Artificial intelligence-based production safety management system, method, and storage medium

Abstract

The application discloses an artificial intelligence-based production safety (including occupational health) management system and method and a storage medium. The production safety management platform is configured to perform the following operations: determining hazard information related to a production process according to production element information; determining a corresponding question vector according to the hazard information by using a question vector model based on natural language processing, the question vector being used to indicate a question in the production process corresponding to the hazard information; determining a solution vector corresponding to the question vector by using a solution vector model based on natural language processing; and determining suggestion information corresponding to the hazard information according to the solution vector. Thus, the technical effect of fully exerting the control mechanism, learning mechanism and system dynamics mechanism of artificial intelligence technology in production safety management is achieved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based production safety management system, method, and storage medium. Background Technology

[0002] Currently, production safety has received increasing attention in production management. For example, the internationally recognized HSE management system is being used more and more widely, and production safety management systems based on HSE management standards are also being developed vigorously. These systems utilize positioning systems, environmental monitoring systems, and other detection systems to collect information related to production safety, and by comparing the collected information with HSE management system standards, identify potential health and safety hazards that could lead to dangers during production.

[0003] The published patent CN115271407A provides an industrial internet data processing method and system based on artificial intelligence. It obtains a first event capture report by capturing risk events from the current cloud-based business interaction information of a digital intelligent production line. Then, based on the prior capture report indications of the digital intelligent production line, the first event capture report is adjusted to obtain a first tracking analysis report. Finally, the first tracking analysis report is verified according to the knowledge unit relationship network corresponding to the prior capture report indications of potential risk interaction events, resulting in a first verified analysis report. This enables the linked analysis and processing of the first event capture report and the prior capture report indications of the digital intelligent production line, and fully explores the correlation characteristics between the two. Therefore, the resulting first verified analysis report can record and output the descriptive vector of potential risk interaction events as accurately, completely, and reliably as possible.

[0004] The published patent CN115022598A proposes an artificial intelligence video monitoring system for hydropower safety, characterized by comprising: an electronic fence alarm system, a personnel clothing recognition system, a leak detection system, a smoke and fire alarm system, an instrument recognition system, an operational risk monitoring system, an emergency command large screen system, and an artificial intelligence video monitoring platform for hydropower safety production; wherein the electronic fence alarm system, personnel clothing recognition system, leak detection system, smoke and fire alarm system, instrument recognition system, and operational risk monitoring system are all connected to the artificial intelligence video monitoring platform for hydropower safety production via an internal network.

[0005] However, current production safety management systems based on HSE management standards are not performing ideally in practice. These systems primarily identify health and safety hazards in the production process by comparing the results to established management standards. Therefore, they cannot make judgments regarding aspects not covered by HSE management standards. In other words, HSE management standards fail to fully utilize the control, learning, and system dynamics mechanisms of artificial intelligence (AI) in production safety management. This limits the in-depth application of current AI technology in production safety management.

[0006] There are currently no effective solutions to the technical problems that existing production safety management systems cannot flexibly determine health and safety hazards in the production process, and that artificial intelligence technology cannot be deeply applied to production management systems. Summary of the Invention

[0007] The embodiments of this disclosure provide an artificial intelligence-based production safety management system, method, and storage medium to at least solve the technical problems existing in the prior art where production safety management systems cannot make flexible judgments on production hazards and where artificial intelligence technology cannot be deeply applied to production management systems.

[0008] According to one aspect of this disclosure, an artificial intelligence-based production safety management system is provided, including a production information acquisition system and a production safety management platform. The production information acquisition system is communicatively connected to the production safety management platform and is used to collect production element information related to the production process and send the collected production element information to the production safety management platform. The production safety management platform is configured to perform the following operations: determine hazard information related to the production process based on the production element information; determine corresponding problem vectors based on the hazard information using a natural language processing-based problem vector model, whereby the problem vectors indicate problems corresponding to the hazard information in the production process; determine solution vectors corresponding to the problem vectors using a natural language processing-based solution vector model; and determine suggestion information corresponding to the hazard information based on the solution vectors.

[0009] According to another aspect of the present disclosure, an artificial intelligence-based production safety management method is provided, comprising: determining hazard information related to the production process based on production factor information related to the production process; determining corresponding problem vectors based on the hazard information using a problem vector model based on natural language processing, wherein the problem vectors are used to indicate problems corresponding to the hazard information in the production process; determining solution vectors corresponding to the problem vectors using a solution vector model based on natural language processing; and determining suggestion information corresponding to the hazard information based on the solution vectors, and sending the suggestion information to the terminal devices of relevant personnel.

[0010] According to another aspect of the present disclosure, a storage medium is also provided, the storage medium including a stored program, wherein the methods described above are executed by a processor when the program is running.

[0011] According to another aspect of the present disclosure, an artificial intelligence-based production safety management device is also provided, comprising: a hazard information determination module, configured to determine hazard information related to the production process based on production factor information related to the production process; a problem vector determination module, configured to determine a corresponding problem vector based on the hazard information using a problem vector model based on natural language processing, wherein the problem vector indicates a problem in the production process corresponding to the hazard information; a solution vector determination module, configured to determine a solution vector corresponding to the problem vector using a solution vector model based on natural language processing; and a suggestion information determination module, configured to determine suggestion information corresponding to the hazard information based on the solution vector and send the suggestion information to the terminal device of relevant personnel.

[0012] According to another aspect of the present disclosure, an artificial intelligence-based production safety management device is also provided, including: a processor; and a memory connected to the processor, configured to provide the processor with instructions to process the following steps: determining hazard information related to the production process based on production element information related to the production process; determining a corresponding problem vector based on the hazard information using a problem vector model based on natural language processing, wherein the problem vector is used to indicate a problem in the production process corresponding to the hazard information; determining a solution vector corresponding to the problem vector using a solution vector model based on natural language processing; and determining suggestion information corresponding to the hazard information based on the solution vector, and sending the suggestion information to the terminal device of relevant personnel.

[0013] The technical solution disclosed herein is based on natural language processing (NLP) technology and introduces problem vectors and solution vectors. The problem vectors indicate the problems corresponding to existing hazards in the production process, while the solution vectors are associated with suggested information corresponding to the hazard information. Furthermore, the problem vectors can constitute a problem space, and the solution vectors can constitute a solution space that maps to the problem space. Both the problem vectors and solution vectors are generated using a NLP-based model. Therefore, according to the technical solution disclosed herein, the NLP-based problem vector model and solution vector model can be trained based on actual hazard information, making the problem space constructed by the problem vectors and the solution space constructed by the solution vectors more flexibly adaptable to current production safety management. This fully leverages the control, learning, and system dynamics mechanisms of artificial intelligence technology in production safety management. It solves the technical problems of existing production safety management systems' inability to flexibly determine production hazards and the limited application of artificial intelligence technology in production management systems. Attached Figure Description

[0014] The accompanying drawings, which are included to provide a further understanding of this disclosure and form part of this application, illustrate exemplary embodiments of the present disclosure and are used to explain the disclosure, but do not constitute an undue limitation thereof. In the drawings:

[0015] Figure 1 This is a schematic diagram of an artificial intelligence-based production safety management system according to the first aspect of Embodiment 1 of this disclosure;

[0016] Figure 2A This is a hardware structure block diagram of the production safety management platform according to the first aspect of Embodiment 1 of this disclosure;

[0017] Figure 2B This is a block diagram of an artificial intelligence-based production safety management system according to the first aspect of Embodiment 1 of this disclosure;

[0018] Figure 2C This is a block diagram of the production safety management platform according to the first aspect of Embodiment 1 of this disclosure;

[0019] Figure 3 This is a schematic diagram of the operation process of the production safety management platform according to the first aspect of Embodiment 1 of this disclosure;

[0020] Figure 4 This is a schematic diagram illustrating the approach to production safety management based on the problem space and solution space disclosed in this technical solution.

[0021] Figure 5A This is a schematic diagram of the feature extraction module according to the first aspect of Embodiment 1 of this disclosure;

[0022] Figure 5B This is a schematic diagram of the various modules and units for generating a solution vector based on hazard information according to the first aspect of Embodiment 1 of this disclosure;

[0023] Figure 5C This is a schematic diagram of various models for generating solution vectors based on hazard information according to the first aspect of Embodiment 1 of this disclosure;

[0024] Figure 6A This is a block diagram of a production safety management platform according to a variation of the first aspect of Embodiment 1 of this disclosure;

[0025] Figure 6B This is a schematic diagram of a feature extraction module according to a variation of the first aspect of Embodiment 1 of this disclosure;

[0026] Figure 6C This is a schematic diagram of the various modules and units that generate a solution vector based on hazard information in a variation of the first aspect of Embodiment 1 of this disclosure;

[0027] Figure 6D This is a schematic diagram of various models for generating solution vectors based on hazard information in a variation of the first aspect of Embodiment 1 of this disclosure;

[0028] Figure 7 This is a flowchart illustrating the production safety management method based on artificial intelligence according to the second aspect of Embodiment 1 of this disclosure;

[0029] Figure 8 This is a schematic diagram of an artificial intelligence-based production safety management device according to Embodiment 2 of this disclosure; and

[0030] Figure 9 This is a schematic diagram of an artificial intelligence-based production safety management device according to Embodiment 3 of this disclosure. Detailed Implementation

[0031] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this disclosure.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0033] Example 1

[0034] According to this embodiment, an artificial intelligence-based production safety management system is provided. Figure 1 A schematic diagram of an AI-based production safety management system according to this embodiment is shown. Figure 2B A block diagram of the production safety management system is shown. (Reference) Figure 1 and Figure 2B As shown, the production safety management system includes: a production information acquisition system 100, a production safety management platform 200, and terminal devices 300 for relevant personnel. The production information acquisition system 100 is communicatively connected to the production safety management platform 200, and is used to collect production element information related to the production process and send the collected production element information to the production safety management platform 200. The production safety management platform 200 can receive the production element information collected by the production information acquisition system 100, and based on the collected production element information, generate production safety-related suggestions and send these suggestions to the terminal devices 300 for relevant personnel.

[0035] Furthermore, Figure 2A The hardware structure block diagram of the production safety management platform is shown. (For example...) Figure 2A As shown, the production safety management platform 200 may include one or more processors (processors may include, but are not limited to, microprocessors such as MCUs or programmable logic devices such as FPGAs), a memory for storing data, a transmission device for communication functions, and input / output interfaces. The memory, transmission device, and input / output interfaces are connected to the processor via a bus. Those skilled in the art will understand that... Figure 2A The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, the production safety management platform 200 may also include... Figure 2A The more or fewer components shown, or having the same Figure 2A The different configurations shown.

[0036] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits can be implemented wholly or partially as software, hardware, firmware, or any other combination. Furthermore, the data processing circuits can be a single, independent processing module, or wholly or partially integrated into any other element within the production safety management platform 200. As involved in the embodiments of this disclosure, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0037] The memory can be used to store application software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory may include high-speed random access memory (RAM) and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, which can be connected to the production safety management platform 200 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks (LANs), mobile communication networks, and combinations thereof.

[0038] The transmission device is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication vendor of the production safety management platform 200. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0039] It should be noted here that, in some optional embodiments, the above... Figure 2A The production safety management platform 200 shown may include hardware components (including circuitry), software components (including computer code stored on a computer-readable medium), or a combination of both hardware and software components. It should be noted that... Figure 2A This is just one example of a specific instance and is intended to illustrate the types of components that may exist in the aforementioned production safety management platform.

[0040] In addition, refer to Figure 1 and Figure 2B As shown, the production information collection system 100 includes on-site production element detection systems 110-150, employee information entry terminals 160, and edge devices 170.

[0041] Among them, the on-site production factor detection system 110-150 is installed in the workplace to detect on-site production factor information related to the production process. Specifically, refer to... Figure 1 and Figure 2B As shown, the on-site production element detection system may include, for example,: a hazardous substance detection system 110, used to detect the status information of hazardous substances (such as toxic gases) on the production site, such as the presence of hazardous substances and their concentration; a hazardous energy detection system 120, used to detect the status information of hazardous energy (such as thermal energy and electrical energy) on the production site, such as the presence of hazardous energy, electrical energy, etc., and the magnitude of hazardous energy; and a work equipment detection system 130, used to detect the status information of work equipment on the production site (e.g., the strength, rigidity, sealing, and operating status of components during production, such as speed and pressure). The system includes: a hazardous organism detection system 140, used to detect the presence of hazardous organisms (e.g., pathogenic microorganisms, bacteria, viruses, fungi, dangerous animals, and dangerous plants) in the production site, such as the presence and density of hazardous organisms; and a work layout detection system 150, used to detect the status of the work layout in the production site (e.g., whether the indoor floor is slippery, the indoor work area is cluttered, ladder defects, and building foundation settlement). These detection systems can be implemented using sensors installed in the production site or by personnel carrying the appropriate detection equipment. Additionally, see reference [link to relevant documentation]. Figure 1 and Figure 2B As shown, the edge device 170 is set up at the production site and is used to transmit the on-site production element information detected by the on-site production element detection system 110-150 to the production safety management platform 200.

[0042] In addition, the employee information entry terminal 160 is communicatively connected to the production safety management platform 200, and is used to enter employee production element information related to the production process into the production safety management platform 200. For example, staff can enter employee production element information related to the production process through the employee information entry terminal 160, including: the employee's work intensity, working hours, health information, emotional information, and hazardous behavior information in the work process.

[0043] also Figure 2C A schematic diagram of the program modules of the production safety management platform 200 is further shown. See also Figure 2CAs shown, the production safety management platform 200 includes an on-site production element database 210, an employee production element database 220, a statistical information database 230, a feature extraction module 240, a problem vector module 250, a solution vector module 260, and a suggestion information generation module 270. These program modules will be described in detail below.

[0044] also, Figure 3 A flowchart illustrating the operation of the production safety management platform 200 is shown. (Reference) Figure 3 As shown, the production safety management platform 200 is configured to perform the following operations:

[0045] S302: Based on production factor information, determine the hazard information related to the hazardous process;

[0046] S304: Using a natural language processing-based question vector model, the corresponding question vector is determined based on the hazard information. The question vector is used to indicate the problems in the production process that correspond to the hazard information.

[0047] S306: Utilize a solution vector model based on natural language processing to determine the solution vector corresponding to the problem vector; and

[0048] S308: Determine the suggested information corresponding to the hazard information based on the solution vector, and send the suggested information to the terminal equipment of relevant personnel.

[0049] Specifically, see Figure 2C As shown, the production safety management platform 200 receives on-site production element information collected by the on-site production element detection systems 110-150 from the edge device 170 and stores it in the on-site production element database 210; it also receives employee production element information related to the production process from the employee information entry terminal 160 and stores it in the employee production element database 220. Then, the statistical information database 230 performs statistical analysis and organization on the information in the on-site production element database 210 and the employee production element database 220. Thus, the feature extraction module 240 can extract hazard information corresponding to a specific work unit, department, or employee from the statistical information database 230 (S302) and generate corresponding feature vectors based on the hazard information.

[0050] Specifically, refer to Figure 5A As shown, the feature extraction module 240 includes an information extraction unit 241, a first feature encoding unit 242a, a second feature encoding unit 242b, a feature fusion unit 243, and a feature vector unit 244. The information extraction unit 241 extracts hazard information corresponding to a specific work unit, department, or employee from the statistical information database 230.

[0051] Then, the question vector module 250 uses a natural language processing-based question vector model to determine the question vector corresponding to the hazard information (S304). This question vector indicates the problem corresponding to the hazard information during the production process. Furthermore, this question vector is generated, for example, based on a fused feature vector, thus containing semantic features related to the problem corresponding to the hazard information. Further details regarding the question vector module 250 and the question vector will be explained in detail later.

[0052] Then, the solution vector module 260 receives the question vector generated by the question vector module 250 and uses a solution vector model based on natural language processing to generate a solution vector corresponding to the question vector (S306). This solution vector is associated with the suggestion information corresponding to the danger information. This solution vector is also generated, for example, based on a fused feature vector. Further details about the solution vector module 260 and the solution vector will be explained in detail later.

[0053] Then, the suggestion information generation module 270 determines the suggestion information corresponding to the solution vector generated by the solution vector module 260 (S308). The production safety management platform sends the suggestion information to the terminal devices 300 of relevant personnel to guide them in making improvements, so as to eliminate hazards in the production process as soon as possible and avoid subsequent production safety accidents.

[0054] As described in the background section, current production safety management systems based on HSE (Health, Safety, and Environment) management standards are not performing ideally during operation. These systems primarily identify safety hazards in the production process by comparing the results to established management standards. Therefore, they cannot make appropriate judgments regarding aspects not covered by HSE management standards. In other words, HSE management standards fail to fully leverage the control, learning, and system dynamics mechanisms of artificial intelligence (AI) technology in production safety management. Consequently, current AI technology cannot be deeply applied to production safety management.

[0055] In view of this, the technical solution disclosed herein is based on natural language processing technology and introduces question vectors and solution vectors. Among them... Figure 4 A schematic diagram illustrating the concept of the technical solution of this disclosure is shown. (Reference) Figure 4As shown, according to the technical solution of this disclosure, production factor information (e.g., on-site production factor information and / or employee production factor information) is first obtained from information on the controlled object (e.g., energy / hazardous substances, employee actions / behaviors / statuses, etc.). Then, a corresponding problem vector is determined in the problem space based on the hazard information. Next, based on the spatial mapping between the vector space and the solution space, a solution vector corresponding to the hazard information is determined according to the problem vector. Then, suggested information corresponding to the hazard information is determined according to the solution vector. Finally, the suggested information is provided to relevant personnel for appropriate management of the controlled object to avoid production safety accidents.

[0056] The problem vector indicates the problem corresponding to the hazard information existing in the production process, while the solution vector is associated with the corresponding suggestion information. Furthermore, the problem vectors can form a problem space, and the solution vectors can form a solution space that has a mapping relationship with the problem space. Both the problem vectors and solution vectors are generated using a natural language processing (NLP)-based model. Therefore, according to the technical solution of this disclosure, the NLP-based problem vector model and solution vector model can be trained based on actual hazard information, making the problem space formed by the problem vectors and the solution space formed by the solution vectors more flexibly adaptable to current production safety management. This fully leverages the control mechanism, learning mechanism, and system dynamics mechanism of artificial intelligence technology in production safety management. It solves the technical problems of existing production safety management systems being unable to make flexible judgments on production hazards and the inability of artificial intelligence technology to be deeply applied to production management systems.

[0057] Optionally, the operation of determining hazard information related to the production process based on production factor information includes: statistically analyzing production factor information associated with the target object from the production factor information according to preset statistical rules; and determining hazard information related to the production process of the target object based on the production factor information associated with the target object.

[0058] Specifically, see Figure 2C As shown, the production safety management platform 200 is equipped with a statistical information database 230. The statistical information database 230 receives on-site production factor information detected by the on-site production factor detection systems 110-150 from the on-site production factor database 210, and receives employee production factor information related to the production process from the employee production factor database 220, and then sorts and compiles the data.

[0059] Therefore, the statistical information database 230 can, according to preset statistical rules, collect on-site production factor information and / or employee production factor information associated with the target object. For example, the statistical information database 230 can collect on-site production factor information and employee production factor information associated with a specified unit, or it can collect on-site production factor information and employee production factor information associated with a specified department, or it can collect on-site production factor information and / or employee production factor information associated with a specified employee.

[0060] Thus, the information extraction unit 241 of the feature extraction module 240 can extract on-site production factor information and employee production factor information associated with a specified unit, department or employee from the statistical information database 230, as production factor information of the specified unit, department or employee in the production process.

[0061] Then, the information extraction unit 241 determines the hazard information from the extracted production element information according to preset rules. For example, there are hazardous substances on site and / or the concentration of hazardous substances reaches a predetermined threshold; there is hazardous energy on site and / or the magnitude of hazardous energy reaches a predetermined threshold; the strength, rigidity, or sealing of the on-site work equipment is lower than a predetermined threshold; there are hazardous organisms on site and / or the density of hazardous organisms reaches a predetermined threshold; the indoor floor is slippery, the indoor work area is cluttered, the ladders are defective, or the building foundation is subsiding. Thus, the hazard information extraction module can determine the hazard information associated with a designated unit, department, or employee from the production element information for further analysis.

[0062] Therefore, based on the technical solution disclosed herein, targeted hazard information can be collected and analyzed for specific targets (such as designated units, departments, or employees), and further production safety management can be carried out.

[0063] Optionally, the operation of determining the hazard information related to the production process of the target object based on the production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on the on-site production factor information associated with the target object, thereby forming a hazard information sequence.

[0064] The on-site production factor information includes information on production factors other than employees during the production process. Preferably, the on-site production factor information includes at least one of the following: the status information of hazardous substances on the production site; the status information of hazardous energy on the production site; the status information of working equipment on the production site; the status information of hazardous organisms present on the production site; and the status information of the work layout on the production site.

[0065] Preferably, the information extraction unit 241 of the feature extraction module 240 can extract on-site production factor information associated with a designated unit, department, or employee within a preset period from the statistical information database 230, according to a preset period. For example, the information extraction unit 241 can extract on-site production factor information associated with a designated unit, department, or employee within the most recent month. This on-site production factor information may be, for example, information detected by the on-site production factor detection systems 110-150. This includes: status information of hazardous substances (such as toxic gases) detected at the production site by the hazardous substance detection system 110, such as the presence and concentration of hazardous substances; status information of hazardous energy (such as thermal and electrical energy) detected at the production site by the hazardous energy detection system 120, such as the presence and magnitude of hazardous energy; status information of work equipment at the production site detected by the work equipment detection system 130 (e.g., the strength, rigidity, sealing, and operating status of components during production, such as speed and pressure); status information of hazardous organisms (such as pathogenic microorganisms, bacteria, viruses, fungi, dangerous animals, and dangerous plants) detected at the production site by the hazardous organism detection system 140, such as the presence and density of hazardous organisms; and status information of the work layout at the production site detected by the work layout detection system 150 (e.g., whether the indoor floor is slippery, the clutter of the indoor work area, defects in ladders and scaffolds, and the settlement of the building foundation).

[0066] Then, the information extraction unit 241 determines the hazard information associated with a designated unit, department, or employee from the extracted on-site production element information. For example, hazard information occurring in a production site managed by a designated unit or department, or hazard information arising from equipment or layout work in a production site managed by a designated employee.

[0067] Then, the information extraction unit 241 sorts the determined danger information in chronological order to form a danger information sequence.

[0068] Therefore, according to the technical solution disclosed herein, the occurrence of hazard information of a target object within a predetermined period (e.g., within one month, one week, or one quarter) can be statistically analyzed in chronological order and sorted accordingly. This helps to identify the problems related to the hazard information based on the correlation between the hazard information occurring in the target object within the predetermined period, and facilitates the formulation of more targeted recommendations.

[0069] Optionally, a question vector model based on natural language processing is used to determine corresponding question vectors based on hazard information. The question vectors are used to indicate the operations related to the hazard information in the production process, including: determining first feature codes corresponding to different hazard information in the hazard information sequence from a pre-set hot unique code encoding library; determining the employees corresponding to the hazard information; determining second feature codes corresponding to the determined employee's production factor information from the hot unique code encoding library; fusing the first and second feature codes to generate fused feature codes corresponding to different hazard information, thereby forming a fused feature code sequence; generating a fused feature vector sequence corresponding to the fused feature code sequence through a word embedding-based word vector model; and generating a question vector sequence corresponding to the fused feature vector sequence through an attention-based question vector model.

[0070] Specifically, after the information extraction unit 241 of the feature extraction module 240 extracts the hazard information from the on-site production element information, the first feature encoding unit 242a determines the hot unique code (i.e. the first feature code) corresponding to the hazard information from the pre-set hot unique code encoding library according to the hazard information in the hazard information sequence.

[0071] Table 1 below shows the types of different danger information set in the hot unique code encoding library and their corresponding hot unique codes:

[0072] Table 1

[0073]

[0074]

[0075] Where, r i (i≥1) represents the hot unique code corresponding to the danger information, for example:

[0076] The hot unique code corresponding to the danger message "Insufficient equipment strength" is:

[0077] r1 = [1 0 0 0 0 0......] T

[0078] The thermal unique code corresponding to the hazard message "Insufficient equipment rigidity" is:

[0079] r2 = [0 1 0 0 0 0......] T

[0080] And so on.

[0081] Therefore, the first feature encoding unit 242a generates a hot unique code (i.e., the first feature code) corresponding to the danger information in the danger information sequence extracted by the information extraction unit 241. For example, the danger information sequence received by the first feature encoding unit 242a from the information extraction unit 241 is shown in Table 2 below:

[0082] Table 2

[0083]

[0084]

[0085] Then, the first feature encoding unit 242a can determine the hot unique code sequence Rs corresponding to the danger information sequence Is={I1,I2,I3,I4,I5,I6,I7,...} from the hot unique code encoding library shown in Table 1:

[0086] Rs={R1,R2,R3,R4,R5,R6,R7,...}={r1,r4,r3,r4,r 50 ,r 54 ,r 56 ,...},

[0087] in

[0088] R1=r1, R2=r4, R3=r3, R4=r4, R5=r 50 R6 = r 54 R7 = r 56 ,......

[0089] In this way, the first feature coding unit 242a can convert each danger information in the danger information sequence Is into a corresponding hot unique code, thereby generating a feature code sequence Rs corresponding to the danger information sequence.

[0090] In addition, the information extraction unit 241 also determines the employee production element information corresponding to the hazard information in the hazard information sequence Is.

[0091] For example, regarding hazard information I1 (i.e., insufficient strength of equipment A), the information extraction unit 241 extracts employee production factor information related to employees (e.g., employee 1 and employee 2) from the statistical information database 230 when equipment A exhibits a hazard state corresponding to hazard information I1. For example, employee 1's employee production factor information is "normal working hours," and employee 2's employee production factor information is also "normal working hours." Furthermore, the hot unique code encoding library of this disclosed technical solution also includes different types of employee production factor information and their corresponding hot unique codes. Employee production factor information and on-site production factor information are encoded in the same hot unique code encoding library so that subsequent feature code fusion can be performed on employee production factor information and on-site production factor information.

[0092] Table 3 below shows the different types of employee production element information and their corresponding hot unique codes in the hot unique code encoding library:

[0093] Table 3

[0094]

[0095]

[0096] Therefore, the second feature encoding unit 242b can determine the hot unique code (i.e., the second feature code) corresponding to the employee production factor information related to employee A and employee B from the hot unique codes shown in Table 2. For example, the employee production factor information of employee 1 is "normal working hours", and the employee production factor information of employee 2 is "normal working hours". Therefore, the second feature encoding unit 242b determines that the hot unique code Rm1 corresponding to the employee production factor information of employee 1 is rs3 and determines that the hot unique code Rm2 corresponding to the employee production factor information of employee 2 is also rs3.

[0097] Then, the feature fusion unit 243 combines the hot unique code (i.e., the first feature code) R1 output by the first feature encoding unit 242a that is associated with the hazard information I1 and the hot unique code Rm corresponding to the employee production factor information of employee 1 and employee 2 associated with the hazard information I1. 1,1 and Rm 1,2 (i.e., the second feature code) is fused to generate the fused feature code Rf1 corresponding to the danger information I1.

[0098] Specifically, for example, feature fusion unit 243 can fuse feature codes by directly summing the hot unique code (i.e., the first feature code) R1 associated with the hazard information I1 and the hot unique code (i.e., the second feature code) corresponding to the employee production factor information. This yields the fused feature code Rf1 corresponding to the hazard information I1.

[0099] Then, for the danger information sequence IS Other hazard information I i You can refer to the above-described method and generate the fusion feature code Rf corresponding to this hazard information according to the following formula. i :

[0100]

[0101] Where: R i It is related to the i-th danger information I i The corresponding hot unique code (i.e., the first feature code), Rm i,j It is related to the i-th danger information I i The hot unique code (i.e. the second feature code) of the employee production factor information of the associated j-th employee.

[0102] Thus, through the above methods, feature fusion unit 243 generates a sequence I containing hazard information. S The corresponding fusion feature code sequence Rfs = {Rf1, Rf2, Rf3, ...}. Here, Rf1 is the fusion feature code corresponding to danger information I1; Rf2 is the fusion feature code corresponding to danger information I2; Rf3 is the fusion feature code corresponding to danger information I3; and so on.

[0103] Then, the feature vector unit 244 receives the fused feature code sequence Rfs from the feature fusion unit 243, and uses the word embedding-based fusion feature vector model to convert each fused feature code in the fused feature code sequence Rfs into a corresponding fused feature vector, thereby generating a fused feature vector sequence Ws corresponding to the fused feature code sequence Rfs: {W1,W2,W3,W4,W5,W6,W7,...}.

[0104] Among them, W1 is the fusion feature vector corresponding to the fusion feature code Rf1 generated based on word embedding; W2 is the fusion feature vector corresponding to the fusion feature code Rf2 generated based on word embedding; W3 is the fusion feature vector corresponding to the fusion feature code Rf3 generated based on word embedding; W4 is the fusion feature vector corresponding to the fusion feature code Rf4 generated based on word embedding; W5 is the fusion feature vector corresponding to the fusion feature code Rf5 generated based on word embedding; W6 is the fusion feature vector corresponding to the fusion feature code Rf6 generated based on word embedding; W7 is the fusion feature vector corresponding to the fusion feature code Rf7 generated based on word embedding, and so on.

[0105] One example of a fusion feature vector model is a word embedding vector model based on CBOW. During training, the hazard information in Table 1 and the employee production factor information in Table 3 can be combined as corpus to train the word embedding vector model. For example, "insufficient equipment strength" in Table 1 can be combined with "normal physical load" in Table 3 as a training sample for CBOW; or "inadequate protection" in Table 1 can be combined with "violation of regulations" in Table 3 as a training sample for CBOW. This process can be repeated to obtain the corpus for training CBOW. Specific training methods will not be elaborated here.

[0106] Then, the question vector module 250 receives the fused feature vector sequence Ws from the feature vector unit 244, and then uses the attention-based question vector model to generate the question vector sequence Qs corresponding to the fused feature vector sequence Ws.

[0107] The present disclosure proposes the concept of a problem space. This problem space consists of multiple problem vectors, and different problem vectors correspond to different problem information in production safety management. Thus, the problem vectors can reflect the problems existing in the production process. For example, Table 4 below shows the various problems constituting the problem space and their corresponding problem vectors:

[0108] Table 4

[0109] 1 Employees working too long hours <![CDATA[q1]]> 2 Employees' workload is too high <![CDATA[q2]]> 3 Lack of high-temperature protection facilities for employees <![CDATA[q3]]> 4 Lack of equipment maintenance <![CDATA[q4]]> 5 Employees are complacent about the dangers of high temperatures. <![CDATA[q5]]> 6 There are deficiencies in employee management. <![CDATA[q6]]> ... ... ...

[0110] Therefore, the question vector unit 353 can utilize an attention-based question vector model to generate a question vector sequence Qs corresponding to the fused feature vector sequence Ws: {Q1,Q2,Q3,Q4,Q5,Q6,Q7,...}. The question vectors in the question vector sequence Qs correspond to different question information shown in Table 4.

[0111] In this case, as an example of a problem vector model, an encoder based on Transformer layers is used as the problem vector model. The number of Transformer layers can be set according to specific needs, which will not be elaborated here.

[0112] Therefore, the technical solution disclosed herein generates question vectors corresponding to hazard information by adopting a question vector model based on an attention mechanism. This enables the semantic features of hazard information to be deeply mined based on the correlation between hazard information collected within a predetermined period, and to more accurately analyze the problems and causes of hazards, providing targeted suggestions to relevant personnel.

[0113] Optionally, the operation of determining the solution vector corresponding to the problem vector using a solution vector model based on natural language processing includes: using an attention-based solution vector model to generate a corresponding solution vector sequence based on the problem vector sequence.

[0114] The present disclosure proposes the concept of a solution space. This solution space consists of multiple solution vectors, and different solution vectors correspond to different suggestions in production safety management. For example, Table 5 below shows the various suggestions constituting the solution space and their corresponding solution vectors:

[0115] Table 5

[0116]

[0117]

[0118] Therefore, the solution vector module 260 can utilize an attention-based solution vector model to generate a solution vector sequence As corresponding to the problem vector sequence Qs: {A1,A2,A3,A4,A5,A6,A7,...}. The solution vectors in the solution vector sequence As correspond to the different suggestion information shown in Table 5.

[0119] In this example, as a solution vector model, this disclosure uses an encoder based on Transformer layers as the problem vector model. The number of Transformer layers can be set according to specific needs, which will not be elaborated here.

[0120] Therefore, the technical solution disclosed herein utilizes an attention-based solution vector model to fully exploit the semantic features of the question vector. For example, when generating a corresponding solution vector for question vector Q7, it not only extracts the features of question vector Q7, but also extracts features from other question vectors in the vector sequence Qs that are highly correlated with question vector Q7 to generate the corresponding solution vector. In this way, the generated solution vector can combine the semantic features of the question vector in the entire question sequence to generate a more targeted suggestion.

[0121] in Figure 5B A schematic diagram of the various modules and units for generating a solution vector sequence based on a sequence of hazard information is shown. Further, Figure 5C This diagram illustrates the various models used to generate a sequence of solution vectors from a sequence of hazard information. (Reference) Figure 5B and Figure 5CAs shown, the information extraction unit 241 of the feature extraction module 240 first inputs the hazard information of the hazard information sequence Is into the first feature encoding unit 242a, thereby generating a hot unique code (i.e., a first feature code) corresponding to the hazard information. Furthermore, the information extraction unit 241 obtains employee production factor information corresponding to the hazard information of the hazard information sequence Is from the statistical information database 230, and uses the second feature encoding unit to generate a hot unique code (i.e., a second feature code) corresponding to each employee production factor information.

[0122] Then, the feature fusion unit fuses the hot unique codes of the hazard information and the corresponding hot unique codes of the employee production factor information to generate a fused feature code corresponding to the hazard information. Furthermore, the feature vector unit uses a CBOW-based word embedding vector model to generate a fused feature vector corresponding to the fused feature code. Thus, this fused feature vector not only incorporates the semantic features of the hazard information but also the semantic features of the employee's work status associated with the hazard information. In addition, the fused feature vectors corresponding to different hazard information constitute a fused feature vector sequence.

[0123] Then, the problem vector module uses an attention-based problem vector model to generate corresponding problem vectors based on the input fused feature vectors. Thus, through this attention-based model, when generating problem vectors from the fused feature codes, the module considers not only the fused feature vectors of the current hazard information but also the fused feature vectors of previous hazard information. This allows the generated problem vectors to more comprehensively reflect the problems existing in the production process.

[0124] Then, the solution vector module uses an attention-based solution vector model to generate corresponding solution vectors based on the input problem vector. Referring to Table 5, the solution vectors correspond to different suggestion information. Therefore, the suggestion generation module generates suggestion information corresponding to the hazard information based on the correspondence shown in Table 5 and the solution vectors generated by the solution vector module. Furthermore, because an attention-based solution vector model is used, the solution vector module generates solution vectors not only based on the current problem vector but also on previous problem vectors. This allows the generated solution vectors to more comprehensively consider the problems existing in the production process and provide matching solution vectors.

[0125] also, Figure 6A A schematic diagram of another variation of the production management platform according to this disclosure is shown. Figure 6B A schematic diagram of the feature extraction module 240 according to this variant is shown. Figure 6C A schematic diagram of each model for generating a sequence of solution vectors from a sequence of hazard information, according to this variation, is further shown. Figure 6DFurther schematic diagrams of the various models used according to this variation are shown.

[0126] refer to Figure 6A As shown, the production safety management platform 200 also includes a suggestion information receiving module 280, which is used to receive suggestion information adopted by relevant staff and transmit the suggestion information adopted by relevant staff to the feature extraction module.

[0127] Further reference Figure 6B As shown, the feature extraction module 240 includes a third feature encoding unit, used to generate a corresponding hot unique code (i.e., a third feature code) based on the suggestion information adopted by relevant staff. Furthermore, in the technical solution of this disclosure, the suggestion information, employee production factor information, and on-site production factor information are encoded in the same hot unique code encoding library, so that subsequent feature code fusion can be performed on the suggestion information, employee production factor information, and on-site production factor information.

[0128] For example, Table 6 below shows the different types of suggestion information set in the hot unique code encoding library and their corresponding hot unique codes:

[0129] Table 6

[0130] 1 Reduce employee working hours <![CDATA[rc1]]> 2 Reduce the labor intensity of employees <![CDATA[rc2]]> 3 Improve employee protection facilities <![CDATA[rc3]]> 4 Strengthen training and awareness among employees regarding the dangers of high temperatures. <![CDATA[rc4]]> 5 Strengthen employee safety awareness training <![CDATA[rc5]]> ... .... ....

[0131] Therefore, the third feature encoding unit 242b can determine the hot unique code (i.e., the third feature code) corresponding to the suggestion information adopted by the relevant staff from the hot unique codes shown in Table 6. For example, in this variant, the adopted suggestion information is "reduce employee working hours", so the second feature encoding unit 242b determines the hot unique code corresponding to the adopted suggestion information as rc1.

[0132] Therefore, for example, feature fusion unit 243 can fuse feature codes by directly summing the hot unique code (i.e., the first feature code) R1 associated with the hazard information I1, the hot unique code (i.e., the second feature code) corresponding to the employee production factor information, and the hot unique code (i.e., the third feature code) corresponding to the suggestion information adopted by relevant staff within the preset period. This yields the fused feature code Rf1 corresponding to the hazard information I1.

[0133] Then, for the danger information sequence I S Other hazard information I i You can refer to the method described above and generate the fusion feature code Rf corresponding to the hazard information according to the following formula. i :

[0134]

[0135] Where: R i It is related to the i-th danger information Ii The corresponding hot unique code (i.e., the first feature code), Rm i,j It is related to the i-th danger information I i The hot unique code (i.e., the second feature code) of the associated j-th employee, Rc k It is the hot unique code (i.e., the third feature code) of the suggestion information adopted within the preset period.

[0136] Thus, through the above methods, feature fusion unit 243 generates a sequence I containing hazard information. S The corresponding fusion feature code sequence Rfs = {Rf1, Rf2, Rf3, ...}. Here, Rf1 is the fusion feature code corresponding to danger information I1; Rf2 is the fusion feature code corresponding to danger information I2; Rf3 is the fusion feature code corresponding to danger information I3; and so on.

[0137] The subsequent process can then follow the steps described above. Therefore, based on the variation, the process of generating the fusion feature code not only considers hazard information and relevant employee production factor information, but also the suggestions already adopted and implemented within the preset period. Thus, the fusion feature code generated from these three aspects contains deeper semantic information. The resulting subsequent question vector can reflect deeper reasons and problems regarding the effectiveness of the proposed suggestions, thereby helping to propose more effective suggestions and avoiding simple repetition of suggestions lacking practical effect.

[0138] refer to Figure 6C and Figure 6D As shown, the information extraction unit 241 of the feature extraction module 240 first inputs the hazard information of the hazard information sequence Is into the first feature encoding unit 242a, thereby generating a hot unique code (i.e., a first feature code) corresponding to the hazard information. Furthermore, the information extraction unit 241 retrieves employee production factor information corresponding to the hazard information of the hazard information sequence Is from the statistical information database 230, and uses the second feature encoding unit to generate a hot unique code (i.e., a second feature code) corresponding to each employee production factor information. Additionally, the suggestion information receiving module 200 of the production safety management platform 200 receives suggestion information adopted and implemented within a preset period and transmits the received suggestion information to the feature extraction module 240. Thus, the feature extraction module 240 uses the third feature encoding unit to generate a hot unique code (i.e., a third feature code) corresponding to the suggestion information.

[0139] Then, the feature fusion unit fuses the hot unique codes of hazard information, corresponding employee production factor information, and suggestion information to generate a fused feature code corresponding to the hazard information. Furthermore, the feature vector unit uses a CBOW-based word embedding vector model to generate a fused feature vector corresponding to the fused feature code. Thus, this fused feature vector not only incorporates the semantic features of the hazard information but also the semantic features of the employee's work status associated with the hazard information and the semantic features of the suggestions adopted and implemented within a preset period. In addition, the fused feature vectors corresponding to different hazard information constitute a fused feature vector sequence.

[0140] Then, the problem vector module uses an attention-based problem vector model to generate corresponding problem vectors based on the input fused feature vectors. Thus, through this attention-based model, when generating problem vectors from the fused feature codes, the module considers not only the fused feature vectors of the current hazard information but also the fused feature vectors of previous hazard information. This allows the generated problem vectors to more comprehensively reflect the problems existing in the production process.

[0141] Then, the solution vector module uses an attention-based solution vector model to generate corresponding solution vectors based on the input problem vector. Referring to Table 5, the solution vectors correspond to different suggestion information. Therefore, the suggestion generation module generates suggestion information corresponding to the hazard information based on the correspondence shown in Table 5 and the solution vectors generated by the solution vector module. Furthermore, because an attention-based solution vector model is used, the solution vector module generates solution vectors not only based on the current problem vector but also on previous problem vectors. This allows the generated solution vectors to more comprehensively consider the problems existing in the production process and provide matching solution vectors.

[0142] Furthermore, according to a second aspect of this embodiment, an artificial intelligence-based production safety management method is provided, the method comprising: Figure 1 The production safety management platform 200 shown is implemented. Figure 7 A flowchart illustrating the method is shown below. (Refer to...) Figure 7 As shown, the method includes:

[0143] S702: Determine hazard information related to the production process based on information about production factors related to the production process;

[0144] S704: Using a natural language processing-based question vector model, the corresponding question vector is determined based on the hazard information. The question vector is used to indicate the problem corresponding to the hazard information in the production process.

[0145] S706: Utilize a solution vector model based on natural language processing to determine the solution vector corresponding to the problem vector; and

[0146] S708: Determine the suggested information corresponding to the hazard information based on the solution vector, and send the suggested information to the terminal equipment of relevant personnel.

[0147] Optionally, the operation of determining hazard information related to the production process based on production factor information includes: statistically analyzing production factor information associated with the target object from the production factor information according to preset statistical rules; and determining hazard information related to the production process of the target object based on the production factor information associated with the target object.

[0148] Optionally, the operation of determining the hazard information related to the production process of the target object based on the production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on the on-site production factor information associated with the target object, forming a hazard information sequence, wherein the on-site production factor information includes at least one of the following: the state information of hazardous substances at the production site; the state information of hazardous energy at the production site; the state information of working equipment at the production site; the state information of hazardous organisms present at the production site; and the state information of the work layout at the production site.

[0149] Optionally, the operation of determining the corresponding question vector based on the hazard information using a natural language processing-based question vector model includes: generating a fused feature vector corresponding to the hazard information and forming a fused feature vector sequence corresponding to the hazard information sequence, wherein the fused feature vector contains semantic features of on-site production factor information and employee production factor information associated with the hazard information, wherein the employee production factor information includes the employee's status information during the production process; and generating a question vector sequence corresponding to the feature vector sequence through an attention-based question vector model.

[0150] Optionally, the operation of generating a fused feature vector corresponding to the hazard information includes: generating a first feature code associated with the hazard information, wherein the first feature code contains the semantic features of the hazard information; determining the employee production factor information associated with the hazard information and generating a second feature code associated with the employee production factor information, wherein the second feature code contains the semantic features of the employee production factor information; fusing the first feature code and the second feature code to generate a fused feature code; and generating a fused feature vector corresponding to the hazard information based on the fused feature code.

[0151] Optionally, the operation of generating a fused feature vector corresponding to the hazard information includes: generating a first feature code associated with the hazard information, wherein the first feature code contains the semantic features of the hazard information; determining employee production factor information associated with the hazard information and generating a second feature code associated with the employee production factor information, wherein the second feature code contains the semantic features of the employee production factor information; generating a third feature code associated with the suggestion information adopted within a preset period, wherein the third feature code contains the semantic features of the suggestion information; fusing the first feature code, the second feature code, and the third feature code to generate a fused feature code; and generating a fused feature vector corresponding to the hazard information based on the fused feature code.

[0152] Optionally, the operation of determining the solution vector corresponding to the problem vector using a solution vector model based on natural language processing includes: using an attention-based solution vector model to generate a corresponding solution vector sequence based on the problem vector sequence.

[0153] Optionally, the operation of determining the suggested information corresponding to the hazard information based on the solution vector includes: accessing a pre-set solution vector-suggestion information mapping table and determining the suggested information corresponding to the solution vector as the suggested information corresponding to the hazard information.

[0154] In addition, refer to Figure 1 As shown, according to a third aspect of this embodiment, a storage medium is provided. The storage medium includes a stored program, wherein, when the program is executed, a processor performs any of the methods described above.

[0155] The technical solution disclosed herein is based on natural language processing (NLP) technology and introduces problem vectors and solution vectors. The problem vectors indicate the problems corresponding to hazardous information existing in the production process, while the solution vectors are associated with the suggested information corresponding to that hazardous information. Furthermore, the problem vectors can constitute a problem space, and the solution vectors can constitute a solution space that has a mapping relationship with the problem space. Both the problem vectors and solution vectors are generated using a NLP-based model. Therefore, according to the technical solution disclosed herein, the NLP-based problem vector model and solution vector model can be trained based on actual hazardous information, making the problem space constructed by the problem vectors and the solution space constructed by the solution vectors more flexibly adaptable to current production safety management. This fully leverages the control mechanism, learning mechanism, and system dynamics mechanism of artificial intelligence technology in production safety management. It solves the technical problems that existing production safety management systems cannot make flexible judgments on production hazards and that artificial intelligence technology cannot be deeply applied to production management systems.

[0156] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0157] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0158] Example 2

[0159] Figure 8 An artificial intelligence-based production safety management device 800 according to this embodiment is shown, which corresponds to the method described according to the second aspect of Embodiment 1. (See reference...) Figure 8 As shown, the device 800 includes: a hazard information determination module 810, used to determine hazard information related to the production process based on production factor information related to the production process; a problem vector determination module 820, used to determine the corresponding problem vector based on the hazard information using a problem vector model based on natural language processing, wherein the problem vector is used to indicate the problem corresponding to the hazard information in the production process; a solution vector determination module 830, used to determine the solution vector corresponding to the problem vector using a solution vector model based on natural language processing; and a suggestion information determination module 840, used to determine the suggestion information corresponding to the hazard information based on the solution vector and send the suggestion information to the terminal device of relevant personnel.

[0160] Optionally, the hazard information determination module 810 includes: a statistics submodule, used to collect production factor information associated with the target object from production factor information according to preset statistical rules; and a hazard information determination submodule, used to determine hazard information related to the production process of the target object based on the production factor information associated with the target object.

[0161] Optionally, the hazard information determination submodule also includes a hazard information determination unit, which is used to determine multiple hazard information related to the production process of the target object within a preset period based on the on-site production element information associated with the target object, thereby forming a hazard information sequence.

[0162] Optionally, the question vector determination module 820 includes: a first feature code generation unit, used to determine first feature codes corresponding to different hazard information in the hazard information sequence from a pre-set hot unique code encoding library; a hazard information determination unit, used to determine the employees corresponding to the hazard information; a second feature code generation unit, used to determine second feature codes corresponding to the determined employee's production factor information from the hot unique code encoding library; a fusion feature code generation unit, used to fuse the first and second feature codes to generate fusion feature codes corresponding to different hazard information, thereby forming a fusion feature code sequence; a fusion feature sequence generation unit, used to generate a fusion feature vector sequence corresponding to the fusion feature code sequence through a word vector model based on word embedding; and a question vector sequence generation unit, used to generate a question vector sequence corresponding to the fusion feature vector sequence through a question vector model based on an attention mechanism.

[0163] Optionally, the solution vector determination module 830 includes: a solution vector determination submodule, used to generate a corresponding solution vector sequence based on the problem vector sequence using an attention-based solution vector model.

[0164] Optionally, the suggestion information determination module 840 includes: a suggestion information determination submodule, used to access a pre-set solution vector-suggestion information mapping table and determine the suggestion information corresponding to the solution vector as the suggestion information corresponding to the danger information.

[0165] Therefore, the technical solution of this disclosure, based on natural language processing (NLP) technology, introduces problem vectors and solution vectors. The problem vectors indicate the problems corresponding to hazardous information existing in the production process, while the solution vectors are associated with the suggested information corresponding to that hazardous information. Furthermore, the problem vectors can constitute a problem space, and the solution vectors can constitute a solution space that has a mapping relationship with the problem space. Both the problem vectors and solution vectors are generated using a NLP-based model. Thus, according to the technical solution of this disclosure, the NLP-based problem vector model and solution vector model can be trained based on actual hazardous information, making the problem space constructed by the problem vectors and the solution space constructed by the solution vectors more flexibly adaptable to current production safety management. This fully leverages the control mechanism, learning mechanism, and system dynamics mechanism of artificial intelligence technology in production safety management. It solves the technical problems that existing production safety management systems cannot make flexible judgments on production hazards and that artificial intelligence technology cannot be deeply applied to production management systems.

[0166] Example 3

[0167] Figure 9 An artificial intelligence-based production safety management device 900 according to this embodiment is shown, which corresponds to the method described according to the second aspect of Embodiment 1. (See reference...) Figure 9 As shown, the device 900 includes: a processor 910; and a memory 920 connected to the processor 910, for providing the processor 910 with instructions to process the following steps: determining hazard information related to the production process based on production factor information related to the production process; determining a corresponding problem vector based on the hazard information using a problem vector model based on natural language processing, the problem vector indicating the problem corresponding to the hazard information in the production process; determining a solution vector corresponding to the problem vector using a solution vector model based on natural language processing; and determining suggestion information corresponding to the hazard information based on the solution vector, and sending the suggestion information to the terminal device of relevant personnel.

[0168] Optionally, the operation of determining hazard information related to the production process based on production factor information includes: statistically analyzing production factor information associated with the target object from the production factor information according to preset statistical rules; and determining hazard information related to the production process of the target object based on the production factor information associated with the target object.

[0169] Optionally, the operation of determining the hazard information related to the production process of the target object based on the production factor information associated with the target object also includes: determining multiple hazard information related to the production process of the target object within a preset period based on the on-site production factor information associated with the target object, thereby forming a hazard information sequence.

[0170] Optionally, a question vector model based on natural language processing is used to determine corresponding question vectors based on hazard information. The question vectors are used to indicate the operations related to the hazard information in the production process, including: determining first feature codes corresponding to different hazard information in the hazard information sequence from a pre-set hot unique code encoding library; determining the employees corresponding to the hazard information; determining second feature codes corresponding to the determined employee's production factor information from the hot unique code encoding library; fusing the first and second feature codes to generate fused feature codes corresponding to different hazard information, thereby forming a fused feature code sequence; generating a fused feature vector sequence corresponding to the fused feature code sequence through a word embedding-based word vector model; and generating a question vector sequence corresponding to the fused feature vector sequence through an attention-based question vector model.

[0171] Optionally, the operation of determining the solution vector corresponding to the problem vector using a solution vector model based on natural language processing includes: using an attention-based solution vector model to generate a corresponding solution vector sequence based on the problem vector sequence.

[0172] Optionally, the operation of determining the suggested information corresponding to the hazard information based on the solution vector includes: accessing a pre-set solution vector-suggestion information mapping table and determining the suggested information corresponding to the solution vector as the suggested information corresponding to the hazard information.

[0173] Therefore, the technical solution of this disclosure, based on natural language processing (NLP) technology, introduces problem vectors and solution vectors. The problem vectors indicate the problems corresponding to hazardous information existing in the production process, while the solution vectors are associated with the suggested information corresponding to that hazardous information. Furthermore, the problem vectors can constitute a problem space, and the solution vectors can constitute a solution space that has a mapping relationship with the problem space. Both the problem vectors and solution vectors are generated using a NLP-based model. Thus, according to the technical solution of this disclosure, the NLP-based problem vector model and solution vector model can be trained based on actual hazardous information, making the problem space constructed by the problem vectors and the solution space constructed by the solution vectors more flexibly adaptable to current production safety management. This fully leverages the control mechanism, learning mechanism, and system dynamics mechanism of artificial intelligence technology in production safety management. It solves the technical problems that existing production safety management systems cannot make flexible judgments on production hazards and that artificial intelligence technology cannot be deeply applied to production management systems.

[0174] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0175] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0176] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0177] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0178] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0179] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0180] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A production safety management system based on artificial intelligence, characterized in that, include: Production information acquisition system (100) and production safety management platform (200), among which The production information collection system (100) is communicatively connected to the production safety management platform (200) and is used to collect production element information related to the production process and send the collected production element information to the production safety management platform (200). The production safety management platform (200) is configured to perform the following operations: Based on the production factor information, determine the hazard information related to the production process; Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. The question vector is used to indicate the problem in the production process that corresponds to the hazard information. Using a solution vector model based on natural language processing, the solution vector corresponding to the problem vector is determined; as well as Based on the solution vector, suggested information corresponding to the hazard information is determined, and the suggested information is sent to the terminal devices of relevant personnel. The operation of determining hazard information related to the production process based on the production factor information includes: According to preset statistical rules, statistical analysis is performed on the production factor information to identify production factor information associated with the target object; and Based on the production factor information associated with the target object, hazard information related to the production process of the target object is determined, wherein, The operation of determining hazard information related to the production process of the target object based on production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on on-site production factor information associated with the target object, forming a hazard information sequence, wherein... Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. This question vector indicates the operation for the problem corresponding to the hazard information during the production process, including: Determine the first feature code corresponding to different danger information in the danger information sequence from the pre-set hot unique code encoding library; Identify the employees corresponding to the aforementioned hazard information; Determine the second feature code corresponding to the determined employee production element information from the hot unique code encoding library; The first feature code and the second feature code are fused to generate a fused feature code corresponding to different danger information, thereby forming a fused feature code sequence; A fused feature vector sequence corresponding to the fused feature code sequence is generated using a word embedding-based word vector model; and A question vector sequence corresponding to the fused feature vector sequence is generated using a question vector model based on an attention mechanism.

2. The system according to claim 1, characterized in that, The operation of determining the solution vector corresponding to the problem vector using a solution vector model based on natural language processing includes: Using an attention-based solution vector model, a corresponding solution vector sequence is generated based on the problem vector sequence.

3. The system according to claim 1, characterized in that, The operation of determining the suggested information corresponding to the hazard information based on the solution vector includes: accessing a pre-set solution vector-suggested information mapping table, and determining the suggested information corresponding to the solution vector as the suggested information corresponding to the hazard information.

4. A production safety management method based on artificial intelligence, characterized in that, include: Based on information about production factors related to the production process, determine the hazard information related to the production process; Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. The question vector is used to indicate the problem in the production process that corresponds to the hazard information. Using a solution vector model based on natural language processing, the solution vector corresponding to the problem vector is determined; as well as Based on the solution vector, suggested information corresponding to the hazard information is determined, and the suggested information is sent to the terminal devices of relevant personnel. The operation of determining hazard information related to the production process based on the production factor information includes: According to preset statistical rules, statistical analysis is performed on the production factor information to identify production factor information associated with the target object; and Based on the production factor information associated with the target object, hazard information related to the production process of the target object is determined, wherein, The operation of determining hazard information related to the production process of the target object based on production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on on-site production factor information associated with the target object, forming a hazard information sequence, wherein... Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. This question vector indicates the operation for the problem corresponding to the hazard information during the production process, including: Determine the first feature code corresponding to different danger information in the danger information sequence from the pre-set hot unique code encoding library; Identify the employees corresponding to the aforementioned hazard information; Determine the second feature code corresponding to the determined employee production element information from the hot unique code encoding library; The first feature code and the second feature code are fused to generate a fused feature code corresponding to different danger information, thereby forming a fused feature code sequence; A fused feature vector sequence corresponding to the fused feature code sequence is generated using a word embedding-based word vector model; and A question vector sequence corresponding to the fused feature vector sequence is generated using a question vector model based on an attention mechanism.

5. A storage medium, characterized in that, The storage medium includes a stored program, wherein the method of claim 4 is executed by a processor when the program is run.

6. A production safety management device based on artificial intelligence, characterized in that, include: The hazard information determination module is used to determine hazard information related to the production process based on production factor information related to the production process. The problem vector determination module is used to determine the corresponding problem vector based on the hazard information using a problem vector model based on natural language processing. The problem vector is used to indicate the problem in the production process that corresponds to the hazard information. The solution vector determination module is used to determine the solution vector corresponding to the problem vector using a solution vector model based on natural language processing. as well as The suggestion information determination module is used to determine the suggestion information corresponding to the hazard information based on the solution vector, and send the suggestion information to the terminal device of relevant personnel. The operation of determining hazard information related to the production process based on the production factor information includes: According to preset statistical rules, statistical analysis is performed on the production factor information to identify production factor information associated with the target object; and Based on the production factor information associated with the target object, hazard information related to the production process of the target object is determined, wherein, The operation of determining hazard information related to the production process of the target object based on production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on on-site production factor information associated with the target object, forming a hazard information sequence, wherein... Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. This question vector indicates the operation for the problem corresponding to the hazard information during the production process, including: Determine the first feature code corresponding to different danger information in the danger information sequence from the pre-set hot unique code encoding library; Identify the employees corresponding to the aforementioned hazard information; Determine the second feature code corresponding to the determined employee production element information from the hot unique code encoding library; The first feature code and the second feature code are fused to generate a fused feature code corresponding to different danger information, thereby forming a fused feature code sequence; A fused feature vector sequence corresponding to the fused feature code sequence is generated using a word embedding-based word vector model; and A question vector sequence corresponding to the fused feature vector sequence is generated using a question vector model based on an attention mechanism.

7. A production safety management device based on artificial intelligence, characterized in that, include: processor; as well as A memory, connected to the processor, for providing the processor with instructions to perform the following processing steps: Based on information about production factors related to the production process, determine the hazard information related to the production process; Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. The question vector is used to indicate the problem in the production process that corresponds to the hazard information. Using a solution vector model based on natural language processing, the solution vector corresponding to the problem vector is determined; as well as Based on the solution vector, suggested information corresponding to the hazard information is determined, and the suggested information is sent to the terminal devices of relevant personnel. The operation of determining hazard information related to the production process based on the production factor information includes: According to preset statistical rules, statistical analysis is performed on the production factor information to identify production factor information associated with the target object; and Based on the production factor information associated with the target object, hazard information related to the production process of the target object is determined, wherein, The operation of determining hazard information related to the production process of the target object based on production factor information associated with the target object further includes: determining multiple hazard information related to the production process of the target object within a preset period based on on-site production factor information associated with the target object, forming a hazard information sequence, wherein... Using a natural language processing-based question vector model, a corresponding question vector is determined based on the hazard information. This question vector indicates the operation for the problem corresponding to the hazard information during the production process, including: Determine the first feature code corresponding to different danger information in the danger information sequence from the pre-set hot unique code encoding library; Identify the employees corresponding to the aforementioned hazard information; Determine the second feature code corresponding to the determined employee production element information from the hot unique code encoding library; The first feature code and the second feature code are fused to generate a fused feature code corresponding to different danger information, thereby forming a fused feature code sequence; A word vector model based on word embedding is used to generate a fusion feature vector sequence corresponding to the fusion feature code sequence; and a question vector sequence corresponding to the fusion feature vector sequence is generated using a question vector model based on attention mechanism.

Images