An industrial park dangerous behavior identification monitoring system based on a 5G private network

Abstract

The application relates to the technical field of image processing, in particular to an industrial park dangerous behavior identification monitoring system based on a 5G private network. The system comprises a data acquisition module, which is used for acquiring multi-source data and aligning and constructing associated data clusters; an abnormality detection module, which is used for identifying data abnormal points and generating abnormal events; a heterogeneous data review module, which is used for cross-verification of the abnormal events, determining whether the abnormal events are normal fluctuations or generating to-be-diagnosed abnormalities; an external factor exclusion module, which is used for layer-by-layer attribution analysis of external interference and output of diagnosis results or to-be-audited identification; an inspection behavior auditing module, which is used for verification of inspection data authenticity and output of auditing results; and a risk determination module, which is used for triggering device abnormality alarms or human-made false alarm. Through the joint means of multi-source data alignment, heterogeneous review, external attribution and behavior auditing, the application realizes accurate tracing and true-false identification of inspection data abnormalities, reduces the false alarm rate and improves the reliability of dangerous behavior identification.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a dangerous behavior identification and monitoring system for industrial parks based on a 5G private network. Background Technology

[0002] With increasingly stringent safety requirements in industrial parks, hazardous behavior identification and monitoring systems have become crucial for ensuring park safety. Traditional industrial park monitoring systems primarily rely on fixed cameras for video capture and manual monitoring. However, these systems suffer from several problems in practical applications: First, industrial parks have widely distributed equipment and numerous inspection points, leading to blind spots for fixed cameras and hindering comprehensive real-time monitoring. Second, when inspection data is abnormal, traditional systems can only trigger alarms without in-depth analysis, resulting in frequent false alarms caused by communication interference, poor equipment condition, or environmental factors. This severely strains the time and energy of safety management personnel and reduces the efficiency of responding to real risks. Third, existing systems lack methods to verify the authenticity of inspection behavior itself, making it difficult to effectively distinguish between genuine equipment malfunctions, location fraud, or human data tampering caused by inspection robots not following preset routes.

[0003] Chinese Patent Publication No. CN115834844A discloses an intelligent early warning system for unsafe behaviors in industrial sites based on multi-source video surveillance. The algorithm module and the safety behavior control module are connected via a 5G network communication module. The algorithm module includes an environmental modeling algorithm module and a behavior recognition algorithm module. The safety behavior control module includes a hardware device module and a video monitoring module. The 5G network communication module includes a split transmission module and a MEC network architecture module. This invention achieves real-time monitoring by quickly and accurately modeling and analyzing behavioral data based on algorithms, reducing the workload of monitoring personnel and minimizing blind spots. The system's alarm function provides intelligent alerts to personnel exhibiting unsafe behaviors, serving as a warning and reminder function, strengthening safety supervision. The system's storage and facial recognition functions quickly identify specific personnel, facilitating timely control and reducing the difficulty of supervision and management.

[0004] Therefore, the existing technology has the following problems: The system does not cover multi-source verification of abnormal inspection data, elimination of external interference, and auditing of the authenticity of inspection activities. It cannot distinguish the true cause of data anomalies, resulting in a high false alarm rate and difficulty in distinguishing between genuine equipment anomalies and human fraud. Summary of the Invention

[0005] To address this, the present invention provides an industrial park hazardous behavior identification and monitoring system based on a 5G private network, which overcomes the problems of existing technologies that do not involve multi-source verification of inspection data anomalies, elimination of external interference, and auditing of the authenticity of inspection behavior, making it impossible to distinguish the real cause of data anomalies, resulting in high false alarm rates and difficulty in distinguishing between genuine equipment anomalies and human fraud.

[0006] To achieve the above objectives, this invention provides a hazardous behavior identification and monitoring system for industrial parks based on a 5G private network, comprising: The data acquisition module is used to acquire multi-source data and, based on the multi-source data, align it according to a unified spatiotemporal reference to construct a related data cluster; The anomaly detection module is used to identify inspection points with data anomalies based on the inspection data in the associated data cluster and the baseline data in the distributed control system, and to generate anomaly events. The heterogeneous data verification module is used to respond to the abnormal event by calling multi-source data in the associated data cluster to cross-validate the abnormal event, and to determine whether the data fluctuation is normal or generate an abnormal event to be diagnosed based on the cross-validation result. The external factor exclusion module is used to respond to the abnormal event to be diagnosed, extract the external interference parameters at the corresponding time in the associated data cluster, perform layer-by-layer attribution analysis on the external interference parameters, and output the external factor diagnosis result or audit mark based on the analysis results. The inspection behavior audit module is used to respond to the auditable identifier, verify the inspection data according to the positioning trajectory of the inspection robot, and output the audit result; The risk assessment module is used to determine whether the alarm is triggered by abnormal equipment or by human-induced fraud based on the audit results.

[0007] Furthermore, the data acquisition module includes: A multi-source data acquisition unit is used to acquire multi-source data including the inspection data, the video images, the environmental parameters, and the location information; the inspection data is collected by an explosion-proof inspection robot according to a preset inspection route and inspection frequency, the video images are collected by a fixed monitoring camera, the environmental parameters are collected by an environmental sensor, and the location information is collected by a personnel positioning terminal. A spatiotemporal alignment unit is used to spatiotemporally align the multi-source data according to a unified timestamp and spatial coordinate system. The data association unit is used to establish an association relationship between the spatiotemporally aligned multi-source data according to the inspection point and the collection time, and generate the associated data cluster.

[0008] Furthermore, the anomaly detection module includes: The deviation calculation unit is used to perform difference calculation between the inspection data and the reference data of the corresponding inspection point and the corresponding acquisition time in the distributed control system to obtain the deviation value. An abnormal event generation unit is used to determine that there is a data abnormality at the corresponding inspection point when the deviation value is greater than a preset deviation threshold, and to generate an abnormal event containing the inspection point identifier, the collection time and the deviation value.

[0009] Furthermore, the heterogeneous data verification module includes: The adjacent point verification unit is used to extract the adjacent point inspection data of the inspection point corresponding to the abnormal event, compare the adjacent point inspection data with the benchmark data for trend consistency, and generate trend consistency comparison results. The video verification unit is used to extract the video image corresponding to the time of the abnormal event, compare the video image with the inspection data for reading consistency, and generate reading consistency comparison results. The verification result determination unit is used to determine that the data fluctuates normally and terminate the processing when the trend consistency comparison result indicates that the trend is consistent and the reading consistency comparison result indicates that the readings are consistent; when the trend consistency comparison result indicates that the trend is inconsistent or the reading consistency comparison result indicates that the readings are inconsistent, an abnormal event to be diagnosed is generated.

[0010] Furthermore, the external factor elimination module includes: The communication link diagnostic unit is used to extract the communication quality parameters at the time corresponding to the abnormal event, and output the communication link abnormality diagnostic result when the communication quality parameters are abnormal. The device status diagnosis unit is used to extract the device perception status parameters at the time corresponding to the abnormal event when the communication link diagnosis unit does not output a diagnosis result, and to output the device perception abnormality diagnosis result when the device perception status parameters are abnormal. An environmental disturbance diagnosis unit is used to extract the environmental disturbance parameters at the time corresponding to the abnormal event when neither the communication link diagnosis unit nor the device status diagnosis unit outputs a diagnosis result, and to output the environmental disturbance diagnosis result when the environmental disturbance parameters are abnormal. The attribution result output unit is used to output an auditable identifier when the communication link diagnosis unit, the device status diagnosis unit, and the environmental disturbance diagnosis unit have not outputted any diagnosis results.

[0011] Furthermore, the inspection behavior audit module includes: The trajectory consistency audit unit is used to spatially match the robot's positioning trajectory with the preset inspection route, calculate the trajectory deviation distance, and output an audit result indicating that the inspection data is not true when the trajectory deviation distance is greater than the preset trajectory deviation threshold. The historical data auditing unit is used to statistically compare the inspection data with historical inspection data under the same working conditions when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold, calculate the statistical deviation value, and output an audit result indicating that the inspection data is not true when the statistical deviation value is greater than the preset statistical deviation threshold. The authenticity determination unit is used to output an audit result that indicates the authenticity of the inspection data when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold and the statistical deviation value is less than or equal to a preset statistical deviation threshold.

[0012] Furthermore, the risk assessment module includes: The fraud alarm unit is used to trigger a fraud alarm when the audit results indicate that the inspection data is not authentic. The equipment alarm unit is used to trigger an equipment anomaly alarm when the audit results indicate that the inspection data is true.

[0013] Furthermore, the communication link diagnostic unit, the device status diagnostic unit, and the environmental disturbance diagnostic unit perform diagnostics sequentially in the order of communication link layer, device perception layer, and environmental coupling layer. When the diagnostic unit of the previous layer determines an abnormality, the subsequent diagnostics are terminated and the corresponding diagnostic results are output.

[0014] Furthermore, the trajectory consistency auditing unit aligns the robot positioning trajectory with the preset inspection route according to the acquisition time, calculates the spatial Euclidean distance point by point, and takes the maximum spatial Euclidean distance of each inspection point as the trajectory deviation distance.

[0015] Furthermore, the historical data auditing unit obtains a historical inspection data sample set of the inspection points corresponding to the inspection data under the same working conditions and the same collection period, calculates the mean and standard deviation of the historical inspection data sample set, and uses the ratio of the absolute value of the difference between the mean and the inspection data to the standard deviation as the statistical deviation value.

[0016] Compared with existing technologies, the beneficial effects of this invention are as follows: This embodiment constructs a related data cluster through spatiotemporal alignment of multi-source data, providing a unified data foundation for subsequent anomaly detection and verification; the anomaly detection module enables rapid location and event-based analysis of data anomalies; the heterogeneous data verification module calls inspection data and video images from adjacent points for cross-verification, effectively eliminating false alarms caused by normal data fluctuations; the external factor elimination module performs layer-by-layer attribution analysis of communication links, equipment status, and environmental disturbances, achieving precise location of anomaly causes and avoiding false alarms caused by communication interference, equipment failure, or environmental changes; the inspection behavior audit module verifies the robot's positioning trajectory, effectively identifying the authenticity of inspection data and distinguishing between genuine equipment anomalies and human-induced falsification; and the risk judgment module enables differentiated alarms, improving the intelligence level of industrial park security monitoring and the accuracy of risk response.

[0017] Furthermore, this embodiment uses a deviation calculation unit to perform difference calculations between the inspection data and the baseline data of the distributed control system to obtain the deviation value, thereby realizing quantitative monitoring of the operating status of equipment at each inspection point in the industrial park. By setting the preset deviation threshold in the transition range between the normal operating deviation distribution and the abnormal deviation distribution, both detection sensitivity and processing efficiency are taken into account, avoiding misjudging normal fluctuations as abnormalities and preventing the missed detection of real abnormalities. When the deviation value exceeds the threshold, the abnormal event generation unit generates a standardized abnormal event containing the location, time and degree of deviation of the abnormality, providing a unified event triggering basis for subsequent verification of heterogeneous data and elimination of external factors.

[0018] Furthermore, this embodiment extracts the time-series data sequences of abnormal points and their adjacent points through the adjacent point verification unit, and compares the slope signs and numerical change directions of the two, realizing rapid judgment of whether the deviation of abnormal points belongs to the overall operation trend change of the region, avoiding the misjudgment of overall data fluctuations caused by changes in operating conditions as equipment abnormalities; the video verification unit extracts the target meter readings at the location of the inspection robot from the video image and compares them with the inspection data for numerical consistency, realizing independent verification of the authenticity of the inspection data collection and transmission process, effectively identifying false abnormalities caused by transmission errors or data forgery; the verification result judgment unit makes a comprehensive judgment based on the comparison results of trend consistency and reading consistency. When both comparison results are consistent, the data is judged as normal fluctuation and the processing is terminated, avoiding unnecessary subsequent diagnostic processes. When either comparison result is inconsistent, an abnormal event to be diagnosed is generated, accurately screening out abnormalities that cannot be eliminated through verification and entering the next stage, realizing the effective diversion of abnormal events and the rational allocation of diagnostic resources. Furthermore, this embodiment utilizes a communication link diagnostic unit to output communication link anomaly diagnostic results when communication quality parameters are abnormal, enabling rapid attribution of data anomalies caused by abnormal 5G private network signals or packet loss rates. A device status diagnostic unit jointly assesses positioning accuracy, gimbal stability, and image clarity when the communication link is normal, achieving accurate identification of data distortion caused by poor device perception. An environmental disturbance diagnostic unit assesses vibration amplitude and temperature / humidity data when both the first two are normal, effectively identifying measurement deviations caused by on-site environmental interference. A cause-attribution mechanism, where the three diagnostic units execute sequentially layer by layer, terminating subsequent diagnoses upon the output of the previous layer's diagnostic result, enables rapid location of anomaly causes and efficient allocation of diagnostic resources. By setting each preset value within a transitional range between normal and abnormal distributions, the detection sensitivity and risk of misjudgment of each diagnostic unit are considered. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the hazardous behavior identification and monitoring system for industrial parks based on a 5G private network, as described in this embodiment. Figure 2 This is a schematic diagram of the data acquisition module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network in this embodiment; Figure 3 This is a schematic diagram of the anomaly detection module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network, as described in this embodiment. Figure 4 This is a flowchart of the heterogeneous data verification module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network in this embodiment. Detailed Implementation

[0020] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0021] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0022] Please see Figure 1 As shown, it is a schematic diagram of the industrial park dangerous behavior identification and monitoring system based on 5G private network in this embodiment.

[0023] This embodiment provides a hazardous behavior identification and monitoring system for industrial parks based on a 5G private network, including: The data acquisition module is used to acquire multi-source data and, based on the multi-source data, align it according to a unified spatiotemporal reference to construct a related data cluster; The anomaly detection module is used to identify inspection points with data anomalies based on the inspection data in the associated data cluster and the baseline data in the distributed control system, and to generate anomaly events. The heterogeneous data verification module is used to respond to the abnormal event by calling multi-source data in the associated data cluster to cross-validate the abnormal event, and to determine whether the data fluctuation is normal or generate an abnormal event to be diagnosed based on the cross-validation result. The external factor exclusion module is used to respond to the abnormal event to be diagnosed, extract the external interference parameters at the corresponding time in the associated data cluster, perform layer-by-layer attribution analysis on the external interference parameters, and output the external factor diagnosis result or audit mark based on the analysis results. The inspection behavior audit module is used to respond to the auditable identifier, verify the inspection data according to the positioning trajectory of the inspection robot, and output the audit result; The risk assessment module is used to determine whether the alarm is triggered by abnormal equipment or by human-induced fraud based on the audit results.

[0024] This embodiment constructs a related data cluster through spatiotemporal alignment of multi-source data, providing a unified data foundation for subsequent anomaly detection and verification. The anomaly detection module enables rapid location and event-based analysis of data anomalies. The heterogeneous data verification module cross-verifies data from adjacent inspection points and video footage, effectively eliminating false alarms caused by normal data fluctuations. The external factor elimination module performs layer-by-layer attribution analysis of communication links, equipment status, and environmental disturbances, achieving precise location of anomaly causes and avoiding false alarms due to communication interference, equipment failure, or environmental changes. The inspection behavior audit module verifies the robot's positioning trajectory, effectively identifying the authenticity of inspection data and distinguishing between genuine equipment anomalies and human-induced falsification. The risk assessment module implements differentiated alarms, improving the intelligence level of industrial park security monitoring and the accuracy of risk response.

[0025] Please see Figure 2 As shown, it is a schematic diagram of the data acquisition module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network in this embodiment.

[0026] Specifically, the data acquisition module includes: A multi-source data acquisition unit is used to acquire multi-source data including the inspection data, the video images, the environmental parameters, and the location information; the inspection data is collected by an explosion-proof inspection robot according to a preset inspection route and inspection frequency, the video images are collected by a fixed monitoring camera, the environmental parameters are collected by an environmental sensor, and the location information is collected by a personnel positioning terminal. A spatiotemporal alignment unit is used to spatiotemporally align the multi-source data according to a unified timestamp and spatial coordinate system. The data association unit is used to establish an association relationship between the spatiotemporally aligned multi-source data according to the inspection point and the collection time, and generate the associated data cluster.

[0027] In this embodiment, the multi-source data acquisition unit collects inspection data through an explosion-proof inspection robot according to a preset inspection route and frequency. That is, the robot automatically inspects equipment and facilities and reads meter readings according to a predetermined path and cycle. It also collects video images through fixed monitoring camera equipment, that is, the monitoring camera captures images of the work area in real time. It collects environmental parameters through environmental sensors, that is, the temperature and humidity sensors and vibration sensors monitor the environmental conditions of the work site. Finally, it collects location information through personnel positioning terminals, that is, the real-time spatial coordinates of the workers are obtained through UWB or Beidou positioning equipment.

[0028] The spatiotemporal alignment unit aligns the multi-source data, which comes from different sources and has varying formats, using a unified timestamp and spatial coordinate system. This synchronizes and matches data from different sources at the same time and spatial location, eliminating time discrepancies and coordinate differences between the various acquisition devices. The data association unit then establishes associations between the spatiotemporally aligned multi-source data and the acquisition time. Using the inspection point as an index, it binds inspection data, video images, environmental parameters, and location information from the same acquisition time, generating associated data clusters. Please see Figure 3 As shown, it is a schematic diagram of the anomaly detection module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network in this embodiment.

[0029] Specifically, the anomaly detection module includes: The deviation calculation unit is used to perform difference calculation between the inspection data and the reference data of the corresponding inspection point and the corresponding acquisition time in the distributed control system to obtain the deviation value. An abnormal event generation unit is used to determine that there is a data abnormality at the corresponding inspection point when the deviation value is greater than a preset deviation threshold, and to generate an abnormal event containing the inspection point identifier, the collection time and the deviation value.

[0030] In this embodiment, the deviation calculation unit is used to perform a difference calculation between the inspection data and the baseline data at the corresponding inspection point and the corresponding acquisition time in the distributed control system to obtain the deviation value. The distributed control system is a process control system for an industrial park, used to collect and control the operating parameters of production equipment in real time. Its baseline data is the standard operating value of the equipment collected by the distributed control system at the same time and the same inspection point, which represents the normal state parameters of the equipment under this working condition. The inspection data is the measured reading collected by the explosion-proof inspection robot at the inspection point. The deviation value is used to characterize the direction and degree of deviation between the measured reading and the standard value.

[0031] In this embodiment, the preset deviation threshold is a critical value for determining whether the data is abnormal. This preset deviation threshold is determined based on the historical statistical distribution characteristics of the deviation values ​​at the inspection point under normal operating conditions: historical deviation value samples are collected from the inspection point over a continuous preset time period under normal operating conditions; statistical analysis is performed on the historical deviation value samples to obtain the distribution range and fluctuation range of the deviation values ​​under normal operating conditions; and the preset deviation threshold is set in the transition range between the normal operating deviation distribution and the abnormal deviation distribution. If the preset deviation threshold is set too low, normal data fluctuations are easily misjudged as data anomalies, increasing the processing burden of heterogeneous source verification and attribution analysis; if the preset deviation threshold is set too high, true abnormal data is difficult to identify in a timely manner, reducing the sensitivity of anomaly detection.

[0032] In this embodiment, the abnormal event generation unit compares the deviation value with a preset deviation threshold. When the deviation value is greater than the preset deviation threshold, it determines that there is a data anomaly at the inspection point and generates an abnormal event containing the inspection point identifier, the collection time, and the deviation value. The abnormal event is a structured data object containing information in three dimensions: the location of the anomaly, the time of the anomaly, and the degree of deviation. It is used to provide a standardized event triggering basis for the verification of heterogeneous data and the elimination of external factors.

[0033] This embodiment uses a deviation calculation unit to perform difference calculations between inspection data and distributed control system baseline data to obtain deviation values, thereby achieving quantitative monitoring of the operating status of equipment at each inspection point in the industrial park. By setting a preset deviation threshold in the transition range between normal operating deviation distribution and abnormal deviation distribution, it balances detection sensitivity and processing efficiency, avoiding misjudging normal fluctuations as abnormalities and preventing the missed detection of real anomalies. When the deviation value exceeds the threshold, the abnormal event generation unit generates a standardized abnormal event containing the location, time, and degree of deviation of the abnormality, providing a unified event triggering basis for subsequent verification of heterogeneous data and elimination of external factors.

[0034] Please see Figure 4As shown, this is a flowchart of the heterogeneous data verification module in the industrial park hazardous behavior identification and monitoring system based on a 5G private network in this embodiment.

[0035] Specifically, the heterogeneous data verification module includes: The adjacent point verification unit is used to extract the adjacent point inspection data of the inspection point corresponding to the abnormal event, compare the adjacent point inspection data with the benchmark data for trend consistency, and generate trend consistency comparison results. The video verification unit is used to extract the video image corresponding to the time of the abnormal event, compare the video image with the inspection data for reading consistency, and generate reading consistency comparison results. The verification result determination unit is used to determine that the data fluctuates normally and terminate the processing when the trend consistency comparison result indicates that the trend is consistent and the reading consistency comparison result indicates that the readings are consistent; when the trend consistency comparison result indicates that the trend is inconsistent or the reading consistency comparison result indicates that the readings are inconsistent, an abnormal event to be diagnosed is generated.

[0036] In this embodiment, the adjacent point verification unit extracts the inspection data of adjacent points corresponding to the abnormal event, compares the adjacent point inspection data with the baseline data for trend consistency, and generates a trend consistency comparison result. Adjacent point inspection data refers to the inspection data of at least one inspection point spatially adjacent to the abnormal inspection point on the same process pipeline or equipment link. The trend consistency comparison is performed as follows: taking the current abnormal acquisition time as the center, a preset number of sampling points are taken forward and backward to form the time-series data sequences of the abnormal point and its adjacent points; the slope signs of the abnormal point's time-series data sequence and the adjacent point's time-series data sequence are calculated respectively. When the slope signs are the same and the corresponding sampling point values ​​change in the same direction, the trend consistency comparison result is determined to indicate trend consistency; when the slope signs are different or the corresponding sampling point values ​​change in opposite directions, the trend consistency comparison result is determined to indicate trend inconsistency.

[0037] The video verification unit extracts the video footage corresponding to the moment of the abnormal event, compares the video footage with the inspection data for consistency, and generates a consistency comparison result. The video footage is the view of the inspection robot's location captured by fixed monitoring cameras at the moment of abnormal data collection. The reading consistency comparison is performed as follows: The position of the inspection robot is detected from the video image using image recognition technology to determine the target meter area corresponding to the robot at that moment; the meter reading is extracted from the target meter area. This extraction process specifically includes: image segmentation of the target meter area to extract the meter dial area; detection of scale lines and pointers on the dial area; and conversion of the reading based on the pointer's position on the dial and the scale markings; or, digit detection on the dial area, and recognition of the displayed digits using an optical character recognition algorithm to obtain the reading; the meter reading extracted from the video image is used as the video reading, and this video reading is compared with the inspection data returned by the inspection robot; when the difference between the video reading and the inspection data is less than a preset reading difference threshold, the reading consistency comparison result indicates that the readings are consistent; when the difference between the video reading and the inspection data is greater than or equal to the preset reading difference threshold, the reading consistency comparison result indicates that the readings are inconsistent.

[0038] The verification result determination unit receives the trend consistency comparison result and the reading consistency comparison result. When the trend consistency comparison result indicates that the trend is consistent and the reading consistency comparison result indicates that the reading is consistent, it is determined that the data fluctuates normally and the processing is terminated. When the trend consistency comparison result indicates that the trend is inconsistent or the reading consistency comparison result indicates that the reading is inconsistent, an abnormal event to be diagnosed is generated.

[0039] This embodiment extracts the time-series data sequences of abnormal points and their adjacent points through an adjacent point verification unit, and compares the slope signs and numerical change directions of the two, enabling rapid determination of whether the deviation of abnormal points belongs to the overall operational trend change of the region, avoiding misjudgment of overall data fluctuations caused by changes in operating conditions as equipment anomalies. The video verification unit extracts the target meter readings at the location of the inspection robot from video images and compares them with the inspection data for numerical consistency, achieving independent verification of the authenticity of the inspection data collection and transmission process, effectively identifying false anomalies caused by transmission errors or data forgery. The verification result judgment unit makes a comprehensive judgment based on the comparison results of trend consistency and reading consistency. When both comparison results are consistent, the data is judged as normal fluctuation and the processing is terminated, avoiding unnecessary subsequent diagnostic procedures. When either comparison result is inconsistent, a diagnostic anomaly event is generated, accurately filtering out anomalies that cannot be eliminated through verification and proceeding to the next stage, achieving effective diversion of anomalies and reasonable allocation of diagnostic resources. Specifically, the external factor elimination module includes: The communication link diagnostic unit is used to extract the communication quality parameters at the time corresponding to the abnormal event, and output the communication link abnormality diagnostic result when the communication quality parameters are abnormal. The device status diagnosis unit is used to extract the device perception status parameters at the time corresponding to the abnormal event when the communication link diagnosis unit does not output a diagnosis result, and to output the device perception abnormality diagnosis result when the device perception status parameters are abnormal. An environmental disturbance diagnosis unit is used to extract the environmental disturbance parameters at the time corresponding to the abnormal event when neither the communication link diagnosis unit nor the device status diagnosis unit outputs a diagnosis result, and to output the environmental disturbance diagnosis result when the environmental disturbance parameters are abnormal. The attribution result output unit is used to output an auditable identifier when the communication link diagnosis unit, the device status diagnosis unit, and the environmental disturbance diagnosis unit have not outputted any diagnosis results.

[0040] In this embodiment, the communication link diagnostic unit extracts the communication quality parameters corresponding to the abnormal event at that moment, and outputs the communication link anomaly diagnostic result when the communication quality parameters are abnormal. The communication quality parameters include the 5G private network signal strength and data packet loss rate at the time when the explosion-proof inspection robot transmits inspection data. The 5G private network signal strength is characterized by the reference signal received power value reported in real time by the 5G communication module of the inspection robot; the data packet loss rate is characterized by the percentage of data packets lost by the inspection robot during data transmission relative to the total number of data packets sent. The preset signal strength threshold is determined based on the historical statistical distribution characteristics of the signal strength in the area where the inspection point is located under normal communication conditions. Signal strength samples are collected from the area where the inspection point is located for a consecutive preset number of days during normal working hours. Statistical analysis is performed on the signal strength samples to obtain the distribution range and fluctuation range of the signal strength under normal communication conditions. The preset signal strength threshold is set in the transition range between the normal signal strength distribution and the abnormal signal strength distribution.

[0041] When the preset signal strength threshold is set too high, normal signal fluctuations are easily misjudged as communication anomalies, increasing unnecessary diagnostic processing burdens. When the preset signal strength threshold is set too low, genuine communication link anomalies are difficult to identify in a timely manner, leading to data anomalies that cannot be correctly attributed. The preset packet loss rate threshold is determined based on the historical statistical distribution characteristics of packet loss rates in the area where the inspection point is located under normal communication conditions. Packet loss rate samples are collected from the area where the inspection point is located over a preset number of consecutive days during normal working hours. Statistical analysis of the packet loss rate samples is performed to obtain the distribution range and fluctuation range of packet loss rates under normal communication conditions. The preset packet loss rate threshold is set in the transition range between the normal packet loss rate distribution and the abnormal packet loss rate distribution. When the preset packet loss rate threshold is set too low, normal occasional packet loss is easily misjudged as communication anomalies. When the preset packet loss rate threshold is set too high, genuine communication link anomalies are difficult to identify in a timely manner. When the 5G private network signal strength is lower than the preset signal strength threshold or the data packet loss rate is higher than the preset packet loss threshold, communication quality parameters are abnormal, and a communication link anomaly diagnosis result is output.

[0042] The equipment status diagnostic unit is activated when the communication link diagnostic unit does not output diagnostic results. It extracts the equipment perception status parameters corresponding to the abnormal event at that moment, and outputs the equipment perception anomaly diagnostic result when the equipment perception status parameters are abnormal. The equipment perception status parameters include the positioning accuracy, gimbal stability, and image clarity of the inspection robot. Positioning accuracy is characterized by the deviation between the positioning coordinates output in real time by the positioning module on the inspection robot and the coordinates of the preset inspection route. Gimbal stability is characterized by the offset of the pitch and roll angles relative to the reference angles output in real time by the gimbal attitude sensor of the inspection robot. Image clarity is obtained by evaluating the clarity of the video frame images returned by the inspection robot. A clarity evaluation algorithm based on the gradient energy function is used to calculate the proportion of high-frequency components in the video frame images to the total frequency components for quantitative characterization.

[0043] The preset positioning accuracy range is determined based on the historical statistical distribution characteristics of the positioning signal coverage quality in the area where the inspection point is located. Positioning deviation samples are collected from the area under normal operating conditions for a continuous preset duration. Statistical analysis of these samples yields the distribution range and fluctuation range of positioning deviations under normal operating conditions. The upper and lower limits of the preset positioning accuracy range are set within the transition range between normal and abnormal positioning deviation distributions. Similarly, the preset stable angle range is determined based on the historical statistical distribution characteristics of the gimbal attitude angles of the inspection robot under normal operating conditions. Gimbal pitch and roll angle samples are collected from the inspection robot under normal operating conditions for a continuous preset duration. The distribution range and fluctuation range of attitude angles under normal operating conditions are obtained. The upper and lower limits of the preset stable angle range are set within the transition range between normal and abnormal attitude angle distributions.

[0044] The preset sharpness threshold is determined based on the historical statistical distribution characteristics of image sharpness of the inspection robot under normal working conditions. Video frame sharpness samples are collected over a continuous preset time period during normal working conditions to obtain the distribution range and fluctuation range of sharpness under normal working conditions. The preset sharpness threshold is set within the transition range between normal and abnormal sharpness distributions. When any of the following conditions are met: positioning accuracy deviates from the preset positioning accuracy range, gimbal stability exceeds the preset stability angle range, or image sharpness falls below the preset sharpness threshold, the device's perception status parameters are abnormal, and a device perception anomaly diagnosis result is output.

[0045] The environmental disturbance diagnosis unit is activated when neither the communication link diagnosis unit nor the equipment status diagnosis unit outputs diagnostic results. It extracts the environmental disturbance parameters corresponding to the time of the abnormal event and outputs the environmental disturbance diagnosis result when the environmental disturbance parameters are abnormal. The environmental disturbance parameters include the vibration amplitude and temperature and humidity data at the time of the abnormal event. The vibration amplitude is characterized by the vibration acceleration amplitude collected by vibration sensors installed on the inspection robot or near the inspection point; the temperature and humidity data are characterized by the temperature and humidity values ​​collected by environmental temperature and humidity sensors installed near the inspection point.

[0046] The preset vibration threshold is determined based on the historical statistical distribution characteristics of the vibration amplitude in the area where the inspection point is located under normal working conditions. Vibration amplitude samples are collected in the area where the inspection point is located under normal working conditions for a continuous preset time period. The vibration amplitude samples are statistically analyzed to obtain the distribution range and fluctuation range of the vibration amplitude under normal working conditions. The preset vibration threshold is set in the transition range between normal vibration distribution and abnormal vibration distribution.

[0047] The preset operating temperature and humidity range is determined based on the historical statistical distribution characteristics of the allowable ambient temperature and humidity range for normal operation of the inspection equipment in the area where the inspection point is located. Temperature and humidity samples are collected continuously for a preset duration under normal operating conditions in the area where the inspection point is located. Statistical analysis of the temperature and humidity samples is performed to obtain the distribution range and fluctuation range of temperature and humidity under normal operating conditions. The upper and lower limits of the preset operating temperature and humidity range are set in the transition zone between normal and abnormal temperature and humidity distributions. When the vibration amplitude exceeds the preset vibration threshold or the temperature and humidity data exceed the preset operating temperature and humidity range, the environmental disturbance parameters are abnormal, and the environmental disturbance diagnosis result is output.

[0048] When the communication link diagnosis unit, equipment status diagnosis unit, and environmental disturbance diagnosis unit fail to output diagnosis results, the attribution result output unit outputs an audit pending flag. This audit pending flag is used to trigger the start of the inspection behavior audit module to further determine whether the inspection data anomaly is caused by human factors.

[0049] This embodiment uses a communication link diagnostic unit to output communication link anomaly diagnostic results when communication quality parameters are abnormal, enabling rapid attribution of data anomalies caused by abnormal 5G private network signals or packet loss rates. A device status diagnostic unit, when the communication link is normal, jointly assesses positioning accuracy, gimbal stability, and image clarity, enabling accurate identification of data distortion caused by poor device perception. An environmental disturbance diagnostic unit assesses vibration amplitude and temperature / humidity data when the first two are normal, effectively identifying measurement deviations caused by on-site environmental interference. The attribution mechanism, where the three diagnostic units execute sequentially layer by layer, with the output of the previous layer terminating subsequent diagnoses, enables rapid location of anomaly causes and efficient allocation of diagnostic resources. By setting each preset value within a transitional range between normal and abnormal distributions, the detection sensitivity and risk of misjudgment of each diagnostic unit are balanced.

[0050] Specifically, the inspection behavior audit module includes: The trajectory consistency audit unit is used to spatially match the robot's positioning trajectory with the preset inspection route, calculate the trajectory deviation distance, and output an audit result indicating that the inspection data is not true when the trajectory deviation distance is greater than the preset trajectory deviation threshold. The historical data auditing unit is used to statistically compare the inspection data with historical inspection data under the same working conditions when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold, calculate the statistical deviation value, and output an audit result indicating that the inspection data is not true when the statistical deviation value is greater than the preset statistical deviation threshold. The authenticity determination unit is used to output an audit result that indicates the authenticity of the inspection data when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold and the statistical deviation value is less than or equal to a preset statistical deviation threshold.

[0051] In this embodiment, the trajectory consistency audit unit spatially matches the robot's positioning trajectory with the preset inspection route, calculates the trajectory deviation distance, and outputs an audit result indicating that the inspection data is inaccurate when the trajectory deviation distance exceeds the preset trajectory deviation threshold. The robot's positioning trajectory is represented by a series of positioning coordinate points continuously reported by the positioning module on the inspection robot at a preset frequency during the execution of the inspection task, reflecting the complete movement path of the inspection robot in the actual inspection process. The preset inspection route is the theoretical travel path that the inspection robot should follow when performing the inspection task at that inspection point.

[0052] Spatial matching is performed as follows: Each positioning coordinate point in the robot's positioning trajectory is aligned with the timestamp of the acquisition time and the preset inspection route. The spatial Euclidean distance between each positioning coordinate point and its corresponding reference coordinate point on the preset inspection route is calculated point by point. The maximum spatial Euclidean distance calculated for each inspection point is used as the trajectory deviation distance. The preset trajectory deviation threshold is determined based on the historical statistical distribution characteristics of the positioning trajectory deviation in the area where the inspection point is located under normal inspection conditions. Positioning trajectory deviation samples are collected from the area where the inspection point is located during a preset number of consecutive inspection tasks under normal inspection conditions. Statistical analysis is performed on the positioning trajectory deviation samples to obtain the distribution range and fluctuation range of trajectory deviation under normal inspection conditions. The preset trajectory deviation threshold is set in the transition range between the normal trajectory deviation distribution and the abnormal trajectory deviation distribution.

[0053] When the preset trajectory deviation threshold is set too small, normal positioning errors are easily misjudged as trajectory anomalies; when the preset trajectory deviation threshold is set too large, genuine positioning fraud is difficult to identify in a timely manner. When the trajectory deviation distance is greater than the preset trajectory deviation threshold, it indicates that the actual movement path of the inspection robot deviates significantly from the preset inspection route, suggesting that the inspection robot may not be performing its inspection task according to the prescribed route or that the positioning signal may have been forged. Therefore, an audit result indicating that the inspection data is inaccurate is output.

[0054] The historical data audit unit is activated when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold. It statistically compares the inspection data with historical inspection data under the same operating conditions, calculates the statistical deviation value, and outputs an audit result indicating that the inspection data is inaccurate when the statistical deviation value exceeds the preset statistical deviation threshold. Historical inspection data refers to the sample set of inspection data for the inspection point under the same operating conditions and the same data collection period, under normal historical conditions. The statistical comparison is performed as follows: The historical inspection data sample set for the inspection point under the same operating conditions and the same data collection period is obtained; the mean and standard deviation of the historical inspection data sample set are calculated; and the ratio of the absolute value of the difference between the mean and the current inspection data to the standard deviation is used as the statistical deviation value.

[0055] The preset statistical deviation threshold is determined based on the statistical distribution characteristics of historical inspection data for the inspection point under the same working conditions. Historical inspection data samples are collected from the inspection point for a preset number of consecutive times under the same working conditions. The statistical deviation value of each sample relative to the historical mean is calculated to obtain the distribution range and fluctuation range of the statistical deviation value under normal working conditions. The preset statistical deviation threshold is set in the transition range between the normal and abnormal statistical deviation distributions. If the preset statistical deviation threshold is set too small, normal data fluctuations may be easily misjudged as data tampering; if the preset statistical deviation threshold is set too large, it is difficult to identify actual data tampering in a timely manner. When the statistical deviation value is greater than the preset statistical deviation threshold, it indicates that the inspection data significantly deviates from the historical normal level under the same working conditions, and there is a possibility that the inspection data has been tampered with. Therefore, an audit result indicating that the inspection data is not authentic is output.

[0056] When the trajectory deviation distance is less than or equal to the preset trajectory deviation threshold and the statistical deviation value is less than or equal to the preset statistical deviation threshold, the authenticity determination unit outputs an audit result that represents the authenticity of the inspection data. This indicates that the movement trajectory of the inspection robot conforms to the specifications and the inspection data is consistent with the historical normal level. This anomaly belongs to the real equipment anomaly, and the risk determination module is triggered to alarm the equipment anomaly.

[0057] This embodiment uses a trajectory consistency audit unit and a historical data audit unit to sequentially perform spatial matching between the robot's positioning trajectory and the preset route, and to statistically compare the inspection data with historical data under the same working conditions. When there are anomalies at the trajectory or data level, the inspection data is determined to be untrue. When both verifications pass, the data is determined to be true. By setting various thresholds in the transition range between normal and abnormal distributions, the detection sensitivity and the risk of misjudgment are taken into account, realizing comprehensive authenticity verification from behavioral trajectory to data content, and providing a reliable basis for risk assessment.

[0058] Specifically, the risk assessment module includes: The fraud alarm unit is used to trigger a fraud alarm when the audit results indicate that the inspection data is not authentic. The equipment alarm unit is used to trigger an equipment anomaly alarm when the audit results indicate that the inspection data is true.

[0059] In this embodiment, the fraud alarm unit receives the audit results output by the inspection behavior audit module. When the audit results indicate that the inspection data is not authentic, a fraud alarm is triggered. This alarm is used to notify the park's safety management personnel that there is a suspected tampering of inspection data or that the inspection task has not been performed as required, so that manual verification and accountability procedures can be initiated in a timely manner. The equipment alarm unit receives the audit results output by the inspection behavior audit module. When the audit results indicate that the inspection data is authentic, an equipment anomaly alarm is triggered. This alarm is used to notify the park's equipment maintenance personnel that there is a genuine equipment malfunction at the inspection point, so that repair and maintenance work can be arranged in a timely manner.

[0060] Specifically, the communication link diagnostic unit, the device status diagnostic unit, and the environmental disturbance diagnostic unit perform diagnostics in the order of communication link layer, device perception layer, and environmental coupling layer. When the diagnostic unit of the previous layer determines an abnormality, the subsequent diagnostics are terminated and the corresponding diagnostic results are output.

[0061] In this embodiment, the communication link diagnostic unit, the device status diagnostic unit, and the environmental disturbance diagnostic unit perform diagnostics sequentially according to the communication link layer, the device perception layer, and the environmental coupling layer. Since the inspection data needs to go through three key stages from collection to transmission, namely, the inspection robot's own perception and collection, the on-site environment measurement, and the 5G private network wireless transmission, any abnormality in any stage may lead to the distortion of the inspection data. Therefore, the inspection is carried out layer by layer according to the physical link order in which the data is generated.

[0062] When the communication link diagnostic unit determines that the communication quality parameters are abnormal, it indicates that the cause of the abnormality has been located in the wireless transmission link. The unit directly outputs the communication link abnormality diagnostic result and terminates subsequent diagnostics, ceasing the diagnostics of the device perception layer and the environment coupling layer to avoid invalid calculations. When the communication link diagnostic unit does not output a diagnostic result, but the device status diagnostic unit determines that the device perception status parameters are abnormal, it indicates that the cause of the abnormality has been located in the inspection robot's own perception device link. The unit directly outputs the device perception abnormality diagnostic result and terminates subsequent diagnostics, ceasing the diagnostics of the environment coupling layer. When neither the communication link diagnostic unit nor the device status diagnostic unit outputs a diagnostic result, but the environmental disturbance diagnostic unit determines that the environmental disturbance parameters are abnormal, it indicates that the cause of the abnormality has been located in the on-site environmental interference link, and the environmental disturbance diagnostic result is output. When none of the three diagnostic units output a diagnostic result, it indicates that the cause of the abnormality cannot be attributed to any external factor in the communication link, the device's own status, or the on-site environment. The attribution result output unit outputs an audit pending flag, and the analysis is transferred to the behavior authenticity audit link.

[0063] Specifically, the trajectory consistency auditing unit aligns the robot positioning trajectory with the preset inspection route according to the collection time, calculates the spatial Euclidean distance point by point, and takes the maximum spatial Euclidean distance of each inspection point as the trajectory deviation distance.

[0064] In this embodiment, the trajectory consistency auditing unit aligns the robot positioning trajectory with the preset inspection route according to the acquisition time, calculates the spatial Euclidean distance point by point, and takes the maximum spatial Euclidean distance of each inspection point as the trajectory deviation distance. The robot positioning trajectory consists of multiple positioning coordinate points continuously reported by the inspection robot at a preset sampling frequency during the execution of the inspection task. Each positioning coordinate point has a timestamp of the acquisition time. The preset inspection route is composed of multiple preset trajectory points connected sequentially according to the inspection order. Each preset trajectory point has a timestamp of the corresponding theoretical arrival time. The timestamp alignment is performed as follows: each positioning coordinate point in the robot positioning trajectory is paired one by one with the preset trajectory point in the preset inspection route with the corresponding theoretical arrival time, ensuring that the measured position and the theoretical position at the same time correspond. The spatial Euclidean distance is calculated as follows: for each pair of registered positioning coordinate points and preset trajectory points, the Euclidean distance value between them in the spatial coordinate system is calculated. The maximum spatial Euclidean distance calculated at each inspection point is taken as the trajectory deviation distance. This maximum value reflects the maximum degree of deviation of the inspection robot from the preset inspection route during the entire inspection task execution process.

[0065] Specifically, the historical data auditing unit obtains a historical inspection data sample set of the inspection points corresponding to the inspection data under the same working conditions and the same collection period, calculates the mean and standard deviation of the historical inspection data sample set, and uses the ratio of the absolute value of the difference between the mean and the inspection data to the standard deviation as the statistical deviation value.

[0066] In this embodiment, the historical data auditing unit acquires a sample set of historical inspection data for the inspection points corresponding to the inspection data under the same operating conditions and the same collection period. It calculates the mean and standard deviation of this sample set, using the ratio of the absolute value of the difference between the mean and the inspection data to the standard deviation as the statistical deviation value. The historical inspection data sample set originates from multiple sets of inspection data recorded at the inspection point under the same operating conditions and collection period during historical normal operation. This data is used to construct a data distribution model for the inspection point under normal operating conditions. The mean characterizes the central trend of the historical data for the inspection point under the same operating conditions, reflecting the typical numerical level under normal operating conditions; the standard deviation characterizes the dispersion of the historical data for the inspection point under the same operating conditions, reflecting the fluctuation range of the data under normal operating conditions.

[0067] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A hazardous behavior identification and monitoring system for industrial parks based on a 5G private network, characterized in that, include: The data acquisition module is used to acquire multi-source data and, based on the multi-source data, align it according to a unified spatiotemporal reference to construct a related data cluster; The anomaly detection module is used to identify inspection points with data anomalies based on the inspection data in the associated data cluster and the baseline data in the distributed control system, and to generate anomaly events. The heterogeneous data verification module is used to respond to the abnormal event by calling multi-source data in the associated data cluster to cross-validate the abnormal event, and to determine whether the data fluctuation is normal or generate an abnormal event to be diagnosed based on the cross-validation result. The external factor exclusion module is used to respond to the abnormal event to be diagnosed, extract the external interference parameters at the corresponding time in the associated data cluster, perform layer-by-layer attribution analysis on the external interference parameters, and output the external factor diagnosis result or audit mark based on the analysis results. The inspection behavior audit module is used to respond to the auditable identifier, verify the inspection data according to the positioning trajectory of the inspection robot, and output the audit result; The risk assessment module is used to determine whether the alarm is triggered by abnormal equipment or by human-induced fraud based on the audit results.

2. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The data acquisition module includes: A multi-source data acquisition unit is used to acquire multi-source data including the inspection data, the video images, the environmental parameters, and the location information; the inspection data is collected by an explosion-proof inspection robot according to a preset inspection route and inspection frequency, the video images are collected by a fixed monitoring camera, the environmental parameters are collected by an environmental sensor, and the location information is collected by a personnel positioning terminal. A spatiotemporal alignment unit is used to spatiotemporally align the multi-source data according to a unified timestamp and spatial coordinate system. The data association unit is used to establish an association relationship between the spatiotemporally aligned multi-source data according to the inspection point and the collection time, and generate the associated data cluster.

3. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The anomaly detection module includes: The deviation calculation unit is used to perform difference calculation between the inspection data and the reference data of the corresponding inspection point and the corresponding acquisition time in the distributed control system to obtain the deviation value. An abnormal event generation unit is used to determine that there is a data abnormality at the corresponding inspection point when the deviation value is greater than a preset deviation threshold, and to generate an abnormal event containing the inspection point identifier, the collection time and the deviation value.

4. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The heterogeneous data verification module includes: The adjacent point verification unit is used to extract the adjacent point inspection data of the inspection point corresponding to the abnormal event, compare the adjacent point inspection data with the benchmark data for trend consistency, and generate trend consistency comparison results. The video verification unit is used to extract the video image corresponding to the time of the abnormal event, compare the video image with the inspection data for reading consistency, and generate reading consistency comparison results. The verification result determination unit is used to determine that the data fluctuates normally and terminate the processing when the trend consistency comparison result indicates that the trend is consistent and the reading consistency comparison result indicates that the readings are consistent; when the trend consistency comparison result indicates that the trend is inconsistent or the reading consistency comparison result indicates that the readings are inconsistent, an abnormal event to be diagnosed is generated.

5. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The external factor elimination module includes: The communication link diagnostic unit is used to extract the communication quality parameters at the time corresponding to the abnormal event, and output the communication link abnormality diagnostic result when the communication quality parameters are abnormal. The device status diagnosis unit is used to extract the device perception status parameters at the time corresponding to the abnormal event when the communication link diagnosis unit does not output a diagnosis result, and to output the device perception abnormality diagnosis result when the device perception status parameters are abnormal. An environmental disturbance diagnosis unit is used to extract the environmental disturbance parameters at the time corresponding to the abnormal event when neither the communication link diagnosis unit nor the device status diagnosis unit outputs a diagnosis result, and to output the environmental disturbance diagnosis result when the environmental disturbance parameters are abnormal. The attribution result output unit is used to output an auditable identifier when the communication link diagnosis unit, the device status diagnosis unit, and the environmental disturbance diagnosis unit have not outputted any diagnosis results.

6. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The inspection behavior audit module includes: The trajectory consistency audit unit is used to spatially match the robot's positioning trajectory with the preset inspection route, calculate the trajectory deviation distance, and output an audit result indicating that the inspection data is not true when the trajectory deviation distance is greater than the preset trajectory deviation threshold. The historical data auditing unit is used to statistically compare the inspection data with historical inspection data under the same working conditions when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold, calculate the statistical deviation value, and output an audit result indicating that the inspection data is not true when the statistical deviation value is greater than the preset statistical deviation threshold. The authenticity determination unit is used to output an audit result that indicates the authenticity of the inspection data when the trajectory deviation distance is less than or equal to a preset trajectory deviation threshold and the statistical deviation value is less than or equal to a preset statistical deviation threshold.

7. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 1, characterized in that, The risk assessment module includes: The fraud alarm unit is used to trigger a fraud alarm when the audit results indicate that the inspection data is not authentic. The equipment alarm unit is used to trigger an equipment anomaly alarm when the audit results indicate that the inspection data is true.

8. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 5, characterized in that, The communication link diagnostic unit, device status diagnostic unit, and environmental disturbance diagnostic unit perform diagnostics sequentially in the order of communication link layer, device perception layer, and environmental coupling layer. When the diagnostic unit of the previous layer determines an abnormality, the subsequent diagnostics are terminated and the corresponding diagnostic results are output.

9. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 6, characterized in that, The trajectory consistency auditing unit aligns the robot positioning trajectory with the preset inspection route according to the acquisition time, calculates the spatial Euclidean distance point by point, and takes the maximum spatial Euclidean distance of each inspection point as the trajectory deviation distance.

10. The industrial park hazardous behavior identification and monitoring system based on a 5G private network according to claim 6, characterized in that, The historical data auditing unit obtains a historical inspection data sample set corresponding to the inspection points under the same working conditions and the same collection period, calculates the mean and standard deviation of the historical inspection data sample set, and uses the ratio of the absolute value of the difference between the mean and the inspection data to the standard deviation as the statistical deviation value.

Images