CN-121745704-B - Laboratory security risk intelligent early warning method, system, medium and product

CN121745704BCN 121745704 BCN121745704 BCN 121745704BCN-121745704-B

Abstract

A laboratory security risk intelligent early warning method, system, medium and product relate to the field of risk early warning. The method comprises the steps of calibrating a laboratory digital twin model based on real-time monitoring data, calculating and generating initial probability distribution based on the laboratory digital twin model and fusing a preset equipment fault probability model and a personnel behavior uncertainty model, deducing probability density distribution change of the initial probability distribution at a plurality of time points in the future through a numerical algorithm to obtain corresponding state prediction probability distribution, obtaining probability values of developing to dangerous states at each time point in the future based on the state prediction probability distribution of each time point in the future, and solving an optimal control operator reversely through an optimization algorithm when the estimated probability value of any dangerous state exceeds a preset threshold value, and executing intervention actions corresponding to the optimal control operator. By implementing the technical scheme, the laboratory safety risk can be predicted in advance.

Inventors

TAN XINXING
HUANG QINGSONG
LIU YINGCHUN
LIN GUANCHUN
XU YANG

Assignees

中理检验有限公司

Dates

Publication Date: 20260512
Application Date: 20260302

Claims (8)

1. The intelligent early warning method for the laboratory security risk is characterized by comprising the following steps of: Acquiring real-time monitoring data of a multi-source heterogeneous sensor deployed in a laboratory, calibrating a preset laboratory digital twin model based on the real-time monitoring data, and generating an initial probability distribution representing the current uncertainty of a system by means of calculation based on the state information of the calibrated laboratory digital twin model and fusing a preset equipment fault probability model and a personnel behavior uncertainty model; deducing probability density distribution changes of the initial probability distribution at a plurality of time points in the future through a numerical algorithm to obtain corresponding state prediction probability distribution; Respectively carrying out projection calculation on the state prediction probability distribution of each future time point to a plurality of predefined dangerous state modes to obtain probability values of developing to dangerous states at each future time point and outputting the probability values; when the estimated probability value of any dangerous state exceeds a preset threshold value, the optimal control operator is reversely solved through an optimization algorithm, the intervention action corresponding to the optimal control operator is executed, The method specifically comprises the steps of calculating and generating initial probability distribution representing current uncertainty of a system based on the state information of the calibrated laboratory digital twin model, and fusing a preset equipment fault probability model and a personnel behavior uncertainty model, wherein the initial probability distribution specifically comprises the following steps: Extracting and mapping the current state parameters of all entities in the calibrated laboratory digital twin model into a system state space to form a state vector, and generating a multi-element Gaussian probability distribution representing a state confidence range based on the state vector to serve as a base point of initial probability distribution; Invoking the equipment fault probability model matched with the equipment model and the current working condition of the target equipment in the laboratory digital twin model, inputting the current operation time length and the load rate of the target equipment into the equipment fault probability model, and receiving a conditional fault probability density function output by the equipment fault probability model, wherein the conditional fault probability density function defines the continuous probability that the current equipment state drifts in a health state neighborhood, and the target equipment is any one of the laboratory digital twin model; Invoking the personnel behavior uncertainty model matched with the role of a target personnel and the current operation task in the laboratory digital twin model, inputting the current task complexity and the environment interference degree to the personnel behavior uncertainty model, and receiving a discrete probability distribution list output by the personnel behavior uncertainty model, wherein the discrete probability distribution list defines possible discrete branches of the behavior state of the target personnel, and the target personnel is any one personnel in the laboratory digital twin model; Based on the physical and logical association relationship between entities defined in the laboratory digital twin model, carrying out joint distribution calculation on the base points, the conditional fault probability density functions of all devices and the discrete probability distribution list of all personnel in a system state space to obtain the initial probability distribution, Based on the physical and logical association relationship between entities defined in the laboratory digital twin model, the joint distribution calculation is performed on the base point, the conditional fault probability density functions of all devices and the discrete probability distribution list of all personnel in a system state space to obtain the initial probability distribution, which specifically comprises: Modeling the base points, the conditional fault probability density functions of all devices and the discrete probability distribution lists of all personnel as independent factor nodes in a probability tensor network respectively; Defining a correlation edge between each factor node of the probability tensor network according to the physical connection, the personnel-equipment operation relationship and the environment coupling relationship between the equipment defined in the laboratory digital twin model so as to construct a network topology structure; and according to the network topology structure, performing tensor contraction operation on the factor nodes which are connected with each other along the shared associated edge, eliminating all intermediate variables through summation to obtain a joint probability distribution tensor, and performing normalization processing on the joint probability distribution tensor to obtain the initial probability distribution.
2. The method according to claim 1, wherein the deriving, by a numerical algorithm, a probability density distribution change of the initial probability distribution at a plurality of future time points to obtain a corresponding state prediction probability distribution, specifically comprises: Discretizing state variables within the laboratory into a set of state basis vectors in a system state space, the state variables including environmental parameters, equipment states, personnel positions and actions; Constructing potential energy operators for representing internal constraints of a system based on standard operation rules, physical laws and equipment safety thresholds of the laboratory, constructing kinetic energy operators for representing state transition trend of the system based on physical connection and linkage logic among the laboratory equipment, and linearly combining the potential energy operators and the kinetic energy operators to form an equivalent Hamiltonian operator; taking the initial probability distribution as an initial state vector, taking the equivalent Hamiltonian as dynamic driving, and solving a time evolution equation by a numerical integration method to obtain state vectors at a plurality of time points in the future; And calculating the modular square of the state vector at a plurality of time points in the future to obtain the corresponding state prediction probability distribution.
3. The method according to claim 1, wherein said projecting the state prediction probability distribution at each future point in time to a predefined plurality of dangerous state modes, respectively, to obtain probability values for the development of each dangerous state at each future point in time, and outputting the probability values, specifically comprises: performing association backtracking analysis on the laboratory historical monitoring data flow and the recorded security events at regular intervals to acquire a system state abnormal evolution mode which is not predefined; Converting the abnormal evolution mode of the system state into a new dangerous eigenstate subspace in a state space, and writing the new dangerous eigenstate subspace into a dynamically updated dangerous state mode library; Carrying out projection calculation on all dangerous state modes currently stored in the dangerous state mode library in parallel by carrying out state prediction probability distribution of each future time point in real time when carrying out risk assessment each time; and outputting the result of projection calculation as a probability value set of each type of dangerous state developed into the mode library by the system at each future time point to form a full risk probability spectrum.
4. A method according to claim 3, wherein said deriving and outputting probability values for developing to dangerous conditions at future points in time comprises: Binding the probability value of each future time point with the corresponding spatial position attribute and the corresponding influence range attribute of the corresponding dangerous state in the laboratory digital twin model to generate a risk data object with spatial coordinates and risk intensity; Dynamically calculating corresponding visual characterization parameters for each risk data object according to the probability value and the evolution stage of the dangerous state, wherein the visual characterization parameters comprise colors, transparency, basic geometric dimensions and dynamic deformation parameters for expressing risk diffusion or accumulation trend; generating corresponding three-dimensional graphic elements based on the visual characterization parameters, and overlapping all the generated three-dimensional graphic elements as independent rendering layers onto a basic three-dimensional scene of the digital twin model of the laboratory in real time to form a three-dimensional dynamic risk visualization picture.
5. The method of claim 1, wherein said solving the optimal control operators back through an optimization algorithm comprises: Setting a sum of projection integral values of the state prediction probability distribution on the dangerous state mode to be minimum within a preset future time range as a convergence objective function of the optimization algorithm; Constructing a control space by using the full controllable parameters of the laboratory digital twin model, and performing global iterative search in the control space by using a gradient-based optimization algorithm to find a control parameter combination capable of reducing the value of the convergence objective function; And identifying the control parameter adjustment quantity with the highest contribution degree to reducing the value of the convergence objective function from the global search result, and determining the mathematical expression corresponding to the control parameter adjustment quantity as the optimal control operator.
6. A computer system comprising a memory, a processor and a computer program stored on the memory, characterized in that the processor executes the computer program to carry out the steps of the method according to any one of claims 1-5.
7. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any of claims 1-5.
8. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1-5.

Description

Laboratory security risk intelligent early warning method, system, medium and product Technical Field The application relates to the field of risk early warning, in particular to an intelligent early warning method, system, medium and product for laboratory safety risk. Background In the prior art, in order to ensure the safety of a laboratory, a method is generally adopted, which comprises the steps of arranging a special person to periodically patrol the laboratory, checking whether the running condition, the environmental parameters and the like of equipment are normal, installing sensors such as a smoke sensor, a temperature sensor and the like, and sending an alarm when abnormal conditions are detected. The method can monitor and manage the safety condition of the laboratory to a certain extent, and plays a certain role in preventing laboratory safety accidents. However, conventional laboratory security management methods have significant drawbacks. The special person can not check the existing time interval, so that real-time monitoring is difficult, sudden potential safety hazards can be missed, the traditional sensor can only monitor single environmental parameters and cannot comprehensively and comprehensively evaluate the safety risk of a laboratory, moreover, the methods cannot predict the safety risk possibly occurring in the future, cannot take effective intervention measures in advance, and cannot meet the increasingly complex safety management requirements of modern laboratories. Disclosure of Invention The application provides an intelligent early warning method, system, medium and product for laboratory security risk, which can predict laboratory security risk in advance, display the risk in a three-dimensional dynamic visualization manner, and timely take intervention measures to reduce the occurrence probability of the risk. In a first aspect, the application provides an intelligent early warning method for laboratory security risk, which comprises the following steps: Acquiring real-time monitoring data of a multi-source heterogeneous sensor deployed in a laboratory, calibrating a preset laboratory digital twin model based on the real-time monitoring data, and generating an initial probability distribution representing the current uncertainty of a system by means of calculation based on the state information of the calibrated laboratory digital twin model and fusing a preset equipment fault probability model and a personnel behavior uncertainty model; deducing probability density distribution changes of the initial probability distribution at a plurality of time points in the future through a numerical algorithm to obtain corresponding state prediction probability distribution; Respectively carrying out projection calculation on the state prediction probability distribution of each future time point to a plurality of predefined dangerous state modes to obtain probability values of developing to dangerous states at each future time point and outputting the probability values; and when the estimated probability value of any dangerous state exceeds a preset threshold value, solving the optimal control operator reversely through an optimization algorithm, and executing the intervention action corresponding to the optimal control operator. By adopting the technical scheme, the laboratory environment and equipment data are acquired in real time through the multi-source heterogeneous sensor, and the high-precision depiction and uncertainty quantification of the current system state are realized by combining the fusion analysis of the digital twin model and the probability model, so that the accuracy and the comprehensiveness of safety monitoring are improved. The probability density deduction is carried out on the future state of the system by adopting a numerical algorithm, so that the development trend of various dangerous situations can be predicted in advance, the transition from passive response to active early warning is realized, and the safety handling time window is obviously prolonged. The occurrence probability of various risks is quantitatively evaluated by projecting and calculating the prediction state to the predefined dangerous mode, and the automatic risk classification is realized by combining with the preset threshold value, so that a scientific and visual basis is provided for safety decision. When the risk probability exceeds a threshold value, the system can automatically solve the optimal control strategy through an optimization algorithm and execute corresponding intervention actions, so that closed-loop management from risk early warning to active treatment is realized, accident development is effectively restrained, and the autonomous protection level of laboratory safety is improved. In some embodiments, the calculating and generating an initial probability distribution representing the current uncertainty of the system based on the state information of the calibrated