Laboratory risk adaptive evaluation method and system based on AI algorithm

Abstract

The invention relates to the technical field of laboratory risk management and predictive analysis, in particular to a laboratory risk adaptive evaluation method and system based on an AI algorithm. Comprising the following steps: S1, obtaining fusion data based on a key security interface and an experimental activity unit, and determining first linkage data and second linkage data based on the fusion data; s2, constructing a first linkage data time sequence based on the first linkage data, and dividing the first linkage data time sequence into N stages; s3, obtaining the in-situ matching influence degree of each stage based on the N stages, and obtaining the dislocation matching influence degree of each stage based on the N stages; and comparing the co-location matching influence degree with the maximum dislocation matching influence degree in the same stage, and determining a dominant factor. According to the method, spatio-temporal data are fused, direct and indirect accurate risk traceability is realized, hidden system risks in a complex experiment environment of colleges and universities are dynamically evaluated, and the early recognition and self-adaptive early warning capabilities are improved.

Description

Technical Field

[0001] This invention relates to the field of laboratory risk management and predictive analysis technology, and in particular to a laboratory risk adaptive assessment method and system based on AI algorithms. Background Technology

[0002] Artificial intelligence algorithms, through learning from historical data and modeling real-time information, have been gradually applied to the dynamic perception of laboratory safety risks. Adaptive laboratory risk assessment refers to a system's ability to autonomously adjust risk assessment model parameters and warning thresholds based on changes in experimental progress and facility status. Existing AI-based assessment methods typically focus on a single laboratory unit, building predictive models from monitoring data of specific equipment or operational procedures. These methods demonstrate certain advantages in identifying abnormal fluctuations within known patterns, providing an effective path to improve the automation level of laboratory safety management.

[0003] However, existing methods have significant limitations in addressing the complex environments of university laboratories that involve multidisciplinary collaboration and shared infrastructure. Current solutions largely rely on isolated monitoring point data and static risk assessment rules, making it difficult to effectively characterize the dynamic and time-delayed interactions between different experimental activity units and shared support facility systems. Furthermore, they fail to identify cross-domain risk transmission caused by indirect coupling. This results in traditional methods lacking the ability to perceive and warn of systemic and latent risks arising from mismatches between facility rhythms and experimental cycles, or from indirect cross-laboratory influences.

[0004] Therefore, there is an urgent need to construct a risk assessment method that can adapt to the characteristics of multi-source heterogeneity and spatiotemporal coupling in university laboratories. Through in-depth correlation and collaborative analysis of key safety interface data and data across the entire experimental activity chain, it is possible to trace the source of complex risk paths and conduct forward-looking adaptive assessments. Summary of the Invention

[0005] To overcome the shortcomings of multi-source dynamic risk coupling assessment, this invention provides a laboratory risk adaptive assessment method and system based on AI algorithms.

[0006] The technical implementation scheme of the present invention is: an adaptive laboratory risk assessment method based on AI algorithms, comprising the following steps: S1: Obtain fused data based on key security interfaces and experimental activity units, and determine the first linkage data and the second linkage data based on the fused data; S2: Construct a first linkage data time sequence based on the first linkage data and divide the first linkage data time sequence into N stages; S3: Obtain the influence of same-position matching in each of the N stages, and obtain the influence of mismatched matching in each of the N stages; compare the influence of same-position matching in the same stage with the influence of the maximum mismatched matching to determine the dominant factor; S4: Calculate the cycle matching degree based on the dominant factors, establish a comprehensive risk judgment matrix with the cycle matching degree value range as the vertical axis and the stage risk dominant factor type as the horizontal axis, and determine the stage overall risk value according to the comprehensive risk judgment matrix.

[0007] Preferably, the process of obtaining fused data based on key security interfaces and experimental activity units includes: The time difference between the critical safety interface and the experimental activity unit is obtained. Based on the time difference, the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit are aligned by time offset and matched by spatial position to obtain fused data. The critical safety interface refers to the functional interface that ensures the exchange of matter, energy, or information between the facility system and the experimental operating space. The continuous monitoring data refers to a set of real-time time-series parameters that reflect the operating status of critical security interfaces; The dynamic data refers to the set of real-time process variables that characterize the progress and internal state changes of experimental activities.

[0008] Preferably, determining the first linkage data and the second linkage data based on the fused data includes: Based on fused data, if there is a coupling relationship between the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit, the continuous monitoring data and the dynamic data will be combined into the first linkage data. Based on fused data, if the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit are not coupled, the continuous monitoring data and the experimental background information are combined into the second linkage data; the experimental background information refers to the set of basic state parameters of the laboratory environment and facilities that are independent of the dynamic changes of a specific experimental activity unit.

[0009] Preferably, the step of constructing a first linkage data time series based on the first linkage data and dividing the first linkage data time series into N stages includes: Based on the chronological order of the timestamps of the data records in the first linkage data, a time sequence of the first linkage data is constructed; If the dynamic data of the experimental activity unit contains predefined key operation transition point identifiers, then the first linkage data time sequence is divided into N stages based on the key operation transition point identifiers. If the dynamic data of the experimental activity unit does not contain predefined key operation transition point identifiers, then an unsupervised clustering analysis based on a sliding window is performed on the time series sequence of the first linkage data to obtain the clustering analysis results. Based on the boundaries of the clustering analysis results, the time series sequence of the first linkage data is divided into N stages; where each stage corresponds to a time window with relatively consistent process states.

[0010] Preferably, obtaining the degree of influence of co-matching at each of the N stages includes: Based on the type and intensity parameters of the experimental activity units in each stage, a predictable baseline sequence of continuous monitoring data for key safety interfaces is constructed. Obtain the actual coupling sequence of key safety interfaces within a phase; the actual coupling sequence refers to the sequence of continuous monitoring data of key safety interfaces extracted from the first linkage data that directly corresponds to the current experimental activity unit; Calculate the sum of squared residuals between the coupled actual sequence and the expected baseline sequence of continuous monitoring data; Calculate the ratio of the sum of squared residuals to the sum of squared deviations of the coupled actual sequence, and normalize the ratio to obtain the normalized residual ratio. Calculate the difference between 1 and the normalized residual ratio to obtain the degree of influence of the co-matching at each stage; The closer the degree of influence of the same position matching is to 1, the greater the expected contribution of the experimental activity unit to the state of the critical security interface within the stage.

[0011] Preferably, obtaining the degree of mismatch impact at each of the N stages includes: A misaligned matching topology model is constructed using all critical security interfaces as nodes and the physical connection paths of the security facility system as edges. The misaligned matching topology model refers to a directed network graph that labels the transmission direction and characteristic time delay parameters. Based on the transmission direction and characteristic time delay parameters provided by the mismatched topology model, a time-delayed translation sequence of dynamic data for each upstream experimental activity unit is constructed. Obtain the actual transmission sequence of the target critical security interface at the current stage; if the target critical security interface has second linkage data at the current stage, then the continuous monitoring data sequence in the second linkage data is preferentially used as the actual transmission sequence; otherwise, the regular monitoring data is used as the actual transmission sequence; the actual transmission sequence is used to analyze the existing cross-domain risk transmission relationship; Calculate the peak value of the time-delay cross-correlation function between each time-delay shift sequence and the actual transmission sequence; The peak value of the time-delay cross-correlation function was subjected to statistical significance testing and normalization. The peak value of the normalized time-delay cross-correlation function that passes the significance test is defined as the degree of mismatch influence of the corresponding upstream experimental activity unit; The closer the degree of mismatched influence is to 1, the higher the weight of the non-directly coupled cross-domain transmission influence generated by the upstream experimental activity unit on the target's critical security interface through the facility network within the stage.

[0012] Preferably, the step of comparing the influence of identical matches and the influence of the maximum mismatch in the same stage to determine the dominant factor includes: If the impact of same-position matching within the same stage is greater than the sum of the impact of maximum mismatched matching and the preset advantage threshold, then the stage risk is determined to be dominated by same-position matching factors. If the maximum impact of mismatch within the same stage is greater than the sum of the impact of same-position matching and the preset advantage threshold, then the stage risk is determined to be dominated by mismatch factors. If the absolute value of the difference between the degree of influence of same-position matching and the degree of influence of maximum mismatch within the same stage is less than or equal to the preset advantage threshold, then the stage risk is determined to be affected by both same-position matching factors and mismatch matching factors. When the risk of a stage is determined to be dominated by mismatch factors, the upstream experimental activity unit that provides the maximum degree of mismatch impact in the stage is identified, and the end time of the active stage of the upstream experimental activity unit is no later than the start time of the stage.

[0013] Preferably, the calculation of the periodic matching degree based on the dominant factor includes: Based on historical continuous monitoring data, a time series periodic detection algorithm is used to extract the dominant period length and phase offset of the operating status of each critical safety interface, and generate a set of operating period parameters for the critical safety interfaces. Based on the historical dynamic data and corresponding metadata of the experimental activity units, a time series period detection algorithm is used to extract the dominant period length and phase offset of each type of experimental activity unit in terms of execution frequency and operation intensity, and generate a set of execution period parameters for the experimental activity units. Calculate the cycle matching degree for the current experimental activity unit and the current critical safety interface: Obtain the execution cycle parameters of the current experimental activity unit type and the runtime parameters of the current critical safety interface; Calculate the harmonic ratio between the execution cycle length of the current experimental activity unit type and the runtime length of the current critical security interface; Calculate the absolute value of the phase difference between the execution cycle phase of the current experimental activity unit type and the current running cycle phase of the critical safety interface at the current moment; The harmonic ratio and the absolute value of the phase difference are normalized and weighted and fused to output the period matching degree value. The period matching degree value range is set to [0, 1]. The closer the period matching degree value is to 1, the higher the degree of matching between the execution cycle of the experimental activity unit and the operation cycle of the key safety interface. The closer the period matching degree value is to 0, the lower the degree of matching.

[0014] Preferably, the step of establishing a comprehensive risk assessment matrix with the periodic matching degree value range as the vertical axis and the type of dominant risk factor for a given stage as the horizontal axis, and determining the overall risk value for a stage based on the comprehensive risk assessment matrix, includes: Based on the cycle matching degree value of the current experimental activity unit and the type of risk-dominant factors at the current stage, the comprehensive risk level is determined by mapping in the comprehensive risk judgment matrix; The mapping rules of the comprehensive risk assessment matrix include: pre-setting a basic risk level for each type of risk-dominant factor in each stage; when the cycle matching degree value is lower than the set threshold for cycle matching degree, the comprehensive risk level is increased based on the basic risk level; when the cycle matching degree value is higher than the set threshold for cycle matching degree, the comprehensive risk level remains at the basic risk level. The risk assessment parameters are dynamically adjusted based on the periodic matching degree value: when the periodic matching degree value is detected to be lower than the set threshold for M consecutive periods, the preset advantage threshold is reduced. Based on the different intervals in which the periodic matching degree value is located, differentiated weighting coefficients are assigned to the degree of influence of same-position matching and the degree of influence of misaligned matching, and the overall risk value of the stage is dynamically calculated after weighting.

[0015] Preferably, an AI-based adaptive risk assessment system for laboratories includes: The data fusion and linkage generation module is used to obtain the time difference between the key safety interface and the experimental activity unit, align the continuous monitoring data and dynamic data to obtain fused data, and generate the first linkage data or the second linkage data according to the coupling relationship. The time-series segmentation module is used to construct a time-series sequence of the first linkage data based on the first linkage data, and to divide the sequence into N stages according to the key operation transition point identifiers or cluster analysis results. The matching impact analysis and dominant factor determination module is used to calculate the degree of influence of same-position matching and misaligned matching at each stage, and to determine the dominant risk factor at each stage by comparing the two. The cycle matching and comprehensive risk assessment module is used to calculate the cycle matching degree, establish a comprehensive risk judgment matrix, and dynamically adjust parameters to determine the overall risk value of the stage.

[0016] Beneficial Effects: This invention constructs a data foundation capable of supporting risk assessment along both direct interaction and cross-domain transmission paths by accurately integrating and classifying the spatiotemporal relationships of monitoring data from key safety interfaces and experimental activity units. Through time-series staged analysis and quantitative calculation of the impact of matching and mismatching, it achieves precise tracing and responsibility separation for risks directly caused by experimental operations and risks indirectly transmitted through facility networks. Furthermore, it introduces periodic matching degree assessment of experiments and facilities and establishes a comprehensive risk judgment matrix to dynamically adjust judgment thresholds and weights, enabling the system to perceive macro-level operational rhythms and adaptively adjust its sensitivity to micro-level risks. Ultimately, it forms a comprehensive assessment capability from micro-events to macro-level trends, from direct attribution to indirect tracing, significantly improving the early identification and adaptive warning level of implicit and time-delayed system risks in complex scenarios involving multidisciplinary collaboration and shared facilities in universities. Attached Figure Description

[0017] Figure 1 This is a flowchart of the AI-based adaptive risk assessment method for laboratories according to the present invention. Figure 2 This is a structural diagram of the AI-based adaptive laboratory risk assessment system of the present invention. Detailed Implementation

[0018] The present invention will be further described below with reference to specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0019] Example 1: An AI-based adaptive risk assessment method for laboratories, such as... Figure 1 As shown, it includes the following steps: S1-1: Obtain fused data based on key security interfaces and experimental activity units, including: The time difference between the critical safety interface and the experimental activity unit is obtained. Based on the time difference, the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit are aligned by time offset and matched by spatial position to obtain fused data. The critical safety interface refers to the functional interface that ensures the exchange of matter, energy, or information between the facility system and the experimental operating space. The continuous monitoring data refers to a set of real-time time-series parameters that reflect the operating status of critical security interfaces; The dynamic data refers to the set of real-time process variables that characterize the progress and internal state changes of experimental activities.

[0020] It is important to note that this step aims to establish a precise spatiotemporal data foundation to overcome the limitations of traditional monitoring methods that separate facility status from experimental activity data. In practice, the operating parameters of critical safety interfaces are continuously collected through pre-deployed IoT sensors, forming continuous monitoring data. For example, the real-time time series of duct velocity and hazardous gas concentration monitored at the exhaust duct interface of a laboratory fume hood (a critical safety interface) constitutes continuous monitoring data. The progress information of experimental activity units is automatically recorded through experimental logs or equipment control interfaces, generating dynamic data such as the temperature of the reactor, heating power, and reagent addition rate—a set of variables characterizing real-time changes in the experimental process. The characteristic time difference between the two needs to be calculated and compensated; this difference typically reflects the time taken for matter or energy to travel along a specific path. Under optimal facility conditions, this time difference remains relatively stable; if the monitored time difference is abnormally prolonged or irregular, it often indicates obstructed conduction paths or equipment malfunction. The raw monitoring data and experimental dynamic data are not synchronized in time and space; direct use of these data could lead to causal misjudgments. Therefore, temporal offset and spatial location matching are necessary to precisely align the causal relationship between "a reagent injection behavior" and "a specific vent pollutant concentration increase event" in time and space. The fused data generated after alignment and matching constitutes a sequence of related events that reflects "who, when, where, and what kind of impact." This provides a crucial analytical starting point for solving the problem pointed out in the background art, where isolated data cannot characterize cross-unit interaction relationships.

[0021] S1-2: Determining the first linkage data and the second linkage data based on the fused data, including: Based on fused data, if there is a coupling relationship between the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit, the continuous monitoring data and the dynamic data will be combined into the first linkage data. Based on fused data, if the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit are not coupled, the continuous monitoring data and the experimental background information are combined into the second linkage data; the experimental background information refers to the set of basic state parameters of the laboratory environment and facilities that are independent of the dynamic changes of a specific experimental activity unit.

[0022] It should be noted that, to overcome the shortcomings of traditional methods that conflate various monitoring signals, this step categorizes the fused data based on the inherent correlations between the data. Specifically, if the continuous monitoring data of a critical safety interface can be interpreted by the dynamic data of a specific experimental activity unit, then a coupling relationship is determined between the two. This type of data with direct causal or strong temporal correlation is combined into the first linkage data, the core purpose of which is to quantitatively analyze the immediate impact and explicit risks of specific experimental operations on their dependent facilities.

[0023] Conversely, when changes in continuous monitoring data cannot be traced back to any ongoing experimental activity—for example, when abnormal fluctuations in water pipe pressure occur in a laboratory during a school-wide shutdown—it indicates that the signal originates from the facility's background condition or other indirect interference. In this case, the continuous monitoring data is combined with experimental background information to form a second set of linked data. This second set of linked data primarily reflects the health status of the facility system itself and the level of background noise. By distinguishing and constructing these two types of data sets, this method effectively separates risks "directly caused by experiments" from risks "rooted in the facility's background," thereby solving the prominent problem in background techniques where it is difficult to trace the true origin of risks due to data confounding.

[0024] S2: Construct a first linkage data time series based on the first linkage data and divide the first linkage data time series into N stages, including: Based on the chronological order of the timestamps of the data records in the first linkage data, a time sequence of the first linkage data is constructed; If the dynamic data of the experimental activity unit contains predefined key operation transition point identifiers, then the first linkage data time sequence is divided into N stages based on the key operation transition point identifiers. If the dynamic data of the experimental activity unit does not contain predefined key operation transition point identifiers, then an unsupervised clustering analysis based on a sliding window is performed on the time series sequence of the first linkage data to obtain the clustering analysis results. Based on the boundaries of the clustering analysis results, the time series sequence of the first linkage data is divided into N stages; where each stage corresponds to a time window with relatively consistent process states.

[0025] It should be noted that sorting the initial linked data by timestamp to form a time-series sequence aims to reconstruct discrete related events into a continuous process trajectory, thereby supporting dynamic analysis of risk evolution as the experiment progresses. In practice, the first step is to detect predefined key operation transition point markers in the dynamic data of the experimental activity units. These markers are usually pre-set in standardized experimental procedures, such as explicit instructions or status flags like "start stirring" or "transfer to heat preservation." If such markers exist, the time-series sequence is directly segmented using them as boundaries to obtain logically clear stage divisions.

[0026] For experimental processes lacking a predefined structure, an unsupervised clustering method based on a sliding window is used to automatically identify stage boundaries. This method continuously calculates the distribution of data within a multidimensional feature space by sliding an analysis window along the time sequence, and automatically aggregates the sequence into several internally stable segments based on distribution similarity. The dividing points between the clustered segments are identified as naturally formed stage boundaries. Each stage must meet the requirement of relative consistency of process states, that is, the core mode of experimental operation and the response characteristics of key safety interfaces remain stable within that period. For example, in the "constant pressure distillation" stage, the heating power and the fraction collection rate maintain a relatively fixed matching relationship.

[0027] By combining rule-based segmentation with data-driven clustering, this method can flexibly adapt to the analytical needs of different types of activities in university laboratories, ranging from highly standardized teaching experiments to free-exploration research projects. This overcomes the limitation of traditional static risk assessment in failing to depict how risk evolves within the experiment at different operational stages, providing the necessary temporal framework for achieving refined process risk tracking.

[0028] S3-1: Obtain the degree of influence of co-matching at each stage based on N stages, including: Based on the type and intensity parameters of the experimental activity units in each stage, a predictable baseline sequence of continuous monitoring data for key safety interfaces is constructed. Obtain the actual coupling sequence of key safety interfaces within a phase; the actual coupling sequence refers to the sequence of continuous monitoring data of key safety interfaces extracted from the first linkage data that directly corresponds to the current experimental activity unit; Calculate the sum of squared residuals between the coupled actual sequence and the expected baseline sequence of continuous monitoring data; Calculate the ratio of the sum of squared residuals to the sum of squared deviations of the coupled actual sequence, and normalize the ratio to obtain the normalized residual ratio. Calculate the difference between 1 and the normalized residual ratio to obtain the degree of influence of the co-matching at each stage; The closer the degree of influence of the same position matching is to 1, the greater the expected contribution of the experimental activity unit to the state of the critical security interface within the stage.

[0029] It should be noted that the core objective of this step is to quantitatively evaluate the degree of conformity between experimental operations and the dynamic interactions between them and the critical safety interfaces they directly depend on. First, based on the type and intensity parameters of the experimental activity units in the current phase, a predicted baseline sequence of continuous monitoring data for the critical safety interfaces under this condition is retrieved and constructed from historical data models. This sequence represents the ideal response pattern summarized from past safe operations. Simultaneously, the coupled actual sequence recorded by the monitoring probe of the same interface within this phase is acquired concurrently. This sequence reflects the immediate and interpretable response of the experimental operations to the critical safety interfaces they directly depend on.

[0030] The residual sum of squares is obtained by calculating the sum of squares of the differences between corresponding data points in the coupled actual sequence and the expected baseline sequence. This value directly measures the overall deviation between the two sequences. To more accurately interpret the meaning of this deviation, it is necessary to further calculate the proportion of the residual sum of squares to the total variation of the coupled actual sequence. The residual sum of squares refers to the sum of squares of the differences between each data point in the coupled actual sequence and the mean of that sequence. This proportion reveals how much of the observed data fluctuations stems from deviations from the expected safe pattern, rather than normal systemic fluctuations. Subsequently, this proportion is normalized to a standard metric between 0 and 1, namely the normalized residual ratio. This ratio intuitively expresses the severity of the expected mismatch.

[0031] Finally, the degree of influence of the matching is obtained by calculating the difference between 1 and the ratio of the normalized residual. This design means subtracting the mismatch portion from the ideal state of perfect matching, thereby obtaining a positive measure of the degree of matching. The higher the value of the degree of influence of the matching, the greater the contribution of the experimental activity to the actual impact on the state of the critical safety interface in this stage, and the lower the possibility of risk caused by this direct coupling relationship; conversely, it indicates that there has been an unexpected anomaly in the experimental operation or facility response, and its contribution has decreased. The calculation process transforms the traditional qualitative judgment-based operational status assessment into a continuous and comparable quantitative indicator, thereby improving the objectivity and sensitivity of the risk assessment of direct interaction, and effectively making up for the deficiency of existing technologies in detecting gradual anomalies due to the lack of refined quantitative methods.

[0032] S3-2: Obtain the degree of mismatch impact at each stage based on N stages, including: A misaligned matching topology model is constructed using all critical security interfaces as nodes and the physical connection paths of the security facility system as edges. The misaligned matching topology model refers to a directed network graph that labels the transmission direction and characteristic time delay parameters. Based on the transmission direction and characteristic time delay parameters provided by the mismatched topology model, a time-delayed translation sequence of dynamic data for each upstream experimental activity unit is constructed. Obtain the actual transmission sequence of the target critical security interface at the current stage; if the target critical security interface has second linkage data at the current stage, then the continuous monitoring data sequence in the second linkage data is preferentially used as the actual transmission sequence; otherwise, the regular monitoring data is used as the actual transmission sequence; the actual transmission sequence is used to analyze the existing cross-domain risk transmission relationship; Calculate the peak value of the time-delay cross-correlation function between each time-delay shift sequence and the actual transmission sequence; The peak value of the time-delay cross-correlation function was subjected to statistical significance testing and normalization. The peak value of the normalized time-delay cross-correlation function that passes the significance test is defined as the degree of mismatch influence of the corresponding upstream experimental activity unit; The closer the degree of mismatched influence is to 1, the higher the weight of the non-directly coupled cross-domain transmission influence generated by the upstream experimental activity unit on the target's critical security interface through the facility network within the stage.

[0033] It should be noted that, to identify the indirect impacts of experimental activities through shared infrastructure, this step first constructs a mismatched topology model. This model abstracts the physical connection structure of the ventilation ducts, electrical buses, and drainage network support system as a directed graph, where nodes represent key safety interfaces, edges represent connection paths, and the direction of material or energy conduction and the characteristic time delay parameter required for its transmission are labeled. The characteristic time delay parameter refers to the typical time it takes for material or energy to be conducted from an upstream node to a downstream node along a specific path in the support system. This parameter is obtained through calculations using fluid dynamics / circuit theory, actual measurements under no-load or calibration conditions, or estimation based on historical normal operation data using system identification methods.

[0034] Based on this model, all upstream experimental activity units that affect the target's critical security interface via physical paths can be identified. For each upstream unit, its dynamic data sequence is shifted along the time axis by a corresponding delay based on the characteristic time delay parameters provided by the model, generating a time-delayed shift sequence. This operation simulates the time required for the impact of the unit's activity to propagate to the target interface.

[0035] The key to this step lies in the selection strategy for the actual transmission sequence of the target interface. Priority is given to identifying the second linkage data associated with this interface at the current stage. The second linkage data represents a special state: the continuous monitoring data of this interface has undergone anomaly changes that cannot be directly explained by any experimental activities in this laboratory. This "uncoupled anomaly" is a typical indication of potential cross-domain risk transmission. Therefore, if second linkage data exists, the continuous monitoring data sequence it contains will be directly used as the actual transmission sequence; if it does not exist, regular monitoring data will be used as the actual transmission sequence. This strategy ensures that mismatch analysis can accurately focus on those suspicious signals with unclear origins that most require tracing.

[0036] Subsequently, the peak value of the time-delay cross-correlation function between each time-delay translation sequence and the actual transmission sequence is calculated. The peak value of the time-delay cross-correlation function refers to the maximum absolute value of the cross-correlation function between the dynamic data sequence of the upstream unit and the actual transmission sequence of the target interface after considering the time-delay translation of the upstream unit, and the peak value is taken as the maximum absolute value of this function in the region near zero delay. This peak value measures the degree of temporal matching between the upstream activity mode and the downstream abnormal state after considering the physical propagation delay. A high peak value suggests a transmission correlation between the two. To ensure the statistical reliability of the conclusions, each peak value needs to be tested for significance and normalized. Finally, the normalized peak value that passes the test is defined as the degree of mismatch influence of the corresponding upstream experimental activity unit.

[0037] This indicator directly quantifies the influence weight or contribution of the activities of a specific upstream unit on the abnormal states observed at the target interface in this stage. This design achieves a closed loop of data within the method: the "uncoupled anomalies" marked by the second linkage data become clues driving risk tracing in this step; while mismatch analysis uses topological models and related algorithms to find the most likely upstream sources for these anomalies. This mechanism directly addresses the core challenge of "cross-domain transmission of uncoupled risks," which traditional methods cannot handle at all. Through data-driven tracing analysis, it transforms implicit, cross-domain systemic risks into clear and identifiable risk responsibility relationships.

[0038] S3-3: Compare the influence of identical matches and the influence of the largest mismatches in the same stage to determine the dominant factors, including: If the impact of same-position matching within the same stage is greater than the sum of the impact of maximum mismatched matching and the preset advantage threshold, then the stage risk is determined to be dominated by same-position matching factors. If the maximum impact of mismatch within the same stage is greater than the sum of the impact of same-position matching and the preset advantage threshold, then the stage risk is determined to be dominated by mismatch factors. If the absolute value of the difference between the degree of influence of same-position matching and the degree of influence of maximum mismatch within the same stage is less than or equal to the preset advantage threshold, then the stage risk is determined to be affected by both same-position matching factors and mismatch matching factors. When the risk of a stage is determined to be dominated by mismatch factors, the upstream experimental activity unit that provides the maximum degree of mismatch impact in the stage is identified, and the end time of the active stage of the upstream experimental activity unit is no later than the start time of the stage.

[0039] It should be noted that, to ensure the clarity and stability of the risk attribution conclusions, this step introduces a preset dominance threshold as a decision boundary. This threshold is typically calibrated based on historical safe operation data by statistically analyzing the distribution of the difference between the two types of influence, taking a specific quantile, or by determining a significance boundary through hypothesis testing. The decision logic follows the principle of dominance: if the influence of the current stage's matching exceeds the sum of the influence of the maximum mismatch and the threshold, it sufficiently demonstrates that the intensity of the direct interaction risk has significantly suppressed any potential indirect transmission effects, and therefore the risk should be attributed to the experiment itself. Conversely, if the influence of the maximum mismatch shows an overwhelming advantage, it means that the observed anomaly is more likely to originate from the cross-domain interaction of external units.

[0040] Calculating the absolute value of the difference between the two provides a direct measure of the degree of disparity in influence. A large value indicates a high confidence level in the judgment pointing to a single dominant factor; a small value suggests that the risk is caused by multiple factors working together, thus being judged as a joint influence. When the mismatch factor is identified as dominant, it is necessary to further anchor the specific source of responsibility. By tracing and confirming the upstream experimental activity unit that provides the greatest degree of mismatch influence, this method completes the key leap from identifying "external influence" to locating "specific source of influence." This mechanism fundamentally solves the dilemma of ambiguous attribution and unclear responsibility in traditional methods when facing intertwined multi-source risks, providing a direct decision-making basis for implementing precise risk intervention and process optimization.

[0041] S4-1: Calculate the periodic matching degree based on the aforementioned dominant factors, including: Based on historical continuous monitoring data, a time series periodic detection algorithm is used to extract the dominant period length and phase offset of the operating status of each critical safety interface, and generate a set of operating period parameters for the critical safety interfaces. Based on the historical dynamic data and corresponding metadata of the experimental activity units, a time series period detection algorithm is used to extract the dominant period length and phase offset of each type of experimental activity unit in terms of execution frequency and operation intensity, and generate a set of execution period parameters for the experimental activity units. Calculate the cycle matching degree for the current experimental activity unit and the current critical safety interface: Obtain the execution cycle parameters of the current experimental activity unit type and the runtime parameters of the current critical safety interface; Calculate the harmonic ratio between the execution cycle length of the current experimental activity unit type and the runtime length of the current critical security interface; Calculate the absolute value of the phase difference between the execution cycle phase of the current experimental activity unit type and the current running cycle phase of the critical safety interface at the current moment; The harmonic ratio and the absolute value of the phase difference are normalized and weighted and fused to output the period matching degree value. The period matching degree value range is set to [0, 1]. The closer the period matching degree value is to 1, the higher the degree of matching between the execution cycle of the experimental activity unit and the operation cycle of the key safety interface. The closer the period matching degree value is to 0, the lower the degree of matching.

[0042] It should be noted that the core of this step lies in assessing the coordination of the macro-rhythms between experimental activity planning and facility operation and maintenance from a time perspective, in order to identify potential long-term systemic risks. To this end, it is first necessary to extract the periodic patterns of key safety interfaces and various experimental activity units from historical data.

[0043] Specifically, by applying a time-series periodicity detection algorithm to analyze historical continuous monitoring data, the algorithm identifies patterns such as the daily load fluctuation cycle of the power system and the operational cycle of critical safety interfaces for cooling water unit maintenance intervals. For each critical safety interface, the algorithm outputs a periodic parameter pair, which consists of two components: the dominant cycle length and the phase offset. The cycle length describes the time interval of the cycle (e.g., 24 hours), and the phase offset describes the relative position of the current moment within that cycle (e.g., 0.25 indicates that the cycle is one-quarter complete). The periodic parameter pairs of all critical safety interfaces together constitute the operational cycle parameter set of the critical safety interfaces. This set is essentially a mapping table that associates each interface with its corresponding (cycle length, phase offset) parameter pair.

[0044] Similarly, by applying the aforementioned time-series periodicity detection algorithm to analyze the historical dynamic data and metadata of experimental activity units, we can extract patterns such as the weekly patterns of teaching experiments and the distribution of densely used periods for large-scale instrument reservations, revealing the execution cycle characteristics. For each type of experimental activity unit, the algorithm also outputs a pair of cycle parameters consisting of the dominant cycle length and phase offset. The cycle parameter pairs of all experimental activity unit types collectively constitute the execution cycle parameter set of the experimental activity unit, which is also used as a mapping table.

[0045] When evaluating a specific current experimental activity unit and critical safety interface, the corresponding periodic parameter pairs are retrieved from the two sets mentioned above. It's important to clarify the conceptual connection here: the retrieved data consists of two complete periodic parameter objects, and subsequent calculations will utilize the specific components within these two objects.

[0046] The calculation process unfolds in two dimensions, corresponding to the two components of the periodic parameter pair: Cycle Length Coordination Assessment (Macro-Range Rhythm Matching): Extract the execution cycle length component from the current experimental activity unit parameter pair and the operating cycle length component from the current critical safety interface parameter pair, calculating the coordination ratio between the two. This ratio reflects the fundamental rhythmic mathematical compatibility between the experimental demand rhythm and the facility supply rhythm. For example, it aims to ensure that the two cycles are simple integer multiples of each other, thereby avoiding long-term incoordination. For instance, a chemical experiment is conducted daily at fixed times (cycle 24 hours), and the building's power system it relies on also has a clear daily load cycle (24 hours). The cycle length ratio between the two is 1:1, which is a coordinated state. If the experiment is changed to be conducted every 36 hours, forming a non-integer multiple relationship with the 24-hour power cycle, it constitutes a fundamental rhythmic incoordination, and thus faces the risk of long-term power supply phase mismatch.

[0047] Phase Synchronization Assessment (Instantaneous Time Alignment): Extract the execution cycle phase component from the current experimental activity unit parameter pair and the operating cycle phase component from the current critical safety interface parameter pair, calculating the absolute value of the phase difference between the two at the current moment. This absolute value of the phase difference directly reflects the degree of synchronization between the real-time demand time of the experimental activity and the real-time supply status of the facility operation; the smaller the value, the higher the degree of synchronization. For example, in a precision instrument experiment requiring a stable cooling water supply, the active period is scheduled for 10:00 AM, when the cooling water system pressure is most stable and has minimal fluctuations throughout the day (assuming this is the "ideal phase" of system operation). If the experiment actually starts at 10:00 AM, the difference between its execution phase and the system's ideal phase is close to zero. This extremely small "phase difference" will serve as a high synchronization indicator when calculating the cycle matching degree, contributing the most to improving the final "cycle matching degree" value. Conversely, if the experiment is delayed until 1:00 PM (a "non-ideal phase" of system pressure fluctuations), a significant phase difference will occur, and this large "phase difference" will directly lower the calculated cycle matching degree. Therefore, the "phase difference" calculation in this step aims to objectively measure the degree of fit between the experiment and the facility at a micro-moment from the perspective of time alignment accuracy. The smaller the value, the more accurate the time alignment, and the better the rhythm matching basis provided for subsequent macro-risk assessment.

[0048] To integrate information from these two dimensions, the harmonic ratio and the absolute value of the phase difference are normalized to eliminate dimensional differences, and corresponding weights are assigned according to the management strategy for fusion calculation, ultimately generating a periodic matching degree value between 0 and 1.

[0049] Regarding weight determination: Weight coefficients are initially set based on expert experience, or determined through reverse optimization based on the correlation between periodic matching degree and known risk events in historical data. During system operation, the weight coefficients are periodically recalibrated based on newly generated risk assessment feedback data to achieve adaptive optimization of the fusion strategy.

[0050] The cycle matching degree, as a comprehensive indicator, directly represents the level of risk at the macro level. A high matching degree means that the demand rhythm of experimental activities and the supply capacity of the facility system are well coordinated on both long-term (cycle length coordination) and short-term (phase alignment) scales, and the system is operating in a coordinated state. A low matching degree, on the other hand, is a clear warning signal, indicating that the experimental arrangements are continuously in a period of weak facility support, or that there is a fundamental conflict between the two rhythms, thus constituting a long-term systemic risk requiring intervention. This assessment mechanism fills the gap in traditional risk management regarding the lack of quantitative assessment of time planning factors and chronic risks.

[0051] S4-2: Establish a comprehensive risk assessment matrix with the periodic matching degree value range as the vertical axis and the type of dominant risk factor for each stage as the horizontal axis, and determine the overall risk value for each stage based on the comprehensive risk assessment matrix, including: Based on the cycle matching degree value of the current experimental activity unit and the type of risk-dominant factors at the current stage, the comprehensive risk level is determined by mapping in the comprehensive risk judgment matrix; The mapping rules of the comprehensive risk assessment matrix include: pre-setting a basic risk level for each type of risk-dominant factor in each stage; when the cycle matching degree value is lower than the set threshold for cycle matching degree, the comprehensive risk level is increased based on the basic risk level; when the cycle matching degree value is higher than the set threshold for cycle matching degree, the comprehensive risk level remains at the basic risk level. The risk assessment parameters are dynamically adjusted based on the periodic matching degree value: when the periodic matching degree value is detected to be lower than the set threshold for M consecutive periods, the preset advantage threshold is reduced. Based on the different intervals in which the periodic matching degree value is located, differentiated weighting coefficients are assigned to the degree of influence of same-position matching and the degree of influence of misaligned matching, and the overall risk value of the stage is dynamically calculated after weighting.

[0052] It should be noted that the design of the comprehensive risk assessment matrix enables a synergistic evaluation of both the macroscopic rhythmic state and microscopic risk events in laboratory operations. The vertical axis of the matrix is ​​divided according to the numerical range of cycle matching degree, used to characterize the coordination level between experimental activities and facility support in terms of time planning; the horizontal axis is based on the type of dominant risk factor at each stage, reflecting the category of direct risks or externally transmitted risks arising from the current operation. By cross-locating the inputs from these two dimensions within the matrix, a comprehensive risk level can be mapped, which comprehensively considers both systemic chronic stress and event-specific acute threats.

[0053] The pre-defined mapping rules in the matrix follow a core logic: the stability of the macro environment significantly affects the severity of micro risks. Therefore, when the periodic matching degree value is lower than the set threshold representing the boundary between coordination and misalignment, it means that the experimental activity is in a vulnerable period of system security. In this macro-fragile context, even if the risk level indicated by the dominant factors at a given stage is not high, the actual threat is amplified, thus requiring an increase in the overall risk level from the original baseline risk level. Specifically, this increase can be based on the specific range or value of the periodic matching degree below the set threshold, setting clear rules for improvement, such as increasing by at least one level, or increasing by multiple levels in cases of severe misalignment, thereby quantifying the amplifying effect of macro-fragility on immediate risks. Conversely, if macro-coordination is good (the periodic matching degree value is higher than the set threshold), the baseline risk level judgment for micro events is maintained. The determination of the periodic matching degree threshold relies on the analysis of the time series of periodic matching degrees in historical operational data, for example, by identifying the long-term mean or median of the matching degree sequence, or by detecting abrupt transitions from a 'high matching' state to a 'low matching' state.

[0054] To respond to the dynamic evolution of risks, this step further introduces a parameter adaptation mechanism. The continuous stage counting parameter M is a configurable system parameter. Its value is set based on historical risk assessment feedback and system stability requirements. For example, by statistically analyzing all event sequences in historical data that meet the condition of 'periodic matching degree continuously below a set threshold,' the average or median number of stages experienced from the beginning to the first associated 'recordable risk event' is calculated, serving as a reference benchmark for parameter M. If the periodic matching degree value remains below the set threshold for M consecutive stages, it indicates that the systemic imbalance has formed a stable trend. As an adaptive defense strategy, parameter adjustment is implemented: automatically lowering the preset advantage threshold for determining the dominant factor. The direct technical effect of this adjustment is that it becomes more sensitive to the difference between the influence of identical matching and mismatched matching. Specifically, a lower threshold allows the system to clearly determine the dominant risk factor based on smaller differences, thereby enabling earlier capture and response to subtle risk signal changes when the macro environment remains fragile.

[0055] Simultaneously, based on the specific numerical range of the periodic matching degree, the weighting of the influence of identical matching and mismatched matching is dynamically adjusted. For example, in ranges with severe macro-coordination imbalances, a higher weighting coefficient is assigned to the influence of mismatched matching, which characterizes the risk transmission across laboratories, because the facility network has become a more sensitive channel for risk diffusion. Finally, the dynamically generated weighting coefficients are used for weighted fusion calculations to generate a quantitative value that comprehensively and dynamically reflects the overall risk situation at the current stage—the stage-wide overall risk value.

[0056] This entire mechanism breaks through the limitations of the static separation between macro background and micro events in traditional assessments, achieving dynamic linkage between the two. It can not only perceive the long-term trend of the health of the macro system, but also adaptively adjust the judgment criteria and sensitivity to micro risks accordingly.

[0057] Example 2: Based on Example 1, a laboratory risk adaptive assessment system based on AI algorithms, such as... Figure 2 As shown, it includes: The data fusion and linkage generation module is used to obtain the time difference between the key safety interface and the experimental activity unit, align the continuous monitoring data and dynamic data to obtain fused data, and generate the first linkage data or the second linkage data according to the coupling relationship. The time-series segmentation module is used to construct a time-series sequence of the first linkage data based on the first linkage data, and to divide the sequence into N stages according to the key operation transition point identifiers or cluster analysis results. The matching impact analysis and dominant factor determination module is used to calculate the degree of influence of same-position matching and misaligned matching at each stage, and to determine the dominant risk factor at each stage by comparing the two. The cycle matching and comprehensive risk assessment module is used to calculate the cycle matching degree, establish a comprehensive risk judgment matrix, and dynamically adjust parameters to determine the overall risk value of the stage.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An adaptive risk assessment method for laboratories based on AI algorithms, characterized in that, Includes the following steps: S1: Obtain fused data based on key security interfaces and experimental activity units, and determine the first linkage data and the second linkage data based on the fused data; S2: Construct a first linkage data time sequence based on the first linkage data and divide the first linkage data time sequence into N stages; S3: Obtain the impact of same-position matching in each stage based on N stages, and obtain the impact of misaligned matching in each stage based on N stages; By comparing the influence of identical matching and the influence of maximum mismatch within the same stage, the dominant factors can be identified. S4: Calculate the cycle matching degree based on the dominant factors, establish a comprehensive risk judgment matrix with the cycle matching degree value range as the vertical axis and the stage risk dominant factor type as the horizontal axis, and determine the stage overall risk value according to the comprehensive risk judgment matrix.

2. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The fused data obtained based on key security interfaces and experimental activity units includes: The time difference between the critical safety interface and the experimental activity unit is obtained. Based on the time difference, the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit are aligned by time offset and matched by spatial position to obtain fused data. The critical safety interface refers to the functional interface that ensures the exchange of matter, energy, or information between the facility system and the experimental operating space. The continuous monitoring data refers to the set of real-time time-series parameters that reflect the operating status of critical security interfaces; The dynamic data refers to the set of real-time process variables that characterize the progress and internal state changes of experimental activities.

3. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The step of determining the first linkage data and the second linkage data based on the fused data includes: Based on fused data, if there is a coupling relationship between the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit, the continuous monitoring data and the dynamic data will be combined into the first linkage data. Based on fused data, if there is no coupling relationship between the continuous monitoring data of the critical safety interface and the dynamic data of the experimental activity unit, the continuous monitoring data and the experimental background information are combined into the second linkage data; the experimental background information refers to the set of basic state parameters of the laboratory environment and facilities that are independent of the dynamic changes of a specific experimental activity unit.

4. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The step of constructing a first linkage data time series based on the first linkage data and dividing the first linkage data time series into N stages includes: Based on the chronological order of the timestamps of the data records in the first linkage data, a time sequence of the first linkage data is constructed; If the dynamic data of the experimental activity unit contains predefined key operation transition point identifiers, then the first linkage data time sequence is divided into N stages based on the key operation transition point identifiers; If the dynamic data of the experimental activity unit does not contain predefined key operation transition point identifiers, then an unsupervised clustering analysis based on a sliding window is performed on the time series sequence of the first linkage data to obtain the clustering analysis results. Based on the boundaries of the clustering analysis results, the time series sequence of the first linkage data is divided into N stages; where each stage corresponds to a time window with relatively consistent process states.

5. The adaptive laboratory risk assessment method based on AI algorithm as described in claim 1, characterized in that, The method of obtaining the degree of influence of co-matching at each stage based on N stages includes: Based on the type and intensity parameters of the experimental activity units in each stage, a predictable baseline sequence of continuous monitoring data for key safety interfaces is constructed. Obtain the actual coupling sequence of key safety interfaces within a phase; the actual coupling sequence refers to the sequence of continuous monitoring data of key safety interfaces extracted from the first linkage data that directly corresponds to the current experimental activity unit; Calculate the sum of squared residuals between the coupled actual sequence and the expected baseline sequence of continuous monitoring data; Calculate the ratio of the sum of squared residuals to the sum of squared deviations of the coupled actual sequence, and normalize the ratio to obtain the normalized residual ratio. Calculate the difference between 1 and the normalized residual ratio to obtain the degree of influence of the co-matching at each stage; The closer the degree of influence of the same position matching is to 1, the greater the expected contribution of the experimental activity unit to the state of the critical security interface within the stage.

6. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The method of obtaining the degree of mismatch impact at each stage based on N stages includes: A misaligned matching topology model is constructed using all critical security interfaces as nodes and the physical connection paths of the security facility system as edges. The misaligned matching topology model refers to a directed network graph that labels the transmission direction and characteristic time delay parameters. Based on the transmission direction and characteristic time delay parameters provided by the mismatched topology model, a time-delayed translation sequence of dynamic data for each upstream experimental activity unit is constructed. Obtain the actual transmission sequence of the target critical security interface at the current stage; if the target critical security interface has second linkage data at the current stage, then the continuous monitoring data sequence in the second linkage data is preferentially used as the actual transmission sequence; otherwise, the regular monitoring data is used as the actual transmission sequence; the actual transmission sequence is used to analyze the existing cross-domain risk transmission relationship; Calculate the peak value of the time-delay cross-correlation function between each time-delay shift sequence and the actual transmission sequence; The peak value of the time-delay cross-correlation function was subjected to statistical significance testing and normalization. The peak value of the normalized time-delay cross-correlation function that passes the significance test is defined as the degree of mismatch influence of the corresponding upstream experimental activity unit; The closer the degree of mismatched influence is to 1, the higher the weight of the non-directly coupled cross-domain transmission influence generated by the upstream experimental activity unit on the target's critical security interface through the facility network within the stage.

7. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The comparison of the influence of identical matching and the influence of maximum mismatch in the same stage to determine the dominant factors includes: If the impact of same-position matching within the same stage is greater than the sum of the impact of maximum mismatched matching and the preset advantage threshold, then the stage risk is determined to be dominated by same-position matching factors. If the maximum impact of mismatch within the same stage is greater than the sum of the impact of same-position matching and the preset advantage threshold, then the stage risk is determined to be dominated by mismatch factors. If the absolute value of the difference between the degree of influence of same-position matching and the degree of influence of maximum mismatch within the same stage is less than or equal to the preset advantage threshold, then the stage risk is determined to be affected by both same-position matching factors and mismatch matching factors. When the risk of a stage is determined to be dominated by mismatch factors, the upstream experimental activity unit that provides the maximum degree of mismatch impact in the stage is identified, and the end time of the active stage of the upstream experimental activity unit is no later than the start time of the stage.

8. The adaptive laboratory risk assessment method based on AI algorithm as described in claim 1, characterized in that, The calculation of the periodic matching degree based on the dominant factors includes: Based on historical continuous monitoring data, a time series periodic detection algorithm is used to extract the dominant period length and phase offset of the operating status of each critical safety interface, and generate a set of operating period parameters for the critical safety interfaces. Based on the historical dynamic data and corresponding metadata of the experimental activity units, a time series period detection algorithm is used to extract the dominant period length and phase offset of each type of experimental activity unit in terms of execution frequency and operation intensity, and generate a set of execution period parameters for the experimental activity units. Calculate the cycle matching degree for the current experimental activity unit and the current critical safety interface: Obtain the execution cycle parameters of the current experimental activity unit type and the runtime parameters of the current critical safety interface; Calculate the harmonic ratio between the execution cycle length of the current experimental activity unit type and the runtime length of the current critical security interface; Calculate the absolute value of the phase difference between the execution cycle phase of the current experimental activity unit type and the current running cycle phase of the critical safety interface at the current moment; The harmonic ratio and the absolute value of the phase difference are normalized and weighted and fused to output the period matching degree value. The period matching degree value range is set to [0, 1]. The closer the period matching degree value is to 1, the higher the degree of matching between the execution cycle of the experimental activity unit and the operation cycle of the key safety interface. The closer the period matching degree value is to 0, the lower the degree of matching.

9. The AI-based adaptive risk assessment method for laboratories as described in claim 1, characterized in that, The establishment of a comprehensive risk assessment matrix with the periodic matching degree value range as the vertical axis and the type of dominant risk factor in a given stage as the horizontal axis, and the determination of the overall risk value for a stage based on the comprehensive risk assessment matrix, includes: Based on the cycle matching degree value of the current experimental activity unit and the type of risk-dominant factors at the current stage, the comprehensive risk level is determined by mapping in the comprehensive risk judgment matrix; The mapping rules of the comprehensive risk assessment matrix include: pre-setting a basic risk level for each type of risk-dominant factor in each stage; when the cycle matching degree value is lower than the set threshold for cycle matching degree, the comprehensive risk level is increased based on the basic risk level; when the cycle matching degree value is higher than the set threshold for cycle matching degree, the comprehensive risk level remains at the basic risk level. The risk assessment parameters are dynamically adjusted based on the periodic matching degree value: when the periodic matching degree value is detected to be lower than the set threshold for M consecutive periods, the preset advantage threshold is reduced. Based on the different intervals in which the periodic matching degree value is located, differentiated weighting coefficients are assigned to the degree of influence of same-position matching and the degree of influence of misaligned matching, and the overall risk value of the stage is dynamically calculated after weighting.

10. A laboratory risk adaptive assessment system based on an AI algorithm, used to implement the laboratory risk adaptive assessment method based on an AI algorithm as described in any one of claims 1-9, characterized in that, include: The data fusion and linkage generation module is used to obtain the time difference between the key safety interface and the experimental activity unit, align the continuous monitoring data and dynamic data to obtain fused data, and generate the first linkage data or the second linkage data according to the coupling relationship. The time-series segmentation module is used to construct a time-series sequence of the first linkage data based on the first linkage data, and to divide the sequence into N stages according to the key operation transition point identifiers or cluster analysis results. The matching impact analysis and dominant factor determination module is used to calculate the degree of influence of same-position matching and misaligned matching at each stage, and to determine the dominant risk factor at each stage by comparing the two. The cycle matching and comprehensive risk assessment module is used to calculate the cycle matching degree, establish a comprehensive risk judgment matrix, and dynamically adjust parameters to determine the overall risk value of the stage.

Images