Search

CN-121789384-B - Laboratory line safety early warning method and device

CN121789384BCN 121789384 BCN121789384 BCN 121789384BCN-121789384-B

Abstract

The application discloses a laboratory line-of-sight safety early warning method and device, which are used for forming line-of-sight basic data with a time stamp and an identity tag by collecting related personnel positioning and behavior monitoring data streams in real time, realizing personnel time-space track reconstruction and abnormal line segment accurate identification by combining a standard template, and providing reliable support for risk assessment. By combining the basic risk coefficients corresponding to the personnel features and the identities, the personal risk contribution value is calculated, and a personal risk field is generated, so that the accurate quantification and differential assessment of the personnel risks are realized. And constructing a risk coupling network can pre-judge the conduction effect of a high risk region, dynamically determine an early warning threshold value by combining regional personnel and activity types, effectively improve early warning accuracy and reduce false alarm and missing report. In summary, the application realizes the precise whole-flow management and control of laboratory line safety and the advanced risk treatment, and remarkably improves the level of refinement and intellectualization of safety management.

Inventors

  • XIONG WEI
  • XU WEI
  • YANG PAN

Assignees

  • 香港科技大学(广州)

Dates

Publication Date
20260508
Application Date
20260224

Claims (10)

  1. 1. A laboratory line safety precaution method, comprising: collecting positioning data streams and behavior monitoring data streams of all personnel in a laboratory area in real time, analyzing and correlating to form line base data with time stamps and personnel identity labels; Based on the continuous space-time track of the line-moving basic data reconstruction personnel, the abnormal line-moving segments are identified by combining the comparison of corresponding standard safety line-moving templates, and space-time characteristics and violation characteristics at least comprising a moving speed sequence, a regional stay time length and a line-moving node sequence are extracted; According to the space-time characteristics and the violation characteristics of each person and the basic risk coefficients corresponding to the person identity labels, calculating personal risk contribution values of the person at each point in space, and generating a personal risk field, wherein the personal risk field comprises the following steps: Fusing the space-time features, the violation features and the basic risk coefficients to construct comprehensive risk features of personnel; according to the comprehensive risk characteristics, calculating to obtain personal risk contribution values representing the risk levels of behaviors at the current moment and the spatial position through a risk evaluation algorithm; Performing radiation diffusion calculation on the personal risk contribution value to a peripheral grid according to a preset spatial attenuation rule by taking the real-time position of the person as a center, and generating a personal risk field which is continuously distributed in space; performing spatial superposition and smoothing on personal risk fields of all people in a specific period of time to generate a dynamic group risk thermodynamic diagram reflecting the real-time risk distribution of a laboratory universe; the laboratory functional area is taken as a node, the flow path of personnel and materials is taken as an edge to construct a risk coupling network, and the risk conduction intensity between the nodes is calculated based on historical data; Monitoring risk values of all areas in the dynamic population risk thermodynamic diagram, and predicting the conduction effect of a local high-risk area by combining the risk coupling network; Triggering a line safety early warning when the risk value of a first area exceeds a first threshold determined according to the current personnel identity composition and the activity type of the first area and the risk gain of the first area, which is conducted to a target area, exceeds a second threshold; Wherein the calculating, by a risk assessment algorithm, the personal risk contribution value includes: Setting a time sliding window and a geographic grid division rule; accumulating the residence time of the personnel in each grid in each time sliding window to obtain an original residence heating power value; And introducing an environmental risk coefficient to carry out weighted correction on the original resident thermal value according to the preset risk level of the stored article or the running equipment in the grid, and taking the weighted correction result value as the personal risk contribution value of the personnel corresponding to the grid.
  2. 2. The method of claim 1, wherein spatially superimposing and smoothing the personal risk fields of all persons within a particular time period to generate a dynamic group risk thermodynamic diagram reflecting a laboratory global real-time risk distribution, comprising: Acquiring a preset fusion weight matched with each personnel identity label, wherein the fusion weight set by an external contractor is determined according to the historical violation rate and the current operation permission level; Based on the fusion weight, weighting and superposing personal risk fields of all people in a specific time period to obtain initial group risk distribution; and processing the initial group risk distribution by using a space smoothing algorithm, eliminating local noise, forming a continuous risk intensity curved surface, and rendering to generate a dynamic group risk thermodynamic diagram.
  3. 3. The method of claim 1, wherein calculating risk conduction intensity between nodes based on historical data comprises: extracting a pairing sequence of the events which occur in succession to the target functional area node in a preset time window after the events occur to the source functional area node from the historical data; Counting and calculating the proportion of the number of the paired sequences to the total number of the source node events to obtain a basic association degree; and carrying out time attenuation correction on the basic association degree according to the average time interval of the occurrence of the events in the paired sequences, and taking the calculated result as the risk conduction intensity between the nodes.
  4. 4. The method of claim 1, wherein the identifying of abnormal line segments in combination with the corresponding standard safety line template alignment is performed by a first implementation, comprising: carrying out Kalman filtering on the continuous space-time track to smooth the track, and mapping the track point sequence onto a path network of a laboratory map by adopting a map matching algorithm to obtain a standard path sequence; carrying out space-time alignment on the standard path sequence and the standard safety line template, and determining the geometric deviation of the path by calculating the dynamic time warping distance between the actual path and the standard path; determining the area intrusion by checking the actually entered area sequence and the electronic fence authority matrix; identifying timing misalignment by comparing actual timing of the critical operational node with standard timing requirements; and identifying and marking the abnormal moving line segment by integrating the geometric deviation of the path, the intrusion of the area and the criterion of the time sequence dislocation.
  5. 5. The method of claim 1, wherein the identifying abnormal line segments in combination with the corresponding standard safety line template alignment is performed by a second implementation manner, comprising: Coding a static risk level preset in each region into a static feature vector; Based on the continuous space-time track, extracting regional personnel aggregation density and violation frequency, and respectively encoding into dynamic aggregation feature vectors and dynamic violation feature vectors; Fusing the static feature vector, the dynamic aggregation feature vector and the dynamic violation feature vector, and inputting the fused static feature vector, the dynamic aggregation feature vector and the dynamic violation feature vector into a multi-layer perceptron network comprising an attention mechanism to obtain dynamic risk re-estimation values of all areas output by the multi-layer perceptron network; And judging and positioning the abnormal dynamic line segment according to the dynamic risk re-estimation value.
  6. 6. The method of claim 1, wherein monitoring risk values for regions in the dynamic population risk thermodynamic diagram and predicting conduction effects for local high risk regions in conjunction with the risk coupling network comprises: Acquiring risk values of grid cells in the dynamic group risk thermodynamic diagram in real time; inquiring all conducting paths taking the first area as a starting point in the risk coupling network when the risk value of the first area is monitored to be increased; and iteratively calculating potential intensities and ranges of risks propagating along the risk coupling network to adjacent and subsequent nodes according to the risk conduction intensity on each conduction path to form quantitative prediction of the conduction effect.
  7. 7. The method of claim 1, wherein the determining of the risk gain for risk conduction of the first region to the target region comprises: Searching all communication paths from the first area corresponding node to the target area corresponding node in the risk coupling network; for each communication path, multiplying the risk conduction intensity of each side on the communication path to obtain path conduction efficiency; and taking the maximum conduction efficiency in all the communication paths as the risk gain from the first region to the target region.
  8. 8. The method of claim 1, further comprising, prior to calculating the personal risk contribution of the person at each point in space: slicing the dynamic population risk thermodynamic diagram according to preset space-time granularity, and extracting thermodynamic value feature vectors of all areas; Space-time aggregation is carried out on the historical violation event data, and violation frequency feature vectors with the same dimensionality as the thermal value feature vectors are generated; And calculating the dynamic coupling weight between the thermodynamic value feature vector and the violation frequency feature vector through association analysis, and taking the dynamic coupling weight as an adjustment factor into the personal risk contribution value calculation.
  9. 9. The method of claim 1, further comprising, after triggering the line-of-wire security pre-warning: Constructing a causal knowledge graph containing nodes of illegal behaviors, equipment states, environmental parameters and potential safety accidents; inputting the abnormal dynamic line segment and the violation features which are currently identified as query conditions into the causal knowledge graph; and calculating the occurrence probability of each potential safety accident node by using a graph reasoning algorithm, and supplementing at least one type of accidents with the occurrence probability ranked at the front as predicted linkage risks into early warning information.
  10. 10. The laboratory line safety early warning device is characterized by comprising a memory and a processor; the memory is used for storing programs; The processor for executing the program to perform the steps of the laboratory line safety precaution method of any one of claims 1 to 9.

Description

Laboratory line safety early warning method and device Technical Field The application relates to the technical field of safety early warning, in particular to a laboratory line safety early warning method and device. Background The laboratory is used as a core place for scientific research experiments, sample detection and hazardous reagent treatment, the standardization of personnel flow, material transfer and test operation is directly related to personnel safety and test order, and line safety is one of the core links of laboratory safety management. Along with the expansion of laboratory scale and the complicating of test projects, personnel types are diversified, the flow frequency of materials and equipment is greatly increased, and the security accidents such as fire disaster, reagent leakage, equipment damage and the like caused by line violations are frequent, so that an accurate and real-time laboratory line security early warning system is constructed, the advanced pre-judgment and timely treatment of risks are realized, the urgent requirement of the current laboratory security management is met, and the system is a necessary measure for guaranteeing the life security of personnel, reducing property loss and maintaining the normal development of tests. In the prior art, laboratory line safety management mainly relies on manual video monitoring, single-dimension positioning early warning and fixed-area risk monitoring equipment, wherein manual monitoring relies on real-time supervision of management staff, has the problems of monitoring blind areas, response delay and susceptibility to human negligence, and cannot realize global and all-weather accurate monitoring, a single positioning technology only can acquire rough position information of staff, cannot correlate staff behavior data and is difficult to identify abnormal fragments in continuous lines, and traditional risk monitoring is only specific to single-area parameters, and risk judgment is more unilateral. The prior art is easy to have false alarm and missing alarm, is difficult to predict risks in advance, and cannot meet the requirements of fine and intelligent line safety management in a laboratory. Therefore, a new laboratory line safety early warning method is needed to solve the defects of the prior art, realize the whole process accurate monitoring, risk early judgment and intelligent early warning of laboratory line safety, improve the refinement and intelligent level of laboratory line safety management, and effectively prevent the occurrence of line-related safety accidents. Disclosure of Invention The application provides a laboratory line safety early warning method and device, which can realize the whole flow and multidimensional accurate management and control of laboratory line safety, pre-judge the risk diffusion trend in advance and trigger line early warning in time, comprehensively promote the refinement and intelligent level of laboratory line safety management, effectively prevent the occurrence of line related safety accidents and meet the safety management requirement of laboratory diversification. A laboratory line-of-sight safety warning method, comprising: collecting positioning data streams and behavior monitoring data streams of all personnel in a laboratory area in real time, analyzing and correlating to form line base data with time stamps and personnel identity labels; Based on the continuous space-time track of the line-moving basic data reconstruction personnel, the abnormal line-moving segments are identified by combining the comparison of corresponding standard safety line-moving templates, and space-time characteristics and violation characteristics at least comprising a moving speed sequence, a regional stay time length and a line-moving node sequence are extracted; calculating personal risk contribution values of each person at each point in space according to the space-time characteristics and the violation characteristics of each person and basic risk coefficients corresponding to the person identity labels, and generating a personal risk field; performing spatial superposition and smoothing on personal risk fields of all people in a specific period of time to generate a dynamic group risk thermodynamic diagram reflecting the real-time risk distribution of a laboratory universe; the laboratory functional area is taken as a node, the flow path of personnel and materials is taken as an edge to construct a risk coupling network, and the risk conduction intensity between the nodes is calculated based on historical data; Monitoring risk values of all areas in the dynamic population risk thermodynamic diagram, and predicting the conduction effect of a local high-risk area by combining the risk coupling network; and triggering a line safety early warning when the risk value of the first area exceeds a first threshold value determined according to the current personnel identity composition and the activity