Search

CN-122022476-A - Enterprise production security risk assessment and management system based on big data

CN122022476ACN 122022476 ACN122022476 ACN 122022476ACN-122022476-A

Abstract

The invention discloses an enterprise production security risk assessment and management system based on big data, which relates to the technical field of enterprise security production intelligent management and control and comprises a data acquisition module, a data preprocessing module, a risk pattern recognition module, a risk situation deduction module and a closed-loop management module which are connected in sequence. The method comprises the steps of collecting multi-source heterogeneous safety data, carrying out standardized processing, carrying out rolling window scanning and depth mode mining on a multi-dimensional feature sequence by utilizing a time sequence association analysis engine so as to identify a potential composite risk mode, carrying out logic association and situation evolution simulation on risk events by utilizing a knowledge graph model, and predicting a risk development path. The system realizes closed-loop management from data to decision, and can actively discover hidden risks and forecast evolution situations thereof, thereby improving the initiative and the accuracy of enterprise security risk management.

Inventors

  • LUO JINGMING
  • YANG YUXIONG

Assignees

  • 深圳广宏盈信网络科技有限公司

Dates

Publication Date
20260512
Application Date
20260129

Claims (10)

  1. 1. An enterprise production security risk assessment and management system based on big data, the system comprising: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for continuously acquiring safety data streams of heterogeneous data sources of an enterprise production site, and the heterogeneous data sources comprise an equipment operation log sensor, an environment state monitoring sensor, a video monitoring image stream and a personnel operation behavior recording terminal; The data preprocessing module is used for establishing a dynamic data preprocessing pipeline for carrying out real-time cleaning and normalization processing on the safety data stream and generating a standardized multi-dimensional safety characteristic sequence; the risk pattern recognition module is used for deploying a security risk pattern recognition engine based on time sequence association analysis, performing rolling window scanning and depth pattern mining on the standardized multi-dimensional security feature sequence, and outputting a potential risk event set and a risk pattern label; The risk situation deduction module is used for constructing a risk situation deduction model based on a knowledge graph, and the risk situation deduction model carries out logic association and situation evolution simulation according to the potential risk event set and the historical accident case library to generate a risk situation deduction report; and the closed-loop management module automatically generates a closed-loop management instruction set containing specific risk control measures and resource configuration suggestions according to the risk mode labels and the risk situation deduction report.
  2. 2. The big data based enterprise production security risk assessment and management system of claim 1, wherein the secure data stream for continuously collecting enterprise production site heterogeneous data sources comprises: establishing communication connection with the equipment operation log sensor, and acquiring time sequence data streams of equipment vibration, temperature, current and pressure parameters at a preset sampling frequency; Establishing communication connection with the environmental state monitoring sensor, and collecting environmental parameter data streams of toxic and harmful gas concentration, dust concentration, temperature and humidity and noise decibel values in a production area in real time; establishing an access channel with the video monitoring image stream, acquiring a real-time video stream covering a key production area and an operation node by taking an image frame as a unit, and carrying out structural analysis on the real-time video stream; Establishing a data synchronization interface with the personnel operation behavior recording terminal, and receiving a personnel position coordinate, an operation action record and a behavior log data stream of the wearing state of the personal protective equipment, which are reported by the personnel operation behavior recording terminal; And performing time stamp alignment and data encapsulation on the time sequence data stream, the environment parameter data stream, the real-time video stream subjected to structural analysis and the behavior log data stream to form the safety data stream with a unified transmission format.
  3. 3. The big data based enterprise production security risk assessment and management system of claim 1, wherein the creating a dynamic data preprocessing pipeline for real-time cleaning and normalization of the security data stream generates a standardized multi-dimensional security feature sequence comprising: performing missing value interpolation and outlier smoothing on the time sequence data stream in the safety data stream, and eliminating inherent noise of equipment through a moving average algorithm to generate a clean equipment state signal; Performing dimension unified conversion and threshold rationality verification on the environment parameter data streams in the safety data streams, removing abrupt data points generated by sensor transient faults, and generating a standard environment monitoring signal; extracting visual feature vectors of a personnel activity area, an equipment running state area and a material stacking area in each frame of image from the real-time video stream subjected to structural analysis in the safety data stream to generate a standardized visual feature stream; Coding discrete operation action records into continuous operation sequence vectors for the behavior log data streams in the safety data streams, and performing regional gridding mapping on personnel position coordinates to generate normalized behavior feature sequences; and carrying out feature alignment and splicing on the clean equipment state signal, the standard environment monitoring signal, the standardized visual feature stream and the standardized behavior feature sequence according to a unified time sequence standard to form the standardized multidimensional safety feature sequence.
  4. 4. The big data based enterprise production security risk assessment and management system of claim 1, wherein the deploying a security risk pattern recognition engine based on time series correlation analysis, rolling window scanning and depth pattern mining the normalized multi-dimensional security feature sequence, outputting a set of potential risk events and a risk pattern tag, comprises: Setting a rolling time window with adjustable length, continuously sliding on the standardized multi-dimensional security feature sequence, and intercepting feature data fragments in each window; Inputting each characteristic data segment into a pre-trained risk mode classification network, wherein the risk mode classification network consists of a plurality of time sequence convolution kernels and an attention mechanism layer and is used for extracting time sequence dependency relations among the characteristics; the risk mode classification network outputs probability distribution of each characteristic data segment belonging to a predefined risk mode, wherein the predefined risk mode comprises an equipment abnormal wear mode, an environment parameter exceeding mode, a personnel violation operation mode and a multi-factor coupling risk mode; When the probability of the risk mode corresponding to any one of the characteristic data fragments exceeds a preset confidence coefficient threshold, triggering a risk event alarm, and recording the starting and ending time, the related equipment identifier, the related environment area and the related personnel information corresponding to the characteristic data fragment to form a risk event instance; And collecting all triggered risk event instances in the current rolling window period, and labeling the leading risk patterns identified by the risk pattern classification network for the risk event instances to form the potential risk event set and the corresponding risk pattern labels.
  5. 5. The system for evaluating and managing the production security risk of enterprises based on big data according to claim 1, wherein the building of a risk situation deduction model based on a knowledge graph, the risk situation deduction model performs logic association and situation evolution simulation according to the set of potential risk events and a historical accident case library, and generates a risk situation deduction report, includes: the historical accident case library stores a historical risk event chain, final accident results and structural records of key cause nodes; mapping each risk event instance in the potential risk event set to a corresponding entity node in the risk situation deduction model, wherein the entity node type comprises a device entity, an environment entity, a personnel entity and a working activity entity; Activating a predefined risk propagation path for connecting different entity nodes in the risk situation deduction model according to the risk mode label, wherein the risk propagation path is defined by an accident cause logic rule; Simulating a diffusion process of risks along the risk propagation path in the risk situation deduction model by taking the potential risk event set as an initial input, and deducting a secondary risk event node to be triggered and a final accident result type; Summarizing all risk event nodes, risk propagation paths and final accident result types passing through in the simulation deduction process, and generating a risk situation deduction report comprising a risk evolution chain, key risk nodes and recommended intervention points.
  6. 6. The big data based enterprise production security risk assessment and management system of claim 4, wherein triggering a risk event alert when the risk pattern probability corresponding to any of the feature data segments exceeds a preset confidence threshold comprises: Monitoring risk mode probability flows output by the risk mode classification network in real time; setting independent confidence coefficient threshold values for each of the predefined risk modes, wherein the confidence coefficient threshold values are dynamically adjusted according to historical false alarm rates and false alarm rates; when the duration that the probability value of a certain predefined risk mode continuously exceeds the corresponding confidence coefficient threshold value in the risk mode probability flow reaches the preset minimum duration, judging that the alarm is effective; When the effective alarm is generated, synchronously recording the time point of triggering the alarm, the characteristic vector snapshot in the related standardized multi-dimensional security characteristic sequence and the change curve of the risk mode probability flow; and packaging the effective alarm, the time point, the feature vector snapshot and the change curve as core content of the risk event instance.
  7. 7. The big-data-based enterprise production security risk assessment and management system of claim 5, wherein the simulating the risk diffusion process along the risk propagation path in the risk situation deduction model with the set of potential risk events as initial inputs, deducting the secondary risk event nodes to be triggered and the final accident outcome type comprises: setting each risk event instance in the potential risk event set as an active source node in the risk situation deduction model; according to the entity type of each active source node and the risk mode labels, retrieving all effective risk propagation paths taking the active source nodes as starting points in a knowledge graph; Traversing each effective risk propagation path in turn according to the weight and the confidence level of the path, activating the entity node corresponding to the path end point into a new secondary risk event node, and recording the conditions and time delay required by activation; checking whether the newly activated secondary risk event node meets the judging conditions of the accident consequence node, wherein the judging conditions comprise node type, risk state severity and association strength with other nodes; If yes, marking the accident consequence node as the accident consequence node of the deduction, and backtracking and recording a complete transmission chain from an active source node to the accident consequence node; and (3) incorporating the secondary risk event nodes, the deduced accident result nodes and the complete propagation chain obtained by traversing into a simulated deduction result.
  8. 8. The big data based enterprise production security risk assessment and management system of claim 1, wherein automatically generating a closed loop management instruction set comprising specific risk control measures and resource configuration suggestions based on the risk pattern tag and the risk situation deduction report comprises: analyzing the risk mode label, and determining a standard countermeasure template library corresponding to the current dominant risk mode; Analyzing the risk situation deduction report, and extracting key risk nodes and recommended intervention point information in the risk situation deduction report; Matching and instantiating the key risk nodes and recommended intervention point information with the measure items in the standard counter measure template library to generate risk control measure rules for specific equipment, specific areas or specific personnel; evaluating the type, the quantity and the response priority of the needed emergency resources according to the complexity and the urgency of a risk evolution chain in the risk situation deduction report; Combining the current available resource state of the enterprise, and carrying out resource binding and scheduling planning on the risk control measure rule to form an executable resource configuration suggestion; integrating the risk control measure rule with the resource allocation proposal, and endowing a unique instruction number and an execution time window to form a complete instruction in the closed-loop management instruction set.
  9. 9. The big data based enterprise production security risk assessment and management system of claim 8, wherein the matching and instantiating the key risk nodes and recommended intervention point information with the measure entries in the standard counter measure template library to generate risk control measure rules for a specific device, a specific area or a specific person comprises: performing multi-level index searching in the standard counter measure template library according to entity types and risk modes in the key risk nodes and recommended intervention point information; finding a matched abstract measure template, wherein the abstract measure template describes general action steps and targets; Extracting specific equipment identifiers, specific area coordinates or specific personnel identity information contained in the key risk nodes and the recommended intervention point information; replacing placeholder variables in the abstract measure templates by using the specific equipment identifiers, specific area coordinates or specific personnel identity information, and converting the general steps into specific and operable action descriptions; And combining the risk propagation logic mentioned in the risk situation deduction report, adding the preconditions of execution and the blocking effect expected to be achieved for each action description, and forming the risk control measure rule.
  10. 10. The big data based enterprise production security risk assessment and management system of claim 1, further comprising a security data archiving and model optimization module; The safety data archiving and model optimizing module is used for carrying out encryption compression and associated storage on the safety data stream processed every day, the standardized multi-dimensional safety feature sequence, the potential risk event set, the risk situation deduction report and the execution feedback of the closed-loop management instruction set; The security data archiving and model optimizing module is used for carrying out incremental learning and optimizing updating on parameters of the dynamic data preprocessing pipeline, a risk pattern classification network in the security risk pattern recognition engine and risk propagation path weights in the risk situation deduction model based on newly archived historical data regularly.

Description

Enterprise production security risk assessment and management system based on big data Technical Field The invention belongs to the technical field of intelligent management and control of enterprise safety production, and particularly relates to an enterprise production safety risk assessment and management system based on big data. Background The current enterprise production security risk management mainly adopts an alarm technology based on independent monitoring and a static threshold value. In the prior art, sensors and video monitoring are deployed in various devices and environment areas, and operation parameters and environment indexes are collected in real time. The supervision of personnel operation behavior depends on an access control system and an operation recording terminal. The above-described systems typically set a fixed alarm threshold for a single type of data source, and trigger an alarm when the monitored data exceeds the threshold. The prior art scheme has the defects of data isolation and risk cognition on one side. The multisource heterogeneous data lacks effective fusion, and the safety information is in an island state. The static threshold alert mechanism can only respond to explicit abrupt changes of a single parameter, and cannot identify a composite risk pattern consisting of implicit correlations across multiple dimensions. Meanwhile, the existing system generally lacks the dynamic analysis capability of the risk evolution process. The function of the method is limited to the presentation of the current abnormal state, and the chain reaction and the final situation possibly caused by a risk point can not be simulated and deduced based on the logic relation among the risk elements. This results in the security management being in passive response mode for a long period of time. There is a need for a technique that enables multi-source secure data depth fusion and cross-timing correlation analysis to actively discover potentially risky patterns driven by multi-factor coupling. A technique is needed that can perform simulation deduction on the risk dynamic evolution path based on historical experience and logic rules to support prospective research and judgment on the security situation. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; therefore, the invention provides an enterprise production security risk assessment and management system based on big data, which comprises the following steps: the system comprises a data acquisition module, a data processing module and a data processing module, wherein the data acquisition module is used for continuously acquiring safety data streams of heterogeneous data sources of an enterprise production site, and the heterogeneous data sources comprise an equipment operation log sensor, an environment state monitoring sensor, a video monitoring image stream and a personnel operation behavior recording terminal; The data preprocessing module is used for establishing a dynamic data preprocessing pipeline for carrying out real-time cleaning and normalization processing on the safety data stream and generating a standardized multi-dimensional safety characteristic sequence; the risk pattern recognition module is used for deploying a security risk pattern recognition engine based on time sequence association analysis, performing rolling window scanning and depth pattern mining on the standardized multi-dimensional security feature sequence, and outputting a potential risk event set and a risk pattern label; The risk situation deduction module is used for constructing a risk situation deduction model based on a knowledge graph, and the risk situation deduction model carries out logic association and situation evolution simulation according to the potential risk event set and the historical accident case library to generate a risk situation deduction report; and the closed-loop management module automatically generates a closed-loop management instruction set containing specific risk control measures and resource configuration suggestions according to the risk mode labels and the risk situation deduction report. Further, the secure data stream for continuously collecting heterogeneous data sources in an enterprise production site includes: establishing communication connection with the equipment operation log sensor, and acquiring time sequence data streams of equipment vibration, temperature, current and pressure parameters at a preset sampling frequency; Establishing communication connection with the environmental state monitoring sensor, and collecting environmental parameter data streams of toxic and harmful gas concentration, dust concentration, temperature and humidity and noise decibel values in a production area in real time; establishing an access channel with the video monitoring image stream, acquiring a real-time video stream covering a key production area an