A pathogenic microorganism laboratory environment monitoring collaborative control system

Abstract

The application relates to the technical field of intelligent control, and relates to a pathogenic microorganism laboratory environment monitoring collaborative control system, which comprises a data acquisition module, an association evaluation module, a model construction module, a risk early warning module, an early warning strategy generation module and a control feedback module, acquires multi-source time sequence data flow of a laboratory, carries out multi-modal association evaluation on environment parameter data, reagent state data and personnel operation behavior data, obtains a risk characteristic vector, constructs an intelligent risk evaluation model, carries out link evolution on the risk characteristic vector, obtains a comprehensive risk score and an early warning level, maps the comprehensive risk score and the early warning level to a preset risk early warning strategy library, obtains early warning information and a disposal scheme, encodes the early warning information and the disposal scheme, responds to the early warning information and the disposal scheme, updates feedback parameters monitored in the response process to historical experimental data, and realizes monitoring collaborative control of the laboratory; and the application can improve the pathogenic microorganism laboratory environment monitoring collaborative control efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, and in particular to a collaborative control system for monitoring the laboratory environment of pathogenic microorganisms. Background Technology

[0002] Existing technologies struggle to efficiently integrate and process environmental parameter data, reagent status data, and personnel operational behavior data in pathogenic microorganism laboratory environmental monitoring. Due to the lack of a unified time-series alignment and precise noise reduction coupling mechanism, multi-source data often exhibits decentralized and non-standardized characteristics, failing to form a coherent and reliable multi-source time-series data stream. This results in the fragmentation of risk-related information contained within the data, making it difficult to provide comprehensive and accurate data support for risk assessment and limiting the ability to detect potential risks in the laboratory early on.

[0003] In the risk assessment and collaborative control phases, existing technologies lack intelligent model building schemes based on historical experimental data and expert knowledge bases, failing to accurately identify key risk signals in multi-source data and apply scientific weighting. Furthermore, the depth of multimodal correlation assessment of environmental, reagent, and personnel data is insufficient, and the extraction of risk feature vectors is not comprehensive enough. This leads to discrepancies between the comprehensive risk score and the early warning level determination, making it impossible to generate targeted response plans. Monitoring and control are disconnected, and the real-time nature and effectiveness of collaborative control are insufficient, making it difficult to quickly respond to and resolve various risks and hidden dangers in laboratory operations. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a collaborative control system for monitoring the laboratory environment of pathogenic microorganisms, characterized in that the system includes a data acquisition module, a correlation assessment module, a model construction module, a risk early warning module, an early warning strategy generation module, and a control feedback module, wherein: The data acquisition module is used to de-noise and couple environmental parameter data, reagent status data and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream of the laboratory. The correlation assessment module is used to perform multimodal correlation assessments on environmental parameter data, reagent status data, and personnel operation behavior data to obtain the laboratory's risk feature vector; The model building module is used to construct an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams through an attention mechanism based on historical experimental data of pathogenic microorganisms and an expert knowledge base. The risk warning module is used to perform link evolution on the risk feature vector based on the intelligent risk assessment model in order to obtain the laboratory's comprehensive risk score and warning level. The early warning strategy generation module is used to map the comprehensive risk score and early warning level to the preset risk early warning strategy library to obtain the laboratory's early warning information and response plan; The control feedback module is used to encode and respond to early warning information and response plans, and update the feedback parameters monitored during the response process to historical experimental data to achieve collaborative monitoring and control of the laboratory.

[0005] In a preferred embodiment, when the data acquisition module performs noise reduction and coupling of environmental parameter data, reagent status data, and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream from the laboratory, it is specifically used for: Collect environmental parameter data, reagent status data, and personnel operation behavior data of pathogenic microorganisms in the laboratory; Based on a unified timestamp, environmental parameter data, reagent status data and personnel operation behavior data are time-series aligned to obtain aligned environmental time-series data, reagent status time-series data and personnel operation time-series data. Outlier removal was performed on environmental time-series data, reagent status time-series data, and personnel operation time-series data to obtain clean environmental time-series data, clean reagent status time-series data, and clean personnel operation time-series data for the laboratory. The time-series data of the clean environment, the time-series data of the test reagent status, and the time-series data of the clean personnel operation are timestamped to obtain a multi-source time-series data stream of the laboratory.

[0006] In a preferred embodiment, when the correlation assessment module performs a multimodal correlation assessment of environmental parameter data, reagent status data, and personnel operation behavior data to obtain the laboratory's risk feature vector, it is specifically used for: Semantic analysis was performed on environmental parameter data, reagent status data, and personnel operation behavior data to obtain experimentally relevant semantic label data for multi-source time-series data streams; Based on the entity attribute knowledge base of pathogenic microorganisms, the experimental-related semantic tag data is retrieved and matched to obtain structured risk instance data of pathogenic microorganisms. By correlating and statistically analyzing structured risk instance data with laboratory operational status data, we can obtain environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators for the laboratory. By combining environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators into tensors, a risk feature vector for the laboratory is obtained.

[0007] In a preferred embodiment, when the model building module executes the process of constructing an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams using an attention mechanism based on historical experimental data of pathogenic microorganisms and an expert knowledge base, it is specifically used for: Based on historical experimental data of pathogenic microorganisms, key factor analysis was performed on historical risk events in the laboratory and their corresponding environmental and operational parameters to obtain the laboratory's historical risk feature set. The constraints of the expert knowledge base in pathogenic microorganisms are analyzed in a structured manner to obtain the risk rule set of the laboratory; Based on the historical risk feature set and risk rule set, pattern recognition screening is performed on multi-source time series data streams to obtain a set of candidate key risk signals for the laboratory; Based on the correlation strength between different features and risk events in the historical risk feature set, the candidate key risk signal set is weighted and assigned to obtain the laboratory's weighted key risk signal. Based on the weighted key risk signals, the parameter structure and decision logic of the preset initial risk assessment model are configured to obtain the laboratory's intelligent risk assessment model.

[0008] In a preferred embodiment, when the model building module performs weighted evaluation and allocation of candidate key risk signal sets based on the correlation strength between different features and risk events in the historical risk feature set to obtain the laboratory's weighted key risk signal, it is specifically used for: Extract co-occurrence pairs between historical risk features and corresponding risk events; By performing probability statistics on the co-occurrence frequency and conditions of co-occurrence pairs, the quantitative correlation strength value of the historical risk feature set is obtained; Risk features are sorted in descending order based on their quantitative correlation strength values ​​to obtain a ranking table of the importance of risk features. Based on the risk feature importance ranking table, a mapping relationship from risk features to preset weight intervals is established to obtain the feature weight mapping table of the laboratory. Based on the feature weight mapping table, the risk features in the candidate key risk signal set are weighted and assigned values ​​to obtain the initial weighted risk signal of the laboratory. The initial weighted risk signal is normalized and its consistency is verified to obtain the laboratory's weighted critical risk signal.

[0009] In a preferred embodiment, when the risk warning module performs a link evolution of the risk feature vector based on an intelligent risk assessment model to obtain the laboratory's comprehensive risk score and warning level, it is specifically used for: Based on the intelligent risk assessment model, a potential causal correlation analysis is performed on the risk feature vector to obtain the preliminary risk evolution path of the risk feature vector; Based on the initial risk evolution path, a dynamic causal network graph of risk feature vectors is constructed with risk signals as nodes and causal relationships as edges. Risk signals are extrapolated from the link transfer probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph; Based on the potential propagation sequence and scope of impact, the initial risk evolution path is prioritized to obtain the key risk propagation path with risk feature vectors; The risk signal strength and corresponding link transfer probability of nodes in the key risk propagation path are quantitatively evaluated to obtain the laboratory's comprehensive risk score; The laboratory's early warning level is obtained by comparing the comprehensive risk score with the risk threshold.

[0010] In a preferred embodiment, when the risk warning module executes the dynamic causal network graph that constructs risk feature vectors based on the initial risk evolution path, using risk signals as nodes and causal relationships as edges, it is specifically used for: The risk signals describing the risk state in the initial risk evolution path are traversed and calibrated to obtain the risk signal node set of the initial risk evolution path; Based on the causal rule base stored in the intelligent risk assessment model, confidence matching is performed on the causal relationships between nodes in the risk signal node set to obtain the causal relationship edge set of the risk signal node set; By topologically connecting the risk signal node set with the causal relationship edge set, the initial causal network structure of the risk feature vector is obtained. Based on the evolution time sequence of risk signals in the initial risk evolution path, timestamp attributes are injected into the corresponding nodes and edges in the initial causal network topology to obtain an enhanced causal network graph of risk feature vectors. Redundant edges are removed and isolated nodes are filtered from the enhanced causal network graph to obtain a dynamic causal network graph of risk feature vectors.

[0011] In a preferred embodiment, when the risk warning module performs risk signal deduction on the link transition probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph, it is specifically used for: Read the link transition probability matrix describing the causal relationship between nodes in different risk states from the rule base of the intelligent risk assessment model; Using the risk state of the risk evolution state set as the initial signal source, iterative state transitions are performed based on the link transition probability matrix to obtain the risk propagation path of the risk evolution state set; The path probability and cumulative impact of risk propagation paths are evaluated and screened to obtain a set of critical propagation paths for risk propagation. By integrating the risk status nodes traversed by different paths in the critical propagation path set, we obtain the set of affected nodes of the critical propagation path set; Based on the set of affected nodes, the transition state of risky nodes is assessed to obtain the potential propagation sequence and impact range of the dynamic causal network graph.

[0012] In a preferred embodiment, when the early warning strategy generation module maps the comprehensive risk score and early warning level to a preset risk early warning strategy library to obtain the laboratory's early warning information and response plan, it is specifically used for: Based on a pre-set risk warning strategy library, the warning levels are matched to obtain preliminary warning strategies for each warning level. Based on the standard risk threshold range in the risk warning strategy library, the deviation of the comprehensive risk score is compared to obtain the warning information to be filled in the comprehensive risk score. Based on the current state data of the risk feature vector and the intelligent risk assessment model, the dynamic variables in the early warning information to be filled are filled with contextual information to obtain the laboratory's early warning information. Based on the risk type and level identifier in the early warning information, a related search is performed in the risk early warning strategy database to obtain the basic treatment plan for the laboratory. By combining the risk evolution state set and the risk transmission path of the dynamic causal network diagram, the basic response plan is adjusted and optimized to adapt to the contingency plan, resulting in the laboratory response plan.

[0013] In a preferred embodiment, when the control feedback module encodes and responds to early warning information and response plans, and updates the feedback parameters monitored during the response process to historical experimental data to achieve collaborative monitoring and control of the laboratory, it is specifically used for: The early warning information and the response plan are converted into commands to obtain the control command sequence for the laboratory. The control command sequence is distributed to the corresponding experimental equipment and personnel message terminals in the laboratory to initiate the execution process of the control command sequence; During the execution process, comprehensive status data is collected in real time through a monitoring sensor network as feedback parameters for the laboratory. The feedback parameters, early warning information, and response plans are associated and packaged to obtain the laboratory's response feedback record; The feedback records will be updated to the repository of historical experimental data to complete the closed loop of laboratory monitoring and collaborative control.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves efficient integration and in-depth mining of multi-source laboratory data through a comprehensive data acquisition and correlation assessment mechanism. The data acquisition process, through time-series alignment, outlier removal, and coupling, yields a high-quality multi-source time-series data stream, providing a reliable data foundation for risk analysis. The correlation assessment process, through semantic parsing, knowledge base matching, and tensor synthesis, comprehensively extracts relevant risk indicators related to the environment, reagents, and personnel, resulting in a risk feature vector with complete dimensions and accurate information, effectively improving the comprehensiveness and accuracy of risk identification.

[0015] 2. This invention, relying on an intelligent risk assessment model and a closed-loop collaborative control mechanism, significantly enhances the scientific rigor and efficiency of risk early warning and response. The intelligent model, constructed by weighting key risk signals through an attention mechanism, combined with link evolution and dynamic causal network analysis, can accurately quantify comprehensive risks and determine early warning levels. The targeted early warning information and response plans, through coded responses and real-time feedback updates, form a complete monitoring and control closed loop. This not only improves the timeliness of risk response and the adaptability of responses but also continuously optimizes the model and data reserves, steadily improving the overall effectiveness of collaborative control for laboratory environmental monitoring. Attached Figure Description

[0016] Figure 1 A system architecture diagram of a collaborative control system for monitoring the laboratory environment of pathogenic microorganisms is provided in one embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0019] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0020] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0021] In practice, the server-side equipment deployed in a pathogenic microorganism laboratory environment monitoring and collaborative control system may consist of one or more devices. This pathogenic microorganism laboratory environment monitoring and collaborative control system can be implemented as: a business instance, a virtual machine, or hardware devices. For example, this pathogenic microorganism laboratory environment monitoring and collaborative control system can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this pathogenic microorganism laboratory environment monitoring and collaborative control system can be understood as software deployed on a cloud node, used to provide a pathogenic microorganism laboratory environment monitoring and collaborative control system to various user terminals. Alternatively, this pathogenic microorganism laboratory environment monitoring and collaborative control system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing various user terminals. Alternatively, this pathogenic microorganism laboratory environment monitoring and collaborative control system can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide a pathogenic microorganism laboratory environment monitoring and collaborative control system to various user terminals.

[0022] In terms of implementation, the collaborative control system for monitoring the laboratory environment of pathogenic microorganisms and the user terminal are mutually compatible. That is, if the collaborative control system for monitoring the laboratory environment of pathogenic microorganisms is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the collaborative control system for monitoring the laboratory environment of pathogenic microorganisms is implemented as a website, then the user terminal is implemented as a webpage; or if the collaborative control system for monitoring the laboratory environment of pathogenic microorganisms is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.

[0023] like Figure 1 The diagram shown is a system architecture diagram of a collaborative control system for monitoring the laboratory environment of pathogenic microorganisms, provided in an embodiment of the present invention.

[0024] The pathogenic microorganism laboratory environment monitoring collaborative control system 100 described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed as a website. Depending on the functions implemented, the pathogenic microorganism laboratory environment monitoring collaborative control system 100 may include a data acquisition module 101, a correlation assessment module 102, a model building module 103, a risk early warning module 104, an early warning strategy generation module 105, and a control feedback module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by an electronic device's processor and can perform a fixed function, stored in the electronic device's memory.

[0025] In this embodiment of the invention, in a collaborative control system for monitoring the environment of a pathogenic microorganism laboratory, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the collaborative control system for monitoring the environment of a pathogenic microorganism laboratory provided by this embodiment of the invention, the applicable scope of the collaborative control system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion, so as to achieve the purpose of quickly and flexibly expanding the collaborative control system for monitoring the environment of a pathogenic microorganism laboratory. In practical applications, the above-mentioned modules can be set in the same device or different devices, or they can be set in a virtual device, such as a service instance in a cloud server.

[0026] The following describes, with reference to specific embodiments, each component and its specific workflow of a collaborative control system for monitoring the laboratory environment of pathogenic microorganisms: Data acquisition module 101 is used to de-noise and couple environmental parameter data, reagent status data and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream of the laboratory; In this embodiment of the invention, when the data acquisition module performs noise reduction and coupling of environmental parameter data, reagent status data, and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream from the laboratory, it is specifically used for: Collect environmental parameter data, reagent status data, and personnel operation behavior data of pathogenic microorganisms in the laboratory; Based on a unified timestamp, environmental parameter data, reagent status data and personnel operation behavior data are time-series aligned to obtain aligned environmental time-series data, reagent status time-series data and personnel operation time-series data. Outlier removal was performed on environmental time-series data, reagent status time-series data, and personnel operation time-series data to obtain clean environmental time-series data, clean reagent status time-series data, and clean personnel operation time-series data for the laboratory. The time-series data of the clean environment, the time-series data of the test reagent status, and the time-series data of the clean personnel operation are timestamped to obtain a multi-source time-series data stream of the laboratory.

[0027] Temperature and humidity sensors, light sensors, temperature sensors built into reagent storage cabinets, high-definition cameras, weight sensors, high-definition surveillance cameras, access control systems, and workbench pressure sensors are deployed in the pathogen culture area, reagent storage area, and operation area of ​​the laboratory. The temperature and humidity sensors and light sensors collect temperature, humidity, and light intensity data for their respective areas every 5 minutes; this data represents the environmental parameters of the pathogens. The temperature sensors built into the reagent storage cabinets collect the reagent storage temperature in real time. The high-definition cameras capture the appearance of the reagents every 10 minutes, and the weight sensors provide real-time feedback on the remaining reagent quantity; this data represents the reagent status. The high-definition surveillance cameras capture personnel actions in real time, the access control system records the time personnel enter and exit the laboratory, and the workbench pressure sensors record the time personnel spend on the workbench; this data represents personnel behavior data. Each set of collected raw data is assigned a unique initial identifier, including the collecting device number, collection location information, and collection time information. This completes the collection of environmental parameter data, reagent status data, and personnel behavior data for pathogens in the laboratory.

[0028] Using the time of the laboratory's central clock server as the reference timestamp (format: year-month-day-hour-minute-second-millisecond), the initial collection time information in the collected environmental parameter data, reagent status data, and personnel operation behavior data is compared and calibrated with the reference timestamp of the central clock server. For any set of data from the three categories, if the deviation between its collection time and the reference timestamp is within 1 millisecond, the reference timestamp is directly used to replace the original collection time. If the deviation exceeds 1 millisecond, time compensation is performed based on the time synchronization logs between the acquisition device and the central clock server. The compensation method involves reverse-correcting the systematic error of the collection time through the historical synchronization records between the device and the central clock, ensuring the time dimension of the three types of data (environmental parameters, reagent status, and personnel operation) is consistent. This provides a reliable time reference for subsequent outlier removal and multi-source time-series data stream splicing. The compensated time is used as the final timestamp of the data set. After calibration, environmental parameter data, reagent status data, and personnel operation behavior data with the same final timestamp are categorized and integrated to obtain aligned environmental time-series data, reagent status time-series data, and personnel operation time-series data.

[0029] In the preset environmental time-series data, the normal ranges for temperature are 20°C to 25°C, humidity is 40% to 60%, and light intensity is 100 lux to 300 lux. Each data point in the environmental time-series data is compared to its corresponding normal range. Data points outside these ranges are identified as outliers, and the arithmetic mean of the three adjacent normal data points replaces the outlier. The replaced data point is then included in the clean environment time-series data. In the preset reagent status time-series data, the normal ranges for reagent storage temperature are 2°C to 8°C, the normal appearance of the reagent is no turbidity, no sediment, and no discoloration, and the normal trend of the remaining reagent quantity is a steady decrease. Each data point in the reagent status time-series data is compared to its corresponding normal range. The data points are compared against normal judgment criteria. Data points that do not meet the criteria are identified as outliers. The outlier is replaced by the value or status of the preceding normal data point. The replaced data point is then included in the clean reagent status time series data. The preset criteria for personnel operation time series data are that the single stay time of personnel does not exceed 120 minutes, the operation steps comply with the laboratory standard operating procedures, and the operation actions do not include unauthorized contact with pathogenic microorganism samples. Each data point in the personnel operation time series data is compared with the corresponding normal judgment criteria. Data points that do not meet the criteria are identified as outliers and are directly removed. The remaining data points after removal are included in the clean personnel operation time series data. This yields the laboratory clean environment time series data, clean reagent status time series data, and clean personnel operation time series data.

[0030] Using the final timestamps from the clean environment time-series data, clean reagent status time-series data, and clean personnel operation time-series data as the concatenation index, the three types of clean data with the same final timestamp are integrated by field consolidation. During consolidation, all fields of the clean environment time-series data, clean reagent status time-series data, and clean personnel operation time-series data are retained. Each data entry after consolidation contains clean environment parameter information, clean reagent status information, and clean personnel operation behavior information under the corresponding timestamp. All the consolidated timestamped data are sorted according to the order of the final timestamps. After sorting, a continuous and ordered dataset is formed, thus obtaining the laboratory's multi-source time-series data stream.

[0031] The correlation assessment module 102 is used to perform multimodal correlation assessment on environmental parameter data, reagent status data and personnel operation behavior data to obtain the laboratory's risk feature vector; In this embodiment of the invention, when the correlation assessment module performs a multimodal correlation assessment of environmental parameter data, reagent status data, and personnel operation behavior data to obtain the laboratory's risk feature vector, it is specifically used for: Semantic analysis was performed on environmental parameter data, reagent status data, and personnel operation behavior data to obtain experimentally relevant semantic label data for multi-source time-series data streams; Based on the entity attribute knowledge base of pathogenic microorganisms, the experimental-related semantic tag data is retrieved and matched to obtain structured risk instance data of pathogenic microorganisms. By correlating and statistically analyzing structured risk instance data with laboratory operational status data, we can obtain environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators for the laboratory. By combining environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators into tensors, a risk feature vector for the laboratory is obtained.

[0032] Based on the environmental parameter data, reagent status data, and personnel operation behavior data obtained from the data acquisition module, the core information of each data point is extracted one by one. The environmental parameter data extracts the collection area, temperature and humidity values, light intensity, and collection time. The reagent status data extracts the reagent number, storage temperature, appearance characteristics, remaining weight, and collection time. The personnel operation behavior data extracts the operator's identity, entry and exit time, operation action details, table dwell time, and collection time. At the same time, a preset semantic tag library is called. This tag library has been pre-entered with the core information and fixed tags of various types of data. For example, temperature and humidity values ​​correspond to the tags "culture area temperature" and "storage area humidity", reagent appearance characteristics correspond to the tag "reagent appearance status", and operation action details correspond to the tag "operation action type". The core information of each extracted data point is compared and matched with the tag library one by one, and a unique set of semantic tags is bound to each raw data point. After all data has been tagged, it is integrated to form experimental-related semantic tag data corresponding to the multi-source time-series data stream.

[0033] The entity attribute knowledge base of pathogenic microorganisms is pre-classified and stored according to the type of pathogenic microorganism. Each entry includes clear environmental control standards, reagent compatibility requirements, standard operating procedures and judgment criteria. For example, the environmental control standards specify that the temperature of the culture area corresponding to the specific pathogenic microorganism is 20-25 degrees Celsius and the humidity is 40%-60%. The reagent compatibility requirements specify that the storage temperature of the corresponding reagent is 2-8 degrees Celsius, the shelf life is 30 days, and the appearance is free of turbidity, precipitation and discoloration. The standard operating procedures specify that the time spent on the workbench at one time shall not exceed 120 minutes and the operator must have the corresponding qualifications. Each item of the experimentally relevant semantic tag data is compared and retrieved with the corresponding pathogenic microorganism entries in the knowledge base. If all the content of the semantic tag data meets the requirements of the corresponding entries in the knowledge base, structured risk instance data containing only the pathogenic microorganism type and semantic tag content is generated. If at least one item in the semantic tag data does not meet the requirements of the corresponding entries in the knowledge base, structured risk instance data containing the pathogenic microorganism type, semantic tag content, corresponding knowledge base standards, and specific deviation content is generated. After all entries have been retrieved and matched, all results are summarized to form complete structured risk instance data of pathogenic microorganisms.

[0034] The laboratory's operational status data includes specific information such as the cumulative runtime of each data acquisition device, reagent arrival date, list of qualified operators, and last calibration date of the equipment. Structured risk instance data is then linked to this operational status data one by one. The number of environmental anomalies in the structured risk instance data is counted, divided by the total number of environmental parameter data entries, and then the cumulative runtime of the corresponding data acquisition devices in the area is checked. If the cumulative runtime exceeds 800 hours, 0.05 is added to the above percentage result. The final value is the environmental stability index. Similarly, the number of reagent anomalies in the structured risk instance data is counted, divided by the total number of reagent status data entries, and then the corresponding reagent arrival date is checked. If the arrival date is more than 30 days from the current date, 0.03 is added to the above percentage result. The final value is the reagent compliance index. The number of abnormal personnel operation entries in the structured risk instance data is counted. This number is divided by the total number of personnel operation behavior data entries. Then, it is checked whether the operator is on the qualification registration list. If not, 0.1 is added to the above percentage result. The final value is the personnel operation risk indicator.

[0035] Using environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators as three independent data dimensions, these indicators are arranged and integrated in the above fixed order. The specific values ​​of each indicator are then concatenated to form a one-dimensional data sequence. During the concatenation process, the original values ​​of each indicator and the corresponding dimension identifier are retained to ensure that the data sequence fully reflects the actual situation of the three indicators. This one-dimensional data sequence is the laboratory's risk characteristic vector.

[0036] The model building module 103 is used to build an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams through an attention mechanism based on historical experimental data of pathogenic microorganisms and an expert knowledge base. In this embodiment of the invention, when the model building module executes the process of constructing an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams based on historical experimental data of pathogenic microorganisms and an expert knowledge base through an attention mechanism, it is specifically used for: Based on historical experimental data of pathogenic microorganisms, key factor analysis was performed on historical risk events in the laboratory and their corresponding environmental and operational parameters to obtain the laboratory's historical risk feature set. The constraints of the expert knowledge base in pathogenic microorganisms are analyzed in a structured manner to obtain the risk rule set of the laboratory; Based on the historical risk feature set and risk rule set, pattern recognition screening is performed on multi-source time series data streams to obtain a set of candidate key risk signals for the laboratory; Based on the correlation strength between different features and risk events in the historical risk feature set, the candidate key risk signal set is weighted and assigned to obtain the laboratory's weighted key risk signal. Based on the weighted key risk signals, the parameter structure and decision logic of the preset initial risk assessment model are configured to obtain the laboratory's intelligent risk assessment model.

[0037] The model building module, when performing weighted evaluation and allocation of candidate key risk signal sets based on the correlation strength between different features and risk events in the historical risk feature set to obtain the laboratory's weighted key risk signal, is specifically used for: Extract co-occurrence pairs between historical risk features and corresponding risk events; By performing probability statistics on the co-occurrence frequency and conditions of co-occurrence pairs, the quantitative correlation strength value of the historical risk feature set is obtained; Risk features are sorted in descending order based on their quantitative correlation strength values ​​to obtain a ranking table of the importance of risk features. Based on the risk feature importance ranking table, a mapping relationship from risk features to preset weight intervals is established to obtain the feature weight mapping table of the laboratory. Based on the feature weight mapping table, the risk features in the candidate key risk signal set are weighted and assigned values ​​to obtain the initial weighted risk signal of the laboratory. The initial weighted risk signal is normalized and its consistency is verified to obtain the laboratory's weighted critical risk signal.

[0038] The formula for calculating the quantitative correlation strength value is as follows: ; In the formula, For the aforementioned risk characteristics Risk events The quantitative correlation strength value between them, For the aforementioned risk characteristics Risk events Co-occurrence frequency, Risk event The total frequency of occurrence in the historical risk characteristic set, These are the preset base weight coefficients. The preset time decay rate, The time decay interval, Risk events in the expert knowledge base The severity level weight.

[0039] The co-occurrence frequency is derived from traversing the laboratory's historical risk feature set and all historical risk event records. The risk features corresponding to each risk event record are checked one by one, and the number of times that risk feature f and risk event e occur simultaneously and have a causal relationship is counted. During the statistical process, only records in which risk event e occurs within 24 hours after the occurrence of risk feature f are included, and cases of simultaneous occurrence without causal relationship are excluded. The final count is the co-occurrence frequency.

[0040] The total frequency of risk event e is derived from traversing all records of risk events in the historical risk feature set. Regardless of whether risk feature f is present, as long as the record is risk event e, it is included in the statistical scope. Each item is checked and counted to calculate the total number of times risk event e occurs in the historical data, which is the total frequency of risk event e.

[0041] The basic weighting coefficient is determined by domain experts in combination with expert knowledge base and past experimental data. Based on the stability of the inherent correlation between risk characteristics and risk events in historical data, the experts set the basic weighting coefficient to 0.7 when the impact of risk characteristics on risk events is not affected by time changes, and set it to 0.3 if it is significantly affected by time. This is used as a fixed preset value to balance the influence of inherent correlation and time factors.

[0042] The time decay rate is determined based on the pattern of the impact of risk characteristics changing over time in historical data. By statistically analyzing the probability changes of risk characteristics triggering the same risk event again after different time intervals, it was found that the impact of risk characteristics decays exponentially with the extension of time. Based on this, the time decay rate is preset to 0.05 to ensure that the longer the time interval, the more significant the decay effect.

[0043] The time decay interval is derived from the difference between the time when the risk feature f was first detected in historical experimental records and the actual time when the corresponding risk event e occurred. It is uniformly expressed in hours. The difference is calculated after extracting two time records and is accurate to the integer hour, serving as the core basis for time decay calculation.

[0044] The severity level weights are derived from the severity classification results of risk event e in the expert knowledge base. Experts classify risk events into three levels based on indicators such as the duration of experimental interruption, the degree of equipment damage, and the risk of pathogen leakage caused by the risk event. Level 1 events that cause experimental interruption for more than 24 hours or pathogen leakage are classified as Level 1, with a weight of 1.0; Level 2 events that cause experimental interruption for 4 to 24 hours are classified as Level 2, with a weight of 0.6; and Level 3 events that cause experimental interruption for less than 4 hours are classified as Level 3, with a weight of 0.3. The weight values ​​of the corresponding levels are directly extracted as the severity level weights.

[0045] This formula integrates multiple factors to accurately calculate the quantitative correlation strength between risk feature f and risk event e. The ratio of co-occurrence frequency to the total frequency of risk events reflects the basic correlation probability between the two. The combination of basic weight coefficient and time decay term balances the inherent correlation and the impact of time decay. The severity level weight highlights the degree of harm of the risk event itself. The final quantitative correlation strength value reflects both the closeness of the correlation between the two and takes into account the time effect and the severity of the event. It provides an accurate and reproducible quantitative basis for subsequent risk feature importance ranking and weight mapping, ensuring a high degree of consistency with the core requirements of weight evaluation and allocation in the model construction module.

[0046] Based on historical experimental data of pathogenic microorganisms, all environmental and operational parameters corresponding to each historical risk event in the laboratory were collected. Environmental parameters included temperature, humidity, ventilation rate, and air pressure difference in the laboratory. Operational parameters included the operation steps, operation duration, and type and method of use of experimental equipment by the experimenters. For each historical risk event, the values ​​of environmental and operational parameters at the time of the risk event were compared one by one with the parameter values ​​when no risk event occurred in the same experimental scenario. A parameter difference threshold was set at ±20% of the average value of parameters when no risk event occurred in the same experimental scenario. Environmental and operational parameters with differences exceeding this threshold were screened out. These screened parameters were identified as the key influencing factors of the corresponding historical risk events. All key influencing factors of historical risk events were classified and organized according to the type of risk event to form a historical risk feature set of the laboratory.

[0047] The risk event types in the document are categorized using a core logic of "risk trigger source + core impact dimension," consistent with the key impact factors in the historical risk feature set and the risk type identifiers in the early warning strategy library. Specific classifications are determined by considering the causes, affected objects, and associated parameter characteristics of the risk events, primarily including three core types: pathogenic microorganism leaks, where the core triggers are often related to abnormal environmental parameters, equipment sealing failures, or operational violations, and are associated with abnormal differences in environmental and operational parameters; experimental equipment failures, where the triggers are concentrated on abnormal equipment operating conditions, and are associated with relevant environmental adaptation parameters or operational parameters; and operational procedure violations, where the core trigger is non-compliance of personnel operating parameters, such as omitted operating steps, exceeding operating time limits, or incorrect equipment usage, and are only associated with abnormal differences in operational parameters. After classification, key impact factors within the same type are integrated to form a "type-key impact factor" correspondence, ultimately constituting a structured historical risk feature set, providing a clear classification basis for subsequent risk rule set construction and key risk signal screening.

[0048] The constraints of the expert knowledge base for pathogenic microorganisms are analyzed in a structured manner. All constraints related to laboratory risks are extracted from the expert knowledge base. Each constraint is broken down into three parts: preconditions, judgment conditions, and risk conclusions. The preconditions are clearly defined as the specific experimental scenario, the type of pathogenic microorganism, and the stage of the experiment. The judgment conditions are clearly defined as the specific numerical range of environmental or operational parameters. The risk conclusions are clearly defined as the corresponding risk type and risk level. All the decomposed clauses are checked one by one to remove clauses with duplicate content or conflicting judgment conditions. The checked clauses are then classified according to the category of preconditions to form a set of risk rules for the laboratory.

[0049] Based on historical risk feature sets and risk rule sets, pattern recognition and screening are performed on multi-source time-series data streams. These multi-source time-series data streams include real-time collected laboratory temperature, humidity, ventilation rate, laboratory personnel operation behavior, and pathogen growth status data. The multi-source time-series data streams are divided into different data stream subsets according to data type. Each data stream subset corresponds to a key influencing factor in the historical risk feature set or a judgment condition in the risk rule set. Time-series pattern matching is performed on each data stream subset. The matching is based on the parameter range of the key influencing factors in the historical risk feature set and the judgment condition in the risk rule set. When the value of a certain segment of time-series data is continuously within the parameter range of the corresponding key influencing factor in the historical risk feature set and meets the judgment condition in the risk rule set, the signal corresponding to that segment of time-series data is marked as a risk candidate signal. All marked risk candidate signals are summarized, and duplicate signals are removed to form a candidate key risk signal set for the laboratory.

[0050] Based on the correlation strength between different features and risk events in the historical risk feature set, a weight evaluation and allocation is performed on the candidate key risk signal set. The total number of times each key feature in the historical risk feature set appears in historical experimental data is counted, and the number of times the corresponding risk event occurs when the key feature appears is also counted. The number of risk events corresponding to a certain key feature is divided by the total number of times the feature appears to obtain the correlation strength value between the feature and the risk event. The correlation strength value ranges from 0 to 1. Each signal in the candidate key risk signal set is matched one-to-one with the key feature in the historical risk feature set, and the correlation strength value of the corresponding key feature is directly assigned to the candidate key risk signal as the weight value of the signal. All candidate key risk signals with weight values ​​are integrated to form the laboratory's weighted key risk signal.

[0051] Based on the weighted key risk signals, the parameter structure and decision logic of the preset initial risk assessment model are configured. The weighted key risk signals are arranged in descending order of weight value, and a weight value threshold is set. Signals with a weight value higher than 0.7 are set as high-priority risk signals, signals with a weight value between 0.3 and 0.7 are set as medium-priority risk signals, and signals with a weight value lower than 0.3 are set as low-priority risk signals. The risk conclusions in the risk rule set are divided into high-risk, medium-risk, and low-risk levels according to risk level. Decision logic is set for risk signals of different priorities. That is, when a high-priority risk signal appears, the model directly determines it as a high-risk level. When only medium-priority or low-priority risk signals appear, the model counts the number of signals and the sum of the weight values ​​of all signals. When the number of signals exceeds 3 or the sum of the weight values ​​exceeds 1.5, it is determined as a medium-risk level; otherwise, it is determined as a low-risk level. The above parameter structure and decision logic are written into the preset initial risk assessment model to complete the model configuration and obtain the laboratory's intelligent risk assessment model.

[0052] Iterate through all risk features in the historical risk feature set and all historical risk events recorded in the laboratory, pair each risk feature with the corresponding historical risk event when the feature appears, remove duplicate pairings, and form a unique co-occurrence relationship pair between the risk feature and the corresponding risk event.

[0053] For each pair of co-occurrence relationships between a risk feature and its corresponding risk event, the total number of times the risk feature appears in historical experimental data is counted, and the number of times the corresponding risk event occurs when the risk feature appears is also counted. The number of times the corresponding risk event occurs is divided by the total number of times the risk feature appears to obtain the conditional probability value of the pair of co-occurrence relationships. This conditional probability value is directly determined as the quantitative correlation strength value between the risk feature and its corresponding risk event.

[0054] All risk features in the historical risk feature set are arranged in descending order of their respective quantitative correlation strength values. For risk features with the same quantitative correlation strength value, they are arranged in descending order of the total number of occurrences of their corresponding risk events in the historical experimental data. The names of the risk features and their corresponding quantitative correlation strength values ​​are listed in order to form a risk feature importance ranking table.

[0055] Three weight ranges are preset: a high-weight range of 0.8 to 1.0, a medium-weight range of 0.5 to 0.79, and a low-weight range of 0.1 to 0.49. Risk features with a quantitative correlation strength value of 0.8 or higher are assigned to the high-weight range, risk features with a quantitative correlation strength value between 0.5 and 0.79 are assigned to the medium-weight range, and risk features with a quantitative correlation strength value between 0.1 and 0.49 are assigned to the low-weight range. Each risk feature in the risk feature importance ranking table is bound to its corresponding weight range to form a feature weight mapping table for the laboratory.

[0056] Iterate through each risk signal in the candidate key risk signal set, extract the risk features contained in each risk signal, compare the extracted risk features with the risk features in the feature weight mapping table, find the matching risk features, extract the weight interval corresponding to the risk feature, take the median value of the weight interval as the initial weight value of the risk signal, and summarize all risk signals with initial weight values ​​to form the laboratory's initial weighted risk signal.

[0057] The sum of the initial weights of all risk signals in the initial weighted risk signal is calculated. The initial weight of each risk signal is divided by this sum to obtain the normalized weight value of that risk signal. Then, a consistency check is performed. The check criteria are that the normalized weight values ​​of all risk signals are in the range of 0 to 1 and the sum of all normalized weight values ​​is 1. For risk signals that do not meet the check criteria, the median value of their corresponding weight range is readjusted and normalization is calculated again until all risk signals meet the check criteria. The risk signals with normalized weight values ​​that meet the check criteria are identified as the weighted critical risk signals of the laboratory.

[0058] The risk warning module 104 is used to perform link evolution on the risk feature vector based on the intelligent risk assessment model in order to obtain the laboratory's comprehensive risk score and warning level. In this embodiment of the invention, when the risk warning module performs a link evolution of the risk feature vector based on an intelligent risk assessment model to obtain the laboratory's comprehensive risk score and warning level, it is specifically used for: Based on the intelligent risk assessment model, a potential causal correlation analysis is performed on the risk feature vector to obtain the preliminary risk evolution path of the risk feature vector; Based on the initial risk evolution path, a dynamic causal network graph of risk feature vectors is constructed with risk signals as nodes and causal relationships as edges. Risk signals are extrapolated from the link transfer probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph; Based on the potential propagation sequence and scope of impact, the initial risk evolution path is prioritized to obtain the key risk propagation path with risk feature vectors; The risk signal strength and corresponding link transfer probability of nodes in the key risk propagation path are quantitatively evaluated to obtain the laboratory's comprehensive risk score; The laboratory's early warning level is obtained by comparing the comprehensive risk score with the risk threshold.

[0059] When the risk warning module executes the dynamic causal network graph that constructs risk feature vectors based on the initial risk evolution path, using risk signals as nodes and causal relationships as edges, it is specifically used for: The risk signals describing the risk state in the initial risk evolution path are traversed and calibrated to obtain the risk signal node set of the initial risk evolution path; Based on the causal rule base stored in the intelligent risk assessment model, confidence matching is performed on the causal relationships between nodes in the risk signal node set to obtain the causal relationship edge set of the risk signal node set; By topologically connecting the risk signal node set with the causal relationship edge set, the initial causal network structure of the risk feature vector is obtained. Based on the evolution time sequence of risk signals in the initial risk evolution path, timestamp attributes are injected into the corresponding nodes and edges in the initial causal network topology to obtain an enhanced causal network graph of risk feature vectors. Redundant edges are removed and isolated nodes are filtered from the enhanced causal network graph to obtain a dynamic causal network graph of risk feature vectors.

[0060] When the risk warning module performs risk signal deduction on the link transition probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph, it is specifically used for: Read the link transition probability matrix describing the causal relationship between nodes in different risk states from the rule base of the intelligent risk assessment model; Using the risk state of the risk evolution state set as the initial signal source, iterative state transitions are performed based on the link transition probability matrix to obtain the risk propagation path of the risk evolution state set; The path probability and cumulative impact of risk propagation paths are evaluated and screened to obtain a set of critical propagation paths for risk propagation. By integrating the risk status nodes traversed by different paths in the critical propagation path set, we obtain the set of affected nodes of the critical propagation path set; Based on the set of affected nodes, the transition state of risky nodes is assessed to obtain the potential propagation sequence and impact range of the dynamic causal network graph.

[0061] Based on the risk rule set and historical risk feature set stored in the intelligent risk assessment model, the historical risk event records corresponding to each risk feature in the risk feature vector are retrieved. The order of occurrence of each risk feature and the corresponding risk event is checked. The time correlation threshold is set to 24 hours. When the occurrence time of the risk feature is earlier than the occurrence time of the risk event and the time difference is within 24 hours, it is determined that there is a potential causal relationship between the risk feature and the corresponding risk event. According to the occurrence order and causal relationship of the risk features, the transmission path from the initial risk feature to the final risk event is sorted out. All the sorted transmission paths are summarized to form the preliminary risk evolution path of the risk feature vector.

[0062] Based on the initial risk evolution path, each risk signal contained in the path is set as an independent node, and the potential causal relationship between risk signals is set as an edge between nodes. The direction of the causal relationship is specified in the edge labeling information, that is, from the risk signal node as the cause to the risk signal node as the result. All risk signal nodes and directional causal relationship edges are integrated according to the structure of the initial risk evolution path to form a dynamic causal network graph of risk feature vectors.

[0063] The link transition probability stored in the intelligent risk assessment model is retrieved. This link transition probability is determined based on the proportion of the number of transitions between adjacent risk signal nodes in historical data to the total number of occurrences of the preceding risk signal node. Starting from each risk signal node in the dynamic causal network graph, the subsequent possible risk signal nodes are deduced according to the magnitude of the link transition probability. All subsequent nodes that can be deduced from each starting node and their order are recorded to form the potential propagation sequence of the dynamic causal network graph. The number of nodes contained in each potential propagation sequence is counted, and this number of nodes is the influence range of the corresponding potential propagation sequence.

[0064] Based on the impact range of potential propagation sequences and the severity level of the terminal risk events of the corresponding propagation sequences in the expert knowledge base, where the "severity level" comes from the expert knowledge base, experts will classify risk events into three levels according to indicators such as the duration of experimental interruption, the degree of equipment damage, and the risk of pathogen leakage caused by the risk event: Level 1 is caused by experimental interruption exceeding 24 hours or pathogen leakage, Level 2 is caused by experimental interruption of 4 to 24 hours, and Level 3 is caused by experimental interruption of less than 4 hours. Each level corresponds to a fixed severity level weight, providing a basis for prioritizing risk sequences. Priority screening criteria are set, that is, when the number of nodes included in the impact range exceeds 5, or the severity level of the terminal risk event is Level 1, the corresponding potential propagation sequence is determined to be a high-priority sequence. The preliminary risk evolution paths corresponding to all high-priority sequences are screened out, the evolution paths with duplicate path structures are removed, and the remaining evolution paths are summarized to form the key risk propagation paths of the risk feature vector.

[0065] The risk signal intensity corresponding to each node in the critical risk propagation path is extracted. This risk signal intensity is taken from the weight value in the weighted critical risk signal obtained by the model building module. At the same time, the link transition probability corresponding to each link is extracted. The risk signal intensity of each node is associated with the corresponding link transition probability and assigned a value. The results of the association assignment of the risk signal intensity of all nodes in the critical risk propagation path with the corresponding link transition probability are accumulated, and the accumulated value is the laboratory's comprehensive risk score.

[0066] The correlation assignment process involves first extracting the risk signal strength of each node in the critical risk propagation path, then extracting the link transition probability corresponding to that node, and finally multiplying the risk signal strength of the same node with the corresponding link transition probability to complete the correlation assignment.

[0067] The result of the correlation assignment is a quantified risk impact value. This value comprehensively reflects the risk intensity of the current node itself and the probability of the risk propagating downwards through the corresponding link, intuitively reflecting the contribution of a single "node-link" combination to the overall risk. The next step is to accumulate the correlation assignment results, that is, to summarize the correlation assignment results of all "node-link" combinations in the key risk propagation path, sum all the quantified values, and the final accumulated value is the laboratory's comprehensive risk score, which is used for subsequent comparison with risk thresholds to determine the warning level.

[0068] The comprehensive risk score is pre-defined into three threshold ranges: a comprehensive risk score between 0 and 30 is considered a low-risk threshold range, between 31 and 70 is a medium-risk threshold range, and between 71 and 100 is a high-risk threshold range. The obtained comprehensive risk score of the laboratory is compared with each of the three threshold ranges to determine the threshold range to which the comprehensive risk score belongs. The risk level of the corresponding threshold range is the warning level of the laboratory.

[0069] Traverse all risk signals describing risk states in the initial risk evolution path, extract the feature information corresponding to each risk signal one by one. The feature information includes the type of risk signal, the corresponding environmental or operational parameters, and the associated risk event type. Assign a unique node identifier to each extracted risk signal. The identifier content corresponds one-to-one with the feature information of the risk signal. Remove the risk signal entries that appear repeatedly during the extraction process, and integrate all risk signals with unique node identifiers to form the risk signal node set of the initial risk evolution path.

[0070] The causal rule base stored in the intelligent risk assessment model is retrieved. This causal rule base is formed by integrating the historical risk feature set and the causal association clauses in the risk rule set. Every combination of two nodes in the risk signal node set is traversed. The risk signal features corresponding to the two nodes in the combination are compared one by one with the clauses in the causal rule base. When the risk signal features of two nodes completely meet the preconditions and result conditions of a certain causal association clause, it is determined that there is a causal relationship between the two nodes. The threshold for determining the confidence of the causal relationship is set to 0.6. This threshold is taken from the average value of the correlation strength between risk features and risk events in the historical risk feature set. Node combinations that meet the causal relationship and have a confidence level higher than the threshold are determined as causal association pairs. A corresponding causal relationship edge is configured for each causal association pair. The edge information includes the direction of the causal relationship and the confidence level value. All configured causal relationship edges are integrated to form the causal relationship edge set of the risk signal node set.

[0071] Each node in the risk signal node set is taken as a vertex of the network, and each causal edge in the causal edge set is taken as a connection link between vertices. The connection direction between vertices is determined according to the association direction marked by the causal edge. All vertices and connection links are integrated in an orderly manner according to the causal association logic to ensure that the start and end points of each causal edge correspond to nodes in the risk signal node set, forming an initial causal network structure of risk feature vector with a clear topology.

[0072] The time point and duration of the first detection of each risk signal in the initial risk evolution path are extracted, and this time point and duration are injected as timestamp attributes into the corresponding nodes in the initial causal network topology. At the same time, the time interval between the risk signal corresponding to the cause node and the risk signal corresponding to the result node in each causal relationship pair is extracted, and this time interval is injected as a timestamp attribute into the corresponding causal relationship edge. This ensures that each node and each edge in the initial causal network topology has a unique and corresponding timestamp attribute, forming an enhanced causal network graph of risk feature vectors.

[0073] The criteria for redundant edges are defined as edges with a causal confidence level below 0.6 and edges with repeated causal relationships. All edges in the enhanced causal network graph are traversed, and edges that meet the criteria for redundancy are removed one by one. At the same time, the criteria for isolated nodes are defined as nodes that are not connected by any causal relationship edges. All nodes in the enhanced causal network graph are traversed, and nodes that meet the criteria for isolated nodes are filtered one by one. The causal logic coherence of the network structure after removing redundant edges and filtering isolated nodes is checked. After the check is passed, a dynamic causal network graph with risk feature vectors is formed.

[0074] The rule base of the intelligent risk assessment model is retrieved to locate the link transition probability matrix describing the causal relationship between nodes in different risk states. This matrix is ​​constructed based on the transition records of risk state nodes in the historical risk feature set. The row dimension of the matrix corresponds to the predecessor risk state node, and the column dimension corresponds to the successor risk state node. The values ​​in the matrix are the probability values ​​of the predecessor node transitioning to the successor node. This probability value is obtained by dividing the number of times the successor node appears after the predecessor node appears in the historical data by the total number of times the predecessor node appears. During the reading process, the integrity of the matrix needs to be verified to ensure that each risk state node has a corresponding row and column in the matrix, and there is no missing or redundant node information.

[0075] The currently monitored risk state in the risk evolution state set is identified as the initial signal source. This initial signal source is an active node with a timestamp attribute marked in the dynamic causal network graph. Based on the probability value of the row corresponding to the initial signal source in the link transition probability matrix, the successor risk state node with a non-zero probability value is selected as the target node for the first state transition. The target node of the first transition is used as the new current node, and the above operation of selecting successor nodes based on the matrix is ​​repeated. The maximum number of iteration transition steps is set to 10 steps, which is determined based on the average length of the historical risk evolution path. The iteration stops when the number of iteration steps reaches the maximum number of steps or when all the probability values ​​of the row corresponding to the current node are zero. The node transition order in each iteration process is recorded to form the risk propagation path of the risk evolution state set.

[0076] The path probability screening threshold is set to 0.3, which is taken from the average probability of risk propagation paths that trigger actual risk events in historical data. At the same time, the cumulative impact assessment criterion is set as the severity level of the risk event corresponding to the risk status node at the end of the path. Level 1 risk events correspond to high cumulative impact, and Level 2 and Level 3 risk events correspond to progressively lower cumulative impact. The path probability of each risk propagation path is calculated, which is the product of the transfer probabilities of all nodes in the path. After the calculation is completed, the path probability is compared with the screening threshold, and risk propagation paths with path probabilities higher than the threshold and high cumulative impact are screened out. All paths that meet the conditions are summarized to form the critical propagation path set of risk propagation paths.

[0077] Traverse each risk propagation path in the critical propagation path set, extract all risk status nodes contained in each path, record the unique identifier information of each risk status node, compare the node's unique identifier information, remove duplicate risk status nodes, and organize all the deduplicated risk status nodes according to the order in which the nodes appear in the path to form the affected node set of the critical propagation path set.

[0078] The occurrence frequency of each risk state node in the affected node set within the critical propagation path set is counted. The paths in the critical propagation path set are arranged in descending order of path probability. The resulting path sequence is the potential propagation sequence of the dynamic causal network graph. The total number of risk state nodes in the affected node set is counted, and this number represents the influence range of the dynamic causal network graph. Simultaneously, the transition state of each affected node is assessed, i.e., the number of successor nodes when the node acts as a predecessor node. The transition state information is then labeled to the corresponding node, thus completing the determination of the potential propagation sequence and influence range.

[0079] The early warning strategy generation module 105 is used to map the comprehensive risk score and early warning level to the preset risk early warning strategy library to obtain the laboratory's early warning information and handling plan. In this embodiment of the invention, when the early warning strategy generation module maps the comprehensive risk score and early warning level to a preset risk early warning strategy library to obtain the laboratory's early warning information and handling plan, it is specifically used for: Based on a pre-set risk warning strategy library, the warning levels are matched to obtain preliminary warning strategies for each warning level. Based on the standard risk threshold range in the risk warning strategy library, the deviation of the comprehensive risk score is compared to obtain the warning information to be filled in the comprehensive risk score. Based on the current state data of the risk feature vector and the intelligent risk assessment model, the dynamic variables in the early warning information to be filled are filled with contextual information to obtain the laboratory's early warning information. Based on the risk type and level identifier in the early warning information, a related search is performed in the risk early warning strategy database to obtain the basic treatment plan for the laboratory. By combining the risk evolution state set and the risk transmission path of the dynamic causal network diagram, the basic response plan is adjusted and optimized to adapt to the contingency plan, resulting in the laboratory response plan.

[0080] The system retrieves a pre-defined risk warning strategy library, which is built based on historical risk management cases and expert knowledge bases in the laboratory. The library is divided into three strategy subsets: high, medium, and low, according to the warning level. Each subset corresponds to a unique warning level label. The system iterates through the three strategy subsets in the library and accurately matches the warning level obtained from the risk warning module with the level label of the strategy subset. The matching criterion is that the level label and the text description of the warning level must be completely consistent. After a successful match, all warning strategy content in the corresponding strategy subset is extracted and integrated to form a preliminary warning strategy for the warning level.

[0081] The system retrieves preset standard risk threshold ranges from the risk warning strategy library. These ranges correspond one-to-one with warning levels, specifically divided into low-risk (0-30), medium-risk (31-70), and high-risk (71-100) standard threshold ranges. It calculates the difference between the comprehensive risk score and the midpoint value of the corresponding warning level standard threshold range. The difference is the comprehensive risk score minus the midpoint value. Based on the sign and absolute value of the difference, the position of the comprehensive risk score within the corresponding range is determined. This position information is compared with the preset deviation descriptions in the strategy library. The warning information entries containing dynamic variable placeholders corresponding to the comparison results are extracted to form the warning information to be filled for the comprehensive risk score.

[0082] The preset deviation descriptions are qualitative descriptions pre-defined in the risk warning strategy library. They are corely linked to the difference range between the comprehensive risk score and the midpoint of the corresponding warning level's standard threshold interval, clearly indicating the specific deviation of the comprehensive risk score within its respective threshold interval. Based on the standard risk threshold intervals and difference calculation logic specified in the document, they are categorized as follows: When the difference is negative and the absolute value is large, it is described as "significantly lower than the midpoint of the corresponding risk level interval, with the risk level at a low level within the interval"; when the difference is close to 0, it is described as "close to the midpoint of the corresponding risk level interval, with the risk level at a medium level within the interval"; when the difference is positive and the absolute value is large, it is described as "significantly higher than the midpoint of the corresponding risk level interval, with the risk level at a high level within the interval". These deviation descriptions are pre-stored in the risk warning strategy library. By comparing them with the actual difference in the comprehensive risk score, corresponding descriptions can be quickly matched and embedded into warning information entries containing dynamic variable placeholders, providing a basis for generating subsequent warning information to be filled in.

[0083] Extract all risk feature names, feature occurrence times, and corresponding parameter values ​​from the risk feature vector. Simultaneously, extract the current state data of the intelligent risk assessment model, which includes the risk rules currently matched by the model, the stability status of the model operation, and the latest risk feature identification results. Traverse all dynamic variable placeholders in the warning information to be filled. Each placeholder corresponds to a unique associated data type. Fill the placeholder with the specific content of the corresponding data type. After filling, check the semantic coherence of the warning information to ensure that the information is accurately expressed and that no variables are omitted, thus forming the laboratory's warning information.

[0084] The risk types and level indicators clearly marked in the early warning information are extracted. Risk types include pathogen leakage risk, experimental equipment failure risk, and operational procedure violation risk. Level indicators include high, medium, and low levels. Using risk type as the first search condition, a subset of treatment plans containing that risk type is retrieved from the risk warning strategy database. Then, using level indicator as the second search condition, treatment plan entries containing that level indicator are retrieved from the retrieved treatment plan subset. The search criteria are that the text descriptions of risk type and level indicator are completely consistent with the entries. All treatment measures in the successfully retrieved treatment plan entries are extracted to form the laboratory's basic treatment plan.

[0085] Extract the current stage of risk evolution from the risk evolution state set. This stage includes the initial risk stage, the development stage, and the diffusion stage. Simultaneously, extract the risk transmission path from the dynamic causal network diagram. This path includes all risk state nodes and the causal relationship directions between nodes. Adjust the execution order of the basic response plan measures according to the risk evolution stage. Prioritize preventive measures in the initial risk stage, control measures in the development stage, and elimination measures in the diffusion stage. Adjust the coverage of measures according to the number of nodes in the risk transmission path. Expand the coverage when there are more than 5 nodes and narrow the coverage when there are fewer than 5 nodes. Conduct practical verification of the adjusted and optimized response plan to ensure that all measures comply with the laboratory's safety management regulations, thus forming the laboratory's response plan.

[0086] The control feedback module 106 is used to encode and respond to early warning information and response plans, and update the feedback parameters monitored during the response process to historical experimental data, so as to realize the monitoring and collaborative control of the laboratory.

[0087] In this embodiment of the invention, when the control feedback module encodes and responds to the early warning information and the response plan, and updates the feedback parameters monitored during the response process to historical experimental data to achieve collaborative monitoring and control of the laboratory, it is specifically used for: The early warning information and the response plan are converted into commands to obtain the control command sequence for the laboratory. The control command sequence is distributed to the corresponding experimental equipment and personnel message terminals in the laboratory to initiate the execution process of the control command sequence; During the execution process, comprehensive status data is collected in real time through a monitoring sensor network as feedback parameters for the laboratory. The feedback parameters, early warning information, and response plans are associated and packaged to obtain the laboratory's response feedback record; The feedback records will be updated to the repository of historical experimental data to complete the closed loop of laboratory monitoring and collaborative control.

[0088] The risk level, risk type, and core impact parameters in the early warning information are traversed. At the same time, the specific measures, implementing entities, and operational requirements in the disposal plan are broken down. The textual early warning information is converted into operation instructions that the equipment can recognize, and the disposal plan is converted into step-by-step execution instructions. Among them, the equipment instructions correspond to the specific operating parameter adjustments of the experimental equipment, and the personnel instructions correspond to the textual operation guidelines. After the conversion is completed, the consistency between the instructions and the early warning information and disposal plan is verified one by one to ensure that there are no instruction deviations or omissions, thus forming a control instruction sequence for the laboratory.

[0089] According to the execution subject marked in each instruction in the control instruction sequence, equipment-type instructions are accurately distributed to the control terminal of the corresponding experimental equipment based on the unique ID of the laboratory equipment, and personnel-type instructions are distributed to the dedicated message terminal of the experimental personnel based on their work number. After distribution, the system waits for confirmation feedback. The feedback timeout threshold is set to 1 minute. If no confirmation is received within the timeout period, the instruction is re-distributed. After confirming that all instructions have been received, the execution start program of the control instruction sequence is triggered to start the instruction execution process.

[0090] Relying on a monitoring sensor network covering the entire laboratory area, which includes environmental sensor nodes, equipment operation sensor nodes, and personnel operation sensor nodes, data is collected in real time every 30 seconds during the execution of control commands. Environmental sensor nodes collect parameters such as temperature, humidity, and air pressure; equipment operation sensor nodes collect parameters such as equipment power and operating status after command execution; and personnel operation sensor nodes collect data on the progress and completion status of personnel in executing commands. During the collection process, invalid data exceeding ±30% of the normal parameter range is removed, and the valid data is integrated into the laboratory's feedback parameters.

[0091] Extract equipment operation data, environmental data, and personnel operation data from the feedback parameters, associate and bind them with the corresponding early warning information and response plans, add auxiliary information such as instruction execution start time, data collection period, and feedback parameter statistical results, and encapsulate them into a structured document according to the logical order of "early warning - response - feedback". Each document is marked with a unique identification number to form the laboratory's response feedback record.

[0092] The repository for historical experimental data is opened. This repository stores data in a hierarchical structure of "risk type - occurrence date". The handling feedback records are stored in the corresponding category directory according to their corresponding risk type and execution date. Before storage, the consistency of the record format with the repository specification is verified. If there is a discrepancy, the record is returned for correction and re-storage. If the record is consistent, the write operation is completed. At the same time, the risk event handling case set and feedback parameter dataset in the historical experimental data are updated synchronously. This provides data support for the subsequent optimization of the intelligent risk assessment model and the supplementation of the risk feature set, realizing a closed loop of monitoring and collaborative control in the laboratory.

[0093] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0094] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A collaborative control system for monitoring and controlling the laboratory environment of pathogenic microorganisms, characterized in that, The system includes a data acquisition module, a correlation assessment module, a model building module, a risk warning module, a warning strategy generation module, and a control feedback module, wherein: The data acquisition module is used to de-noise and couple environmental parameter data, reagent status data and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream of the laboratory. The correlation assessment module is used to perform multimodal correlation assessments on environmental parameter data, reagent status data, and personnel operation behavior data to obtain the laboratory's risk feature vector; The model building module is used to construct an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams through an attention mechanism based on historical experimental data of pathogenic microorganisms and an expert knowledge base. The risk warning module is used to perform link evolution on the risk feature vector based on the intelligent risk assessment model in order to obtain the laboratory's comprehensive risk score and warning level. The early warning strategy generation module is used to map the comprehensive risk score and early warning level to the preset risk early warning strategy library to obtain the laboratory's early warning information and response plan; The control feedback module is used to encode and respond to early warning information and response plans, and update the feedback parameters monitored during the response process to historical experimental data to achieve collaborative monitoring and control of the laboratory.

2. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, The data acquisition module, when performing noise reduction and coupling of environmental parameter data, reagent status data, and personnel operation behavior data collected in the laboratory to obtain a multi-source time-series data stream from the laboratory, is specifically used for: Collect environmental parameter data, reagent status data, and personnel operation behavior data of pathogenic microorganisms in the laboratory; Based on a unified timestamp, environmental parameter data, reagent status data and personnel operation behavior data are time-series aligned to obtain aligned environmental time-series data, reagent status time-series data and personnel operation time-series data. Outlier removal was performed on environmental time-series data, reagent status time-series data, and personnel operation time-series data to obtain clean environmental time-series data, clean reagent status time-series data, and clean personnel operation time-series data for the laboratory. The time-series data of the clean environment, the time-series data of the test reagent status, and the time-series data of the clean personnel operation are timestamped to obtain a multi-source time-series data stream of the laboratory.

3. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, When performing a multimodal correlation assessment of environmental parameter data, reagent status data, and personnel operation behavior data to obtain the laboratory's risk feature vector, the correlation assessment module is specifically used for: Semantic analysis was performed on environmental parameter data, reagent status data, and personnel operation behavior data to obtain experimentally relevant semantic label data for multi-source time-series data streams; Based on the entity attribute knowledge base of pathogenic microorganisms, the experimental-related semantic tag data is retrieved and matched to obtain structured risk instance data of pathogenic microorganisms. By correlating and statistically analyzing structured risk instance data with laboratory operational status data, we can obtain environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators for the laboratory. By combining environmental stability indicators, reagent compliance indicators, and personnel operation risk indicators into tensors, a risk feature vector for the laboratory is obtained.

4. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, The model building module, when executing the process of constructing an intelligent risk assessment model for the laboratory by weighting key risk signals in multi-source time-series data streams using an attention mechanism based on historical experimental data of pathogenic microorganisms and an expert knowledge base, is specifically used for: Based on historical experimental data of pathogenic microorganisms, key factor analysis was performed on historical risk events in the laboratory and their corresponding environmental and operational parameters to obtain the laboratory's historical risk feature set. The constraints of the expert knowledge base in pathogenic microorganisms are analyzed in a structured manner to obtain the risk rule set of the laboratory; Based on the historical risk feature set and risk rule set, pattern recognition screening is performed on multi-source time series data streams to obtain a set of candidate key risk signals for the laboratory; Based on the correlation strength between different features and risk events in the historical risk feature set, the candidate key risk signal set is weighted and assigned to obtain the laboratory's weighted key risk signal. Based on the weighted key risk signals, the parameter structure and decision logic of the preset initial risk assessment model are configured to obtain the laboratory's intelligent risk assessment model.

5. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 4, characterized in that, The model building module, when performing weighted evaluation and allocation of candidate key risk signal sets based on the correlation strength between different features and risk events in the historical risk feature set to obtain the laboratory's weighted key risk signal, is specifically used for: Extract co-occurrence pairs between historical risk features and corresponding risk events; By performing probability statistics on the co-occurrence frequency and conditions of co-occurrence pairs, the quantitative correlation strength value of the historical risk feature set is obtained; Risk features are sorted in descending order based on their quantitative correlation strength values ​​to obtain a ranking table of the importance of risk features. Based on the risk feature importance ranking table, a mapping relationship from risk features to preset weight intervals is established to obtain the feature weight mapping table of the laboratory. Based on the feature weight mapping table, the risk features in the candidate key risk signal set are weighted and assigned values ​​to obtain the initial weighted risk signal of the laboratory. The initial weighted risk signal is normalized and its consistency is verified to obtain the laboratory's weighted critical risk signal.

6. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, When the risk warning module executes a link evolution of the risk feature vector based on the intelligent risk assessment model to obtain the laboratory's comprehensive risk score and warning level, it is specifically used for: Based on the intelligent risk assessment model, a potential causal correlation analysis is performed on the risk feature vector to obtain the preliminary risk evolution path of the risk feature vector; Based on the initial risk evolution path, a dynamic causal network graph of risk feature vectors is constructed with risk signals as nodes and causal relationships as edges. Risk signals are extrapolated from the link transfer probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph; Based on the potential propagation sequence and scope of impact, the initial risk evolution path is prioritized to obtain the key risk propagation path with risk feature vectors; The risk signal strength and corresponding link transfer probability of nodes in the key risk propagation path are quantitatively evaluated to obtain the laboratory's comprehensive risk score; The laboratory's early warning level is obtained by comparing the comprehensive risk score with the risk threshold.

7. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 6, characterized in that, When the risk warning module executes the dynamic causal network graph that constructs risk feature vectors based on the initial risk evolution path, using risk signals as nodes and causal relationships as edges, it is specifically used for: The risk signals describing the risk state in the initial risk evolution path are traversed and calibrated to obtain the risk signal node set of the initial risk evolution path; Based on the causal rule base stored in the intelligent risk assessment model, confidence matching is performed on the causal relationships between nodes in the risk signal node set to obtain the causal relationship edge set of the risk signal node set; By topologically connecting the risk signal node set with the causal relationship edge set, the initial causal network structure of the risk feature vector is obtained. Based on the evolution time sequence of risk signals in the initial risk evolution path, timestamp attributes are injected into the corresponding nodes and edges in the initial causal network topology to obtain an enhanced causal network graph of risk feature vectors. Redundant edges are removed and isolated nodes are filtered from the enhanced causal network graph to obtain a dynamic causal network graph of risk feature vectors.

8. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 6, characterized in that, When the risk warning module performs risk signal deduction on the link transition probabilities stored in the intelligent risk assessment model to obtain the potential propagation sequence and impact range of the dynamic causal network graph, it is specifically used for: Read the link transition probability matrix describing the causal relationship between nodes in different risk states from the rule base of the intelligent risk assessment model; Using the risk state of the risk evolution state set as the initial signal source, iterative state transitions are performed based on the link transition probability matrix to obtain the risk propagation path of the risk evolution state set; The path probability and cumulative impact of risk propagation paths are evaluated and screened to obtain a set of critical propagation paths for risk propagation. By integrating the risk status nodes traversed by different paths in the critical propagation path set, we obtain the set of affected nodes of the critical propagation path set; Based on the set of affected nodes, the transition state of risky nodes is assessed to obtain the potential propagation sequence and impact range of the dynamic causal network graph.

9. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, When the early warning strategy generation module maps the comprehensive risk score and early warning level to a preset risk early warning strategy library to obtain the laboratory's early warning information and response plan, it is specifically used for: Based on a pre-set risk warning strategy library, the warning levels are matched to obtain preliminary warning strategies for each warning level. Based on the standard risk threshold range in the risk warning strategy library, the deviation of the comprehensive risk score is compared to obtain the warning information to be filled in the comprehensive risk score. Based on the current state data of the risk feature vector and the intelligent risk assessment model, the dynamic variables in the early warning information to be filled are filled with contextual information to obtain the laboratory's early warning information. Based on the risk type and level identifier in the early warning information, a related search is performed in the risk early warning strategy database to obtain the basic treatment plan for the laboratory. By combining the risk evolution state set and the risk transmission path of the dynamic causal network diagram, the basic response plan is adjusted and optimized to adapt to the contingency plan, resulting in the laboratory response plan.

10. The pathogenic microorganism laboratory environment monitoring and collaborative control system as described in claim 1, characterized in that, The control feedback module, when encoding and responding to early warning information and response plans, and updating the monitored feedback parameters to historical experimental data to achieve collaborative monitoring and control of the laboratory, is specifically used for: The early warning information and the response plan are converted into commands to obtain the control command sequence for the laboratory. The control command sequence is distributed to the corresponding experimental equipment and personnel message terminals in the laboratory to initiate the execution process of the control command sequence; During the execution process, comprehensive status data is collected in real time through a monitoring sensor network as feedback parameters for the laboratory. The feedback parameters, early warning information, and response plans are associated and packaged to obtain the laboratory's response feedback record; The feedback records will be updated to the repository of historical experimental data to complete the closed loop of laboratory monitoring and collaborative control.

Images