Laboratory environment regulation and control auxiliary system based on artificial intelligence

Abstract

The invention discloses a laboratory environment regulation and control auxiliary system based on artificial intelligence. The system comprises a data acquisition module, a laboratory scene recognition module, a laboratory environment suitability evaluation module, a laboratory environment regulation and control strategy generation module and an environment intelligent regulation and control auxiliary module. The invention relates to the technical field of data processing, in particular to a laboratory environment regulation and control auxiliary system based on artificial intelligence, which innovatively introduces a laboratory scene recognition module to recognize a laboratory scene in real time and effectively improve the environment regulation and control precision; a clustering center based on mahalanobis distance is introduced to initialize, a laboratory scene recognition objective function fused with prior reinforcement constraints is constructed, a weak correlation clustering center recognition strategy and membership redistribution optimization are used for carrying out clustering algorithm improvement, and the accuracy of laboratory scene recognition is improved; and a historical optimal position memory mechanism is introduced, and a dimension learning neighborhood search strategy is adopted to improve an optimization algorithm, so that the precision of a laboratory environment regulation and control strategy is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an artificial intelligence-based laboratory environment control auxiliary system. Background Technology

[0002] The AI-based laboratory environment control auxiliary system refers to an intelligent management system for laboratory environments built on artificial intelligence. It is applied to real-time monitoring, risk assessment, and control decision support of laboratory environment operation status. The system collects multi-source data from the laboratory, uses artificial intelligence algorithms to process data and generate intelligent control strategies, and executes the control strategies to achieve intelligent collaborative management and optimization decision-making for laboratory environmental quality, thereby improving the safety, comfort, and energy efficiency of laboratory operation.

[0003] However, traditional laboratory environmental control systems suffer from several technical problems. They lack the ability to identify laboratory experimental scenarios and cannot adaptively adjust control strategies to dynamically change with the experimental process. This leads to lagging environmental control and higher risks in laboratory operation. Existing models for laboratory scenario identification suffer from issues such as random initialization of cluster centers, insufficient utilization of historical annotations and prior information, and lack of differentiation for weakly correlated cluster centers. These problems result in clustering results easily getting trapped in local optima and low scenario identification accuracy, leading to unstable laboratory scenario segmentation and inaccurate triggering of environmental control strategies. Furthermore, existing algorithms for optimizing laboratory environmental control parameters suffer from problems such as the optimization process easily getting trapped in local optima, insufficient population diversity, low search efficiency, and long time consumption in obtaining the optimal environmental control strategy, resulting in slow response times in laboratory environmental control. Summary of the Invention

[0004] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides an AI-based laboratory environment control auxiliary system. Addressing the technical problems of traditional laboratory environment control systems, such as a lack of ability to identify laboratory experimental scenarios and the inability of control strategies to adapt dynamically to changes in the experimental process, leading to lagging environmental control and high operational risks, this solution innovatively introduces a laboratory scenario identification module. This module performs real-time classification and identification of laboratory scenarios, enabling differentiated environmental control strategies based on different experimental scenarios. This improves the foresight, precision, and intelligence of environmental control, effectively enhancing its accuracy and significantly strengthening the laboratory's responsiveness to high-precision experimental scenarios, thus achieving safe, stable, and efficient laboratory environment operation. Furthermore, addressing the technical problems in existing laboratory scenario identification models, such as random initialization of cluster centers, insufficient utilization of historical annotation prior information, and lack of differentiation of weakly correlated cluster centers, which lead to clustering results easily getting trapped in local optima and low scenario identification accuracy, resulting in unstable laboratory scenario segmentation and inaccurate triggering of environmental control strategies, this solution innovatively introduces a Mahalanobis distance-based initialization and construction of cluster centers. This paper improves the clustering algorithm by integrating a laboratory scene recognition objective function with prior reinforcement constraints, a weakly correlated cluster center recognition strategy, and membership redistribution optimization. This improves the rationality and stability of cluster center initialization, enhances the utilization of prior information from historical scene annotations, improves the discriminative power of key scene cluster centers and the efficiency of clustering computation, effectively suppresses the interference of weakly correlated scene centers on membership assignment, and significantly improves the accuracy and reliability of laboratory scene recognition. This provides a more reliable foundation for subsequent environmental suitability assessment and environmental control strategy generation. Addressing the technical problems of existing algorithms for optimizing laboratory environmental control parameters, such as easy getting trapped in local optima, insufficient population diversity, low search efficiency, and long time consumption in obtaining the optimal environmental control strategy, leading to slow response in laboratory environmental control, this solution innovatively introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm. This effectively improves the stability, convergence speed, and solution quality of the environmental control parameter optimization process, significantly improves the rationality, accuracy, and robustness of laboratory environmental control strategies, and enables rapid and intelligent control of the laboratory environment under complex experimental conditions.

[0005] The technical solution adopted by the present invention is as follows: The laboratory environment control auxiliary system based on artificial intelligence provided by the present invention includes a data acquisition module, a laboratory scene recognition module, a laboratory environment suitability assessment module, a laboratory environment control strategy generation module, and an environment intelligent control auxiliary module;

[0006] The data acquisition module specifically obtains optimized laboratory environment control data through data acquisition and data optimization operations.

[0007] The laboratory scene recognition module is used to automatically identify the current laboratory scene type, providing scene basis for laboratory environment suitability assessment and laboratory environment control strategy generation. Specifically, it introduces cluster center initialization based on Mahalanobis distance, constructs a laboratory scene recognition objective function that integrates prior reinforcement constraints, a weak correlation cluster center recognition strategy, and membership degree reassignment optimization to improve the clustering algorithm. The improved clustering algorithm is used to construct a laboratory scene recognition model, and real-time data is input into the model to obtain real-time laboratory scene recognition results.

[0008] The laboratory environment suitability assessment module is used to assess whether the current laboratory environment is suitable for the target experiment based on the real-time laboratory scene recognition results. Specifically, it constructs and trains a laboratory environment suitability assessment model, then performs a real-time laboratory environment suitability assessment, outputs the real-time laboratory environment suitability assessment results, and makes a laboratory environment suitability judgment based on the assessment results to obtain the real-time laboratory environment status.

[0009] The laboratory environment control strategy generation module is used to generate a laboratory environment control strategy adapted to the current laboratory scene based on the real-time laboratory scene recognition results and the laboratory environment suitability judgment results. Specifically, it first selects a laboratory environment control template strategy, then sets the boundary range of the environment control parameters, then introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm. The improved optimization algorithm optimizes and searches for the environment control parameters to obtain the optimal combination of environment control parameters for the current laboratory scene. Finally, it generates the optimal environment control strategy for the current laboratory scene based on the optimal combination of environment control parameters.

[0010] The intelligent environmental control auxiliary module specifically adjusts the environmental control strategy based on the real-time laboratory scene recognition results and the real-time laboratory environment status, and executes the control measures to achieve intelligent control of the laboratory environment.

[0011] Furthermore, the data acquisition module specifically involves obtaining raw data on laboratory environmental control through data acquisition operations in the laboratory management platform, and performing data optimization processing on the raw data on laboratory environmental control to obtain optimized data on laboratory environmental control.

[0012] The raw data for laboratory environment control includes historical laboratory scene identification data, real-time laboratory scene identification data, historical laboratory environment suitability assessment data, and real-time laboratory environment suitability assessment data.

[0013] The data optimization process includes data cleaning, data normalization, data encoding, and data feature selection.

[0014] Furthermore, the laboratory scene recognition module specifically includes the following steps:

[0015] The construction of a laboratory scene recognition model includes the following steps:

[0016] Cluster center initialization involves calculating the Mahalanobis distance from each sample data point to the origin of the laboratory scene feature space, sorting all sample data points in ascending order of Mahalanobis distance, and then dividing the sorted data points into K groups. Finally, the average position of all sample data points in each group is used as the initial cluster center for that group, resulting in an initial cluster center set. Based on the initial cluster centers, the initial membership matrix is ​​calculated using the membership formula in the standard fuzzy C-means clustering algorithm.

[0017] A laboratory scene recognition objective function is constructed, specifically by constructing a laboratory scene recognition objective function that integrates prior reinforcement constraints; the formula used is as follows:

[0018] ;

[0019] In the formula, This represents the objective function value for clustering in the laboratory scene recognition at the t-th iteration. This represents the weight coefficient of the feature clustering item in the unsupervised scenario. Indicates the number of sample data points. This represents the membership value of the i-th sample data point to the j-th laboratory scenario at the t-th iteration. This represents the i-th sample data point. This represents the cluster center of the j-th type of laboratory scene at the t-th iteration. This represents the weight coefficient of the semi-supervised prior reinforcement constraint term. Indicates the number of cluster categories. This represents the prior membership value of the i-th sample data point to the j-th laboratory scenario at the t-th iteration. Represents the binary identifier of the sample data point label;

[0020] The membership matrix is ​​iteratively updated by combining the objective function for laboratory scene recognition with the core constraints of fuzzy clustering, constructing an unconstrained Lagrangian function using the Lagrange multiplier method, transforming the constrained optimization problem into an unconstrained optimization problem, and introducing a prior reinforcement bias derivative term to iteratively update the membership matrix; the formula used is as follows:

[0021] ;

[0022] In the formula, Indicates the first In the next iteration, the membership value of the i-th sample data point to the j-th type of laboratory scenario is... This represents the cluster center of the k-th type of laboratory scene at the t-th iteration. This represents the prior membership value of the i-th data point to the k-th laboratory scenario at the t-th iteration. This represents a prior reinforcement bias derivative;

[0023] The cluster center matrix is ​​updated by updating the membership matrix and updating the cluster center for each class of laboratory scenarios.

[0024] Weak correlation cluster center identification, specifically, first for each sample data point Calculate the nearest cluster center in the t-th iteration, and then calculate the sample data points respectively. The feature difference of the nearest cluster center, the iterative position deviation of each cluster center and the nearest cluster center are used to finally select weakly correlated cluster centers that have no significant impact on the laboratory scene attribution result of the sample data point, and construct a set of weakly correlated cluster centers.

[0025] The membership degree redistribution optimization is specifically achieved by performing differential scaling on the membership degrees of weakly correlated cluster centers of different types of sample data points based on the binary identifiers of the sample data point labels and the set of weakly correlated cluster centers, and then normalizing the membership degrees of strongly correlated centers by combining the corrected membership degrees to obtain the optimized membership degree matrix.

[0026] ;

[0027] In the formula, Indicates the first In the next iteration, the optimized membership value of the i-th sample data point belonging to the j-th type of laboratory scenario is determined. Indicates the first In the next iteration, the corrected membership value of the i-th sample to the c-th weakly correlated cluster center is determined, where c represents the scene category index within the set of weakly correlated cluster centers. Indicates the first In the next iteration, the membership value of the i-th sample to the c-th weakly correlated cluster center is... Indicates the first In the next iteration, the set of weakly correlated cluster centers corresponding to the i-th sample data point;

[0028] The clustering iteration terminates, specifically by recalculating the objective function for identifying the laboratory scene. The difference between the objective function value and the objective function value of the previous iteration is calculated to obtain the change value of the objective function. Finally, the iteration termination judgment is made according to the preset termination condition. If any clustering iteration termination condition is met, the clustering iteration is stopped and the final cluster center matrix and the final optimized membership matrix are output.

[0029] The scene recognition clustering results are output as follows: based on the final optimized membership matrix, for each sample data point, the optimized membership value matching criterion is adopted, and the scene category with the largest value among the optimized membership values ​​of the corresponding laboratory scenes is selected as the final scene clustering category of the sample. In this way, all sample data points are divided into multiple independent clusters according to their respective scene categories, and the scene recognition clustering results are obtained.

[0030] Real-time laboratory scene recognition involves inputting the laboratory scene recognition feature dataset into the laboratory scene recognition model to generate real-time scene recognition clustering results. Based on the real-time scene recognition clustering results, the cluster labels in each cluster are statistically analyzed, and the cluster label with the highest frequency in each cluster is selected as the laboratory scene recognition label for that cluster, thus obtaining the real-time laboratory scene recognition result.

[0031] Furthermore, the laboratory environment suitability assessment module specifically includes the following steps:

[0032] A laboratory environment suitability assessment model was constructed and trained. Specifically, the model was built based on a bidirectional long short-term memory neural network. The historical laboratory environment suitability assessment feature dataset and historical laboratory scene recognition results were used as training data for the assessment model to obtain the trained laboratory environment suitability assessment model.

[0033] Real-time laboratory environment suitability assessment specifically involves inputting a real-time laboratory environment suitability assessment feature dataset and real-time laboratory scene recognition results into a trained laboratory environment suitability assessment model to obtain real-time laboratory environment suitability assessment results.

[0034] The laboratory environment suitability assessment is specifically based on the real-time laboratory environment suitability assessment results and preset environmental suitability thresholds and environmental hazard thresholds to classify the laboratory environment status and obtain the real-time laboratory environment status.

[0035] Furthermore, the laboratory environment control strategy generation module specifically includes the following steps:

[0036] The laboratory environment control template strategy is selected. Specifically, based on the real-time laboratory scene recognition results, the laboratory environment control strategy data corresponding to the current laboratory scene is matched in the preset laboratory scene environment control template strategy library, and the typical set values ​​of the environmental control parameters are extracted from them as the initial reference values ​​for the optimization of the environmental control parameters.

[0037] Setting the boundary range of environmental control parameters involves retrieving the upper and lower limits of each environmental control parameter from the laboratory scene boundary mapping table based on the real-time laboratory scene recognition results, and generating the dynamic boundary range of each environmental control parameter accordingly.

[0038] Optimizing environmental control parameters involves using an improved optimization algorithm to search for the optimal combination of environmental control parameters within the boundary range of each parameter, thereby obtaining the best combination of parameters suitable for the current laboratory environment. This includes the following steps:

[0039] The initialization of the search population specifically involves using the laboratory environmental control strategy as the individual position vector in the optimization algorithm, and generating the vector by perturbation based on the initial reference value of the environmental control parameters as the center, combined with the boundary interval of each environmental control parameter. Individual location vectors for each search Thus, the initial search population is obtained;

[0040] The individual fitness value is calculated by using the individual environmental suitability assessment result as the individual fitness value, sorting all search individuals in the population in descending order according to the fitness value, and defining three types of guide individuals and their positions.

[0041] The initial update of individual locations involves guiding individuals based on three categories and assigning contribution weights to their location information using a dynamic weighting strategy. This initial update yields the initial location of the i-th search individual in the i-th search position. Initial position during iteration ;

[0042] Individual memory position update specifically involves introducing a historical optimal position memory mechanism. Based on the individual's historical movement speed from the previous iteration, the influence of historical trends is controlled through inertia weights. Simultaneously, the deviation between the individual's local movement position and its initial position is calculated using three types of guidance. Combined with learning factors and random parameters, the individual's movement speed direction is corrected. Finally, the individual memory position is updated to obtain the position of the i-th search individual in the i-th iteration. Memorizing and updating positions during iteration ;

[0043] Individual dimensional position updates are achieved by employing a dimensional learning neighborhood search strategy to calculate... and The Euclidean distance is used as the neighborhood search radius for the individual. Then, based on the neighborhood search radius, the population is traversed except for the individual. All individuals other than those selected, those with... Individuals whose Euclidean distance is less than the individual's neighborhood search radius constitute the individual's neighborhood set. Then, for each dimension of each individual, the position update of the individual's single dimension is completed by combining the dimensional information of the neighborhood individuals and the random individuals in the population. After traversing all dimensions, the update results of each dimension are combined to generate the complete dimensional update position of the individual.

[0044] Iterative dual-position filtering, specifically, calculating the position of the i-th search individual at the i-th position. In the next iteration, the fitness values ​​of the memory update position and the dimension update position are compared. The fitness values ​​of the two candidate positions are then selected based on the principle of better fitness as the position for that individual in the [number]th iteration. Final position of the next iteration This completes the position optimization closed loop for a single iteration.

[0045] The search for the optimal position of an individual involves evaluating the fitness values ​​of all individuals in the current population after each iteration. If the fitness of an individual's position is better than that of the current global optimal individual position, then that individual's global optimal position is updated.

[0046] The search iteration terminates when the individual's fitness value is higher than the fitness threshold or when the maximum number of iterations is reached, and the search is terminated and the individual's global optimal position is obtained. The individual's global optimal position specifically refers to the optimal combination of environmental control parameters in the current laboratory scenario.

[0047] The optimal environmental control strategy for the laboratory is output, specifically, based on the optimal combination of environmental control parameters for the current laboratory scenario, the optimal environmental control strategy for the current laboratory scenario is generated and output.

[0048] Furthermore, the intelligent environmental control auxiliary module specifically monitors the real-time laboratory scene recognition results and the real-time laboratory environment status, and intelligently adjusts the laboratory environment control strategy according to different triggering conditions, and performs coordinated control of the laboratory environment according to the current environmental control strategy.

[0049] The beneficial effects achieved by the present invention using the above solution are as follows:

[0050] (1) In view of the technical problems that traditional laboratory environmental control systems lack the ability to identify laboratory experimental scenarios and cannot adaptively adjust laboratory control strategies according to the dynamic changes in the experimental process, resulting in lagging environmental control and high laboratory operation risks, this solution innovatively introduces a laboratory scenario identification module to classify and identify laboratory scenarios in real time. This enables environmental control to adopt differentiated environmental control strategies according to different experimental scenarios, improves the foresight, precision and intelligence of environmental control, effectively improves the accuracy of environmental control, and significantly enhances the laboratory's response to high-precision experimental scenarios, so as to achieve safe, stable and efficient operation of the laboratory environment.

[0051] (2) To address the technical problems in existing laboratory scene recognition models, such as random initialization of cluster centers, insufficient utilization of prior information from historical annotations, and lack of differentiation of weakly correlated cluster centers, which lead to clustering results easily falling into local optima and low scene recognition accuracy, resulting in unstable laboratory scene division and inaccurate triggering of environmental control strategies, this solution innovatively introduces cluster center initialization based on Mahalanobis distance, constructs a laboratory scene recognition objective function that integrates prior reinforcement constraints, a weakly correlated cluster center recognition strategy, and membership degree redistribution optimization to improve the clustering algorithm. This improves the rationality and stability of cluster center initialization, enhances the ability to utilize prior information from historical scene annotations, improves the discriminability of key scene cluster centers and the efficiency of clustering calculation, effectively suppresses the interference of weakly correlated scene centers on membership degree allocation, and significantly improves the accuracy and reliability of laboratory scene recognition, providing a more reliable scene recognition foundation for subsequent environmental suitability assessment and environmental control strategy generation.

[0052] (3) In view of the technical problems in the existing optimization algorithms for laboratory environmental control parameters, such as easy getting trapped in local optima, insufficient population diversity, low search efficiency, and long time to obtain the optimal environmental control strategy, which leads to slow response of laboratory environmental control, this scheme innovatively introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm. This effectively improves the stability, convergence speed and solution quality of the environmental control parameter optimization process, significantly improves the rationality, accuracy and robustness of the laboratory environmental control strategy, and realizes rapid and intelligent control of the laboratory environment under complex experimental conditions. Attached Figure Description

[0053] Figure 1 A schematic diagram of the modules of the artificial intelligence-based laboratory environment control auxiliary system provided by the present invention;

[0054] Figure 2 A flowchart illustrating the process of building a laboratory scene recognition model in the laboratory scene recognition module;

[0055] Figure 3 A flowchart illustrating the module for generating laboratory environment control strategies;

[0056] Figure 4 A flowchart illustrating the process of optimizing environmental control parameters in the laboratory environmental control strategy generation module;

[0057] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0058] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0059] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0060] Example 1, see Figure 1 The artificial intelligence-based laboratory environment control auxiliary system provided by the present invention includes a data acquisition module, a laboratory scene recognition module, a laboratory environment suitability assessment module, a laboratory environment control strategy generation module, and an intelligent environment control auxiliary module.

[0061] The data acquisition module specifically obtains laboratory environment control optimization data through data acquisition and data optimization operations, and sends the data to the laboratory scene recognition module and the laboratory environment suitability assessment module.

[0062] The laboratory scene recognition module receives data sent by the data acquisition module to automatically identify the current laboratory scene type, providing scene basis for laboratory environment suitability assessment and laboratory environment control strategy generation. Specifically, it introduces cluster center initialization based on Mahalanobis distance, constructs a laboratory scene recognition objective function that integrates prior reinforcement constraints, a weak correlation cluster center recognition strategy, and membership degree redistribution optimization to improve the clustering algorithm. The improved clustering algorithm is used to construct a laboratory scene recognition model, and real-time data is input into the model to perform real-time laboratory scene recognition, obtain real-time laboratory scene recognition results, and send the data to the laboratory environment suitability assessment module, the laboratory environment control strategy generation module, and the environmental intelligent control auxiliary module.

[0063] The laboratory environment suitability assessment module receives data from the data acquisition module, the laboratory scene recognition module, and the laboratory environment control strategy generation module. Based on the real-time laboratory scene recognition results, it assesses whether the current experimental environment is suitable for the target experiment. Specifically, it constructs and trains a laboratory environment suitability assessment model, then performs a real-time laboratory environment suitability assessment, outputs the real-time laboratory environment suitability assessment results, makes a laboratory environment suitability judgment based on the assessment results, obtains the real-time laboratory environment status, and sends the data to the laboratory environment control strategy generation module and the environmental intelligent control auxiliary module.

[0064] The laboratory environment control strategy generation module receives data sent by the laboratory environment suitability assessment module. Based on the real-time laboratory scene identification results and laboratory environment suitability judgment results, it generates a laboratory environment control strategy adapted to the current laboratory scene. Specifically, it first selects a laboratory environment control template strategy, then sets the boundary range of environmental control parameters, then introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm. Under the parameter boundary constraints, the improved optimization algorithm optimizes and searches for the environmental control parameters to obtain the optimal combination of environmental control parameters for the current laboratory scene. Finally, it generates the optimal environmental control strategy for the current laboratory scene based on the optimal environmental control parameter group and sends the data to the laboratory environment suitability assessment module and the environmental intelligent control auxiliary module.

[0065] The intelligent environmental control auxiliary module receives data from the laboratory scene recognition module, the laboratory environment suitability assessment module, and the laboratory environment control strategy generation module. Specifically, it dynamically adjusts the environmental control strategy based on the real-time laboratory scene recognition results and the real-time laboratory environment status, and performs coordinated control of the laboratory environment to achieve intelligent control of the laboratory environment.

[0066] By performing the above operations, this solution addresses the technical problems of traditional laboratory environmental control systems, such as the lack of ability to identify laboratory experimental scenarios and the inability of laboratory control strategies to adapt to dynamic changes in the experimental process, resulting in lagging environmental control and high laboratory operational risks. It innovatively introduces a laboratory scenario identification module to classify and identify laboratory scenarios in real time. This enables environmental control to adopt differentiated environmental control strategies based on different experimental scenarios, improving the foresight, precision, and intelligence of environmental control, effectively enhancing the accuracy of environmental control, and significantly strengthening the laboratory's responsiveness to high-precision experimental scenarios, thus achieving safe, stable, and efficient operation of the laboratory environment.

[0067] Example 2, see Figure 1This embodiment is based on the above embodiment. Specifically, the data acquisition module obtains the original data of laboratory environment control through data acquisition operation in the laboratory management platform, and performs data optimization processing on the original data of laboratory environment control to obtain optimized data of laboratory environment control.

[0068] The raw data for laboratory environment control includes historical laboratory scene identification data, real-time laboratory scene identification data, historical laboratory environment suitability assessment data, and real-time laboratory environment suitability assessment data.

[0069] The data optimization process includes data cleaning, data normalization, data encoding, and data feature selection.

[0070] Both the historical laboratory scene recognition data and the real-time laboratory scene recognition data include current experimental information data, current experimental personnel data, and current experimental equipment data.

[0071] The historical laboratory scene identification data also includes historical laboratory scene identification results; the laboratory scene identification results refer to the scene determination information of the current laboratory, and this is used as a cluster label, including idle standby experimental scene, personnel on-site preparation scene, routine experimental operation scene, local pollutant generation scene, and precision experimental operation scene.

[0072] The current experimental information data includes the experiment number, experiment area number, experiment type, experiment hazard level, experiment reservation information, experiment start and end time, and experiment execution stage;

[0073] The current experimental personnel data includes the number of experimental personnel, their distribution location, personnel density, experimental personnel level, personnel stay duration, and entry and exit records;

[0074] The data on the equipment used in the current experiment includes the types of key equipment involved in the experiment, equipment operating status, operating mode, operating power, and usage time;

[0075] Both the historical laboratory environment suitability assessment data and the real-time laboratory environment suitability assessment data include laboratory environment data and laboratory internal data;

[0076] The laboratory environmental data include temperature, humidity, air exchange rate, carbon dioxide concentration, light intensity, and particulate matter concentration.

[0077] The data inside the laboratory includes the density of personnel in the laboratory, the heat generated by all equipment in the laboratory, and the duration of the experiment.

[0078] The historical laboratory environment suitability assessment data includes the historical laboratory environment suitability assessment results and the historical laboratory scene identification results;

[0079] The environmental suitability assessment result of the historical laboratory is specifically a value between 1 and 100;

[0080] The data cleaning is used to eliminate errors, missing values ​​and inconsistencies in the data, specifically by filling in missing values ​​and removing outliers from the original data on laboratory environmental control.

[0081] The missing value imputation specifically involves filling in missing values ​​using the mean imputation method; the outlier removal specifically involves detecting and removing extreme values ​​and logical outliers in the original data using the interquartile range method.

[0082] The data normalization is used to adjust the scale and range of the data and eliminate the adverse effects of different units and numerical ranges on model training. Specifically, it is based on the Z-Score normalization method to standardize all continuous variables to ensure that the data is within a uniform range.

[0083] The data encoding process is used to convert non-numerical data into a numerical format that can be processed by the model. Specifically, it uses a label encoding method to map the category field in the original data to the corresponding integer value according to the category value, thereby realizing the numerical representation of the category feature.

[0084] The data feature selection is used to screen out key features that contribute to laboratory scene identification and laboratory environment suitability assessment, reduce data dimensionality, and reduce the impact of redundant information and noisy features on model performance. Specifically, correlation analysis and variance thresholding methods are used to measure the correlation between each candidate feature and the target variable and the information content of the feature itself, and the features are sorted and screened to retain feature subsets whose correlation and importance both meet the preset threshold conditions, namely the laboratory scene identification feature set and the laboratory environment suitability assessment feature set.

[0085] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. Specifically, the laboratory scene recognition module constructs a laboratory scene recognition model using an improved clustering algorithm, inputs data into the model, performs real-time laboratory scene recognition, and obtains real-time laboratory scene recognition results. The specific steps include:

[0086] The construction of a laboratory scene recognition model includes the following steps:

[0087] Cluster center initialization is used to provide an accurate and stable initial clustering benchmark for laboratory scene recognition, avoiding the problems of local optima and scene recognition bias caused by traditional random initialization. Specifically, the Mahalanobis distance from each sample data point to the origin of the laboratory scene feature space is calculated, and all sample data points are sorted in ascending order of Mahalanobis distance. Then, the sorted data points are evenly divided into K groups. Finally, the average position of all sample data points in each group is used as the initial cluster center for that group, resulting in the initial cluster center set. Based on the initial cluster centers, the initial membership matrix is ​​calculated using the membership formula in the standard fuzzy C-means clustering algorithm.

[0088] The sample data points are specifically single feature vectors extracted from laboratory scene identification feature datasets from laboratory environment control and optimization data. This feature dataset is an effective set of feature data obtained by performing data optimization operations on historical laboratory scene identification data and real-time laboratory scene identification data.

[0089] The origin of the laboratory scene feature space is the reference coordinate point of all sample data points, and the value of this point is 0 in all feature dimensions.

[0090] The average position of all sample data points is specifically the feature vector formed by the average value of all sample data points across each feature dimension.

[0091] The formula used is as follows:

[0092] ;

[0093] ;

[0094] In the formula, Represents sample data points The Mahalanobis distance to the origin of the feature space of the laboratory scene. This represents the i-th sample data point. The covariance matrix represents the feature set for identifying laboratory scenes. Denotes the inverse matrix of the covariance matrix. This represents the membership degree of the i-th sample data point to the j-th class of laboratory scene clusters, with a value range of... Furthermore, it must satisfy the constraint that the sum of the initial membership degrees of a single sample to all scene categories is 1. This indicates the number of clusters, that is, the total number of clusters in the laboratory scenario. This represents the initial cluster center of the j-th type of laboratory scenario. This represents the initial cluster centers for the k-th type of laboratory scenario. This represents the Euclidean distance calculation function, where m represents the fuzzy membership index, used to control the degree of fuzziness in membership, and its value range is... ;

[0095] A laboratory scene recognition objective function is constructed to achieve synergistic optimization of scene feature compactness and prior label constraint compliance. This addresses the technical pain points of traditional laboratory scene recognition, such as insufficient utilization of historical prior label information, weak constraints in laboratory scene recognition, and poor fusion of real-time and historical scene data. Specifically, a laboratory scene recognition objective function integrating enhanced prior constraints is constructed; the formula used is as follows:

[0096] ;

[0097] In the formula, Let represent the objective function value of the laboratory scene recognition clustering at the t-th iteration. The smaller the value, the better the compactness of the scene features and the better the conformity of the clustering to the prior label constraints. This represents the weight coefficient of the unsupervised scene feature clustering term, used to balance the contribution of scene feature compactness and semi-supervised prior label constraints, and has a value of 4. Indicates the number of sample data points. This represents the membership value of the i-th sample data point to the j-th laboratory scenario at the t-th iteration. This represents the cluster center of the j-th type of laboratory scene at the t-th iteration. This represents the weight coefficient of the semi-supervised prior reinforcement constraint term, used to control the constraint strength of the prior labels from the laboratory's historical scenarios on the clustering process, and its value is 2. This represents the prior membership value of the i-th sample data point to the j-th class of laboratory scenarios at the t-th iteration. If the sample is a historically labeled sample, the target scenario's... In other scenarios, the value is 0. If it is a real-time unlabeled sample, Initially 0, This represents a binary identifier for the sample data point label. If , it means that the sample is a sample with historical scene annotations. This represents a real-time sample to be identified without historical labels, enabling accurate differentiation of semi-supervised data;

[0098] The membership matrix is ​​iteratively updated to achieve accurate iteration of sample membership in laboratory scene recognition. This strengthens the guiding role of historical prior labels in determining laboratory scene membership, ensuring that membership assignment aligns with both real-time scene characteristics and historical operational experience. Specifically, for the objective function of laboratory scene recognition, combined with the core constraints of fuzzy clustering, an unconstrained Lagrange function is constructed using the Lagrange multiplier method, transforming the constrained optimization problem into an unconstrained optimization problem. Furthermore, prior reinforcement bias derivatives are introduced to iteratively update the membership matrix. ;

[0099] The a prior reinforcement bias derivative Specifically, it strengthens the guiding role of prior labels for samples with historical annotations, enhances prior label constraints, and solves the pain point of constraint failure when the membership degree approaches the prior value in traditional algorithms, especially improving the accuracy of membership degree determination in laboratory scenarios.

[0100] The core constraint of the fuzzy clustering is that the sum of the membership degrees of a single sample to the K classes of scenarios is 1, that is... ;

[0101] The formula used is as follows:

[0102] ;

[0103] ;

[0104] In the formula, This represents the value of the unconstrained Lagrangian function at the t-th iteration. Represents the Lagrange multiplier for the i-th sample data point. Indicates the first In the next iteration, the membership value of the i-th sample data point to the j-th type of laboratory scenario is... This represents the cluster center of the k-th type of laboratory scene at the t-th iteration. This represents the prior membership value of the i-th data point to the k-th laboratory scenario at the t-th iteration.

[0105] The cluster center matrix update is used to achieve accurate iteration of cluster centers for laboratory scenarios. Specifically, it updates the cluster centers for each class of laboratory scenarios based on the updated membership matrix; the formula used is as follows:

[0106] ;

[0107] In the formula, Indicates the first In the next iteration, the cluster center of the j-th type of laboratory scene;

[0108] Weakly correlated cluster center identification is used to eliminate the interference of invalid membership degrees of weakly correlated scene centers, compress the computational dimension, and retain the effective feature associations of strongly correlated scene centers. This enhances the discriminative power of core scene features, avoids the decrease in recognition accuracy caused by the dilution of invalid information, improves model iteration efficiency, and meets the response requirements of real-time recognition in laboratory scenarios. Specifically, it first processes each sample data point... Calculate the nearest cluster center in the t-th iteration, and then calculate the sample data points respectively. The feature difference to the nearest cluster center, the iteration position deviation of each cluster center and the nearest cluster center, and finally the weakly correlated cluster centers that have no significant impact on the laboratory scene attribution result of the sample data point are selected to construct a set of weakly correlated cluster centers; the formula used is as follows:

[0109] ;

[0110] ;

[0111] ;

[0112] ;

[0113] ;

[0114] In the formula, Indicates the first In the next iteration, the set of weakly correlated cluster centers corresponding to the i-th sample data point is used. Since the weakly correlated cluster centers in this set have no substantial impact on the scene attribution of the sample, they can be removed from subsequent invalid calculations. Indicates the first In the next iteration, the nearest scene cluster center corresponding to the i-th sample data point is the scene cluster center whose features best match those of the sample. Indicates the first In the next iteration, the feature difference between the i-th sample data point and its nearest cluster center is the core benchmark threshold for determining weakly correlated cluster centers. Indicates the iteration from the t-th iteration to the t-th iteration. In the next iteration, the positional deviation of the cluster center for the j-th type of laboratory scene is, i.e., the iteration variation range of the cluster center. Indicates the iteration from the t-th iteration to the t-th iteration. In the next iteration, the positional deviation of the nearest scene cluster center for the i-th sample data point is used as an auxiliary threshold for determining the weakly correlated cluster center.

[0115] Membership degree reassignment optimization is used to eliminate the interference of invalid membership degrees corresponding to weakly correlated scene centers, compress the computational dimension of the model, and strengthen the effective feature association of affinity scene centers. This ensures the discriminativeness and rationality of the membership degree assignment of laboratory scenes, and achieves the reduction of invalid computational dimensions and the preservation of effective membership degree information. Specifically, based on the binary identifier of the sample data point labels, combined with the set of weakly correlated cluster centers, differential scaling processing is performed on the membership degrees of weakly correlated cluster centers of different types of sample data points. Then, combined with the modified membership degree, the membership degree of strongly correlated centers is normalized to obtain the optimized membership degree matrix.

[0116] ;

[0117] ;

[0118] In the formula, Indicates the first In the next iteration, the optimized membership value of the i-th sample data point belonging to the j-th type of laboratory scenario is determined. Indicates the first In the next iteration, the corrected membership value of the i-th sample to the c-th weakly correlated cluster center is used to calibrate the normalization coefficient of the affinity center membership, ensuring the rationality of the membership assignment. c represents the scene category index within the set of weakly correlated cluster centers. Indicates the first In the next iteration, the membership value of the i-th sample to the c-th weakly correlated cluster center;

[0119] The clustering iteration terminates, specifically by recalculating the objective function for identifying the laboratory scene. and the objective function value of the previous iteration The comparison is performed, and the difference between the two is calculated to obtain the change value of the objective function. Finally, the iteration is terminated according to the preset termination condition. If any clustering iteration termination condition is met, the clustering iteration is stopped, and the final cluster center matrix and the final optimized membership matrix are output.

[0120] The clustering iteration termination conditions include the objective function change value being less than the clustering convergence threshold and the number of iterations reaching the maximum number of iterations.

[0121] The scene recognition clustering results are output as follows: Based on the final optimized membership matrix, for each sample data point, the optimized membership value matching criterion is used to select the scene category with the largest optimized membership value among all types of laboratory scenes as the final scene clustering category of the sample. This divides all sample data points into multiple independent clusters according to their respective scene categories, thus obtaining the scene recognition clustering results. The scene clustering category specifically refers to the classification of laboratory scene categories based on the scene recognition clustering results for laboratory scene recognition data.

[0122] Real-time laboratory scene recognition involves inputting the laboratory scene recognition feature dataset into the laboratory scene recognition model to generate real-time scene recognition clustering results. Based on the real-time scene recognition clustering results, the cluster labels in each cluster are statistically analyzed, and the cluster label with the highest frequency in each cluster is selected as the laboratory scene recognition label for that cluster, thus obtaining the real-time laboratory scene recognition result.

[0123] By performing the above operations, this solution addresses the technical problems in existing laboratory scene recognition models, such as random initialization of cluster centers, insufficient utilization of historical annotation prior information, and lack of differentiation of weakly correlated cluster centers. These problems lead to clustering results easily getting trapped in local optima and low scene recognition accuracy, resulting in unstable laboratory scene segmentation and inaccurate triggering of environmental control strategies. This solution innovatively introduces Mahalanobis distance-based cluster center initialization, constructs a laboratory scene recognition objective function that integrates prior reinforcement constraints, implements a weakly correlated cluster center recognition strategy, and optimizes membership degree reassignment to improve the clustering algorithm. This improves the rationality and stability of cluster center initialization, enhances the ability to utilize historical scene annotation prior information, improves the discriminativeness of key scene cluster centers and the efficiency of clustering computation, effectively suppresses the interference of weakly correlated scene centers on membership degree assignment, and significantly improves the accuracy and reliability of laboratory scene recognition. This provides a more reliable foundation for subsequent environmental suitability assessment and environmental control strategy generation.

[0124] Example 4, see Figure 1 This embodiment is based on the above embodiment. The laboratory environment suitability assessment module is used to assess whether the current experimental environment is suitable for carrying out the target experiment based on the real-time laboratory scene recognition results. Specifically, it includes the following steps:

[0125] The laboratory environment suitability assessment module is used to assess whether the current laboratory environment is suitable for conducting the target experiment based on real-time laboratory scene recognition results. Specifically, it includes the following steps:

[0126] A laboratory environment suitability assessment model was constructed and trained. Specifically, the model was built based on a bidirectional long short-term memory neural network. The historical laboratory environment suitability assessment feature dataset and historical laboratory scene recognition results were used as training data for the assessment model to obtain the trained laboratory environment suitability assessment model.

[0127] The bidirectional long short-term memory neural network can simultaneously utilize environmental change information from historical and future time segments to analyze the dynamic change patterns of environmental parameters. Compared with traditional static models, it can more accurately reflect the overall suitability of the laboratory environment for the target experiment.

[0128] The historical laboratory environmental suitability assessment feature dataset is a set of effective feature data obtained by performing data optimization operations on historical laboratory environmental suitability assessment data;

[0129] The evaluation model training specifically involves using mean squared error as the loss function to optimize the gap between the prediction results and the true values ​​of the evaluation model. The parameters of the evaluation model are jointly optimized through the backpropagation algorithm. The loss function value is continuously monitored and optimized until it converges, at which point the model training is stopped, and the trained laboratory environment suitability evaluation model is obtained.

[0130] Real-time laboratory environment suitability assessment specifically involves inputting a real-time laboratory environment suitability assessment feature dataset and real-time laboratory scene recognition results into a trained laboratory environment suitability assessment model to obtain real-time laboratory environment suitability assessment results.

[0131] The real-time laboratory environment suitability assessment feature dataset is a set of effective feature data obtained by performing data optimization operations on real-time laboratory environment suitability assessment data;

[0132] Laboratory environment suitability assessment is used to determine the suitability level of the current laboratory environment, providing a basis for the generation of subsequent environmental control strategies and alarm handling. Specifically, it is based on real-time laboratory environment suitability assessment results. Suitable threshold for preset environment and environmental hazard thresholds The laboratory environment is classified to obtain the real-time laboratory environment status.

[0133] The laboratory environment status includes three categories: suitable, unsuitable, and hazardous. Specifically, these are further divided into: If so, the current laboratory environment is judged to be in a suitable state. If so, the current laboratory environment is determined to be unsuitable. If so, the current laboratory environment is judged to be in a dangerous state.

[0134] Example 5, see Figure 1 , Figure 3 and Figure 4 This embodiment is based on the above embodiment. The laboratory environment control strategy generation module is used to generate a laboratory environment control strategy adapted to the current experimental scenario based on the real-time laboratory scenario identification results and the laboratory environment suitability judgment results; specifically, it includes the following steps:

[0135] The laboratory environment control template strategy is selected to provide an initial reference scheme for the optimization algorithm. Specifically, based on the real-time laboratory scene recognition results, the laboratory environment control strategy data corresponding to the current laboratory scene is matched in the preset laboratory scene environment control template strategy library, and typical set values ​​of the environmental control parameters are extracted from them as the initial reference values ​​for the optimization of the environmental control parameters.

[0136] The laboratory environment control strategy consists of six environmental control parameters: temperature, humidity, air exchange rate, carbon dioxide concentration, light intensity, and particulate matter concentration.

[0137] Each laboratory environmental control strategy in the environmental control template strategy library is based on a set of typical environmental parameter values ​​predefined by experts for a specific laboratory scenario, which are used as an initial strategy reference for optimization calculation.

[0138] Setting boundary intervals for environmental control parameters is used to limit the adjustable range of environmental control parameters. Specifically, based on the real-time laboratory scene recognition results, the upper and lower limits corresponding to each environmental control parameter are retrieved from the laboratory scene boundary mapping table, and the dynamic boundary interval of each environmental control parameter is generated accordingly.

[0139] The laboratory scenario boundary mapping table predefines the mapping relationship between different laboratory scenarios and the upper and lower limits of corresponding environmental control parameters, which is used to constrain the changes of environmental control parameters within the range of safety, comfort and energy-saving requirements;

[0140] Optimizing environmental control parameters involves using an improved optimization algorithm to search for the optimal combination of environmental control parameters within the boundary range of each parameter, thereby obtaining the best combination of parameters suitable for the current laboratory environment. This includes the following steps:

[0141] The initialization of the search population specifically involves using the laboratory environmental control strategy as the individual position vector in the optimization algorithm, and generating the vector by perturbation based on the initial reference value of the environmental control parameters as the center, combined with the boundary interval of each environmental control parameter. Individual location vectors for each search Thus, the initial search population is obtained;

[0142] Each dimension of the individual location vector corresponds to six environmental control parameters: temperature, humidity, air exchange rate, carbon dioxide concentration, light intensity, and particulate matter concentration.

[0143] The individual fitness value is calculated by using the individual environmental suitability assessment result as the individual fitness value, sorting all search individuals in the population in descending order according to the fitness value, and defining three types of guide individuals and their positions.

[0144] The three types of guiding individuals and their positions are specifically the individuals with the best fitness and their positions, the second-best fitness individuals and their positions, and the third-best fitness individuals and their positions.

[0145] The individual environmental suitability assessment result is specifically obtained by converting the individual location vector, which represents the laboratory environment control strategy, into corresponding environmental data, replacing the environmental data in the laboratory environmental suitability assessment feature dataset, and then combining it with the laboratory scene recognition result. The result is then input into the trained laboratory environmental suitability assessment model to obtain the individual's laboratory environmental suitability assessment result.

[0146] The initial update of individual locations involves guiding individuals based on three categories and assigning contribution weights to their location information using a dynamic weighting strategy. This initial update yields the initial location of the i-th search individual in the i-th search position. Initial position during iteration The formula used is as follows:

[0147] ;

[0148] ;

[0149] ;

[0150] ;

[0151] ;

[0152] In the formula, Indicates the i-th individual's direction The new location Indicates the i-th individual's direction The new location Indicates the i-th individual's direction The new location Indicates that the i-th search individual is in the first position. Initial position during iteration Represents the i-th individual. Position during iteration , and This represents a randomness parameter within the range [0,1]. This represents the individual with the best fitness. This indicates the second-best fitter individual. This indicates the individual with the third best fitness. This indicates the location of the individual with the best fitness. This indicates the position of the second-best fitter individual. This indicates the position of the third-best fitter individual. Indicates the first The convergence factor of the iteration. This represents the fitness value calculation function, i.e., the trained laboratory environment suitability assessment model. express The individual fitness value, express The individual fitness value, express The individual fitness value, , and They represent , , Dynamic weights;

[0153] Individual memory position updates are used to preserve the optimal position information during the individual's evolution and to correct the current search direction using the superior features of multiple guide bodies, avoiding the loss of effective positions due to deviations in a single update, thus balancing the algorithm's global exploration and local convergence capabilities. Specifically, a historical optimal position memory mechanism is introduced, based on the individual's historical movement speed from the previous iteration. The influence of historical trends is controlled through inertia weights. Simultaneously, the deviation between the local movement position of the three types of guide individuals and the individual's initial position is calculated. Combined with learning factors and random parameters, the individual's movement speed direction is corrected. Finally, the individual memory position is updated to obtain the position of the i-th search individual in the i-th iteration. Memorizing and updating positions during iteration The formula used is as follows:

[0154] ;

[0155] In the formula, Represents the i-th individual Motion speed during iteration Represents the i-th individual Motion speed during iteration Indicates that the i-th search individual is in the first position. Remember the update position during iteration. This represents the inertia weight, used to balance the influence of an individual's historical velocity with the influence of new guidance information; its value is typically between [value missing]. , , and They represent , and Guiding weights for speed updates, , and This represents a randomness parameter within the range [0,1]. Indicates that the i-th search individual is in the first position. Remember the update position during iteration;

[0156] Individual dimensional position updates are used to generate new solutions by adaptively defining the neighborhood range and fusing local high-quality information with global random information at the dimensional level. This improves population diversity and avoids the algorithm getting trapped in local optima. Specifically, it employs a dimensional learning neighborhood search strategy and calculates... and The Euclidean distance is used as the neighborhood search radius for the individual. Then, based on the neighborhood search radius, the population is traversed except for the individual. All individuals other than those selected, those with... Individuals whose Euclidean distance is less than their neighborhood search radius are considered as their neighborhood set. Then, for each dimension of each individual, the position of that individual is updated by combining the dimensional information of neighboring individuals and random individuals in the population. After traversing all dimensions, the update results are combined to generate the complete dimensional update position of that individual. The formula used is as follows:

[0157] ;

[0158] ;

[0159] ;

[0160] In the formula, This indicates that the i-th search individual is in the i-th position. The neighborhood search radius in the next iteration. This indicates that the i-th search individual is in the i-th position. The neighborhood set of the next iteration This indicates that the position of the d-th dimension of an individual is randomly selected from the neighborhood set. This indicates that the j-th search individual is in the first position. The position of the next iteration. This indicates that the i-th search individual is in the i-th position. The position after updating the d-th dimension in the next iteration. This represents the position of a random individual in the population along the d-th dimension. This represents a randomness parameter within the range [0,1]. This indicates that the i-th search individual is in the i-th position. The position in the d-th dimension of the next iteration;

[0161] Iterative dual-position selection integrates the advantages of dual-optimized positions from previous iterations, balancing the preservation of historical high-quality information with the improvement of population diversity, and ultimately locks the optimal individual position in a single iteration. Specifically, it calculates the position of the i-th search individual in the i-th iteration. In the next iteration, the fitness values ​​of the memory update position and the dimension update position are compared. The fitness values ​​of the two candidate positions are then selected based on the principle of better fitness as the position for that individual in the [number]th iteration. Final position of the next iteration This completes the position optimization loop for a single iteration; the formula used is as follows:

[0162] ;

[0163] In the formula, This indicates that the i-th search individual is in the i-th position. The position of dimension update in the next iteration Represents the i-th individual. Position during iteration;

[0164] The search for the optimal position of an individual involves evaluating the fitness values ​​of all individuals in the current population after each iteration. If the fitness of an individual's position is better than that of the current global optimal individual position, then that individual's global optimal position is updated.

[0165] The search iteration terminates when the individual's fitness value is higher than the fitness threshold or when the maximum number of iterations is reached, and the search is terminated and the individual's global optimal position is obtained. The individual's global optimal position specifically refers to the optimal combination of environmental control parameters in the current laboratory scenario.

[0166] The optimal environmental control strategy for the laboratory is output, specifically, based on the optimal combination of environmental control parameters for the current laboratory scenario, the optimal environmental control strategy for the current laboratory scenario is generated and output.

[0167] By performing the above operations, this solution addresses the technical problems of existing algorithms for optimizing laboratory environmental control parameters, such as the tendency to get trapped in local optima, insufficient population diversity, low search efficiency, and long time consumption in obtaining the optimal environmental control strategy, which leads to slow response in laboratory environmental control. This solution innovatively introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm. This effectively enhances the stability, convergence speed, and solution quality of the environmental control parameter optimization process, significantly improves the rationality, accuracy, and robustness of the laboratory environmental control strategy, and realizes rapid and intelligent control of the laboratory environment under complex experimental conditions.

[0168] Example 6, see Figure 1 This embodiment is based on the above embodiment. Specifically, the intelligent environmental control auxiliary module continuously monitors the real-time laboratory scene recognition results and the real-time laboratory environment status, and intelligently adjusts the laboratory environment control strategy according to different triggering conditions. Based on the current environmental control strategy, it performs coordinated control of the laboratory environment. Specifically, it includes the following steps:

[0169] Real-time laboratory scene recognition result monitoring is used to trigger environmental suitability reassessment and control strategy update when the laboratory scene changes. Specifically, if the real-time laboratory scene recognition result changes, the laboratory environment suitability assessment module is called to assess the current laboratory environment suitability. If the real-time laboratory environment status is determined to be unsuitable or dangerous, the laboratory environment control strategy generation module is triggered to generate a laboratory environment control strategy, obtain the optimal environmental control strategy for the current laboratory scene, and perform linkage control on the laboratory environment. If the real-time laboratory environment status is suitable, only the scene information is updated, and the existing environmental control strategy remains unchanged.

[0170] Real-time laboratory environment status monitoring is used to trigger adjustments to environmental control strategies when environmental conditions deteriorate. Specifically, it periodically acquires the output results of the laboratory environment suitability assessment module. If the real-time laboratory environment status is determined to be unsuitable or dangerous, the laboratory environment control strategy generation module is invoked based on the current laboratory scenario identification results to obtain the optimal environmental control strategy for the current laboratory scenario. This optimal strategy is then used to replace the existing environmental control strategy and to perform coordinated control of the laboratory environment. If the real-time laboratory environment status is suitable, the existing environmental control strategy is maintained.

[0171] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0172] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0173] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based laboratory environment control auxiliary system, characterized in that: It includes a data acquisition module, a laboratory scene recognition module, a laboratory environment suitability assessment module, a laboratory environment control strategy generation module, and an environmental intelligent control auxiliary module; The data acquisition module specifically obtains optimized laboratory environment control data through data acquisition and data optimization operations. The laboratory scene recognition module specifically introduces cluster center initialization based on Mahalanobis distance, constructs a laboratory scene recognition objective function that integrates prior reinforcement constraints, adopts a weak correlation cluster center recognition strategy, and optimizes membership degree reassignment to improve the clustering algorithm. The improved clustering algorithm is used to construct a laboratory scene recognition model, and real-time data is input into the model to obtain real-time laboratory scene recognition results. The laboratory environment suitability assessment module specifically involves constructing and training a laboratory environment suitability assessment model, then performing a real-time assessment of laboratory environment suitability, outputting the real-time laboratory environment suitability assessment results, and determining the suitability of the laboratory environment based on the assessment results to obtain the real-time laboratory environment status. The laboratory environment control strategy generation module first selects a laboratory environment control template strategy, then sets the boundary range of environment control parameters, then introduces a historical optimal position memory mechanism and adopts a dimensional learning neighborhood search strategy to improve the optimization algorithm, and then optimizes the environmental control parameters through the improved optimization algorithm to obtain the optimal combination of environmental control parameters for the current laboratory scenario. Finally, it generates the optimal environmental control strategy for the current laboratory scenario based on the optimal environmental control parameter group. The intelligent environmental control auxiliary module specifically adjusts the environmental control strategy based on the real-time laboratory scene recognition results and the real-time laboratory environment status, and executes the control measures to achieve intelligent control of the laboratory environment.

2. The artificial intelligence-based laboratory environment control auxiliary system according to claim 1, characterized in that: The laboratory scene recognition module specifically includes the following steps: The laboratory scene recognition model is constructed, specifically including cluster center initialization, construction of the laboratory scene recognition objective function, membership matrix iterative update, cluster center matrix update, weak correlation cluster center recognition, membership degree reassignment optimization, cluster iteration termination, and scene recognition clustering result output; The cluster center matrix update specifically involves updating the cluster centers for each type of laboratory scenario based on the updated membership matrix. The termination of the clustering iteration specifically involves recalculating the objective function for laboratory scene recognition. The difference between the objective function value and the objective function value of the previous iteration is calculated to obtain the change value of the objective function. Finally, the iteration termination judgment is made according to the preset termination condition. If any clustering iteration termination condition is met, the clustering iteration is stopped and the final cluster center matrix and the final optimized membership matrix are output. The scene recognition clustering result output is specifically as follows: based on the final optimized membership matrix, for each sample data point, the optimized membership value matching criterion is adopted, and the scene category with the largest value among the optimized membership values ​​of the corresponding laboratory scenes is selected as the final scene clustering category of the sample. In this way, all sample data points are divided into multiple independent clusters according to their respective scene categories, and the scene recognition clustering result is obtained. Real-time laboratory scene recognition involves inputting the laboratory scene recognition feature dataset into the laboratory scene recognition model to generate real-time scene recognition clustering results. Based on the real-time scene recognition clustering results, the cluster labels in each cluster are statistically analyzed, and the cluster label with the highest frequency in each cluster is selected as the laboratory scene recognition label for that cluster, thus obtaining the real-time laboratory scene recognition result.

3. The artificial intelligence-based laboratory environment control auxiliary system according to claim 2, characterized in that: The initialization of the cluster centers specifically involves calculating the Mahalanobis distance from each sample data point to the origin of the laboratory scene feature space, sorting all sample data points in ascending order of Mahalanobis distance, and then dividing the sorted data points into K groups. Finally, the average position of all sample data points in each group is used as the initial cluster center for that group, resulting in an initial set of cluster centers. Based on the initial cluster centers, the initial membership matrix is ​​calculated using the membership formula in the standard fuzzy C-means clustering algorithm. The construction of the laboratory scene recognition objective function specifically involves constructing a laboratory scene recognition objective function that integrates prior reinforcement constraints; the formula used is as follows: ； In the formula, This represents the objective function value for clustering in the laboratory scene recognition at the t-th iteration. This represents the weight coefficient of the feature clustering item in the unsupervised scenario. Indicates the number of sample data points. This represents the membership value of the i-th sample data point to the j-th laboratory scenario at the t-th iteration. This represents the i-th sample data point. This represents the cluster center of the j-th type of laboratory scene at the t-th iteration. This represents the weight coefficient of the semi-supervised prior reinforcement constraint term. Indicates the number of cluster categories. This represents the prior membership value of the i-th sample data point to the j-th laboratory scenario at the t-th iteration. Represents the binary identifier of the sample data point label. Indicates the first In the next iteration, the cluster center of the j-th type of laboratory scene.

4. The artificial intelligence-based laboratory environment control auxiliary system according to claim 2, characterized in that: The membership matrix iterative update specifically involves constructing an unconstrained Lagrangian function using the Lagrange multiplier method for the objective function of laboratory scene recognition, combined with the core constraints of fuzzy clustering. This transforms the constrained optimization problem into an unconstrained optimization problem. Furthermore, a prior reinforcement bias term is introduced to iteratively update the membership matrix. The formula used is as follows: ； In the formula, Indicates the first In the next iteration, the membership value of the i-th sample data point to the j-th type of laboratory scenario is... This represents the cluster center of the k-th type of laboratory scene at the t-th iteration. Let represent the prior membership value of the i-th data point belonging to the k-th laboratory scenario at the t-th iteration.

5. The artificial intelligence-based laboratory environment control auxiliary system according to claim 2, characterized in that: The identification of weakly correlated cluster centers specifically involves first identifying each sample data point... Calculate the nearest cluster center in the t-th iteration, and then calculate the sample data points respectively. The feature difference of the nearest cluster center, the iterative position deviation of each cluster center and the nearest cluster center are used to finally select weakly correlated cluster centers that have no significant impact on the laboratory scene attribution result of the sample data point, and construct a set of weakly correlated cluster centers. The membership degree redistribution optimization specifically involves performing differential scaling on the membership degrees of weakly correlated cluster centers for different types of sample data points based on the binary identifiers of the sample data point labels and the set of weakly correlated cluster centers, and then normalizing the membership degrees of strongly correlated centers by combining the corrected membership degrees to obtain the optimized membership degree matrix. ； In the formula, Indicates the first In the next iteration, the optimized membership value of the i-th sample data point belonging to the j-th type of laboratory scenario is determined. Indicates the first In the next iteration, the corrected membership value of the i-th sample to the c-th weakly correlated cluster center is determined, where c represents the scene category index within the set of weakly correlated cluster centers. Indicates the first In the next iteration, the membership value of the i-th sample to the c-th weakly correlated cluster center is... Indicates the first In the next iteration, the set of weakly correlated cluster centers corresponding to the i-th sample data point.

6. The artificial intelligence-based laboratory environment control auxiliary system according to claim 1, characterized in that: The laboratory environment suitability assessment module specifically includes the following steps: A laboratory environment suitability assessment model was constructed and trained. Specifically, the model was built based on a bidirectional long short-term memory neural network. The historical laboratory environment suitability assessment feature dataset and historical laboratory scene recognition results were used as training data for the assessment model to obtain the trained laboratory environment suitability assessment model. Real-time laboratory environment suitability assessment specifically involves inputting a real-time laboratory environment suitability assessment feature dataset and real-time laboratory scene recognition results into a trained laboratory environment suitability assessment model to obtain real-time laboratory environment suitability assessment results. The laboratory environment suitability assessment is specifically based on the real-time laboratory environment suitability assessment results and preset environmental suitability thresholds and environmental hazard thresholds to classify the laboratory environment status and obtain the real-time laboratory environment status.

7. The artificial intelligence-based laboratory environment control auxiliary system according to claim 1, characterized in that: The laboratory environment control strategy generation module specifically includes the following steps: The laboratory environment control template strategy is selected. Specifically, based on the real-time laboratory scene recognition results, the laboratory environment control strategy data corresponding to the current laboratory scene is matched in the preset laboratory scene environment control template strategy library, and the typical set values ​​of the environmental control parameters are extracted from them as the initial reference values ​​for the optimization of the environmental control parameters. Setting the boundary range of environmental control parameters involves retrieving the upper and lower limits of each environmental control parameter from the laboratory scene boundary mapping table based on the real-time laboratory scene recognition results, and generating the dynamic boundary range of each environmental control parameter accordingly. Optimize environmental control parameters; The optimal environmental control strategy for the laboratory is output, specifically, based on the optimal combination of environmental control parameters for the current laboratory scenario, the optimal environmental control strategy for the current laboratory scenario is generated and output.

8. The artificial intelligence-based laboratory environment control auxiliary system according to claim 7, characterized in that: The optimization of environmental control parameters specifically includes the following steps: The initialization of the search population specifically involves using the laboratory environmental control strategy as the individual position vector in the optimization algorithm, and generating the vector by perturbation based on the initial reference value of the environmental control parameters as the center, combined with the boundary interval of each environmental control parameter. Individual location vectors for each search Thus, the initial search population is obtained; The individual fitness value is calculated by using the individual environmental suitability assessment result as the individual fitness value, sorting all search individuals in the population in descending order according to the fitness value, and defining three types of guide individuals and their positions. The three types of guiding individuals and their positions are specifically the individuals with the best fitness and their positions, the second-best fitness individuals and their positions, and the third-best fitness individuals and their positions. Initial location updates for individual search individuals involve providing location guidance based on three categories of individuals, and assigning contribution weights to the location information of these three categories using a dynamic weighting strategy. This yields the location of the i-th search individual in the i-th search position. Initial position during iteration ; Individual memory position update specifically involves introducing a historical optimal position memory mechanism. Based on the individual's historical movement speed from the previous iteration, the influence of historical trends is controlled through inertia weights. Simultaneously, the deviation between the individual's local movement position and its initial position is calculated using three types of guidance. Combined with learning factors and random parameters, the individual's movement speed direction is corrected. Finally, the individual memory position is updated to obtain the position of the i-th search individual in the i-th iteration. Memorizing and updating positions during iteration ; Individual dimensional position updates are achieved by employing a dimensional learning neighborhood search strategy to calculate... and The Euclidean distance is used as the neighborhood search radius for the individual. Then, based on the neighborhood search radius, the population is traversed except for the individual. All individuals other than those selected, those with... Individuals whose Euclidean distance is less than the individual's neighborhood search radius constitute the individual's neighborhood set. Then, for each dimension of each individual, the position update of the individual's single dimension is completed by combining the dimensional information of the neighborhood individuals and the random individuals in the population. After traversing all dimensions, the update results of each dimension are combined to generate the complete dimensional update position of the individual. Iterative dual-position filtering, specifically, calculating the position of the i-th search individual at the i-th position. In the next iteration, the fitness values ​​of the memory update position and the dimension update position are compared. The fitness values ​​of the two candidate positions are then selected based on the principle of better fitness as the position for that individual in the [number]th iteration. Final position of the next iteration This completes the position optimization closed loop for a single iteration. The search for the optimal position of an individual involves evaluating the fitness values ​​of all individuals in the current population after each iteration. If the fitness of an individual's position is better than that of the current global optimal individual position, then that individual's global optimal position is updated. The search iteration terminates when the individual's fitness value is higher than the fitness threshold or when the maximum number of iterations is reached, and the search is terminated and the individual's global optimal position is obtained. The individual's global optimal position specifically refers to the optimal combination of environmental control parameters in the current laboratory scenario.

9. The artificial intelligence-based laboratory environment control auxiliary system according to claim 1, characterized in that: The intelligent environmental control auxiliary module specifically monitors the real-time laboratory scene recognition results and the real-time laboratory environment status, and intelligently adjusts the laboratory environment control strategy according to different triggering conditions. Based on the current environmental control strategy, it performs coordinated control of the laboratory environment.

10. The artificial intelligence-based laboratory environment control auxiliary system according to claim 1, characterized in that: The data acquisition module specifically involves obtaining raw data on laboratory environmental control through data collection operations within the laboratory management platform, and then performing data optimization processing on this raw data to obtain optimized laboratory environmental control data. The raw data on laboratory environmental control includes historical laboratory scene identification data, real-time laboratory scene identification data, historical laboratory environmental suitability assessment data, and real-time laboratory environmental suitability assessment data. The data optimization processing includes data cleaning, data normalization, data encoding, and data feature selection.

Images