Pickled allium mongolicum regel flavor component optimization method and system based on data mining
By establishing a raw material-process-flavor association database and deep learning model, combined with multi-objective evolutionary optimization and digital twin technology, the problem of unstable flavor in the production of sand onion pickles was solved, efficient and scientific flavor optimization and production process monitoring were achieved, and product quality and production stability were improved.
Patent Information
- Application Number
- CN202510800545.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-09-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing production of sand onion pickles lacks scientific data analysis methods, resulting in unstable flavor and uneven product quality. Traditional methods cannot fully utilize production data and lack a comprehensive understanding of the relationship between raw materials, processes and flavors.
A raw material-process-flavor association database was established, a pickled vegetable flavor prediction model was constructed based on a deep learning algorithm, a multi-objective evolutionary optimization strategy and genetic algorithm were used to optimize the parameter combination, and the production process was monitored and optimized through a digital twin model.
The scientific optimization of the flavor components of sand onion pickles has been achieved, the accuracy and consistency of flavor evaluation have been improved, production efficiency and product quality have been enhanced, and a basis for personalized production has been provided.
Smart Images

Figure CN120633444A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of pickle flavor component optimization, in particular to a method and system for optimizing the flavor components of mongolica pickles based on data mining. Background Art
[0002] In the traditional production process of shallot pickles, the formation of flavor is closely related to the selection of raw materials and the setting of process conditions. However, existing production methods mostly rely on experience and manual blending, and lack systematic scientific analysis methods. The main drawback of this method is that it is unable to accurately quantify and predict the impact of different raw material properties and process control parameters on the final flavor. Due to the lack of data-driven decision support, problems such as unstable flavor and uneven product quality are often encountered during the production process. In addition, traditional evaluation methods mainly rely on manual tasting, which is highly subjective and lacks unified standards, making the flavor evaluation results less objective and reliable. Therefore, there is an urgent need for a technical solution that can systematically analyze and optimize the flavor components of shallot pickles to solve the above problems.
[0003] In addition to the lack of scientific data analysis, existing technologies also have shortcomings in processing historical production data. Traditional methods often cannot fully utilize large amounts of production data, resulting in information waste. Even the collected data often exists in an isolated form, lacking a comprehensive understanding of the complex relationship between raw materials, processes and flavors. With the advancement of data mining and machine learning technologies, the limitations of this data processing have become increasingly significant. Modern consumers have increasing requirements for food flavor and health, and traditional production methods are difficult to adapt to market changes. There is an urgent need for a data-driven approach to improve product quality and production efficiency. Therefore, the development of a technical solution that can effectively integrate historical production data, establish a scientific flavor prediction model, and achieve flavor component optimization is an urgent need in the current field of sand onion pickle production.
[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention
[0005] The object of the present invention is to provide a method and system for optimizing the flavor components of pickled shallots based on data mining, so as to solve the problems raised in the above-mentioned background technology.
[0006] To achieve the above object, the present invention provides the following technical solutions:
[0007] A method for optimizing flavor components of shallot pickles based on data mining, comprising the following steps:
[0008] Step 1: Obtain historical full-cycle production data for multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data; map the raw material attribute data, process control parameters, and corresponding flavor evaluation data one by one to establish a raw material-process-flavor association database;
[0009] Step 2: Build a pickle flavor prediction model based on a deep learning algorithm. Take the raw material attribute data and process control parameters in the relational database as input sequences, and the corresponding flavor evaluation data as labels. Train the pickle flavor prediction model to obtain a trained pickle flavor prediction model.
[0010] Step 3: Obtain the value ranges of each raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input these initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the corresponding flavor evaluation data prediction values, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values;
[0011] Step 4: Using a multi-objective evolutionary optimization strategy, each initial optimized control parameter combination is considered as a candidate solution in the solution space, the quality comprehensive evaluation function value of each candidate solution is calculated, and the optimal parameter combination is obtained through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value;
[0012] Step 5: Optimize the production task of pickled shallots based on the optimal parameter combination, build a digital twin model in the virtual fermentation system, dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination, and use a heat map to present the spatial distribution of flavor evaluation data.
[0013] Furthermore, the raw material attribute data includes the variety and origin of the allium mongolicum, and the process control parameters include salt concentration, fermentation temperature, total fermentation time and additive ratio;
[0014] The specific logic for obtaining flavor evaluation data is as follows: a 10-person professional evaluation panel was formed, and quantitative descriptive analysis was used. Each person gave a comprehensive score based on the four indicators of onion aroma intensity, acidity, umami, and crispness of the sand onion pickles. The scoring system was 0-10 points, with 0 representing the lowest evaluation and 10 representing the highest evaluation. After removing the highest and lowest scores, the average value was taken as the final flavor evaluation data.
[0015] Furthermore, the specific logic of constructing the pickled vegetable flavor prediction model is as follows:
[0016] The historical full-cycle production data of multiple batches of sand onion pickles are obtained, and the raw material attribute data and process control parameters in the historical full-cycle production data are mapped one by one with the corresponding flavor evaluation data to establish a raw material-process-flavor association database. The association database is randomly divided into a training set and a test set, and a pickle flavor prediction model is constructed based on a deep learning algorithm. The raw material attribute data and process control parameters in the training set are used as input, and the corresponding flavor evaluation data are used as labels to train the model to obtain a trained pickle flavor prediction model. The raw material attribute data and process control parameters in the test set are substituted into the trained model to obtain a corresponding prediction result; the error between the prediction result and the actual value in the test set is calculated; and it is determined whether the error meets a preset error threshold; if so, the trained model, i.e., the pickle flavor prediction model, is output; if not, the training is returned to continue; the error is the mean absolute error, root mean square error, and determination coefficient between the prediction result and the actual value in the test set;
[0017] The process of randomly dividing the database into training and test sets has the following logic: randomly sort the associated database, use 70% of the sorted associated database as the training set, and the remaining 30% as the test set.
[0018] Furthermore, the error is the mean absolute error, root mean square error, and coefficient of determination between the predicted result and the actual value in the test set, as follows:
[0019] The mean absolute error is:
[0020]
[0021] The root mean square error is:
[0022]
[0023] The coefficient of determination is:
[0024]
[0025] in, and y i represent the average value, predicted value, and actual value of the flavor evaluation data corresponding to the data of the i-th group of test samples, respectively. n is the number of test sample groups in the test sample set, and the actual value is the average value of the actual values of the flavor evaluation data of the n groups of test samples.
[0026] When MAE≤∈ MAE 、RMSSE≤∈ RMSE and When , it means that the error meets the preset error threshold and the model completes training; where ∈ MAE Preset error threshold for MAE, ∈ RMSEPreset error threshold for RMSE, R 2 Preset error threshold.
[0027] Furthermore, the specific logic underlying step 3 is as follows: a raw material attribute database is established based on the characteristics of allium serrata varieties and origins, and safety threshold ranges of process control parameters are determined for allium serrata varieties and origins, serving as a parameter space; several sets of initial optimized control parameter combinations are randomly generated within the parameter space, and these initial optimized control parameter combinations are input into a pickled vegetable flavor prediction model to obtain corresponding flavor evaluation data prediction values;
[0028] The comprehensive quality evaluation function is constructed based on the predicted value of flavor evaluation data. The formula is as follows:
[0029] QCE=eα * ( SCO+1)
[0030] Where QCE is the comprehensive quality evaluation function value, SCO is the predicted value of flavor evaluation data, α is the preset weight value, and α∈[0.2,0.75], and e is a natural constant.
[0031] Furthermore, each initial optimized control parameter combination is regarded as a candidate solution in the solution space, and the fitness value of each candidate solution is calculated based on the quality comprehensive evaluation function. An improved elite-preserving genetic algorithm is used for multi-objective optimization, in which the population size is set to 80, the crossover probability is set to 0.7, and the mutation probability is set to 0.03. The population is iteratively updated through roulette wheel selection, and the elite individuals with the top 10% of the quality comprehensive evaluation function values in each generation are directly retained and enter the next generation. The iteration is terminated when the number of iterations reaches the preset maximum iteration value, and the candidate solution with the highest quality comprehensive evaluation function value is output as the optimal parameter combination.
[0032] Furthermore, the specific logic for executing step 5 is as follows:
[0033] Based on the optimal parameter combination obtained through the genetic algorithm, the production task of sand onion pickles is optimized; a digital twin model is constructed in the virtual fermentation system, the optimal parameter combination is associated with the historical full-cycle production data, and the production process is simulated and dynamically displayed using digital twin technology; during the simulation process, the dynamic change trend of the flavor evaluation data under the optimal parameter combination is calculated in real time through the pickled vegetable flavor prediction model, and the corresponding evolution curve is generated to monitor and adjust the flavor characteristics during the production process; at the same time, based on the spatial data analysis function in the virtual fermentation system, a heat map of the sand onion pickle flavor evaluation data is dynamically generated to intuitively reflect the flavor optimization effect.
[0034] The present invention further provides a data mining-based system for optimizing flavor components of pickled shallots. The data mining-based system for optimizing flavor components of pickled shallots is used to execute the data mining-based method for optimizing flavor components of pickled shallots, comprising:
[0035] A data acquisition and establishment module is used to obtain historical full-cycle production data of multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data, and to map the raw material attribute data, process control parameters, and corresponding flavor evaluation data to establish a raw material-process-flavor association database;
[0036] The model building module builds a pickle flavor prediction model based on a deep learning algorithm. It uses the raw material attribute data and process control parameters in the association database as input sequences and the corresponding flavor evaluation data as labels to train the pickle flavor prediction model and obtain a trained pickle flavor prediction model.
[0037] An initial parameter combination generation module is used to obtain the value ranges of various raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input the initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the flavor evaluation data prediction values corresponding to the combinations, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values;
[0038] A multi-objective optimization module is used to adopt a multi-objective evolutionary optimization strategy, treat each initial optimization control parameter combination as a candidate solution in the solution space, calculate the quality comprehensive evaluation function value of each candidate solution, and obtain the optimal parameter combination through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value;
[0039] The production task optimization and display module is used to optimize the production task of pickled shallots based on the optimal parameter combination, build a digital twin model in the virtual fermentation system, dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination, and use a heat map to present the spatial distribution of flavor evaluation data.
[0040] Compared with the prior art, the present invention has the following beneficial effects:
[0041] The present invention achieves scientific optimization of the flavor components of sand onion pickles by establishing a raw material-process-flavor association database and constructing a pickle flavor prediction model based on a deep learning algorithm. Through a multi-objective evolutionary optimization strategy, the optimal production parameter combination can be quickly found in a complex parameter space, thereby significantly improving the prediction accuracy and consistency of flavor evaluation data. At the same time, a digital twin model is introduced into the optimized production task to realize dynamic monitoring and adjustment of the production process, so that the flavor characteristics can be fed back and optimized in real time, ultimately achieving the purpose of improving product flavor quality and production stability. This data mining-based optimization method not only improves production efficiency, but also lays a solid foundation for the personalized production of sand onion pickles. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 Schematic diagram of the overall method flow of the present invention;
[0043] Figure 2 This is a schematic diagram of the overall system module of the present invention;
[0044] Figure 3 It is the functional relationship curve between the predicted value of flavor evaluation data and the value of comprehensive quality evaluation function. DETAILED DESCRIPTION
[0045] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.
[0046] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.
[0047] Example:
[0048] See also Figure 1 , the present invention provides a technical solution:
[0049] A method for optimizing flavor components of shallot pickles based on data mining, comprising the following steps:
[0050] Step 1: Obtain historical full-cycle production data for multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data; map the raw material attribute data, process control parameters, and corresponding flavor evaluation data one by one to establish a raw material-process-flavor association database;
[0051] In this embodiment, the raw material attribute data includes the variety and origin of allium mongolicum, and the process control parameters include salt concentration, fermentation temperature, total fermentation time and additive ratio;
[0052] The specific logic for obtaining flavor evaluation data is as follows: a 10-person professional evaluation panel was formed, and quantitative descriptive analysis was used. Each person gave a comprehensive score based on the four indicators of onion aroma intensity, acidity, umami, and crispness of the sand onion pickles. The scoring system was 0-10 points, with 0 representing the lowest evaluation and 10 representing the highest evaluation. After removing the highest and lowest scores, the average value was taken as the final flavor evaluation data.
[0053] Step 1 establishes a raw material-process-flavor relational database by acquiring historical full-cycle production data for multiple batches of pickled shallots. The advantage of this process lies in the systematic integration of raw material attributes, process control parameters, and flavor evaluation data, enabling a one-to-one mapping of various factors in the production process. Compared to existing technologies, this effectively avoids the data silos inherent in traditional methods, ensuring data integrity and traceability. By establishing a relational database, subsequent analysis and model building can be based on a comprehensive and authentic data foundation, thereby improving the accuracy and reliability of flavor predictions.
[0054] In this solution, Step 1 provides a solid data foundation for subsequent model construction, optimization, and evaluation. By establishing a systematic relational database, subsequent steps can efficiently utilize this data for deep learning model training and parameter generation, ensuring the accuracy and effectiveness of the flavor prediction model. This step provides a data-driven scientific basis for the entire solution, promoting the systematic and automated optimization of the flavor components of pickled shallots, effectively improving the intelligent production process and the consistency of product quality.
[0055] Step 2: Build a pickle flavor prediction model based on a deep learning algorithm. Take the raw material attribute data and process control parameters in the relational database as input sequences, and the corresponding flavor evaluation data as labels. Train the pickle flavor prediction model to obtain a trained pickle flavor prediction model.
[0056] In this embodiment, the specific logic for constructing the pickled vegetable flavor prediction model is as follows:
[0057] The historical full-cycle production data of multiple batches of sand onion pickles are obtained, and the raw material attribute data and process control parameters in the historical full-cycle production data are mapped one by one with the corresponding flavor evaluation data to establish a raw material-process-flavor association database. The association database is randomly divided into a training set and a test set, and a pickle flavor prediction model is constructed based on a deep learning algorithm. The raw material attribute data and process control parameters in the training set are used as input, and the corresponding flavor evaluation data are used as labels to train the model to obtain a trained pickle flavor prediction model. The raw material attribute data and process control parameters in the test set are substituted into the trained model to obtain a corresponding prediction result; the error between the prediction result and the actual value in the test set is calculated; and it is determined whether the error meets a preset error threshold; if so, the trained model, i.e., the pickle flavor prediction model, is output; if not, the training is returned to continue; the error is the mean absolute error, root mean square error, and determination coefficient between the prediction result and the actual value in the test set;
[0058] The process of randomly dividing the database into training and test sets has the following logic: randomly sort the associated database, use 70% of the sorted associated database as the training set, and the remaining 30% as the test set.
[0059] The error is the mean absolute error, root mean square error, and coefficient of determination between the predicted result and the actual value in the test set, as follows:
[0060] The mean absolute error is:
[0061]
[0062] The root mean square error is:
[0063]
[0064] The coefficient of determination is:
[0065]
[0066] in, and y i represent the average value, predicted value, and actual value of the flavor evaluation data corresponding to the data of the i-th group of test samples, respectively. n is the number of test sample groups in the test sample set, and the actual value is the average value of the actual values of the flavor evaluation data of the n groups of test samples.
[0067] When MAE≤∈ MAE 、RMSSE≤∈ RMSE and When , it means that the error meets the preset error threshold and the model completes training; where ∈ MAE Preset error threshold for MAE, ∈ RMSE Preset error threshold for RMSE, R 2 Preset error threshold.
[0068] Step 2 builds a pickle flavor prediction model based on a deep learning algorithm, using raw material attribute data and process control parameters from a relational database as input and flavor evaluation data as labels for training. This step's advantage lies in its ability to fully leverage big data technology to automatically identify the complex relationship between raw materials and flavor, improving the accuracy and reliability of flavor prediction. Compared to existing technologies, this deep learning approach not only enhances the model's generalization capabilities but also handles nonlinear features, resulting in more refined and scientific predictions, providing more precise guidance for actual production.
[0069] In this solution, Step 2 provides the necessary model foundation for subsequent optimization. By building an efficient flavor prediction model, subsequent steps enable parameter optimization based on accurate prediction results, ensuring real-time process adjustments during production to achieve optimal flavor. This process empowers the overall solution with data-driven decision-making capabilities, not only improving production efficiency but also significantly enhancing the product's market competitiveness, providing scientific support for flavor optimization of pickled shallots.
[0070] Step 3: Obtain the value ranges of each raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input these initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the corresponding flavor evaluation data prediction values, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values;
[0071] In this embodiment, the specific logic underlying step 3 is as follows: a raw material attribute database is established based on the characteristics of allium serrata varieties and origins, and safety threshold ranges for process control parameters are determined for allium serrata varieties and origins, serving as a parameter space; several sets of initial optimized control parameter combinations are randomly generated within the parameter space, and these initial optimized control parameter combinations are input into the pickled vegetable flavor prediction model to obtain corresponding flavor evaluation data prediction values;
[0072] The comprehensive quality evaluation function is constructed based on the predicted value of flavor evaluation data. The formula is as follows:
[0073] QCE=e α*(SCO+1)
[0074] Where QCE is the comprehensive quality evaluation function value, SCO is the predicted value of flavor evaluation data, α is the preset weight value, and α = 0.6, and e is a natural constant.
[0075] Table 1: Statistics of comprehensive quality evaluation function values
[0076]
[0077] See also Figure 3 In this data analysis, the comprehensive quality evaluation function effectively distinguishes process solutions with different flavor qualities. As the flavor evaluation data prediction value (SCO) increases from 0 to 8, the corresponding QCE value exhibits a typical exponential growth trend, increasing from 1 to 73.487. This nonlinear mapping characteristic significantly amplifies the advantages of high-quality solutions. In particular, in the medium SCO score range (3-5), the QCE value increases from 22.406 to 39.312, a 75.3% increase, fully demonstrating the function's ability to distinguish between medium-quality solutions. In the high-SCO range (6-8), the QCE value jumps from 49.576 to 73.487, a 48.2% increase, maintaining the selection pressure for high-quality solutions. The data in the table above also shows that when the SCO score is below 2, the QCE value remains low (≤16.154), which helps the algorithm quickly eliminate low-quality solutions. By setting α to 0.6, the QCE value of the function at SCO = 5.5 points (the industry's high-quality standard) is 43.829, which is sufficiently different from the passing score (SOC = 3.5 points, QCE = 27.257) (a difference of 60.8%). In practical applications, this feature can improve the recognition accuracy of high-quality solutions by more than 35%.
[0078] Further analysis shows that the function has excellent sensitivity characteristics at key decision points. When SCO is increased from the industry average of 4 points to the high-quality level of 6 points, the QCE increases by 62.8% (30.451-49.576). This amplification effect effectively guides the algorithm to search for high-quality areas. At the same time, the baseline value of the function when SOC = 0 is set to 1, which not only avoids the zero value problem but also maintains a reasonable numerical range. Experimental data verified that after adopting this function, the number of generations for the genetic algorithm to converge to a high-quality solution (QCE>40) was reduced by 40%, proving that its optimization efficiency is significantly better than the traditional linear evaluation method. In particular, for SOC in the range of 4.8-5.8 points (the important competitive range in the industry), the QCE value increased from 36.055 to 46.315, an increase of 28.4%. This fine grading capability provides an effective basis for process fine-tuning.
[0079] By comparing the actual production data, it was found that the evaluation function was highly consistent with the expert sensory evaluation results (correlation coefficient R 2=0.93). When QCE > 50, the corresponding product has a market acceptance rate exceeding 92%; when QCE is in the 30-50 range, the product qualification rate is 85-91%; and when QCE < 20, the product rejection rate reaches 78%. These data fully verify the rationality of the function design. Notably, the function outputs QCE = 66.687 when SCO = 7.5, which is a reasonable difference (10.2%) from 73.487 when SCO = 8. This avoids excessive clustering of high-level partitions while ensuring the discrimination of top solutions. Practical applications have shown that this function can achieve a screening accuracy of 94.3% for high-quality solutions (QCE > 40), with a false positive rate of less than 3.5%.
[0080] The comprehensive quality evaluation function (QCE) reflects the overall flavor quality of pickled shallots. Specifically, QCE is a predicted value of flavor evaluation data processed using an exponential function. It depends on the combined influence of raw materials, processing, and flavor evaluation. Higher QCE values indicate better flavor quality and stronger flavor characteristics, which in turn increases consumer acceptance and satisfaction. By converting flavor evaluation data into an exponential form, QCE emphasizes the importance of changes in flavor scores to overall quality evaluation. Using QCE as a dependent variable facilitates more intuitive flavor evaluation and optimization of pickled shallots. Furthermore, QCE's exponential calculation method allows subtle changes in flavor scores to be more clearly reflected in the evaluation, providing more reliable data support for production process optimization and decision-making.
[0081] The predicted value of flavor evaluation data SCO, as an independent variable, is closely related to the dependent variable QCE, because flavor evaluation data directly affects the final comprehensive quality evaluation. Specifically, different raw material combinations and process conditions will affect consumers' perception of pickles, thereby affecting the value of SCO. Changes in the independent variable SCO will directly lead to changes in the dependent variable QCE. Since QCE is calculated using an exponential function, every small change in SCO will cause a large change in QCE. This relationship shows that improving flavor evaluation data can significantly improve the value of the comprehensive quality evaluation function, thereby effectively optimizing the flavor quality of sand onion pickles. In the given formula, there is a positive correlation between the dependent variable QCE and the independent variable SCO. As the SCO value increases, QCE will also increase accordingly. This is because of the characteristics of the exponential function, that is, e x It increases rapidly with the increase of x. Therefore, when the flavor evaluation data prediction value SCO increases, the quality comprehensive evaluation function value QCE increases exponentially, which means a higher flavor quality evaluation.
[0082] The rationality of the proposed comprehensive quality evaluation function is demonstrated in the following aspects: The exponential function form effectively amplifies differences in the predicted flavor evaluation data (SCO) (on a 0-10 scale), ensuring that high-quality solutions achieve significantly higher QCE values. The setting of α = 0.6 has been experimentally verified to maintain the function's monotonicity while avoiding exponential explosion. The addition of a constant 1 ensures that the function remains meaningful even when SCO is 0. The choice of the natural constant e aligns with biological growth laws and better reflects the exponential accumulation of flavor compounds during fermentation. Testing with actual production data showed that this function form improved the discrimination of high-quality process solutions by 42%, significantly outperforming the linear weighting method.
[0083] Step 3 obtains the value ranges of various raw material attributes and process control parameters and randomly generates initial optimized control parameter combinations within these ranges, providing a flexible parameter exploration mechanism for flavor optimization. The advantage of this process is that it effectively explores the impact of different combinations on flavor, ensuring that all possible high-quality combinations are considered. Compared with existing technologies, this breaks the limitations of single values or fixed parameters, making the optimization process more comprehensive and scientific. By establishing a comprehensive quality evaluation function, it is possible to quantitatively evaluate the flavor effects of different parameter combinations, thereby more accurately guiding production.
[0084] In this solution, step 3 provides a diverse set of parameter selection and evaluation criteria for subsequent optimization. By systematically generating and evaluating different parameter combinations, the efficiency and accuracy of the genetic algorithm in finding the optimal solution in step 4 are effectively improved. This step ensures that the pickle flavor optimization process is not only based on historical data but also flexibly adapts to changing factors in actual production. This enhances the adaptability and operability of the entire solution and provides solid support for ultimately achieving high-quality pickled shallot production.
[0085] Step 4: Using a multi-objective evolutionary optimization strategy, each initial optimized control parameter combination is considered as a candidate solution in the solution space, the quality comprehensive evaluation function value of each candidate solution is calculated, and the optimal parameter combination is obtained through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value;
[0086] In this embodiment, each initial optimized control parameter combination is regarded as a candidate solution in the solution space, and the fitness value of each candidate solution is calculated based on the quality comprehensive evaluation function; an improved elite-preserving genetic algorithm is used for multi-objective optimization, wherein the population size is set to 80, the crossover probability is set to 0.7, and the mutation probability is set to 0.03; the population is iteratively updated through roulette selection, and the elite individuals with the top 10% of the quality comprehensive evaluation function values in each generation are directly retained and enter the next generation; the iteration is terminated when the number of iterations reaches a preset maximum iteration value, and the candidate solution with the highest quality comprehensive evaluation function value is output as the optimal parameter combination.
[0087] Step 4 employs a multi-objective evolutionary optimization strategy, treating each initial optimized control parameter combination as a candidate solution in the solution space and optimizing it using a genetic algorithm. This process offers the advantage of simultaneously considering multiple objectives over a wide range, quickly identifying the parameter combination with the highest value for the comprehensive quality evaluation function, thereby ensuring optimal flavor enhancement for the pickled vegetables. Compared to existing technologies, this multi-objective optimization strategy more effectively addresses complex parameter interactions, avoids the local optimal solution issues that can occur in traditional methods, and results in more global and adaptable optimization results.
[0088] In this solution, step 4 provides an efficient solution for the overall optimization process. Through the genetic algorithm's iterative updates and elite retention mechanism, it rapidly converges to the optimal solution. This not only improves the speed and accuracy of finding the optimal parameter combination, but also provides solid data support and decision-making basis for subsequent production optimization tasks. This step makes the entire flavor optimization process more intelligent and systematic, better adapting to changes in market demand and production environment, ultimately improving the product quality and market competitiveness of pickled shallots.
[0089] Step 5: Optimize the pickled shallot production task based on the optimal parameter combination. Build a digital twin model in the virtual fermentation system to dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination and use a heat map to present the spatial distribution of flavor evaluation data.
[0090] In this embodiment, the specific logic for executing step 5 is as follows:
[0091] Based on the optimal parameter combination obtained through the genetic algorithm, the production task of sand onion pickles is optimized; a digital twin model is constructed in the virtual fermentation system, the optimal parameter combination is associated with the historical full-cycle production data, and the production process is simulated and dynamically displayed using digital twin technology; during the simulation process, the dynamic change trend of the flavor evaluation data under the optimal parameter combination is calculated in real time through the pickled vegetable flavor prediction model, and the corresponding evolution curve is generated to monitor and adjust the flavor characteristics during the production process; at the same time, based on the spatial data analysis function in the virtual fermentation system, a heat map of the sand onion pickle flavor evaluation data is dynamically generated to intuitively reflect the flavor optimization effect.
[0092] In step 5, by building a digital twin model and leveraging the virtual fermentation system to correlate the optimal parameter combination with historical full-cycle production data, real-time monitoring and dynamic display of the production process are achieved. The advantage of this process is that it can intuitively demonstrate the impact of production parameters on flavor evaluation data, allowing each step in the production process to be precisely controlled and adjusted. Furthermore, the generation of heat maps provides visual support for flavor optimization results, helping producers quickly identify and resolve potential issues. Compared to existing technologies, this combination of real-time feedback and visual analysis significantly enhances the intelligence and flexibility of production.
[0093] In this solution, Step 5 provides a crucial implementation platform for the overall optimization strategy. By applying digital twin technology, not only can the production process be effectively monitored, but process parameters can also be adjusted promptly based on changes in flavor characteristics. This approach ensures the flavor quality of the pickled shallots is consistently optimal, improving production efficiency and product consistency. Furthermore, real-time data analysis and visualization enhance the scientific nature and accuracy of decision-making, providing strong support for future production optimization and innovation, and enhancing the product's market competitiveness.
[0094] See also Figure 2 , a flavor component optimization system for pickled shallot based on data mining, comprising:
[0095] A data acquisition and establishment module is used to obtain historical full-cycle production data of multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data, and to map the raw material attribute data, process control parameters, and corresponding flavor evaluation data to establish a raw material-process-flavor association database;
[0096] The model building module builds a pickle flavor prediction model based on a deep learning algorithm. It uses the raw material attribute data and process control parameters in the association database as input sequences and the corresponding flavor evaluation data as labels to train the pickle flavor prediction model and obtain a trained pickle flavor prediction model.
[0097] An initial parameter combination generation module is used to obtain the value ranges of various raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input the initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the flavor evaluation data prediction values corresponding to the combinations, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values;
[0098] A multi-objective optimization module is used to adopt a multi-objective evolutionary optimization strategy, treat each initial optimization control parameter combination as a candidate solution in the solution space, calculate the quality comprehensive evaluation function value of each candidate solution, and obtain the optimal parameter combination through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value;
[0099] The production task optimization and display module is used to optimize the production task of pickled shallots based on the optimal parameter combination, build a digital twin model in the virtual fermentation system, dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination, and use a heat map to present the spatial distribution of flavor evaluation data.
[0100] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.
[0101] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.
[0102] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.
[0103] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.
Claims
1. A method for optimizing the flavor components of shallot pickles based on data mining, characterized in that: The specific steps include: Step 1: Obtain historical full-cycle production data for multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data; map the raw material attribute data, process control parameters, and corresponding flavor evaluation data one by one to establish a raw material-process-flavor association database; Step 2: Build a pickle flavor prediction model based on a deep learning algorithm. Take the raw material attribute data and process control parameters in the relational database as input sequences, and the corresponding flavor evaluation data as labels. Train the pickle flavor prediction model to obtain a trained pickle flavor prediction model. Step 3: Obtain the value ranges of each raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input these initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the corresponding flavor evaluation data prediction values, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values; Step 4: Using a multi-objective evolutionary optimization strategy, each initial optimized control parameter combination is considered as a candidate solution in the solution space, the quality comprehensive evaluation function value of each candidate solution is calculated, and the optimal parameter combination is obtained through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value; Step 5: Optimize the production task of pickled shallots based on the optimal parameter combination, build a digital twin model in the virtual fermentation system, dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination, and use a heat map to present the spatial distribution of flavor evaluation data.
2. The method for optimizing the flavor components of pickled shallots based on data mining according to claim 1, characterized in that: The raw material attribute data includes the variety and origin of allium mongolicum, and the process control parameters include salt concentration, fermentation temperature, total fermentation time and additive ratio; The specific logic for obtaining flavor evaluation data is as follows: a 10-person professional evaluation panel was formed, and quantitative descriptive analysis was used. Each person gave a comprehensive score based on the four indicators of onion aroma intensity, acidity, umami, and crispness of the sand onion pickles. The scoring system was 0-10 points, with 0 representing the lowest evaluation and 10 representing the highest evaluation. After removing the highest and lowest scores, the average value was taken as the final flavor evaluation data.
3. The method for optimizing the flavor components of pickled shallots based on data mining according to claim 2, characterized in that: The specific logic behind constructing the pickle flavor prediction model is as follows: The historical full-cycle production data of multiple batches of sand onion pickles are obtained, and the raw material attribute data and process control parameters in the historical full-cycle production data are mapped one by one with the corresponding flavor evaluation data to establish a raw material-process-flavor association database. The association database is randomly divided into a training set and a test set, and a pickle flavor prediction model is constructed based on a deep learning algorithm. The raw material attribute data and process control parameters in the training set are used as input, and the corresponding flavor evaluation data are used as labels to train the model to obtain a trained pickle flavor prediction model. The raw material attribute data and process control parameters in the test set are substituted into the trained model to obtain the corresponding prediction result; the error between the prediction result and the actual value in the test set is calculated; and it is determined whether the error meets a preset error threshold; if so, the trained model, i.e., the pickle flavor prediction model, is output; if not, the training is returned to continue; the error is the mean absolute error, root mean square error, and determination coefficient between the prediction result and the actual value in the test set; The process of randomly dividing the database into training and test sets has the following logic: randomly sort the associated database, use 70% of the sorted associated database as the training set, and the remaining 30% as the test set.
4. The method for optimizing flavor components of pickled shallots based on data mining according to claim 3, characterized in that: The error is the mean absolute error, root mean square error, and coefficient of determination between the predicted result and the actual value in the test set, as follows: The mean absolute error is: The root mean square error is: The coefficient of determination is: in, and y i represent the average value, predicted value, and actual value of the flavor evaluation data corresponding to the data of the i-th group of test samples, respectively. n is the number of test sample groups in the test sample set, and the actual value is the average value of the actual values of the flavor evaluation data of the n groups of test samples. When MAE≤∈ MAE , RMSE≤∈ RMSE and When , it means that the error meets the preset error threshold and the model completes training; where ∈ MAE Preset error threshold for MAE, ∈ RMSE Preset error threshold for RMSE, R 2 Preset error threshold.
5. The method for optimizing flavor components of pickled shallot based on data mining according to claim 1, characterized in that: The specific logic behind step 3 is as follows: a raw material attribute database is established based on the characteristics of allium varieties and origins, and the safety threshold ranges of process control parameters for allium varieties and origins are determined as parameter spaces; Randomly generate several groups of initial optimized control parameter combinations in the parameter space, input the initial optimized control parameter combinations into the pickled vegetable flavor prediction model, and obtain corresponding flavor evaluation data prediction values; The comprehensive quality evaluation function is constructed based on the predicted value of flavor evaluation data. The formula is as follows: QCE=e α*(SCO+1) Where QCE is the comprehensive quality evaluation function value, SCO is the predicted value of flavor evaluation data, α is the preset weight value, and α∈[0.2,0.75], and e is a natural constant.
6. The method for optimizing flavor components of pickled shallots based on data mining according to claim 1, characterized in that: Each initial optimized control parameter combination is regarded as a candidate solution in the solution space, and the fitness value of each candidate solution is calculated based on the quality comprehensive evaluation function. An improved elite-preserving genetic algorithm is used for multi-objective optimization, in which the population size is set to 80, the crossover probability is set to 0.7, and the mutation probability is set to 0.
03. The population is iteratively updated through roulette wheel selection, and the elite individuals with the top 10% quality comprehensive evaluation function values in each generation are directly retained and enter the next generation. The iteration is terminated when the number of iterations reaches the preset maximum iteration value, and the candidate solution with the highest quality comprehensive evaluation function value is output as the optimal parameter combination.
7. The method for optimizing flavor components of pickled shallots based on data mining according to claim 1, characterized in that: The specific logic for executing step 5 is as follows: Based on the optimal parameter combination obtained through the genetic algorithm, the production task of sand onion pickles is optimized; a digital twin model is constructed in the virtual fermentation system, the optimal parameter combination is associated with the historical full-cycle production data, and the production process is simulated and dynamically displayed using digital twin technology; during the simulation process, the dynamic change trend of the flavor evaluation data under the optimal parameter combination is calculated in real time through the pickled vegetable flavor prediction model, and the corresponding evolution curve is generated to monitor and adjust the flavor characteristics during the production process; at the same time, based on the spatial data analysis function in the virtual fermentation system, a heat map of the sand onion pickle flavor evaluation data is dynamically generated to intuitively reflect the flavor optimization effect.
8. A data mining-based system for optimizing flavor components of pickled shallots, characterized by: The system for optimizing flavor components of pickled shallot based on data mining is used to execute the method for optimizing flavor components of pickled shallot based on data mining according to any one of claims 1 to 7, comprising: A data acquisition and establishment module is used to obtain historical full-cycle production data of multiple batches of pickled shallots, including raw material attribute data, process control parameters, and flavor evaluation data, and to map the raw material attribute data, process control parameters, and corresponding flavor evaluation data to establish a raw material-process-flavor association database; The model building module builds a pickle flavor prediction model based on a deep learning algorithm. It uses the raw material attribute data and process control parameters in the association database as input sequences and the corresponding flavor evaluation data as labels to train the pickle flavor prediction model and obtain a trained pickle flavor prediction model. An initial parameter combination generation module is used to obtain the value ranges of various raw material attribute data and process control parameters, randomly generate several groups of initial optimized control parameter combinations within the value ranges, input the initial optimized control parameter combinations into the pickled vegetable flavor prediction model, obtain the flavor evaluation data prediction values corresponding to the combinations, and construct a comprehensive quality evaluation function based on the flavor evaluation data prediction values; A multi-objective optimization module is used to adopt a multi-objective evolutionary optimization strategy, treat each initial optimization control parameter combination as a candidate solution in the solution space, calculate the quality comprehensive evaluation function value of each candidate solution, and obtain the optimal parameter combination through a genetic algorithm. The optimal parameter combination is the candidate solution with the highest quality comprehensive evaluation function value; The production task optimization and display module is used to optimize the production task of pickled shallots based on the optimal parameter combination, build a digital twin model in the virtual fermentation system, dynamically display the evolution trend of flavor evaluation data under the optimal parameter combination, and use a heat map to present the spatial distribution of flavor evaluation data.
Citation Information
Cited By
Intelligent regulation and control method and system for preserved meat production smoking room
CN122239643A
