Safety monitoring management method and system for chemical product production workshop, and medium

Abstract

The invention discloses a safety monitoring management method and system for a chemical product production workshop and a medium, and relates to the technical field of workshop safety management, and the method comprises the steps: dividing the chemical product production workshop according to regional functions, and building a regional conduction topological structure diagram; arranging multiple types of monitors, and collecting multi-source detection data; training a classifier, performing security identification classification, and positioning a security identification result; according to the regional conduction topological structure diagram, regional fixed-point identification and regional conduction risk prediction are carried out, an early warning level is generated, and graded early warning and product production process emergency management control are carried out according to the early warning level. The technical problems that in the prior art, safety monitoring of a chemical product production workshop is not comprehensive, recognition is not accurate, risks are difficult to effectively pre-judge, the accident occurrence probability is high, and production safety is difficult to guarantee are solved, graded early warning and emergency management control in chemical product production process management are achieved, and the safety of the chemical product production workshop is improved. And the workshop safety monitoring management level is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of workshop safety management, and particularly to a safety monitoring and management method, system and medium for chemical product production workshops. Background Art

[0002] During the production process of chemical products, due to the involvement of numerous flammable, explosive, toxic and harmful chemicals, as well as complex processes such as high temperature and high pressure, there are many potential hidden dangers in the production workshop. In the prior art, the safety monitoring and management of chemical workshops mostly rely on manual inspections and single-type sensors, which not only have problems such as limited monitoring scope and incomplete data collection, but also are difficult to accurately identify and analyze abnormal hidden dangers in complex environments. At the same time, there is a lack of a systematic risk conduction prediction mechanism, and it is impossible to predict in advance the impact of risks on surrounding areas, resulting in difficulty in timely response when accidents occur, seriously threatening production safety and personnel safety.

[0003] The prior art has technical problems such as incomplete safety monitoring, inaccurate identification in chemical product production workshops, difficulty in effectively predicting risks, resulting in a relatively high probability of accidents and difficulty in ensuring production safety. Summary of the Invention

[0004] The present application provides a safety monitoring and management method, system and medium for chemical product production workshops, which are used to solve the technical problems in the prior art that the safety monitoring in chemical product production workshops is incomplete, the identification is inaccurate, it is difficult to effectively predict risks, resulting in a relatively high probability of accidents and difficulty in ensuring production safety.

[0005] In view of the above problems, the present application provides a safety monitoring and management method, system and medium for chemical product production workshops.

[0006] In the first aspect of the embodiments of the present application, a safety monitoring and management method for chemical product production workshops is provided. The method includes:

[0007] Dividing the chemical product production workshop according to regional functions, establishing a regional conduction topology structure diagram; arranging multi-type monitors at the functional area nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facilities to collect multi-source detection data; training a classifier based on the safety level division rules, and performing safety identification and classification through the classifier according to the multi-source detection data to locate the safety identification results; according to the regional conduction topology structure diagram, combining the safety identification results to locate for regional fixed-point identification and regional conduction risk prediction, generating a warning level, and performing hierarchical warning and emergency management control during the product production process according to the warning level.

[0008] In the second aspect of the embodiments of the present application, a safety monitoring and management system for chemical product production workshops is provided. The system includes:

[0009] A topology structure diagram building module, which is used to divide the chemical product production workshop according to regional functions and build a regional conduction topology structure diagram; a multi-source detection data acquisition module, which is used to deploy multi-type monitors at the functional area nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facilities to collect multi-source detection data; a safety identification result positioning module, which is used to train a classifier based on safety level division rules, and perform safety identification classification through the classifier according to the multi-source detection data to locate the safety identification result; a warning level generation module, which is used to perform regional fixed-point identification and regional conduction risk prediction according to the regional conduction topology structure diagram in combination with the safety identification result positioning, generate a warning level, and perform hierarchical warning and emergency management control during the product production process according to the warning level.

[0010] In the third aspect of the embodiments of the present application, the present application provides a computer-readable storage medium storing a computer program for executing the safety monitoring and management method for a chemical product production workshop provided by the present application.

[0011] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0012] Divide the chemical product production workshop according to regional functions and build a regional conduction topology structure diagram; deploy multi-type monitors to collect multi-source detection data; train a classifier, perform safety identification classification through the classifier according to the multi-source detection data, and locate the safety identification result; according to the regional conduction topology structure diagram, perform regional fixed-point identification and regional conduction risk prediction in combination with the safety identification result positioning, generate a warning level, and perform hierarchical warning and emergency management control during the product production process according to the warning level. It achieves the technical effects of hierarchical warning and emergency management control in the chemical product production process management and improves the technical level of workshop safety monitoring and management. Description of the Drawings

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0014] Figure 1 It is a schematic flowchart of the safety monitoring and management method for a chemical product production workshop provided by the embodiments of the present application.

[0015] Figure 2 It is a schematic structural diagram of the safety monitoring and management system for a chemical product production workshop provided by the embodiments of the present application.

[0016] Description of the attached drawing reference numerals: Topology structure building module 10, multi-source detection data acquisition module 20, security identification result positioning module 30, warning level generation module 40. Specific implementation mode

[0017] The present application provides a safety monitoring and management method, system and medium for a chemical product production workshop, which is used to solve the technical problems in the prior art that the safety monitoring of a chemical product production workshop is incomplete, the identification is inaccurate, it is difficult to effectively predict risks, resulting in a relatively high probability of accidents and difficult to guarantee production safety.

[0018] Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0019] Embodiment 1, as Figure 1 shown, the present application provides a safety monitoring and management method for a chemical product production workshop, and the method includes:

[0020] Step S100: Divide the chemical product production workshop according to the regional function to establish a regional conduction topology structure diagram.

[0021] Specifically, the workshop is divided into regional functions by comprehensively considering multiple factors, including process flow, sorting out the processing sequence of materials from raw material input to finished product output; equipment relevance, clarifying the collaborative relationship and energy transfer path between each equipment; spatial position relationship, determining the actual location of each area based on the architectural layout of the workshop; material properties, analyzing the flammable, explosive, toxic and harmful characteristics of materials stored or used in each area; personnel and energy flow, mastering the daily activity routes of personnel and the direction of energy transmission. After the division is completed, each functional area is defined as a node, and node attributes containing basic information such as node number, function type, material or energy risk level are assigned to it. Then, directed connection edges are established between nodes according to process flow, equipment relevance, and spatial position relationship, and the source node, target node, connection type, material hazard attribute and hazardous material conduction probability of each edge are recorded in detail. Through these data, the conduction risk weight of each edge is calculated and the edge weight is assigned, and then a weighted directed graph is constructed to form a regional conduction topological structure diagram. At the same time, according to the real-time process status, it will dynamically analyze the process reaction risk level, configure the regional risk adjustment boundary, reconstruct the functional zoning boundary, adjust the adjacent edge weight according to the span of the boundary reconstruction area, and dynamically update the regional conduction topology structure diagram in a timely manner, so that it can accurately reflect the real-time status of the workshop, and provide an accurate and effective infrastructure for subsequent multi-source data collection, safety identification and classification, risk prediction, early warning and emergency management based on the diagram.

[0022] Step S200: deploy multiple types of monitors at nodes in the reaction unit, raw material storage area, high temperature and high pressure equipment area, and power auxiliary facility functional area to collect multi-source detection data.

[0023] Specifically, in the reaction unit, since the chemical reaction process is complex and prone to produce combustible gas, toxic substances and heat changes, combustible gas detectors, toxic gas detectors, temperature sensors, etc. will be deployed in a targeted manner according to specific process requirements. Combustible gas detectors monitor in real time whether there is a combustible gas leak to prevent fire and explosion accidents; toxic gas detectors protect personnel from toxic substances; temperature sensors can detect temperature anomalies in the reaction process in a timely manner to prevent the reaction from getting out of control.

[0024] In the raw material storage area, in addition to combustible gas and toxic gas detectors, pressure sensors are also installed to take into account the dangers of stored materials. Because some raw materials may cause pressure changes during storage due to factors such as temperature and chemical reactions, pressure sensors can monitor pressure values in real time to ensure storage safety.

[0025] The high-temperature and high-pressure equipment area is a high-incidence area for accidents, where vibration sensors play an important role. It can monitor the vibration of equipment and judge whether there are potential faults in the equipment by analyzing vibration data. At the same time, temperature sensors and pressure sensors are also essential for monitoring the temperature and pressure during equipment operation to ensure the stable operation of the equipment.

[0026] The power auxiliary facilities provide power support for the entire workshop, and the monitoring of their operating status is equally important. Video monitors can directly observe the operating conditions of the equipment, facilitating the timely discovery of abnormalities. In addition, temperature sensors and pressure sensors are also equipped to ensure that the power auxiliary facilities operate within safe parameter ranges.

[0027] By deploying multiple types of monitors at the nodes of these functional areas, multi-source detection data such as combustible gas concentration, toxic substance leakage, temperature, pressure, and equipment vibration can be collected. These data not only provide rich information for training classifiers based on safety level classification rules in the follow-up, but also serve as an important basis for regional fixed-point identification, regional conduction risk prediction, and then generating warning levels and implementing hierarchical early warnings and emergency management, comprehensively ensuring the safe production of chemical product production workshops.

[0028] Step S300: Train a classifier based on safety level classification rules, and perform safety identification and classification through the classifier according to the multi-source detection data to locate the safety identification results.

[0029] Specifically, historical accident record data is collected and industry standards are referred to in order to set safety classification rules. Based on the safety sample characteristics, types, and corresponding safety level annotations in the historical accident records, a dataset for training and validation is constructed. The training dataset is used to train the selected classifier, and the validation dataset is used to evaluate the performance of the classifier and ensure its accuracy. During this process, if new monitored hazardous material information (i.e., a situation different from the safety types in the historical accident record data) is encountered, hazardous characteristics are collected based on this new information, and the corresponding safety types and safety levels are annotated. An incremental training set and a validation set are constructed, and then the initially trained classifier is incrementally learned to continuously optimize the classifier until the validation convergence goal is achieved, thereby obtaining a classifier with excellent performance that can accurately identify various hazards. After the classifier training is completed, the multi-source detection data collected at the nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facility functional area is input into this classifier. The classifier deeply analyzes and processes this data according to the learned rules and patterns, quickly determines the type of safety, such as whether it is a combustible gas leak, a toxic gas leak, or an abnormality caused by equipment failure, etc., and determines its safety level, accurately locating the safety identification result. These results will serve as the key basis for subsequent area fixed-point identification, area conduction risk prediction, generating warning levels, and implementing classified early warning and emergency management in combination with the area conduction topology structure diagram, providing strong support for ensuring the safety of the chemical production workshop.

[0030] Step S400: According to the area conduction topology structure diagram, in combination with the safety identification result, conduct area fixed-point identification and area conduction risk prediction, generate a warning level, and conduct classified early warning and emergency management control for emergency management during the product production process according to the warning level.

[0031] Specifically, when conducting area fixed-point identification and area conduction risk prediction, first map the initial risk nodes to the dynamic area conduction topology structure diagram. In terms of calculating the dynamic conduction weight, according to the real-time process parameters (pressure change rate, material flow rate) and environmental parameters (wind speed, obstacle density), use the formula to correct the conduction weight of the connection edge. Among them, w ij represents the initial edge weight, which is the basic weight value of the risk conduction from node i to node j without considering the change of real-time parameters; ΔP ij is the pressure gradient between adjacent nodes, reflecting the change of pressure between adjacent nodes. The greater the pressure gradient, the greater the impact on the risk conduction weight; v wind is the real-time wind speed, reflecting the speed condition of the wind in the current environment. The wind speed will affect the diffusion of hazardous substances, etc., and thus affect the conduction weight; v refis the reference wind speed, serving as a comparison benchmark to measure the degree of influence of the real-time wind speed on the conduction weight; α and β are weight coefficients used to adjust the influence degree of the pressure gradient and wind speed on the conduction weight; w‘ ij (t) is the conduction weight of the connection edge after being corrected by real-time parameters at time t, comprehensively reflecting the weight of risk conduction between nodes under the current working conditions. In the risk diffusion simulation, starting from the initial risk node, an improved graph propagation algorithm combined with Monte Carlo random walk is used, and through the formula R j (t) = ∑ i∈邻接节点 R i (t - 1)·w‘ ij (t)·λ ij (t) to predict the risk conduction path and risk value. Among them, R i (t - 1) represents the risk value of the adjacent node i at time t - 1, which is the basic data for calculating the risk value of the current node; w‘ ij (t) is the corrected conduction weight from node i to node j at time t calculated above, determining the proportion of risk conduction from node i to node j; λ ij (t) is the dynamic attenuation factor, which is inversely proportional to the obstacle distribution density. The denser the obstacle distribution, the smaller the value of λ ij (t), meaning that the risk is more hindered and attenuated during the conduction process; R j (t) is the risk value calculated for the target node j at time t, obtained by weighted summation of the relevant data of the adjacent nodes.

[0032] Meanwhile, the risk center point is determined by combining the safety recognition result positioning, and the area fixed-point recognition is completed and the fixed-point risk level is divided. The early warning level is generated according to the preset fuzzy logic rule table by integrating the area fixed-point recognition risk level and the area conduction risk prediction result. When the early warning level is red (global high risk), with the help of the automatic control function of the distributed control system (DCS), the valves of the material supply pipelines in the dangerous area are cut off and the valves of the associated equipment are closed; the global spray system and the inert gas injection device are activated through the industrial automation system; an alternative production path is generated based on the simulation ability of the digital twin model, and the intelligent algorithm is used to adjust the material conveying system to bypass the high-risk area; the reinforcement learning algorithm is used to optimize the equipment control system to ensure that the equipment in the non-dangerous area starts and stops according to the new sequence, maintaining more than 50% of the production capacity. If it is an orange early warning (regional medium risk), physical isolation facilities are used to isolate the dangerous area, and the reaction temperature, flow and other parameters of the upstream and downstream process equipment are adjusted through the industrial parameter adjustment system; relying on the data processing and real-time control capabilities of the edge computing node, the operation parameters of the equipment are collected and optimized; the explosion-proof robot is commanded to enter the dangerous area to perform tasks through the robot scheduling system. For yellow or blue early warnings (local low risk), the monitoring frequency is increased to 1 time per second by upgrading the sensors and monitoring system, and an audible and visual alarm device is used to give a warning; information is pushed to the patrol personnel with the help of the visualization and positioning functions of the AR device; according to the results of the risk conduction prediction model, the emergency resources are deployed to the high-probability risk area in advance through logistics scheduling means, and finally accurate hierarchical early warning and emergency management control during the product production process are realized.

[0033] In a possible implementation manner, step S200 further includes:

[0034] Step S210: The multi-type monitors include: combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, video monitors, and the monitoring device layout nodes in the functional area include one or more of the combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, video monitors.

[0035] Specifically, to comprehensively and accurately monitor the operation status of the workshop, a variety of monitoring devices are selected, such as combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, and video monitors. Due to their different characteristics, different functional areas face different risks, so the layout of monitors in each area will vary. In the raw material storage area, since most of the stored raw materials are flammable, explosive, or toxic substances, combustible gas detectors and toxic gas detectors are installed simultaneously to monitor in real time whether there is leakage of combustible or toxic gases; temperature sensors and pressure sensors will also be set as needed because changes in temperature and pressure may affect the stability of raw materials and even trigger safety accidents. In the high-temperature and high-pressure equipment area, vibration sensors are crucial. It can detect potential equipment failure hazards in a timely manner by monitoring data such as the vibration amplitude and frequency of the equipment; at the same time, to prevent the equipment from operating dangerously due to over-temperature and over-pressure, temperature sensors and pressure sensors are also essential. Because the chemical reactions in the reaction unit are complex, in addition to combustible and toxic gas detectors, the temperature changes during the reaction process need to be focused on. Therefore, temperature sensors are usually the standard monitoring equipment in this area. In the power auxiliary facilities area, video monitors can visually present the operating status of the equipment, facilitating operators to grasp the situation at any time; temperature sensors and pressure sensors are used to monitor the key parameters during equipment operation to ensure its stable operation. In short, each functional area will select one or more of these multiple types of monitors for reasonable layout according to its own risk characteristics, so as to achieve comprehensive and effective monitoring of the chemical product production workshop and provide a strong guarantee for safe production.

[0036] In a possible implementation manner, step S100 further includes:

[0037] Step S110: Analyze the characteristic relationships of each functional area according to the process flow, equipment correlation, spatial position relationship, material properties, personnel and energy flow.

[0038] Step S120: Define each functional area as a node, and the node attributes include node number, function type, and basic information of the material or energy risk level.

[0039] Step S130: Establish a directed connection edge between the nodes according to the process flow, equipment correlation, and spatial position relationship. Each edge includes the source node and the target node, connection type, material hazard attribute, and conduction probability of the hazardous substance.

[0040] Step S140: Calculate the conduction risk weight of each edge according to the material safety level and the conduction probability of the hazardous substance in the connection edge, assign a weight to the edge, and establish a weighted directed graph based on the nodes, edges, and the weights of the edges to obtain the regional conduction topology diagram.

[0041] Specifically, first, deeply study the process flow, carefully sort out every step from raw material input to product output, clarify the sequence and role of each functional area in the entire production chain, as well as the transfer path and transformation process of materials between different areas. At the same time, based on the equipment correlation, analyze the collaborative operation relationship between the equipment within each area, judge which equipment has connections such as material transportation, energy transfer or signal interaction, and then determine the associations generated by equipment cooperation between functional areas. Starting from the spatial position relationship, accurately consider the actual distribution of each functional area in the workshop, including the relative position, distance and layout characteristics between areas, which will affect the diffusion range and speed of hazardous substances. Regarding the material properties, carefully analyze the chemical properties of the substances involved in each area, such as whether they are flammable, explosive, toxic, harmful, strongly corrosive, etc., and the stability and reactivity of these substances under different conditions. In addition, also closely pay attention to personnel and energy flow, carefully record the daily activity trajectories, operation processes and frequently active areas of personnel in each area, and at the same time analyze the transmission direction, transformation form and concentrated areas of energy in the workshop. Through the comprehensive analysis of the process flow, equipment correlation, spatial position relationship, material properties, personnel and energy flow, deeply explore the internal connections and interactions between each functional area, laying a solid foundation for the subsequent establishment of the regional conduction topology structure diagram.

[0042] Based on the comprehensive analysis of each area in the workshop in the early stage, each functional area is separately defined as a node, aiming to provide a clear and independent research object for subsequent risk analysis and management. Each node is given specific attribute information, where the node number is the unique identifier of the node, which can quickly and accurately locate and distinguish different functional areas whether in data processing or subsequent operations. Taking a chemical product production workshop as an example, each node is given specific attribute information: Node number: Adopt the rule of "the first letter of the area type - serial number", such as the raw material storage area is numbered R - 001, and the reaction unit is numbered P - 003, which serves as the unique identifier of the node and is convenient to quickly retrieve the relevant parameters of the ethanol storage area through R - 001 in the data management system. The function type clarifies the specific tasks undertaken by this area in the chemical production process, such as whether it is a raw material storage area responsible for raw material storage, a reaction unit for chemical reactions, or a power auxiliary facility area that provides power support for the entire production process, etc., which helps to grasp the nature and role of each area from a macro level. The basic information of the material or energy risk level is the preliminary risk judgment based on the comprehensive evaluation of the material properties (such as flammability, explosiveness, toxicity, etc.) and energy flow characteristics (such as whether there is a high-temperature and high-pressure energy source, etc.) in the area. Taking a chemical product production workshop as an example, for the raw material storage area (R - 001): Ethanol is a flammable liquid (flash point 12°C, risk level M), and chlorine is a highly toxic gas (LC50 = 90mg / m 3, Risk level H), the basic risk level of this node is comprehensively determined to be H; Reaction unit (P-003): High-temperature environment (200 °C) accompanied by chlorine leakage risk, energy risk level is M, substance risk level is H, and the comprehensive basic risk level is H; Power auxiliary facilities (A-005): High-pressure steam (10 MPa) belongs to an energy hazard source (risk level M), and there is no storage of highly toxic substances. The comprehensive basic risk level is M. This information provides basic data for subsequent risk conduction analysis based on the topological structure, making the analysis of the risk status of the entire workshop more targeted and accurate.

[0043] According to the technological process, along the path of material flow, determine the connection sequence between each node. For example, from the raw material storage area node to the reaction unit node, reflecting the process trend of material input into production. At the same time, combined with the equipment correlation, sort out the energy transfer, material transportation and other correlations between equipment, and establish connections between the corresponding nodes. If two pieces of equipment are located in different functional areas and there is close cooperation, a connection edge is constructed between the nodes representing these two areas. When considering the spatial position relationship, analyze the nodes in adjacent areas or areas that are not adjacent but may have risk conduction, and establish connections for them. Each connection edge contains rich key information: The source node and the target node clearly define the starting and ending areas of the connection, clarifying the direction of risk conduction; The connection type is used to indicate the specific nature of the association between nodes, whether it is a material transmission connection, an energy conduction connection or other functional connections; The hazardous properties of substances detail the hazardous characteristics of the substances involved in the connection, such as whether they are flammable, toxic, etc.; The conduction probability of hazardous substances is determined through data statistics, reflecting the likelihood of hazardous substances being conducted from the source node area to the target node area. By establishing these directed connection edges containing multi-dimensional information, the risk conduction relationship between each functional area of the chemical product production workshop can be accurately presented, providing a key basis for subsequent risk assessment and management.

[0044] For each connecting edge, extract the information on the material safety level and the conduction probability of hazardous substances contained therein. The material safety level is a pre-defined grading standard based on various factors such as the chemical properties and hazard levels of the substances themselves. The higher the value, the higher the degree of danger. The conduction probability of hazardous substances is determined through various methods such as statistical analysis of past accident data, simulation experiments in similar chemical environments, and the experience judgment of professionals. It reflects the likelihood of hazardous substances spreading from the starting area (source node) to the ending area (target node) of the connecting edge. Then, by combining the material safety level and the conduction probability of hazardous substances, using a simple multiplication method, that is, the conduction risk weight is equal to the material safety level multiplied by the conduction probability of hazardous substances. In this way, the obtained weight value comprehensively considers two key factors: material hazard and transmission possibility. After obtaining the conduction risk weight for each edge, assign these weights to the corresponding connecting edges, so that the connecting edges can not only reflect the connection relationship between regions but also intuitively show the risk levels borne by different connections. Finally, based on the nodes with defined attributes, the edges with connection relationships, and the weight information assigned to the edges, construct a weighted directed graph. In this graph, the nodes represent the various functional areas of the chemical product production workshop, the directed edges between the nodes clearly show the possible conduction directions of risks, and the weights of the edges present the relative severity of risk conduction in a quantitative form. Through this method, a regional conduction topology structure diagram is successfully obtained, which comprehensively and intuitively depicts the risk conduction network between the various functional areas in the workshop.

[0045] In a possible implementation manner, step S100 further includes:

[0046] Step S150: Dynamically analyze the risk level of the process reaction according to the real-time process state, and configure the regional risk adjustment boundary.

[0047] Step S160: Reconstruct the boundary of the functional partition according to the regional risk adjustment boundary, and adjust the weights of the adjacent edges according to the span range of the boundary reconstruction area.

[0048] Step S170: Dynamically update the regional temporalization of the regional conduction topology structure diagram according to the adjustment results of the reconstructed boundary of the node area and the weights of the adjacent edges.

[0049] Specifically, first, the process is carefully divided, and multi-stage state recognition logic is set up, covering stages such as reaction start-up, stable reaction, exothermic or overpressure, and cooling. Different stages correspond to different risk levels and regional impact radius parameters, which are determined based on a large amount of historical data, process characteristic research, and professional evaluations. For example, during the reaction start-up stage, various reaction conditions are in the initial establishment state, the reactant concentration and reaction rate are relatively stable, so the risk level is low and the regional impact radius is small; while in the exothermic or overpressure stage, the chemical reaction is intense, a large amount of heat accumulates or the pressure exceeds the normal range, which is likely to cause danger, the risk level is high, and the regional impact radius is large. Process parameters such as temperature, pressure, flow rate, and material concentration are collected in real time through various sensors. Based on these real-time data, the current process state can be accurately judged. Once the state is determined, the risk level of the process reaction is dynamically analyzed according to the pre-set rules. For example, during the reaction exothermic stage, when it is monitored that the temperature rises above a certain threshold and continues to rise, it is determined that the risk level increases, and the warning range of the reaction unit is automatically expanded, increasing the warning radius by 1.5 - 2 times, so as to configure the regional risk adjustment boundary and clarify the possible range of risk spread.

[0050] Taking the chlorination reaction process in a chemical workshop as an example, its multi-stage state recognition logic and parameter configuration are as follows: Reaction startup stage (0 - 30 min): The process characteristics are that ethanol (initial concentration 99.5%) and chlorine gas (initial pressure 0.5 MPa) are injected into the reactor in proportion, the temperature gradually rises to 120 °C, and the reaction rate is 0.1 mol / (L·min). Based on historical data, the risk level in this stage is L (low), and the default regional influence radius is 5 meters (centered on the reactor). For example, startup records show that when the temperature rises to 100 °C and the pressure stabilizes at 0.3 MPa, the risk level remains at L, and the influence radius does not expand. Stable reaction stage (30 - 180 min): The process parameters are maintained at a temperature of 180 - 220 °C, a pressure of 1.2 MPa, the reaction rate is stable at 0.8 mol / (L·min), and the purity of the product chloroethane is 98%. The risk level in this stage is M (medium), and the influence radius expands to 8 meters. Historical data shows that when the temperature fluctuates within ±5 °C and the pressure fluctuates within ±0.1 MPa, the risk level remains at M. For example, during stable operation, with a temperature of 210 °C and a pressure of 1.15 MPa, the influence radius does not change. Exothermic or overpressure stage (abnormally triggered): When the sensor real-time collects that the temperature exceeds 220 °C and the heating rate per minute > 5 °C, or the pressure exceeds 1.5 MPa and the pressure increase rate > 0.2 MPa / min, it is determined to enter the high-risk stage. For example, in an abnormal event, the temperature suddenly rises from 210 °C to 240 °C (heating rate 6 °C / min) within 5 minutes, and the pressure reaches 1.8 MPa. According to the rules, the risk level is dynamically analyzed and upgraded to H (high). The warning radius of the reaction unit is automatically expanded from 8 meters to 15 meters, and the adjacent raw material storage area (20 meters away from the reactor) is included in the risk adjustment boundary. Cooling stage (after the reaction terminates): The process uses circulating cooling water (inlet temperature 25 °C, flow rate 50 m 3 / h) Lower the reactor temperature below 80°C, return the pressure to atmospheric pressure, and gradually reduce the risk level to L. The influence radius shrinks to 5 meters. During the cooling process, for example, when the temperature drops to 70°C and the pressure is 0.1 MPa, the risk level returns to L and the boundary shrinks synchronously. Real-time collect process parameters such as temperature (accuracy ±1°C), pressure (accuracy ±0.05 MPa), flow rate (ethanol flow meter range 0 - 300 L / h, accuracy ±1%), and material concentration (chlorine detector range 0 - 1000 ppm, resolution 1 ppm) through the distributed control system (DCS). For example, when receiving the temperature sensor data of 235°C (exceeding the threshold of 220°C) and the heating rate of 7°C / min, combined with the pressure sensor data of 1.6 MPa (exceeding the threshold of 1.5 MPa) and the pressure increase rate of 0.3 MPa / min, it is determined that the exothermic overpressure stage is entered, and the risk level upgrade logic is triggered. At this time, according to the preset rules, the regional risk adjustment boundary of the reaction unit is expanded from an 8-meter radius centered on the original coordinates (X25 - Y15) to a 15-meter radius, covering the power auxiliary facility area 10 meters to the east. At the same time, a risk warning is sent to the adjacent raw material storage area (20 meters away from the reactor, outside the 15-meter boundary but subsequent conduction needs to be monitored). This dynamic configuration process is realized through the PLC (programmable logic controller), and the response time from parameter anomaly to boundary adjustment is ≤2 seconds to ensure the accurate definition of the risk diffusion range.

[0051] After determining the regional risk adjustment boundary, start the reconstruction of the functional partition boundary. Take the critical abnormal stage of a sharp increase in pressure as an example. At this time, according to the risk adjustment boundary, temporarily merge the adjacent functional units into the high-risk area for unified management. This is because in such a high-risk situation, the possibility of risk conduction to adjacent areas is extremely high, and including them in the unified management scope can more effectively prevent and control risks. At the same time, adjust the weight of the adjacent edge according to the span range of the boundary reconstruction area. The larger the span range, the greater the potential impact of risk conduction. For example, for the connecting edge between two adjacent functional areas, after the boundary reconstruction, if one area is included in the high-risk area and the span range is large, the risk weight of this adjacent edge will be increased by more than 50% according to the rules. Accurately calculate the weight adjustment range according to specific span values, the degree of association between regions, etc., so as to quantify the change of risk conduction between regions and make the risk assessment more accurate.

[0052] Based on the reconstructed boundaries of node regions and the adjustment results of adjacent edge weights, update the regional conduction topology structure diagram. In the topology structure diagram, nodes represent each functional region, and their attributes reflect the risk status of the region, such as risk level, whether it is in a high-risk state, etc.; the connecting edges represent the associations between regions, and the weights reflect the degree of risk conduction. Dynamically redraw the node attributes and the weights of the connecting edges. For example, for a node that was originally in a normal state, after the process state changes and enters the high-risk stage, the attributes such as the color and shape of the node may change to visually display the change in its risk status; the thickness or color of the connecting edge will also change according to the weight adjustment. The higher the weight, the thicker the edge or the more prominent the color. Each adjustment brought about by the change in the process state is synchronized to the topology diagram in real time, forming a regional conduction diagram with time series characteristics that can reflect the risk evolution process. In this way, the safety recognition model and the early warning decision-making module can obtain the latest and accurate risk distribution information, and effectively identify hazards and make reasonable early warning decisions based on this to ensure the safe and stable operation of chemical production.

[0053] In a possible implementation manner, step S300 further includes:

[0054] Step S310: Set safety classification rules using historical accident record data and industry standards.

[0055] Step S320: Construct a training data set and a validation data set according to the safety sample characteristics, safety types, and safety level annotations in the historical accident record data.

[0056] Step S330: Train and validate the convergence of the initial classifier using the training data set and the validation data set to obtain the classifier, which is used to input the monitoring data characteristics and output the safety type and safety level.

[0057] Specifically, historical accident record data are collected. These data are sourced from various production scenarios within the chemical industry, covering accident information that occurred during the past production processes of chemical enterprises of different scales and different process types, including details such as the time and location of the accident, the source that triggered the accident, and the degree of harm caused by the accident. At the same time, strict reference is made to industry standards, which are formulated by industry associations, regulatory departments, etc., and comprehensively consider the safety, standardization of chemical production, and the prevention and control requirements for various risks. On this basis, safety grading is carried out according to the laws presented by the historical accident record data and the normative guidance of industry standards. During the grading process, multiple key attributes of safety are key considerations. For example, for those hazards where the substance itself has a high degree of flammability, explosiveness or strong toxicity, and is extremely likely to trigger catastrophic consequences such as large-scale explosions, fires or serious environmental pollution once out of control during the production process, they will be classified as high-level. For those hazards that are relatively stable under normal production conditions and are only likely to cause minor impacts or minor hazards even in case of abnormal situations, they are classified as low-level. Through the comprehensive weighing of various factors such as the nature of safety, the potential degree of harm, and the likelihood of accidents, a set of scientific, reasonable and practical safety grading rules for production are set.

[0058] Based on the set safety grading rules, in-depth mining and sorting of the historical accident record data are carried out. Safety sample features are extracted one by one from the massive data. These features cover multi-dimensional information such as substance attributes (such as chemical composition, ignition point, toxicity concentration threshold), equipment operation parameters (abnormal fluctuations in temperature, pressure, rotation speed), and process operation conditions (material ratio deviation, reaction time out of control). At the same time, according to the actual situation of the accident, the safety type corresponding to each sample is accurately marked, such as chemical reaction out-of-control type, equipment failure type, human operation error type, etc., as well as the safety level determined according to the grading rules, divided into high, medium, and low levels. Subsequently, when constructing the training data set and the validation data set, the ratio is set to 7:3. The training data set is mainly used for the subsequent training of the classifier. Through a large number of sample data, the classifier learns the mapping relationship between different features and safety types and safety levels; the validation data set undertakes the important task of testing the training effect. During the training process of the classifier, the validation data set is used to test it to judge the prediction accuracy and stability of the classifier, ensuring that the classifier can accurately identify the safety type and judge the safety level in actual applications, thus providing a reliable guarantee for chemical production safety.

[0059] The gradient boosting decision tree (GBDT) algorithm is used to train and optimize the classifier. First, the safety sample features in the training dataset (such as abnormal fluctuations in reaction temperature, over-limit amplitude of pressure, toxic concentration of materials, vibration frequency of equipment, etc.), the labeled safety types (such as chemical reaction out-of-control type, equipment failure type, etc.) and safety levels (high, medium, low) are used as inputs. The GBDT algorithm is based on the additive model and the forward stagewise algorithm. Starting from an initial simple decision tree, in each round of iteration, it fits the residual between the prediction result of the previous round and the true value. By minimizing the loss function (such as the logarithmic loss function, suitable for classification problems), a new decision tree is constructed to correct the previous prediction bias. During the training process, aiming at the dangerous characteristics in chemical production, the feature weights related to the dangerous properties of substances and the fluctuations of process parameters are mainly optimized. For example, when accident samples caused by sudden increases in reaction temperature appear multiple times in the training data, the algorithm will increase the weight of the temperature feature in the decision tree splitting to highlight its key role in safety classification. After each decision tree is constructed, the validation dataset is used to verify the effect. The validation data is input into the current model to obtain the prediction results of safety types and safety levels, and the model performance is evaluated by calculating indicators such as accuracy rate and confusion matrix. If the recognition accuracy rate of the high-risk level of the model on the validation set is low, or there are misjudgments for different types of safety, the parameters of the GBDT algorithm, such as the depth of the tree, learning rate, subsampling ratio, etc., are adjusted, and training and validation are carried out again. After multiple rounds of iterative optimization, when the evaluation indicators of the model on the validation dataset no longer improve significantly and tend to be stable, it is considered that the model reaches the convergence state. At this time, the trained GBDT model is the required classifier. In the actual chemical production scenario, by inputting the data features such as temperature, pressure, and material concentration obtained by real-time monitoring into this classifier, the model can quickly and accurately output the corresponding safety types and safety levels, providing a scientific basis for the safety risk warning and control of chemical enterprises.

[0060] In a possible implementation manner, step S320 further includes:

[0061] Step S321: Obtain the monitored dangerous substance information, where the monitored dangerous substance information is different from the safety types in the historical accident record data.

[0062] Step S322: Collect dangerous features based on the monitored dangerous substance information and label safety types and safety levels to construct an incremental training set and a validation set.

[0063] Step S323: Use the incremental training set and the validation set to perform incremental learning on the classifier until the validation convergence target is reached.

[0064] Specifically, through diverse monitoring devices deployed in production workshops and warehousing areas, such as highly sensitive gas detectors, infrared thermal imagers, vibration sensors, etc., production environment data is collected in real time. At the same time, with the help of on-line analytical instruments, the composition of newly introduced chemical raw materials and the characteristics of intermediate products during the reaction process are analyzed. For example, when a new type of nanomaterial is used as a catalyst in the reaction, information such as the unique physical and chemical properties of the material and its activity changes during the reaction process has no corresponding type in the historical accident record data; or due to process upgrades, a high-temperature and high-pressure continuous flow reaction process is introduced, and the resulting risks such as abnormal fluid dynamics and local overheating are also different from previous accident types. Through the data monitoring and analysis of these new scenarios, new substances, and new processes, it is possible to accurately identify and obtain monitoring hazardous material information different from historical safety types, laying a foundation for subsequent risk prevention and control.

[0065] For the monitored hazardous materials, a variety of professional detection means are used, such as gas chromatography-mass spectrometry (GC-MS) and X-ray diffractometers, etc., to deeply analyze their material composition and physical and chemical properties; with the help of high-precision sensors, process parameters such as temperature, pressure, flow rate, and material concentration during the production process are collected in real time, as well as operation status data such as the vibration frequency, current, and voltage of the equipment. For example, for newly introduced special chemical raw materials, not only their inherent hazardous characteristics such as flash point, ignition point, and toxicity need to be detected, but also their dynamic characteristics such as concentration changes during the reaction process and compatibility with other substances need to be monitored. After collecting the hazardous characteristic data, an expert safety team is organized to label the monitored hazardous materials according to the set safety classification rules, combined with industry standards and actual production experience. For example: if the risk of reaction runaway is caused by the special reaction activity of the new raw material, it is labeled as a new raw material reaction runaway type; if there are mechanical failure hazards due to the new structure of the equipment, it is labeled as a special equipment failure type. At the same time, according to factors such as the potential hazard degree and the possibility of accidents of the hazardous materials, their safety levels are divided into three levels: high, medium, and low. Finally, according to the ratio of 7:3, the labeled sample data is respectively divided into an incremental training set and a validation set. The incremental training set is used to let the initial classifier learn the characteristics of new types of safety, and the validation set is used to test the prediction accuracy of the initial classifier after learning new knowledge, ensuring that the model can effectively identify newly emerging risks, and finally completing the training to obtain the classifier.

[0066] The gradient boosting decision tree (GBDT) algorithm is used to perform incremental learning on the initial classifier. First, the incremental training set is input into the existing GBDT classifier. GBDT is an iterative decision tree ensemble algorithm. Based on the idea of the forward stagewise algorithm, a new decision tree is generated in each round of iteration to fit the residuals of the previous round of model. For each sample in the incremental training set, it includes the previously collected dangerous features, the labeled safety types, and safety levels. When constructing a new decision tree in each round of iteration, GBDT randomly selects a part of these features and uses metrics such as the Gini index or mean squared error to determine the optimal splitting point, divides the samples into different child nodes, and gradually constructs the structure of the decision tree. After each round of training is completed, a validation set is used to evaluate the performance of the current model. For the samples in the validation set, they are input into the current GBDT model, and the model makes predictions according to the internal decision tree structure and outputs the corresponding safety types and safety levels. Then, the error between the prediction result and the true label is calculated. Common evaluation metrics include accuracy, recall, F1 value, etc. If the evaluation result does not reach the preset validation convergence target, such as the accuracy being lower than a certain threshold, or the F1 value not having a significant improvement, some hyperparameters of GBDT will be adjusted, such as the learning rate, the maximum depth of the tree, the sample sampling ratio of each tree, etc. After adjustment, the incremental training set is used again for a new round of training and learning, generating a new decision tree and adding it to the model. This process will be repeated continuously to continuously optimize and improve the model. Each round of training enables the model to better learn the features and classification rules of new types of safety. As the training progresses, the performance of the model on the validation set will gradually stabilize. When the evaluation metrics no longer have a significant improvement and reach the preset validation convergence target, the incremental learning process ends. At this time, the obtained GBDT model is the updated classifier, which can more accurately identify newly emerging safety types and safety levels.

[0067] In a possible implementation manner, step S400 further includes:

[0068] Step S410: Locate the dangerous area based on the monitor layout position and the effective monitoring range.

[0069] Step S420: Project the safety recognition result onto the area conduction topology diagram according to the location, and perform adjacent area risk conduction prediction based on the area conduction path and the conduction weight relationship to obtain the area conduction risk.

[0070] Specifically, based on the pre - reasonably planned layout positions of the monitors and the effective coverage range of each monitor, the dangerous area is accurately located. In complex environments such as chemical production workshops, various monitors, such as gas sensors, temperature sensors, pressure sensors, etc., are installed according to the technological process, equipment layout, and the distribution of potential hazardous substances. These monitors collect data in real - time. Once an abnormal data is detected by a certain monitor, such as an excessive concentration of harmful gases, or an abnormal increase in temperature or pressure, the specific area where danger may exist is immediately determined by combining the position of the monitor and its effective monitoring range. For example, if a gas monitor in a certain area detects a toxic gas leak, the approximate area where the leak source is located can be quickly locked based on the position of the monitor and the radius range it can effectively monitor.

[0071] The graph propagation algorithm combined with the Bayesian network idea is used to complete the prediction of risk conduction in adjacent areas. First, the safety recognition results are accurately mapped into the regional conduction topology structure diagram according to their positioning information. This diagram is represented in a graph structure, where nodes represent each functional area, edges represent the connection relationships between areas, and the weights on the edges reflect the possibility and degree of risk conduction. Based on the graph propagation algorithm, risk information starts to spread from the node determined to be dangerous. Initially, the risk value of the dangerous node is set to 100% (which can be quantified according to the actual situation), and the risk values of other nodes are 0. Then, iterative propagation is carried out according to the regional conduction path and conduction weight. In each round of iteration, each node distributes its own risk value to adjacent nodes according to the weight ratio of the edges connected to it. For example, if node A is connected to nodes B and C, the weight of edge A - B is 0.3, the weight of edge A - C is 0.7, and the risk value of node A is RA, then in this round of iteration, the risk increment obtained by node B from node A is RA×0.3, and the risk increment obtained by node C is RA×0.7. At the same time, the idea of the Bayesian network is introduced to consider the conditional probability of risk during the conduction process. For example, some areas may have protective measures or specific technological processes, which reduce the probability of harm after the risk is conducted to this area. Through a pre - set conditional probability table, the risk value during the propagation process is corrected. After each iteration ends, the risk value of each node is updated until the risk values of all nodes no longer change significantly, that is, the convergence state is reached. Finally, the risk value corresponding to each node is the regional conduction risk of the adjacent area. In this way, the risk degree that each adjacent area may face can be predicted more accurately, providing a scientific basis for the safety management of chemical production.

[0072] In a possible implementation manner, step S420 further includes:

[0073] Step S421: According to the layout positions of the monitors and their effective monitoring ranges, conduct cross-thermal distribution analysis of the monitoring data, determine the diffusion distribution characteristics of the risks, and identify the risk center points.

[0074] Step S422: Take the risk center point as the initial node, and in the regional conduction topology structure diagram in the current segmentation state, establish a set of directed propagation paths starting from the initial node.

[0075] Step S423: According to the risk propagation attenuation factor, combined with the edge weights of the regional conduction topology structure diagram, calculate the conduction risk values of the directed propagation paths to obtain the regional conduction risks.

[0076] Specifically, based on the specific layout positions of the monitors in the factory area and their respective effective monitoring ranges, conduct cross-thermal distribution analysis on the multi-source monitoring data collected. Integrate the temperature, pressure, gas concentration and other types of data collected in real time by the sensor network through data fusion technology. Subsequently, based on the geographic information system (GIS), divide the monitoring area into regular grids, and calculate the comprehensive values of various monitoring indicators in each grid. Use the kernel density estimation method to process these values to generate a cross-thermal distribution map, where the depth of the color in the map intuitively reflects the level of danger. Analyze the distribution pattern and gradient change of the colors in the thermal map to determine the diffusion distribution characteristics such as the diffusion direction, speed and influence range of the risks. Finally, with the help of the clustering analysis algorithm, find the area where the values are most concentrated and the highest in the map, and the central position of this area is the risk center point, providing accurate positioning for subsequent risk prevention and control and emergency response.

[0077] After determining the risk center point, take this center point as the key initial node and conduct analysis in the regional conduction topology structure diagram that has been constructed and is in the current segmentation state. The regional conduction topology structure diagram uses nodes to represent different functional areas in chemical production, and the connecting edges between nodes represent the risk conduction relationships existing between regions, and the weights of the edges reflect the possibility and intensity of risk conduction. Starting from the node corresponding to the risk center point, according to the connection relationships of the nodes in the diagram, search along the direction of the edges, and use the depth-first search (DFS) or breadth-first search (BFS) algorithm in graph theory to traverse all reachable nodes, thereby constructing a set of directed propagation paths starting from the initial node. These directed propagation paths clearly show the possible risk conduction from the center point to which adjacent regions and farther regions, providing a structured data basis for subsequent quantitative evaluation of the conduction risks faced by each region, helping to intuitively grasp the potential paths of risk diffusion, so as to formulate targeted prevention and control strategies in advance.

[0078] First, it is clear that the risk propagation attenuation factor reflects the degree of weakening of the risk intensity affected by factors such as distance, obstacles, and protective measures during the propagation process. The edge weight in the regional conduction topology diagram reflects the possibility and difficulty of risk conduction between two regions. For the obtained set of directed propagation paths, starting from the risk center point as the initial node, set its risk value to the highest preset value. Along each path, when the next node receives the risk, first multiply the current node's risk value by the weight of the connection edge between this node and the next node to obtain the preliminary conduction risk value, and then multiply it by the propagation attenuation factor to correct it and calculate the actual conduction risk value of the next node. In this way, the risk value calculation for all nodes on the path is completed. When the risk values of all nodes on all paths are calculated, if there are multiple paths pointing to the same node, the weighted summation method is used to aggregate the risk value of this node. Finally, by synthesizing the final risk values of all nodes, the conduction risk of the entire region is obtained, providing a quantitative basis for the safety management of chemical production.

[0079] Table 1: An example of calculating the conduction risk value of the directed propagation path based on the risk propagation attenuation factor and combining the edge weight of the regional conduction topology diagram to obtain the regional conduction risk value:

[0080] Table 1

[0081]

[0082] Example 2, based on the same inventive concept as the safety monitoring and management method for the chemical product production workshop in the foregoing embodiment, as Figure 2 shown, this application provides a safety monitoring and management system for the chemical product production workshop. The system in the embodiment of this application and the method embodiment are based on the same inventive concept. Among them, the system includes:

[0083] A topology diagram establishment module 10, configured to establish a regional conduction topology diagram by dividing the chemical product production workshop according to regional functions.

[0084] A multi-source detection data acquisition module 20, configured to deploy multiple types of monitors at the nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facility functional areas to acquire multi-source detection data.

[0085] A safety recognition result positioning module 30, configured to train a classifier based on the safety level division rules, and perform safety recognition classification through the classifier according to the multi-source detection data to locate the safety recognition result.

[0086] An early warning level generation module 40, configured to perform regional fixed-point recognition and regional conduction risk prediction according to the regional conduction topology diagram in combination with the safety recognition result positioning, generate an early warning level, and perform hierarchical early warning and emergency management control during the product production process according to the early warning level.

[0087] Furthermore, the system is also used to implement the following functions:

[0088] The multi-type monitors include: combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, and video monitors. The monitor layout nodes in the functional areas include one or more of the combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, and video monitors.

[0089] Furthermore, the system is also used to implement the following functions:

[0090] Analyze the characteristic relationships of each functional area according to the technological process, equipment relevance, spatial position relationship, material attributes, personnel, and energy flow; define each functional area as a node, and the node attributes include node number, function type, and basic information on the material or energy risk level; establish directed connection edges between the nodes according to the technological process, equipment relevance, and spatial position relationship. Each edge includes a source node and a target node, connection type, material hazard attribute, and probability of dangerous substance conduction; calculate the conduction risk weight of each edge according to the material safety level and the probability of dangerous substance conduction in the connection edge, assign a weight to the edge, and establish a weighted directed graph based on the nodes, edges, and weights of the edges to obtain the regional conduction topology diagram.

[0091] Furthermore, the system is also used to implement the following functions:

[0092] According to the real-time process state, dynamically analyze the process reaction risk level, and configure the regional risk adjustment boundary; reconstruct the functional area boundary according to the regional risk adjustment boundary, and adjust the weights of adjacent edges according to the span range of the boundary reconstruction area; perform regional temporal dynamic update on the regional conduction topology diagram according to the adjustment results of the node area reconstruction boundary and the weights of adjacent edges.

[0093] Furthermore, the system is also used to implement the following functions:

[0094] Use historical accident record data and industry standards to set safety grading rules; construct a training data set and a validation data set according to the safety sample characteristics, safety types, and safety level annotations in the historical accident record data; use the training data set and the validation data set to train and validate the convergence of the initial classifier to obtain the classifier, which is used to input monitoring data characteristics and output safety types and safety levels.

[0095] Furthermore, the system is also used to implement the following functions:

[0096] Obtain information on monitored hazards, where the information on monitored hazards is different from the safety types in the historical accident record data; collect hazard characteristics based on the information on monitored hazards and perform safety type and safety level annotation to construct an incremental training set and a validation set; use the incremental training set and the validation set to perform incremental learning on the classifier until the validation convergence target is reached.

[0097] Furthermore, the system is also used to implement the following functions:

[0098] Based on the layout position of the monitors and the effective monitoring range, locate the hazardous area; project the safety recognition results onto the area conduction topology diagram according to the location, and perform adjacent area risk conduction prediction based on the area conduction path and the conduction weight relationship to obtain the area conduction risk.

[0099] Furthermore, the system is also used to implement the following functions:

[0100] According to the layout position of the monitors and the effective monitoring range, perform cross-thermal distribution analysis of the monitoring data to determine the diffusion distribution characteristics of the risk and identify the risk center point; use the risk center point as the initial node, and establish a set of directed propagation paths starting from the initial node in the area conduction topology diagram in the current segmentation state; according to the risk propagation attenuation factor, combine the edge weights of the area conduction topology diagram to calculate the conduction risk value of the directed propagation path to obtain the area conduction risk.

[0101] Embodiment 3. Based on the same inventive concept as the safety monitoring and management method in the chemical product production workshop in the foregoing embodiments, this embodiment provides a computer-readable storage medium, which can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the safety monitoring and management method in the chemical product production workshop in the embodiments of the present application. The processor executes various functional applications and data processing of the computer device by running the software programs, instructions, and modules stored in the memory, that is, implements the above-mentioned safety monitoring and management method for the chemical product production workshop.

[0102] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above description of specific embodiments of this specification is made. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0103] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0104] This specification and the accompanying drawings are merely illustrative of the present application and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the present application. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalent technologies, the present application is intended to include these changes and modifications.

Claims

1. Safety monitoring and management method for chemical product production workshop, characterized in that, Including: Dividing the chemical product production workshop according to regional functions to establish a regional conduction topological structure diagram; Deploying multiple types of monitors at the functional area nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facilities to collect multi-source detection data; Training a classifier based on the safety level division rules, and performing safety identification and classification through the classifier according to the multi-source detection data to locate the safety identification results; According to the regional conduction topological structure diagram, combining the safety identification results for regional fixed-point identification and regional conduction risk prediction, generating a warning level, and performing hierarchical warning and emergency management control during the product production process according to the warning level.

2. The safety monitoring and management method for a chemical product production workshop according to claim 1, wherein, The multiple types of monitors include: combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, and video monitors. The monitor deployment nodes in the functional area include one or more of the combustible gas detectors, toxic gas detectors, temperature sensors, pressure sensors, vibration sensors, and video monitors.

3. The safety monitoring and management method for a chemical product production workshop according to claim 2, characterized in that Establishing a regional conduction topological structure diagram includes: Analyzing the characteristic relationships of each functional area according to the process flow, equipment correlation, spatial position relationship, material attributes, personnel, and energy flow; Defining each functional area as a node, and the node attributes include node number, functional type, and basic information on the material or energy risk level; Establishing directed connection edges between nodes according to the process flow, equipment correlation, and spatial position relationship. Each edge includes a source node and a target node, connection type, material hazard attribute, and hazardous material conduction probability; Calculating the conduction risk weight of each edge according to the material safety level and hazardous material conduction probability in the connection edge, assigning weights to the edges, and establishing a weighted directed graph based on the nodes, edges, and the weights of the edges to obtain the regional conduction topological structure diagram.

4. The safety monitoring and management method for a chemical product production workshop according to claim 3, characterized in that, Establishing a regional conduction topological structure diagram also includes: Dynamically analyzing the process reaction risk level according to the real-time process state and configuring the regional risk adjustment boundary; Reconstructing the functional area boundary according to the regional risk adjustment boundary and adjusting the weights of adjacent edges according to the boundary reconstruction area span range; Performing regional temporal dynamic update on the regional conduction topological structure diagram according to the adjustment results of the node area reconstruction boundary and the weights of adjacent edges.

5. The safety monitoring and management method for a chemical product production workshop according to claim 2, characterized in that Training a classifier based on the safety level division rules includes: Using historical accident record data and industry standards to set safety classification rules; Constructing a training data set and a validation data set according to the safety sample characteristics, safety types, and safety level annotations in the historical accident record data; Training and validating the convergence of the initial classifier using the training data set and the validation data set to obtain the classifier, which is used to input the monitoring data characteristics and output the safety type and safety level.

6. The safety monitoring and management method for a chemical product production workshop according to claim 5, characterized in that, Constructing a training data set and a validation data set includes: Obtaining monitoring hazardous substance information, which is different from the safety types in the historical accident record data; Collecting hazard characteristics based on the monitoring hazardous substance information and performing safety type and safety level annotations to construct an incremental training set and a validation set. Perform incremental learning on the classifier using the incremental training set and the validation set until the validation convergence target is reached.

7. The safety monitoring and management method for a chemical product production workshop according to claim 3, characterized in that, Generate a warning level, and perform hierarchical warning and emergency management control during the product production process according to the warning level, including: Locate the safety risk area based on the monitor layout position and the effective monitoring range; Project the safety identification results onto the regional conduction topology diagram according to the located safety risk areas, and perform adjacent area risk conduction prediction based on the regional conduction path and the conduction weight relationship to obtain the regional conduction risk; Perform a joint determination based on the fixed-point safety identification risk and the regional conduction risk of the safety risk area to determine the warning level, where the warning level is the cumulative result of the fixed-point warning level and the conduction risk area; Based on the fixed-point safety identification risk and the regional conduction risk, for the product production and processing area, with the goal of maximizing the response effect and minimizing the impact on the product production process, perform dynamic product time series management control and diffusion time series response control according to the risk identification results to ensure that each functional area meets the preset safety target.

8. The safety monitoring and management method for a chemical product production workshop according to claim 7, characterized in that, Obtaining the regional conduction risk further includes: Perform cross-thermal distribution analysis of the monitoring data according to the monitor layout position and the effective monitoring range to determine the diffusion distribution characteristics of the risk and identify the risk center point; Using the risk center point as the initial node, establish a set of directed propagation paths starting from the initial node in the regional conduction topology diagram in the current segmentation state; According to the risk propagation attenuation factor, combined with the edge weights of the regional conduction topology diagram, calculate the conduction risk value for the directed propagation paths to obtain the regional conduction risk.

9. Safety monitoring and management system for chemical product production workshop, characterized in that, The system is used to implement the safety monitoring and management method for a chemical product production workshop according to any one of claims 1-8, and the system includes: A topology diagram establishment module for establishing a regional conduction topology diagram by dividing the chemical product production workshop according to regional functions; A multi-source detection data acquisition module for arranging multiple types of monitors at the functional area nodes of the reaction unit, raw material storage area, high-temperature and high-pressure equipment area, and power auxiliary facilities to collect multi-source detection data; A safety identification result positioning module for training a classifier based on the safety level division rules, and performing safety identification classification through the classifier according to the multi-source detection data to locate the safety identification results; A warning level generation module for generating a warning level according to the regional conduction topology diagram, combining the safety identification result positioning for regional fixed-point identification and regional conduction risk prediction, and performing hierarchical warning and emergency management control during the product production process according to the warning level.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the safety monitoring and management method for a chemical product production workshop according to any one of claims 1-8.

Images