A Dynamic Monitoring and Analysis Platform for the Inhibitory Effect of the Maillard Reaction

Abstract

The present invention discloses a dynamic monitoring and analysis platform for Maillard reaction inhibition effect, comprising: a data acquisition module for collecting a monitoring data set during the fermentation process of fermented tofu; a data preprocessing module for generating a preprocessed monitoring data set; a process parameter optimization module for searching and determining optimal process parameters; a Maillard reaction state recognition module for recognizing the Maillard reaction state and abnormal data points at different fermentation stages; an abnormal data detection module for constructing a dynamic monitoring model of the Maillard reaction inhibition effect; a dynamic monitoring and feedback module for realizing dynamic adjustment of the fermented tofu fermentation process parameters through a closed-loop feedback control module; and a control execution module for realizing full-process dynamic monitoring of the Maillard reaction inhibition effect of fermented tofu. The present invention effectively inhibits the excessive progress of the Maillard reaction, ensuring that the color, flavor and safety of the fermented tofu product are always maintained in the optimal state.

Description

Technical Field

[0001] The present invention relates to the technical field of monitoring and analysis of fermented bean curd production, and particularly to a dynamic monitoring and analysis platform for the inhibitory effect of Maillard reaction. Background Art

[0002] With the intelligent development of the food industry, the monitoring of the production process and quality control of fermented foods have gradually become the research focus. As a typical fermented food, the flavor, color, and safety of fermented bean curd are greatly affected by the Maillard reaction during the fermentation process. The Maillard reaction is an important non-enzymatic browning reaction in food processing. During the fermentation process of fermented bean curd, an appropriate Maillard reaction can enhance the product flavor, but excessive reaction may lead to too dark color, flavor deterioration, and even the generation of potential harmful substances.

[0003] Currently, the monitoring of the Maillard reaction during the fermentation process of fermented bean curd mainly relies on off-line detection and manual experience judgment. Usually, laboratory analysis methods such as high-performance liquid chromatography, spectral analysis, or chemical titration are used to detect the concentrations of key reactants and products. Although these methods can provide accurate data, the detection cycle is relatively long, which cannot meet the requirements of real-time monitoring and dynamic regulation. In addition, manual experience judgment has great subjectivity and lag, and it is difficult to accurately regulate the complex fermentation process, resulting in a lack of a scientific feedback adjustment mechanism in the production process.

[0004] In recent years, with the development of food intelligent detection technology, some on-line monitoring systems have been introduced into the fermented food industry. For example, sensor technology combined with data analysis methods is used to automatically monitor the fermentation environment. However, the existing on-line monitoring systems have exposed multiple problems in practical applications. First, the data acquisition system usually only focuses on single or limited parameters, such as temperature, humidity, or pH value, and cannot comprehensively reflect the actual progress of the Maillard reaction, resulting in the one-sidedness of the monitoring data. Second, the existing monitoring systems lack effective data processing and analysis capabilities, mainly relying on traditional statistical methods or simple rule threshold determination, and it is difficult to establish an accurate monitoring model for complex non-linear reaction processes. In addition, the detection of abnormal reactions is mainly based on fixed thresholds, and intelligent analysis algorithms are not introduced, making it easy to produce false judgments or missed judgments when the process changes, affecting the monitoring accuracy.

[0005] In terms of process parameter optimization, traditional fermentation control methods often use empirical methods or single-factor test methods to determine the optimal process parameters. The existing methods are not only inefficient but also difficult to adapt to the complex environmental changes during the fermentation process of fermented bean curd. In addition, the existing optimization methods lack global search and adaptive adjustment capabilities, resulting in the difficulty of dynamically adjusting the optimization of process parameters and the inability to accurately control the inhibitory effect of the Maillard reaction.

[0006] Therefore, there is an urgent need for a dynamic monitoring method that combines intelligent optimization algorithms and clustering analysis techniques to achieve precise monitoring and real-time control of the Maillard reaction inhibition effect, so as to improve the quality stability and production efficiency of fermented bean curd products. Summary of the Invention

[0007] An object of the present invention is to provide a dynamic monitoring and analysis platform for the Maillard reaction inhibition effect. The present invention effectively inhibits the excessive progress of the Maillard reaction, ensures that the color, flavor and safety of fermented bean curd products always remain in the optimal state, and improves the consistency and stability of production.

[0008] A dynamic monitoring and analysis platform for the Maillard reaction inhibition effect according to an embodiment of the present invention includes the following modules:

[0009] A data acquisition module for continuously acquiring a historical normal monitoring data set during the fermentation process of fermented bean curd;

[0010] A data preprocessing module for preprocessing the historical normal monitoring data set during the fermentation process of fermented bean curd to generate a preprocessed monitoring data set;

[0011] A process parameter optimization module for globally optimizing the process parameters in the preprocessed monitoring data set by using the ant colony optimization algorithm to search for and determine the optimal process parameters;

[0012] A Maillard reaction state recognition module for combining the optimal process parameters with the monitoring data set and using the local density peak clustering method to identify the Maillard reaction states and abnormal data points at different fermentation stages;

[0013] An abnormal data detection module for screening and fusing the identified abnormal data points and the optimal process parameters based on the clustering analysis results to construct a dynamic monitoring model for the Maillard reaction inhibition effect;

[0014] A dynamic monitoring and feedback module for comparing the dynamic monitoring model with the monitoring data collected in real time during the fermentation process of fermented bean curd, updating the optimal process parameters in real time, and realizing the dynamic adjustment of the fermentation process parameters of fermented bean curd through a closed-loop feedback control module;

[0015] A control execution module for generating a dynamic monitoring report according to the real-time dynamic adjustment result to realize the full-process dynamic monitoring of the Maillard reaction inhibition effect of fermented bean curd.

[0016] The present invention also provides a dynamic monitoring and analysis method for the Maillard reaction inhibition effect, which is applied to the above-mentioned dynamic monitoring and analysis platform for the Maillard reaction inhibition effect, and includes the following steps:

[0017] S1. Continuously collect the historical normal monitoring data set during the fermentation of fermented bean curd, preprocess the historical normal monitoring data set during the fermentation of fermented bean curd, and generate a preprocessed monitoring data set;

[0018] S2. Using the ant colony optimization algorithm to globally optimize the process parameters in the pre-processed monitoring data set, search and determine the optimal process parameters of temperature control parameters, humidity control parameters and pH value control parameters during the fermentation of fermented bean curd;

[0019] S3. The optimal process parameters and the pre-processed monitoring data set were processed using the local density peak clustering method, and the Maillard reaction process of fermented bean curd was clustered and analyzed to identify the Maillard reaction states and abnormal data points at different fermentation stages;

[0020] S4. Based on the cluster analysis results, the abnormal data points identified were screened and integrated with the optimal process parameters to construct a dynamic monitoring model for the Maillard reaction inhibition effect;

[0021] S5. The dynamic monitoring model is compared with the monitoring data collected in real time during the fermentation process of fermented bean curd, and the optimal process parameters are updated in real time, and the dynamic adjustment of the fermented bean curd process parameters is achieved through a closed-loop feedback control module;

[0022] S6. Based on the real-time dynamic adjustment results, a dynamic monitoring report including the Maillard reaction progress curve, abnormal warning and process parameter optimization suggestions is generated to achieve full-process dynamic monitoring of the Maillard reaction inhibition effect of fermented bean curd.

[0023] Optionally, the S1 includes the following steps:

[0024] S11. During the fermentation of fermented bean curd, the key monitoring parameters of the fermentation environment and matrix of the fermented bean curd are continuously collected based on the multi-source sensor system, wherein the key monitoring parameters include temperature, humidity, pH value, browning degree and concentration of Maillard reaction intermediates, and a monitoring data set of the fermentation process of the fermented bean curd is constructed in chronological order;

[0025] S12. Check the data integrity of the monitoring data set during the fermentation of fermented bean curd, remove missing values, duplicate values and outliers caused by sensor failure or environmental interference during data collection, and form a monitoring data set during the fermentation of fermented bean curd after data integrity screening;

[0026] S13. Performing time series alignment on the monitoring data set of the fermented bean curd after screening, aligning the asynchronous data sampled by different sensors to a uniform time interval, and constructing a monitoring data set of the fermented bean curd after time alignment;

[0027] S14. Standardize the data format of the monitoring data set during the fermented tofu process, normalize all monitoring parameters to a unified dimension, and construct a standardized monitoring data set for the fermented tofu process;

[0028] S15. Store the monitoring data set during the standardized fermented tofu process, construct a monitoring data storage structure for the fermented tofu process based on the time dimension, and mark the key process states to form a preprocessed monitoring data set 。

[0029] Optionally, the S2 includes the following literal steps:

[0030] S21. Based on the preprocessed monitoring data set , construct a globally optimal objective function for evaluating the Maillard reaction inhibition effect during the fermented tofu process , the globally optimal objective function uses the temperature control parameter T, the humidity control parameter H, and the pH value control parameter pH as optimization variables, and introduces weight coefficients and , which are used to measure the deviation between the currently predicted browning degree and the concentration of Maillard reaction intermediates and the predetermined target state, and form a comprehensive evaluation index reflecting the overall state of the Maillard reaction;

[0031] S22. Randomly initialize a group of ant populations between the minimum allowable value and the maximum allowable value of each pre-defined process parameter. Each ant obtains a set of initial parameter combinations, and the initial parameter combinations of each ant in the ant population are within the range allowed by the fermented tofu process;

[0032] S23. Use an adaptive search strategy to update the initial parameter combinations of each ant. According to the deviation between the current comprehensive evaluation index and the target state, dynamically adjust the search step size of each ant. When the deviation is greater than the preset value, reduce the search step size to achieve local search. When the deviation is less than the preset value, increase the search step size to accelerate global convergence;

[0033] S24. During the ant colony search process, use an adaptive pheromone update strategy based on the Maillard reaction state to dynamically weight each search path. Evaluate the deviation degree of each ant in the current search round according to the pheromone evaporation parameter ρ, the constant factor Q, and the sensitivity coefficient κ. When the predicted browning degree and the concentration of intermediate products deviate from the target state by more than the preset value, the pheromone of the corresponding search path is strengthened to guide subsequent searches to focus on the deviation area below the threshold. When the predicted browning degree and the concentration of intermediate products deviate from the target state by less than the preset value, reduce the cumulative effect of pheromone to inhibit ineffective searches;

[0034] S25. Repeat the execution of the adaptive search strategy and pheromone update until the set convergence criterion is met, that is, the change in the global optimal objective function value within consecutive rounds is lower than the predetermined threshold ε. When this convergence condition is reached, determine the optimal process parameter combination and regard the optimal process parameter combination as the optimal process parameters for the temperature control parameter, humidity control parameter, and pH value control parameter during the fermented bean curd fermentation process .

[0035] Optionally, the S3 includes the following steps:

[0036] S31. According to the optimal process parameters combined with the preprocessed monitoring data set , extract the temperature control parameter , humidity control parameter , pH value control parameter , browning degree and the concentration of Maillard reaction intermediate products as the key monitoring variables, and construct a time series-based data matrix of the Maillard reaction state of fermented bean curd ;

[0037] S32. Adopt an adaptive density calculation method based on time weights, define a time dynamic decay factor to make the influence of new data in the fermentation process larger, optimize the sensitivity of clustering analysis to real-time changes, and set the time weight density calculation function:

[0038]

[0039] where is the time weight density of the data point , is the time weight factor, is the truncation kernel function to control the effective range of time weight density calculation, represents the data matrix of the Maillard reaction state of fermented bean curd inside the data point and the data point the Euclidean distance between represents the truncation distance, represents a certain data point in the data matrix of the Maillard reaction state of fermented bean curd ;

[0040] S33. On the basis of calculating the time weight density, adopt a minimum distance calculation method based on Maillard reaction risk assessment to comprehensively evaluate the abnormality of data points:

[0041]

[0042] Among them, is the weighted minimum distance from the data point to the data point with a higher density than it, is the data point risk coefficient in the Maillard reaction inhibition evaluation, is the weight coefficient, is the data point time - weighted density;

[0043] S34. Based on the time - weighted density and weighted minimum distance calculated according to steps S32 and S33, construct an improved decision diagram, use as the coordinate axis for plotting, and select the cluster - center data points based on the clustering stability optimization mechanism to set the optimal cluster - center set C;

[0044] S35. Attribute all non - cluster - center data points in the optimal cluster - center set C to the optimal cluster center, construct a hierarchical clustering model of the Maillard reaction state of fermented bean curd, complete the identification of the Maillard reaction state in different fermentation stages, and form an evolution trend of the fermentation stage in combination with time factors;

[0045] S36. For the abnormal data points identified in the Maillard reaction state, calculate the state deviation between the abnormal data points and the cluster center:

[0046]

[0047] Among them, is the minimum weighted distance from the data point to all cluster centers, is the process parameter corresponding to the data point , is the parameter deviation adjustment factor, represents the Euclidean distance between the data point and a certain cluster center c;

[0048] When the minimum weighted distance exceeds the set threshold , this data point is determined as an abnormal data point, forming an abnormal data set for the Maillard reaction process. The abnormal data set contains the abnormal states corresponding to all abnormal data points that cause the quality decline of fermented bean curd.

[0049] Optionally, the said S4 includes the following steps:

[0050] S41. According to the hierarchical clustering results of the hierarchical clustering model of the Maillard reaction state, extract the combination of cluster - center parameters in each fermentation stage, and combine with the optimal process parameters and the abnormal data set , construct the optimal process parameter set for each fermentation stage in the fermented tofu process ;

[0051] S42. Screen the abnormal data set , extract the deviation information between the abnormal data points and the optimal process parameters, and construct the abnormal data deviation matrix ;

[0052] S43. Construct a dynamic monitoring model for the Maillard reaction inhibition effect. The dynamic monitoring model for the Maillard reaction inhibition effect uses the optimal process parameter set at different stages in the fermentation process as the target control parameters, and adjusts the parameter adaptive optimization rule through the abnormal data deviation matrix to set the output structure of the dynamic monitoring model

[0053] Optionally, the output structure of the dynamic monitoring model includes:

[0054] Normal fermentation area: Set the normal parameter range for each fermentation stage according to the optimal process parameter set ;

[0055] Abnormal detection area: Calculate the degree to which the current process parameters deviate from the optimal parameters according to the abnormal data deviation matrix and output the abnormal level

[0056] Dynamic feedback mechanism: Adjust the process parameters based on the real-time monitoring data. If the deviation of the current process parameters exceeds the threshold , trigger the Maillard reaction inhibition optimization mechanism

[0057] Optionally, the S5 includes the following steps:

[0058] S51. According to the dynamic monitoring model for the Maillard reaction inhibition effect, compare the monitoring data set collected in real time during the fermented tofu process with the optimal process parameter set and the abnormal data deviation matrix to calculate the deviation between the current process parameters and the target parameters, and establish a real-time process parameter deviation mapping matrix ;

[0059] S52. According to the real-time process parameter deviation mapping matrix , divide the process parameter state of the fermented tofu process into a stable state area, a warning state area, and an abnormal state area

[0060] S53. According to the results of the process parameter state division, for the temperature control parameter , the humidity control parameter and pH value regulation parameters Optimize and adjust them, and set a dynamic adjustment strategy:

[0061] Small - scale adaptive adjustment: When the process parameters are in the early warning state area, adopt an adaptive fine - tuning mechanism, and adjust the process parameters by less than the preset amplitude according to the historical monitoring data set, gradually guiding the parameters back to the target range;

[0062] Quick recovery adjustment: When the process parameters are in the abnormal state area, determine the optimal recovery path according to the abnormal data deviation matrix and adopt a dynamic step - size control strategy to adjust the parameters by more than the preset amplitude to quickly recover to the optimal process parameter interval;

[0063] Pheromone enhancement mechanism: Combine the pheromone update mechanism of the ant colony optimization algorithm, and strengthen the pheromone of the historical optimal recovery path during the process parameter adjustment;

[0064] S54. According to the execution result of the adjustment strategy, evaluate the Maillard reaction inhibition effect in real - time, and conduct secondary optimization according to the evaluation result.

[0065] Optionally, the stable state area, early warning state area and abnormal state area specifically include:

[0066] Stable state area: When the real - time deviation of the process parameters in the process parameter deviation mapping matrix satisfies , it is considered that the fermentation process is in a stable state, and the system maintains the current process parameters without adjustment;

[0067] Early warning state area: When the real - time deviation of the process parameters in the process parameter deviation mapping matrix , it is considered that there are deviations in the fermentation process but still within the controllable range. The system records the current deviation information and conducts trend prediction to decide whether to perform adjustment;

[0068] Abnormal state area: When the real - time deviation of the process parameters in the process parameter deviation mapping matrix , it is considered that the fermentation process is in an abnormal state. The system immediately performs dynamic adjustment of the process parameters and stores the abnormal state data in the abnormal data set ;

[0069] Among them, is the small - parameter fluctuation range during the fermentation process, is the maximum deviation during the fermentation process.

[0070] Optionally, the secondary optimization includes setting the evaluation criteria for the adjusted process parameters:

[0071] Target achievement status: When the real-time process parameter deviation after adjustment maps the real-time deviation of the process parameters in the matrix When the system remains stable, the adjustment is successful and the system enters the stable monitoring mode;

[0072] Continuous deviation state: When the adjusted process parameters are still in the warning state area or abnormal state area, re-execute the dynamic adjustment strategy and optimize the optimal recovery path;

[0073] Abnormal lock state: When the process parameters are adjusted continuously for more than the set rounds and still fail to meet the standards, the abnormal lock state is entered, triggering manual intervention and outputting an adjustment failure analysis report.

[0074] The beneficial effects of the present invention are:

[0075] (1) The present invention adopts an improved ant colony optimization algorithm to perform global search and adaptive optimization of key process parameters in the fermentation process of fermented bean curd. By introducing a dynamic step size control strategy, the ant individuals adjust the search range according to the change of the Maillard reaction state during the parameter search process, thereby improving the optimization efficiency. The Maillard reaction deviation factor is introduced into the pheromone update mechanism to make the enhancement direction of the pheromone consistent with the actual reaction inhibition target, thereby accelerating the optimization convergence speed, improving the matching degree of process parameters, ensuring that the fermentation process is always in the optimal state, and reducing the quality fluctuation between product batches.

[0076] (2) The present invention adopts a local density peak clustering method, combined with time weighted density calculation and weighted minimum distance calculation based on Maillard reaction risk assessment, to achieve hierarchical identification of the Maillard reaction state at different fermentation stages. The latest collected data is given a higher weight through the time dynamic attenuation factor to improve the response capability to real-time changes. At the same time, the deviation factor based on the Maillard reaction risk is used to classify the data points for abnormality to ensure the accuracy of the identification of abnormal points. Compared with traditional methods, the present invention can accurately capture potential abnormal fluctuations in the fermentation process and avoid monitoring errors caused by data lag or unreasonable fixed threshold setting.

[0077] (3) The present invention constructs a closed-loop control system that integrates data collection, intelligent analysis and real-time feedback. The dynamic monitoring model is used to compare and analyze the process data collected in real time, and the process parameters are dynamically adjusted according to the state division. In the parameter adjustment process, the pheromone enhancement mechanism is used to optimize the adjustment path, and the parameter correction strategy is optimized through self-learning of historical adjustment data to improve the adjustment accuracy. The closed-loop feedback control mechanism can achieve accurate and rapid parameter adjustment during the fermentation process, effectively inhibit the excessive progress of the Maillard reaction, ensure that the color, flavor and safety of the fermented bean curd product are always maintained in the optimal state, and improve the consistency and stability of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, but do not constitute a limitation to the present invention. In the accompanying drawings:

[0079] Figure 1 It is a flowchart of a dynamic monitoring and analysis platform for the inhibitory effect of Maillard reaction proposed by the present invention. Specific embodiments

[0080] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only showing the basic structure of the present invention in a schematic way, so they only show the components related to the present invention.

[0081] Reference Figure 1 , a dynamic monitoring and analysis platform for the inhibitory effect of Maillard reaction, includes the following modules:

[0082] The data acquisition module is used to continuously acquire the monitoring data set during the fermentation process of fermented bean curd.

[0083] The data preprocessing module is used to preprocess the monitoring data set during the fermentation process of fermented bean curd to generate a preprocessed monitoring data set.

[0084] The process parameter optimization module is used to globally optimize the process parameters in the preprocessed monitoring data set by using the ant colony optimization algorithm, and search for and determine the optimal process parameters.

[0085] The Maillard reaction state recognition module is used to combine the optimal process parameters with the monitoring data set, and use the local density peak clustering method to identify the Maillard reaction states and abnormal data points at different fermentation stages.

[0086] The abnormal data detection module is used to screen and fuse the identified abnormal data points and the optimal process parameters based on the clustering analysis results to construct a dynamic monitoring model for the inhibitory effect of Maillard reaction.

[0087] The dynamic monitoring and feedback module is used to compare the dynamic monitoring model with the monitoring data collected in real time during the fermentation process of fermented bean curd, update the optimal process parameters in real time, and realize the dynamic adjustment of the process parameters of fermented bean curd fermentation through the closed-loop feedback control module.

[0088] The control execution module is used to generate a dynamic monitoring report according to the real-time dynamic adjustment results, and realize the full-process dynamic monitoring of the inhibitory effect of Maillard reaction of fermented bean curd.

[0089] A dynamic monitoring and analysis method for the inhibitory effect of Maillard reaction, applied to a dynamic monitoring and analysis platform for the inhibitory effect of Maillard reaction, includes the following steps:

[0090] S1. During the fermenting process of fermented tofu, continuously collect the historical normal monitoring dataset during the fermenting process of fermented tofu, preprocess the historical normal monitoring dataset during the fermenting process of fermented tofu, and generate a preprocessed monitoring dataset;

[0091] S2. Use the ant colony optimization algorithm to globally optimize the process parameters in the preprocessed monitoring dataset, search for and determine the optimal process parameters of the temperature control parameter, humidity control parameter, and pH value control parameter during the fermenting process of fermented tofu;

[0092] S3. Process the optimal process parameters and the preprocessed monitoring dataset through the local density peak clustering method, perform clustering analysis on the Maillard reaction process of fermented tofu, and identify the Maillard reaction states and abnormal data points in different fermentation stages;

[0093] S4. Based on the clustering analysis results, screen and fuse the identified abnormal data points and the optimal process parameters to construct a dynamic monitoring model for the inhibitory effect of the Maillard reaction;

[0094] S5. Compare the dynamic monitoring model with the monitoring data collected in real time during the fermenting process of fermented tofu, update the optimal process parameters in real time, and achieve dynamic adjustment of the process parameters of the fermented tofu through the closed-loop feedback control module;

[0095] S6. According to the real-time dynamic adjustment results, generate a dynamic monitoring report including the Maillard reaction process curve, abnormal warning, and process parameter optimization suggestions, and achieve full-process dynamic monitoring of the inhibitory effect of the Maillard reaction of fermented tofu.

[0096] During the fermenting process of fermented tofu, after the monitoring data collected in real time is preprocessed, process parameter optimized, state identified, anomaly detected, and dynamically regulated, the system generates a comprehensive dynamic monitoring report based on the final adjustment results to achieve full-process dynamic monitoring of the inhibitory effect of the Maillard reaction.

[0097] The system automatically draws the Maillard reaction process curve based on time series data. The Maillard reaction process curve shows the changing trends of key parameters (temperature, humidity, pH value, amino acid concentration, reducing sugar concentration, browning degree, and Maillard reaction intermediate product concentration) over time during the entire fermentation cycle, and combines the optimal process parameters to set the reference interval, and visually marks the deviation between the actual monitoring value and the target value.

[0098] Based on the abnormal data points identified by the anomaly detection module, the system generates abnormal warning information, records the time of the anomaly occurrence, the production batches involved, the specific deviation parameters, and the anomaly level. The abnormal warning information is displayed in a classified manner:

[0099] Low-level anomaly (yellow warning): The parameter deviates from the optimal range but is still within the safe interval, and it is recommended to pay attention;

[0100] Intermediate abnormality (orange warning): The parameters exceed the safety range, but have not yet triggered the Maillard reaction to intensify. The system automatically adjusts the process parameters and records the adjustment path.

[0101] Advanced abnormality (red warning): Severe deviation of parameters may cause the Maillard reaction to get out of control, triggering emergency adjustments and sending an early warning report to the production management system.

[0102] At the same time, the system predicts future parameter trends based on the historical adjustment path of the ant colony optimization algorithm and combines it with the dynamic monitoring model, and generates process parameter optimization suggestions. The optimization suggestions include:

[0103] 1. Short-term adjustment suggestions: According to the real-time adjustment plan for the current batch, reduce the temperature by 0.5-1℃, adjust the humidity by ±2%, and add an appropriate amount of alkaline regulator to maintain pH balance;

[0104] 2. Long-term optimization suggestions: Based on the historical data of multiple production batches, an optimized fermentation parameter setting plan is provided. It is recommended that future batches adopt the optimal parameter range, temperature 27.2-27.8℃, humidity 77%-81%, pH 5.6-5.8, to maximize the inhibition of the adverse effects of the Maillard reaction and improve product consistency.

[0105] The system outputs dynamic monitoring reports in the form of data tables, trend curves, abnormal analysis and optimization suggestions, and stores them in the database to support subsequent production optimization and quality traceability, realize full-process dynamic monitoring of the Maillard reaction inhibition effect, and improve the intelligent management level of the fermented tofu production process.

[0106] In this implementation, S1 includes the following steps:

[0107] S11. During the fermentation of fermented bean curd, the key monitoring parameters of the fermentation environment and matrix of the fermented bean curd are continuously collected based on the multi-source sensor system. The key monitoring parameters include temperature, humidity, pH value, browning degree and concentration of Maillard reaction intermediates. In practical applications, amino acid concentration and reducing sugar concentration can also be monitored, and a monitoring data set of the fermentation process of fermented bean curd is constructed in chronological order.

[0108] S12. Check the data integrity of the monitoring data set during the fermentation of fermented bean curd, remove missing values, duplicate values and outliers caused by sensor failure or environmental interference during data collection, and form a monitoring data set during the fermentation of fermented bean curd after data integrity screening;

[0109] S13. Performing time series alignment on the monitoring data set of the fermented bean curd after screening, aligning the asynchronous data sampled by different sensors to a uniform time interval, and constructing a monitoring data set of the fermented bean curd after time alignment;

[0110] S14. Standardize the data format of the monitoring data set during the fermented bean curd fermentation process, normalize all monitoring parameters to a unified dimension, and construct a standardized monitoring data set for the fermented bean curd fermentation process;

[0111] S15. Store the monitoring data set during the standardized fermented bean curd fermentation process, construct a monitoring data storage structure for the fermented bean curd fermentation process based on the time dimension, and mark the key process states to form a preprocessed monitoring data set 。

[0112] In this embodiment, S2 includes the following literal steps:

[0113] S21. Based on the preprocessed monitoring data set , construct a global optimal objective function for evaluating the Maillard reaction inhibition effect during the fermented bean curd fermentation process , the global optimal objective function uses the temperature control parameter T, the humidity control parameter H, and the pH value control parameter pH as optimization variables, and introduces weight coefficients and , which are used to measure the deviation between the currently predicted browning degree and the concentration of Maillard reaction intermediate products and the predetermined target state, and form a comprehensive evaluation index reflecting the overall state of the Maillard reaction:

[0114] ;

[0115] Among them, is the browning degree predicted based on the current process parameters, is the predicted concentration of Maillard reaction intermediate products, and are the set target browning degree and target intermediate product concentration respectively, and are weight coefficients used to balance the importance of each index;

[0116] The optimization principle of the objective function is to minimize the deviation between the browning degree and the concentration of Maillard reaction intermediate products, ensure that the Maillard reaction is effectively inhibited, and thus maintain the best color and flavor of the fermented bean curd.

[0117] S22. Within the process parameter search space , use the uniform random number to initialize the process parameter variable set , generate an initial ant population, and the initial parameter combination of each ant :

[0118] ;

[0119] Among them, and are the minimum and maximum feasible values of temperature , humidity and pH value respectively, is the ant serial number, is the weight coefficient;

[0120] S23. Update the initial parameter combinations of each ant using an adaptive search strategy. According to the deviation between the current comprehensive evaluation index and the target state, dynamically adjust the search step size of each ant. When the deviation is greater than the preset value, reduce the search step size to achieve local search, and when the deviation is less than the preset value, increase the search step size to accelerate global convergence:

[0121] ;

[0122] Among them, is the parameter combination of the th ant in the th round, is the parameter combination of the th ant in the th round, is the standard search step size vector, is the dynamic step size factor, whose value decreases as the global optimal objective function decreases, is the step size adjustment sensitivity coefficient, so that when the Maillard reaction deviates greatly from the target state, the search step size is appropriately reduced to achieve more refined local exploration;

[0123] When the gap between the current solution and the optimal solution in the formula is large, the step size becomes larger to accelerate global search. When the current solution is close to the optimal solution, the step size is reduced to improve local search accuracy.

[0124] S24. During the ant colony search process, adopt an adaptive pheromone update strategy based on the Maillard reaction state to dynamically weight each search path. Evaluate the deviation degree of each ant in the current search round according to the pheromone evaporation parameter ρ, the constant factor Q, and the sensitivity coefficient κ. When the predicted browning degree and the intermediate product concentration deviate from the target state by more than the preset value, the pheromone of the corresponding search path is strengthened to guide subsequent searches to focus on the deviation area below the threshold. When the predicted browning degree and the intermediate product concentration deviate from the target state by less than the preset value, reduce the cumulative effect of pheromone to inhibit ineffective search:

[0125] ;

[0126] Among them, represents the path and the path in the The cumulative pheromone concentration of the wheel, indicating the path at the cumulative pheromone concentration in the +1-th round, is the pheromone evaporation coefficient, is the constant factor, is the total number of ants, is the pheromone sensitivity coefficient, indicating the comprehensive deviation between the parameter combination of the -th ant and the target state:

[0127] ;

[0128] The adaptive pheromone update strategy of the formula, by combining the global optimal objective function and the Maillard reaction state deviation , makes the update of pheromone more in line with the requirements of Maillard reaction inhibition during the fermented bean curd fermentation process, avoids the local convergence problem of the traditional ant colony algorithm, improves the optimization efficiency, and finally ensures that the Maillard reaction is accurately controlled, enhancing the consistency and stability of product quality.

[0129] S25. Repeat the adaptive search strategy and pheromone update until the set convergence criterion is met, that is, the change in the global optimal objective function value within consecutive rounds is lower than the predetermined threshold ε. When this convergence condition is reached, determine the optimal process parameter combination and regard the optimal process parameter combination as the optimal process parameters for the temperature control parameter, humidity control parameter, and pH value control parameter during the fermented bean curd fermentation process :

[0130] ;

[0131] This embodiment uses an improved ant colony optimization algorithm to globally search and adaptively optimize the key process parameters during the fermented bean curd fermentation process. By introducing a dynamic step size control strategy, the ant individuals adjust the search range according to the change in the Maillard reaction state during the parameter search process, improving the optimization efficiency. A Maillard reaction deviation factor is introduced into the pheromone update mechanism, making the enhancement direction of the pheromone consistent with the actual reaction inhibition target, thereby accelerating the optimization convergence speed, improving the process parameter matching degree, ensuring that the fermentation process is always in the optimal state, and reducing the quality fluctuation between product batches.

[0132] In this embodiment, S3 includes the following steps:

[0133] S31. According to the optimal process parameters combined with the preprocessed monitoring data set , extract the temperature control parameter during the fermented bean curd fermentation process , Humidity control parameters , pH value control parameters , Browning degree And the concentration of Maillard reaction intermediate products As the key monitoring variables, construct a Maillard reaction state data matrix of fermented bean curd based on time series ;

[0134] S32. Adopt an adaptive density calculation method based on time weight, define a time dynamic decay factor to make the influence of new data in the fermentation process larger, optimize the sensitivity of clustering analysis to real-time changes, and set the time weight density calculation function:

[0135]

[0136] Among them, Is the time weight density of the data point , Is the time weight factor, Is the truncated kernel function, which controls the effective range of time weight density calculation, Represents the data point in the Maillard reaction state data matrix of fermented bean curd Inside And the data point The Euclidean distance between Represents the truncated distance, Represents the Maillard reaction state data matrix of fermented bean curd A certain data point in;

[0137] In the formula, the latest data point is given greater influence, the response ability to real-time Maillard reaction changes is improved, interference of outdated data on clustering results is avoided, and the accuracy of clustering analysis is improved.

[0138] S33. On the basis of calculating the time weight density, adopt a minimum distance calculation method based on Maillard reaction risk assessment to comprehensively evaluate the abnormality of data points:

[0139]

[0140] Among them, Is the weighted minimum distance from the data point To the data point with higher density than it, Is the risk coefficient of the data point In the Maillard reaction inhibition assessment, Is the weight coefficient, Is the data point Time weight density of;

[0141] The formula calculates the distance of data points by adjusting with risk coefficients, making data points with a higher Maillard reaction risk more likely to be judged as outliers and improving the accuracy of outlier detection.

[0142] S34. The time-weighted density and weighted minimum distance calculated according to steps S32 and S33 are used to construct an improved decision graph. is used as the coordinate axis for plotting, and the cluster center data points are selected based on the cluster stability optimization mechanism to set the optimal cluster center set C;

[0143] S35. All non-cluster center data points in the optimal cluster center set C are attributed to the optimal cluster center to construct a hierarchical clustering model of the Maillard reaction state of fermented bean curd, complete the identification of the Maillard reaction state at different fermentation stages, and form an evolution trend of the fermentation stage in combination with time factors;

[0144] S36. For the outlier data points identified in the Maillard reaction state, calculate the state deviation between the outlier data points and the cluster center:

[0145]

[0146] where is the minimum weighted distance from the data point to all cluster centers, is the process parameter corresponding to the data point , is the parameter deviation adjustment factor, represents the Euclidean distance between the data point and a certain cluster center c;

[0147] When the minimum weighted distance exceeds the set threshold , the data point is judged as an outlier data point to form an outlier data set of the Maillard reaction process. The outlier data set contains the abnormal states corresponding to all outlier data points that cause the quality decline of fermented bean curd.

[0148] This embodiment adopts the local density peak clustering method, combines the time-weighted density calculation and the weighted minimum distance calculation based on the Maillard reaction risk assessment to achieve hierarchical identification of the Maillard reaction states at different fermentation stages. A time dynamic decay factor is used to assign higher weights to the latest collected data, improving the response ability to real-time changes. At the same time, a deviation factor based on the Maillard reaction risk is used to classify abnormal data points to ensure the accuracy of abnormal point identification. Compared with traditional methods, the present invention can accurately capture potential abnormal fluctuations during the fermentation process, avoiding monitoring errors caused by data lag or unreasonable setting of fixed thresholds.

[0149] In this embodiment, S4 includes the following steps:

[0150] S41. According to the hierarchical clustering results of the hierarchical clustering model of the Maillard reaction state, extract the clustering center parameter combinations at each fermentation stage, and combine the optimal process parameters and the abnormal data set to construct the optimal process parameter set at each fermentation stage during the fermented bean curd fermentation process ;

[0151] S42. Screen the abnormal data set to extract the deviation information of the abnormal data points from the optimal process parameters, and construct an abnormal data deviation matrix ;

[0152] S43. Construct a dynamic monitoring model for the Maillard reaction inhibition effect. The dynamic monitoring model for the Maillard reaction inhibition effect uses the optimal process parameter sets at different stages during the fermentation process as the target control parameters, and adjusts the parameter adaptive optimization rule through the abnormal data deviation matrix to set the output structure of the dynamic monitoring model.

[0153] In this embodiment, the output structure of the dynamic monitoring model includes:

[0154] Normal fermentation area: Set the normal parameter intervals at each fermentation stage according to the optimal process parameter set ;

[0155] Abnormal detection area: Calculate the degree to which the current process parameters deviate from the optimal parameters according to the abnormal data deviation matrix and output the abnormal level;

[0156] Dynamic feedback mechanism: Adjust the process parameters based on real-time monitoring data. If the deviation of the current process parameters exceeds the threshold , trigger the Maillard reaction inhibition optimization mechanism.

[0157] In this embodiment, S5 includes the following steps:

[0158] S51. According to the dynamic monitoring model for Maillard reaction inhibition, compare the monitoring data set collected in real time during the fermented tofu process with the optimal process parameter set and the abnormal data deviation matrix to calculate the deviation between the current process parameters and the target parameters, and establish a real-time process parameter deviation mapping matrix ;

[0159] S52. According to the real-time process parameter deviation mapping matrix , divide the process parameter state during the fermented tofu process into a stable state area, a warning state area, and an abnormal state area;

[0160] S53. According to the result of the process parameter state division, optimize and adjust the temperature control parameter , the humidity control parameter and the pH value control parameter , and set a dynamic adjustment strategy:

[0161] Small-amplitude adaptive adjustment: When the process parameters are in the warning state area, adopt an adaptive fine-tuning mechanism to adjust the process parameters by less than the preset amplitude according to the historical monitoring data set, and gradually guide the parameters back to the target range;

[0162] Fast recovery adjustment: When the process parameters are in the abnormal state area, determine the optimal recovery path according to the abnormal data deviation matrix and adopt a dynamic step size control strategy to adjust the parameters by more than the preset amplitude to quickly recover to the optimal process parameter interval;

[0163] Pheromone enhancement mechanism: Combine the pheromone update mechanism of the ant colony optimization algorithm to enhance the pheromone of the historical optimal recovery path during the process parameter adjustment to improve the optimization efficiency of the future optimal recovery path;

[0164] S54. According to the execution result of the adjustment strategy, evaluate the Maillard reaction inhibition effect in real time, and perform secondary optimization according to the evaluation result.

[0165] In this embodiment, the stable state area, the warning state area, and the abnormal state area specifically include:

[0166] Stable state area: When the real-time deviation of the process parameters in the process parameter deviation mapping matrix satisfies , it is considered that the fermentation process is in a stable state, and the system maintains the current process parameters without adjustment;

[0167] Warning state area: When the real-time deviation of the process parameters in the process parameter deviation mapping matrix When it is determined that there is a deviation in the fermentation process but it is still within the controllable range, the system records the current deviation information and conducts trend prediction to decide whether to perform adjustments;

[0168] Abnormal state area: When the real-time deviation of the process parameters within the process parameter deviation mapping matrix it is considered that the fermentation process is in an abnormal state. The system immediately performs dynamic adjustment of the process parameters and stores the abnormal state data in the abnormal data set ;

[0169] Among them, is the range of small parameter fluctuations during the fermentation process, is the maximum deviation during the fermentation process.

[0170] In this embodiment, the secondary optimization includes setting evaluation criteria for the adjusted process parameters:

[0171] Target achievement state: When the real-time deviation of the process parameters within the adjusted real-time process parameter deviation mapping matrix and remains stable, the adjustment is successful and enters the stable monitoring mode;

[0172] Continuous deviation state: When the adjusted process parameters are still in the early warning state area or abnormal state area, the dynamic adjustment strategy is re-executed and the optimal recovery path is optimized;

[0173] Abnormal locking state: When the process parameters still do not meet the standards after continuous adjustment exceeding the set number of rounds, it enters the abnormal locking state, triggers manual intervention and outputs an analysis report of the adjustment failure.

[0174] The present invention constructs a closed-loop control system integrating data acquisition, intelligent analysis and real-time feedback, uses a dynamic monitoring model to compare and analyze the real-time collected process data, dynamically adjusts the process parameters according to the state division, adopts a pheromone enhancement mechanism to optimize the adjustment path during the parameter adjustment process, and optimizes the parameter correction strategy through self-learning of historical adjustment data to improve the adjustment accuracy. The closed-loop feedback control mechanism can achieve precise and rapid parameter adjustment during the fermentation process, effectively inhibit the excessive progress of the Maillard reaction, ensure that the color, flavor and safety of the fermented bean curd products always remain in the optimal state, and improve the consistency and stability of production.

[0175] Example 1:

[0176] In March 2024, a food processing enterprise found that there were large differences in color and flavor among recent batches of fermented bean curd products during the production of fermented bean curd. The color of some batches of fermented bean curd was too dark, the flavor was bitter, and there was even a slight burnt smell, seriously affecting the product quality. The technical team of the enterprise decided to use the method of the present invention to optimize the production process and improve the product quality stability.

[0177] The enterprise installed sensor modules on six fermented bean curd production lines in the fermentation workshop, including temperature sensors (model: TMP36), humidity sensors (model: DHT22), pH sensors (model: SEN0161), online chromaticity sensors (model: OPL530), amino acid concentration detection modules (using biosensing technology), and reducing sugar concentration sensors, to collect key process parameters during the fermentation of fermented bean curd in real time and upload the data to the central monitoring system.

[0178] On March 10, the system started running, and data was collected every 10 minutes on each production line to form a standardized monitoring data set. After five days of stable operation, the average browning degree of the production batches was between 0.18 and 0.22, and the concentration of Maillard reaction intermediate products was between 0.45 and 0.55 mmol / L, indicating that the Maillard reaction was within the normal range.

[0179] On March 16, the system detected an abnormal Maillard reaction on production line 3 (batch number: TL-0316-3):

[0180] 15:30: Sensor data showed that the pH value dropped from 5.6 to 5.3, the temperature rose slightly to 29.2°C (the set temperature was 28°C), the reducing sugar concentration increased from 1.2% to 1.9%, and the browning degree reached 0.31, significantly higher than the average value of the previous few days, and this trend continued to rise.

[0181] 15:40: The system used the local density peak clustering method for real-time data analysis and found that compared with normal fermentation batches, the distribution of data points of this batch in the Maillard reaction state clustering model deviated from the normal area and belonged to abnormal clustering points.

[0182] 15:45: Combining the optimal process parameters calculated by the ant colony optimization algorithm, the system determined that the Maillard reaction of this batch entered the overreaction risk interval, generated a "Maillard reaction abnormal" warning (warning number: ALRT-0316-3), and pushed it to the production management terminal.

[0183] 15:50: The production person in charge received the warning, viewed the detailed analysis report, and found that the key factors leading to the abnormality were high temperature and a drop in pH value. The system recommended adjusting the fermentation room temperature to 27.5°C and appropriately adding an alkaline regulator (sodium carbonate solution) to restore the pH value to 5.7.

[0184] 16:00: The production personnel confirmed the adjustment strategy in the central control system. The system automatically adjusted the fermentation room temperature to 27.5°C and instructed to add 0.02% sodium carbonate solution.

[0185] 16:30: The monitoring data shows that the pH value has risen back to 5.6, the temperature has dropped to the set value, the browning degree has stabilized at 0.23, and the Maillard reaction has returned to the normal range. The system records this adjustment as an effective optimization case and stores it in the optimization path database for intelligent optimization in future similar situations.

[0186] This embodiment fully demonstrates the effectiveness of the present invention in the dynamic monitoring of Maillard reaction inhibition. Compared with traditional methods, the advantages of the present invention are reflected in the following aspects:

[0187] 1. Strong real-time performance, capable of quickly detecting and providing feedback in the early stage when the Maillard reaction anomaly occurs, avoiding the lag of traditional methods in detecting problems at the terminal.

[0188] 2. High adaptive optimization ability, using the ant colony optimization algorithm to dynamically adjust process parameters, improving the accuracy of Maillard reaction inhibition, and making the color, flavor and consistency of the product significantly better than traditional methods.

[0189] 3. Reduction of abnormal batches and improvement of production stability. The method of the present invention can reduce more than 70% of fermentation anomalies through anomaly detection and closed-loop feedback, and the product qualification rate has increased by nearly 20%.

[0190] The experimental results show that the intelligent optimization monitoring system of the present invention can effectively solve the problems of lag, insufficient accuracy and slow regulation response existing in traditional process monitoring methods, greatly improving the automation level and product quality stability of fermented bean curd production, and having significant industrial application value.

[0191] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A dynamic monitoring and analysis platform for Maillard reaction inhibition effect, characterized in that Including: A data acquisition module for continuously acquiring the historical normal monitoring data set during the fermented bean curd fermentation process; A data preprocessing module for preprocessing the historical normal monitoring data set during the fermented bean curd fermentation process to generate a preprocessed monitoring data set; A process parameter optimization module for globally optimizing the process parameters in the preprocessed monitoring data set by using an improved ant colony optimization algorithm, and searching for and determining the optimal process parameters; The improved ant colony optimization algorithm includes an adaptive search strategy and an adaptive pheromone update strategy; The adaptive search strategy is to update the initial parameter combinations of each ant, and dynamically adjust the search step of each ant according to the deviation between the current comprehensive evaluation index and the target state. When the deviation is greater than the preset value, the search step is reduced to achieve local search. When the deviation is less than the preset value, the search step is increased to accelerate global convergence; For the comprehensive evaluation index, the temperature control parameter T, the humidity control parameter H, and the pH value control parameter pH are used as optimization variables: ； Among them, is the browning degree predicted based on the current process parameters, is the predicted concentration of Maillard reaction intermediate products, and are the set target browning degree and target intermediate product concentration respectively, and are weight coefficients used to balance the importance of various indicators; The adaptive pheromone update strategy is to dynamically weight each search path during the ant colony search process, and evaluate the deviation degree of each ant in the current search round according to the pheromone evaporation parameter ρ, the constant factor Q, and the sensitivity coefficient κ. When the predicted browning degree and the intermediate product concentration deviate from the target state by more than the preset value, the pheromone of the corresponding search path is strengthened, and subsequent searches focus on the deviation area below the threshold. When the predicted browning degree and the intermediate product concentration deviate from the target state by less than the preset value, the cumulative effect of the pheromone is reduced; A Maillard reaction state recognition module for combining the optimal process parameters and the monitoring data set, and using the local density peak clustering method to identify the Maillard reaction states and abnormal data points at different fermentation stages; An abnormal data detection module for screening and fusing the identified abnormal data points and the optimal process parameters based on the clustering analysis results to construct a dynamic monitoring model for the Maillard reaction inhibition effect; A dynamic monitoring and feedback module for comparing the dynamic monitoring model with the monitoring data collected in real time during the fermented bean curd fermentation process, updating the optimal process parameters in real time, and realizing the dynamic adjustment of the fermented bean curd fermentation process parameters through a closed-loop feedback control module; A control execution module for generating a dynamic monitoring report according to the real-time dynamic adjustment result, and realizing the full-process dynamic monitoring of the Maillard reaction inhibition effect of the fermented bean curd.

2. A dynamic monitoring and analysis method for Maillard reaction inhibition effect, applied to the dynamic monitoring and analysis platform for Maillard reaction inhibition effect described in claim 1, characterized in that, Including the following steps: S1. Acquire the historical normal monitoring data set during the fermented bean curd fermentation process, and preprocess the historical normal monitoring data set during the fermented bean curd fermentation process to generate a preprocessed monitoring data set; S2. Globally optimize the process parameters in the preprocessed monitoring data set by using an improved ant colony optimization algorithm, and search for and determine the optimal process parameters of the temperature control parameter, the humidity control parameter, and the pH value control parameter during the fermented bean curd fermentation process; S3. Use the local density peak clustering method to process the optimal process parameters and the preprocessed monitoring data set, perform clustering analysis on the Maillard reaction process of the fermented bean curd, and identify the Maillard reaction states and abnormal data points at different fermentation stages; S4. Based on the cluster analysis results, the abnormal data points identified were screened and integrated with the optimal process parameters to construct a dynamic monitoring model for the Maillard reaction inhibition effect; S5. Compare the dynamic monitoring model with the monitoring data collected in real time during the fermentation process of fermented bean curd, update the optimal process parameters in real time, and dynamically adjust the process parameters of fermented bean curd; S6. Based on the real-time dynamic adjustment results, a dynamic monitoring report including the Maillard reaction progress curve, abnormal warning and process parameter optimization suggestions is generated to achieve full-process dynamic monitoring of the Maillard reaction inhibition effect of fermented bean curd.

3. A method for dynamically monitoring and analyzing the Maillard reaction inhibition effect according to claim 2, characterized in that, The S1 comprises the following steps: S11. During the fermentation of fermented bean curd, the key monitoring parameters of the fermentation environment and matrix of the fermented bean curd are continuously collected based on the multi-source sensor system, wherein the key monitoring parameters include temperature, humidity, pH value, browning degree and concentration of Maillard reaction intermediates, and a monitoring data set of the fermentation process of the fermented bean curd is constructed in chronological order; S12. Check the data integrity of the monitoring data set during the fermentation of fermented bean curd, remove missing values, duplicate values and outliers caused by sensor failure or environmental interference during data collection, and form a monitoring data set during the fermentation of fermented bean curd after data integrity screening; S13. Performing time series alignment on the monitoring data set of the fermented bean curd after screening, aligning the asynchronous data sampled by different sensors to a uniform time interval, and constructing a monitoring data set of the fermented bean curd after time alignment; S14. Standardizing the data format of the aligned monitoring data set during the fermentation process of fermented bean curd, normalizing all monitoring parameters to a unified dimension, and constructing a standardized monitoring data set during the fermentation process of fermented bean curd; S15. Store the monitoring data set during the standardized fermented tofu process, construct a monitoring data storage structure for the fermented tofu process based on the time dimension, and label the key process states to form a preprocessed monitoring data set 。 4. A method for dynamically monitoring and analyzing the Maillard reaction inhibition effect according to claim 3, characterized in that The S3 comprises the following steps: S31. According to the optimal process parameters combined with the preprocessed monitoring data set , extract the temperature control parameters , humidity control parameters , pH value control parameters , browning degree and the concentration of Maillard reaction intermediates as the key monitoring variables, and construct a time-series based data matrix of the Maillard reaction state of fermented bean curd ; S32. Adopting the adaptive density calculation method based on time weight, defining the time dynamic attenuation factor, making the influence of new data in the fermentation process larger, optimizing the sensitivity of cluster analysis to real-time changes, and setting the time weight density calculation function: S33. Based on the calculation of time weight density, the minimum distance calculation method based on Maillard reaction risk assessment is used to comprehensively evaluate the abnormality of data points; The time-weighted density calculated according to steps S32 and S33 and the weighted minimum distance , construct an improved decision diagram, and use as the coordinate axis for plotting, and select the cluster center data points based on the cluster stability optimization mechanism, and set the optimal cluster center set C; S35. All non-cluster center data points of the optimal cluster center set C are attributed to the optimal cluster center, a hierarchical clustering model of the Maillard reaction state of fermented bean curd is constructed, the Maillard reaction state of different fermentation stages is identified, and the fermentation stage evolution trend is formed in combination with the time factor; S36. For the abnormal data points identified by the Maillard reaction state, the state deviation between the abnormal data points and the cluster center is calculated: When the data point to the minimum weighted distance to all cluster centers exceeds the set threshold the data point is determined as an abnormal data point, forming an abnormal data set of the Maillard reaction process The abnormal data set contains the abnormal states corresponding to all abnormal data points that cause the quality decline of fermented tofu.

5. The dynamic monitoring and analysis method for the Maillard reaction inhibition effect according to claim 4, characterized in that, The S4 comprises the following steps: S41. According to the hierarchical clustering results of the hierarchical clustering model based on the Maillard reaction state, extract the combination of clustering center parameters at each fermentation stage, and combine with the optimal process parameters and the abnormal data set , to construct the optimal process parameter set at each fermentation stage during the fermented tofu production process ; S42. For the abnormal data set perform screening, extract the deviation information between the abnormal data points and the optimal process parameters, and construct an abnormal data deviation matrix ; S43. Construct a dynamic monitoring model for the Maillard reaction inhibition effect. The dynamic monitoring model for the Maillard reaction inhibition effect uses the set of optimal process parameters at different stages during the fermentation process as the target control parameters, and adjusts the parameter adaptive optimization rule through the abnormal data deviation matrix to set the output structure of the dynamic monitoring model.

6. A method for dynamically monitoring and analyzing the Maillard reaction inhibition effect according to claim 5, characterized in that, The output structure of the dynamic monitoring model includes: Normal fermentation area: Based on the optimal process parameter set Set the normal parameter ranges for each fermentation stage; Abnormal detection area: Based on the abnormal data deviation matrix Calculate the degree to which the current process parameters deviate from the optimal parameters and output the abnormal level; Dynamic feedback mechanism: Adjust process parameters based on real-time monitoring data. If the deviation of the current process parameters exceeds the threshold , trigger the Maillard reaction inhibition optimization mechanism.

7. A method for dynamically monitoring and analyzing the Maillard reaction inhibition effect according to claim 5, characterized in that, The S5 comprises the following steps: S51. According to the dynamic monitoring model for Maillard reaction inhibition, the monitoring data set collected in real time during the fermenting process of fermented bean curd is compared with the set of optimal process parameters and the abnormal data deviation matrix to calculate the deviation between the current process parameters and the target parameters, and establish a real-time process parameter deviation mapping matrix ; S52. According to the real-time process parameter deviation mapping matrix , the process parameter state during the fermented tofu production process is divided into a stable state area, a warning state area, and an abnormal state area; S53. According to the classification result of process parameter status, optimize and adjust the temperature control parameter , humidity control parameter and pH value control parameter to set a dynamic adjustment strategy: Small adaptive adjustment: When the process parameters are in the warning state area, the adaptive fine-tuning mechanism is used to adjust the process parameters by a smaller amount than the preset amount based on the historical monitoring data set, gradually guiding the parameters to return to the target range; Quick recovery adjustment: When the process parameters are in the abnormal state area, based on the abnormal data deviation matrix determine the optimal recovery path, and adopt a dynamic step size control strategy to adjust the parameters by more than the preset amplitude to quickly recover to the optimal process parameter range; Pheromone enhancement mechanism: Combined with the pheromone update mechanism of the improved ant colony optimization algorithm, pheromone enhancement is performed on the historical optimal recovery path during the process parameter adjustment; S54. Based on the execution results of the adjustment strategy, the Maillard reaction inhibition effect is evaluated in real time, and secondary optimization is carried out according to the evaluation results.

8. A dynamic monitoring and analysis method for the Maillard reaction inhibition effect according to claim 7, characterized in that The stable state region includes: when the real-time deviation of process parameters in the process parameter deviation mapping matrix satisfies it is considered that the fermentation process is in a stable state, and the system maintains the current process parameters without adjustment; The warning status area includes: when there is a real-time deviation of process parameters in the process parameter deviation mapping matrix it is considered that there is a deviation in the fermentation process but it is still within the controllable range. The system records the current deviation information and conducts trend prediction to determine whether to execute adjustment; The abnormal state area includes: when the real-time deviation of the process parameters in the process parameter deviation mapping matrix is such that it is considered that the fermentation process is in an abnormal state, the system immediately performs dynamic adjustment of the process parameters and stores the abnormal state data in the abnormal data set ; Among them, is the fluctuation range of minute parameters during the fermentation process, is the maximum deviation during the fermentation process.

9. A method for dynamically monitoring and analyzing the Maillard reaction inhibition effect according to claim 7, characterized in that, the secondary optimization includes setting evaluation criteria for the adjusted process parameters: Target achievement status: When the real-time process parameter deviation in the real-time process parameter deviation mapping matrix after adjustment and remains stable, the adjustment is successful and enters the stable monitoring mode; Continuous deviation state: When the adjusted process parameters are still in the early warning state area or the abnormal state area, the dynamic adjustment strategy is re-executed, and the optimal recovery path is optimized. Abnormal locking state: When the process parameters still do not meet the standards after continuous adjustment exceeds the set number of rounds, enter the abnormal locking state, trigger manual intervention and output an analysis report on the adjustment failure.

Images