CN-122020252-A - Laboratory risk self-adaptive assessment method and system based on AI algorithm

CN122020252ACN 122020252 ACN122020252 ACN 122020252ACN-122020252-A

Abstract

The invention relates to the technical field of laboratory risk management and predictive analysis, in particular to a laboratory risk self-adaptive assessment method and system based on an AI algorithm. The method comprises the following steps of S1, obtaining fusion data based on a key safety interface and an experimental activity unit, 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 parity matching influence degree of each stage based on the N stages, obtaining the dislocation matching influence degree of each stage based on the N stages, comparing the parity matching influence degree in the same stage with the maximum dislocation matching influence degree, and determining a dominant factor. The invention fuses space-time data, realizes accurate tracing of direct and indirect risks, dynamically evaluates hidden system risks in complex experimental environments of colleges and universities, and improves early recognition and self-adaptive early warning capability.

Inventors

ZHANG DAWEI
HUANG FENGYUAN
WEN BOYU
GUAN HONGQIANG
YU KAIMING
HAN LIN
LIU XIN
XU XIAOZHOU
GAO FANG
TENG HONGJUN

Assignees

辽东学院
沈阳久成科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260202

Claims (10)

1. The laboratory risk self-adaptive assessment method based on the AI algorithm is characterized by comprising the following steps of: s1, acquiring fusion data based on a key safety 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 parity matching influence degree of each stage based on N stages, and obtaining the dislocation matching influence degree of each stage based on N stages; And S4, calculating the period matching degree based on the dominant factors, establishing a comprehensive risk judgment matrix taking a period matching degree numerical interval as a vertical axis and taking a stage risk dominant factor type as a horizontal axis, and determining a stage overall risk value according to the comprehensive risk judgment matrix.
2. The method for adaptive assessment of laboratory risk based on AI algorithm of claim 1, wherein the obtaining of fusion data based on the critical security interface and the experimental activity unit comprises: Acquiring a time difference between a key safety interface and an experimental activity unit, and performing time offset alignment and spatial position matching on continuity monitoring data of the key safety interface and dynamic data of the experimental activity unit according to the time difference to acquire fusion data; The key safety interface is a functional interface for ensuring the exchange of substances, energy or information between the facility system and the experimental operation space; the continuity monitoring data refers to a real-time sequence parameter set reflecting the operation state of the key safety interface; The dynamic data refers to a set of real-time process variables that characterize the course of experimental activity and internal state changes.
3. The AI-algorithm-based laboratory risk adaptive assessment method of claim 1, wherein the determining the first linkage data and the second linkage data based on the fusion data comprises: based on the fusion data, if the continuity monitoring data of the key safety interface and the dynamic data of the experimental activity unit have a coupling relation, combining the continuity monitoring data and the dynamic data into first linkage data; Based on the fusion data, if the continuity monitoring data of the key safety interface and the dynamic data of the experimental activity unit have no coupling relation, the continuity monitoring data and experimental background information are combined into second linkage data, wherein the experimental background information refers to a laboratory environment and facility basic state parameter set which are independent of the dynamic change of the specific experimental activity unit.
4. The AI algorithm-based laboratory risk adaptive assessment method of claim 1, wherein constructing a first linkage data timing sequence based on the first linkage data and dividing the first linkage data timing sequence into N stages comprises: Constructing a first linkage data time sequence based on the time stamp sequence of the data record in the first linkage data; If the dynamic data of the experimental activity unit contains a predefined key operation conversion point identifier, dividing the first linkage data time sequence into N stages according to the key operation conversion point identifier; If the dynamic data of the experimental activity unit does not contain the predefined key operation conversion point identification, performing unsupervised cluster analysis based on a sliding window on the first linkage data time sequence to obtain a cluster analysis result, and dividing the first linkage data time sequence into N stages according to the boundary of the cluster analysis result, wherein each stage corresponds to a time window with relative consistency of process states.
5. The AI-algorithm-based laboratory risk adaptive assessment method of claim 1, wherein the obtaining the parity matching influence degree for each stage based on the N stages comprises: according to the type parameter and the intensity parameter of the experimental activity unit in each stage, constructing an expected reference sequence of the continuity monitoring data of the key safety interface; Acquiring a coupling actual sequence of a key safety interface in a stage, wherein the coupling actual sequence is a key safety interface continuity monitoring data sequence which is extracted from first linkage data and directly corresponds to a current experimental activity unit; calculating a residual square sum of the coupling actual sequence of the continuity monitoring data and the expected reference sequence; Calculating the ratio of the sum of squares of the residual errors to the sum of squares of the deviations of the coupling actual sequences, and normalizing the comparison value to obtain a normalized residual error ratio; calculating the difference value of the ratio of the 1 to the normalized residual error to obtain the parity matching influence degree of each stage; the closer the parity matching influence degree is to 1, the greater the expected influence contribution degree of the experimental activity unit in the stage to the key safety interface state is indicated.
6. The AI algorithm-based laboratory risk adaptive assessment method of claim 1, wherein the obtaining the degree of influence of each stage mismatch based on the N stages comprises: Constructing a dislocation matching topology model by taking all key safety interfaces as nodes and taking an entity connection path of a guarantee facility system as an edge, wherein the dislocation matching topology model is a directed network diagram of index note conduction directions and characteristic time delay parameters; Constructing a time-lag translation sequence of dynamic data of each upstream experimental activity unit according to the conduction direction and the characteristic time delay parameter provided by the dislocation matching topology model; the method comprises the steps of acquiring a conduction actual sequence of a target key safety interface in a current stage, preferentially adopting a continuity monitoring data sequence in second linkage data as the conduction actual sequence if the target key safety interface has the second linkage data in the current stage, otherwise adopting conventional monitoring data as the conduction actual sequence, wherein the conduction actual sequence is used for analyzing the existing cross-domain risk conduction relation; Calculating the time lag cross correlation function peak value of each time lag translation sequence and the conduction actual sequence; Carrying out statistical significance test and normalization treatment on the time-lag cross-correlation function peak value; defining a peak value of a normalized time-lapse cross-correlation function passing through the significance test as a dislocation matching influence degree corresponding to an upstream experimental activity unit; The closer the dislocation matching influence degree is to 1, the higher the cross-domain conduction influence weight of the upstream experimental activity unit in the stage, which is generated by the facility network on the target key safety interface and is indirectly coupled, is shown.
7. The AI algorithm-based laboratory risk adaptive assessment method of claim 1, wherein comparing the parity matching impact level with the maximum misalignment matching impact level in the same stage, determining the dominant factor, comprises: if the influence degree of the co-located matching in the same stage is larger than the sum of the maximum dislocation matching influence degree and a preset dominance threshold, judging that the stage risk is dominated by the co-located matching factor; If the maximum dislocation matching influence degree in the same stage is greater than the sum of the dislocation matching influence degree and a preset dominance threshold, judging that the stage risk is dominated by dislocation matching factors; If the absolute value of the difference between the parity matching influence degree and the maximum dislocation matching influence degree in the same stage is smaller than or equal to a preset dominant threshold, judging that the stage risk is influenced by the parity matching factor and the dislocation matching factor together; And when judging that the stage risk is dominated by the dislocation matching factor, confirming an upstream experimental activity unit for providing the stage maximum dislocation matching influence degree, wherein the ending time of the active stage of the upstream experimental activity unit is not later than the stage starting time.
8. The AI-algorithm-based laboratory risk adaptive assessment method of claim 1, wherein the calculating the period matching degree based on the dominant factor comprises: Based on the historical continuity monitoring data, extracting the dominant period length and the phase offset of the operation state of each key safety interface by adopting a time sequence period detection algorithm, and generating an operation period parameter set of the key safety interface; based on historical dynamic data and corresponding metadata of the experimental activity units, extracting dominant period length and phase offset of each type of experimental activity units on the execution frequency and the operation intensity by adopting a time sequence period detection algorithm, and generating an execution period parameter set of the experimental activity units; aiming at the current experimental activity unit and the current key safety interface, calculating the period matching degree: Acquiring an execution period parameter of the type of the current experimental activity unit and an operation period parameter of the current key safety interface; calculating the harmony ratio between the execution period length of the type of the current experimental activity unit and the running period length of the current key safety interface; calculating the absolute value of a phase difference value between the phase of an execution period of the type to which the current experimental activity unit belongs at the current moment and the phase of an operation period of the current key safety interface; And carrying out normalization processing and weighted fusion calculation on the absolute value of the harmonic ratio and the phase difference value to output a period matching degree value, wherein the period matching degree value range is set to be 0 and 1, the closer the period matching degree value is to 1, the higher the matching degree between the execution period of the experimental activity unit and the operation period of the key safety interface is, and the closer the period matching degree value is to 0, the lower the matching degree is.
9. The method for adaptively evaluating laboratory risk based on AI algorithm according to claim 1, wherein said establishing a comprehensive risk decision matrix with a period matching degree value interval as a vertical axis and a stage risk leading factor type as a horizontal axis, and determining a stage total risk value according to the comprehensive risk decision matrix, comprises: according to the cycle matching degree value of the current experimental activity unit and the type of the risk leading factor of the current stage, mapping and determining the comprehensive risk level in a comprehensive risk judgment matrix; The mapping rule of the comprehensive risk judgment matrix comprises the steps of presetting a basic risk level for each stage of risk leading factor type, lifting the comprehensive risk level on the basic risk level when the period matching degree value is lower than a period matching degree set threshold value, and maintaining the basic risk level when the period matching degree value is higher than the period matching degree set threshold value; dynamically adjusting risk assessment parameters according to the period matching degree value, wherein when M continuous periods of the period matching degree value are monitored to be lower than a set threshold value, a preset dominant threshold value is reduced; and according to different intervals where the period matching degree value is located, different weight coefficients are distributed for the parity matching influence degree and the dislocation matching influence degree, and the weighted stage overall risk value is dynamically calculated.
10. An AI algorithm-based laboratory risk adaptive assessment system for implementing the AI algorithm-based laboratory risk adaptive assessment method as claimed in any one of claims 1 to 9, comprising: The data fusion and linkage generation module is used for acquiring the time difference between the key safety interface and the experimental activity unit, aligning the continuity monitoring data with the dynamic data to acquire fusion data, and generating first linkage data or second linkage data according to the coupling relation; The time sequence stage dividing module is used for constructing a first linkage data time sequence based on the first linkage data and dividing the sequence into N stages according to key operation conversion point identification or cluster analysis results; The matching influence analysis and leading factor judgment module is used for calculating the parity matching influence degree and the dislocation matching influence degree of each stage and determining a stage risk leading factor by comparing the parity matching influence degree and the dislocation matching influence degree; The period matching and comprehensive risk assessment module is used for calculating period matching degree, establishing a comprehensive risk judgment matrix and dynamically adjusting parameters to determine a stage overall risk value.

Description

Laboratory risk self-adaptive assessment method and system based on AI algorithm Technical Field The invention relates to the technical field of laboratory risk management and predictive analysis, in particular to a laboratory risk self-adaptive assessment method and system based on an AI algorithm. Background Artificial intelligence algorithms have been gradually applied to the field of dynamic perception of laboratory security risk through learning of historical data and modeling of real-time information. The laboratory risk self-adaptive assessment refers to that the system can automatically adjust risk assessment model parameters and early warning thresholds according to the changes of experimental progress and facility states. The existing AI-based assessment method is generally focused on the inside of a single laboratory unit, and by establishing a prediction model for monitoring data of specific equipment or operation flow, certain advantages are presented in the aspect of identifying abnormal fluctuation of a known mode, and an effective path is provided for improving the automation level of laboratory safety management. However, existing approaches have significant limitations in dealing with complex environments of university laboratories shared by multidisciplinary intersections and infrastructure. The existing scheme depends on isolated monitoring point data and static risk assessment rules, so that dynamic and time-lag interaction relations between different experimental activity units and shared security facility systems are difficult to effectively describe, and cross-domain risk conduction caused by indirect coupling cannot be identified. This results in the traditional approach lacking in sensing and early warning capabilities for systematic, implicit risks arising from facility rhythms not matching experimental periods, or from indirect effects across the laboratory. Therefore, it is needed to construct a risk assessment method capable of adapting to multi-source heterogeneous and space-time coupling characteristics of a college laboratory, and achieve traceability and prospective adaptive assessment of complex risk paths through deep association and collaborative analysis of key safety interface data and experimental activity full-chain data. Disclosure of Invention In order to overcome the defect of insufficient multi-source dynamic risk coupling evaluation, the invention provides a laboratory risk self-adaptive evaluation method and system based on an AI algorithm. The technical embodiment of the invention is a laboratory risk self-adaptive assessment method based on an AI algorithm, which comprises the following steps: s1, acquiring fusion data based on a key safety 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 parity matching influence degree of each stage based on N stages, and obtaining the dislocation matching influence degree of each stage based on N stages; And S4, calculating the period matching degree based on the dominant factors, establishing a comprehensive risk judgment matrix taking a period matching degree numerical interval as a vertical axis and taking a stage risk dominant factor type as a horizontal axis, and determining a stage overall risk value according to the comprehensive risk judgment matrix. Preferably, the obtaining the fusion data based on the key security interface and the experimental activity unit includes: Acquiring a time difference between a key safety interface and an experimental activity unit, and performing time offset alignment and spatial position matching on continuity monitoring data of the key safety interface and dynamic data of the experimental activity unit according to the time difference to acquire fusion data; The key safety interface is a functional interface for ensuring the exchange of substances, energy or information between the facility system and the experimental operation space; the continuity monitoring data refers to a real-time sequence parameter set reflecting the operation state of the key safety interface; The dynamic data refers to a set of real-time process variables that characterize the course of experimental activity and internal state changes. Preferably, the determining the first linkage data and the second linkage data based on the fusion data includes: based on the fusion data, if the continuity monitoring data of the key safety interface and the dynamic data of the experimental activity unit have a coupling relation, combining the continuity monitoring data and the dynamic data into first linkage data; Based on the fusion data, if the continuity monitoring data of the key safety interface and the dynamic data of the experimental activity unit have no coupling relatio