Biological reaction intelligent monitoring system and method

Abstract

The invention discloses an intelligent biological reaction monitoring system and method, and belongs to the field of intelligent monitoring, and the system comprises a reaction stage dividing unit which is used for dividing and determining multiple reaction stages; the variable relation mining unit is used for determining a dominant variable and an auxiliary variable, mining a variable linear relation and a collaborative insensitive loss function, and supervising and training the soft monitoring module; the continuous monitoring acquisition unit is used for configuring a monitoring equipment group, performing continuous monitoring and determining reaction monitoring data; the data analyzing and screening unit is used for preprocessing the data, performing linear analysis and abnormal reaction positioning and determining effective monitoring data; the game equilibrium decision-making unit is used for performing game equilibrium decision-making on the abnormal reaction identifier and determining a reaction adjustment strategy; and the reaction monitoring management unit is used for responding to the adjustment strategy, carrying out feedback monitoring analysis and reaction monitoring management, improving the biological reaction monitoring precision, realizing real-time online monitoring and carrying out intelligent regulation and control aiming at abnormal conditions.

Description

Technical Field

[0001] The present invention relates to the field of intelligent monitoring, and in particular to a biological reaction intelligent monitoring system and method. Background Art

[0002] Due to the complexity and dynamics of the biological reaction process, the reaction needs to be monitored and regulated in real time and accurately. In the prior art, biological reaction monitoring is mainly carried out by offline sampling and analysis, which has the problems of low monitoring accuracy and poor real-time performance. Offline sampling and analysis require manual operation, and the sampling frequency is limited, making it difficult to monitor the reaction in real time and continuously, resulting in incompleteness and lag in monitoring data. At the same time, since the biological reaction process involves multiple parameters and indicators, the existing offline analysis methods are difficult to achieve accurate identification and quantitative analysis of key variables, affecting the monitoring accuracy. In addition, in the biological reaction process, due to the influence of factors such as raw material fluctuations, equipment failures, and environmental disturbances, abnormal operating conditions often occur, causing the reaction to deviate from the optimal state. The prior art lacks effective intelligent decision-making means, making it difficult to quickly and accurately diagnose and regulate abnormal situations, affecting the biological reaction process. Therefore, the prior art has the technical problems of low accuracy, poor real-time performance, and difficulty in effectively regulating abnormal reactions in biological reaction monitoring. Summary of the invention

[0003] The present application provides a biological reaction intelligent monitoring system and method, aiming to solve the technical problems in the prior art of low biological reaction monitoring accuracy, poor real-time performance, and difficulty in effectively regulating abnormal reactions.

[0004] In view of the above problems, the present application provides a biological reaction intelligent monitoring system and method.

[0005] The first aspect disclosed in the present application provides an intelligent monitoring system for biological reactions, which includes: a reaction stage division unit, which is used to interact with the reaction process of the target biological reaction and divide and determine multiple reaction stages; a variable relationship mining unit, which is used to traverse multiple reaction stages, determine the dominant variables and auxiliary variables, mine the linear relationship of variables, coordinate the insensitive loss function, and supervise the training of the soft monitoring module; a continuous monitoring acquisition unit, which is used to traverse the dominant variables and auxiliary variables, configure a monitoring device group for directly monitoring the variables, perform continuous monitoring along with the reaction process, and determine the reaction monitoring data; a data analysis and screening unit, which is used to pre-process the reaction monitoring data at the front end, transmit it back to the soft monitoring module for linear analysis and abnormal reaction location, and determine the effective monitoring data, and the effective monitoring data includes the stage variable data and the abnormal reaction identification; a game equilibrium decision unit, which is used to traverse the effective monitoring data, make a game equilibrium decision on the abnormal reaction identification, and determine the reaction adjustment strategy; a reaction monitoring management unit, which is used for the reaction container to respond to the reaction adjustment strategy, and combines the soft monitoring module to perform feedback monitoring analysis to perform reaction monitoring management of the target biological reaction.

[0006] Another aspect disclosed in the present application provides a method for intelligent monitoring of biological reactions, which includes: dividing and determining multiple reaction stages according to the reaction process of an interactive target biological reaction; traversing multiple reaction stages, determining dominant variables and auxiliary variables, mining linear relationships between variables, coordinating insensitive loss functions, and supervising and training soft monitoring modules; traversing dominant variables and auxiliary variables, configuring a monitoring device group that directly monitors the variables, performing continuous monitoring along with the reaction process, and determining reaction monitoring data, wherein the reaction monitoring data is identified with stage nodes; preprocessing the reaction monitoring data at the front end, transmitting the data back to the soft monitoring module for linear analysis and abnormal reaction location, and determining effective monitoring data, wherein the effective monitoring data includes stage variable data and abnormal reaction identification; traversing the effective monitoring data, making a game equilibrium decision on the abnormal reaction identification, and determining a reaction adjustment strategy; the reaction container responds to the reaction adjustment strategy, and performs feedback monitoring and analysis in combination with the soft monitoring module to perform reaction monitoring management of the target biological reaction.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By adopting the reaction stage division unit, the reaction process of the interactive target biological reaction is divided and determined into multiple reaction stages, which can describe the reaction process more precisely and provide a basis for subsequent variable analysis and monitoring control; through the variable relationship mining unit, multiple reaction stages are traversed to determine the dominant variables and auxiliary variables, and the linear relationship between variables is mined, and the insensitive loss function is coordinated to supervise the training of the soft monitoring module. With a data-driven approach, the intrinsic connection between the key variables in the biological reaction process is revealed to improve the monitoring accuracy; through the continuous monitoring and acquisition unit, the dominant variables and auxiliary variables are traversed, and a monitoring equipment group for directly monitoring the variables is configured. Continuous monitoring is performed along with the reaction process to determine the reaction monitoring data, realize real-time online monitoring of the biological reaction process, and improve the timeliness and integrity of data collection; the reaction monitoring data is pre-processed at the front end through the data analysis and screening unit and transmitted back to the soft monitoring The module performs linear analysis and abnormal reaction positioning, determines effective monitoring data, including stage variable data and abnormal reaction identification, automatically screens out high-quality monitoring data, and provides a reliable basis for subsequent decision-making and control; traverses the effective monitoring data through the game equilibrium decision unit, performs game equilibrium decision on the abnormal reaction identification, determines the reaction adjustment strategy, and generates the optimal abnormal handling strategy through multi-objective optimization and equilibrium analysis to achieve intelligent reaction regulation; controls the response adjustment strategy of the reaction vessel through the reaction monitoring management unit, and combines the soft monitoring module for feedback monitoring and analysis to achieve a technical solution for the full-process monitoring and management of the target biological reaction, which solves the technical problems of low accuracy, poor real-time performance, and difficulty in effectively regulating abnormal reactions in the existing technology, and achieves the technical effect of improving the accuracy of biological reaction monitoring, realizing real-time online monitoring, and performing intelligent regulation for abnormal situations.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A structural schematic diagram of a biological reaction intelligent monitoring system is provided for an embodiment of the present application.

[0010] Figure 2 A schematic flow chart of a method for intelligent monitoring of biological reactions is provided for an embodiment of the present application; Explanation of the accompanying drawings: reaction stage division unit 11, variable relationship mining unit 12, continuous monitoring and collection unit 13, data analysis and screening unit 14, game equilibrium decision unit 15, reaction monitoring management unit 16. DETAILED DESCRIPTION

[0011] The overall idea of ​​the technical solution provided by this application is as follows: The embodiment of the present application provides a system and method for intelligent monitoring of biological reactions. First, a reaction stage division unit is used to perform a refined description and stage division of the biological reaction process. Secondly, a variable relationship mining unit is used to reveal the intrinsic connection between key variables and train a soft monitoring module. Furthermore, real-time online monitoring of the biological reaction process is achieved through a continuous monitoring and acquisition unit, and intelligent data processing and abnormality positioning are performed using a data analysis and screening unit. Furthermore, a game equilibrium decision unit is introduced to perform multi-objective optimization decisions on abnormal reactions and generate an optimal control strategy. Afterwards, a reaction monitoring management unit is used to achieve closed-loop optimization control and full-process management of the biological reaction process, thereby achieving the purpose of improving the accuracy of biological reaction monitoring, realizing real-time online monitoring, and performing intelligent control for abnormal situations.

[0012] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0013] Embodiment 1 like Figure 1 As shown, the embodiment of the present application provides a biological reaction intelligent monitoring system, the system comprising: The reaction stage division unit 11 is used to divide and determine multiple reaction stages of the reaction process of the interactive target biological reaction.

[0014] Specifically, the reaction stage division unit 11 first obtains and analyzes the reaction process of the target biological reaction, and divides the complete reaction process into multiple reaction stages according to the characteristics and laws of the biological reaction at different stages, such as the dynamic changes of parameters such as biomass, substrate concentration, and metabolite concentration. Each reaction stage corresponds to a certain time period and is associated with a specific step in the reaction process. By dividing the target biological reaction into stages, the reaction stage division unit 11 outputs the reaction stage division result, i.e., multiple reaction stages, including the start and end time points, duration, and other information of each reaction stage, laying the foundation for subsequent intelligent monitoring and control of each reaction stage.

[0015] Through the reaction stage division unit 11, the stage characteristics of the biological reaction are fully considered, so as to achieve more refined and targeted stage-by-stage monitoring and regulation, and improve the accuracy of the biological reaction process optimization.

[0016] The variable relationship mining unit 12 is used to traverse multiple reaction stages, determine the dominant variables and auxiliary variables, mine the linear relationship of variables, coordinate the insensitive loss function, and supervise the training of the soft monitoring module.

[0017] Specifically, the variable relationship mining unit 12 identifies the key influencing factors of each reaction stage by analyzing historical production data and expert experience, and selects a series of dominant variables, such as temperature, pH value, dissolved oxygen content, etc. At the same time, it is considered that some dominant variables may not be able to be monitored online continuously due to the limitation of detection means. To this end, mathematical statistics methods, such as correlation analysis, regression analysis, etc., are further used to discover auxiliary variables with significant linear correlation with each dominant variable from a large number of process parameters. Subsequently, the variable relationship mining unit 12 makes full use of the linear relationship between the mined dominant variables and the auxiliary variables to construct a soft monitoring model. Through the soft monitoring model, the dominant variables that are difficult to monitor directly can be indirectly estimated and predicted with the help of auxiliary variables that are easy to monitor online continuously, thereby realizing real-time monitoring of all dominant variables. In the process of constructing the soft monitoring model, the variable relationship mining unit 12 also introduces an insensitive loss function to describe the degree of influence on the accuracy of the monitoring results when the measured values ​​of the dominant variables or auxiliary variables deviate. By coordinating the insensitive loss function, the robustness and anti-interference ability of the soft monitoring model are further improved. Afterwards, the determined dominant variables, auxiliary variables, variable linear relationships, and insensitive loss functions are used to construct a soft monitoring module, which is then supervised and trained. The trained and optimized soft monitoring module can accurately predict the state of the dominant variable based on the auxiliary variable data collected online, and timely reflect the dynamic changes of the biological reaction process.

[0018] Through the variable relationship mining unit 12, the coordinated monitoring of the dominant variables and the auxiliary variables is realized, overcoming the difficulty that some key parameters cannot be continuously monitored online, and by introducing an insensitive loss function, the adaptability and reliability of the soft monitoring module are further enhanced, thereby effectively improving the accuracy and real-time performance of biological reaction process monitoring.

[0019] The continuous monitoring acquisition unit 13 is used to traverse the dominant variables and the auxiliary variables, configure a monitoring device group for directly monitoring the variables, perform continuous monitoring along with the reaction process, and determine the reaction monitoring data.

[0020] Specifically, the continuous monitoring and acquisition unit 13 selects appropriate sensors, such as temperature sensors, pH electrodes, dissolved oxygen probes, etc., according to the physicochemical properties of each direct monitoring variable in the dominant variable and the auxiliary variable, and configures the installation position and quantity of the sensors in combination with the structural layout of the reaction device to form a complete monitoring equipment group. In the biological reaction process, the continuous monitoring and acquisition unit 13 uses the configured monitoring equipment group to perform real-time and continuous online monitoring of each direct monitoring variable. The raw data signal collected by the monitoring equipment group is pre-processed, such as filtering, amplification, A / D conversion, etc., and converted into usable digital quantitative monitoring data. The continuous monitoring and acquisition unit 13 summarizes and organizes the digital quantitative monitoring data collected by the monitoring equipment group to obtain a series of reaction monitoring data reflecting the real-time state of the biological reaction process, which is presented in the form of a time series, and each time point corresponds to a set of measured values ​​of the direct monitoring variables.

[0021] Through the continuous monitoring and acquisition unit 13, a complete monitoring equipment group is established for the direct monitoring variables in the biological reaction process, and the real-time continuous acquisition of process parameters is realized, providing reliable data support for subsequent data analysis and process optimization control.

[0022] The data analysis and screening unit 14 is used for front-end preprocessing reaction monitoring data, and transmits it back to the soft monitoring module for linear analysis and abnormal reaction positioning to determine effective monitoring data, which includes stage variable data and abnormal reaction identification.

[0023] Specifically, the data analysis and screening unit 14 first performs front-end preprocessing on the received reaction monitoring data. The preprocessing process includes a series of data quality control operations such as data cleaning, missing value interpolation, noise removal, and data normalization to improve the integrity, accuracy, and consistency of the data. Then, the data analysis and screening unit 14 returns the preprocessed reaction monitoring data to the soft monitoring module constructed by the variable relationship mining unit 12. The soft monitoring module uses the received reaction monitoring data to perform linear analysis and abnormal reaction location on the leading variables and auxiliary variables. By comparing the deviation between the actual monitoring value of each variable and the expected normal value, the soft monitoring module promptly detects abnormal fluctuations or situations beyond the control range in the biological reaction process, and generates corresponding abnormal reaction identification.

[0024] After receiving the analysis results of the soft monitoring module, the data analysis and screening unit 14 comprehensively considers the quality of the reaction monitoring data and the credibility of the soft monitoring results, performs secondary screening on the data, and determines the effective monitoring data with high confidence. The effective monitoring data is encapsulated in a structured form, which contains two types of key information: one is the variable data of each stage after the reaction stage is divided, and the other is the abnormal reaction identification identified by the soft monitoring module.

[0025] Through the data analysis and screening unit 14, the effective monitoring data presents the key information and intelligent analysis results of the biological reaction process in a clearer, more accurate and more efficient manner, laying a data foundation for realizing intelligent monitoring and optimized control.

[0026] The game equilibrium decision unit 15 is used to traverse the effective monitoring data, make a game equilibrium decision on the abnormal reaction mark, and determine the reaction adjustment strategy.

[0027] Specifically, the game equilibrium decision unit 15 is connected to the data analysis and screening unit 14 to receive the effective monitoring data output by the data analysis and screening unit 14, perform intelligent decision analysis on the abnormal reaction identification therein, and determine the reaction adjustment strategy. The game equilibrium decision unit 15 traverses the effective monitoring data transmitted by the data analysis and screening unit 14, and extracts the abnormal reaction identification information therein. The abnormal reaction identification is usually associated with the abnormal fluctuation of one or some key process parameters, and characterizes the potential problems or risks in the biological reaction process. For each abnormal reaction identification, the game equilibrium decision unit 15 matches the abnormal situation with the preset strategy set, and preliminarily screens out a series of possible response strategies. These strategies correct abnormal deviations by adjusting relevant process parameters, such as temperature, pH value, stirring rate, ventilation volume, etc., so that the reaction process returns to normal. However, in complex biological reaction processes, there are often interactions and game relationships between the various adjustment parameters. Simply adjusting a parameter may cause secondary offsets of other parameters, causing the problem to worsen. Therefore, when screening response strategies, the game equilibrium decision unit 15 introduces the idea of ​​game theory, regards the abnormal reaction process as a multi-party game process, and seeks the optimal equilibrium decision plan and determines the reaction adjustment strategy by weighing the interests of all parties and analyzing the benefits and costs of the strategy combination.

[0028] Through the game equilibrium decision unit 15, a systematic game analysis and equilibrium solution are carried out on the abnormal reaction identification, and the optimal decision plan is quickly locked in among the massive strategy combinations, which not only considers the solution of local problems, but also takes into account the optimization of the global process, thereby improving the flexibility, efficiency and robustness of dealing with abnormalities.

[0029] The reaction monitoring management unit 16 is used for the reaction container to respond to the reaction adjustment strategy, and combines with the soft monitoring module to perform feedback monitoring and analysis to perform reaction monitoring management of the target biological reaction.

[0030] Specifically, the reaction monitoring management unit 16 is connected to the game equilibrium decision unit 15 and receives the reaction adjustment strategy output by the reaction monitoring management unit 16. The reaction adjustment strategy includes a series of optimized control instructions for abnormal reactions, and clarifies the key parameters that need to be adjusted as well as the direction and amplitude of the adjustment. The reaction monitoring management unit 16 parses the reaction adjustment strategy into a series of executable control commands, and through digital / analog conversion, signal amplification, isolation protection and other interface processing, the control commands are sent to the various actuators of the biological reaction container, such as heaters, coolers, pH adjustment pumps, ventilation devices, etc. These actuators accurately adjust the relevant parameters of the reaction container according to the received control commands, so that the reaction process runs according to the guidance of the optimization strategy. While the strategy is being executed, the reaction monitoring management unit 16 collects various feedback signals of the reaction container in real time, such as temperature, pH value, dissolved oxygen, etc., and transmits them to the soft monitoring module. The soft monitoring module uses these feedback data to dynamically update its internal mathematical model and state estimation, and track and evaluate the effect of strategy execution. The soft monitoring module feeds back the evaluation results to the reaction monitoring management unit 16, including the actual changes in each key parameter, the degree of elimination of abnormal conditions, etc. The reaction monitoring management unit 16 determines the effectiveness and completion of the current strategy execution based on the feedback information, so that the reaction process has returned to normal and the key quality indicators meet the predetermined requirements.

[0031] Through the reaction monitoring management unit 16, real-time monitoring and analysis, dynamic strategy optimization and continuous feedback adjustment of the target biological reaction are achieved to improve the operating efficiency and product quality of the biological reaction process.

[0032] Furthermore, the embodiment of the present application also includes: Traversing the reaction process, and determining a first division result based on the process steps; Traversing the reaction process, and determining a second division result based on biological reaction rules and growth and metabolic trends; The first division result and the second division result are integrated to determine the multi-reaction stage.

[0033] In a feasible implementation, the reaction stage division unit 11 first traverses the complete reaction process of the target biological reaction, and identifies the conversion nodes and durations between different steps by analyzing the various steps in the process flow, such as feeding, heating, stirring, cooling, sampling, etc., and thereby obtains a preliminary segmentation of the biological reaction process to form a first division result. Then, the reaction stage division unit 11 traverses the reaction process again, focusing on analyzing the inherent regularity of biological reactions and the dynamic characteristics of microbial growth and metabolism. By consulting relevant literature and historical data, the reaction stage division unit 11 summarizes the general laws of such biological reactions, such as the lag phase, logarithmic growth phase, stable phase, decay phase, etc., as well as the typical change trends of microbial populations in terms of quantity, activity, material conversion, etc. at different stages. Based on this information, the reaction stage division unit 11 performs a secondary division of the reaction process from a biological perspective to obtain a second division result. Among them, the second division result reflects the inherent stage of the biological reaction process, and its division nodes and duration may not completely overlap with the process steps. Afterwards, the reaction stage division unit 11 comprehensively considers the first division result and the second division result, and conducts a deep comparison and fusion of the two results. By analyzing the correspondence between process steps and biological reaction stages and identifying the commonalities and differences between the two division results, the reaction stage division unit 11 combines the reaction process with biological laws to determine a unified, multi-level reaction stage division scheme including multiple reaction stages.

[0034] By making full use of information from two different sources, process information and biological knowledge, we consider both the external operating characteristics of the reaction process and the internal change laws of the biological system. Compared with the single-dimensional division method, the fusion division strategy adopted can obtain more comprehensive, reasonable and precise reaction stage division results, laying a solid foundation for realizing phased intelligent monitoring and control.

[0035] Furthermore, the embodiment of the present application also includes: Based on the first reaction stage, the stage monitoring parameters are determined as the dominant variables, and one group of dominant variables for direct monitoring and two groups of dominant variables for indirect monitoring are divided; Traversing the two groups of dominant variables, determining auxiliary variables and mining variable linear relationships based on variable correlations, wherein the auxiliary variables include at least one item; The set of dominant variables and the linear relationship of the variables based on the two sets of dominant variables are integrated to obtain training samples, and supervised training of the first reaction stage is performed based on sample drive.

[0036] In a preferred embodiment, the variable relationship mining unit 12 first obtains the key monitoring parameters that can represent the biological reaction state of the first reaction stage given by the reaction stage division unit 11 by analyzing the process characteristics and key influencing factors of the first reaction stage, and defines them as stage monitoring parameters, and uses them as the leading variables for establishing the soft monitoring model. Then, the variable relationship mining unit 12 divides the leading variables into two groups according to the physicochemical properties of each leading variable and the available monitoring means. One group of leading variables refers to variables that can be directly monitored by online sensors and other devices, such as temperature, pH value, dissolved oxygen, etc.; the second group of leading variables refers to variables that are difficult to be directly monitored online, such as cell concentration, substrate concentration, product concentration, etc. For the two groups of leading variables, since it is impossible to directly obtain real-time monitoring data, the variable relationship mining unit 12 traverses each variable, and by analyzing the correlation between it and a group of leading variables and other process parameters, it determines the parameters that are significantly linearly correlated with the target variable, and uses them as auxiliary variables of the variable. Auxiliary variables can be elements in a group of leading variables, or other process parameters that are easy to monitor online. The variable relationship mining unit 12 uses mathematical statistics methods, such as partial correlation analysis, stepwise regression, etc., to quantitatively characterize the linear relationship between the dominant variable and the auxiliary variable.

[0037] After obtaining a set of real-time monitoring data of dominant variables and the linear relationship between two sets of dominant variables and their auxiliary variables, the variable relationship mining unit 12 integrates the two to construct a complete soft monitoring training sample. Subsequently, the variable relationship mining unit 12 uses a suitable machine learning algorithm, such as a support vector machine, a neural network, etc., based on the training sample to supervise the soft monitoring model. Through the training process driven by sample data, the soft monitoring model can automatically learn and optimize the complex nonlinear relationship between each dominant variable and the auxiliary variable, forming an intelligent measurement for the first reaction stage.

[0038] Through soft monitoring model training, the problem of some key variables in the biological reaction process being difficult to monitor online was solved, and online estimation of difficult-to-measure variables was achieved using easy-to-measure variables, thus expanding the online monitoring capabilities of biological processes.

[0039] Furthermore, the embodiment of the present application also includes: Determining variable insensitivity coefficients based on the set of leading variables and auxiliary variables; Based on the insensitivity coefficient, a variable loss ratio is set, where the variable loss ratio is a threshold value of the difference between the observed data and the predicted output data of each sample in the sample data, and the insensitivity coefficient corresponds to the variable one by one; When the observed data and predicted output data of the sample are less than the variable loss ratio, it is determined that the sample has no loss; When the observed data and predicted output data of the sample are greater than or equal to the variable loss ratio, the vector value based on the variable loss ratio is used as the loss compensation amount.

[0040] In a preferred embodiment, when constructing the linear relationship between the dominant variable and the auxiliary variable, the variable relationship mining unit 12 introduces an insensitive loss function to quantify the influence of the variable measurement error on the model output and improve the robustness of the model. To obtain the insensitive loss function, the variable relationship mining unit 12 first calculates the insensitivity coefficient of each variable based on a set of dominant variables and auxiliary variables that have been determined. The insensitivity coefficient reflects the sensitivity of the model output to the measurement error of the variable. The larger the value, the stronger the model's dependence on the variable, and the greater the influence of the measurement error. Then, the variable relationship mining unit 12 sets the corresponding variable loss ratio according to the insensitivity coefficient of each variable. The variable loss ratio is defined as the threshold value of the ratio of the absolute value of the difference between the observed value of a certain variable and the predicted output value of the model in each sample in the sample data to the observed amplitude of the variable. The variable loss ratio corresponds to the insensitivity coefficient one by one. The variable with a larger insensitivity coefficient has a smaller loss ratio set to achieve key monitoring of sensitive variables. In the model verification stage, the variable relationship mining unit 12 traverses each sample data, calculates the percentage difference between the observed value of each variable and the model output value, and compares it with the preset variable loss ratio. If the percentage is less than the loss ratio, it is determined that the variable has no significant loss on the sample, and the corresponding loss function value is 0; if the percentage is greater than or equal to the loss ratio, the part exceeding the loss ratio is used as the loss compensation amount of the variable, and the corresponding loss function value is a certain measure of the compensation amount, such as the absolute value, square value, etc. Through the obtained insensitive loss function, the variable error that has a greater impact on the model performance can be accurately captured, and its loss contribution can be amplified through nonlinear transformation, while the slight error of the insensitive variable is ignored. The loss function is embedded in the model training target, making the model more robust in dealing with measurement noise and error, and reducing the impact of measurement errors of certain key variables on the model output.

[0041] Furthermore, the embodiment of the present application also includes: Traversing the dominant variables, determining point monitoring variables and continuous monitoring variables, wherein the continuous monitoring variables are marked with time zone identifiers; The continuous monitoring variables are traversed, and combined with corresponding auxiliary variables, long- and short-term memory compensation learning is performed on the soft monitoring module.

[0042] In a preferred embodiment, after completing the linear relationship mining between the dominant variable and the auxiliary variable and the training of the soft monitoring module, the variable relationship mining unit 12 performs supplementary learning of the time series characteristics of the trained soft monitoring module according to the dynamic characteristics of each variable. To this end, the variable relationship mining unit 12 first traverses all dominant variables and divides them into two categories: point monitoring variables and continuous monitoring variables according to their measurement characteristics and response requirements. Point monitoring variables refer to those variables that only need to obtain instantaneous values ​​at certain critical moments, such as temperature and pH value during the conversion of key process steps; continuous monitoring variables refer to those variables that need to continuously track their dynamic change trends, such as biomass concentration, substrate consumption rate, etc. For continuous monitoring variables, the variable relationship mining unit 12 marks the time zone to which it belongs to characterize the span characteristics of the variable on the time scale.

[0043] After determining the classification of the monitoring variables, the variable relationship mining unit 12 further traverses all the continuous monitoring variables, and combines the corresponding auxiliary variables to perform long and short-term memory compensation learning on the trained soft monitoring module. Specifically, the variable relationship mining unit 12 takes the sequence data of the continuous monitoring variables and their auxiliary variables in a certain historical period as input, and introduces the long and short-term memory neural network unit in the model to train the model to learn and extract the evolution law and related patterns of the variables on different time scales. Among them, different types of continuous monitoring variables have different related time scales and memory cycles. For example, for variables such as organism concentration that reflect the growth state of cells, it is necessary to trace their changing trends across a long period of time in order to accurately grasp the current state; while for process parameters with relatively short response cycles, it is only necessary to refer to the historical data of the recent period. The variable relationship mining unit 12 matches the optimal memory time window for different continuous monitoring variables through time zone identification, and then carries out targeted long and short-term memory compensation training based on this.

[0044] By performing long and short-term memory compensation learning, a soft monitoring module that integrates the ability to extract time series features is obtained. It can not only accurately characterize the nonlinear mapping relationship between the dominant variable and the auxiliary variable, but also adaptively mine and utilize the dynamic time series characteristics of the variables, presenting better prediction performance in continuous monitoring tasks.

[0045] Furthermore, the embodiment of the present application also includes: Traversing the abnormal reaction identifiers to determine an adjustment strategy set, wherein the adjustment strategy set includes an initial strategy and an extended strategy; Traversing the abnormal reaction identifiers, combining the correlation of abnormal reaction parameters with the multi-level subordinate divisions, and determining a multi-level game tree; Based on the adjustment strategy set and the multi-level game tree, a decision is made to determine the reaction adjustment strategy.

[0046] In a feasible implementation, the game equilibrium decision unit 15 first traverses the abnormal reaction identification sent by the data analysis and screening unit 14, and matches and screens a series of targeted adjustment strategies from the pre-set strategy library according to the attributes such as identification type and severity, forms an initial strategy, and incorporates it into the adjustment strategy set. At the same time, the game equilibrium decision unit 15 expands and derives a deformation strategy or a combination strategy according to the characteristics of each initial strategy, and incorporates it into the adjustment strategy set to form a complete candidate strategy space.

[0047] After obtaining the adjustment strategy set, the game equilibrium decision unit 15 traverses the abnormal reaction identification, analyzes the dynamic correlation characteristics between the key parameters involved in the abnormal reaction, and divides the abnormal reaction process into multiple levels accordingly. Specifically, the game equilibrium decision unit 15 extracts the first-level parameter group of the abnormal reaction according to the spatiotemporal correlation of the key parameters; then, within each first-level parameter group, further explores the higher-dimensional parameter correlation pattern and extracts the second-level, third-level and even lower-level parameter groups. Through nested correlation deconstruction, the game equilibrium decision unit 15 divides the abnormal reaction into several multi-level, tree-like expanded parameter game units to form a complete multi-level game tree.

[0048] After generating the adjustment strategy set and the multi-level game tree, the game equilibrium decision unit 15 starts to make dynamic game decisions on the adjustment strategy based on the two. Specifically, the game equilibrium decision unit 15 first maps each strategy in the adjustment strategy set to the corresponding node of the multi-level game tree, and assigns a corresponding utility function to each strategy node according to attributes such as strategy complexity and scope of influence. Then, the game equilibrium decision unit 15 deduce and analyze the multi-level game tree layer by layer, examines the trade-offs of different strategy combinations in each game unit, and uses game equilibrium solving tools such as Nash equilibrium and Pareto optimality to determine the local optimal strategy of each game unit. Afterwards, the game equilibrium decision unit 15 integrates the local equilibrium strategies of each game unit, and finally determines the global optimal response adjustment strategy through cross-layer trade-offs and top-down coordination.

[0049] Through intelligent regulation of abnormal reactions based on dynamic game of strategy sets, the multi-parameter interaction characteristics of abnormal reactions are fully considered, and intelligent trade-offs between adjustment strategies at different spatiotemporal scales and different optimization objectives are achieved, thereby improving the effectiveness and global optimality of abnormal regulation.

[0050] Furthermore, the embodiment of the present application also includes: Initializing the multi-level game tree based on the adjustment strategy set; Based on the initialized first-level game layer of the multi-level game tree, combined with the profit function, a first-level strategy point is determined, wherein the first-level game layer is determined from bottom to top; Based on the primary strategy point, a multi-level game is conducted from bottom to top to determine the reaction adjustment strategy.

[0051] In a preferred embodiment, first, based on the generated regulation strategy set, the multi-level game tree is initialized and configured. Traverse each strategy in the regulation strategy set and map it to different level nodes of the multi-level game tree according to its complexity and abstraction. Among them, the more abstract and wider the overall strategy is, the higher the level of the strategy node is placed, while the more specific and targeted sub-strategies are merged into the relatively lower level strategy nodes. Through the hierarchical and recursive strategy configuration, the step-by-step expansion from coarse-grained regulation to fine-grained regulation is realized, and the hierarchical decision-making framework of the multi-level game tree is initially established. After completing the initialization of the multi-level game tree, the game decision-making process is deduced from the bottom to the top. First, focus on the bottom layer of the game tree, that is, the first-level game layer. In this layer, the specific regulation sub-strategies attached to each bottom node are examined respectively, and a benefit function is set for each sub-strategy node based on the historical data and expert experience of the reaction process. The benefit function quantitatively evaluates the expected utility that can be achieved by different combinations of regulation behaviors, and is the basis for conducting strategic game analysis. By solving the Nash equilibrium point of each node strategy, the optimal strategy selection in each primary game unit is determined, and these local optimal solutions are marked as primary strategy points.

[0052] After determining the first-level strategy point, multi-level game decision-making is advanced upward layer by layer. For each high-level game node, it is regarded as a multi-party game problem composed of its subordinate subtrees. By feeding back the first-level strategy points and their utility information of the lower layer to the current layer, the comprehensive benefits of different overall strategies are accurately evaluated, and the equilibrium control scheme of each game node is determined by using methods such as Stackelberg game and hierarchical analysis. This iteration is repeated until the root node of the game tree is reached, and finally the globally optimal top-level control strategy is determined, and it is expanded from top to bottom with the equilibrium point strategies of each layer to form a complete multi-level linkage control scheme as a reaction regulation strategy.

[0053] Through bottom-up local equilibrium solving and top-down global collaborative optimization, dynamic trade-offs and integrated decision-making of regulatory behaviors are achieved at multiple strategy granularity levels, fully utilizing information and resources at different decision-making levels to achieve a system-level balance between local benefits and overall effectiveness, thereby achieving a comprehensive improvement in overall performance.

[0054] In summary, the biological reaction intelligent monitoring system provided by the embodiment of the present application has the following technical effects: The reaction stage division unit is used to divide and determine the reaction process of the interactive target biological reaction, and provide a more refined process description for subsequent variable analysis and monitoring control. The variable relationship mining unit is used to traverse multiple reaction stages, determine the dominant variables and auxiliary variables, mine the linear relationship of variables, coordinate the insensitive loss function, supervise the training of the soft monitoring module, and improve the accuracy of the soft monitoring module. The continuous monitoring acquisition unit is used to traverse the dominant variables and auxiliary variables, configure the monitoring equipment group that directly monitors the variables, and continuously monitors the reaction process to determine the reaction monitoring data, so as to improve the timeliness and integrity of data collection. The data analysis and screening unit is used to pre-process the reaction monitoring data at the front end, and transmit it back to the soft monitoring module for linear analysis and abnormal reaction location, determine the effective monitoring data, and the effective monitoring data includes the stage variable data and the abnormal reaction identification. The effective monitoring data containing the stage variable data and the abnormal reaction identification is screened out to provide a reliable basis for subsequent decision-making and control. The game equilibrium decision unit is used to traverse the effective monitoring data, make game equilibrium decisions on the abnormal reaction identification, determine the reaction adjustment strategy, and realize intelligent abnormal reaction regulation through multi-objective optimization and equilibrium analysis. The reaction monitoring management unit is used to respond to the reaction container reaction adjustment strategy, and combines with the soft monitoring module to conduct feedback monitoring and analysis, conduct reaction monitoring management of the target biological reaction, realize the full-process monitoring and management of the target biological reaction, improve the accuracy of biological reaction monitoring, realize real-time online monitoring, and perform intelligent regulation and control for abnormal situations.

[0055] Embodiment 2 Based on the same inventive concept as the biological reaction intelligent monitoring system in the aforementioned embodiment, Figure 2 As shown, the embodiment of the present application provides a method for intelligent monitoring of biological reactions, the method comprising: The reaction process of interactive target biological reactions is divided into multiple reaction stages; Traversing the multi-reaction stages, determining the dominant variables and the auxiliary variables, mining the linear relationship of the variables, coordinating the insensitive loss function, and supervising the training of the soft monitoring module; Traversing the leading variables and the auxiliary variables, configuring a monitoring device group for directly monitoring the variables, continuously monitoring the reaction process, and determining reaction monitoring data, wherein the reaction monitoring data is marked with stage nodes; The front end pre-processes the reaction monitoring data and transmits it back to the soft monitoring module for linear analysis and abnormal reaction location to determine effective monitoring data, which includes stage variable data and abnormal reaction identification; Traversing the effective monitoring data, making a game equilibrium decision on the abnormal reaction identifier, and determining a reaction adjustment strategy; The reaction container responds to the reaction regulation strategy and performs feedback monitoring and analysis in conjunction with the soft monitoring module to perform reaction monitoring management of the target biological reaction.

[0056] Further, the division determines multiple reaction stages, including: Traversing the reaction process, and determining a first division result based on the process steps; Traversing the reaction process, and determining a second division result based on biological reaction rules and growth and metabolic trends; The first division result and the second division result are integrated to determine the multi-reaction stage.

[0057] Furthermore, the supervised training soft monitoring module includes: Based on the first reaction stage, the stage monitoring parameters are determined as the dominant variables, and one group of dominant variables for direct monitoring and two groups of dominant variables for indirect monitoring are divided; Traversing the two groups of dominant variables, determining auxiliary variables and mining variable linear relationships based on variable correlations, wherein the auxiliary variables include at least one item; The set of dominant variables and the linear relationship of the variables based on the two sets of dominant variables are integrated to obtain training samples, and supervised training of the first reaction stage is performed based on sample drive.

[0058] Furthermore, an insensitive loss function is obtained, including: Determining variable insensitivity coefficients based on the set of leading variables and auxiliary variables; Based on the insensitivity coefficient, a variable loss ratio is set, where the variable loss ratio is a threshold value of the difference between the observed data and the predicted output data of each sample in the sample data, and the insensitivity coefficient corresponds to the variable one by one; When the observed data and predicted output data of the sample are less than the variable loss ratio, it is determined that the sample has no loss; When the observed data and predicted output data of the sample are greater than or equal to the variable loss ratio, the vector value based on the variable loss ratio is used as the loss compensation amount.

[0059] Furthermore, after the supervised training soft monitoring module, the following steps are included: Traversing the dominant variables, determining point monitoring variables and continuous monitoring variables, wherein the continuous monitoring variables are marked with time zone identifiers; The continuous monitoring variables are traversed, and combined with corresponding auxiliary variables, long- and short-term memory compensation learning is performed on the soft monitoring module.

[0060] Further, the determining of the reaction adjustment strategy includes: Traversing the abnormal reaction identifiers to determine an adjustment strategy set, wherein the adjustment strategy set includes an initial strategy and an extended strategy; Traversing the abnormal reaction identifiers, combining the correlation of abnormal reaction parameters with the multi-level subordinate divisions, and determining a multi-level game tree; Based on the adjustment strategy set and the multi-level game tree, a decision is made to determine the reaction adjustment strategy.

[0061] Further, based on the adjustment strategy set and the multi-level game tree, the reaction adjustment strategy is determined by decision, including: Initializing the multi-level game tree based on the adjustment strategy set; Based on the initialized first-level game layer of the multi-level game tree, combined with the profit function, a first-level strategy point is determined, wherein the first-level game layer is determined from bottom to top; Based on the primary strategy point, a multi-level game is conducted from bottom to top to determine the reaction adjustment strategy.

[0062] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0063] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these changes and variations.

Claims

1. A biological reaction intelligent monitoring system, characterized in that: The system comprises: A reaction stage division unit is used to divide and determine multiple reaction stages in the reaction process of the interactive target biological reaction; Variable relationship mining unit, used to traverse multiple reaction stages, determine the dominant variables and auxiliary variables, mine the linear relationship of variables, coordinate insensitive loss functions, and supervise the training of soft monitoring modules; The continuous monitoring acquisition unit is used to traverse the leading variables and auxiliary variables, configure the monitoring equipment group for directly monitoring the variables, continuously monitor the reaction process, and determine the reaction monitoring data; The data analysis and screening unit is used to pre-process the reaction monitoring data at the front end, and transmit it back to the soft monitoring module for linear analysis and abnormal reaction location, and determine the effective monitoring data. The effective monitoring data includes stage variable data and abnormal reaction identification; The game equilibrium decision unit is used to traverse the effective monitoring data, make game equilibrium decisions on abnormal reaction identification, and determine the reaction adjustment strategy; The reaction monitoring management unit is used to monitor the reaction container response reaction adjustment strategy, and combines with the soft monitoring module to perform feedback monitoring and analysis to perform reaction monitoring management of the target biological reaction.

2. A biological reaction intelligent monitoring system as claimed in claim 1, characterized in that: The division determines multiple reaction stages, including: Traversing the reaction process, and determining a first division result based on the process steps; Traversing the reaction process, and determining a second division result based on biological reaction rules and growth and metabolic trends; The first division result and the second division result are integrated to determine the multi-reaction stage.

3. A biological reaction intelligent monitoring system as claimed in claim 1, characterized in that: The supervised training soft monitoring module comprises: Based on the first reaction stage, the stage monitoring parameters are determined as the dominant variables, and one group of dominant variables for direct monitoring and two groups of dominant variables for indirect monitoring are divided; Traversing the two groups of dominant variables, determining auxiliary variables and mining variable linear relationships based on variable correlations, wherein the auxiliary variables include at least one item; The set of dominant variables and the linear relationship of the variables based on the two sets of dominant variables are integrated to obtain training samples, and supervised training of the first reaction stage is performed based on sample drive.

4. A biological reaction intelligent monitoring system as claimed in claim 3, characterized in that: Get insensitive loss functions, including: Determining variable insensitivity coefficients based on the set of leading variables and auxiliary variables; Based on the insensitivity coefficient, a variable loss ratio is set, where the variable loss ratio is a threshold value of the difference between the observed data and the predicted output data of each sample in the sample data, and the insensitivity coefficient corresponds to the variable one by one; When the observed data and predicted output data of the sample are less than the variable loss ratio, it is determined that the sample has no loss; When the observed data and predicted output data of the sample are greater than or equal to the variable loss ratio, the vector value based on the variable loss ratio is used as the loss compensation amount.

5. The biological reaction intelligent monitoring system according to claim 1, characterized in that: After the supervised training soft monitoring module, it includes: Traversing the dominant variables, determining point monitoring variables and continuous monitoring variables, wherein the continuous monitoring variables are marked with time zone identifiers; The continuous monitoring variables are traversed, and combined with corresponding auxiliary variables, long- and short-term memory compensation learning is performed on the soft monitoring module.

6. A biological reaction intelligent monitoring system as claimed in claim 1, characterized in that: The determining of the reaction adjustment strategy comprises: Traversing the abnormal reaction identifiers to determine an adjustment strategy set, wherein the adjustment strategy set includes an initial strategy and an extended strategy; Traversing the abnormal reaction identifiers, combining the correlation of abnormal reaction parameters with the multi-level subordinate divisions, and determining a multi-level game tree; Based on the adjustment strategy set and the multi-level game tree, a decision is made to determine the reaction adjustment strategy.

7. A biological reaction intelligent monitoring system as claimed in claim 6, characterized in that: Based on the adjustment strategy set and the multi-level game tree, the reaction adjustment strategy is determined by decision, including: Initializing the multi-level game tree based on the adjustment strategy set; Based on the initialized first-level game layer of the multi-level game tree, combined with the profit function, a first-level strategy point is determined, wherein the first-level game layer is determined from bottom to top; Based on the primary strategy point, a multi-level game is conducted from bottom to top to determine the reaction adjustment strategy.

8. A biological reaction intelligent monitoring method, characterized in that: The method comprises: The reaction process of interactive target biological reactions is divided into multiple reaction stages; Traversing the multi-reaction stages, determining the dominant variables and the auxiliary variables, mining the linear relationship of the variables, coordinating the insensitive loss function, and supervising the training of the soft monitoring module; Traversing the leading variables and the auxiliary variables, configuring a monitoring device group for directly monitoring the variables, continuously monitoring the reaction process, and determining reaction monitoring data, wherein the reaction monitoring data is marked with stage nodes; The front end pre-processes the reaction monitoring data and transmits it back to the soft monitoring module for linear analysis and abnormal reaction location to determine effective monitoring data, which includes stage variable data and abnormal reaction identification; Traversing the effective monitoring data, making a game equilibrium decision on the abnormal reaction identifier, and determining a reaction adjustment strategy; The reaction container responds to the reaction regulation strategy and performs feedback monitoring and analysis in conjunction with the soft monitoring module to perform reaction monitoring management of the target biological reaction.

Images