Safety production risk dynamic assessment and hierarchical management and control method based on mobile management and control system

Abstract

The invention discloses a safety production risk dynamic assessment and hierarchical management and control method based on a mobile management and control system, and relates to the technical field of mobile management and control systems, and the method comprises the steps: building an assessment model, and constructing a risk assessment model through a big data analysis and machine learning algorithm; real-time evaluation calculation: the mobile management and control system substitutes the data collected in real time into a risk evaluation model for dynamic calculation; according to low risk level definition, medium risk level definition and high risk level definition, a hierarchical management and control strategy is established, technologies such as big data analysis and machine learning algorithms can be utilized to realize real-time monitoring and dynamic evaluation of various risks in the production process, and accurate hierarchical management and control are implemented according to evaluation results. And the scientificity and effectiveness of safety production management are improved.

Description

Technical Field

[0001] This invention relates to the field of mobile management and control system technology, and specifically to a method for dynamic assessment and hierarchical management and control of safety production risks based on a mobile management and control system. Background Technology

[0002] Existing methods for dynamic assessment and tiered management of safety risks based on mobile control systems still have the following drawbacks in practical use:

[0003] Existing risk assessment algorithms may struggle to adapt to complex and ever-changing production environments. Some mobile control systems may fail to comprehensively cover all factors affecting safe production, primarily due to deficiencies in risk assessment models. Control strategies are often based on fixed risk levels, without fully considering the company's own production characteristics, resource availability, and changes in the external environment. For example, some companies have different personnel and equipment configurations during peak and off-peak seasons, but their control strategies are not adjusted accordingly. This leads to difficulties in implementing high-risk control measures during peak seasons and resource waste during off-peak seasons. Furthermore, there is insufficient coordinated control of multiple risks. In actual production, multiple risks often coexist and influence each other, but existing control strategies are mostly designed for single risks, lacking consideration for the synergistic effects of multiple risks. For instance, in petrochemical companies, fire and explosion risks may be interrelated, but the coupling effect between the two may not be fully considered when formulating control strategies, resulting in poor control effectiveness. Summary of the Invention

[0004] The purpose of this invention is to provide a method for dynamic assessment and hierarchical control of safety production risks based on a mobile management and control system, in order to solve the above-mentioned problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for dynamic assessment and hierarchical control of safety production risks based on a mobile control system, including establishing an assessment model and using big data analysis and machine learning algorithms to construct a risk assessment model;

[0006] Real-time assessment and calculation: The mobile control system dynamically calculates risks based on real-time collected data and inputs it into the risk assessment model.

[0007] A tiered control strategy is established based on the definitions of low-risk, medium-risk, and high-risk levels.

[0008] Furthermore, the establishment of the assessment model, utilizing big data analysis and machine learning algorithms, specifically includes:

[0009] Collect and organize historical data on equipment failures, data on personnel operational proficiency, and data on environmental factors;

[0010] Historical equipment failure data: Extract equipment failure records from the enterprise's equipment management database, including information such as failure occurrence time, failure type, failure location, maintenance measures, and maintenance duration. Clean this data to remove duplicate and erroneous records and ensure data accuracy.

[0011] Personnel proficiency data: Collect personnel proficiency data through multiple channels such as employee training records, operation assessment scores, and historical violations. Quantify training course hours and assessment scores into corresponding proficiency indicators, and also track employee violations during the production process.

[0012] Environmental factor data: By utilizing various environmental sensors deployed at the production site, environmental data such as temperature, humidity, noise, and concentration of harmful gases are collected. Time series analysis is then performed on this data to understand the changing trends of environmental parameters over time.

[0013] Furthermore, the establishment of the assessment model, utilizing big data analysis and machine learning algorithms to construct the risk assessment model, also includes:

[0014] Feature extraction and selection are performed on the collected multi-dimensional data. For equipment failure history data, features such as failure frequency, failure interval time, and average maintenance time under different failure modes are extracted. For personnel operation proficiency data, features such as training completion rate, average operation assessment score, and frequency of violation are extracted. For environmental factor data, statistical features such as maximum, minimum, average, and standard deviation of environmental parameters are extracted.

[0015] Furthermore, the establishment of the assessment model, utilizing big data analysis and machine learning algorithms to construct the risk assessment model, also includes:

[0016] Based on the random forest algorithm: Let the total number of data samples be , and divide it into training sets. and test set If the dataset is divided into 70% training set and 30% test set, the number of samples in the training set and the number of samples in the test set are randomly selected from the total dataset. The training set consists of 10 samples. The remaining The test set consists of 10 samples. ;

[0017] A random forest consists of multiple decision trees, each built from a random subset of the training set. Let the i-th decision tree be... Decision Tree From the training set A bootstrap sample set is generated through bootstrap sampling. The sample size is also [number]. When splitting nodes in the decision tree, from the feature set Randomly select a feature subset ,generally ( (Indicates the number of elements in the set);

[0018] For classification problems, the information gain ratio or Gini index is often used as a node splitting criterion. Taking the Gini index as an example, let the node... The included sample set is The sample category set is, the first Class samples in The proportion in is Then the node The Gini index is: ,

[0019] Assuming in features The above threshold To split, Divided into left and right child nodes and The Gini index after splitting is: ,

[0020] Choose to Minimum feature and threshold Perform node splitting;

[0021] Number of trees The stability and generalization ability of the model are affected by... Increasing the depth reduces model variance but increases computational cost, and also increases the maximum depth. The growth depth of the decision tree is limited to prevent overfitting. Let the current depth of the decision tree be... ,when When to stop splitting, minimum number of samples This represents the minimum number of samples a child node must contain when a node splits. When the number of samples in a child node is less than... The splitting stops at that time;

[0022] For input samples Each decision tree in a random forest Make a prediction on it and obtain the prediction result. The final prediction results of the random forest It is obtained through majority voting (for classification problems) or averaging (for regression problems). For classification problems, assuming the class set is , each tree predicts samples. Category The number of votes received was Then the random forest predicts the samples Category ,in ;

[0023] Accuracy: (This refers to the percentage of the test set that is accurate.) The total number of samples is The number of correctly predicted samples is The accuracy rate is: ;

[0024] Recall: For a given category, the percentage of students who actually belong to the test set. The number of samples is , and the correct prediction is The number of samples is Then the recall rate for this category is: ;

[0025] F1 score: The F1 score is a metric that takes into account both precision and recall. For categories... Its F1 value is: ,in, For category The precision rate is calculated similarly to the accuracy rate, but the molecule is the one correctly predicted. The number of samples, the denominator is all those predicted as The number of samples is used to calculate the macro F1 score, which is the average of the F1 scores for all classes, while the micro F1 score is calculated based on the precision and recall of all samples.

[0026] Furthermore, the real-time assessment calculation, whereby the mobile control system dynamically calculates based on the real-time collected data and inputs it into the risk assessment model, specifically includes:

[0027] The mobile control system continuously acquires equipment operating status data (such as temperature, pressure, speed, etc.), personnel operation data (such as operation instructions, operation time, etc.), and environmental data (such as temperature, humidity, concentration of harmful gases, etc.) through real-time communication with on-site sensors and mobile terminal devices.

[0028] The real-time collected data is preprocessed, including data cleaning and normalization. For any abnormal data that may occur, such as sensor measurements that exceed the reasonable range, a data smoothing algorithm is used to correct it.

[0029] The pre-processed real-time data is fed into the established risk assessment model for dynamic calculation. The model reassesses the equipment failure risk level based on changes in equipment operating parameters, personnel operation behavior, and environmental factors in the real-time data.

[0030] Furthermore, the establishment of a tiered control strategy based on the definitions of low-risk, medium-risk, and high-risk levels specifically includes:

[0031] For low-risk management strategies: Regular inspections are scheduled, with a rigorous inspection plan. Professional inspection personnel conduct a comprehensive daily inspection of production equipment and the environment. The inspection route covers all equipment within the production area, including core equipment, auxiliary equipment, and surrounding facilities. Inspection personnel are equipped with mobile devices and a dedicated inspection app. During inspections, they use the app to photograph and document the equipment's appearance according to a standardized inspection checklist, and meticulously check various operating parameters such as temperature, pressure, and speed. This data is then entered into the app in real time. The app has a built-in intelligent analysis function that automatically compares the entered data with preset standard values. If any anomalies are detected, an immediate warning is issued.

[0032] Strengthen daily supervision: Establish a sound daily supervision system. Managers can view the inspection records and equipment operation data uploaded by the inspectors in real time through the back-end management interface of the mobile control system. Every day, a dedicated person will conduct a detailed spot check on some inspection records, focusing on verifying the accuracy and completeness of the data.

[0033] For medium-risk management strategies: Establish a dedicated person in charge mechanism, assign an experienced person in charge to each medium-risk area or equipment. The person in charge should have at least 5 years of relevant industry safety production management experience and hold a registered safety engineer certificate. The person in charge is fully responsible for the risk management work of the area or equipment, including formulating rectification plans, organizing and implementing rectification measures, coordinating relevant resources, and reporting and communicating with higher-level departments.

[0034] A detailed rectification plan was developed. Based on the risk assessment results, the person in charge formulated a comprehensive rectification plan, clearly defining the rectification objectives, measures, timelines, and responsible parties. The rectification objectives were specific and measurable, such as reducing the equipment failure rate to below 1% and fully restoring environmental parameters to normal standard ranges. The rectification measures were targeted and feasible. For equipment failure issues, these measures might include comprehensive equipment overhaul, replacement of aging parts, and optimization of equipment operating parameters. For personnel operation issues, these measures might include organizing specialized training and strengthening on-site supervision and guidance.

[0035] In the mobile management system, a countdown reminder function is set up for each task in the rectification plan. Three days before the task completion deadline, the system automatically sends SMS reminders to the person in charge and relevant responsible persons. One day before the deadline, another SMS reminder is sent, and a prominent red warning pop-up appears on the mobile terminal APP. A rectification work progress report meeting is held once a week. The person in charge uploads the rectification work progress through the mobile terminal APP, including completed tasks, the current task completion progress, problems encountered and solutions.

[0036] For high-risk management strategies: once a high-risk situation is identified, the pre-established emergency plan will be activated immediately. The emergency plan covers all aspects of accident prevention, emergency response, emergency rescue, and accident recovery. An emergency notification will be sent to all relevant personnel through the mobile management system. The notification will include key information such as the risk type, risk level, location of the incident, and emergency response requirements. All relevant personnel will be notified within 3 minutes and required to take swift action in strict accordance with the provisions of the emergency plan. Within 30 minutes, professional technical personnel and emergency rescue personnel will be quickly dispatched to the scene.

[0037] Establish emergency resource dispatch procedures and, through the resource dispatch module of the mobile control system, allocate various internal resources in real time, including human, material, and financial resources. Coordinate personnel from other departments to support emergency response work, such as drawing personnel from the production department to assist in evacuation, urgently procuring necessary emergency relief supplies from the purchasing department, and arranging for the finance department to prepare emergency funds within one hour to ensure timely availability. At the same time, maintain close contact with external rescue forces, such as fire departments, medical institutions, and professional rescue organizations, and request external support when necessary. When coordinating with external rescue forces, share on-site risk information and rescue progress in a timely manner through the mobile control system to improve rescue efficiency and minimize losses caused by high risks.

[0038] Compared with existing technologies, the method for dynamic assessment and hierarchical control of safety production risks based on a mobile control system provided by this invention has the following beneficial effects:

[0039] This method for dynamic assessment and hierarchical control of safety production risks based on a mobile control system establishes an assessment model and utilizes big data analysis and machine learning algorithms to construct a risk assessment model. Real-time assessment and calculation are performed by the mobile control system using real-time collected data, which is then fed into the risk assessment model for dynamic calculation. Based on the definitions of low-risk, medium-risk, and high-risk levels, a hierarchical control strategy is established. This method leverages big data analysis and machine learning algorithms to achieve real-time monitoring and dynamic assessment of various risks in the production process, and implements precise hierarchical control based on the assessment results, thereby improving the scientific nature and effectiveness of safety production management. Attached Figure Description

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0041] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0042] Figure 2 This is a schematic diagram of the system structure of the present invention. Detailed Implementation

[0043] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] Please see Figure 1-2 A method for dynamic assessment and hierarchical control of safety production risks based on a mobile control system, including establishing an assessment model and using big data analysis and machine learning algorithms to construct a risk assessment model;

[0045] Real-time assessment and calculation: The mobile control system dynamically calculates risks based on real-time collected data and inputs it into the risk assessment model.

[0046] A tiered control strategy is established based on the definitions of low-risk, medium-risk, and high-risk levels.

[0047] S101. Establish an assessment model, utilizing big data analysis and machine learning algorithms to construct a risk assessment model, specifically including:

[0048] Collect and organize historical data on equipment failures, data on personnel operational proficiency, and data on environmental factors;

[0049] Historical equipment failure data: Extract equipment failure records from the enterprise's equipment management database, including information such as failure occurrence time, failure type, failure location, maintenance measures, and maintenance duration. Clean this data to remove duplicate and erroneous records and ensure data accuracy.

[0050] Personnel proficiency data: Collect personnel proficiency data through multiple channels such as employee training records, operation assessment scores, and historical violations. Quantify training course hours and assessment scores into corresponding proficiency indicators, and also track employee violations during the production process.

[0051] Environmental factor data: By utilizing various environmental sensors deployed at the production site, environmental data such as temperature, humidity, noise, and concentration of harmful gases are collected. Time series analysis is then performed on this data to understand the changing trends of environmental parameters over time.

[0052] The assessment model is built by using big data analysis and machine learning algorithms, and also includes:

[0053] Feature extraction and selection are performed on the collected multi-dimensional data. For equipment failure history data, features such as failure frequency, failure interval time, and average maintenance time under different failure modes are extracted. For personnel operation proficiency data, features such as training completion rate, average operation assessment score, and frequency of violation are extracted. For environmental factor data, statistical features such as maximum, minimum, average, and standard deviation of environmental parameters are extracted.

[0054] The assessment model is built by using big data analysis and machine learning algorithms, and also includes:

[0055] Based on the random forest algorithm: Let the total number of data samples be , and divide it into training sets. and test set If the dataset is divided into 70% training set and 30% test set, the number of samples in the training set and the number of samples in the test set are randomly selected from the total dataset. The training set consists of 10 samples. The remaining The test set consists of 10 samples. ;

[0056] A random forest consists of multiple decision trees, each built from a random subset of the training set. Let the i-th decision tree be... Decision Tree From the training set A bootstrap sample set is generated through bootstrap sampling. The sample size is also [number]. When splitting nodes in the decision tree, from the feature set Randomly select a feature subset ,generally ( (Indicates the number of elements in the set);

[0057] For classification problems, the information gain ratio or Gini index is often used as a node splitting criterion. Taking the Gini index as an example, let the node... The included sample set is The sample category set is, the first Class samples in The proportion in is Then the node The Gini index is: ,

[0058] Assuming in features The above threshold To split, Divided into left and right child nodes and The Gini index after splitting is: ,

[0059] Choose to Minimum feature and threshold Perform node splitting;

[0060] Number of trees The stability and generalization ability of the model are affected by... Increasing the depth reduces model variance but increases computational cost, and also increases the maximum depth. The growth depth of the decision tree is limited to prevent overfitting. Let the current depth of the decision tree be... ,when When to stop splitting, minimum number of samples This represents the minimum number of samples a child node must contain when a node splits. When the number of samples in a child node is less than... The splitting stops at that time;

[0061] For input samples Each decision tree in a random forest Make a prediction on it and obtain the prediction result. The final prediction results of the random forest It is obtained through majority voting (for classification problems) or averaging (for regression problems). For classification problems, assuming the class set is , each tree predicts samples. Category The number of votes received was Then the random forest predicts the samples Category ,in ;

[0062] Accuracy: (This refers to the percentage of the test set that is accurate.) The total number of samples is The number of correctly predicted samples is The accuracy rate is: ;

[0063] Recall: For a given category, the percentage of students who actually belong to the test set. The number of samples is , and the correct prediction is The number of samples is Then the recall rate for this category is: ;

[0064] F1 score: The F1 score is a metric that takes into account both precision and recall. For categories... Its F1 value is: ,in, For category The precision rate is calculated similarly to the accuracy rate, but the molecule is the one correctly predicted. The number of samples, the denominator is all those predicted as The number of samples is used to calculate the macro F1 score, which is the average of the F1 scores for all classes, while the micro F1 score is calculated based on the precision and recall of all samples.

[0065] S102. Real-time assessment and calculation: The mobile control system dynamically calculates risks based on real-time collected data, inputting it into the risk assessment model. This includes:

[0066] The mobile control system continuously acquires equipment operating status data (such as temperature, pressure, speed, etc.), personnel operation data (such as operation instructions, operation time, etc.), and environmental data (such as temperature, humidity, concentration of harmful gases, etc.) through real-time communication with on-site sensors and mobile terminal devices.

[0067] The real-time collected data is preprocessed, including data cleaning and normalization. For any abnormal data that may occur, such as sensor measurements that exceed the reasonable range, a data smoothing algorithm is used to correct it.

[0068] The pre-processed real-time data is fed into the established risk assessment model for dynamic calculation. The model reassesses the equipment failure risk level based on changes in equipment operating parameters, personnel operation behavior, and environmental factors in the real-time data.

[0069] S103. Based on the definitions of low-risk, medium-risk, and high-risk levels, establish a tiered control strategy, specifically including:

[0070] For low-risk management strategies: Regular inspections are scheduled, with a rigorous inspection plan. Professional inspection personnel conduct a comprehensive daily inspection of production equipment and the environment. The inspection route covers all equipment within the production area, including core equipment, auxiliary equipment, and surrounding facilities. Inspection personnel are equipped with mobile devices and a dedicated inspection app. During inspections, they use the app to photograph and document the equipment's appearance according to a standardized inspection checklist, and meticulously check various operating parameters such as temperature, pressure, and speed. This data is then entered into the app in real time. The app has a built-in intelligent analysis function that automatically compares the entered data with preset standard values. If any anomalies are detected, an immediate warning is issued.

[0071] Strengthen daily supervision: Establish a sound daily supervision system. Managers can view the inspection records and equipment operation data uploaded by the inspectors in real time through the back-end management interface of the mobile control system. Every day, a dedicated person will conduct a detailed spot check on some inspection records, focusing on verifying the accuracy and completeness of the data.

[0072] For medium-risk management strategies: Establish a dedicated person in charge mechanism, assign an experienced person in charge to each medium-risk area or equipment. The person in charge should have at least 5 years of relevant industry safety production management experience and hold a registered safety engineer certificate. The person in charge is fully responsible for the risk management work of the area or equipment, including formulating rectification plans, organizing and implementing rectification measures, coordinating relevant resources, and reporting and communicating with higher-level departments.

[0073] A detailed rectification plan was developed. Based on the risk assessment results, the person in charge formulated a comprehensive rectification plan, clearly defining the rectification objectives, measures, timelines, and responsible parties. The rectification objectives were specific and measurable, such as reducing the equipment failure rate to below 1% and fully restoring environmental parameters to normal standard ranges. The rectification measures were targeted and feasible. For equipment failure issues, these measures might include comprehensive equipment overhaul, replacement of aging parts, and optimization of equipment operating parameters. For personnel operation issues, these measures might include organizing specialized training and strengthening on-site supervision and guidance.

[0074] In the mobile management system, a countdown reminder function is set up for each task in the rectification plan. Three days before the task completion deadline, the system automatically sends SMS reminders to the person in charge and relevant responsible persons. One day before the deadline, another SMS reminder is sent, and a prominent red warning pop-up appears on the mobile terminal APP. A rectification work progress report meeting is held once a week. The person in charge uploads the rectification work progress through the mobile terminal APP, including completed tasks, the current task completion progress, problems encountered and solutions.

[0075] For high-risk management strategies: once a high-risk situation is identified, the pre-established emergency plan will be activated immediately. The emergency plan covers all aspects of accident prevention, emergency response, emergency rescue, and accident recovery. An emergency notification will be sent to all relevant personnel through the mobile management system. The notification will include key information such as the risk type, risk level, location of the incident, and emergency response requirements. All relevant personnel will be notified within 3 minutes and required to take swift action in strict accordance with the provisions of the emergency plan. Within 30 minutes, professional technical personnel and emergency rescue personnel will be quickly dispatched to the scene.

[0076] Establish emergency resource dispatch procedures and, through the resource dispatch module of the mobile control system, allocate various internal resources in real time, including human, material, and financial resources. Coordinate personnel from other departments to support emergency response work, such as drawing personnel from the production department to assist in evacuation, urgently procuring necessary emergency relief supplies from the purchasing department, and arranging for the finance department to prepare emergency funds within one hour to ensure timely availability. At the same time, maintain close contact with external rescue forces, such as fire departments, medical institutions, and professional rescue organizations, and request external support when necessary. When coordinating with external rescue forces, share on-site risk information and rescue progress in a timely manner through the mobile control system to improve rescue efficiency and minimize losses caused by high risks.

[0077] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A safety production risk dynamic assessment and hierarchical control method based on a mobile management system, characterized in that, The method comprises the following steps: Establishing an evaluation model, using big data analysis and machine learning algorithms to build a risk assessment model; Real-time evaluation and calculation: the mobile management and control system dynamically calculates the risk assessment model based on real-time data collection; 2. The safety production risk dynamic assessment and hierarchical control method based on the mobile management system according to claim 1, characterized in that, Establishing a hierarchical control strategy according to the low-risk level definition, medium-risk level definition, and high-risk level definition. The establishment of the evaluation model, the use of big data analysis and machine learning algorithms, and the construction of the risk assessment model specifically include: Collect and organize equipment failure history data, personnel operation proficiency data, and environmental factor data; Equipment failure history data: extract equipment failure records from the enterprise's equipment management database, including failure time, failure type, failure location, repair measures, and repair duration, etc. Clean these data to remove duplicate and incorrect records to ensure data accuracy; Personnel operation proficiency data: collect personnel operation proficiency related data through employee training records, operation examination results, and historical violation operation times, etc. Quantify the training course hours and examination results into corresponding proficiency indicators, and also count the violation operation behaviors of employees in the production process; 3. The safety production risk dynamic assessment and hierarchical control method based on the mobile management system according to claim 2, characterized in that, Environmental factor data: use various environmental sensors deployed in the production site to collect temperature, humidity, noise, and harmful gas concentration, etc. Perform time series analysis on these data to understand the trend of environmental parameters over time. The establishment of the evaluation model, the use of big data analysis and machine learning algorithms, and the construction of the risk assessment model also include:

4. The safety production risk dynamic assessment and hierarchical control method based on the mobile management system according to claim 3, characterized in that, Feature extraction and selection of the collected multi-dimensional data: for equipment failure history data, extract the frequency of failure occurrence, failure interval time, average repair time under different failure modes, etc.; for personnel operation proficiency data, extract training completion rate, operation examination average score, and violation operation frequency, etc.; for environmental factor data, extract statistical features such as maximum value, minimum value, average value, and standard deviation of environmental parameters. Based on the random forest algorithm: Let the total number of data samples be , and divide it into training sets. and test set If the training set is divided into 70% training set and 30% test set, then the number of training set samples is... Number of test set samples Randomly sampled from the total dataset The training set consists of 10 samples. The remaining The test set consists of 10 samples. ; A random forest consists of multiple decision trees, each built from a random subset of the training set. Let the i-th decision tree be... Decision Tree From the training set A bootstrap sample set is generated through bootstrap sampling. The sample size is also [number]. When splitting nodes in the decision tree, from the feature set Randomly select a feature subset ,generally ( (Indicates the number of elements in the set); For classification problems, information gain ratio or Gini index is often used as the node split criterion. Taking Gini index as an example, let the node be the sample set, the sample category set be , the proportion of the th category sample in be , and the Gini index of the node be . , Assume that in the feature threshold Splitting, will into left and right child nodes and Gini index after splitting: , selecting to cause the smallest feature and the threshold perform node splitting; Number of trees Affects the stability and generalization ability of the model, as Increasing, model variance decreases but computational cost increases, maximum depth Limits the growth depth of the decision tree, preventing overfitting, let the current depth of the decision tree be Stop splitting when The minimum number of samples Indicates the minimum number of samples that the child node must contain when the node is split, when the number of samples of the child node is less than Stop splitting; For input samples Each decision tree in a random forest Make a prediction on it and obtain the prediction result. The final prediction results of the random forest It is obtained through majority voting (for classification problems) or averaging (for regression problems). For classification problems, assuming the class set is , each tree predicts samples. Category The number of votes received was Then the random forest predicts the samples Category ,in ; Accuracy: (This refers to the percentage of the test set that is accurate.) The total number of samples is The number of correctly predicted samples is The accuracy rate is: ; Recall: For a certain class, let the number of samples in the test set that actually belong to the class be, and the number of samples that are correctly predicted to belong to the class be. Then the recall of the class is: ;​​​ F1 value (F1-score): The F1 value is an index that comprehensively considers the accuracy and recall rate, and for a class , the F1 value is: , wherein, is the precision of the class , the calculation method is similar to the accuracy, but the numerator is the number of samples correctly predicted as , and the denominator is the number of all samples predicted as , the Macro-F1 value is the average value of the F1 values of all classes, and the Micro-F1 value is calculated based on the accuracy and recall rate of all samples.

5. The safety production risk dynamic assessment and hierarchical control method based on the mobile management system according to claim 1, characterized in that, The establishment of the evaluation model, the use of big data analysis and machine learning algorithms, and the construction of the risk assessment model also include: The real-time evaluation and calculation, the mobile management and control system dynamically calculates the risk assessment model based on real-time data collection, specifically includes: The mobile management and control system continuously acquires equipment operation state data (such as temperature, pressure, and speed, etc.), personnel operation data (such as operation instructions and operation time, etc.), and environmental data (such as temperature and humidity, and harmful gas concentration, etc.) through real-time communication with on-site sensors and mobile terminal devices; Preprocess the real-time collected data, including data cleaning and normalization, etc. For possible abnormal data, such as sensor measurement values exceeding the reasonable range, correct them through data smoothing algorithm; 6. The safety production risk dynamic assessment and hierarchical management and control method based on the mobile management and control system according to claim 1, characterized in that, Put the preprocessed real-time data into the already constructed risk assessment model for dynamic calculation. The model re-evaluates the equipment failure risk level according to the changes of equipment operation parameters, personnel operation behaviors, and environmental factors in real-time data. The establishment of the hierarchical control strategy according to the low-risk level definition, medium-risk level definition, and high-risk level definition specifically includes: For low-risk control strategy: Regular inspection arrangement, develop strict inspection plan, arrange professional inspection personnel to conduct comprehensive inspection on production equipment and environment every day, inspection route covers all equipment in production area, including core equipment, auxiliary equipment and surrounding environment facilities, inspection personnel are equipped with mobile terminal equipment, install special inspection APP, during inspection, according to standardized inspection list, use APP to take pictures of equipment appearance, check equipment operation parameters such as temperature, pressure and speed in detail, and input data into APP, APP has built-in intelligent analysis function, which can automatically compare input data with preset standard value, and issue warning prompt as soon as abnormality is found, Strengthen daily supervision: Establish perfect daily supervision system, management personnel can check inspection records and equipment operation data uploaded by inspection personnel through background management interface of mobile control system, arrange special personnel to check part of inspection records every day, focus on verifying data accuracy and completeness; For medium-risk control strategy: Develop special person responsible mechanism, assign experienced person in charge for each medium-risk area or equipment, person in charge should have at least 5 years of experience in safety production management of related industry, and hold registered safety engineer certificate preferentially, person in charge is responsible for risk control of the area or equipment, including developing rectification plan, organizing and implementing rectification measures, coordinating related resources and reporting to superior department; Develop detailed rectification plan, person in charge develops detailed rectification plan according to risk assessment results, rectification plan specifies rectification target, rectification measures, rectification time node and responsible person, rectification target is specific and measurable, such as reducing equipment failure rate to less than 1%, and restoring environmental parameters to normal standard range, rectification measures are targeted and operable, for equipment failure problem, may include comprehensive maintenance of equipment, replacement of aging parts, optimization of equipment operation parameters, for personnel operation problem, may include organization of special training, strengthening of on-site supervision and guidance, In mobile control system, set countdown reminder function for each task of rectification plan, 3 days before task completion time node, system automatically sends SMS reminder to person in charge and related responsible person, 1 day before, send SMS reminder again, and pop up prominent red warning prompt on mobile terminal APP, hold rectification work progress report meeting once a week, person in charge uploads rectification work progress through mobile terminal APP, including completed tasks, completion progress of current task, problems encountered and solutions; For high-risk control strategy: emergency plan starts, as soon as it is determined to be high risk, start the pre-prepared emergency plan, which covers accident prevention, emergency response, emergency rescue, accident recovery and other aspects. Through the mobile control system, send emergency notification to all relevant personnel, including risk type, risk level, location, emergency response requirements and other key information, ensure that all relevant personnel receive the notification within 3 minutes, and require them to strictly follow the provisions of the emergency plan and take action quickly within 30 minutes. Rapidly deploy professional technical personnel and emergency rescue personnel to the scene; Develop resource emergency scheduling procedures, real-time allocation of various resources within the enterprise, including manpower, material and financial resources, through the resource scheduling module of the mobile control system, coordinate the support of personnel from other departments for emergency response work, such as diverting some personnel from the production department to assist in evacuation, and urgently purchasing emergency rescue materials from the procurement department, arranging the finance department to prepare emergency funds within 1 hour to ensure that funds are in place in time. At the same time, maintain close contact with external rescue forces, such as the fire department, medical institutions and professional rescue agencies, and request external assistance if necessary. When interfacing with external rescue forces, share real-time risk information and rescue progress through the mobile control system to improve rescue efficiency and minimize the loss caused by high risk.

Images