An AI model-based safety production hidden danger auxiliary inspection method and system

Abstract

The application relates to the field of production safety technology, in particular to a safety production hidden danger auxiliary inspection method and system based on an AI model. The method comprises the following steps: constructing an intelligent agent and multiple monitoring points according to original parameters of a production area; obtaining initial monitoring data according to a first-level monitoring strategy set by the intelligent agent, and setting an auxiliary monitoring strategy according to the initial monitoring data; obtaining a feedback data packet according to the auxiliary monitoring strategy; and the intelligent agent generates a hidden danger diagnosis result according to the feedback data packet. Based on the equipment structure parameters of the multiple monitoring points of the production area, comprehensive monitoring of the production area is realized, a correlation diagnosis model is constructed according to historical hidden danger data analysis, a correlation relationship network of the monitoring points and single-type hidden dangers is generated, multiple verifications of single-type hidden dangers are realized, and the misidentification rate of production hidden dangers is reduced. Linkage monitoring of different types of hidden dangers is realized, the overall identification efficiency of production hidden dangers in the production area is improved, and the safe operation in the production area is ensured.

Description

Technical Field

[0001] This application relates to the field of production safety technology, and in particular to a method and system for assisting in the inspection of safety hazards based on an AI model. Background Technology

[0002] Safety hazard identification is a core aspect of accident prevention. Traditional methods mainly rely on manual inspections, fixed sensor monitoring, and basic video surveillance analysis.

[0003] Manual inspections are inefficient, have limited coverage, heavily rely on the inspector's experience and condition, are highly subjective, difficult to standardize, and have weak ability to identify complex or hidden hazards. Sensor monitoring typically targets specific quantitative parameters, has high installation and maintenance costs, poor flexibility, and struggles to cover non-quantifiable hazards such as visual (e.g., equipment damage, improper wearing of protective equipment) and behavioral (e.g., violations of operating procedures) hazards. Basic video analytics has low levels of intelligence, is mostly based on simple rules, has poor environmental adaptability, and suffers from high false alarm and false negative rates. Summary of the Invention

[0004] The purpose of this application is to provide a safety production hazard auxiliary inspection method and system based on AI model in order to solve the above-mentioned technical problems, thereby improving the efficiency of identifying and inspecting production hazards in the production area and ensuring the safe operation of the production area.

[0005] In some embodiments of this application, multiple monitoring points based on the equipment structure parameters of the production area are used to achieve comprehensive monitoring of the production area. A correlation diagnosis model is constructed based on the analysis of historical hidden danger data to generate a correlation network between monitoring points and single-type hidden dangers, thereby achieving multiple verifications for single-type hidden dangers and reducing the false identification rate of production hidden dangers.

[0006] In some embodiments of this application, a network of relationships between various types of hazards is constructed based on the risk diffusion trends among different categories of hazards, enabling coordinated monitoring of different types of hazards and improving the overall efficiency of identifying production hazards within the generated area. This ensures the safe operation of the production area.

[0007] In some embodiments of this application, a safety hazard auxiliary inspection method based on an AI model is provided, including:

[0008] Based on the original parameters of the production area, an intelligent agent and multiple monitoring points are constructed;

[0009] Initial monitoring data is obtained according to the primary monitoring strategy set by the intelligent agent, and auxiliary monitoring strategies are set according to the initial monitoring data.

[0010] The agent obtains feedback data packets based on the auxiliary monitoring strategy and generates hazard diagnosis results based on the feedback data packets.

[0011] In some embodiments of the present application, constructing an agent includes:

[0012] Generating a historical hidden danger data packet according to the original parameters, and generating a plurality of hidden danger events according to the historical hidden danger packet;

[0013] Establishing a hidden danger event sequence A, A = (a1, a2…a i …a n ), where ai i is the i-th hidden danger event; n is the number of hidden danger events;

[0014] Constructing a first-level identification model according to all hidden danger events, and sequentially generating auxiliary sub-models for each hidden danger event;

[0015] Constructing an associated diagnosis model according to all auxiliary sub-models;

[0016] Constructing an agent according to the hidden danger event sequence A, the first-level identification model, and the associated diagnosis model.

[0017] In some embodiments of the present application, constructing a first-level identification model according to all hidden danger events includes:

[0018] Sequentially setting ai as the target hidden danger event according to the hidden danger event sequence A;

[0019] Generating an associated data packet of the target hidden danger event based on the historical hidden danger data packet;

[0020] Generating an identification value b of the target hidden danger event according to the associated data packet;

[0021] b = β i *v i ;

[0022] where θ1 is the number of first-level identification indicators; β i is the influence factor of the i-th first-level identification indicator; v i is the reference value of the i-th first-level identification indicator set based on the associated data packet;

[0023] Presetting an identification value threshold B1;

[0024] If b > B1, setting the target hidden danger event as a first-level hidden danger, and generating an identification sub-model of the target hidden danger event according to the associated data packet;

[0025] If b < B1, setting the target hidden danger event as a second-level hidden danger;

[0026] Sequentially judging each hidden danger event;

[0027] Establishing a first-level hidden danger sequence A1, A1 = (a 11 , a12 …a 1i …a 1n1 ), where a 1i Let n1 be the i-th level 1 hidden danger; n1 is the number of level 1 hidden dangers;

[0028] Obtain the identification sub-models for each Level 1 hazard, and set the Level 1 identification model based on all the identification sub-models.

[0029] In some embodiments of this application, auxiliary sub-models for generating various potential hazard events include:

[0030] Based on the sequence of potential incidents A, a is set sequentially. i Potential hazards to be monitored;

[0031] Select the primary correlation points of the potential hazards to be monitored based on the associated data packets of the potential hazards to be monitored;

[0032] A first-level association table of potential hazards to be monitored is generated based on all first-level association points;

[0033] Generate primary correlation values ​​between the potential hazard events to be monitored and each potential hazard event;

[0034] Establish a first-level related value sequence C, C=(c1, c2…c i …c n ), where c i is the primary correlation value between the monitored hazard event and the i-th hazard event; n is the number of hazard events;

[0035] Preset the first-level correlation threshold C1;

[0036] If c i >C1, define the i-th hidden danger event as a related hidden danger event of the hidden danger event to be monitored;

[0037] A secondary association table is generated based on all associated hazards of the hazard event to be monitored.

[0038] Auxiliary sub-models for potential hazards to be monitored are generated based on the primary and secondary association tables.

[0039] Each potential hazard event auxiliary sub-model is generated sequentially.

[0040] In some embodiments of this application, a secondary association table is generated based on all associated hazards of the hazard event to be monitored, including:

[0041] An initial set of hazards is constructed based on the hazard events to be monitored and all related hazards;

[0042] Determine whether there are any Level 1 hazards in the initial hazard set;

[0043] If they exist, generate a secondary association table based on the initial set of potential hazards;

[0044] If none exist, set all Level 1 hazards as associated hazards of the hazard events to be monitored, generate correction instructions for the initial hazard set, and generate a Level 2 association table based on the correction results.

[0045] In some embodiments of this application, the primary monitoring strategy set by the intelligent agent includes:

[0046] Establish a sequence of sub-models for identification, P, P = (p1, p2, ..., p2). i …p n1 ), where p i Let n1 be the identification sub-model for the i-th Level 1 hazard; n1 is the number of Level 1 hazard.

[0047] Based on the identification sub-model sequence P, p is set sequentially. i Identify sub-models for the target;

[0048] Obtain all feature indicators of the target recognition sub-model, and select all secondary association points of the target recognition model based on all feature indicators;

[0049] The monitoring substructure of the target identification sub-model is constructed based on all secondary correlation points;

[0050] The monitoring substructures of each identification sub-model are constructed sequentially, and a first-level monitoring strategy is generated based on all the monitoring substructures.

[0051] In some embodiments of this application, an auxiliary monitoring strategy is set, including:

[0052] Multiple packages to be identified are generated based on the initial monitoring data, and a mapping table of primary hazards and packages to be identified is constructed.

[0053] Based on the first-level hidden danger sequence A1, a is set sequentially. 1i The target is a Level 1 hazard;

[0054] The primary identification model obtains the associated packets to be identified for the primary hidden dangers of the target.

[0055] The risk probability value f for generating a Level 1 hidden danger of the target;

[0056] f=[ μ i *s i ];

[0057] Where θ2 represents the number of characteristic indicators of the target's primary hidden danger; μ i The influencing factor of the i-th characteristic indicator of the primary hidden danger; s i The matching value of the i-th feature index of the target first-level hidden danger is generated based on the associated package to be identified;

[0058] Preset risk probability threshold F1;

[0059] If f > F1, generate an auxiliary sub-strategy for the first-level hidden danger of the target;

[0060] The system sequentially determines whether to generate auxiliary sub-strategies for each level-one hidden danger, and generates auxiliary monitoring strategies based on all auxiliary sub-strategies.

[0061] In some embodiments of this application, the auxiliary sub-strategy for generating the target level-one hidden danger includes:

[0062] The auxiliary sub-model for the primary hidden danger of the target is set as the target auxiliary model;

[0063] A primary acquisition strategy is generated based on the primary association table in the target auxiliary model;

[0064] The verification sub-data packets of the target's primary hidden dangers are obtained based on the primary acquisition strategy;

[0065] Each associated hazard of the target primary hazard is designated as a hazard to be diagnosed, and a secondary data collection strategy is set for all hazards to be diagnosed.

[0066] The diagnostic sub-data packets for each potential hazard to be diagnosed are obtained according to the secondary acquisition strategy;

[0067] Based on the primary and secondary data collection strategies, auxiliary sub-strategies for identifying primary hidden dangers are set.

[0068] In some embodiments of this application, a safety hazard auxiliary inspection system based on an AI model is provided, including:

[0069] The central control unit is used to construct intelligent agents and multiple monitoring points based on the original parameters of the production area;

[0070] The monitoring unit includes multiple monitoring sub-modules, which are located at various monitoring points.

[0071] The central control unit includes:

[0072] The first processing module is used to construct the intelligent agent;

[0073] The second processing module is used to obtain initial monitoring data according to the primary monitoring strategy set by the intelligent agent, and to set auxiliary monitoring strategies according to the initial monitoring data.

[0074] The third processing module is used to obtain feedback data packets according to the auxiliary monitoring strategy, and the intelligent agent generates hidden danger diagnosis results based on the feedback data packets.

[0075] The constructed intelligent agent includes:

[0076] Generate historical hazard data packets based on the original parameters, and generate multiple hazard events based on the historical hazard data packets;

[0077] Establish a hidden danger event sequence A, A = (a1, a2…a i …a n ), where a i is the i-th hidden danger event; n is the number of hidden danger events;

[0078] Construct a first-level identification model based on all hidden danger events, and sequentially generate auxiliary sub-models for each hidden danger event;

[0079] Construct an associated diagnosis model based on all auxiliary sub-models;

[0080] Construct an agent based on the hidden danger event sequence A, the first-level identification model and the associated diagnosis model.

[0081] In some embodiments of the present application, the first processing module is further configured to:

[0082] Sequentially set ai as the target hidden danger event according to the hidden danger event sequence A;

[0083] Generate an associated data packet for the target hidden danger event based on the historical hidden danger data packet;

[0084] Generate an identification value b for the target hidden danger event according to the associated data packet;

[0085] b = β i *v i ;

[0086] Where θ1 is the number of first-level identification indicators; β i is the influence factor of the i-th first-level identification indicator; v i is the reference value of the i-th first-level identification indicator set based on the associated data packet;

[0087] Preset an identification value threshold B1;

[0088] If b > B1, set the target hidden danger event as a first-level hidden danger, and generate an identification sub-model for the target hidden danger event according to the associated data packet;

[0089] If b < B1, set the target hidden danger event as a second-level hidden danger;

[0090] Sequentially judge each hidden danger event;

[0091] Establish a first-level hidden danger sequence A1, A1 = (a 11 , a 12 …a 1i …a 1n1 ), where a 1i is the i-th first-level hidden danger; n1 is the number of first-level hidden dangers;

[0092] Obtain the identification sub-models for each Level 1 hazard, and set the Level 1 identification model based on all the identification sub-models.

[0093] Compared with existing technologies, the safety hazard auxiliary inspection method and system based on AI models in this application have the following advantages:

[0094] Based on multiple monitoring points of equipment structural parameters in the production area, comprehensive monitoring of the production area is achieved. Based on the analysis of historical hidden danger data, an association diagnosis model is constructed to generate a correlation network between monitoring points and single types of hidden dangers, enabling multiple verifications of single types of hidden dangers and reducing the false identification rate of production hidden dangers.

[0095] By analyzing the risk diffusion trends among different types of hazards, a network of relationships between various hazards is constructed, enabling coordinated monitoring of different types of hazards and improving the overall efficiency of identifying production hazards within the production area. This ensures the safe operation of the production area. Attached Figure Description

[0096] Figure 1 This is a flowchart illustrating a preferred embodiment of a safety production hazard auxiliary inspection method based on an AI model. Detailed Implementation

[0097] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0098] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0099] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0100] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0101] like Figure 1 As shown, a preferred embodiment of this application provides a safety hazard auxiliary inspection method based on an AI model, comprising:

[0102] Based on the original parameters of the production area, an intelligent agent and multiple monitoring points are constructed;

[0103] Initial monitoring data is obtained according to the primary monitoring strategy set by the intelligent agent, and auxiliary monitoring strategies are set according to the initial monitoring data.

[0104] The agent obtains feedback data packets based on the auxiliary monitoring strategy and generates hazard diagnosis results based on the feedback data packets.

[0105] Specifically, building intelligent agents includes:

[0106] Generate historical hazard data packets based on the original parameters, and generate multiple hazard events based on the historical hazard data packets;

[0107] Establish a sequence A of potential incidents, A=(a1,a2…a ... i …a n ), where a i Let be the i-th potential hazard; n is the number of potential hazard events;

[0108] A primary identification model is constructed based on all potential hazards, and auxiliary sub-models for each potential hazard are generated sequentially.

[0109] Construct an associated diagnostic model based on all auxiliary sub-models;

[0110] Based on the sequence of potential incidents A, an intelligent agent is constructed using a primary identification model and a correlation diagnosis model.

[0111] Specifically, the original parameters include historical hazard data and equipment structure parameters of the production area.

[0112] Specifically, by analyzing historical hazard data, multiple hazard events are extracted. Each hazard event includes the hazard category, hazard characteristic parameters, and the required data collection points.

[0113] Specifically, data collection points are relevant monitoring points where characteristic parameters of potential hazards can be collected.

[0114] Specifically, the hazard categories in each hazard event are different. The hazard categories include, but are not limited to, behavior violation types (such as violation of regulations, failure to wear required items, entering restricted areas, improper stacking), equipment operation types (such as equipment damage, equipment aging, equipment not in the specified state, e.g., the box door not closed, the switch not closed as required), environmental types (such as temperature exceeding the standard, abnormal gas exceeding the standard), etc., which are various hazards that can cause production risks.

[0115] Specifically, by comprehensively analyzing each hazard event and combining with the equipment parameters in the production area, multiple monitoring points are set. Among them, each monitoring point corresponds to the data collection points in one or more hazard events.

[0116] In the preferred implementation of the embodiment of this application, a first-level identification model is constructed based on all hazard events, including:

[0117] Set ai as the target hazard event in sequence according to the hazard event sequence A;

[0118] Generate an associated data packet for the target hazard event based on the historical hazard data packet;

[0119] Generate an identification value b for the target hazard event according to the associated data packet;

[0120] b = β i *v i ;

[0121] Among them, θ1 is the number of first-level identification indicators; β i is the influence factor of the i-th first-level identification indicator; v i is the reference value of the i-th first-level identification indicator set based on the associated data packet;

[0122] Preset an identification value threshold B1;

[0123] If b > B1, set the target hazard event as a first-level hazard, and generate an identification sub-model for the target hazard event according to the associated data packet;

[0124] If b < B1, set the target hazard event as a second-level hazard;

[0125] Judge each hazard event in sequence;

[0126] Establish a first-level hazard sequence A1 according to the judgment result, A1 = (a 11 , a 12 … a 1i … a 1n1 ), where a 1i is the i-th first-level hazard; n1 is the number of first-level hazards;

[0127] Obtain the identification sub-models for each Level 1 hazard, and set the Level 1 identification model based on all the identification sub-models.

[0128] Specifically, the identification threshold can be set based on historical parameters.

[0129] Specifically, the associated data package includes recorded data on the historical identification process of the target hazard event, including the type of monitoring data and the corresponding identification accuracy based on each identification of the target hazard event.

[0130] Specifically, the primary identification indicators include images, ambient temperature, and time periodicity. Time periodicity refers to the periodic characteristics of the hazard. The more obvious the periodic characteristics of the target hazard event, the higher the reference value of the corresponding primary identification indicator.

[0131] Specifically, when identifying potential hazards, the judgment is made solely based on primary identification indicators. The higher the accuracy of the judgment, the greater the reference value of the primary identification indicator.

[0132] Specifically, the impact factor of each primary identification indicator is determined based on the difficulty of collecting the data within the production area. The greater the difficulty, the smaller the corresponding impact factor.

[0133] Specifically, Level 1 hazards are those that can be identified by collecting image data, ambient temperature data, or making periodic assessments, and are relatively easy to identify.

[0134] Specifically, auxiliary sub-models for generating various potential incidents include:

[0135] Based on the sequence of potential incidents A, a is set sequentially. i Potential hazards to be monitored;

[0136] Select the primary correlation points of the potential hazards to be monitored based on the associated data packets of the potential hazards to be monitored;

[0137] A first-level association table of potential hazards to be monitored is generated based on all first-level association points;

[0138] Generate primary correlation values ​​between the potential hazard events to be monitored and each potential hazard event;

[0139] Establish a first-level related value sequence C, C=(c1, c2…c i …c n ), where c i is the primary correlation value between the monitored hazard event and the i-th hazard event; n is the number of hazard events;

[0140] Preset the first-level correlation threshold C1;

[0141] If c i>C1, define the i-th hidden danger event as a related hidden danger event of the hidden danger event to be monitored;

[0142] A secondary association table is generated based on all associated hazards of the hazard event to be monitored.

[0143] Auxiliary sub-models for potential hazards to be monitored are generated based on the primary and secondary association tables.

[0144] Each potential hazard event auxiliary sub-model is generated sequentially.

[0145] Specifically, if the current monitoring point is the data collection point required for the potential hazard event to be monitored, then it is set as the first-level correlation point of the potential hazard event to be monitored.

[0146] Specifically, based on the risk diffusion parameters of the potential hazard event to be monitored, the correlation between it and the remaining potential hazard events is determined. The primary correlation value is either zero or one. If the probability of the current potential hazard event occurring exceeds a preset probability threshold when the potential hazard event to be monitored occurs, the primary correlation value between the two is set to one. If the probability of the current potential hazard event occurring does not exceed the preset probability threshold when the potential hazard event to be monitored occurs, the primary correlation value between the two is zero.

[0147] Specifically, the threshold value for the first-level correlation value is between zero and one.

[0148] Specifically, the probability of each potential hazard occurring when the monitored hazard occurs can be obtained by analyzing historical parameters. The probability threshold can be set based on historical parameters.

[0149] Specifically, a secondary association table is generated based on all associated hazards of the hazard event to be monitored, including:

[0150] An initial set of hazards is constructed based on the hazard events to be monitored and all related hazards;

[0151] Determine whether there are any Level 1 hazards in the initial hazard set;

[0152] If they exist, generate a secondary association table based on the initial set of potential hazards;

[0153] If none exist, set all Level 1 hazards as associated hazards of the hazard events to be monitored, generate correction instructions for the initial hazard set, and generate a Level 2 association table based on the correction results.

[0154] Specifically, the correction instruction means that if there are no Level 1 hazards in the initial hazard set corresponding to the hazard event to be monitored, all Level 1 hazards will be added to the initial hazard set corresponding to the hazard event to be monitored. This will generate a Level 2 association table for the hazard event to be monitored.

[0155] It is understood that in the above embodiments, multiple monitoring points based on the equipment structure parameters of the production area are used to achieve comprehensive monitoring of the production area. Based on the analysis of historical hidden danger data, an association diagnosis model is constructed to generate a network of associations between monitoring points and single types of hidden dangers, thereby achieving multiple verifications for single types of hidden dangers and reducing the false identification rate of production hidden dangers.

[0156] In a preferred embodiment of this application, the primary monitoring strategy set by the intelligent agent includes:

[0157] Establish a sequence of sub-models for identification, P, P = (p1, p2, ..., p2). i …p n1 ), where p i Let n1 be the identification sub-model for the i-th Level 1 hazard; n1 is the number of Level 1 hazard.

[0158] Based on the identification sub-model sequence P, p is set sequentially. i Identify sub-models for the target;

[0159] Obtain all feature indicators of the target recognition sub-model, and select all secondary association points of the target recognition model based on all feature indicators;

[0160] The monitoring substructure of the target identification sub-model is constructed based on all secondary correlation points;

[0161] The monitoring substructures of each identification sub-model are constructed sequentially, and a first-level monitoring strategy is generated based on all the monitoring substructures.

[0162] Specifically, the secondary correlation points are the monitoring points for collecting primary identification indicators from the primary correlation points of the primary hidden dangers corresponding to the target identification sub-model.

[0163] Specifically, by constructing a monitoring substructure for each identification sub-model, real-time data parameters of each primary identification indicator are collected in real time, and early warnings are issued for potential primary hazards in a timely manner, thereby reducing the overall monitoring cost while ensuring the monitoring efficiency of the production area.

[0164] In a preferred embodiment of this application, an auxiliary monitoring strategy is set, including:

[0165] Multiple packages to be identified are generated based on the initial monitoring data, and a mapping table of primary hazards and packages to be identified is constructed.

[0166] Based on the first-level hidden danger sequence A1, a is set sequentially. 1i The target is a Level 1 hazard;

[0167] The primary identification model obtains the associated packets to be identified for the primary hidden dangers of the target.

[0168] The risk probability value f for generating a Level 1 hidden danger of the target;

[0169] f=[ μ i *s i ];

[0170] Where θ2 represents the number of characteristic indicators of the target's primary hidden danger; μ i The influencing factor of the i-th characteristic indicator of the primary hidden danger; s i The matching value of the i-th feature index of the target first-level hidden danger is generated based on the associated package to be identified;

[0171] Preset risk probability threshold F1;

[0172] If f > F1, generate an auxiliary sub-strategy for the first-level hidden danger of the target;

[0173] The system sequentially determines whether to generate auxiliary sub-strategies for each level-one hidden danger, and generates auxiliary monitoring strategies based on all auxiliary sub-strategies.

[0174] Specifically, the risk probability threshold can be set based on historical parameters. When the risk probability value exceeds the risk probability threshold, it indicates that there is a level-one hidden danger, requiring timely inspection and repair to eliminate the risk.

[0175] Specifically, a higher risk probability value indicates a greater likelihood of a primary hidden danger to the target at the current point in time.

[0176] Specifically, based on the category of the target primary hazard, corresponding characteristic indicators are set. These characteristic indicators are the parameters required to identify the target primary hazard. For example, for behavioral violations, characteristic indicators include, but are not limited to, parameters such as not wearing equipment or entering prohibited areas. For equipment status hazards, these include parameters such as doors not being closed or equipment being moved. By quantifying each characteristic indicator, a precise analysis of the risk probability of the target primary hazard can be achieved.

[0177] Specifically, the larger the fit value of each primary characteristic indicator, the greater the current state of the corresponding parameter, and the greater the possibility of a primary hidden danger occurring in the target.

[0178] Specifically, the influence factors of each characteristic indicator are set according to their correlation with the target level-one hidden danger. The greater the correlation, the larger the value of the corresponding influence factor.

[0179] Specifically, the auxiliary sub-strategies for generating primary-level hidden dangers include:

[0180] The auxiliary sub-model for the primary hidden danger of the target is set as the target auxiliary model;

[0181] A primary acquisition strategy is generated based on the primary association table in the target auxiliary model;

[0182] The verification sub-data packets of the target's primary hidden dangers are obtained based on the primary acquisition strategy;

[0183] Each associated hazard of the target primary hazard is designated as a hazard to be diagnosed, and a secondary data collection strategy is set for all hazards to be diagnosed.

[0184] The diagnostic sub-data packets for each potential hazard to be diagnosed are obtained according to the secondary acquisition strategy;

[0185] Based on the primary and secondary data collection strategies, auxiliary sub-strategies for identifying primary hidden dangers are set.

[0186] Specifically, when a target level-one hazard exists, relevant data of all level-one related points of the target level-one hazard are collected (including both level-one identification indicators and related equipment operation data). By analyzing multi-source data, it is determined whether there is a misjudgment of the target level-one hazard, thereby improving the diagnostic accuracy of level-one hazard.

[0187] Specifically, by analyzing each sub-data packet to be diagnosed, it is determined whether each potential hazard exists at the current time point, thereby generating corresponding hazard diagnosis results. This facilitates maintenance by management personnel, thereby eliminating the corresponding production hazards and ensuring the safe operation of the production area.

[0188] It is understandable that, in the above embodiments, by analyzing the risk diffusion trends among different types of hazards, a network of relationships between various hazards is constructed, enabling coordinated monitoring of different types of hazards and improving the overall efficiency of identifying production hazards within the generated area. This ensures the safe operation of the production area.

[0189] In another preferred embodiment of the AI-based auxiliary inspection method for safety hazards based on any of the above preferred embodiments, this preferred embodiment provides an AI-based auxiliary inspection system for safety hazards, comprising:

[0190] The central control unit is used to construct intelligent agents and multiple monitoring points based on the original parameters of the production area;

[0191] The monitoring unit includes multiple monitoring sub-modules, which are set up at various monitoring points;

[0192] The central control unit includes:

[0193] The first processing module is used to construct the intelligent agent;

[0194] The second processing module is used to obtain initial monitoring data according to the primary monitoring strategy set by the intelligent agent, and to set auxiliary monitoring strategies according to the initial monitoring data.

[0195] The third processing module is used to obtain feedback data packets according to the auxiliary monitoring strategy, and the intelligent agent generates hidden danger diagnosis results based on the feedback data packets.

[0196] Building an intelligent agent includes:

[0197] Generate historical hazard data packets based on the original parameters, and generate multiple hazard events based on the historical hazard data packets;

[0198] Establish a sequence A of potential incidents, A=(a1,a2…a ... i …a n ), where a i Let be the i-th potential hazard; n is the number of potential hazard events;

[0199] A primary identification model is constructed based on all potential hazards, and auxiliary sub-models for each potential hazard are generated sequentially.

[0200] Construct an associated diagnostic model based on all auxiliary sub-models;

[0201] Based on the sequence of potential incidents A, an intelligent agent is constructed using a primary identification model and a correlation diagnosis model.

[0202] Specifically, the monitoring submodule is preferably composed of various types of data acquisition devices. The corresponding data acquisition device is selected according to the data types required for different monitoring points, thereby obtaining various types of monitoring data.

[0203] In a preferred embodiment of this application, the first processing module is further configured to:

[0204] Based on the sequence of potential hazard events A, ai are sequentially designated as target potential hazard events;

[0205] Generate associated data packets for the target hazard event based on historical hazard data packets;

[0206] Generate the identification value b of the target potential hazard event based on the associated data packets;

[0207] b=[ β i *v i ];

[0208] Where θ1 is the number of primary identification indicators; β i v is the influence factor of the i-th primary identification indicator; i It is a reference value for the i-th primary identification indicator set based on the associated data packets;

[0209] Preset recognition threshold B1;

[0210] If b>B1, the target hidden danger event is set as a level one hidden danger, and an identification sub-model of the target hidden danger event is generated based on the associated data packet;

[0211] If b < B1, set the target potential hazard event as a secondary potential hazard;

[0212] Judge each potential hazard event in turn;

[0213] Establish a primary potential hazard sequence A1 according to the judgment results, A1 = (a 11 , a 12 … a 1i … a 1n1 ), where a 1i is the i-th primary potential hazard; n1 is the number of primary potential hazards;

[0214] Obtain the identification sub-models of each primary potential hazard, and set a primary identification model according to all the identification sub-models.

[0215] According to the first concept of the present application, based on the equipment structure parameters of multiple monitoring points in the production area, comprehensive monitoring of the production area is realized, and an associated diagnosis model is constructed according to the analysis of historical potential hazard data, generating an associated relationship network between the monitoring points and a single type of potential hazard, realizing multiple verification for a single type of potential hazard, and reducing the mis-identification rate of production potential hazards.

[0216] According to the second concept of the present application, through the risk diffusion trend between different types of potential hazards, an associated relationship network between various potential hazards is constructed, realizing the linkage monitoring of different types of potential hazards, improving the overall identification efficiency of production potential hazards in the production area, and ensuring the safe operation in the production area.

[0217] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present application, several improvements and substitutions can be made, and these improvements and substitutions should also be regarded as the protection scope of the present application.

Claims

1. A method for assisting in the inspection of safety hazards based on an AI model, characterized in that, including: Construct an agent and multiple monitoring points according to the original parameters of the production area; Obtain initial monitoring data according to the primary monitoring strategy set by the agent, and set an auxiliary monitoring strategy according to the initial monitoring data; Obtain a feedback data packet according to the auxiliary monitoring strategy, and the agent generates a hidden danger diagnosis result according to the feedback data packet; Construct an agent, including: Generate a historical hidden danger data packet according to the original parameters, and generate multiple hidden danger events according to the historical hidden danger packet; Establish a sequence A of potential incidents, A=(a1,a2…a ... i …a n ), where a i Let be the i-th potential hazard; n is the number of potential hazard events; Construct a primary identification model according to all hidden danger events, and sequentially generate auxiliary sub-models for each hidden danger event; Construct an associated diagnosis model according to all auxiliary sub-models; Construct an agent according to the hidden danger event sequence A, the primary identification model and the associated diagnosis model; Construct a primary identification model according to all hidden danger events, including: Sequentially set ai as the target hidden danger event according to the hidden danger event sequence A; Generate an associated data packet for the target hidden danger event based on the historical hidden danger data packet; Generate an identification value b for the target hidden danger event according to the associated data packet; b=[ b i v i ]; Where θ1 is the number of primary identification indicators; β i v is the influence factor of the i-th primary identification indicator; i It is a reference value for the i-th primary identification indicator set based on the associated data packets; Preset an identification value threshold B1; If b > B1, set the target hidden danger event as a primary hidden danger, and generate an identification sub-model for the target hidden danger event according to the associated data packet; If b < B1, set the target hidden danger event as a secondary hidden danger; Sequentially judge each hidden danger event; Based on the judgment results, establish a first-level hidden danger sequence A1, A1=(a 11 ,a 12 …a 1i …a 1n1 ), where a 1i Let n1 be the i-th level 1 hidden danger; n1 is the number of level 1 hidden dangers; Obtain the identification sub-models of each primary hidden danger, and set the primary identification model according to all identification sub-models; Generate auxiliary sub-models for each hidden danger event, including: Based on the sequence of potential incidents A, a is set sequentially. i Potential hazards to be monitored; Select the primary associated points of the to-be-monitored hidden danger event according to the associated data packet of the to-be-monitored hidden danger event; Generate a primary association table for the to-be-monitored hidden danger event according to all primary associated points; Generate a primary association value between the to-be-monitored hidden danger event and each hidden danger event; Establish a first-level related value sequence C, C=(c1, c2…c i …c n ), where c i is the primary correlation value between the monitored hazard event and the i-th hazard event; n is the number of hazard events; Preset a primary association value threshold C1; If c i >C1, define the i-th hidden danger event as a related hidden danger event of the hidden danger event to be monitored; Generate a secondary association table according to all associated hidden dangers of the to-be-monitored hidden danger event; Generate an auxiliary sub-model for the to-be-monitored hidden danger event according to the primary association table and the secondary association table; Sequentially generate auxiliary sub-models for each hidden danger event.

2. The safety hazard auxiliary inspection method based on an AI model as described in claim 1, characterized in that, Generate a secondary association table according to all associated hidden dangers of the to-be-monitored hidden danger event, including: Construct an initial hidden danger set based on the to-be-monitored hidden danger event and all associated hidden dangers; Judge whether there is a primary hidden danger in the initial hidden danger set; If it exists, generate a secondary association table according to the initial hidden danger set; If not, set all primary hidden dangers as the associated hidden dangers of the to-be-monitored hidden danger event, generate a correction instruction for the initial hidden danger set, and generate a secondary association table according to the correction result.

3. The safety hazard auxiliary inspection method based on an AI model as described in claim 1, characterized in that, The primary monitoring strategy set by the agent includes: Establish a sequence of sub-models for identification, P, P = (p1, p2, ..., p2). i …p n1 ), where p i Let n1 be the identification sub-model for the i-th Level 1 hazard; n1 is the number of Level 1 hazard. Based on the identification sub-model sequence P, p is set sequentially. i Sub-model for target identification; Obtain each characteristic index of the target identification sub-model, and select all secondary associated points of the target identification model according to all characteristic indexes; Construct a monitoring sub-structure of the target identification sub-model according to all secondary associated points; Sequentially construct the monitoring sub-structures of each identification sub-model, and generate a primary monitoring strategy according to all monitoring sub-structures.

4. The safety hazard auxiliary inspection method based on an AI model as described in claim 3, characterized in that, Set an auxiliary monitoring strategy, including: Generate multiple to-be-identified packets according to the initial monitoring data, and construct a primary hidden danger - to-be-identified packet mapping table; Based on the first-level hidden danger sequence A1, a is set sequentially. 1i The target is a Level 1 hazard; The primary identification model obtains the associated to-be-identified packet of the target primary hidden danger; Generate a risk probability value f for the target primary hidden danger; f=[ m i s i ]; Where θ2 represents the number of characteristic indicators of the target's primary hidden danger; μ i The influencing factor of the i-th characteristic indicator of the primary hidden danger; s i The matching value of the i-th feature index of the target first-level hidden danger is generated based on the associated package to be identified; Preset a risk probability value threshold F1; If f > F1, generate an auxiliary sub-strategy for the target primary hidden danger; Successively determine whether to generate the auxiliary sub-strategies for each first-level hidden danger, and generate the auxiliary monitoring strategy based on all the auxiliary sub-strategies.

5. The safety hazard auxiliary inspection method based on an AI model as described in claim 4, characterized in that, Generate the auxiliary sub-strategies for the target first-level hidden danger, including: Set the auxiliary sub-model of the target first-level hidden danger as the target auxiliary model; Generate the first-level acquisition strategy according to the first-level association table in the target auxiliary model; Obtain the verification sub-data packet of the target first-level hidden danger based on the first-level acquisition strategy; Set each associated hidden danger of the target first-level hidden danger as the to-be-diagnosed hidden danger, and set the second-level acquisition strategy for all the to-be-diagnosed hidden dangers; Obtain the diagnosis sub-data packet of each to-be-diagnosed hidden danger according to the second-level acquisition strategy; Set the auxiliary sub-strategy of the target first-level hidden danger according to the first-level acquisition strategy and the second-level acquisition strategy.

6. A safety hazard auxiliary inspection system based on an AI model, employing the safety hazard auxiliary inspection method based on an AI model as described in any one of claims 1-5, characterized in that, Including: The central control unit is used to construct an intelligent agent and multiple monitoring points according to the original parameters of the production area; The monitoring unit includes multiple monitoring sub-modules, and the monitoring sub-modules are set at each monitoring point; The central control unit includes: The first processing module is used to construct an intelligent agent; The second processing module is used to obtain the initial monitoring data according to the first-level monitoring strategy set by the intelligent agent, and set the auxiliary monitoring strategy according to the initial monitoring data; The third processing module is used to obtain the feedback data packet according to the auxiliary monitoring strategy, and the intelligent agent generates the hidden danger diagnosis result according to the feedback data packet; The construction of the intelligent agent includes: Generate the historical hidden danger data packet according to the original parameters, and generate multiple hidden danger events according to the historical hidden danger packet; Establish a sequence A of potential incidents, A=(a1,a2…a ... i …a n ), where a i Let be the i-th potential hazard; n is the number of potential hazard events; Construct the first-level recognition model according to all the hidden danger events, and successively generate the auxiliary sub-models for each hidden danger event; Construct the associated diagnosis model according to all the auxiliary sub-models; Construct an intelligent agent according to the hidden danger event sequence A, the first-level recognition model and the associated diagnosis model.

7. The safety production hazard auxiliary inspection system based on an AI model as described in claim 6, characterized in that, The first processing module is also used for: Successively set ai as the target hidden danger event according to the hidden danger event sequence A; Generate the associated data packet of the target hidden danger event based on the historical hidden danger data packet; Generate the recognition value b of the target hidden danger event according to the associated data packet; b=[ b i v i ]; Where θ1 is the number of primary identification indicators; β i v is the influence factor of the i-th primary identification indicator; i It is a reference value for the i-th primary identification indicator set based on the associated data packets; Preset the recognition value threshold B1; If b > B1, set the target hidden danger event as the first-level hidden danger, and generate the recognition sub-model of the target hidden danger event according to the associated data packet; If b < B1, set the target hidden danger event as the second-level hidden danger; Successively judge each hidden danger event; Based on the judgment results, establish a first-level hidden danger sequence A1, A1=(a 11 ,a 12 …a 1i …a 1n1 ), where a 1i Let n1 be the i-th level 1 hidden danger; n1 is the number of level 1 hidden dangers; Obtain the recognition sub-models of each first-level hidden danger, and set the first-level recognition model according to all the recognition sub-models.

Images