval.

[0158] For each abnormal parameter chain, the dynamic weight of all nodes in the parameter chain is extracted, and the node with the lowest dynamic weight is selected. If the dynamic weights of multiple nodes are the same and are the lowest, further comparison of their historical matching accuracy is performed, and the node with the lowest matching rate is confirmed as a sensitive node. The sensitive node is a key node that causes the stability of the parameter chain to be insufficient, and its monitoring process needs to be focused on or adjusted.

[0159] The core link information of the key parameters and the sensitive node information of the abnormal parameters are associated and stored. The storage content includes the parameter node identification, association density value, and additional monitoring frequency setting of the core link, as well as the dynamic weight, historical matching accuracy, and abnormal parameter chain number of the sensitive node, forming a structured database record to facilitate the generation of subsequent adjustment strategies and model updates.

[0160] In practical applications, this embodiment realizes intelligent screening and analysis of fermentation parameters through systematic algorithm logic and data processing flow. For example, when the time span of a parameter chain is large and the association density value is low, the parameter stability value will decrease accordingly, and it may be marked as an abnormal parameter chain, triggering the investigation of the sensitive node therein.

[0161] Embodiment 4: Generating the initial adjustment strategy set is a key step for intervention and optimization of the fermentation process, which contains sensor calibration scheme, environmental regulation scheme and priority ranking scheme. The implementation thereof will be described in detail below in combination with specific examples.

[0162] Suppose that in the fermentation process of a batch of high-grade rice wine, an abnormal parameter chain is marked by the method of embodiment 3, which involves the correlation between the temperature parameter and the alcohol content parameter in the fermentation tank. Further analysis shows that the dynamic weight of the temperature parameter node is the lowest and the historical matching accuracy is relatively low, which is confirmed as a sensitive node. At this time, a sensor calibration scheme needs to be generated for the abnormal parameter:

[0163] Extract the marked abnormal parameter and the corresponding sensitive node information, i.e. the temperature parameter node and its position in the topology. Then traverse the adjacent nodes of the sensitive node, including the humidity sensor node and the pH sensor node in the same fermentation tank. Screen the adjacent nodes whose redundant monitoring capabilities meet the preset conditions, for example, set the current monitoring load of the adjacent node to be lower than the preset load threshold of 70%, and its sensitivity degree needs to be greater than the current temperature monitoring requirement. Suppose that the current monitoring load of the humidity sensor node is 50% and the sensitivity score is 8 (full score is 10), which meets the screening conditions; while the monitoring load of the pH sensor node is 80%, which does not meet the conditions.

[0164] Calculate the monitoring deviation of the sensitive node, i.e. the current temperature sensor's monitoring value is 32℃, while according to the process standard, the standard temperature of this fermentation stage should be 30℃, the deviation is +2℃. Based on the redundant monitoring capability and the correlation density value of the adjacent nodes, the monitoring requirement is allocated: the correlation density value between the humidity sensor node and the temperature sensor node is 0.7 (full value is 1), which belongs to high correlation. According to the rule, the adjacent node with high correlation density value, i.e. the humidity sensor node, is preferentially selected, and the allocated amount is set to not more than 60% of its redundant monitoring capability. The redundant monitoring capability of the humidity sensor node is to collect 3 additional data every 10 minutes, so it is allocated to assist in collecting 2 temperature data every 10 minutes. Finally, generate the sensor calibration instruction, which includes triggering the additional temperature collection program of the humidity sensor node, setting the collection frequency and synchronizing the calibrated temperature data to the monitoring system, and add the instruction to the initial adjustment strategy set.

[0165] Take the correlation between the temperature and humidity parameters of the fermentation tank and the alcohol content growth rate as an example. Suppose that in the middle of the fermentation, the current environmental parameter values of the core link are temperature 35℃ and humidity 85%, while the preset parameter threshold is temperature ≤33℃ and humidity ≤80%. The current environmental parameter values exceed the preset parameter threshold. At this time, the environmental regulation scheme is executed:

[0166] Extract the marked key parameters and core link information to clarify the strong correlation between temperature and humidity parameters and alcohol content parameters. Execute the parameter optimization strategy and search for alternative parameter chains through the topological structure, looking for links with a correlation density value no less than the original parameter chain (the original correlation density value is 0.8) and a shorter time span. Assume that an alternative parameter chain consisting of temperature parameters, pH value parameters, and alcohol content parameters is found, with a correlation density value of 0.82 and a time span of 80% of the original parameter chain, meeting the conditions. Switch to monitoring on this alternative parameter chain and adjust the monitoring frequency to collect data every 5 minutes.

[0167] If no suitable alternative parameter chain can be found—for example, if the correlation density of all possible alternative parameter chains is less than 0.7 during a particular fermentation phase—the original parameter chain is retained, triggering an early warning signal and generating a control recommendation report. This report may include recommendations such as increasing the number of temperature sensors in the fermenter to improve monitoring density, temporarily activating a backup humidity sensor, or adjusting the sampling period to the cooler early morning hours.

[0168] When implementing the interference avoidance strategy, assume that a humidity sensor node in a core link has a history of 20 interference events and a total of 100 monitoring events, resulting in a calculated interference coefficient of 0.2. The preset interference threshold is 0.15, and the interference coefficient of this node exceeds the threshold. Based on raw material batch records and fermentation tank data, nodes with interference coefficients below 0.15 and adjacent monitoring periods are selected, such as pH sensor nodes (interference coefficient 0.12), as backup nodes. The high-interference humidity sensor node is replaced with a backup pH sensor node, and the correlation strength and time span of the parameter chain after the replacement are recalculated. Assuming the correlation strength after the replacement is 0.75 and the time span is increased by 10%, an optimized environmental control instruction is generated, including activating the pH sensor node's auxiliary humidity monitoring function and adjusting the data fusion algorithm to adapt to the new parameter chain. This instruction is then added to the initial adjustment strategy set.

[0169] Consider the scenario where conflicts in control priorities arise during the fermentation process. For example, in the late fermentation stage, a parameter node receives demands from both temperature control and alcohol content control. Receive updated fermentation data in real time, and parse out that the fermentation stage demand is the late maturation stage, and the raw material characteristics are high-quality glutinous rice with a high starch content. Based on the control priority coefficient generation rules, recalculate the control priority coefficients of the conflicting nodes: the fermentation stage is divided into early, middle, and late stages, and the weight of the late stage demand is quantified to 0.4; the raw material batches are divided into superior and ordinary, and the weight of superior raw material characteristics is 0.3. Assign basic weights to the fermentation stage demand and raw material characteristics respectively. Assume that the basic weight of the fermentation stage is 0.5, the basic weight of the raw material characteristics is 0.3, and the process violation cost ratio is 0.2. The priority coefficient is calculated by weighting the basic weight and the process violation cost ratio. For example, the priority coefficient for temperature control is 0.5 × 0.4 + 0.3 × 0.3 + 0.2 × 0.5 = 0.2 + 0.09 + 0.1 = 0.39, and the priority coefficient for alcohol control is 0.5 × 0.6 (the demand for alcohol control is higher in the later stages) + 0.3 × 0.3 + 0.2 × 0.6 = 0.3 + 0.09 + 0.12 = 0.51. Because alcohol control has a higher priority coefficient, its control needs are prioritized. If multiple control requirements conflict at the same parameter node and the stage requirements share the same classification, for example, both are mid-fermentation, the priority is determined by the weight of the raw material characteristics. If the stage requirement classification and raw material characteristics are the same, they are ranked from highest to lowest by the process violation cost ratio.

[0170] Based on the updated priority coefficients, the control queue order of parameter nodes is dynamically adjusted: the alcohol control request with the highest priority coefficient is placed at the top, and reserved monitoring resources, such as dedicated data transmission channels and computing resources, are allocated to it. For the temperature control request with the second highest priority coefficient, a delay processing suggestion is generated, including an estimated processing time (such as 30 minutes after the alcohol control is completed) and a recommended alternative parameter node (such as an adjacent pressure sensor node that can temporarily assist in temperature monitoring). Finally, the adjusted priority instructions are added to the initial adjustment strategy set, including the execution order of the control requests, the resource allocation plan, and the specific time points for delay processing.

[0171] In practical applications, the generation of the initial set of adjustment strategies is closely dependent on the parameter evaluation results of the fermentation monitoring model. For example, when the parameter stability value of a parameter chain falls below the abnormal threshold, the sensor calibration scheme is triggered, leveraging the redundant capabilities of adjacent nodes to achieve monitoring remediation. When the core link is disturbed by the environment, the environmental control scheme maintains monitoring accuracy by switching parameter chains or replacing interfering nodes. Furthermore, the prioritization scheme ensures that, given limited resources, control requirements with greater impact on the fermentation process are prioritized.

[0172] Example 5: Receiving user interaction instructions and dynamically revising the initial adjustment strategy set, as well as executing the revised strategy and updating the fermentation monitoring model based on feedback, are key steps in achieving the flexibility and adaptability of the monitoring system. Specific implementation methods are as follows:

[0173] When the system receives a user interaction instruction, it first determines the type of adjustment operation based on the instruction content. Adjustment operation types include parameter node adjustment, relationship modification, and dynamic weight update. Different adjustment operation types correspond to different correction logic to achieve precise optimization of the initial adjustment strategy set.

[0174] If the adjustment operation type is parameter node adjustment, for example, the user finds that a temperature sensor node needs to be temporarily removed due to a hardware failure, the system will recalculate the dynamic weights of the affected nodes. The affected nodes include the humidity parameter node and the alcohol parameter node directly connected to the temperature sensor node. When recalculating the dynamic weight, factors such as historical matching accuracy, current control priority coefficient, and parameter stability will be re-evaluated. For example, the historical matching accuracy of the humidity parameter node may be reduced due to the removal of the temperature node, resulting in a decrease in its dynamic weight. Based on the new dynamic weight, the system will update the sensor calibration scheme, such as adjusting the auxiliary monitoring frequency of the temperature parameter by the adjacent pH sensor node to ensure the continuity of the monitoring data.

[0175] If the adjustment operation type is association modification, for example, based on actual fermentation experience, the user believes that the correlation between the pressure parameter and the alcohol content parameter in the fermentation tank needs to be adjusted, the system will adjust the correlation between the two parameter nodes. After the adjustment, the system will re-evaluate the core link. If the association is part of the core link, the re-evaluation may cause the composition of the core link to change. For example, the original core link is temperature-humidity-alcohol content. When the correlation between pressure and alcohol content increases, the core link may be updated to temperature-pressure-alcohol content. Based on the new core link, the system will generate a new environmental control plan, such as assigning a higher monitoring frequency to the pressure parameter node to ensure the accuracy of the new core link.

[0176] If the adjustment operation type is dynamic weight update, the user may directly enter a dynamic weight value for a parameter node based on the needs of a specific fermentation process, such as adjusting the dynamic weight of the alcohol parameter node from 0.3 to 0.5. The system will overwrite the original dynamic weight based on the direct weight value entered by the user and simultaneously adjust the priority sorting scheme. For example, if the dynamic weight of the alcohol parameter node is increased, the corresponding control requirement will be moved up in the priority sorting and will receive priority in resource allocation.

[0177] After dynamically revising the initial set of adjustment strategies, the system executes the revised strategy set, including the sensor calibration plan, environmental control plan, and priority sorting plan. During execution, the system collects real-time feedback data to evaluate the effectiveness of the strategy and update the fermentation monitoring model.

[0178] When collecting feedback data in real time, data collection instructions are triggered at preset intervals, which are dynamically adjusted based on the needs of the fermentation stage. For example, in the early stages of fermentation, the interval is set to 10 minutes; in the middle stages of fermentation, the interval is shortened to 5 minutes to obtain more intensive data. For sensor calibration solutions, the operating status of the calibration equipment is collected, such as whether the calibration was successful and whether any abnormalities occurred during the calibration process; the deviation of the monitoring values ​​before and after calibration is collected, such as the temperature sensor's deviation before calibration was +2°C and after calibration was +0.5°C; and the calibration operation duration, such as this calibration took 3 minutes. For environmental control solutions, the operating parameters of the control equipment are collected, such as the cooling power of the air conditioner; the real-time fluctuation curve of the environmental parameters is collected, such as the temperature change from 35°C to 33°C within 30 minutes after the adjustment; and the control response delay time, such as the time from issuing the control instruction to the device starting operation is 2 minutes. For the priority sorting scheme, collect the actual processing order of the control requirements, such as whether the alcohol control requirement is executed before the temperature control requirement; collect the waiting time of delayed processing requirements, such as the temperature control requirement waited for 25 minutes before starting to be executed; and the monitoring data of the alternative parameter nodes, such as the measurement value of the pressure sensor node when assisting in monitoring the temperature.

[0179] Based on collected feedback data, the system adjusts the dynamic weights of parameter nodes in the fermentation monitoring model. It recalculates the parameter node's historical matching accuracy, current control priority coefficient, and parameter stability, and updates the dynamic weights of each parameter node. If the sensor calibration completion rate falls below the set threshold (e.g., the set threshold is 90% and the actual completion rate is 85%), the weight adjustment factor is increased to strengthen the influence of historical matching accuracy on the dynamic weight, thereby increasing the weight of historical matching accuracy in subsequent dynamic weight calculations. If the controlled value of an environmental parameter deviates beyond the tolerance range (e.g., the temperature control target is 30°C and the actual control value is 32°C, exceeding the 2°C tolerance), the weight adjustment factor is increased to increase the weight of the current control priority coefficient, giving parameter nodes with higher control priorities a higher priority in the dynamic weight calculation. If the monitored load change value falls short of the target (e.g., the target load change value is a 20% decrease and the actual decrease is only 15%), the weight adjustment factor is increased to strengthen the contribution of parameter stability to the dynamic weight, giving nodes with higher parameter stability a higher weight in the dynamic weight calculation.

[0180] Finally, based on the updated dynamic weights, the system updates the topology structure and the correlation density values of the fermentation monitoring model. For example, after the dynamic weight of a certain parameter node is increased, the correlation density values between the parameter node and its adjacent nodes may be adjusted accordingly, and the display priority of the parameter node in the topology structure is also increased, so that the monitoring personnel can pay more attention to the key parameter nodes more intuitively.

[0181] In practical applications, this process realizes the closed-loop optimization of the monitoring system. For example, after the user finds that the monitoring data of a certain parameter is abnormal and adjusts the parameter node, the system can quickly adapt to changes and re-optimize the monitoring scheme through dynamic correction strategies and real-time feedback. For another example, in a long-term fermentation process, through continuous accumulation of feedback data and updating of the model, the system's understanding of the fermentation process will become more and more in-depth, and the accuracy of monitoring and adjustment will gradually improve, thereby better guaranteeing the quality and stability of the fermentation of high-grade rice wine.

[0182] It should be noted that, in this document, the terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or device.

[0183] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring method for the fermentation process of high-proof rice wine based on an intelligent sensor, characterized in that: The following steps are involved: S100, constructing a fermentation monitoring model and updating the fermentation monitoring model based on real-time sensor data; wherein the fermentation monitoring model includes multiple parameter nodes, associations between parameter nodes, and status information of each parameter node, and the real-time sensor data includes temperature fluctuation value, humidity change rate, pH value offset, and alcohol content growth rate; S200, identifying key parameters and abnormal parameters in the fermentation monitoring model and confirming sensitive nodes through a data fusion algorithm; S300, generating an initial adjustment strategy set based on the key parameters and sensitive nodes; The initial adjustment strategy set includes a sensor calibration scheme, an environmental control scheme, and a priority sorting scheme; S400: Receive a user interaction instruction, determine an adjustment operation type according to the user interaction instruction, and dynamically modify the initial adjustment strategy set based on the adjustment operation type; S500: Execute the dynamically revised initial adjustment strategy set, and update the fermentation monitoring model based on the execution result feedback.

2. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 1, characterized in that: The specific implementation of constructing the fermentation monitoring model in S100 includes: S110, collecting historical fermentation data, wherein the historical fermentation data includes raw material batch records, fermentation tank temperature and humidity curves, pH value change logs, and alcohol content test reports; S120: Perform multi-dimensional feature extraction on the historical fermentation data to generate status information of parameter nodes; the status information includes parameter type, sensitivity, control priority, parameter stability, historical matching accuracy, and dynamic weight; wherein: The control priority is dynamically assigned according to the fermentation stage requirements, raw material characteristics and process standards to obtain the corresponding control priority coefficient; The dynamic weight is composed of the weighted sum of three parts, specifically including: Divide the historical matching accuracy of the parameter node by the preset highest matching benchmark value, and then multiply it by the weight adjustment factor to obtain the historical matching ratio as the first part; Divide the current control priority coefficient by the average priority coefficient, and then multiply it by the weight adjustment factor to get the current priority ratio as the second part; Divide the parameter stability by the total monitoring range, and then multiply it by the weight adjustment factor to get the stability ratio as the third part; Adding the first part, the second part and the third part to obtain the dynamic weight; S130, constructing a topological structure of the fermentation monitoring model based on a database; in the topological structure: Parameter nodes are connected by directed edges, which represent the direction of parameter influence and the degree of association; Each parameter node is bound to the state information to form a complete node description including parameter type label, sensitivity value, control priority coefficient, parameter stability, historical matching accuracy and dynamic weight parameters.

3. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 2, characterized in that: The correlation closeness in S130 is specifically: S130.

1. Based on the historical fermentation data collected in S110, calculate the parameter linkage frequency and linkage amplitude between adjacent parameter nodes; S130.

2. Calculate an association closeness value based on the parameter linkage frequency, the parameter linkage amplitude, and the node status information generated in S120. The association closeness value is calculated as follows: Divide the parameter linkage frequency by the set maximum linkage frequency reference value to obtain the normalized linkage frequency ratio; Divide the parameter linkage amplitude by the set maximum linkage amplitude reference value to obtain the normalized linkage amplitude ratio; The average value of the sensitivity of the source node and the target node is taken as the contribution value of the sensitivity to the association closeness; Multiply the above three parts of the results by the preset weight coefficients respectively, and satisfy the condition that the sum of the weight coefficients is 1. The correlation closeness value is the sum of the three parts; S130.

3. Assign labels to directed edges based on the association density value: If the association density value is greater than or equal to the first critical value, a high association density label is assigned to the directed edge; If the association density value is between the second critical value and the first critical value, a medium association density label is assigned to the directed edge; If the association density value is less than the second critical value, a low association density label is assigned to the directed edge; S130.

4. Based on the dynamic weight adjustment threshold in S120, mark the core link: Re-adjusting the first critical value according to the dynamic weight of the parameter node calculated in S120 so that the first critical value increases as the importance of the node increases; Only when the association closeness value reaches or exceeds the adjusted first critical value, the association relationship is marked as a core link.

4. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 2, characterized in that: The S200 specifically includes the following sub-steps: S210, using a breadth-first search algorithm to traverse the fermentation monitoring model and extract all complete parameter chains from initial fermentation to terminal ripening; S220, calculating the parameter stability of each parameter chain; S230, marking key parameters and abnormal parameters based on parameter stability; It includes the following sub-steps: Comparing the parameter stability value with a preset stability threshold interval, where the preset stability threshold interval includes a stability threshold and an abnormality threshold; If the parameter stability value is greater than or equal to the stability threshold, the corresponding parameter chain is marked as a key parameter; If the parameter stability value is less than the abnormal threshold, the corresponding parameter chain is marked as an abnormal parameter; S240: Perform differential analysis on key parameters and abnormal parameters, and identify sensitive nodes; It includes the following sub-steps: Extract the correlation edge with the highest correlation closeness value among the key parameters, mark it as the core link, and allocate additional monitoring frequency to the core link to ensure its accuracy; For each abnormal parameter chain, the dynamic weights of all nodes in the parameter chain are extracted, and the node with the lowest dynamic weight is screened out. If the dynamic weights of multiple nodes are the same and all are the lowest value, their historical matching accuracy rates are further compared, and the node with the lowest matching rate is identified as a sensitive node. The core link information of key parameters and the sensitive node information of abnormal parameters are associated and stored.

5. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 4, characterized in that: The calculation of parameter stability in S220 includes the following sub-steps: Extract the dynamic weights of all nodes in the parameter chain and calculate the geometric mean of the dynamic weights of the nodes; Extract the labels of all associated edges in the parameter chain and calculate the arithmetic mean of the association closeness values; Statistical parameter chain time span, where the time span is defined as the sum of monitoring time intervals between adjacent nodes in the parameter chain; The parameter stability value is calculated using the following comprehensive evaluation logic: Multiply the geometric mean of the node dynamic weights by the arithmetic mean of the association closeness values ​​to obtain a comprehensive evaluation value; Adding the parameter chain time span to a preset zero-proof constant, where the zero-proof constant is a very small positive number to avoid calculation anomalies; The comprehensive evaluation value is divided by the time adjustment value to obtain the parameter stability value; the higher the parameter stability value, the better the monitoring reliability and continuity of the parameter chain.

6. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 5, characterized in that: The specific implementation of generating the initial adjustment strategy set in S300 includes: S310: Generate a sensor calibration plan for abnormal parameters; specifically, the following sub-steps are included: Extracting the abnormal parameters marked in S230 and S240 and their corresponding sensitive node information; Traversing the adjacent nodes of the sensitive node and screening out adjacent nodes whose redundant monitoring capabilities meet preset conditions; the screening process is specifically that the current monitoring load of the adjacent node is lower than the preset load threshold and its sensitivity is greater than the current demand; Calculate the monitoring deviation of the sensitive node, that is, the difference between the current monitoring value and the standard value; Based on the redundant monitoring capabilities and correlation closeness values ​​of adjacent nodes, monitoring requirements are allocated according to the following rules: Prioritize adjacent nodes with high association closeness values; The allocated amount does not exceed a set percentage threshold of the redundant monitoring capacity of adjacent nodes; generating a sensor calibration instruction, and adding the sensor calibration instruction to an initial adjustment strategy set; S320: Generate an environment control plan for the core link; specifically, the plan includes the following sub-steps: Extracting the key parameters and core link information marked in S230; Obtain the current environmental parameter values ​​and external interference data of the core link; If the current environmental parameter value exceeds the preset parameter threshold or the external interference intensity is higher than the preset interference threshold, perform the following adjustments: Execute parameter optimization strategy, specifically: Through the topological structure of S130, a link in the replacement parameter chain is searched for a link whose correlation closeness value is not lower than that of the original parameter chain and whose time span is shorter; if no replacement parameter chain is found, the original parameter chain is retained, but an early warning signal is triggered and a control recommendation report is generated, wherein the control recommendation report includes adding monitoring equipment, temporarily activating backup sensors, or adjusting the sampling period; Implement interference avoidance strategies, specifically: Calculate the interference coefficient of the node in the parameter chain based on the historical interference times and total monitoring times recorded in the raw material batch records and fermentation tank data collected by S110; Nodes with interference coefficients lower than a preset threshold and adjacent monitoring periods are selected as backup nodes; Replace the high-interference node with a backup node and recalculate the correlation density value and time span of the parameter chain after replacement; An optimized environment control instruction is generated, and the environment control instruction is added to the initial adjustment strategy set.

7. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 6, characterized in that: In S120, the control priority is dynamically assigned according to the fermentation stage requirements, raw material characteristics and process standards, and the corresponding control priority coefficient is obtained as follows: The fermentation stage is divided into three stages: early, middle and late, and the stage demand weights are quantified based on process standards; Classify raw material batches into superior and ordinary categories, and assign weights to raw material characteristics based on raw material freshness, starch content, and impurity ratio; The control priority coefficient is calculated using the following priority coefficient generation rules: Assign basic weights to fermentation stage requirements and raw material characteristics respectively; The priority coefficient is generated by weighting and summing the basic weight and the process violation cost ratio; The priority coefficient is adjusted in conjunction with the node dynamic weight to ensure that high-priority parameter nodes are preferentially matched to high-priority control requirements; The specific implementation of generating the initial adjustment strategy set in S300 further includes: S330: Generate a priority sorting scheme for regulating priority conflicts; specifically, the scheme includes the following sub-steps: Receive the fermentation data updated by S110 in real time and analyze the fermentation stage requirements and raw material characteristics; Recalculate the control priority coefficient of the conflicting node based on the priority coefficient generation rule defined in S120; If multiple control requirements conflict on the same parameter node, the following rules apply: Priority is given to ensuring the regulation of higher demand classification during the fermentation stage; If the stage requirements are classified in the same way, the priority is determined based on the weight of the raw material characteristics; If the stage demand classification and raw material characteristics are the same, they are sorted from high to low according to the proportion of process violation costs; Then, according to the updated priority coefficient, the control queue order of the parameter node is dynamically adjusted: The control requirements with the highest priority coefficient are placed at the top and reserved monitoring resources are allocated to them; Generate delayed processing suggestions for the control demand with the second highest priority coefficient, including estimated processing time and alternative parameter node recommendations; Finally, the adjusted priority instructions are added to the initial adjustment policy set.

8. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 2, characterized in that: The adjustment operation types in S400 include parameter node adjustment, association relationship modification and weight dynamic update; Generating the target adjustment strategy set in S400 specifically includes the following steps: S410: If the adjustment operation type is parameter node adjustment, recalculate the dynamic weights of the affected nodes and update the sensor calibration scheme; S420: If the adjustment operation type is association modification, adjust the association closeness value and re-evaluate the core link to generate a new environment control plan; S430: If the adjustment operation type is dynamic weight update, the original dynamic weight is overwritten based on the direct weight value input by the user, and the priority sorting scheme is modified simultaneously.

9. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 2, characterized in that: The S500 specifically includes the following sub-steps: S510, executing the dynamically revised initial adjustment strategy set, including a sensor calibration plan, an environmental control plan, and a priority sorting plan; S520: Real-time collection of feedback data during the execution process, wherein the feedback data includes sensor calibration completion rate, actual control value of environmental parameters, control demand processing time, and node monitoring load change value; S530, adjusting the dynamic weights of parameter nodes in the fermentation monitoring model; S540, updating the topological structure and correlation density value of the fermentation monitoring model based on the updated dynamic weight; The specific implementation of adjusting the dynamic weights of parameter nodes in the fermentation monitoring model includes: Based on the feedback data collected by S520, the historical matching accuracy of the parameter nodes, the current control priority coefficient and the parameter stability are recalculated, and the dynamic weight of each parameter node is updated: If the sensor calibration completion rate is lower than the set threshold, the value of the weight adjustment factor is increased to strengthen the impact of the historical matching accuracy on the dynamic weight; If the deviation of the environmental parameter control value exceeds the tolerance range, the value of the weight adjustment factor is increased to increase the weight ratio of the current control priority coefficient; If the monitored load change value does not reach the target, the value of the weight adjustment factor is increased to enhance the contribution of parameter stability in the dynamic weight.

10. The method for real-time monitoring of the fermentation process of high-proof rice wine based on an intelligent sensor according to claim 9, characterized in that: The specific implementation of the feedback data collected during the execution of real-time collection in S520 includes: Triggering data collection instructions at preset time intervals, which are dynamically adjusted according to the needs of the fermentation stage; For sensor calibration solutions, collect the working status of the calibration equipment, the deviation of the monitoring values ​​before and after calibration, and the calibration operation time; For environmental control schemes, collect the operating parameters of control equipment, real-time fluctuation curves of environmental parameters, and control response delay time; For the priority sorting scheme, the actual processing order of the control requirements, the waiting time of delayed processing requirements and the monitoring data of the alternative parameter nodes are collected.

Images