Method and system for whole-process detection of yogurt strain activity and metabolic products

Abstract

The application discloses a whole-process detection method and system for yogurt strain activity and metabolic products, and relates to the technical field of intelligent analysis and detection.The method comprises the following steps: performing whole-process detection on a yogurt fermentation workshop, obtaining a strain activity and metabolic product detection matrix, performing abnormal fitting detection, establishing an activity metabolic abnormality matrix, performing fermentation accident prediction and tracing based on the matrix, adjusting fermentation decisions and generating an initial fermentation adjustment group, performing multi-party optimization using a strain activity evaluation and metabolic quality evaluation model, and generating final fermentation adjustment instructions.The application solves the technical problems of the prior art, such as the lack of real-time monitoring and precise control of the fermentation process, the instability of the fermentation process, the large fluctuation of product quality, and the low production efficiency, and achieves the technical effects of efficient optimization and stable control of the yogurt fermentation process, and the improvement of product quality and production efficiency through whole-process detection and precise control.

Description

Technical Field

[0001] This invention relates to the field of intelligent analysis and detection technology, specifically to a method and system for the whole-process detection of yogurt strain activity and metabolites. Background Technology

[0002] In the yogurt production industry, controlling the fermentation process is crucial for product quality and production efficiency. Traditional fermentation processes rely heavily on experience, lacking real-time monitoring and precise control of changes in microbial activity and metabolites during fermentation. This results in poor fermentation process stability, large fluctuations in product quality, and difficulty in effectively preventing fermentation accidents, thus impacting production efficiency and economic benefits. Summary of the Invention

[0003] This application provides a method and system for the whole-process detection of yogurt strain activity and metabolites, which is used to solve the technical problems of the lack of real-time monitoring and precise control of the fermentation process in the existing technology, resulting in unstable fermentation process, large fluctuations in product quality and low production efficiency.

[0004] The first aspect of this application provides a method for the full-process detection of yogurt strain activity and metabolites. The method includes: performing full-process detection on a yogurt fermentation workshop to obtain a strain activity detection matrix and a metabolite detection matrix corresponding to the current fermentation decision; performing anomaly fitting detection on the strain activity detection matrix and the metabolite detection matrix based on the current fermentation decision to establish an activity-metabolic anomaly matrix; performing multivariate fermentation accident prediction and tracing on the yogurt fermentation workshop based on the activity-metabolic anomaly matrix to obtain a fermentation accident prediction and tracing map; performing compensation and adjustment on the current fermentation decision based on the fermentation accident prediction and tracing map to obtain an initial fermentation regulation group; performing multivariate optimization on the initial fermentation regulation group based on a strain activity evaluation model and a metabolic quality evaluation model to establish a fermentation regulation multivariate optimization domain; and guiding the initial fermentation regulation group to perform reproduction optimization based on the strain activity evaluation model and the metabolic quality evaluation model, according to the fermentation regulation multivariate optimization domain, to generate fermentation regulation instructions.

[0005] The second aspect of this application provides a full-process detection system for yogurt strain activity and metabolites. The system includes: a full-process detection module for performing full-process detection on a yogurt fermentation workshop to obtain a strain activity detection matrix and a metabolite detection matrix corresponding to the current fermentation decision; an anomaly fitting detection module for performing anomaly fitting detection on the strain activity detection matrix and the metabolite detection matrix based on the current fermentation decision to establish an activity-metabolic anomaly matrix; an accident prediction and tracing module for performing multi-factor fermentation accident prediction and tracing on the yogurt fermentation workshop based on the activity-metabolic anomaly matrix to obtain a fermentation accident prediction and tracing map; a compensation and adjustment module for compensating and adjusting the current fermentation decision based on the fermentation accident prediction and tracing map to obtain an initial fermentation regulation group; a multi-factor optimization module for performing multi-factor optimization on the initial fermentation regulation group based on a strain activity evaluation model and a metabolic quality evaluation model to establish a fermentation regulation multi-factor optimization domain; and a reproduction optimization module for guiding the initial fermentation regulation group to perform reproduction optimization based on the strain activity evaluation model and the metabolic quality evaluation model, according to the fermentation regulation multi-factor optimization domain, to generate fermentation regulation instructions.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The method and system for full-process detection of yogurt strain activity and metabolites provided in this application belong to the field of intelligent analysis and detection technology. It obtains strain activity and metabolite data through full-process detection, performs anomaly fitting detection and fermentation accident prediction and tracing, and generates fermentation regulation instructions based on multi-factor optimization and propagation optimization to achieve precise control of the fermentation process. It solves the technical problems of existing technologies that lack real-time monitoring and precise control of the fermentation process, resulting in unstable fermentation process, large fluctuations in product quality and low production efficiency. It achieves the technical effect of efficient optimization and stable control of yogurt fermentation process through full-process detection and precise control, thereby improving product quality and production efficiency. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a schematic diagram of the entire process detection method for yogurt strain activity and metabolites provided in the embodiments of this application;

[0010] Figure 2This is a schematic diagram of the structure of a full-process detection system for yogurt strain activity and metabolites provided in an embodiment of this application.

[0011] Figure labeling: Full-process detection module 11, Anomaly fitting detection module 12, Accident prediction and tracing module 13, Compensation and adjustment module 14, Multi-party optimization module 15, Propagation optimization module 16. Detailed Implementation

[0012] This application provides a method and system for the whole-process detection of yogurt strain activity and metabolites, which is used to solve the technical problems of the lack of real-time monitoring and precise control of the fermentation process in the existing technology, resulting in unstable fermentation process, large fluctuations in product quality and low production efficiency.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a comprehensive detection method for the activity and metabolites of yogurt strains, which includes:

[0016] P10: Conduct full-process testing in the yogurt fermentation workshop to obtain the strain activity detection matrix and metabolite detection matrix corresponding to the current fermentation decision.

[0017] Furthermore, step P10 in this embodiment of the application also includes:

[0018] P11: Perform current strain activity testing on the yogurt fermentation workshop to obtain strain activity testing data; P12: Clean the strain activity testing data to generate the strain activity testing matrix; P13: Perform current metabolite testing on the yogurt fermentation workshop to obtain metabolite testing data; P14: Clean the metabolite testing data to obtain the metabolite testing matrix.

[0019] It should be understood that conducting full-process testing in the yogurt fermentation workshop to obtain the strain activity detection matrix and metabolite detection matrix corresponding to the current fermentation decision serves as the basis for achieving refined management and optimization of the yogurt fermentation process.

[0020] First, the activity of the microbial strains in the yogurt fermentation workshop is tested. Strain activity refers to the life activities exhibited by microorganisms during fermentation, including key indicators such as growth rate and metabolic capacity. Raw microbial activity data can be obtained by sampling and testing parameters such as growth rate, survival rate, metabolic rate, and extracellular enzyme activity of key functional strains (lactobacterium, Streptococcus thermophilus, Lactobacillus bulgaricus, etc.) at multiple points in the fermentation system. This data reflects the physiological state of the microorganisms at different stages of fermentation and is crucial for subsequent fermentation process control.

[0021] Subsequently, the obtained microbial activity detection data were cleaned, including outlier removal, missing value compensation, unit normalization, and time synchronization, thereby eliminating data deviations caused by instrument drift, sampling errors, or environmental interference. The cleaned multidimensional data was then structured to form a microbial activity detection matrix, where each row represents a time point or a specific fermentation condition, and each column corresponds to a specific indicator of microbial activity, intuitively reflecting the physiological state and activity trends of different strains at specific fermentation stages.

[0022] Next, the current metabolites in the yogurt fermentation workshop were tested. Metabolites refer to various chemical substances produced by the microorganisms during fermentation, such as lactic acid, acetic acid, and flavor compounds. These metabolites not only determine the taste and quality of yogurt but also reflect the metabolic state of the fermentation process. Therefore, the content of various metabolites during fermentation can be determined by analytical techniques such as high-performance liquid chromatography (HPLC) and gas chromatography-mass spectrometry (GC-MS).

[0023] Finally, the metabolite detection data underwent cleaning and standardization, including abnormal peak identification, background signal subtraction, signal intensity normalization, and multi-timepoint alignment, thereby generating a stable and reliable metabolite detection matrix. Similar to the strain activity detection matrix, the metabolite detection matrix is ​​also a structured data table recording the content of various metabolites at different time points and under different conditions. By cleaning and organizing this data, we can gain a clearer understanding of the changing trends of metabolites during fermentation, providing data support for subsequent fermentation process optimization.

[0024] P20: Based on the current fermentation decision, perform anomaly fitting detection on the strain activity detection matrix and the metabolite detection matrix to establish an activity metabolism anomaly matrix.

[0025] Furthermore, step P20 in this embodiment of the application also includes:

[0026] P21: Perform twin modeling based on the yogurt fermentation workshop to obtain a fermentation twin workshop; P22: Perform multi-feature fitting on the fermentation twin workshop based on the current fermentation decision to establish a strain activity prediction matrix and a metabolite prediction matrix; P23: Perform anomaly detection on the strain activity detection matrix based on the strain activity prediction matrix to obtain an activity anomaly detection matrix; P24: Perform anomaly detection on the metabolite detection matrix based on the metabolite prediction matrix to obtain a metabolic anomaly detection matrix; P25: Perform fusion processing on the activity anomaly detection matrix and the metabolic anomaly detection matrix to generate the activity-metabolic anomaly matrix.

[0027] Optionally, based on existing fermentation data and combined with current fermentation decision parameters, anomaly fitting and comparison analysis can be performed on the multidimensional detection matrix of strain activity and metabolites to establish an activity metabolism anomaly matrix that can be used for dynamic monitoring and early warning.

[0028] First, based on the physical model and historical data of the yogurt fermentation workshop, a twin model is created to obtain a fermentation twin workshop. Twin modeling is a method of constructing a virtual model corresponding to a physical entity using digital technology. The generated model can simulate the operating state of the fermentation workshop under different conditions. Specifically, by collecting actual production data, equipment operating parameters, current fermentation decision conditions, and detection data of microbial activity and metabolites from the fermentation workshop, a digital model of the fermentation twin workshop is constructed using machine learning algorithms or physical models. This model can predict changes in microbial activity and metabolites during fermentation based on the input process parameters.

[0029] Next, based on the current fermentation decisions, multi-feature fitting is performed on the fermentation twin workshop to establish a strain activity prediction matrix and a metabolite prediction matrix. Multi-feature fitting refers to modeling and predicting the fermentation process by considering multiple relevant features, such as temperature, pH, and inoculum size. For example, features closely related to changes in strain activity and metabolites are selected from a large amount of data in the fermentation workshop, and trained using a multivariate regression or neural network model. This enables the model to accurately predict changes in strain activity and metabolites based on the input feature values, generating strain activity prediction matrices and metabolite prediction matrices. These matrices record the predicted values ​​of strain activity and metabolites under different fermentation conditions.

[0030] Subsequently, based on the strain activity prediction matrix, anomaly detection was performed on the strain activity detection matrix to obtain an activity anomaly detection matrix. Anomaly detection refers to identifying data points that deviate from the normal range by comparing actual detection data with predicted data. Specifically, the actual data in the strain activity detection matrix and the predicted data in the strain activity prediction matrix were compared point by point; statistical analysis or residual analysis was used to identify significant differences between the actual and predicted data. These differences may indicate abnormalities in the fermentation process. For example, if the difference between the predicted activity and the actual detection results exceeds a certain threshold, it indicates problems such as microbial decline, exogenous contamination, or inadequate environmental control. Finally, the identified abnormal data points were recorded in the activity anomaly detection matrix, which details the location, type, and degree of the abnormal data.

[0031] Next, based on the metabolite prediction matrix, anomaly detection is performed on the metabolite detection matrix to obtain a metabolic anomaly detection matrix. This process is similar to the detection of strain activity anomalies. Specifically, the actual data in the metabolite detection matrix is ​​compared point by point with the predicted data in the metabolite prediction matrix. Through statistical analysis or machine learning algorithms, significant differences between the actual and predicted data are identified. These differences may indicate abnormal accumulation or absence of metabolites during fermentation, such as flavor defects, acidity imbalances, or abnormal metabolic pathways. Finally, the identified abnormal data points are recorded in the metabolic anomaly detection matrix, which details the location, type, and severity of the abnormal data.

[0032] Finally, the activity anomaly detection matrix and the metabolic anomaly detection matrix are fused, which means integrating the abnormal data in the two matrices. The fusion process can be carried out by weighting, analysis of covariance or Bayesian methods to form a comprehensive activity and metabolic anomaly matrix. This matrix records in detail the abnormalities of strain activity and metabolites during fermentation, which can provide an important basis for subsequent fermentation accident prediction and regulation.

[0033] P30: Based on the abnormal activity metabolism matrix, perform multi-factor fermentation accident prediction and tracing in the yogurt fermentation workshop to obtain a fermentation accident prediction and tracing map.

[0034] Furthermore, step P30 in this embodiment of the application also includes:

[0035] P31: Classify the fermentation accident record set in the yogurt fermentation workshop to obtain multiple fermentation accident record blocks corresponding to multiple fermentation accident types; P32: Extract the qth fermentation accident record block corresponding to the qth fermentation accident type from the multiple fermentation accident record blocks, where q is a positive integer; P33: Perform accident tree tracing based on the qth fermentation accident record block to obtain the qth fermentation accident tracing path group, and train the qth fermentation accident prediction initial model based on the qth fermentation accident tracing path group; P34: Based on the accident prediction loss characteristics of the qth fermentation accident prediction initial model, inject perturbation into the qth fermentation accident record block according to the adversarial example generator to obtain the qth fermentation accident perturbation block; P35: Perform robustness enhancement training on the qth fermentation accident prediction initial model based on the qth fermentation accident perturbation block to generate the qth fermentation accident prediction tracing model; P36: Input the active metabolism abnormality matrix into the qth fermentation accident prediction tracing model to obtain the qth fermentation accident prediction tracing result, and add the qth fermentation accident prediction tracing result to the fermentation accident prediction tracing map.

[0036] Specifically, by using the active metabolic anomaly matrix, we can predict and trace multiple fermentation accidents in the yogurt fermentation workshop and generate a fermentation accident prediction and traceability map.

[0037] First, the fermentation accident records from the yogurt fermentation workshop were categorized to obtain multiple fermentation accident record blocks corresponding to various accident types. Each accident type represents a different fermentation problem, such as temperature fluctuations, microbial contamination, or abnormal acidity. Categorization allows historical accidents to be grouped by type, facilitating subsequent analysis and tracing. Furthermore, these accident record blocks include the time of the accident, the affected batches, relevant environmental parameters (such as temperature, pH, and dissolved oxygen), and the state of microbial activity and metabolites at the time.

[0038] Next, from the multiple classified incident record blocks, record blocks related to specific fermentation incident types are extracted to form the q-th fermentation incident record block corresponding to the q-th fermentation incident type, where q is a positive integer representing a specific type of fermentation incident. For example, when q=1, all incident record blocks related to temperature fluctuations are extracted, and when q=2, all record blocks related to acidity anomalies are extracted. In this way, each type of incident has a corresponding record block, ensuring that each type of incident can be traced and analyzed in detail later.

[0039] Then, fault tree tracing is performed based on the record block of the q-th fermentation accident to obtain the fault tree tracing path group for the q-th fermentation accident. Fault tree tracing is a system analysis method used to identify various factors that lead to an accident and their interrelationships. By constructing a fault tree, the various stages and paths from the initial event to the final accident can be clearly displayed, i.e., the accident tracing paths, forming the fault tree tracing path group. Based on this fault tree tracing path group, an initial prediction model for the q-th fermentation accident is trained. Machine learning algorithms can be used to establish a preliminary prediction model based on the paths and data obtained from the fault tree tracing, used to preliminarily predict the probability and impact of similar accidents occurring in the future.

[0040] Subsequently, based on the accident prediction loss characteristics of the initial model for predicting the q-th fermentation accident, an adversarial example generator was used to inject perturbations into the q-th fermentation accident record block. The adversarial example generator generates new perturbation blocks by adjusting some key parameters in the accident data, such as temperature and pH changes, thus obtaining the q-th fermentation accident perturbation block. An adversarial example generator is a tool used to generate adversarial examples; it can improve the robustness of the model by adding small perturbations to the original data to generate new training samples.

[0041] Next, the initial model for predicting the q-th fermentation accident is robustly enhanced using the q-th fermentation accident perturbation block. By continuously adjusting the model parameters, the model's ability to cope with unknown perturbations is improved, enabling it to handle more variables during fermentation while maintaining prediction accuracy. Through multiple training iterations, a retrospective model for predicting the q-th fermentation accident is generated.

[0042] Finally, the abnormal active metabolism matrix is ​​input into the q-th fermentation accident prediction and tracing model to predict and output the q-th fermentation accident prediction and tracing result, which is then added to the fermentation accident prediction and tracing map. In other words, the model's prediction results are integrated into a comprehensive map to facilitate unified management and analysis of different types of fermentation accidents. The fermentation accident prediction and tracing map is a visualization tool that displays the prediction results and tracing paths for different fermentation accidents, providing decision support for fermentation plant operators.

[0043] P40: Based on the fermentation accident prediction and tracing map, the current fermentation decision is compensated and adjusted to obtain the initial fermentation regulation group.

[0044] Optionally, after constructing the fermentation accident prediction and tracing map, the system can visually display the potential paths and influencing factors of different types of fermentation accidents. These paths reveal key variables and factors that may lead to fermentation problems, such as temperature fluctuations, pH changes, and decline in strain activity. By analyzing this accident prediction information, potential risk factors in current fermentation decisions can be identified, especially those factors that have already been identified as potentially causing problems through the prediction and tracing map.

[0045] Therefore, based on the results of the predictive traceability profile, current fermentation decisions can be adjusted to compensate for potential problems. By adjusting parameters that may trigger accidents during fermentation, the system can ensure that the fermentation process remains stable and efficient. For example, if the predictive profile indicates that temperature fluctuations may lead to abnormal acidity, the system may adjust the fermentation temperature range or optimize the temperature control strategy; if the profile indicates that the inoculum activity may decrease, the system may adjust the inoculum size or optimize the stirring rate to improve the growth rate and activity of the inoculum. Through these adjustments, potential risks can be eliminated or reduced, preventing malfunctions or quality problems during fermentation.

[0046] After compensation and adjustment, a new initial fermentation adjustment group can be generated, which contains multiple adjustment schemes for the current fermentation decision. Each scheme corresponds to a different adjustment strategy, such as temperature adjustment, time extension, and stirring rate adjustment, which can provide multiple possible adjustment schemes for subsequent optimization processes, so as to further screen and optimize.

[0047] P50: Based on the strain activity evaluation model and metabolic quality evaluation model, the initial fermentation regulation group is optimized in multiple ways to establish a fermentation regulation multi-way optimization domain.

[0048] Furthermore, step P50 in this embodiment of the application also includes:

[0049] P51: Based on the strain activity evaluation model, the initial fermentation regulation group is optimized by predicting strain activity, establishing a first optimization domain for fermentation regulation; P52: Based on the metabolic quality evaluation model, the initial fermentation regulation group is optimized by predicting metabolic quality, establishing a second optimization domain for fermentation regulation; P53: Based on predetermined energy consumption constraints, the initial fermentation regulation group is optimized by predicting energy consumption, establishing a third optimization domain for fermentation regulation; P54: The first optimization domain, the second optimization domain, and the third optimization domain for fermentation regulation are integrated to generate the multi-faceted optimization domain for fermentation regulation.

[0050] It should be understood that after obtaining the initial fermentation regulator group, the first step is to predict and optimize the strain activity based on the strain activity evaluation model. This model, established based on historical data and experimental results, can predict the activity performance of the strain under different fermentation conditions. Using this model, each regulatory decision in the initial fermentation regulator group is evaluated, predicting its potential impact on strain activity. Specifically, the model considers the adjustment of key process parameters such as fermentation temperature, pH, and inoculum size, as well as the impact of these adjustments on strain growth and metabolic activity. Based on the prediction results, regulatory decisions that can significantly improve strain activity or maintain its stability are selected, thus establishing the first optimization domain for fermentation regulation. This optimization domain contains regulatory schemes that demonstrate excellent performance in terms of strain activity, providing a foundation for subsequent optimization processes.

[0051] Next, metabolic quality prediction and optimization were performed on the initial fermentation regulatory group based on the metabolic quality assessment model. The metabolic quality assessment model focuses on evaluating the quality and yield of metabolites during fermentation, such as the formation of lactic acid and flavor compounds, and predicts the impact of different regulatory decisions on the quality and yield of metabolites by analyzing changes in fermentation conditions. For example, some regulatory decisions may improve the quality of yogurt by optimizing fermentation temperature and time, thereby promoting the formation of specific flavor compounds. By performing metabolic quality prediction and evaluation on each decision in the initial fermentation regulatory group, regulatory schemes that can improve product quality, reduce metabolic byproducts, and optimize flavor were screened, establishing a second optimization domain for fermentation regulation. This optimization domain focuses on optimizing the quality of metabolites, which helps improve the taste and nutritional value of yogurt.

[0052] Furthermore, to ensure energy efficiency optimization of the fermentation process, energy consumption prediction and optimization are performed on the initial fermentation control group based on predetermined energy consumption constraints. Predetermined energy consumption constraints refer to the upper limit of energy consumption set during the fermentation process. By performing energy consumption prediction analysis on each control decision in the initial fermentation control group, its impact on energy consumption while meeting fermentation process requirements can be assessed. For example, some control decisions may reduce equipment operating energy consumption by optimizing fermentation time and temperature. Control decisions that can achieve efficient fermentation under the condition of meeting energy consumption constraints are selected, establishing a third optimization domain for fermentation control. This domain includes schemes that achieve the best fermentation effect while minimizing energy consumption.

[0053] Finally, the first, second, and third optimization domains of fermentation regulation are integrated to generate a multi-faceted optimization domain for fermentation regulation. Specifically, a multi-objective optimization method is used to comprehensively consider the balance between strain activity, metabolic quality, and energy efficiency. Multi-objective optimization algorithms, such as genetic algorithms and particle swarm optimization, are employed to weigh and optimize these regulatory decisions, generating a comprehensive multi-faceted optimization domain for fermentation regulation. This optimization domain includes regulatory schemes that perform well in multiple aspects such as strain activity, metabolic quality, and energy consumption, providing effective decision support for the final fermentation regulation decision.

[0054] Furthermore, step P51 in the embodiments of this application also includes:

[0055] P51-1: Traverse the initial fermentation regulation group and extract the first fermentation regulation decision; P51-2: Perform multiple simulation adjustments on the fermentation twin workshop according to the first fermentation regulation decision to obtain multiple fermentation regulation simulation data; P51-3: Extract the association of strain activity based on the multiple fermentation regulation simulation data to obtain the first regulatory strain activity feature set; P51-4: Perform confidence fusion based on the first regulatory strain activity feature set to obtain the first regulatory strain activity confidence vector; P51-5: Input the first regulatory strain activity confidence vector into the strain activity evaluation model to obtain the first strain activity evaluation coefficient; P51-6: If the first strain activity evaluation coefficient is greater than or equal to the strain activity evaluation threshold, add the first fermentation regulation decision to the first optimization domain of fermentation regulation.

[0056] Specifically, the process of predicting and optimizing strain activity can be further refined to ensure that the best fermentation regulation scheme can be selected to improve strain activity.

[0057] First, the initial fermentation regulation group is traversed, and each fermentation regulation decision is extracted one by one. For example, the first fermentation regulation decision is randomly selected as the initial evaluation target. Next, the fermentation twin workshop is subjected to multiple simulation adjustments based on the first fermentation regulation decision. The fermentation twin workshop is a digital model capable of simulating the operation of an actual fermentation workshop under different regulation decisions. By performing multiple simulation adjustments according to the first fermentation regulation decision, result data from multiple simulation adjustments can be obtained. The central parity of these result data is taken, and the effective data is filtered out to generate multiple, more accurate fermentation regulation simulation data, such as information on strain activity and metabolite generation.

[0058] Then, the activity correlation of these fermentation regulation simulation data is extracted. By analyzing the changes in the activity of the strains in the simulation data, the activity features related to the first fermentation regulation decision are extracted, including the growth rate, metabolic activity, and cell survival rate of the strains, and the first regulatory strain activity feature set is obtained.

[0059] Subsequently, confidence fusion is performed based on the first regulatory strain activity feature set. Confidence fusion integrates information from multiple data sources, enhancing the influence of important features and reducing noise interference through weighting or other methods. Through confidence fusion, multiple strain activity features can be integrated into a comprehensive first regulatory strain activity confidence vector. This vector reflects the reliability of strain activity changes under the first fermentation regulation decision.

[0060] Next, the confidence vector of the first regulatory bacterial species activity is input into the bacterial species activity evaluation model for analysis. The bacterial species activity evaluation model is a pre-trained model that can assess the level of bacterial species activity based on the input confidence vector and obtain the first bacterial species activity evaluation coefficient. This evaluation coefficient is a quantitative indicator used to measure the quality of bacterial species activity.

[0061] Finally, it is determined whether the first strain activity evaluation coefficient is greater than or equal to the strain activity evaluation threshold. The strain activity evaluation threshold is a preset benchmark value used to determine whether the strain activity meets the expected optimization standard. If the first strain activity evaluation coefficient meets the condition, it indicates that the fermentation regulation decision can effectively improve the strain activity and meets the optimization requirements. Therefore, this first fermentation regulation decision is added to the first optimization domain of fermentation regulation as part of the strain activity optimization scheme. Similarly, the initial fermentation regulation group is traversed to evaluate and screen all fermentation regulation decisions, completing the construction of the first optimization domain of fermentation regulation. This optimization domain contains all fermentation regulation decisions that can effectively improve strain activity.

[0062] P60: Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproduction optimization according to the fermentation regulation multi-party optimization domain to generate fermentation regulation instructions.

[0063] Furthermore, step P60 in this embodiment of the application also includes:

[0064] P61: Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the first optimization domain of fermentation regulation to establish a first fermentation regulation space; P62: Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the second optimization domain of fermentation regulation to establish a second fermentation regulation space; P63: Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the third optimization domain of fermentation regulation to establish a third fermentation regulation space; P64: The first fermentation regulation space, the second fermentation regulation space, and the third fermentation regulation space are merged to establish a fourth fermentation regulation space; P65: The fermentation regulation fitness analysis conditions are activated, including the strain activity evaluation weight and the metabolic quality evaluation weight; P66: The fermentation regulation fitness of the fourth fermentation regulation space is iteratively optimized according to the fermentation regulation fitness analysis conditions to obtain the fermentation regulation instruction.

[0065] It should be understood that, firstly, based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulatory group is guided to reproduce and optimize according to the first optimization domain of fermentation regulation, thus establishing the first fermentation regulation space. Specifically, through a reproduction algorithm, the system selects, crosses, and mutates multiple regulatory decisions in the initial regulatory group to generate new regulatory schemes. These new schemes are optimized in terms of strain activity and metabolite production, thereby forming the first fermentation regulation space. This space represents the combination of all possible regulatory schemes under the background of strain activity optimization, ensuring that the decision space that maximizes strain activity is found.

[0066] Next, based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulatory group was guided to reproduce and optimize according to the second optimization domain of fermentation regulation, thus establishing a second fermentation regulatory space. Similar to the first fermentation regulatory space, this process also optimizes the regulatory decisions in the initial fermentation regulatory group through reproduction and mutation operations. However, this optimization process focuses on improving metabolic quality. By improving metabolite production, it ensures an ideal balance of metabolites during fermentation, resulting in regulatory decisions in the generated second fermentation regulatory space that demonstrate superior performance in terms of metabolite quality.

[0067] Subsequently, based on the strain activity evaluation model and metabolic quality evaluation model, the initial fermentation regulatory group was guided to reproduce and optimize according to the third optimization domain of fermentation regulation, thus establishing a third fermentation regulatory space. This process also utilizes reproduction and mutation operations to optimize the regulatory decisions in the initial fermentation regulatory group. This optimization process focuses on reducing energy consumption while ensuring the stability of the fermentation process and product quality. Through evaluation and optimization, the regulatory decisions in the generated third fermentation regulatory space show better performance in terms of energy consumption.

[0068] Then, the first, second, and third fermentation regulation spaces are merged to establish a fourth fermentation regulation space. This space contains regulatory decisions that perform well in multiple aspects such as strain activity, metabolic quality, and energy consumption, and can simultaneously meet the optimization needs of multiple objectives.

[0069] Next, the fermentation regulation fitness analysis conditions are activated, including the weights for strain activity evaluation and metabolic quality evaluation. These conditions serve as the basis for evaluating and selecting the optimal fermentation regulation decisions. The weights for strain activity evaluation and metabolic quality evaluation reflect the importance of strain activity and metabolic quality in the fermentation process, respectively. By adjusting these two weights, the system can flexibly balance the priority between strain activity and metabolic quality according to the specific needs of the fermentation process.

[0070] Finally, the fermentation regulation fitness of the fourth fermentation regulation space is iteratively optimized based on the analytical conditions of fermentation regulation fitness. Iterative optimization algorithms, such as genetic algorithms or particle swarm optimization, are used to evaluate and select regulation decisions in the fourth fermentation regulation space. Based on the analytical conditions of fermentation regulation fitness, the regulation scheme is continuously optimized so that the final generated regulation decision achieves the best fermentation effect while satisfying the requirements of strain activity, metabolic quality, and energy efficiency, thus generating the optimal fermentation regulation instructions. These instructions will guide the production operation in the actual fermentation workshop to achieve a high-efficiency, stable, and high-quality fermentation process.

[0071] Furthermore, step P61 in the embodiments of this application also includes:

[0072] P61-1: Perform difference detection on the initial fermentation regulation group according to the first optimization domain of fermentation regulation to obtain a first fermentation regulation difference feature set; P61-2: Guide the initial fermentation regulation group to reproduce according to the first fermentation regulation difference feature set to obtain a first fermentation regulation reproduction domain; P61-3: Perform strain activity prediction optimization on the first fermentation regulation reproduction domain according to the strain activity evaluation model to establish a second fermentation regulation reproduction domain; P61-4: Perform metabolic quality prediction optimization on the second fermentation regulation reproduction domain according to the metabolic quality evaluation model to establish a third fermentation regulation reproduction domain; P61-5: Perform energy consumption prediction optimization on the third fermentation regulation reproduction domain according to the predetermined energy consumption constraint to generate the first fermentation regulation space.

[0073] Specifically, the reproductive optimization process based on the first optimization domain of fermentation regulation can be further refined.

[0074] First, differential detection was performed on the initial fermentation regulatory group based on the first optimization domain of fermentation regulation to obtain the first fermentation regulation differential feature set. Through detailed analysis of each regulatory decision in the initial fermentation regulatory group, including data preprocessing, differential analysis, and feature extraction, significant differences between regulatory decisions were identified, and key features were extracted to generate the first fermentation regulation differential feature set. These features may include the adjustment range of fermentation parameters, the changing trend of strain activity, etc., which can provide a basis for subsequent propagation operations.

[0075] Next, the initial fermentation regulatory group is guided to reproduce based on the first set of differential features for fermentation regulation. This means that key features within the first set of differential features are used as adjustment directions for reproduction operations, such as moving towards values ​​of key features or focusing on adjusting the values ​​of key features. Reproduction operations are performed on each regulatory decision in the initial fermentation regulatory group to generate new regulatory decisions, thus forming the first fermentation regulatory reproduction domain. These new decisions have a higher probability of exhibiting higher fitness in terms of strain activity, which can improve the accuracy of global optimization.

[0076] Subsequently, the strain activity of the first fermentation regulation and reproduction domain was predicted and optimized based on the strain activity evaluation model. This involved using the model to assess newly generated regulatory decisions, predicting the impact of each decision on strain activity, and generating new strain activity evaluation coefficients. Based on these coefficients, regulatory decisions exhibiting superior strain activity were selected to form the second fermentation regulation and reproduction domain.

[0077] Next, the metabolic quality of the second fermentation regulatory and reproductive domain is predicted and optimized using a metabolic quality assessment model. This involves evaluating the selected regulatory decisions using the model, predicting the impact of each decision on metabolite quality, and generating metabolic quality assessment coefficients. Based on these coefficients, regulatory decisions that perform better in terms of metabolite quality are further selected, forming the third fermentation regulatory and reproductive domain.

[0078] Finally, energy consumption prediction and optimization are performed on the third fermentation regulation and reproduction domain according to predetermined energy consumption constraints to generate the first fermentation regulation space. This involves predicting and evaluating the energy consumption of each regulation decision within the third fermentation regulation and reproduction domain to ensure it meets the predetermined energy consumption constraints. Based on the energy consumption evaluation results, regulation decisions that perform well in terms of energy consumption are selected, ultimately generating the first fermentation regulation space. This process not only considers the optimization of strain activity and metabolite quality but also takes into account energy consumption constraints, providing crucial decision support for subsequent comprehensive optimization.

[0079] In summary, the embodiments of this application have at least the following technical effects:

[0080] This application ensures process stability by detecting bacterial activity and metabolites throughout the entire fermentation process, monitoring key parameters in real time, and using an activity metabolism anomaly matrix and fermentation accident prediction and tracing to identify and warn of potential fermentation accidents in advance, reducing production risks. By combining bacterial activity evaluation models and metabolic quality evaluation models, the fermentation regulation scheme is optimized from multiple perspectives to achieve comprehensive optimization of bacterial activity, metabolites, and energy efficiency. By generating fermentation regulation instructions, the yogurt fermentation process is automatically regulated and optimized, reducing manual intervention and improving production efficiency and product consistency.

[0081] This technology achieves the goal of efficiently optimizing and stably controlling the yogurt fermentation process through full-process testing and precise regulation, thereby improving product quality and production efficiency.

[0082] Example 2, based on the same inventive concept as the full-process detection method for yogurt strain activity and metabolites in the aforementioned examples, such as... Figure 2 As shown, this application provides a full-process detection system for yogurt strain activity and metabolites. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0083] The full-process detection module 11 is used to conduct full-process detection of the yogurt fermentation workshop and obtain the strain activity detection matrix and metabolite detection matrix corresponding to the current fermentation decision.

[0084] The anomaly fitting detection module 12 is used to perform anomaly fitting detection on the strain activity detection matrix and the metabolite detection matrix based on the current fermentation decision, and to establish an activity metabolism anomaly matrix.

[0085] The accident prediction and tracing module 13 is used to perform multi-factor fermentation accident prediction and tracing in the yogurt fermentation workshop based on the active metabolic abnormality matrix, and obtain a fermentation accident prediction and tracing map.

[0086] The compensation and adjustment module 14 is used to compensate and adjust the current fermentation decision based on the fermentation accident prediction and tracing map to obtain an initial fermentation adjustment group.

[0087] The multi-factor optimization module 15 is used to perform multi-factor optimization on the initial fermentation regulation group based on the strain activity evaluation model and the metabolic quality evaluation model, and to establish a fermentation regulation multi-factor optimization domain.

[0088] The propagation optimization module 16 is used to guide the initial fermentation regulation group to perform propagation optimization based on the strain activity evaluation model and the metabolic quality evaluation model, according to the fermentation regulation multi-party optimization domain, and generate fermentation regulation instructions.

[0089] Furthermore, the full-process detection module 11 is also used to perform the following steps:

[0090] The activity of the current strain is detected in the yogurt fermentation workshop to obtain strain activity detection data; the strain activity detection data is cleaned to generate the strain activity detection matrix.

[0091] Furthermore, the full-process detection module 11 is also used to perform the following steps:

[0092] The current metabolites in the yogurt fermentation workshop are detected to obtain metabolite detection data; the metabolite detection data is cleaned to obtain the metabolite detection matrix.

[0093] Furthermore, the anomaly fitting detection module 12 is also used to perform the following steps:

[0094] A twin model of the yogurt fermentation workshop is performed to obtain a fermentation twin workshop. Based on the current fermentation decision, multi-feature fitting is performed on the fermentation twin workshop to establish a strain activity prediction matrix and a metabolite prediction matrix. Anomaly detection is performed on the strain activity detection matrix based on the strain activity prediction matrix to obtain an activity anomaly detection matrix. Anomaly detection is performed on the metabolite detection matrix based on the metabolite prediction matrix to obtain a metabolic anomaly detection matrix. The activity anomaly detection matrix and the metabolic anomaly detection matrix are fused to generate the activity-metabolic anomaly matrix.

[0095] Furthermore, the accident prediction and tracing module 13 is also used to perform the following steps:

[0096] The fermentation accident record set in the yogurt fermentation workshop is classified to obtain multiple fermentation accident record blocks corresponding to multiple fermentation accident types. The qth fermentation accident record block corresponding to the qth fermentation accident type is extracted from these multiple fermentation accident record blocks, where q is a positive integer. An accident tree tracing is performed on the qth fermentation accident record block to obtain a qth fermentation accident tracing path group. Based on the qth fermentation accident tracing path group, an initial prediction model for the qth fermentation accident is trained. Based on the accident prediction loss characteristics of the initial prediction model for the qth fermentation accident, a perturbation injection is performed on the qth fermentation accident record block using an adversarial example generator to obtain a qth fermentation accident perturbation block. The initial prediction model for the qth fermentation accident is robustly reinforced using the qth fermentation accident perturbation block to generate a qth fermentation accident prediction tracing model. The abnormal activity metabolism matrix is ​​input into the qth fermentation accident prediction tracing model to obtain the qth fermentation accident prediction tracing result, and the qth fermentation accident prediction tracing result is added to the fermentation accident prediction tracing map.

[0097] Furthermore, the multi-party optimization module 15 is also used to perform the following steps:

[0098] The initial fermentation regulation group is optimized by predicting and selecting the activity of the microbial strains based on the microbial activity evaluation model, thus establishing a first optimization domain for fermentation regulation. The initial fermentation regulation group is optimized by predicting and selecting the metabolic quality based on the metabolic quality evaluation model, thus establishing a second optimization domain for fermentation regulation. The initial fermentation regulation group is optimized by predicting and selecting the energy consumption based on a predetermined energy consumption constraint, thus establishing a third optimization domain for fermentation regulation. The first optimization domain, the second optimization domain, and the third optimization domain for fermentation regulation are integrated to generate the multi-faceted optimization domain for fermentation regulation.

[0099] Furthermore, the multi-party optimization module 15 is also used to perform the following steps:

[0100] The initial fermentation regulation group is traversed to extract a first fermentation regulation decision; multiple simulations of the fermentation twin workshop are performed based on the first fermentation regulation decision to obtain multiple fermentation regulation simulation data; strain activity correlation is extracted based on the multiple fermentation regulation simulation data to obtain a first regulatory strain activity feature set; confidence fusion is performed based on the first regulatory strain activity feature set to obtain a first regulatory strain activity confidence vector; the first regulatory strain activity confidence vector is input into the strain activity evaluation model to obtain a first strain activity evaluation coefficient; if the first strain activity evaluation coefficient is greater than or equal to the strain activity evaluation threshold, the first fermentation regulation decision is added to the first optimization domain of the fermentation regulation.

[0101] Furthermore, the reproductive optimization module 16 is also used to perform the following steps:

[0102] Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the first optimization domain of fermentation regulation to establish a first fermentation regulation space; based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the second optimization domain of fermentation regulation to establish a second fermentation regulation space; based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproductive optimization according to the third optimization domain of fermentation regulation to establish a third fermentation regulation space; the first fermentation regulation space, the second fermentation regulation space, and the third fermentation regulation space are merged to establish a fourth fermentation regulation space; the fermentation regulation fitness analysis conditions are activated, the fermentation regulation fitness analysis conditions include strain activity evaluation weights and metabolic quality evaluation weights; the fermentation regulation fitness of the fourth fermentation regulation space is iteratively optimized according to the fermentation regulation fitness analysis conditions to obtain the fermentation regulation instructions.

[0103] Furthermore, the reproductive optimization module 16 is also used to perform the following steps:

[0104] Based on the first optimization domain of fermentation regulation, the initial fermentation regulation group is subjected to difference detection to obtain a first fermentation regulation difference feature set; the initial fermentation regulation group is guided to reproduce based on the first fermentation regulation difference feature set to obtain a first fermentation regulation reproduction domain; the first fermentation regulation reproduction domain is optimized for strain activity prediction based on the strain activity evaluation model to establish a second fermentation regulation reproduction domain; the second fermentation regulation reproduction domain is optimized for metabolic quality prediction based on the metabolic quality evaluation model to establish a third fermentation regulation reproduction domain; the third fermentation regulation reproduction domain is optimized for energy consumption prediction based on a predetermined energy consumption constraint to generate the first fermentation regulation space.

[0105] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0106] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0107] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A comprehensive detection method for the activity and metabolites of yogurt strains, characterized in that, The method includes: A full-process test was conducted in the yogurt fermentation workshop to obtain the strain activity detection matrix and metabolite detection matrix corresponding to the current fermentation decision. Based on the current fermentation decision, anomaly fitting detection is performed on the strain activity detection matrix and the metabolite detection matrix to establish an activity metabolism anomaly matrix. Based on the active metabolic anomaly matrix, a multi-factor fermentation accident prediction and tracing was performed in the yogurt fermentation workshop to obtain a fermentation accident prediction and tracing map. The current fermentation decision is compensated and adjusted based on the fermentation accident prediction and tracing map to obtain an initial fermentation regulation group; Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group was optimized in multiple ways to establish a fermentation regulation multi-factor optimization domain. Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproduction optimization according to the fermentation regulation multi-party optimization domain to generate fermentation regulation instructions; The step of performing multi-factor fermentation accident prediction and tracing in the yogurt fermentation workshop based on the active metabolic anomaly matrix to obtain a fermentation accident prediction and tracing map includes: Based on the fermentation accident record set of the yogurt fermentation workshop, multiple fermentation accident record blocks corresponding to multiple fermentation accident types are obtained; Extract the qth fermentation accident record block corresponding to the qth fermentation accident type from the multiple fermentation accident record blocks, where q is a positive integer; Based on the q-th fermentation accident record block, an accident tree tracing is performed to obtain the q-th fermentation accident tracing path group, and based on the q-th fermentation accident tracing path group, an initial prediction model for the q-th fermentation accident is trained. Based on the accident prediction loss characteristics of the initial model for predicting the qth fermentation accident, the qth fermentation accident record block is perturbed and injected according to the adversarial example generator to obtain the qth fermentation accident perturbation block. Based on the qth fermentation accident disturbance block, the initial model for predicting the qth fermentation accident is robustly enhanced and trained to generate a retrospective model for predicting the qth fermentation accident. The abnormal active metabolism matrix is ​​input into the q-th fermentation accident prediction and tracing model to obtain the q-th fermentation accident prediction and tracing result, and the q-th fermentation accident prediction and tracing result is added to the fermentation accident prediction and tracing map; The process, based on the strain activity evaluation model and the metabolic quality evaluation model, guides the initial fermentation regulation population to perform reproductive optimization according to the fermentation regulation multi-way optimization domain, generating fermentation regulation instructions, including: Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproduction optimization according to the first optimization domain of fermentation regulation to establish the first fermentation regulation space; Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproduction optimization according to the second optimization domain of fermentation regulation to establish a second fermentation regulation space; Based on the strain activity evaluation model and the metabolic quality evaluation model, the initial fermentation regulation group is guided to perform reproduction optimization according to the third optimization domain of fermentation regulation to establish the third fermentation regulation space; The first fermentation regulation space, the second fermentation regulation space, and the third fermentation regulation space are merged to establish a fourth fermentation regulation space; Activate the fermentation regulation fitness analysis conditions, which include the strain activity evaluation weight and the metabolic quality evaluation weight; Based on the fermentation regulation fitness analysis conditions, the fermentation regulation fitness of the fourth fermentation regulation space is iteratively optimized to obtain the fermentation regulation command. The process, based on the strain activity evaluation model and the metabolic quality evaluation model, guides the initial fermentation regulation group to perform reproductive optimization according to the first optimization domain of fermentation regulation, establishing a first fermentation regulation space, including: Based on the first optimization domain of fermentation regulation, the initial fermentation regulation group is subjected to difference detection to obtain the first fermentation regulation difference feature set; The initial fermentation regulation group is guided to reproduce based on the first fermentation regulation differential feature set to obtain the first fermentation regulation reproduction domain; Based on the strain activity evaluation model, the strain activity of the first fermentation regulation and reproduction domain is predicted and optimized to establish the second fermentation regulation and reproduction domain. Based on the metabolic quality evaluation model, the metabolic quality of the second fermentation regulation and reproduction domain is predicted and optimized to establish the third fermentation regulation and reproduction domain. Based on predetermined energy consumption constraints, the energy consumption of the third fermentation regulation and reproduction domain is predicted and optimized to generate the first fermentation regulation space.

2. The method for the full-process detection of yogurt strain activity and metabolites as described in claim 1, characterized in that, Based on the current fermentation decision, anomaly fitting detection is performed on the strain activity detection matrix and the metabolite detection matrix to establish an activity metabolism anomaly matrix, including: Based on the yogurt fermentation workshop, a twin model was created to obtain the fermentation twin workshop; Based on the current fermentation decision, multi-feature fitting is performed on the fermentation twin workshop to establish a strain activity prediction matrix and a metabolite prediction matrix. Anomaly detection matrix is ​​obtained by performing anomaly detection on the strain activity detection matrix based on the strain activity prediction matrix. Anomaly detection is performed on the metabolite detection matrix based on the metabolite prediction matrix to obtain a metabolic anomaly detection matrix. The activity abnormality detection matrix and the metabolic abnormality detection matrix are fused to generate the activity metabolic abnormality matrix.

3. The method for the full-process detection of yogurt strain activity and metabolites as described in claim 1, characterized in that, Based on the strain activity evaluation model and metabolic quality evaluation model, the initial fermentation regulatory population was optimized in multiple ways to establish a fermentation regulation multi-way optimization domain, including: Based on the strain activity evaluation model, the initial fermentation regulation group is optimized by predicting strain activity, and a first optimization domain for fermentation regulation is established. Based on the metabolic quality evaluation model, the initial fermentation regulation group is subjected to metabolic quality prediction and optimization to establish a second optimization domain for fermentation regulation. Based on the predetermined energy consumption constraints, the initial fermentation regulation group is optimized by predicting energy consumption, and a third optimization domain for fermentation regulation is established. The fermentation regulation first optimization domain, the fermentation regulation second optimization domain, and the fermentation regulation third optimization domain are integrated to generate the fermentation regulation multi-directional optimization domain.

4. The method for the full-process detection of yogurt strain activity and metabolites as described in claim 3, characterized in that, Based on the strain activity evaluation model, the initial fermentation regulation group is optimized by predicting strain activity, and a first optimization domain for fermentation regulation is established, including: Traverse the initial fermentation regulation group and extract the first fermentation regulation decision; Based on the first fermentation regulation decision, the fermentation twin workshop was simulated and regulated multiple times to obtain multiple fermentation regulation simulation data. Based on the multiple fermentation regulation simulation data, the activity correlation of the strains is extracted to obtain the first regulatory strain activity feature set; Confidence fusion is performed based on the first regulatory bacterial species activity feature set to obtain the first regulatory bacterial species activity confidence vector; The first regulatory strain activity confidence vector is input into the strain activity evaluation model to obtain the first strain activity evaluation coefficient. If the first strain activity evaluation coefficient is greater than or equal to the strain activity evaluation threshold, the first fermentation regulation decision is added to the first optimization domain of fermentation regulation.

5. The method for the full-process detection of yogurt strain activity and metabolites as described in claim 1, characterized in that, A full-process inspection of the yogurt fermentation workshop was conducted, including: The activity of the current bacterial strain in the yogurt fermentation workshop was tested to obtain bacterial strain activity test data; The bacterial activity detection data is cleaned to generate the bacterial activity detection matrix.

6. The method for the full-process detection of yogurt strain activity and metabolites as described in claim 1, characterized in that, A full-process inspection of the yogurt fermentation workshop was conducted, including: The current metabolites in the yogurt fermentation workshop are detected to obtain metabolite detection data; The metabolite detection data is cleaned to obtain the metabolite detection matrix.

7. A comprehensive detection system for yogurt strain activity and metabolites, characterized in that, The system is used to execute the full-process detection method for yogurt strain activity and metabolites as described in any one of claims 1 to 6, the system comprising: The full-process detection module is used to perform full-process detection in the yogurt fermentation workshop and obtain the strain activity detection matrix and metabolite detection matrix corresponding to the current fermentation decision. An anomaly fitting detection module is used to perform anomaly fitting detection on the strain activity detection matrix and the metabolite detection matrix based on the current fermentation decision, and to establish an activity metabolism anomaly matrix. The accident prediction and tracing module is used to perform multi-factor fermentation accident prediction and tracing in the yogurt fermentation workshop based on the active metabolism abnormality matrix, and obtain a fermentation accident prediction and tracing map. The compensation and adjustment module is used to compensate and adjust the current fermentation decision based on the fermentation accident prediction and tracing map to obtain an initial fermentation adjustment group; The multi-factor optimization module is used to perform multi-factor optimization on the initial fermentation regulation group based on the strain activity evaluation model and the metabolic quality evaluation model, and to establish a fermentation regulation multi-factor optimization domain. The propagation optimization module is used to guide the initial fermentation regulation group to perform propagation optimization based on the strain activity evaluation model and the metabolic quality evaluation model, according to the fermentation regulation multi-way optimization domain, and generate fermentation regulation instructions.

Images