Safety management method and system based on material processing industry

Abstract

According to the safety management method and system based on the material processing industry, various data on a production line are monitored in real time, and once any data is found to be abnormal, a joint response mechanism, namely a linkage mechanism, of related data is started immediately according to the property of the abnormal data; according to various linkage mechanisms triggered by abnormal data, corresponding emergency processing schemes are formulated, through comprehensive and systematic safety monitoring, the information island phenomenon is eliminated, safety risks in the production process are effectively recognized and prevented, a solid guarantee is provided for safety supervision and technological process control of factories, and the safety of the factories is improved. The method has important practical value for reducing the requirement of manual inspection, reducing production safety accidents, guaranteeing the safety of equipment and materials and improving the efficiency of overall safety management.

Description

Technical Field

[0001] The present application belongs to the field of safe production technology, and in particular, relates to a safety management method and system based on the material processing industry. Background Art

[0002] In the material processing industry, with the continuous advancement of technology, the scale of production is constantly expanding, the production process is becoming increasingly automated, and the control of production equipment is becoming more sophisticated and complex. In this context, the demand for safety monitoring is also growing. How to enhance the prediction and prevention of potential safety risks, how to quickly identify and solve safety problems in the production process, has become particularly critical. By adopting intelligent monitoring, analysis and management methods, the scientific and technological content and response speed of safety monitoring can be improved, casualties and property losses can be minimized, and the smooth operation of the production process and effective traceability afterwards can be ensured.

[0003] The existing safety monitoring technologies in the material processing industry are relatively simple. These technologies play an important role in improving the safety monitoring level of the material processing industry, but they also have some limitations and challenges.

[0004] 1. Video surveillance system: provides real-time visual monitoring, which helps to detect safety hazards and accidents in a timely manner.

[0005] Disadvantages: Depends on the quality and quantity of cameras, there may be blind spots in monitoring. In addition, the amount of video data is huge, requiring effective data storage and analysis technology.

[0006] 2. Sensors and IoT technology: Monitor the environment and equipment status through various sensors (such as temperature, pressure, and gas sensors) to achieve real-time data collection and analysis.

[0007] Disadvantages: The sensor may be damaged or fail due to environmental factors (such as dust, humidity), and requires regular maintenance and calibration.

[0008] 3. Data analysis and machine learning: Use machine learning algorithms to analyze the collected data and predict potential security issues.

[0009] Disadvantages: A large amount of high-quality data is required for training, and the accuracy of the algorithm is limited by the quality and diversity of the data.

[0010] 4. Automated control system: Automated control can reduce human errors and improve the consistency and safety of operations.

[0011] Disadvantages: The system may fail due to programming errors or hardware failures, requiring complex fault diagnosis and recovery mechanisms.

[0012] 5. Emergency response and dispatch system: In the event of an emergency, it can respond and dispatch resources quickly to reduce the impact of the accident.

[0013] Disadvantages: Depends on the reliability of the system and the training level of the personnel, and may not respond quickly enough in extreme situations.

[0014] 6. Door access and access control systems: Control access to sensitive areas, prevent unauthorized personnel from entering, and reduce security risks.

[0015] Disadvantages: May fail due to technical failure or human negligence, requires regular security audits and maintenance.

[0016] 7. Environmental and energy management system: monitors energy usage and environmental conditions, contributing to energy conservation, emission reduction and environmental safety.

[0017] Disadvantages: Requires high-precision measuring equipment and complex data analysis, and there may be errors in data interpretation.

[0018] 8. Mobile monitoring and notification system: allows managers to remotely monitor and respond to abnormal events through mobile devices.

[0019] Disadvantages: Depends on the stability and coverage of the mobile network, there may be communication delays or interruptions.

[0020] In summary, although the safety monitoring system in the material processing industry is well-equipped, most systems are independent and lack effective integration and coordination, resulting in the dispersion of safety management information and the identification of risk factors is not accurate and rapid enough, which not only affects the overall control of safety conditions, but also fails to make full use of existing technical equipment. In addition, the traditional manual safety management method has many shortcomings and can no longer meet the standards of modern information management. Therefore, there is an urgent need for an innovative intelligent safety monitoring strategy to integrate various safety factors in the production process, achieve rapid response and linkage processing of algorithm warnings, use intelligent technology to reduce human resource costs, and optimize personnel allocation, so as to achieve the goals of reducing enterprise costs, improving efficiency, and realizing refined safety management. Summary of the invention

[0021] The purpose of this application is to solve the problems of the prior art and to provide a safety management method and system based on the material processing industry.

[0022] In order to solve the technical problem, the technical solution of this application is: a safety management method based on the material processing industry, comprising the following steps: Step S101: Collect key data; key data includes equipment data, video surveillance data, production data, environmental and health and safety data, access control data, and safety warning data; Step S102: pre-processing and classifying the collected key data to obtain the warning thresholds of various parameters under different targets; Step S103: Algorithm warning; establishing a linkage model to predict potential safety risks of the production line through the linkage model; Step S1031: Establishing a linkage model: Using a machine learning algorithm to establish a random forest model to learn the linkage relationship between the preprocessed and classified key data; Step S1032: Prediction model training: Based on the importance calculation of the random forest model, a support vector machine (SVM) is used to train the prediction model; Step S1033: Model tuning: adjusting the parameters of the SVM and using cross-validation and grid search techniques to find the optimal parameter combination; Step S1034: Model evaluation: Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, precision, recall, F1 score and ROC-AUC value. Evaluate the performance of the model and optimize it to obtain a trained linkage model, and deploy the trained linkage model to the production environment; Step S104: linkage mechanism; real-time monitoring and analysis of data on the production line; if the real-time data exceeds the warning threshold of step S102, i.e. abnormal data, the alarm system is used to determine the severity and start an alarm; at the same time, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model, and a linkage response is performed; after each linkage strategy, step S103 is repeated again; Step 105: emergency plan; generating and executing a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data; Step 106: Record and analyze detailed information of abnormal events, and perform maintenance and upgrades regularly.

[0023] Preferably, the preprocessing of key data in step S102 includes unifying frequencies and supplementing missing values; The unified frequency specifically includes: when processing different data, using Fourier transform to convert the time series data into the frequency domain and analyzing its frequency components; The Fourier transform formula is:

[0024] in: N is the number of sample points, X[k] is the kth sample in the frequency domain; x[n] is the nth sample in the time domain; The missing value supplementation includes mean filling and decision tree model; When the data is relatively stable, use the mean to fill in missing values; When the data is relatively discrete and irregular, a decision tree model is used to predict missing values. For regression relationships, the prediction missing value formula of the decision tree model is:

[0025] in: It is the data set contained in a leaf node in the decision tree model; y(x) is the target value of data point x; For classification relationships, the prediction missing value formula of the decision tree model is:

[0026] in:

[0027] Is a leaf node The majority of .

[0028] Preferably, the classification of key data in step S102 is as follows: the preprocessed key data is classified according to the data type, and multi-objective optimization is performed by comparing the range of each data when the output is high and the restrictions of the prerequisites to obtain the normal range of each parameter under different objectives, thereby obtaining the warning threshold of each parameter under different objectives.

[0029] Preferably, the step S1031 uses a machine learning algorithm to establish a random forest model as follows: Step S10311: For each tree, calculate the reduction in impurity caused by each feature when the node is split, and the impurity is represented by the Gini index; Step S10312: For node m in tree i, importance It is expressed as:

[0030] in:

[0031] ; Step S10313: Accumulate the contributions of all nodes in each tree to obtain the feature Importance of Tree I:

[0032] Step S10334: The overall importance score of each feature in the random forest is the average of the feature value scores in all trees, where n is the number of trees in the random forest:

[0033] Preferably, the prediction model training in step S1032 is specifically as follows: Step S10321: Data labeling, labeling the data, determining the association rules for each data point, or determining whether each data point is associated with an abnormal event, and calculating the association based on the importance of the random forest model; the association rules are pre-set according to safety specifications and operation manuals, and are used to reveal the mutual influence between different data; Step S10322: Data segmentation: split the data set into a training set and a test set according to an 8:2 ratio; Step S10323: Select model: Select SVM algorithm as the prediction model; first construct the SVM function, select RBF as the kernel function, then introduce the Lagrange multiplier method, find the optimal Lagrange multiplier by solving the dual problem, and then use the decision function for classification to complete the SVM algorithm; Step S10324: Model training: Use the training set data to train the SVM algorithm to obtain a prediction model.

[0034] Preferably, adjusting the parameters of the SVM in step S1033 specifically includes: adjusting the regularization parameter C, the kernel function and the parameters of the kernel function, and using cross-validation and grid search techniques to find the optimal parameter combination.

[0035] Preferably, the alarm of the alarm system in step S104 is divided into three levels: low, medium and high, wherein a value exceeding the warning threshold by 10% is a low-level alarm, a value exceeding the warning threshold by 25% is a medium-level alarm, and a value exceeding the warning threshold by 26% or more is a high-level alarm.

[0036] Preferably, in step S104, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model. Specifically, the production line data is firstly detected for abnormality. Once the data exceeds a preset threshold, it is marked as abnormal data. Then, the association rules are retrieved according to each abnormal data. When multiple data points are abnormal at the same time, the system will evaluate the correlation between them, and only when at least one associated data is not affected, the linkage strategy for non-abnormal associated data will be triggered. The association rules are pre-set according to safety specifications and operating manuals, or obtained according to the linkage model.

[0037] Preferably, a safety management system based on the material processing industry includes a data acquisition module, a data preprocessing and classification, a real-time monitoring and anomaly detection module, a linkage response module, an emergency dispatch module, a video monitoring and viewing angle adjustment module, a door access control and safety warning module, an energy and environment management module, and a mobile response and notification module, wherein the data acquisition module is respectively connected to the data preprocessing and classification module, the real-time monitoring and anomaly detection module, the video monitoring and viewing angle adjustment module, the door access control and safety warning module, the energy and environment management module, and the mobile response and notification module, the real-time monitoring and anomaly detection module is respectively connected to the linkage response module and the emergency dispatch module, and the linkage response module is respectively connected to the emergency dispatch module, the video monitoring and viewing angle adjustment module, the door access control and safety warning module, and the energy and environment management module; The data acquisition module is used to collect key data on the production line; the key data includes equipment data, production data, video monitoring data of the video monitoring and viewing angle adjustment module, environmental and health safety data of the energy and environmental management module, access control data of the access control and safety warning module, and safety warning data of the real-time monitoring and anomaly detection module; The data preprocessing and classification is used to preprocess and classify the collected key data, obtain the warning thresholds of various parameters under different targets, and conduct in-depth analysis and knowledge graph construction; The real-time monitoring and anomaly detection module is used to continuously monitor the data flow on the production line, establish a linkage model, and predict the potential safety risks of the production line through the linkage model; The linkage response module is used to implement the linkage mechanism, monitor and analyze the data on the production line in real time. If the real-time data exceeds the warning threshold of step S102, that is, abnormal data, the alarm system is used to determine the severity and start the alarm; at the same time, according to the type and severity of the abnormal data, the trained linkage model is used to start the linkage strategy and perform the linkage response; The emergency dispatch module is used to generate and execute corresponding emergency plans according to the specific circumstances of abnormal events generated by abnormal data; The video monitoring and viewing angle adjustment module is used to automatically adjust the viewing angle of the surveillance camera through the video monitoring system when the real-time monitoring and anomaly detection module detects an anomaly in a specific area; The access control and security warning module is used to automatically switch the access control system to manual control mode and notify relevant personnel when the real-time monitoring and anomaly detection module detects an anomaly in a specific area; The energy and environment management module uses the energy and environment system to monitor energy usage and environmental conditions in real time; The mobile response and notification module is used to send real-time safety alerts and operation instructions to staff via mobile devices.

[0038] Preferably, the equipment data is generated by industrial equipment sensors and actuators, including temperature, pressure and vibration parameters, and the equipment data is transmitted to the data acquisition module through a gateway or a direct Ethernet connection using Modbus, OPC UA, Profibus industrial communication protocols; The video surveillance data is captured by a surveillance camera and transmitted to a data acquisition module via RTSP and ONVIF protocols; The production data includes output, efficiency, and quality control indicators, which are generated by the production management system and transmitted to the data acquisition module using an API interface or database synchronization technology; The environmental and health safety data is obtained by the energy and environment management module; The access control data is obtained by the access control and security warning module; The safety warning data comes from safety-related sensors and monitoring systems, and is transmitted to the data acquisition module in real time via the industrial field bus.

[0039] Compared with the prior art, the advantages of this application are: (1) This application monitors various data on the production line in real time. Once any data is found to be abnormal, the joint response mechanism of the relevant data, i.e., the linkage mechanism, is immediately initiated according to the nature of the abnormal data; corresponding emergency response plans are formulated based on various linkage mechanisms triggered by abnormal data. Through comprehensive and systematic safety monitoring, not only the information island phenomenon is eliminated, but also safety risks in the production process are effectively identified and prevented, providing a solid guarantee for the factory's safety supervision and process control. This has important practical value in reducing the need for manual inspections, reducing the occurrence of production safety accidents, ensuring the safety of equipment and materials, and improving the efficiency of overall safety management; (2) The system of this application includes the integrated application of various modules. With their support, it not only improves the intelligent level of safety monitoring in the material processing industry, but also greatly enhances the prevention, response and processing capabilities of production safety, providing a strong guarantee for the stable operation of the enterprise and the safety of employees; (3) The present application executes corresponding linkage strategies based on abnormal data and non-abnormal associated data, which not only helps to diagnose problems and determine the source of faults, but also improves the efficiency of linkage response by independently processing abnormal data and linking non-abnormal associated data, ensuring that emergency situations on the production line can be responded to quickly and accurately. This method enables the system to handle data anomalies on the production line more intelligently and achieve rapid and effective linkage response, thereby significantly improving the safety and reliability of the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0041] Figure 1 A flowchart of a safety management method based on the material processing industry is provided for this application; Figure 2 This application provides a structural block diagram of a safety management system based on the material processing industry.

[0042] Description of reference numerals: 1. Data acquisition module, 2. Data preprocessing and classification, 3. Real-time monitoring and anomaly detection module, 4. Linkage response module, 5. Emergency dispatch module, 6. Video monitoring and viewing angle adjustment module, 7. Access control and safety warning module, 8. Energy and environment management module, 9. Mobile response and notification module, 10. Video monitoring system, 11. Access control system, 12. Energy and environment system. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0045] This application proposes a safety management method and system based on the material processing industry, which covers the following key steps: collecting video surveillance information, process parameters, environmental safety data, energy consumption data, access control records and safety warning signals in the production process; real-time monitoring of various data on the production line, and once any data is found to be abnormal, immediately initiating a joint response mechanism for related data based on the nature of the abnormal data; formulating corresponding emergency response plans based on various joint responses triggered by abnormal data, and through comprehensive and systematic safety monitoring, not only the information island phenomenon is eliminated, but also the safety risks in the production process are effectively identified and prevented, providing a solid guarantee for the factory's safety supervision and process flow control, which has important practical value in reducing the need for manual inspections, reducing the occurrence of production safety accidents, ensuring the safety of equipment and materials, and improving the efficiency of overall safety management.

[0046] The present application discloses a safety management method based on the material processing industry, comprising the following steps: Step S101: Collect key data; key data includes equipment data, video surveillance data, production data, environmental and health and safety data, access control data, and safety warning data; Step S102: pre-processing and classifying the collected key data to obtain the warning thresholds of various parameters under different targets; Step S103: Algorithm warning; establishing a linkage model to predict potential safety risks of the production line through the linkage model; Step S1031: Establishing a linkage model: Using a machine learning algorithm to establish a random forest model to learn the linkage relationship between the preprocessed and classified key data; Step S1032: Prediction model training: Based on the importance calculation of the random forest model, a support vector machine (SVM) is used to train the prediction model; Step S1033: Model tuning: adjusting the parameters of the SVM and using cross-validation and grid search techniques to find the optimal parameter combination; Step S1034: Model evaluation: Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, precision, recall, F1 score and ROC-AUC value. Evaluate the performance of the model and optimize it to obtain a trained linkage model, and deploy the trained linkage model to the production environment; Step S104: linkage mechanism; real-time monitoring and analysis of data on the production line; if the real-time data exceeds the warning threshold of step S102, i.e. abnormal data, the alarm system is used to determine the severity and start an alarm; at the same time, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model, and a linkage response is performed; after each linkage strategy, step S103 is repeated again; Step 105: emergency plan; generating and executing a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data; Step 106: Record and analyze detailed information of abnormal events, and perform maintenance and upgrades regularly.

[0047] Preferably, the preprocessing of key data in step S102 includes unifying frequencies and supplementing missing values; The unified frequency specifically includes: when processing different data, using Fourier transform to convert the time series data into the frequency domain and analyzing its frequency components; The Fourier transform formula is:

[0048] in: N is the number of sample points, X[k] is the kth sample in the frequency domain; x[n] is the nth sample in the time domain; The missing value supplementation includes mean filling and decision tree model; When the data is relatively stable, use the mean to fill in missing values; When the data is relatively discrete and irregular, a decision tree model is used to predict missing values. For regression relationships, the prediction missing value formula of the decision tree model is:

[0049] in: It is the data set contained in a leaf node in the decision tree model; y(x) is the target value of data point x; For classification relationships, the prediction missing value formula of the decision tree model is:

[0050] in:

[0051] Is a leaf node The majority of .

[0052] Preferably, the classification of key data in step S102 is as follows: the preprocessed key data is classified according to the data type, and multi-objective optimization is performed by comparing the range of each data when the output is high and the restrictions of the prerequisites to obtain the normal range of each parameter under different objectives, thereby obtaining the warning threshold of each parameter under different objectives.

[0053] Preferably, the step S1031 uses a machine learning algorithm to establish a random forest model as follows: Step S10311: For each tree, calculate the reduction in impurity caused by each feature when the node is split, and the impurity is represented by the Gini index; Step S10312: For node m in tree i, importance It is expressed as:

[0054] in:

[0055] ; Step S10313: Accumulate the contributions of all nodes in each tree to obtain the feature Importance of Tree I:

[0057] Step S10334: The overall importance score of each feature in the random forest is the average of the feature value scores in all trees, where n is the number of trees in the random forest:

[0058] Preferably, the prediction model training in step S1032 is specifically as follows: Step S10321: Data labeling, labeling the data, determining the association rules for each data point, or determining whether each data point is associated with an abnormal event, and calculating the association based on the importance of the random forest model; the association rules are pre-set according to safety specifications and operation manuals, and are used to reveal the mutual influence between different data; Step S10322: Data segmentation: split the data set into a training set and a test set according to an 8:2 ratio; Step S10323: Select model: Select SVM algorithm as the prediction model; first construct the SVM function, select RBF as the kernel function, then introduce the Lagrange multiplier method, find the optimal Lagrange multiplier by solving the dual problem, and then use the decision function for classification to complete the SVM algorithm; Step S10324: Model training: Use the training set data to train the SVM algorithm to obtain a prediction model.

[0059] Preferably, adjusting the parameters of the SVM in step S1033 specifically includes: adjusting the regularization parameter C, the kernel function and the parameters of the kernel function, and using cross-validation and grid search techniques to find the optimal parameter combination.

[0060] The cross validation and grid search techniques are existing methods and will not be described in detail in this application.

[0061] Preferably, the alarm of the alarm system in step S104 is divided into three levels: low, medium and high, wherein a value exceeding the warning threshold by 10% is a low-level alarm, a value exceeding the warning threshold by 25% is a medium-level alarm, and a value exceeding the warning threshold by 26% or more is a high-level alarm.

[0062] Preferably, in step S104, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model. Specifically, the production line data is firstly detected for abnormality. Once the data exceeds a preset threshold, it is marked as abnormal data. Then, the association rules are retrieved according to each abnormal data. When multiple data points are abnormal at the same time, the system will evaluate the correlation between them, and only when at least one associated data is not affected, the linkage strategy for non-abnormal associated data will be triggered. The association rules are pre-set according to safety specifications and operating manuals, or obtained according to the linkage model.

[0063] Preferably, a safety management system based on the material processing industry comprises a data acquisition module 1, a data preprocessing and classification module 2, a real-time monitoring and anomaly detection module 3, a linkage response module 4, an emergency dispatch module 5, a video monitoring and viewing angle adjustment module 6, an access control and safety warning module 7, an energy and environment management module 8 and a mobile response and notification module 9, wherein the data acquisition module 1 is respectively connected to the data preprocessing and classification module 2, the real-time monitoring and anomaly detection module 3, the video monitoring and viewing angle adjustment module 6, the access control and safety warning module 7, the energy and environment management module 8 and the mobile response and notification module 9, the real-time monitoring and anomaly detection module 3 is respectively connected to the linkage response module 4 and the emergency dispatch module 5, and the linkage response module 4 is respectively connected to the emergency dispatch module 5, the video monitoring and viewing angle adjustment module 6, the access control and safety warning module 7 and the energy and environment management module 8; The data acquisition module 1 is used to collect key data on the production line; the key data includes equipment data, production data, video monitoring data of the video monitoring and viewing angle adjustment module 6, environmental and health safety data of the energy and environmental management module 8, access control data of the access control and safety warning module 7, and safety warning data of the real-time monitoring and abnormality detection module 3; The data preprocessing and classification 2 is used to preprocess and classify the collected key data, obtain the warning thresholds of various parameters under different targets, and conduct in-depth analysis and knowledge graph construction; The real-time monitoring and anomaly detection module 3 is used to continuously monitor the data flow on the production line, establish a linkage model, and predict the potential safety risks of the production line through the linkage model; The linkage response module 4 is used to implement the linkage mechanism, monitor and analyze the data on the production line in real time. If the real-time data exceeds the warning threshold of step S102, that is, abnormal data, the alarm system is used to determine the severity and start the alarm; at the same time, according to the type and severity of the abnormal data, the trained linkage model is used to start the linkage strategy and perform the linkage response; The emergency dispatch module 5 is used to generate and execute corresponding emergency plans according to the specific circumstances of the abnormal events generated by the abnormal data; The video monitoring and viewing angle adjustment module 6 is used to automatically adjust the viewing angle of the surveillance camera through the video monitoring system 10 when the real-time monitoring and abnormality detection module 3 detects an abnormality in a specific area; The access control and security warning module 7 is used to automatically switch the access control system 11 to the manual control mode and notify relevant personnel when the real-time monitoring and abnormality detection module 3 detects an abnormality in a specific area; The energy and environment management module 8 uses the energy and environment system 12 to monitor energy usage and environmental conditions in real time; The mobile response and notification module 9 is used to send real-time safety alerts and operation instructions to staff via mobile devices.

[0064] Preferably, the equipment data is generated by industrial equipment sensors and actuators, including temperature, pressure and vibration parameters, and the equipment data is transmitted to the data acquisition module 1 through a gateway or a direct Ethernet connection using Modbus, OPC UA, Profibus industrial communication protocols; The video surveillance data is captured by a surveillance camera and transmitted to the data acquisition module 1 via RTSP and ONVIF protocols; The production data includes output, efficiency, and quality control indicators, which are generated by the production management system and transmitted to the data acquisition module 1 using an API interface or database synchronization technology; The environmental and health safety data is obtained by the energy and environment management module 8; The access control data is obtained by the access control and security warning module 7; The safety warning data comes from safety-related sensors and monitoring systems, and is transmitted to the data acquisition module 1 in real time via the industrial field bus.

[0065] The linkage response module 4 and the emergency dispatch module 5 form an efficient security incident handling process through close collaboration, ensuring that prompt action can be taken to minimize losses and impacts when facing various potential security risks and emergencies. The connection between the two modules is reflected in information sharing on the one hand, that is, after detecting an abnormality and executing preliminary response measures, the linkage response module 4 will pass the detailed information of the event to the emergency dispatch module 5 for further emergency processing; on the other hand, it is also reflected in the coordinated response. The emergency dispatch module 5 will adjust or supplement the emergency measures according to the preliminary processing results of the linkage response module 4 to ensure that the event is fully and effectively controlled. In addition, after executing the emergency measures, the emergency dispatch module 5 will feed back the processing results and subsequent monitoring data to the linkage response module 4 so that the system can continuously monitor and evaluate. Finally, the collaborative work results of the linkage response module 4 and the emergency dispatch module 5 can be used to evaluate the effectiveness of the existing emergency plan, and update and optimize the plan according to the actual response situation.

[0066] Example 1 like Figure 1 As shown, the present invention discloses a safety management method based on the material processing industry, specifically: Step S101: Collect key data on the production line, including equipment data, video surveillance data, production data, environmental and health and safety data, access control data, and safety warning data; Step S102: pre-processing and classifying the collected data, and obtaining the normal range of each parameter under different objectives through multi-objective optimization, and obtaining the warning threshold at the same time; The data were preprocessed to unify the frequency and supplement missing values; Unified frequency: When processing different data, use Fourier transform to convert time series data into frequency domain and analyze its frequency components; Step S103: Algorithm warning to predict potential safety risks of the production line; Step S104: Real-time monitoring and analysis of data on the production line, setting warning thresholds based on normal ranges of various parameters under different targets. If real-time data exceeds the warning threshold, the system starts to alarm and initiates linkage strategies; Step 105: Generate and execute a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data, that is, control different other subsystems through linkage with different severity levels; Step 106: Record and analyze detailed information of abnormal events, and perform maintenance and upgrades regularly.

[0067] Example 2 like Figure 2As shown, the present invention discloses a safety management system based on the material processing industry, including a data acquisition module 1, a data preprocessing and classification module 2, a real-time monitoring and anomaly detection module 3, a linkage response module 4, an emergency dispatch module 5, a video monitoring and viewing angle adjustment module 6, an access control and safety warning module 7, an energy and environment management module 8 and a mobile response and notification module 9, wherein the data acquisition module 1 is respectively connected to the data preprocessing and classification module 2, the real-time monitoring and anomaly detection module 3, the video monitoring and viewing angle adjustment module 6, the access control and safety warning module 7, the energy and environment management module 8 and the mobile response and notification module 9, the real-time monitoring and anomaly detection module 3 is respectively connected to the linkage response module 4 and the emergency dispatch module 5, and the linkage response module 4 is respectively connected to the emergency dispatch module 5, the video monitoring and viewing angle adjustment module 6, the access control and safety warning module 7 and the energy and environment management module 8; The data acquisition module 1 is used to collect key data on the production line; the key data includes equipment data, production data, video monitoring data of the video monitoring and viewing angle adjustment module 6, environmental and health safety data of the energy and environment management module 8, access control data of the access control and safety warning module 7, and safety warning data of the real-time monitoring and abnormality detection module 3; data acquisition module: responsible for collecting key data on the production line, including video monitoring, production parameters, fire protection system status, energy and environment monitoring, access control records and safety warning signals. This module classifies the data by type and performs preliminary processing.

[0068] The data preprocessing and classification 2 is used to deeply analyze the key data collected by the data acquisition module 1 and construct a knowledge graph, deeply analyze the collected video, production, fire protection, energy environment, access control and security warning data, extract useful information, and construct a knowledge graph based on the correlation and attributes between the data, so as to better understand and respond to security incidents, and perform preprocessing and classification; The real-time monitoring and anomaly detection module 3 is used to continuously monitor the data flow on the production line, establish a linkage model, and predict the potential safety risks of the production line through the linkage model; The linkage response module 4 is used to implement the linkage mechanism, monitor and analyze the data on the production line in real time. If the real-time data exceeds the warning threshold of step S102, that is, abnormal data, the alarm system is used to determine the severity and start the alarm; at the same time, according to the type and severity of the abnormal data, the trained linkage model is used to start the linkage strategy and perform the linkage response; The emergency dispatch module 5 is used to generate and execute corresponding emergency plans according to the specific circumstances of the abnormal events generated by the abnormal data; The video monitoring and viewing angle adjustment module 6 is used to automatically adjust the viewing angle of the surveillance camera through the video monitoring system 10 when the real-time monitoring and abnormality detection module 3 detects an abnormality in a specific area; The access control and security warning module 7 is used to automatically switch the access control system 11 to the manual control mode and notify relevant personnel when the real-time monitoring and abnormality detection module 3 detects an abnormality in a specific area; The energy and environment management module 8 uses the energy and environment system 12 to monitor energy usage and environmental conditions in real time; The mobile response and notification module 9 is used to send real-time safety alerts and operation instructions to staff via mobile devices.

[0069] Data flow description: The data acquisition module 1 passes the data to the data preprocessing and classification module 2, the real-time monitoring and anomaly detection module 3, the video monitoring and viewing angle adjustment module 6, the access control and safety warning module 7, the energy and environment management module 8 and the mobile response and notification module 9.

[0070] The data preprocessing and classification 2 passes the parsed data to the real-time monitoring and anomaly detection module 3.

[0071] After detecting an abnormality, the real-time monitoring and abnormality detection module 3 transmits the information to the linkage response module 4 and the emergency dispatch module 5 .

[0072] The linkage response module 4 calls the access control and safety warning module 7, the energy and environment management module 8 and the video monitoring and viewing angle adjustment module 6 according to the abnormal situation.

[0073] The emergency dispatch module 5 generates and executes an emergency plan and notifies the operator at the same time.

[0074] Example 3 The specific steps of this application are as follows: Step S101, collecting key data on the production line, including but not limited to: When building a comprehensive security management platform, it is crucial to be able to efficiently access and manage various types of data. The following is a description of the data access protocol for six different types of data: 1. Device data access agreement: Equipment data is usually generated by industrial equipment sensors and actuators, including parameters such as temperature, pressure, and vibration.

[0075] Adopt industrial communication protocols such as Modbus, OPC UA, Profibus, etc., and transmit data to the data acquisition module 1 through a gateway or direct Ethernet connection. For devices that support OPC UA, take advantage of its security and cross-platform features to achieve secure data collection and real-time monitoring.

[0076] 2. Video surveillance data access protocol: Video surveillance data is captured by surveillance cameras and transmitted via protocols such as RTSP and ONVIF.

[0077] For devices that comply with the GB28181 national standard, registration and access are performed through fixed IP or dynamic DNS.

[0078] For ONVIF-compatible cameras, use the ONVIF protocol to acquire and control media streams, and transcode them when necessary to adapt to the platform standard.

[0079] 3. Production data access agreement: Production data includes output, efficiency, quality control indicators, etc., and is usually generated by production management systems (such as MES).

[0080] Use API interface or database synchronization technology to achieve real-time access and update of production data.

[0081] 4. Environmental and Health Safety (EHS) Data Access Agreement: EHS data covers environmental monitoring and employee health data, such as dust, noise, radiation levels, etc.

[0082] The sensor data is sent to the cloud platform or data acquisition module 1 through the Internet of Things (IoT) protocol such as MQTT, CoAP, etc.

[0083] For environmental alarm data that requires high real-time performance, real-time communication protocols such as WebSocket are used to push data.

[0084] 5. Access control data access agreement: Access control data includes the access control system’s card swiping records, door lock status, alarm information, etc.

[0085] Utilize RFID, NFC, biometrics and other technologies, combined with the access control system's API or database interface for data access.

[0086] For access control systems that require high security, SSL / TLS encryption protocol is used to ensure the security of data transmission.

[0087] 6. Security Warning Data Access Agreement: Safety warning data comes from various safety-related sensors and monitoring systems, such as fire alarms, emergency stop buttons, etc.

[0088] The warning information is transmitted to the control center in real time via industrial fieldbus such as HART, EtherCAT or dedicated safety network protocol.

[0089] For rapid response in emergency situations, edge computing technology is used for local processing and critical information is quickly reported via 5G or dedicated wireless networks.

[0090] This application utilizes the video surveillance, production data, fire data, etc. collected by the data acquisition module, and integrates this information through data fusion technology to reduce monitoring blind spots.

[0091] Step S102, preprocess and classify the collected data for subsequent analysis and application. Data preprocessing is mainly to unify the frequency and supplement missing values. When processing data with different frequencies, we used Fourier Transform to convert time series data into the frequency domain and analyze its frequency components. The advantage of Fourier Transform is that it can process data with different frequencies in the frequency domain, making it easier to identify and filter signals of specific frequencies. The specific formula is as follows:

[0092] Where N is the number of sample points, X[k] is the kth sample in the frequency domain, and x[n] is the nth sample in the time domain; There are two situations when dealing with missing values. For relatively stable data, use mean filling; for relatively discrete and irregular data, use a decision tree model to predict the size of the missing value.

[0093] For regression relationships, the prediction missing value formula of the decision tree is:

[0094] in is the data set contained in a leaf node in the decision tree, and y(x) is the target value of data point x; For classification relationships, the prediction missing value formula of the decision tree is:

[0095] in Is a leaf node The majority of .

[0096] In the steps, firstly, the data of each submodule mentioned above are classified according to the data type; then the video is analyzed according to different data types, and multi-objective optimization is performed by comparing the range of each feature when the output is high and the restrictions of the prerequisites, so as to obtain the normal range of each parameter under different targets and obtain the warning threshold; then, once a parameter exceeds the normal range based on certain prerequisite targets during work, an alarm is issued.

[0097] Step S103, the system monitors and analyzes the data on the production line in real time to ensure the stability of the production process.

[0098] Step S1031: Establishing a linkage model: Using a machine learning algorithm to establish a random forest model to learn the linkage relationship between the preprocessed and classified key data; Random Forest: Build a random forest model to improve the accuracy and robustness of predictions by integrating multiple decision trees. During the training process, the random forest uses the replacement sampling method to select samples from the training data set and randomly selects features for each tree to split; when calculating feature values, the "mean impurity reduction (Mean DecreaseImpurity)" method is used. The specific steps are as follows: For each tree, the reduction in impurity caused by each feature when the node is split is calculated, and the impurity is expressed as the Gini index; For node m in tree i, the importance can be expressed as:

[0099] ; The contribution of all nodes in each tree is accumulated to obtain the feature Importance of Tree I:

[0100] Finally, the overall importance score of each feature in the random forest is the average of the scores of that feature across all trees, where n is the number of trees in the random forest:

[0101] Step S1032: Prediction model training: Based on the importance calculation of the random forest model, a support vector machine (SVM) is used to train the prediction model; supervised learning algorithms, such as support vector machines (SVM), are used to train the prediction model. These models can learn to predict future security risks based on historical data and known security events.

[0102] Prediction model training: The steps to train a predictive model using a supervised learning algorithm, such as a support vector machine (SVM), typically include the following stages: Data labeling: Label the data to determine whether each data point is associated with a security incident. This is based on the importance calculation of the previous random forest model.

[0103] Data segmentation: The dataset is split into training set and test set in an 8:2 ratio.

[0104] Select Model: Select SVM algorithm as the prediction model.

[0105] Model training: Use the training set data to train the SVM algorithm. SVM distinguishes different categories by finding the hyperplane that maximizes the margin.

[0106] The specific mathematical calculation process of the model is as follows: First, construct the SVM function and select RBF as the kernel function. Then introduce the Lagrange multiplier method. By solving the dual problem, find the optimal Lagrange multiplier, and then use the decision function for classification to complete the SVM algorithm.

[0107] The formula involved is as follows: Lagrangian function:

[0108] Dual problem:

[0109] Decision function:

[0110] Step S1033: Model tuning: adjusting the parameters of the SVM and using cross-validation and grid search techniques to find the optimal parameter combination; Adjust the parameters of SVM, such as C (regularization parameter), kernel function (such as linear kernel, polynomial kernel, radial basis kernel) and kernel function parameters (such as γ of RBF kernel).

[0111] Use techniques such as cross validation and grid search to find the optimal parameter combination.

[0112] Step S1034: Model evaluation: Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, precision, recall, F1 score and ROC-AUC value. Evaluate the performance of the model and optimize it to obtain a trained linkage model. Ensure that the model has high accuracy and generalization ability. Deploy the trained linkage model to the production environment, monitor the data flow in real time, and trigger the early warning system when potential risks are detected.

[0113] Step S104, linkage mechanism; real-time monitoring and analysis of data on the production line, if the real-time data exceeds the warning threshold of step S102, that is, abnormal data, the alarm system is used to determine the severity and start the alarm; at the same time, according to the type and severity of the abnormal data, the trained linkage model is used to start the linkage strategy and perform the linkage response; after each linkage strategy, step S103 is repeated again; Once the real-time data exceeds the warning threshold, the system automatically triggers the alarm system's alarm mechanism and activates a series of predefined linkage strategies. The execution of these strategies follows a set of carefully designed algorithms. First, the production line data is detected for anomalies. Once the data exceeds the preset threshold, it is marked as abnormal data. Subsequently, the system retrieves its association rules based on each abnormal data point. These rules are pre-set according to safety specifications and operating manuals to reveal the mutual influence between different data. In the case of multiple data points being abnormal at the same time, the system will evaluate the correlation between them, and only when at least one of the associated data is not affected will the linkage reaction of the non-abnormal associated data be triggered. In addition, if an abnormal data point is associated with another abnormal data point, such as data B with data A, then the anomaly of data B will be processed separately, while the anomaly of data A will trigger the linkage reaction of all related data including data B. The system will execute corresponding linkage reactions based on the abnormal data and the non-abnormal associated data, which not only helps to diagnose the problem and determine the source of the fault, but also improves the efficiency of linkage response by independently processing abnormal data and linking non-abnormal associated data, ensuring that emergency situations on the production line can be responded to quickly and accurately. The application of this algorithm enables the system to handle data anomalies on the production line more intelligently and achieve fast and effective linkage response, thereby significantly improving the safety and reliability of the production line.

[0114] To give two specific examples in practice, suppose that the sensors on the production line detect a drop in production speed, which may be caused by a machine failure or improper operation. When the production data is abnormal, the system automatically triggers an alarm and responds as follows: reduce the supply of raw materials and shut down the feeding equipment to avoid more defective products during the failure; at the same time, the access control system of the production laboratory locks the production line area to prevent unauthorized personnel from entering and causing the leakage of core technology; the energy and environmental system also adjusts energy distribution accordingly to ensure that other production lines are not affected.

[0115] Another example: the smoke detector of the fire protection system detects smoke in a certain area, which may indicate a fire hazard or a fire. At this time, based on the linkage strategy, the system automatically performs the following operations: triggering the automatic sprinkler system, the access control system opens all emergency exits to ensure the safe evacuation of personnel, the production system suspends all equipment, shuts down machinery that may produce sparks, and finally the energy and environmental system shuts down the fuel supply that may fuel the fire.

[0116] After each linkage strategy ensures production safety, we use machine learning algorithms to predict potential safety risks in production. After the linkage is completed, the main steps are as follows: 1. Data collection: Collect various data in the production environment during linkage, including production data, fire data, energy and environmental data, and access control data. These data may come from sensors, monitoring systems, operation logs, etc. Then pre-process the data, including processing missing values, outliers, standardization and normalization, etc., to ensure data quality (processing method is the same as step S102).

[0117] 2. Feature engineering: Extract features from raw data that are helpful in predicting safety risks. For example, in the case of fire, equipment operating parameters can be extracted from production data, alarm records can be extracted from fire data, energy consumption patterns can be extracted from energy and environment data, and access control records can be extracted from access control data; finally, these data are normalized.

[0118] 3. Establish linkage model: Use machine learning algorithm to establish model and learn linkage relationship between different data. In this example, we use random forest to identify which factors will jointly affect security risk in specific situations (the processing method is the same as step S103).

[0119] When abnormal data is detected, the signal of the linkage response mechanism is immediately activated, including but not limited to: The alarm system is triggered to notify on-site staff and managers; the alarms are divided into three levels: low, medium and high. A value exceeding the maximum or minimum value in the normal range by 10% is a low-level alarm, a value exceeding the maximum or minimum value in the normal range by 25% is a medium-level alarm, and a value exceeding the maximum or minimum value in the normal range by 26% or more is a high-level alarm.

[0120] The low level requires maintenance personnel and workers to pay attention and check when they have time today; the middle level requires staff and equipment maintenance personnel to check immediately; the high level requires staff to arrange evacuation under the leadership of management personnel.

[0121] Automatically adjust the operating status of equipment on the production line to prevent accidents from escalating.

[0122] Activate emergency lighting and evacuation indication systems to ensure personnel safety.

[0123] When a high alarm or fire alarm is triggered, access control will become manual to allow staff to leave.

[0124] When a power supply accident is triggered, the main power supply will be cut off.

[0125] For example, once an abnormality is detected in a data indicator, the system will react accordingly based on the predefined association between the data and other indicators. This linkage mechanism is similar to that in a symphony orchestra, once a certain instrument's performance deviates, the entire orchestra will adjust according to the pre-rehearsed coordination movements. If multiple data indicators are abnormal at the same time, the non-abnormal indicators will also trigger a corresponding linkage reaction. Just like in a football team, if a player is injured, other players will immediately adjust their positions and strategies according to the tactical arrangements to maintain the continuity of the game and the coordination of the team.

[0126] For example, when a sensor in the system detects that the temperature is outside the normal range, it triggers the response of other safety measures associated with the sensor. In the fire safety system in this example, once the smoke detector is activated, not only will the sprinkler system be activated, but the access locks of the emergency exits will also be automatically opened to ensure that people can evacuate safely. If another safety system, such as the power safety system, also detects an abnormality, it will trigger its own alarm independently of the sprinkler system, and the power cabinet will coordinate work, such as powering off, to improve the overall safety response efficiency.

[0127] Step S105, generating and executing a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data, that is, controlling different other subsystems through linkage with different severity levels.

[0128] Specifically, with the production monitoring system as the core, the equipment status subsystem, environmental monitoring subsystem, safety monitoring subsystem, production monitoring and alarm system, emergency response subsystem, access control and energy, and decision support subsystem are integrated, and the information exchange and coordination needs between the subsystems are fully considered, and modular and standardized design is carried out. Through the linkage mechanism between the subsystems, a highly integrated intelligent platform is built to achieve integrated processing of multi-source data and comprehensive application across subsystems. It is worth noting that no matter which data point is abnormal, the video stream of the real-time monitoring system needs to be used as a reference.

[0129] In this embodiment, wired and wireless network technologies such as Ethernet are used to access the environmental monitoring subsystem. In the chemical industry, each environmental monitoring point often works independently and lacks effective information integration. By deploying smart gateways and edge computing devices, centralized data management and remote access can be achieved to collect environmental data, including pollutant concentrations, noise levels, radiation intensity and other information.

[0130] In this embodiment, advanced data processing and analysis technologies, such as big data analysis, machine learning, etc., are used to access the decision support subsystem. By analyzing historical data and real-time data, potential safety risks are predicted to provide a scientific basis for decision-making for the management. Once the production data is not within the normal range, an alarm is issued to notify the equipment maintenance personnel and process manager on duty to go to the operation, in order to ensure that the quality of the finished product is the best and the safety hazards are the least.

[0131] Combined with the emergency dispatch module, a standardized emergency response process is developed to ensure a prompt and effective response in an emergency. Regular emergency drills are conducted to improve employees’ familiarity with emergency plans and their ability to execute them.

[0132] Step 106, record and analyze detailed information of the security incident, including the time, location, cause and handling process of the accident, to facilitate subsequent security training and risk assessment, and regularly maintain and upgrade the security monitoring system, including software updates, hardware inspections and employee training, to ensure efficient operation of the system.

[0133] On the basis of the above embodiments, it also includes: conducting comprehensive evaluation and continuous improvement through centralized management of safety training materials, emergency plans, accident records and other information to enhance the prevention, response and recovery capabilities of safety management.

[0134] In this embodiment, all data collection and processing complies with industry standards and regulatory requirements to ensure data accuracy and security. At the same time, the system design takes into account the characteristics of different production environments and processes to achieve a highly customized safety monitoring solution.

[0135] It should be noted that the steps in this application can be adjusted according to actual production needs and site conditions to adapt to different application scenarios in the material processing industry. Through this flexible and efficient safety management process, the safety level of production in the material processing industry can be significantly improved, accidents can be reduced, and the safety of employees' lives and property can be protected.

[0136] Example 4 Implementation case: Production line fire emergency response Background: A fire was caused by an electrical fault on the production line. The system needed to respond quickly to ensure the safety of personnel and control the fire.

[0137] Implementation steps: 1. Data acquisition module: An abnormal temperature rise is detected, triggering the fire alarm system.

[0138] The video surveillance system 10 captured the smoke and flames.

[0139] The access control system 11 records the evacuation of personnel.

[0140] The energy and environment system 12 detects abnormal energy consumption.

[0141] 2. Data preprocessing and classification: Analyze video surveillance data to confirm the location of the fire.

[0142] Extract possible fire causes from production parameters.

[0143] Build a knowledge graph to associate fire locations with production line equipment.

[0144] 3. Real-time monitoring and anomaly detection module: Abnormal temperatures and smoke were detected, confirming the fire.

[0145] Trigger the fire alarm.

[0146] 4. Linkage response module: Automatically activate the fire fighting system and release the fire extinguishing agent.

[0147] Adjust production line equipment, cut off power supply, and prevent the fire from spreading.

[0148] Use the trained linkage model to start the linkage strategy and perform linkage response.

[0149] 5. Emergency dispatch module: Activate emergency response plans based on the location and size of the fire.

[0150] Dispatch firefighting equipment and personnel to the fire scene.

[0151] 6. Video monitoring and viewing angle adjustment module: The camera automatically adjusts its viewing angle to focus on the fire area.

[0152] The monitoring interface highlights the fire area for easy identification by operators.

[0153] 7. Access control and security warning module: The access control system 11 switches to emergency mode to ensure that the evacuation passage is unobstructed.

[0154] Send a safety alert to notify all personnel to evacuate.

[0155] 8. Energy and Environmental Management Module: Monitor energy usage to ensure adequate power for firefighting equipment.

[0156] Monitor environmental conditions, such as toxic gas concentrations, to ensure personnel safety.

[0157] 9. Mobile response and notification module: Send fire alarms and evacuation instructions to production line supervisors and safety personnel.

[0158] Provides fire location and evacuation route information.

[0159] In this case, all modules worked together to ensure the safe evacuation of personnel and the timely control of the fire; the production line equipment was powered off in a timely manner to reduce property losses; through video surveillance, operators were able to monitor the fire and evacuation situation in real time; in addition, the energy and environmental management module provided the necessary data support to ensure the normal operation of the fire-fighting equipment.

[0160] This application integrates data, which reduces blind spots in monitoring; establishes a linkage model, predicts potential safety risks of the production line through the linkage model, and improves the accuracy and speed of identifying potential safety issues.

[0161] This application conducts automation control and system redundancy: the design of automation control systems and their application in reducing human errors; redundant system design and multiple verification mechanisms.

[0162] Redundant system design is a key strategy in automated control systems that enhances the system's fault tolerance by introducing backups in key components and functions. This design allows the backup system to take over seamlessly when the primary system fails, ensuring continuity and stability of operations. For example, in a critical manufacturing process, two independent control units (also called redundant systems, or linkage strategies) may be designed. When a problem occurs in the primary control unit, the backup unit immediately starts to ensure that production is not interrupted.

[0163] The redundant system design is mainly reflected in the following aspects, namely: 1. Hardware redundancy: Critical hardware components such as power supplies, processors, sensors, etc. are backed up to ensure that when the primary hardware fails, the backup can take over seamlessly.

[0164] 2. Software redundancy: Redundancy at the software level may involve running the same program on different servers or systems to achieve failover.

[0165] 3. Data redundancy: Protecting data from a single point of failure by storing copies of the data in multiple locations.

[0166] 4. Control logic redundancy: Implement multiple copies of control logic in the control system so that if one control logic fails, another can take over.

[0167] 5. Communication redundancy: Deploy multiple communication paths in the system to ensure that the system can still maintain communication when there is a problem with the primary communication link.

[0168] In terms of reducing human errors, the preset procedures and automated processes of the automated control system can reduce the intervention of operators, supplemented by redundant design and linkage strategies, thereby reducing the risks caused by improper human operation. Through well-designed redundancy and linkage strategies, it is possible to respond quickly and minimize the impact on the business when a failure or security incident occurs.

[0169] This application monitors various data on the production line in real time. Once any data is found to be abnormal, a joint response mechanism for related data, namely a linkage mechanism, is immediately initiated according to the nature of the abnormal data; corresponding emergency response plans are formulated based on various linkage mechanisms triggered by abnormal data. Through comprehensive and systematic security monitoring, not only the information island phenomenon is eliminated, but also the safety risks in the production process are effectively identified and prevented, providing a solid guarantee for the factory's safety supervision and process flow control. This has important practical value in reducing the need for manual inspections, reducing the occurrence of production safety accidents, ensuring the safety of equipment and materials, and improving the efficiency of overall safety management.

[0170] The system of this application includes the integrated application of various modules. With their support, it not only improves the intelligence level of safety monitoring in the material processing industry, but also greatly enhances the prevention, response and processing capabilities of production safety, providing strong protection for the stable operation of the enterprise and the safety of employees.

[0171] The present application executes corresponding linkage strategies according to abnormal data and non-abnormal associated data, which not only helps to diagnose problems and determine the source of faults, but also improves the efficiency of linkage response by independently processing abnormal data and linking non-abnormal associated data, ensuring that emergencies on the production line can be responded to quickly and accurately. This method enables the system to handle data anomalies on the production line more intelligently and achieve fast and effective linkage response, thereby significantly improving the safety and reliability of the production line.

[0172] The above contents are only specific implementation methods of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A safety management method based on the material processing industry, characterized in that: The following steps are involved: Step S101: Collect key data; key data includes equipment data, video surveillance data, production data, environmental and health and safety data, access control data, and safety warning data; Step S102: pre-processing and classifying the collected key data to obtain the warning thresholds of various parameters under different targets; Step S103: Algorithm warning; Establish a linkage model to predict potential safety risks of the production line; Step S1031: Establishing a linkage model: Using a machine learning algorithm to establish a random forest model to learn the linkage relationship between the preprocessed and classified key data; Step S1032: Prediction model training: Based on the importance calculation of the random forest model, a support vector machine (SVM) is used to train the prediction model; Step S1033: Model tuning: adjusting the parameters of the SVM and using cross-validation and grid search techniques to find the optimal parameter combination; Step S1034: Model evaluation: Use the test set to evaluate the performance of the model. The evaluation indicators include accuracy, precision, recall, F1 score and ROC-AUC value. Evaluate the performance of the model and optimize it to obtain a trained linkage model, and deploy the trained linkage model to the production environment; Step S104: linkage mechanism; real-time monitoring and analysis of data on the production line; if the real-time data exceeds the warning threshold of step S102, i.e. abnormal data, the alarm system is used to determine the severity and start an alarm; at the same time, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model, and a linkage response is performed; after each linkage strategy, step S103 is repeated again; Step 105: emergency plan; generating and executing a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data; Step 106: Record and analyze detailed information of abnormal events, and perform maintenance and upgrades regularly.

2. A safety management method based on the material processing industry according to claim 1, characterized in that: The preprocessing of key data in step S102 includes unifying frequencies and supplementing missing values; The unified frequency specifically includes: when processing different data, using Fourier transform to convert the time series data into the frequency domain and analyzing its frequency components; The Fourier transform formula is: ； in: N is the number of sample points, X[k] is the kth sample in the frequency domain; x[n] is the nth sample in the time domain; The missing value supplementation includes mean filling and decision tree model; When the data is relatively stable, use the mean to fill in missing values; When the data is relatively discrete and irregular, a decision tree model is used to predict missing values. For regression relationships, the prediction missing value formula of the decision tree model is: ； in: It is the data set contained in a leaf node in the decision tree model; y(x) is the target value of data point x; For classification relationships, the prediction missing value formula of the decision tree model is: ； in: Is a leaf node The majority of .

3. A safety management method based on the material processing industry according to claim 1, characterized in that: The classification of key data in step S102 is as follows: the pre-processed key data is classified according to the data type, and multi-objective optimization is performed by comparing the range of each data when the output is high and the restrictions of the prerequisites to obtain the normal range of each parameter under different objectives, thereby obtaining the warning threshold of each parameter under different objectives.

4. A safety management method based on the material processing industry according to claim 1, characterized in that: The specific steps of using the machine learning algorithm to establish a random forest model in step S1031 are: Step S10311: For each tree, calculate the reduction in impurity caused by each feature when the node is split, and the impurity is represented by the Gini index; Step S10312: For node m in tree i, importance It is expressed as: ； in: ； ; Step S10313: Accumulate the contributions of all nodes in each tree to obtain the feature Importance of Tree I: ; Step S10334: The overall importance score of each feature in the random forest is the average of the feature value scores in all trees, where n is the number of trees in the random forest:

5. A safety management method based on the material processing industry according to claim 1, characterized in that: The prediction model training in step S1032 is specifically as follows: Step S10321: Data labeling, labeling the data, determining the association rules for each data point, or determining whether each data point is associated with an abnormal event, and calculating the association based on the importance of the random forest model; the association rules are pre-set according to safety specifications and operation manuals, and are used to reveal the mutual influence between different data; Step S10322: Data segmentation: split the data set into a training set and a test set according to an 8:2 ratio; Step S10323: Select model: Select SVM algorithm as the prediction model; first construct the SVM function, select RBF as the kernel function, then introduce the Lagrange multiplier method, find the optimal Lagrange multiplier by solving the dual problem, and then use the decision function for classification to complete the SVM algorithm; Step S10324: Model training: Use the training set data to train the SVM algorithm to obtain a prediction model.

6. A safety management method based on the material processing industry according to claim 1, characterized in that: The adjustment of the SVM parameters in step S1033 specifically includes: adjusting the regularization parameter C, the kernel function and the parameters of the kernel function, and using cross-validation and grid search technology to find the optimal parameter combination.

7. A safety management method based on the material processing industry according to claim 1, characterized in that: The alarms of the alarm system in step S104 are divided into three levels: low, medium and high. A value exceeding the warning threshold by 10% is a low-level alarm, a value exceeding the warning threshold by 25% is a medium-level alarm, and a value exceeding the warning threshold by 26% or more is a high-level alarm.

8. A safety management method based on the material processing industry according to claim 1, characterized in that: In step S104, the linkage strategy is started according to the type and severity of the abnormal data using the trained linkage model. Specifically, the production line data is firstly detected for abnormality. Once the data exceeds the preset threshold, it is marked as abnormal data. Then, the association rules are retrieved according to each abnormal data. When multiple data points are abnormal at the same time, the system will evaluate the correlation between them, and only when at least one associated data is not affected, the linkage strategy for the non-abnormal associated data will be triggered. The association rules are pre-set according to the safety specifications and operation manual, or obtained according to the linkage model.

9. A safety management system based on the material processing industry, characterized by: The system comprises a data acquisition module (1), a data preprocessing and classification module (2), a real-time monitoring and anomaly detection module (3), a linkage response module (4), an emergency dispatch module (5), a video monitoring and viewing angle adjustment module (6), an access control and safety warning module (7), an energy and environment management module (8), and a mobile response and notification module (9), wherein the data acquisition module (1) is respectively connected to the data preprocessing and classification module (2), the real-time monitoring and anomaly detection module (3), the video monitoring and viewing angle adjustment module (6), the access control and safety warning module (7), the energy and environment management module (8), and the mobile response and notification module (9), the real-time monitoring and anomaly detection module (3) is respectively connected to the linkage response module (4) and the emergency dispatch module (5), and the linkage response module (4) is respectively connected to the emergency dispatch module (5), the video monitoring and viewing angle adjustment module (6), the access control and safety warning module (7), and the energy and environment management module (8); The data acquisition module (1) is used to collect key data on the production line; the key data includes equipment data, production data, video monitoring data of the video monitoring and viewing angle adjustment module (6), environmental and health safety data of the energy and environmental management module (8), access control data of the access control and safety warning module (7), and safety warning data of the real-time monitoring and abnormality detection module (3); The data preprocessing and classification (2) is used to preprocess and classify the collected key data, obtain the warning thresholds of various parameters under different targets, and conduct in-depth analysis and knowledge graph construction; The real-time monitoring and anomaly detection module (3) is used to continuously monitor the data flow on the production line, establish a linkage model, and predict the potential safety risks of the production line through the linkage model; The linkage response module (4) is used to implement a linkage mechanism, monitor and analyze the data on the production line in real time. If the real-time data exceeds the warning threshold of step S102, i.e. abnormal data, the alarm system is used to determine the severity and start an alarm; at the same time, according to the type and severity of the abnormal data, the trained linkage model is used to start the linkage strategy and perform a linkage response; The emergency dispatch module (5) is used to generate and execute a corresponding emergency plan according to the specific circumstances of the abnormal event generated by the abnormal data; The video monitoring and viewing angle adjustment module (6) is used to automatically adjust the viewing angle of the monitoring camera through the video monitoring system (10) when the real-time monitoring and abnormality detection module (3) detects an abnormality in a specific area; The access control and security warning module (7) is used to automatically switch the access control system (11) to a manual control mode and notify relevant personnel when the real-time monitoring and abnormality detection module (3) detects an abnormality in a specific area; The energy and environment management module (8) utilizes the energy and environment system (12) to monitor energy usage and environmental conditions in real time; The mobile response and notification module (9) is used to send real-time safety alerts and operation instructions to staff via mobile devices.

10. A safety management system based on the material processing industry according to claim 9, characterized in that: The equipment data is generated by industrial equipment sensors and actuators, including temperature, pressure and vibration parameters, and is transmitted to the data acquisition module (1) via a gateway or direct Ethernet connection using Modbus, OPC UA, Profibus industrial communication protocols; The video surveillance data is captured by a surveillance camera and transmitted to a data acquisition module (1) via RTSP and ONVIF protocols; The production data includes output, efficiency, and quality control indicators, which are generated by the production management system and transmitted to the data acquisition module (1) using an API interface or database synchronization technology; The environmental and health safety data is obtained by the energy and environment management module (8); The access control data is obtained by the access control and security warning module (7); The safety warning data comes from safety-related sensors and monitoring systems, and the safety warning data is transmitted in real time to the data acquisition module (1) via an industrial field bus.

Images