Early warning system for production safety accidents based on artificial intelligence

By integrating data collection, anomaly detection and emergency response modules in the field of production safety, using IoT sensors and graph neural networks to build a knowledge graph, the existing system's incomplete coverage and inaccurate evaluation in complex environments is solved, and the problem of existing systems is achieved Efficient warning and response to potential hazards of production safety accidents.

CN118333411BActive Publication Date: 2024-11-01FUZHOU WUBIHUAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410768446.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-06-14
Publication Date
2024-11-01
Estimated Expiration
2044-06-14

AI Technical Summary

Technical Problem

The existing artificial intelligence-based production safety accident hazard warning system is difficult to fully cover potential hazards in a complex and changing production environment, and is easily affected by individual subjective factors, resulting in inaccurate risk assessment and difficult to effectively prevent and promptly detect production safety accidents.

Method used

A system including data collection, anomaly detection, accident prediction, emergency response and evaluation optimization module was designed. By deploying IoT sensors to collect data in real time, using natural language processing and graph neural networks to build a knowledge graph, perform abnormal detection and accident prediction. Generate early warning information and formulate emergency measures to optimize system performance.

Benefits of technology

It has achieved comprehensive monitoring of the production environment and automatic identification of potential hidden dangers, improved the intelligence level of accident warning, reduced the risk of accidents, and improved the prevention and response capabilities of production safety accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118333411B_ABST
    Figure CN118333411B_ABST
Patent Text Reader

Abstract

The present invention discloses an artificial intelligence-based early warning system for hidden dangers of production accidents, which relates to the technical field of early warning technology for hidden dangers of production accidents, including a data collection module for deploying sensors to collect production environment monitoring data and historical data in real time, pre-processing the data and integrating and storing the data into a database, and constructing a knowledge graph based on the database; an anomaly detection module for performing anomaly detection based on pre-processed real-time monitoring data; an accident prediction module for performing accident prediction and risk assessment based on the anomaly detection results; an emergency response module for generating early warning information and formulating emergency response measures. The present invention can automatically identify and predict potential accidents and improve the intelligent level of early warning by constructing a knowledge graph, performing anomaly detection on real-time data, predicting accidents and conducting risk assessment, generating early warning information and formulating emergency response measures.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of early warning technology for hidden dangers of production safety accidents, and in particular to an early warning system for hidden dangers of production safety accidents based on artificial intelligence. Background Art

[0002] With the continuous improvement of the global industrialization level, production safety has become a key issue that governments and enterprises are extremely concerned about. Traditional production safety management methods mainly rely on manual inspections and experience judgments. This method is effective in dealing with simple and repetitive problems. However, in the face of complex and changeable production environments and potential safety hazards, these methods often seem to be powerless. On the one hand, manual inspections have blind spots and it is difficult to achieve comprehensive coverage; on the other hand, experience judgments are easily affected by personal subjective factors, making it difficult to achieve objective and accurate risk assessments. Therefore, how to effectively prevent and promptly discover hidden dangers of production safety accidents, reduce the risk of accidents, and improve the prevention and response capabilities of production safety accidents is a major challenge facing the current production safety field. Summary of the invention

[0003] In view of the above-mentioned problems existing in the existing artificial intelligence-based production safety accident hazard early warning system, the present invention is proposed.

[0004] Therefore, the problem to be solved by the present invention is how to effectively prevent and timely discover hidden dangers of production safety accidents, reduce the risk of accidents, and improve the prevention and response capabilities of production safety accidents, which is a major challenge facing the current production safety field.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: an artificial intelligence-based safety production accident hazard early warning system, which includes a data collection module for deploying sensors to collect production environment monitoring data and historical data in real time, pre-processing the data and integrating and storing the data in a database, and constructing a knowledge graph based on the database;

[0006] Anomaly detection module, used to perform anomaly detection based on preprocessed real-time monitoring data;

[0007] The accident prediction module is used to predict accidents and conduct risk assessment based on the anomaly detection results;

[0008] Emergency response module, used to generate early warning information and formulate emergency response measures;

[0009] Evaluation and optimization module, used to evaluate system performance and perform optimization.

[0010] As a preferred solution of the artificial intelligence-based production safety accident hazard early warning system described in the present invention, the deployment of sensors to collect production environment monitoring data and historical data in real time refers to deploying loT sensors to collect production environment monitoring data and equipment parameters in real time, obtaining safety regulations, standards and historical accident reports from safety supervision departments, and obtaining safety logs, operation manuals and equipment maintenance records from the company's internal management system.

[0011] As a preferred solution of the artificial intelligence-based production safety accident hazard early warning system described in the present invention, wherein: the pre-processing of data and then integrating and storing the data into a database refers to pre-processing text data using natural language processing technology, including word segmentation, removal of stop words, part-of-speech tagging and entity recognition, converting the pre-processed text into structured data, normalizing the collected real-time data, using a box plot method to identify and remove outliers, integrating the pre-processed historical data and real-time data into a unified database for storage, and regularly backing up the stored data to the cloud.

[0012] As a preferred solution of the artificial intelligence-based safety production accident hazard early warning system of the present invention, wherein: the construction of the knowledge graph based on the database includes the following steps:

[0013] Use the BERT model to identify key entities in the text of historical data and label the entity types;

[0014] Construct a neural network model based on the attention mechanism for relationship extraction:

[0015] , where is the relation extraction result vector, is the Softmax function, is the number of key entities in the text, is the transpose of the weight vector, is the hyperbolic tangent activation function, is the weight matrix, is the bias term, is the hidden state vector of the qth entity after being encoded by the BERT model, refers to the hidden state vector of the jth entity obtained after encoding the entity, is the weight matrix of the relation extraction task, is an exponential function;

[0016] The graph database is used to construct a knowledge graph in the field of safe production based on the extracted entities and relationships.

[0017] As a preferred solution of the artificial intelligence-based early warning system for production safety accidents and hidden dangers of the present invention, wherein: the abnormality detection based on the pre-processed real-time monitoring data includes the following steps:

[0018] For each data point in the real-time data stream , calculate the weighted moving mean and weighted moving standard deviation of the data points:

[0019] , , where is the weighted moving average of the most recent N data points including the current data point on the dth feature dimension, is the weight coefficient of the ith data point, is the value of the ith data point in the dth feature dimension, It is the weighted moving standard deviation of the most recent N data points including the current data point on the dth feature dimension;

[0020] Compute the anomaly score for each dimension of the data separately:

[0021] , where is the comprehensive anomaly score of the tth data point, D is the number of feature dimensions of the data, is the value of the t-th data point on the d-th feature dimension, is the weighted moving average, is the weighted moving standard deviation;

[0022] The anomaly threshold is defined as:

[0023] , where is the abnormal threshold at time t, is the anomaly detection score of the ith data point, is the number of anomaly scores, is the moving average of the historical anomaly scores, is the regulating factor;

[0024] The abnormal threshold and anomaly detection results Compare and judge the abnormality of data.

[0025] like , it means that the data is abnormal. , it means that there is no abnormality in the data.

[0026] As a preferred solution of the artificial intelligence-based early warning system for production safety accidents, the method comprises the following steps:

[0027] Use autoencoders to extract deep features of abnormal data, use clustering algorithms to classify the extracted features, extract knowledge graph features through graph neural networks, use statistical methods to extract vectors of real-time data, and comprehensively build an accident prediction model:

[0028] , where is the probability of an accident happening, is the activation function, , and They are respectively the results of deep clustering , Knowledge Graph Features Extracted by Graph Neural Network and real-time data features The corresponding weight vector, is the bias term, Represents a transpose operation;

[0029] The abnormal data is input into the accident prediction model, which automatically processes the input data and outputs the predicted probability of an accident;

[0030] Create a risk matrix, derive the severity of the accident through knowledge graph analysis, quantify the probability of an accident and the severity of the accident, and determine different risk levels based on the combination of the two;

[0031] Calculate the risk index of an accident:

[0032] , where is the risk index of the accident, is the probability of an accident happening, is the potential severity of the accident, is the weight coefficient of the lth additional risk factor, is the quantitative value of the lth additional risk factor, is the total number of additional risk factors, is the weight coefficient;

[0033] The overall risk index and risk judgment threshold and Compare and judge the accident risk level, among which < ,like Less than , then it is low risk, if ≤ < , then it is medium risk, if ≥ , then high risk.

[0034] As a preferred solution of the artificial intelligence-based safety production accident hazard early warning system of the present invention, wherein: the generation of early warning information and the formulation of emergency response measures refer to determining the generation of early warning information and the formulation of corresponding emergency response measures according to the risk level,

[0035] If the risk level is low, no warning information will be generated, and daily monitoring of low-risk areas will continue, and detailed information on low-risk situations will be recorded, and safety inspections of low-risk areas will be carried out regularly;

[0036] If the risk level is medium or high, early warning information will be generated, including the type of accident, risk level, cause and response measures. The early warning information will be released to relevant managers and operators through the communication network. Emergency response measures will be formulated according to the recommended measures in the early warning information, including isolation and safety inspection of potential risk areas, emergency shutdown, evacuation of personnel and adjustment of production parameters.

[0037] As a preferred solution of the artificial intelligence-based production safety accident hazard early warning system described in the present invention, the evaluation of system performance and optimization refers to collecting feedback information from relevant personnel and systems after executing emergency response measures, analyzing the implementation status and effects of emergency response measures, evaluating the accuracy and timeliness of early warning information, adjusting the accident prediction model parameters and optimizing the process accordingly based on the feedback analysis results, and updating the feedback analysis results to the knowledge graph.

[0038] A computer device comprises: a memory and a processor; the memory stores a computer program, and the processor implements the steps of an artificial intelligence-based safety production accident hazard early warning system when executing the computer program.

[0039] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an artificial intelligence-based safety production accident hazard early warning system.

[0040] The beneficial effects of the present invention are as follows: the present invention constructs a knowledge graph, performs anomaly detection on real-time data, performs accident prediction and risk assessment, generates early warning information and formulates emergency measures, can automatically identify and predict potential accidents, improve the intelligence level of early warning, reduce the risk of accidents and improve the prevention and response capabilities of production safety accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0042] Figure 1 This is a flow chart of the early warning system for production safety accidents based on artificial intelligence.

[0043] Figure 2 This is a schematic diagram of the implementation of an early warning system for production safety accidents based on artificial intelligence. DETAILED DESCRIPTION

[0044] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings.

[0045] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0047] Example 1

[0048] Reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and this embodiment provides a safety production accident hazard early warning system based on artificial intelligence. The safety production accident hazard early warning system based on artificial intelligence includes the following modules:

[0049] S1, data collection module, used to deploy sensors to collect production environment monitoring data and historical data in real time, integrate and store the data into the database after preprocessing, and build a knowledge graph based on the database;

[0050] Specifically, deploying sensors to collect production environment monitoring data and historical data in real time means deploying loT sensors to collect production environment monitoring data and equipment parameters in real time, obtaining safety regulations, standards and historical accident reports from safety supervision departments, and obtaining safety logs, operation manuals and equipment maintenance records from the company's internal management system.

[0051] By deploying IoT sensors to monitor key parameters (such as temperature, pressure, humidity, etc.) and equipment status in the production environment in real time and collecting historical data, comprehensive and real-time monitoring of the production environment is achieved. In addition, the integration of safety regulations, standards, historical accident reports, safety logs, operating manuals and equipment maintenance records obtained from safety supervision departments and internal enterprise management systems provides a rich data source for subsequent analysis, significantly improves the ability to identify potential safety hazards, and lays a data foundation for accurate early warning, thereby achieving the beneficial effects of improving safety supervision efficiency and preventing accidents.

[0052] Furthermore, preprocessing the data and then integrating and storing the data into a database refers to using natural language processing technology to preprocess the text data, including word segmentation, removal of stop words, part-of-speech tagging and entity recognition, converting the preprocessed text into structured data, normalizing the collected real-time data, using a box plot method to identify and remove outliers, integrating the preprocessed historical data and real-time data into a unified database for storage, and regularly backing up the stored data to the cloud.

[0053] Natural language processing technology is a branch of computer science and artificial intelligence that aims to enable computers to understand, interpret and manipulate human language. The box plot method is a statistical chart used to describe data distribution characteristics. It can intuitively display the median, quartiles, extreme values ​​and other information of the data. It is also an effective outlier detection method. The collected text data is preprocessed using natural language processing technology, including word segmentation, stop word removal, part-of-speech tagging and entity recognition, and real-time monitoring data is normalized and outlier processed, unstructured text is converted into structured data, and all preprocessed data is integrated and stored in a unified database. Regular cloud backup ensures data security and reliability, which not only optimizes the data storage structure and improves data processing efficiency, but also creates conditions for subsequent in-depth analysis and knowledge extraction, thereby achieving the beneficial effect of improving data utilization and analysis accuracy.

[0054] Furthermore, building a knowledge graph based on a database includes the following steps:

[0055] Use the BERT model to identify key entities in the text of historical data and label the entity types;

[0056] Construct a neural network model based on the attention mechanism for relationship extraction:

[0057] , where is the relationship extraction result vector, which represents the probability distribution of various possible relationships. It is the Softmax function, which is used to convert the relationship extraction results into probability distribution. is the number of key entities in the text, is the transpose of the weight vector, which is used to convert The result of the activation function is converted to a real number, is the hyperbolic tangent activation function, is the weight matrix used to transform the hidden state Mapped to a space of lower or equal dimensionality for importance scoring, obtained by training a neural network model, is the bias term, is the hidden state vector of the qth entity after being encoded by the BERT model, refers to the hidden state vector of the jth entity obtained after encoding the entity, It is the weight matrix of the relation extraction task, which is used to map the hidden state to the relation space and is obtained by training the neural network model. is an exponential function;

[0058] The graph database is used to construct a knowledge graph in the field of safe production based on the extracted entities and relationships.

[0059] The BERT model is a pre-trained language representation model based on the Transformer architecture. The nodes in the graph represent entities such as equipment and components, and the edges represent the relationships between entities, such as causes and belongs to. The BERT model is used to identify key entities in historical data and label entity types, and is combined with a neural network model based on the attention mechanism for relationship extraction. Finally, a knowledge graph in the field of safe production is constructed through a graph database. This not only improves the accuracy of entity recognition and relationship extraction, but also uses the knowledge graph constructed by the graph database to enable the safety management system to fully understand and master the professional knowledge in the production field, including equipment characteristics, operating specifications, safety standards, etc. The construction of the knowledge graph enables the safety early warning system to provide more accurate accident predictions based on in-depth domain knowledge, realizes the transformation from data to knowledge, and provides strong support for the formulation of targeted emergency response measures, thereby achieving the beneficial effect of enhancing the decision-making support capabilities for safe production.

[0060] S2, anomaly detection module, used to perform anomaly detection based on preprocessed real-time monitoring data;

[0061] Specifically, anomaly detection based on preprocessed real-time monitoring data includes the following steps:

[0062] For each data point in the real-time data stream , calculate the weighted moving mean and weighted moving standard deviation of the data points:

[0063] , , where is the weighted moving average of the most recent N data points including the current data point on the dth feature dimension, is the weight coefficient of the i-th data point, which can be set according to the distance of the data point from the current moment. is the value of the ith data point in the dth feature dimension, It is the weighted moving standard deviation of the most recent N data points including the current data point on the dth feature dimension;

[0064] Compute the anomaly score for each dimension of the data separately:

[0065] , where is the comprehensive anomaly score of the tth data point, D is the number of feature dimensions of the data, is the value of the t-th data point on the d-th feature dimension, is the weighted moving average, is the weighted moving standard deviation;

[0066] The anomaly threshold is defined as:

[0067] , where is the abnormal threshold at time t, is the anomaly detection score of the ith data point, is the number of anomaly scores used to calculate the moving mean and moving standard deviation, is the moving average of the historical anomaly scores, is the regulating factor;

[0068] The abnormal threshold and anomaly detection results Compare and judge the abnormality of data.

[0069] like , it means that the data is abnormal. , it means that there is no abnormality in the data.

[0070] By performing anomaly detection on pre-processed real-time monitoring data, we can instantly identify potential risks in the production environment. We use weighted moving averages and standard deviations to calculate the anomaly scores of data points, and use historical data and safety standards to set anomaly thresholds, thus achieving a comprehensive evaluation of the data. This not only improves the accuracy of anomaly detection, but also reduces reliance on manual monitoring through real-time monitoring, significantly improving the efficiency and response speed of safety management.

[0071] S3, accident prediction module, used to predict accidents and conduct risk assessment based on anomaly detection results;

[0072] Specifically, accident prediction and risk assessment based on anomaly detection results include the following steps:

[0073] Use autoencoders to extract deep features of abnormal data, use clustering algorithms to classify the extracted features, extract knowledge graph features through graph neural networks, use statistical methods to extract vectors of real-time data, and comprehensively build an accident prediction model:

[0074] , where is the probability of an accident happening, is the activation function, , and They are respectively the results of deep clustering , Knowledge Graph Features Extracted by Graph Neural Network , real-time data features The corresponding weight vector, is a bias term used to adjust the output probability of the model. Represents a transpose operation;

[0075] The abnormal data is input into the accident prediction model, which automatically processes the input data and outputs the predicted probability of an accident;

[0076] Create a risk matrix, derive the severity of the accident through knowledge graph analysis, quantify the probability of an accident and the severity of the accident, and determine different risk levels based on the combination of the two;

[0077] Calculate the risk index of an accident:

[0078] , where is the risk index of the accident, is the probability of an accident happening, is the potential severity of the incident, which can be based on historical data, expert assessments, or standardized risk rating tables, is the weight coefficient of the lth additional risk factor, is the quantitative value of the lth additional risk factor, is the total number of additional risk factors, is the weight coefficient;

[0079] The overall risk index and risk judgment threshold and Compare and judge the accident risk level, among which < ,like Less than , then it is low risk, if ≤ < , then it is medium risk, if ≥ , then high risk.

[0080] Autoencoder is an unsupervised learning algorithm, usually used for feature extraction and dimensionality reduction. Clustering algorithm is an unsupervised learning method for grouping data. Graph neural network is a neural network that processes graph structured data. According to the anomaly detection results, combined with advanced artificial intelligence technologies such as autoencoder, clustering algorithm and graph neural network, it is possible to extract deep features from abnormal data, and predict the probability of accidents by building an accident prediction model. At the same time, by creating a risk matrix, the severity and risk level of the accident are quantified, and a comprehensive assessment of potential accident risks is achieved. It not only improves the early warning system's ability to predict accidents, but also provides a clear basis for subsequent emergency response through the division of risk levels, ensuring the pertinence and effectiveness of early warning measures.

[0081] S4, emergency response module, used to generate early warning information and formulate emergency response measures;

[0082] Specifically, generating early warning information and formulating emergency response measures means determining the generation of early warning information and formulating corresponding emergency response measures according to the risk level.

[0083] If the risk level is low, no warning information will be generated, and daily monitoring of low-risk areas will continue, and detailed information on low-risk situations will be recorded, and safety inspections of low-risk areas will be carried out regularly;

[0084] If the risk level is medium or high, early warning information will be generated, including the type of accident, risk level, cause and response measures. The early warning information will be released to relevant managers and operators through the communication network. Emergency response measures will be formulated according to the recommended measures in the early warning information, including isolation and safety inspection of potential risk areas, emergency shutdown, evacuation of personnel and adjustment of production parameters.

[0085] Early warning information is directly generated according to the risk level, and emergency response measures are formulated according to the recommended measures in the early warning information, achieving rapid response from medium and high risk levels, ensuring timely isolation and safety inspection of potential risk areas, as well as shutdown and evacuation measures in emergency situations. This process greatly reduces the possibility of accidents, improves the company's ability to respond to emergencies, and ensures production safety and the safety of employees' lives.

[0086] S5, evaluation and optimization module, used to evaluate system performance and optimize;

[0087] Specifically, evaluating system performance and optimizing it means collecting feedback information from relevant personnel and systems after executing emergency response measures, analyzing the implementation and effectiveness of emergency response measures, evaluating the accuracy and timeliness of early warning information, adjusting the accident prediction model parameters and optimizing the process accordingly based on the feedback analysis results, and updating the feedback analysis results to the knowledge graph.

[0088] By collecting feedback information after executing emergency response measures, the system performance is evaluated and optimized. Not only the accuracy and timeliness of the warning information are examined, but also the accident prediction model parameters are adjusted according to the feedback analysis results, and the entire warning process is optimized. By updating the feedback results to the knowledge graph, this step realizes the continuous learning and improvement of the early warning system, ensuring that the system continues to adapt to changes in the production environment over time and remains efficient and accurate.

[0089] Example 2

[0090] This is the second embodiment of the present invention, and the difference between this embodiment and the previous embodiment is that:

[0091] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0092] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0093] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0094] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit with a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit with a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

Claims

1. An early warning system for production safety accidents based on artificial intelligence. Features: include, The data collection module is used to deploy sensors to collect production environment monitoring data and historical data in real time, integrate and store the data into the database after preprocessing, and build a knowledge graph based on the database; Anomaly detection module, used to perform anomaly detection based on preprocessed real-time monitoring data; The accident prediction module is used to predict accidents and conduct risk assessment based on the anomaly detection results; Emergency response module, used to generate early warning information and formulate emergency response measures; Evaluation and optimization module, used to evaluate system performance and optimize it; The construction of a knowledge graph based on a database includes the following steps: Use the BERT model to identify key entities in the text of historical data and label the entity types; Construct a neural network model based on the attention mechanism for relationship extraction: , where is the relation extraction result vector, is the Softmax function, is the number of key entities in the text, is the transpose of the weight vector, is the hyperbolic tangent activation function, is the weight matrix, is the bias term, is the hidden state vector of the qth entity after being encoded by the BERT model, refers to the hidden state vector of the jth entity obtained after encoding the entity, is the weight matrix of the relation extraction task, is an exponential function; Use graph database to build knowledge graph in the field of safe production based on extracted entities and relationships; The anomaly detection based on the pre-processed real-time monitoring data includes the following steps: For each data point in the real-time data stream , calculate the weighted moving mean and weighted moving standard deviation of the data points: , , where is the weighted moving average of the most recent N data points including the current data point on the dth feature dimension, is the weight coefficient of the ith data point, is the value of the ith data point in the dth feature dimension, It is the weighted moving standard deviation of the most recent N data points including the current data point on the dth feature dimension; Compute the anomaly score for each dimension of the data separately: , where is the comprehensive anomaly score of the tth data point, D is the number of feature dimensions of the data, is the value of the t-th data point on the d-th feature dimension, is the weighted moving average, is the weighted moving standard deviation; The anomaly threshold is defined as: , where is the abnormal threshold at time t, is the anomaly detection score of the ith data point, is the number of anomaly scores, is the moving average of the historical anomaly scores, is the regulating factor; The abnormal threshold and anomaly detection results Compare and judge the abnormality of data. like , it means that the data is abnormal. , it means that there is no abnormality in the data; The accident prediction and risk assessment based on the abnormal detection results include the following steps: Use autoencoders to extract deep features of abnormal data, use clustering algorithms to classify the extracted features, extract knowledge graph features through graph neural networks, use statistical methods to extract vectors of real-time data, and comprehensively build an accident prediction model: , where is the probability of an accident happening, is the activation function, , and They are respectively the results of deep clustering , Knowledge Graph Features Extracted by Graph Neural Network and real-time data features The corresponding weight vector, is the bias term, Represents a transpose operation; The abnormal data is input into the accident prediction model, which automatically processes the input data and outputs the predicted probability of an accident; Create a risk matrix, derive the severity of the accident through knowledge graph analysis, quantify the probability of an accident and the severity of the accident, and determine different risk levels based on the combination of the two; Calculate the risk index of an accident: , where is the risk index of the accident, is the probability of an accident happening, is the potential severity of the accident, is the weight coefficient of the lth additional risk factor, is the quantitative value of the lth additional risk factor, is the total number of additional risk factors, is the weight coefficient; The overall risk index and risk judgment threshold and Compare and judge the accident risk level, among which < ,like Less than , then it is low risk, if ≤ < , then it is medium risk, if ≥ , then high risk; Generating early warning information and formulating emergency response measures refers to determining the generation of early warning information and formulating corresponding emergency response measures according to the risk level. If the risk level is low, no warning information will be generated, and daily monitoring of low-risk areas will continue, and detailed information on low-risk situations will be recorded, and safety inspections of low-risk areas will be carried out regularly; If the risk level is medium or high, early warning information will be generated, including the type of accident, risk level, cause and response measures. The early warning information will be released to relevant managers and operators through the communication network. Emergency response measures will be formulated according to the recommended measures in the early warning information, including isolation and safety inspection of potential risk areas, emergency shutdown, evacuation of personnel and adjustment of production parameters.

2. The artificial intelligence-based production safety accident hazard early warning system as claimed in claim 1, Features: The deployment of sensors to collect production environment monitoring data and historical data in real time refers to deploying loT sensors to collect production environment monitoring data and equipment parameters in real time, obtaining safety regulations, standards and historical accident reports from safety supervision departments, and obtaining safety logs, operation manuals and equipment maintenance records from the company's internal management system.

3. The artificial intelligence-based production safety accident hazard early warning system as claimed in claim 2, Features: The preprocessing of data and then integrating and storing the data into a database refers to preprocessing text data using natural language processing technology, including word segmentation, removal of stop words, part-of-speech tagging and entity recognition, converting the preprocessed text into structured data, normalizing the collected real-time data, using a box plot method to identify and remove outliers, integrating the preprocessed historical data and real-time data into a unified database for storage, and regularly backing up the stored data to the cloud.

4. The artificial intelligence-based safety production accident hazard early warning system as claimed in claim 3, Features: The evaluation of system performance and optimization refers to collecting feedback information from relevant personnel and systems after executing emergency response measures, analyzing the implementation and effectiveness of emergency response measures, evaluating the accuracy and timeliness of early warning information, adjusting the accident prediction model parameters and optimizing the process accordingly based on the feedback analysis results, and updating the feedback analysis results to the knowledge graph.