A smelting plant personnel behavior anomaly monitoring method and system

Abstract

The application discloses a smelting plant personnel behavior abnormality monitoring method and system, relates to the smelting plant monitoring technical field, constructs a three-dimensional analysis model, sets an abnormal sensitivity coefficient in combination with a work area risk type, provides a quantitative basis for abnormal detection, configures a virtual human body model, captures personnel behavior trajectories and actions by using visual analysis technology, determines basic abnormality parameters, uses a simulation simulation identification library to deeply analyze real-time behaviors of the virtual human body model, dynamically determines observation abnormality parameters, comprehensively calculates abnormality degrees based on the basic abnormality parameters, the observation abnormality parameters and the abnormal sensitivity coefficient, and realizes accurate alarm. The application solves the problems that traditional manual monitoring is easy to be missed and is not accurate, significantly improves the efficiency and reliability of smelting plant safety monitoring by means of automatic and intelligent monitoring means, and provides strong technical support for safety production.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smelting plant monitoring, in particular to a smelting plant personnel behavior anomaly monitoring method and system. BACKGROUND

[0002] In high-risk industrial environments such as smelting plants, personnel behavior safety monitoring is an important link to ensure production and personnel life safety. However, the existing monitoring methods mainly rely on manual monitoring, which has significant drawbacks and needs to be improved. The traditional manual monitoring method has the following main problems: first, due to the complex smelting plant environment and dynamic personnel behavior, it is difficult for monitoring personnel to maintain high attention for a long time, and key abnormal behaviors are easily missed due to fatigue or negligence, resulting in safety hazards that cannot be discovered in time. Secondly, the accuracy of manual monitoring is uneven, and it is difficult to systematically analyze and quantitatively evaluate complex three-dimensional spatial layout and personnel behavior characteristics, resulting in insufficient recognition accuracy of abnormal behavior. In addition, manual monitoring lacks the ability to analyze historical accident data and real-time behavior performance, and cannot dynamically adjust the monitoring strategy, further reducing the effectiveness and timeliness of monitoring. Finally, the traditional method is based on single observation results, lacks a multi-parameter comprehensive evaluation mechanism, resulting in high false positive or false negative rates, and is difficult to meet the high standard requirements of modern industrial safety monitoring. SUMMARY

[0003] The purpose of the present application is to provide a monitoring method and system that can accurately analyze smelting plant personnel behavior anomalies.

[0004] The present application discloses a smelting plant personnel behavior anomaly monitoring method, comprising:

[0005] Obtain the plan design drawing of the smelting plant, and construct a three-dimensional analysis model of the smelting plant based on the plan design drawing, set abnormal sensitivity coefficients for different position space points of the three-dimensional analysis model of the smelting plant based on the work content of different work areas of the smelting plant;

[0006] Obtain the human body features and behavior features of personnel in different work areas of the smelting plant, configure a preset virtual human body model based on the human body features and behavior features, map the corresponding virtual human body model in the three-dimensional analysis model of the smelting plant, and perform initial behavior analysis on the performance of the virtual human body model to determine the basic abnormal parameters;

[0007] A plurality of personnel behavior identification rules are constructed, each of which includes a plurality of behavior description keywords, and the virtual human model is driven to perform simulation performance according to a preset operation mode, and the key behavior segments in the simulation performance are intercepted and marked, and a plurality of description keywords are configured, if the combination of the description keywords in the continuous time period meets the personnel behavior identification rule, the corresponding simulation performance is identified, and the simulation performance and the personnel behavior identification are constructed into a simulation identification library in a corresponding manner;

[0008] The virtual human model in the smelting plant three-dimensional analysis model is subjected to deep behavior analysis by using the simulation identification library, and the observation abnormality parameter is determined;

[0009] The abnormality degree of the virtual human model is determined in combination with the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human model, and an alarm is given based on the abnormality degree.

[0010] In the embodiments disclosed in the present application, based on the work content of different work areas in the smelting plant, the method for setting abnormality sensitivity coefficients for different position space points of the smelting plant three-dimensional analysis model includes:

[0011] According to the positions of different work areas, the smelting plant three-dimensional analysis model is divided into three-dimensional work blocks, including a three-dimensional melting work block, a three-dimensional casting work block, a three-dimensional raw material storage work block and a three-dimensional equipment maintenance block;

[0012] Based on the label of each three-dimensional work block, the risk type label is determined, and the risk performance of each risk type label is analyzed, based on the risk performance, the risk parameter of the different position space points in the three-dimensional equipment work block is determined, and based on the risk parameter, the abnormality sensitivity coefficient of the corresponding position space point is determined.

[0013] In the embodiments disclosed in the present application, the risk type label includes:

[0014] For the three-dimensional melting work block, high-temperature risk and liquid metal splashing risk;

[0015] For the three-dimensional casting work block, mechanical injury risk and high-temperature risk;

[0016] For the three-dimensional raw material storage work block, fire risk, explosion risk and chemical leakage risk;

[0017] For the three-dimensional equipment maintenance block, mechanical injury risk and electric shock risk.

[0018] In the embodiments disclosed in the present application, the method for analyzing the risk performance of each risk type label includes:

[0019] Acquire historical accident data of the smelting plant, determine risk probability and risk consequence severity parameter of different position space points, and calculate the product of the risk probability and the risk consequence severity parameter to obtain a risk parameter;

[0020] Different risk parameters have corresponding abnormal sensitivity coefficients.

[0021] In the embodiments disclosed in the present application, the method for configuring a preset virtual human body model based on human features and behavior features comprises:

[0022] The visual analysis technology is used to determine the positioning of the plant personnel in the monitoring video and determine the identity, clothing, behavior trajectory and body movement of the plant personnel;

[0023] The virtual human body model is constructed, including a trunk expression line, an upper limb expression line and a lower limb expression line, wherein the upper limb expression line and the lower limb expression line are movably connected to the side and bottom of the trunk expression line, an upper limb swing mapping sphere is arranged at the connection between the upper limb expression line and the trunk expression line, and a lower limb swing mapping sphere is arranged at the connection between the lower limb expression line and the trunk expression line.

[0024] If the determined identity and clothing of the plant personnel meet the preset standard, the position change of the virtual human body model in the three-dimensional analysis model of the smelting plant is determined based on the determined behavior trajectory, and the swing change of the upper limb expression line and the lower limb expression line relative to the trunk expression line is determined based on the determined body movement.

[0025] In the embodiments disclosed in the present application, the method for performing initial behavior analysis on the performance of the virtual human body model comprises:

[0026] The historical accident data and the simulation accident data of the smelting plant are analyzed to determine the device working state, the marked behavior trajectory and the marked body movement of the plant personnel when each accident occurs, and the marked behavior trajectory and the marked body movement are analyzed to determine whether there is an abnormal situation, if there is an abnormal situation, the corresponding marked behavior trajectory is identified as an abnormal behavior trajectory, and the corresponding marked body movement is recorded as an abnormal body movement.

[0027] The abnormal behavior trajectory and the abnormal body movement are subjected to key abnormal section division to determine an abnormal behavior trajectory comparison group and an abnormal body movement comparison group, and the method for key abnormal section division comprises:

[0028] The end analysis time node is determined as the end time node of the accident, and the end analysis time node is pushed forward by a preset time period to determine a start analysis time node, and the time period between the start analysis time node and the end analysis time node is recorded as an abnormal analysis time section.

[0029] A plurality of abnormal analysis time nodes are uniformly set for the abnormal analysis time section, and a device working state corresponding to each abnormal analysis time node is determined, and based on the device working state, the standard behavior track set of the factory worker and the standard body action set of the factory worker corresponding to each abnormal analysis time node are determined;

[0030] The labeled behavior track corresponding to the abnormal analysis time node is compared with the standard behavior track set, if no corresponding standard behavior track is found in the standard behavior track set, the labeled behavior track is identified as an abnormal behavior track, and the labeled body action corresponding to the abnormal analysis time node is compared with the standard body action set, if no corresponding standard body action is found in the standard body action set, the labeled body action is identified as an abnormal body action;

[0031] The time relationship analysis is performed on different abnormal behavior tracks and abnormal body actions in the abnormal analysis time section, the sequence, time interval and device state corresponding to the abnormal behavior track of the different abnormal behavior tracks are determined, and are combined into an abnormal behavior track comparison group, the sequence, time interval and corresponding device state of the different abnormal body actions are determined, and are combined into an abnormal body action comparison group;

[0032] In step S206, a plurality of abnormal behavior track comparison groups are combined to obtain an abnormal behavior track comparison set, a plurality of abnormal body action comparison groups are combined to obtain an abnormal body action comparison set, the virtual human body model is compared with the abnormal behavior track comparison set and the abnormal body action comparison set respectively, the first corresponding parameter of the virtual human body model and the different abnormal behavior track comparison groups and the second corresponding parameter of the virtual human body model and the different abnormal body action comparison groups are calculated, the average value of the first corresponding parameter and the average value of the second corresponding parameter are calculated, and the first average corresponding parameter and the second average corresponding parameter are obtained;

[0033] Based on the first average corresponding parameter and the second average corresponding parameter, a basic abnormal parameter is determined.

[0034] In the embodiments disclosed in the present application, the method for calculating the basic abnormal parameter comprises:

[0035] The device working state corresponding to the virtual human body model at the moment is determined, and based on the equivalence of the device working state, the abnormal behavior track comparison group and the abnormal body action comparison group are screened, and the screened abnormal behavior track comparison group and the abnormal body action comparison group are compared with the performance of the virtual human body model respectively, and the track comparison result and the body action comparison result are determined;

[0036] In the trajectory comparison result, a first identical number ratio of a first identical number of identical abnormal behavior trajectories to a first recorded number of abnormal behavior trajectories in the abnormal behavior trajectory comparison group is calculated, and based on the first identical number ratio, a first coincidence parameter is determined; in the body movement comparison result, a second identical number ratio of a second identical number of identical abnormal body movements to a second recorded number of abnormal body movements in the abnormal body movement comparison group is calculated;

[0037] Based on the first identical number ratio, a first coincidence parameter is determined, based on the second identical number ratio, a second coincidence parameter is determined, a first average coincidence parameter and a second average coincidence parameter are calculated, and a basic abnormal parameter is determined;

[0038] Wherein, the expression for calculating the basic abnormal parameter is:

[0039] ;

[0040] Wherein, is the basic abnormal parameter, is an abnormal behavior trajectory comparison weight adjustment coefficient, is an abnormal body movement comparison weight adjustment coefficient, is the first identical number ratio corresponding to the i1th abnormal behavior trajectory comparison group, is a first identical number influence adjustment coefficient, is a first identical number influence adjustment constant, is the total number of abnormal behavior trajectory comparison groups screened out, is the second identical number ratio corresponding to the i2th abnormal body movement comparison group, is a second identical number influence adjustment coefficient, is a second identical number influence adjustment constant, is the total number of abnormal body movement comparison groups screened out.

[0041] In the embodiments disclosed in the present application, the method for performing deep behavior analysis on a virtual human model in a three-dimensional analysis model of a smelting plant by using a simulation recognition library includes:

[0042] According to a preset time length, the current behavior characteristics of the virtual human model are intercepted to obtain a current behavior characteristic performance segment, and the simulation recognition library is used to recognize the current behavior segments in the current behavior characteristic performance segment, and based on the recognition result, each current behavior segment is marked with a description keyword;

[0043] Step S402, analyze the segment proportion of the current behavior segment marked with the description keyword in the current behavior performance segment, and based on the preset proportion interval to which the segment proportion belongs, determine an observation abnormal parameter.

[0044] In the disclosed embodiments, the abnormality degree of the virtual human model is determined according to the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human model, and the expression of the abnormality degree of the virtual human model is:

[0045] ;

[0046] wherein, is the abnormality degree, is the abnormality sensitivity coefficient, is the basic abnormality parameter weight adjustment coefficient, is the observation abnormality parameter weight adjustment coefficient, is the basic abnormality parameter, is the observation abnormality parameter.

[0047] In the disclosed embodiments, a smelting plant personnel behavior abnormality monitoring system is also disclosed, comprising:

[0048] A first module is configured to acquire a plan design drawing of the smelting plant, construct a three-dimensional analysis model of the smelting plant based on the plan design drawing, and set abnormality sensitivity coefficients for different position space points of the three-dimensional analysis model of the smelting plant based on the work contents of different work areas of the smelting plant.

[0049] A second module is configured to acquire human body features and behavior features of personnel in different work areas of the smelting plant, configure preset virtual human models based on the human body features and behavior features, map the corresponding virtual human models in the three-dimensional analysis model of the smelting plant, and perform initial behavior analysis on the virtual human models to determine basic abnormality parameters.

[0050] A third module is configured to construct a plurality of personnel behavior identification rules, each of which includes a plurality of behavior description keywords, and drive the virtual human models to perform simulation performance according to a preset control mode, mark and configure a plurality of description keywords on key behavior segments in the simulation performance, and if the combination of the description keywords in a continuous time period meets the personnel behavior identification rules, the corresponding simulation performance is identified as personnel behavior, and the simulation performance and the personnel behavior identification are constructed into a simulation identification library in a corresponding manner.

[0051] A fourth module is configured to perform deep behavior analysis on the virtual human models in the three-dimensional analysis model of the smelting plant by using the simulation identification library to determine observation abnormality parameters.

[0052] A fifth module is configured to determine the abnormality degree of the virtual human model by combining the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human model, and to perform alarm based on the abnormality degree.

[0053] The application provides a smelting plant personnel behavior anomaly monitoring method and system, relates to the smelting plant monitoring technical field, constructs a three-dimensional analysis model, sets an abnormal sensitive coefficient in combination with a work area risk type, provides a quantitative basis for anomaly detection, configures a virtual human body model, captures personnel behavior trajectories and actions by using visual analysis technology, determines basic anomaly parameters, performs deep analysis on real-time behaviors of the virtual human body model by using a simulation identification library, dynamically determines observation anomaly parameters, comprehensively calculates anomaly degrees based on the basic anomaly parameters, the observation anomaly parameters and the abnormal sensitive coefficient, and realizes accurate alarm. The application solves the problems that traditional manual monitoring is easy to be missed and is not accurate, significantly improves the efficiency and reliability of smelting plant safety monitoring by using automatic and intelligent monitoring means, and provides strong technical support for safety production.

[0054] The technical scheme of the application will be further described below with the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A method step diagram of the smelting plant personnel behavior anomaly monitoring method disclosed in the embodiment of the application. DETAILED DESCRIPTION

[0056] The technical scheme of the application will be further described below with the drawings and examples.

[0057] The technical scheme of the application will be further described below with the drawings and examples.

[0058] Embodiment:

[0059] The application discloses a smelting plant personnel behavior anomaly monitoring method, and refers to Figure 1 , comprising:

[0060] In step S100, a plan design drawing of the smelting plant is acquired, and a three-dimensional analysis model of the smelting plant is constructed based on the plan design drawing; abnormal sensitive coefficients are set for different position space points of the three-dimensional analysis model of the smelting plant based on work contents of different work areas of the smelting plant.

[0061] Firstly, the plan design drawing of the smelting plant is acquired, and a three-dimensional analysis model is constructed based on the design drawing. Then, according to the work content of different work areas of the smelting plant, the three-dimensional model is divided into multiple work blocks, such as a three-dimensional smelting work block, a three-dimensional casting work block, a three-dimensional raw material storage work block, and a three-dimensional equipment maintenance block. For each work block, the potential risk types (such as high temperature, mechanical injury, fire, etc.) are analyzed, and combined with historical accident data, the risk parameters (the product of risk probability and consequence severity) of each position point are calculated. Finally, based on the risk parameters, the abnormal sensitivity coefficient of each position point is set (the higher the risk, the greater the sensitivity coefficient), which provides a spatial dimension risk assessment basis for subsequent personnel behavior anomaly monitoring.

[0062] Step S200, the human body features and behavior features of the personnel in different work areas of the smelting plant are acquired, and a preset virtual human body model is configured based on the human body features and behavior features, and the corresponding virtual human body model is mapped in the three-dimensional analysis model of the smelting plant, and the performance of the virtual human body model is analyzed to determine the basic abnormal parameters.

[0063] Firstly, a plurality of personnel behavior identification rules are constructed, each rule including a plurality of behavior description keywords. Then, the virtual human body model is driven to perform simulation performance, and key behavior segments are intercepted and labeled with description keywords. If the combination of description keywords meets the identification rule, the performance is identified for personnel behavior. Finally, the simulation performance and the corresponding identification results are combined to form a simulation recognition library, which provides identification basis for subsequent deep behavior analysis.

[0064] Step S300, a plurality of personnel behavior identification rules are constructed, each personnel behavior identification rule including a plurality of behavior description keywords, and the virtual human body model is driven to perform simulation performance according to a preset control mode, and key behavior segments in the simulation performance are intercepted and labeled with a plurality of description keywords. If the combination of description keywords meets the personnel behavior identification rule in a continuous time period, the corresponding simulation performance is identified for personnel behavior, and the simulation performance and the personnel behavior identification are constructed to obtain a simulation recognition library in a corresponding manner.

[0065] Step S400, the simulation recognition library is used to perform deep behavior analysis on the virtual human body model in the three-dimensional analysis model of the smelting plant to determine the observation abnormal parameters.

[0066] According to the preset time length, the current behavior characteristics of the virtual human body model are intercepted to obtain a current behavior characteristic performance segment. The behavior segments in the current behavior characteristic performance segment are identified by using a simulation identification library, and each segment is marked with a description keyword. The proportion of the behavior segments marked with the description keyword in the entire performance segment is analyzed, and based on the preset proportion interval to which the segment proportion belongs, an observation abnormality parameter is determined, and the abnormal performance of the virtual human body model is further quantified by comparing the simulation identification library.

[0067] In step S500, the abnormal degree of the virtual human body model is determined by combining the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human body model, and an alarm is given based on the abnormal degree.

[0068] The abnormal degree of the virtual human body model is calculated by combining the basic abnormality parameter, the observation abnormality parameter and the abnormality sensitivity coefficient of the corresponding position point. According to the abnormal degree, it is judged whether to trigger an alarm. The calculation formula of the abnormal degree comprehensively considers the basic abnormality parameter weight adjustment coefficient, the observation abnormality parameter weight adjustment coefficient and the abnormality sensitivity coefficient. Through this method, accurate monitoring and real-time alarm of personnel behavior abnormality are realized.

[0069] In the embodiments disclosed in the present application, based on the work content of different working areas of the smelting plant, the method for setting abnormality sensitivity coefficients for different position space points of the three-dimensional analysis model of the smelting plant includes:

[0070] In step S101, according to the positions of different working areas, three-dimensional working blocks of the three-dimensional analysis model of the smelting plant are demarcated, including three-dimensional melting working blocks, three-dimensional casting working blocks, three-dimensional raw material storage working blocks and three-dimensional equipment maintenance blocks.

[0071] According to the positions of different working areas of the smelting plant, the three-dimensional analysis model is divided into multiple three-dimensional working blocks, including three-dimensional melting working blocks, three-dimensional casting working blocks, three-dimensional raw material storage working blocks and three-dimensional equipment maintenance blocks. The basis for division is the work content and potential risk characteristics of each area, for example, the melting block involves the risk of high temperature and liquid metal splashing, the casting block may have mechanical injury and high temperature risk, the raw material storage block may have fire, explosion and chemical leakage risk, and the equipment maintenance block faces mechanical injury and electric shock risk. By clearly defining the functions and risk characteristics of each working block, a foundation is laid for subsequent risk analysis and abnormality sensitivity coefficient setting.

[0072] Step S102, based on the label corresponding to the risk type of each three-dimensional work block, the risk performance of each risk type label is analyzed, based on the risk performance, the risk parameter of the different position space points in the three-dimensional equipment work block is determined, and based on the risk parameter, the abnormal sensitivity coefficient of the corresponding position space point is determined.

[0073] For each three-dimensional work block, its corresponding risk type label is marked, such as high temperature risk, liquid metal splash risk, mechanical injury risk, fire risk, explosion risk and chemical leakage risk. Based on each risk type label, its risk performance is analyzed, including calculating the risk probability (based on historical accident data) of different position space points and evaluating the severity of risk consequences (such as personnel casualty and equipment damage). The risk probability is multiplied by the consequence severity to obtain the risk parameter of each position space point. Finally, based on the risk parameter, the abnormal sensitivity coefficient of the corresponding position space point is set, and the higher the risk parameter, the greater the abnormal sensitivity coefficient. This method ensures that the abnormal sensitivity coefficient is set to match the actual risk level, providing a scientific and quantitative basis for personnel behavior anomaly monitoring.

[0074] In the embodiments disclosed in the present application, the risk type label includes:

[0075] For the three-dimensional smelting work block, high temperature risk and liquid metal splash risk; for the three-dimensional casting work block, mechanical injury risk and high temperature risk; for the three-dimensional raw material storage work block, fire risk, explosion risk and chemical leakage risk; for the three-dimensional equipment maintenance block, mechanical injury risk and electric shock risk.

[0076] In the embodiments disclosed in the present application, the method of analyzing the risk performance of each risk type label includes:

[0077] Step S1021, obtaining the historical accident data of the smelting plant, determining the risk probability and risk consequence severity parameter of different position space points, and calculating the product of the risk probability and the risk consequence severity parameter to obtain the risk parameter.

[0078] By obtaining the historical accident data of the smelting plant, the risk probability and risk consequence severity parameter of different position space points are determined. The risk probability is calculated based on the frequency of historical accidents, and the risk consequence severity parameter is quantified according to the loss degree (such as personnel casualty and equipment damage) caused by the accident. The risk probability is multiplied by the risk consequence severity parameter to obtain the risk parameter of each position space point. This process provides a specific quantitative basis for the subsequent setting of the abnormal sensitivity coefficient.

[0079] Step S1022, for different risk parameters, set corresponding abnormal sensitivity coefficients.

[0080] The risk parameter calculated for each position space point is set with a corresponding abnormal sensitivity coefficient. The greater the risk parameter, the higher the potential risk of the position point, and therefore the abnormal sensitivity coefficient will also increase accordingly. Through this corresponding relationship, it is ensured that the high-risk area is given higher attention in personnel behavior monitoring, thereby improving the safety management effect.

[0081] In the embodiments disclosed in the present application, the method for configuring a preset virtual human body model based on human body features and behavior features comprises:

[0082] Step S201, using visual analysis technology, the factory personnel in the monitoring video are positioned, and the identity, clothing, behavior trajectory and body action of the factory personnel are determined.

[0083] Through visual analysis technology, the factory personnel in the monitoring video are accurately positioned, and their identity, clothing, behavior trajectory and body action are identified. The identification of identity and clothing ensures the accuracy and compliance of personnel behavior analysis, and the extraction of behavior trajectory and body action provides basic data support for the construction of the virtual human body model.

[0084] Step S202, a virtual human body model is constructed, including a torso expression line, an upper limb expression line and a lower limb expression line, wherein the upper limb expression line and the lower limb expression line are movably connected to the side and bottom of the torso expression line, an upper limb swing mapping sphere is set at the connection between the upper limb expression line and the torso expression line, and a lower limb swing mapping sphere is set at the connection between the lower limb expression line and the torso expression line.

[0085] Based on the body features and behavior features of the factory personnel, a virtual human body model is constructed. The model includes a torso expression line, an upper limb expression line and a lower limb expression line, wherein the upper limb expression line and the lower limb expression line are movably connected to the side and bottom of the torso expression line. An upper limb swing mapping sphere is set at the connection between the upper limb expression line and the torso expression line, and a lower limb swing mapping sphere is set at the connection between the lower limb expression line and the torso expression line. These designs can accurately simulate the limb action of the human body and provide technical support for subsequent behavior anomaly analysis.

[0086] Step S203, if the identity and clothing of the determined factory personnel meet the preset standard, the position change of the virtual human body model in the smelting plant three-dimensional analysis model is determined based on the determined behavior trajectory, and the swing change of the upper limb expression line and the lower limb expression line relative to the torso expression line is determined based on the determined body action.

[0087] If the identity and clothing of the plant personnel meet the preset standards, the position change of the virtual human model in the three-dimensional analysis model of the smelting plant is determined based on the behavior trajectory of the plant personnel. At the same time, the swing change of the upper limb expression line and the lower limb expression line relative to the trunk expression line is determined based on the body action of the plant personnel. This process maps the behavior of the real personnel to the virtual human model, providing dynamic data support for subsequent behavior anomaly detection and risk warning.

[0088] In the embodiments disclosed in the present application, the method for performing initial behavior analysis on the performance of the virtual human model includes:

[0089] In step S204, the historical accident data and the simulated accident data of the smelting plant are analyzed to determine the working state of the equipment, the marked behavior trajectory of the plant personnel and the marked body action at the time of each accident, and the marked behavior trajectory and the marked body action are analyzed to determine whether there is an abnormal situation. If there is an abnormal situation, the corresponding marked behavior trajectory is identified as an abnormal behavior trajectory, and the corresponding marked body action is recorded as an abnormal body action.

[0090] The historical accident data and the simulated accident data of the smelting plant are analyzed to determine the working state of the equipment, the behavior trajectory (marked behavior trajectory) of the plant personnel and the body action (marked body action) at the time of each accident. By analyzing the marked behavior trajectory and the marked body action, it is determined whether there is an abnormal situation. If there is an abnormal situation, the corresponding marked behavior trajectory is identified as an abnormal behavior trajectory, and the corresponding marked body action is recorded as an abnormal body action.

[0091] In step S205, the abnormal behavior trajectory and the abnormal body action are divided into key abnormal sections to determine the abnormal behavior trajectory comparison group and the abnormal body action comparison group. The method for dividing the key abnormal sections includes.

[0092] The abnormal behavior trajectory and the abnormal body action are divided into key abnormal sections to generate the abnormal behavior trajectory comparison group and the abnormal body action comparison group. The specific method includes the following steps:

[0093] In step S2051, the time node of the end of the accident is analyzed, and the end analysis time node is pushed forward by a preset time period to determine the start analysis time node. The time period between the start analysis time node and the end analysis time node is recorded as the abnormal analysis time section.

[0094] The time node of the end of the accident is analyzed, and the end analysis time node is pushed forward by a preset time period (for example, 5 minutes before the accident) to determine the start analysis time node. The time period between the start analysis time node and the end analysis time node is recorded as the abnormal analysis time section.

[0095] Step S2052, a plurality of abnormal analysis time nodes are uniformly set for the abnormal analysis time section, and the device working state corresponding to each abnormal analysis time node is determined, and based on the device working state, the standard behavior trajectory set of the factory worker and the standard body action set of each abnormal analysis time node are determined.

[0096] A plurality of abnormal analysis time nodes are uniformly set in the abnormal analysis time section, and the device working state corresponding to each time node is determined. Based on the device working state, the standard behavior trajectory set and the standard body action set of the factory worker at each time node are determined.

[0097] Step S2053, the labeled behavior trajectory corresponding to the abnormal analysis time node is compared with the standard behavior trajectory set, if no corresponding standard behavior trajectory is found in the standard behavior trajectory set, the corresponding labeled behavior trajectory is identified as an abnormal behavior trajectory, and the labeled body action corresponding to the abnormal analysis time node is compared with the standard body action set, if no corresponding standard body action is found in the standard body action set, the corresponding labeled body action is identified as an abnormal body action.

[0098] The labeled behavior trajectory of each abnormal analysis time node is compared with the standard behavior trajectory set. If no corresponding standard behavior trajectory is found in the standard behavior trajectory set, the corresponding labeled behavior trajectory is identified as an abnormal behavior trajectory. Similarly, the labeled body action of each abnormal analysis time node is compared with the standard body action set, if no corresponding standard body action is found, the corresponding labeled body action is identified as an abnormal body action.

[0099] Step S2054, the time relationship analysis is performed on different abnormal behavior trajectories and abnormal body actions in the abnormal analysis time section, the sequence, time interval and device state corresponding to the abnormal behavior trajectory of different abnormal behavior trajectories are determined, and are combined into an abnormal behavior trajectory comparison group, the sequence, time interval and corresponding device state of different abnormal body actions are determined, and are combined into an abnormal body action comparison group.

[0100] The time relationship analysis is performed on different abnormal behavior trajectories and abnormal body actions in the abnormal analysis time section, including determining the sequence, time interval and corresponding device state of the abnormal behavior trajectory, and combining them into an abnormal behavior trajectory comparison group. At the same time, the abnormal body action is analyzed similarly to generate an abnormal body action comparison group.

[0101] Step S206, combine the plurality of abnormal behavior trajectory comparison groups to obtain an abnormal behavior trajectory comparison set, combine the plurality of abnormal body movement comparison groups to obtain an abnormal body movement comparison set, respectively compare the virtual human body model with the abnormal behavior trajectory comparison set and the abnormal body movement comparison set, calculate the first matching parameter of the virtual human body model and different abnormal behavior trajectory comparison groups, and the second matching parameter of the virtual human body model and different abnormal body movement comparison groups, calculate the average value of the first matching parameter and the average value of the second matching parameter, and obtain the first average matching parameter and the second average matching parameter.

[0102] The plurality of abnormal behavior trajectory comparison groups are combined into an abnormal behavior trajectory comparison set, and the plurality of abnormal body movement comparison groups are combined into an abnormal body movement comparison set. The virtual human body model is compared with the comparison set, and the following parameters are calculated:

[0103] The first matching parameter: the matching degree of the virtual human body model and different abnormal behavior trajectory comparison groups.

[0104] The second matching parameter: the matching degree of the virtual human body model and different abnormal body movement comparison groups.

[0105] The average value of all first matching parameters is calculated to obtain the first average matching parameter, and the average value of all second matching parameters is calculated to obtain the second average matching parameter.

[0106] Step S207, based on the first average matching parameter and the second average matching parameter, determine the basic abnormal parameter.

[0107] Based on the first average matching parameter and the second average matching parameter, the basic abnormal parameter of the virtual human body model is determined. The basic abnormal parameter is used to quantify the matching degree of the virtual human body model and the abnormal behavior trajectory and body movement, and provides a basis for subsequent risk warning and safety management.

[0108] In the embodiments disclosed in the present application, the method for calculating the basic abnormal parameter comprises:

[0109] Step S2061, determine the corresponding device working state of the virtual human body model at the moment, and based on the equivalence of the device working state, screen the abnormal behavior trajectory comparison group and the abnormal body movement comparison group, and compare the screened abnormal behavior trajectory and abnormal body movement comparison group with the performance of the virtual human body model respectively, to determine the trajectory comparison result and the body movement comparison result.

[0110] In the trajectory comparison result, a first identical number ratio of a first identical number of abnormal behavior trajectories to a first recorded number of abnormal behavior trajectories in the abnormal behavior trajectory comparison group is calculated, and based on the first identical number ratio, a first coincidence parameter is determined; in the body movement comparison result, a second identical number ratio of a second identical number of abnormal body movements to a second recorded number of abnormal body movements in the abnormal body movement comparison group is calculated.

[0111] In step 2063, based on the first identical number ratio, the first coincidence parameter is determined, based on the second identical number ratio, the second coincidence parameter is determined, the first average coincidence parameter and the second average coincidence parameter are calculated, and the basic abnormal parameter is determined.

[0112] The expression for calculating the basic abnormal parameter is:

[0113] .

[0114] wherein, the basic abnormal parameter, the abnormal behavior trajectory comparison weight adjustment coefficient, the abnormal body movement comparison weight adjustment coefficient, the first identical number ratio corresponding to the i1th abnormal behavior trajectory comparison group, the first identical number influence adjustment coefficient, the first identical number influence adjustment constant, the total number of the screened abnormal behavior trajectory comparison groups, the second identical number ratio corresponding to the i2th abnormal body movement comparison group, the second identical number influence adjustment coefficient, the second identical number influence adjustment constant, the total number of the screened abnormal body movement comparison groups.

[0115] In the embodiment disclosed in the present application, the method for performing deep behavior analysis on the virtual human model in the three-dimensional analysis model of the smelting plant by using the simulation recognition library comprises:

[0116] In step S401, the current behavior characteristics of the virtual human model are intercepted according to a preset time length to obtain a current behavior characteristic performance segment, and the simulation recognition library is used to recognize the current behavior segment in the current behavior characteristic performance segment, and based on the recognition result, a description keyword is marked for each current behavior segment.

[0117] In step S402, the segment proportion of the current behavior segment marked with the description keyword in the current behavior performance segment is analyzed, and based on the preset proportion interval to which the segment proportion belongs, an observation abnormal parameter is determined.

[0118] In the embodiment disclosed by the application, the expression of the abnormal degree of the virtual human body model is determined in combination with the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human body model, and is as follows:

[0119] .

[0120] wherein, is the abnormal degree, is the abnormality sensitivity coefficient, is the basic abnormality parameter weight adjustment coefficient, is the observation abnormality parameter weight adjustment coefficient, is the basic abnormality parameter, is the observation abnormality parameter.

[0121] In the embodiment disclosed by the application, a smelting plant personnel behavior abnormality monitoring system is also disclosed, comprising:

[0122] The first module is configured to acquire a plan design drawing of the smelting plant, construct a three-dimensional analysis model of the smelting plant based on the plan design drawing, and set abnormality sensitivity coefficients for different position space points of the three-dimensional analysis model of the smelting plant based on the work contents of different work areas of the smelting plant.

[0123] The second module is configured to acquire human body features and behavior features of personnel in different work areas of the smelting plant, configure preset virtual human body models based on the human body features and behavior features, map the corresponding virtual human body models in the three-dimensional analysis model of the smelting plant, perform initial behavior analysis on the performance of the virtual human body models, and determine the basic abnormality parameters.

[0124] The third module is configured to construct a plurality of personnel behavior identification rules, each of which includes a plurality of behavior description keywords, drive the virtual human body models to perform simulated performance according to a preset control mode, mark and configure a plurality of description keywords on key behavior segments in the simulated performance, if the combination of the description keywords in a continuous time period meets the personnel behavior identification rules, identify the personnel behavior of the corresponding simulated performance, and construct a simulated simulation recognition library by the simulated performance and the personnel behavior in a corresponding manner.

[0125] The fourth module is configured to perform deep behavior analysis on the virtual human body models in the three-dimensional analysis model of the smelting plant by using the simulated simulation recognition library, and determine the observation abnormality parameters.

[0126] The fifth module is configured to determine the abnormal degree of the virtual human body model in combination with the basic abnormality parameter, the observation abnormality parameter and the corresponding abnormality sensitivity coefficient of the virtual human body model, and perform alarm based on the abnormal degree.

[0127] The application provides a smelting plant personnel behavior anomaly monitoring method and system, relates to the smelting plant monitoring technical field, constructs a three-dimensional analysis model, sets an abnormal sensitive coefficient in combination with a work area risk type, provides a quantitative basis for anomaly detection, configures a virtual human body model, captures personnel behavior trajectories and actions by using visual analysis technology, determines basic abnormal parameters, utilizes a simulation simulation recognition library to deeply analyze real-time behaviors of the virtual human body model, dynamically determines observation abnormal parameters, comprehensively calculates abnormal degrees based on the basic abnormal parameters, the observation abnormal parameters and the abnormal sensitive coefficient, and realizes accurate alarm.

[0128] Through the description of the above implementation manners, those skilled in the art can clearly understand that the application can be implemented by hardware, or by means of software and a necessary general hardware platform. Based on such understanding, the technical solutions of the application can be embodied in the form of a software product, which can be stored in a nonvolatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation scenarios of the application.

[0129] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the application rather than limit them, and although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: it can still modify or equivalently replace the technical solutions of the application, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the application.

Claims

1. A method for monitoring abnormal behavior of personnel in a smelting plant, characterized in that, include: Obtain the floor plan of the smelting plant and construct a three-dimensional analysis model of the smelting plant based on the floor plan. Based on the work content of different work areas of the smelting plant, set anomaly sensitivity coefficients for different spatial points in the three-dimensional analysis model of the smelting plant. The human body characteristics and behavioral characteristics of personnel in different work areas of the smelting plant are obtained, and a preset virtual human body model is configured based on the human body characteristics and behavioral characteristics. The corresponding virtual human body model is mapped into the 3D analysis model of the smelting plant, and the performance of the virtual human body model is analyzed to determine the basic abnormal parameters. Several rules for identifying human behavior are constructed. Each rule includes several behavioral description keywords. The virtual human model is driven to perform simulation performance according to a preset control method. Key behavioral segments in the simulation performance are extracted and marked, and several descriptive keywords are configured. If the combination of descriptive keywords satisfies the rules for identifying human behavior within a continuous time period, the corresponding simulation performance is identified as human behavior. The simulation performance and the human behavior identification are matched in a corresponding way to construct a simulation recognition library. Using a simulation recognition library, we conducted in-depth behavioral analysis on the virtual human body model in the 3D analysis model of the smelting plant to determine the parameters of observed anomalies. By combining the basic abnormal parameters, observed abnormal parameters, and corresponding abnormality sensitivity coefficients of the virtual human body model, the degree of abnormality of the virtual human body model is determined, and an alarm is triggered based on the degree of abnormality. Methods for configuring pre-defined virtual human body models based on human body features and behavioral characteristics include: Visual analysis technology was used to locate factory personnel in surveillance videos, and to determine their identities, clothing, behavioral patterns, and body movements. A virtual human body model is constructed, including a trunk expression line, an upper limb expression line, and a lower limb expression line. The upper limb expression line and the lower limb expression line are movably connected to the side and bottom of the trunk expression line, respectively. An upper limb swing mapping sphere is set at the connection between the upper limb expression line and the trunk expression line, and a lower limb swing mapping sphere is set at the connection between the lower limb expression line and the trunk expression line. If the identified factory personnel’s identity and clothing meet the preset standards, then based on the determined behavioral trajectory, the positional changes of the virtual human body model in the 3D analysis model of the smelting plant are determined, and based on the determined body movements, the swing changes of the upper limb expression lines and lower limb expression lines relative to the trunk expression lines are determined. Methods for initial behavioral analysis of virtual human models include: The historical accident data and simulated accident data of the smelting plant are analyzed to determine the equipment working status, the marked behavior trajectory and marked body movement of the factory workers at the time of each accident. The marked behavior trajectory and marked body movement are analyzed to determine whether there are any abnormalities. If there are abnormalities, the corresponding marked behavior trajectory is identified as an abnormal behavior trajectory and the corresponding marked body movement is recorded as an abnormal body movement. The abnormal behavioral trajectories and abnormal body movements are divided into key abnormal segments to determine the abnormal behavioral trajectory comparison group and the abnormal body movement comparison group. The methods for dividing the key abnormal segments include: The analysis time node is determined by taking the time of the accident as the end analysis time node and advancing the end analysis time node forward by a preset time period. The start analysis time node is then determined, and the time period between the start analysis time node and the end analysis time node is recorded as the abnormal analysis time segment. Several anomaly analysis time nodes are evenly set for the anomaly analysis time period, and the equipment working status corresponding to each anomaly analysis time node is determined. Based on the equipment working status, the standard behavior trajectory set and standard body movement set of factory workers are determined for each anomaly analysis time node. The marked behavior trajectory at the corresponding anomaly analysis time node is compared with the standard behavior trajectory set. If no matching standard behavior trajectory is found in the standard behavior trajectory set, the corresponding marked behavior trajectory is identified as an abnormal behavior trajectory. The marked body movement at the corresponding anomaly analysis time node is compared with the standard body movement set. If no corresponding standard body movement is found in the standard body movement set, the corresponding marked body movement is identified as an abnormal body movement. Time relationship analysis is performed on different abnormal behavior trajectories and abnormal body movements in the anomaly analysis time period to determine the order, time interval and corresponding equipment status of different abnormal behavior trajectories, and combine them into an abnormal behavior trajectory comparison group. Similarly, the order, time interval and corresponding equipment status of different abnormal body movements are determined and combined into an abnormal body movement comparison group. Several abnormal behavior trajectory comparison groups are combined to obtain an abnormal behavior trajectory comparison set, and several abnormal body movement comparison groups are combined to obtain an abnormal body movement comparison set. The virtual human body model is compared using the abnormal behavior trajectory comparison set and the abnormal body movement comparison set, respectively. The first consistency parameter between the virtual human body model and different abnormal behavior trajectory comparison groups, and the second consistency parameter between the virtual human body model and different abnormal body movement comparison groups are calculated. The average value of the first consistency parameter and the average value of the second consistency parameter are calculated to obtain the first average consistency parameter and the second average consistency parameter. Based on the first average consistent parameter and the second average consistent parameter, the basic abnormal parameters are determined; Methods for conducting in-depth behavioral analysis of virtual human models in a 3D analysis model of a smelting plant using a simulation recognition library include: According to the preset time length, the current behavior characteristics of the virtual human body model are extracted to obtain the current behavior characteristic performance segment. Then, the current behavior fragments in the current behavior characteristic performance segment are identified using the simulation recognition library. Based on the recognition results, each current behavior fragment is marked with descriptive keywords. Analyze the proportion of current behavior segments marked with descriptive keywords to the total current behavior segments, and determine the observed abnormal parameters based on the preset proportion range to which the segment proportion belongs.

2. The method for monitoring abnormal behavior of personnel in a smelting plant according to claim 1, characterized in that, Based on the work content of different work areas in a smelting plant, methods for setting anomaly sensitivity coefficients for different spatial points in the 3D analysis model of the smelting plant include: Based on the location of different work areas, the 3D analysis model of the smelting plant is divided into 3D work blocks, including 3D smelting work blocks, 3D casting work blocks, 3D raw material storage work blocks, and 3D equipment maintenance blocks. Based on the risk type label corresponding to the marker of each three-dimensional working block, the risk performance of each risk type label is analyzed. Based on the risk performance, the risk parameters of different spatial points in the three-dimensional equipment working block are determined, and based on the risk parameters, the anomaly sensitivity coefficient of the corresponding spatial points is determined.

3. The method for monitoring abnormal behavior of personnel in a smelting plant according to claim 2, characterized in that, Risk type labels include: For the three-dimensional melting work area, there are risks of high temperature and molten metal splashing. For the three-dimensional casting work area, there are risks of mechanical injury and high temperature. For the three-dimensional raw material storage work area, there are risks of fire, explosion, and chemical leakage. For the 3D equipment maintenance area, there are risks of mechanical injury and electric shock.

4. The method for monitoring abnormal behavior of personnel in a smelting plant according to claim 2, characterized in that, Methods for analyzing the risk performance of each risk type label include: Obtain historical accident data of the smelting plant, determine the risk probability and severity parameters of different spatial points, and calculate the product of the risk probability and the severity parameters of the risk consequences to obtain the risk parameters. For different risk parameters, corresponding anomaly sensitivity coefficients are set.

5. The method for monitoring abnormal behavior of personnel in a smelting plant according to claim 1, characterized in that, Methods for calculating fundamental anomaly parameters include: Determine the current working state of the device corresponding to the virtual human body model, and based on the equivalence of the working state of the device, filter the abnormal behavior trajectory comparison group and the abnormal body movement comparison group, and compare the filtered abnormal behavior trajectory and abnormal body movement comparison group with the performance of the virtual human body model to determine the trajectory comparison result and the body movement comparison result. In the trajectory comparison results, the first number of identical abnormal behavior trajectories is calculated to the first number of identical abnormal behavior trajectories in the abnormal behavior trajectory comparison group. Based on the first number of identical abnormal behavior trajectories, the first matching parameter is determined. In the body movement comparison results, the second number of identical abnormal body movements is calculated to the second number of identical abnormal body movements in the abnormal body movement comparison group. Based on the first equal number ratio, the first matching parameter is determined; based on the second equal number ratio, the second matching parameter is determined; the first average matching parameter and the second average matching parameter are calculated; and the basic abnormal parameters are determined. The expression for calculating the basic anomaly parameters is as follows: ； in, Based on the anomaly parameter, This is the weight adjustment coefficient for comparing abnormal behavior trajectories. The weighting adjustment coefficient for abnormal body movements. The ratio of the first identical counts corresponding to the i1th abnormal behavior trajectory comparison group. The adjustment factor is the first equivalent number of influences. The first equivalent influence adjustment constant, This represents the total number of groups selected for comparison of abnormal behavior trajectories. The second equivalent count ratio corresponding to the i2th abnormal body movement comparison group. The adjustment factor is the second equivalent number of influences. The second equivalent influence adjustment constant, This represents the total number of abnormal body movement comparison groups selected.

6. The method for monitoring abnormal behavior of personnel in a smelting plant according to claim 1, characterized in that, Combining the basic anomaly parameters, observed anomaly parameters, and corresponding anomaly sensitivity coefficients of the virtual human body model, the expression for determining the degree of anomaly in the virtual human body model is as follows: ； in, To indicate the degree of abnormality, The abnormal sensitivity coefficient, The adjustment coefficient for the weight of the basic abnormal parameter. To observe the adjustment coefficients for outlier parameters, Based on the anomaly parameter, To observe abnormal parameters.

7. A monitoring system for abnormal personnel behavior in a smelting plant, characterized in that, A method for monitoring abnormal behavior of personnel in a smelting plant according to any one of claims 1-6, comprising: The first module is used to obtain the plan design drawings of the smelting plant and construct a three-dimensional analysis model of the smelting plant based on the plan design drawings. Based on the work content of different work areas of the smelting plant, anomaly sensitivity coefficients are set for different spatial points in the three-dimensional analysis model of the smelting plant. The second module is used to acquire the human body characteristics and behavioral characteristics of personnel in different work areas of the smelting plant, configure a preset virtual human body model based on the human body characteristics and behavioral characteristics, map the corresponding virtual human body model into the smelting plant 3D analysis model, and perform initial behavioral analysis on the performance of the virtual human body model to determine the basic abnormal parameters. The third module is used to construct several rules for identifying human behavior. Each rule includes several behavioral description keywords. The virtual human model is driven to perform simulation performance according to a preset control method. Key behavioral segments in the simulation performance are extracted and marked, and several descriptive keywords are configured. If the combination of descriptive keywords satisfies the rules for identifying human behavior within a continuous time period, the corresponding simulation performance is identified as human behavior. The simulation performance and the human behavior identification are matched in a corresponding way to construct a simulation recognition library. The fourth module is used to perform in-depth behavioral analysis on the virtual human body model in the 3D analysis model of the smelting plant using a simulation recognition library, and to determine the observed abnormal parameters. The fifth module is used to determine the degree of abnormality of the virtual human body model by combining the basic abnormal parameters, observed abnormal parameters and corresponding abnormality sensitivity coefficients, and to issue an alarm based on the degree of abnormality.

Images