AI-based petrochemical pipeline leakage steam dynamic identification method and system

Abstract

The invention provides an AI-based petrochemical pipeline leakage steam dynamic identification method and system. The method comprises the following steps: firstly, acquiring multispectral video data; secondly, acquiring operation parameters of the petrochemical pipeline; thermodynamic visualization processing is carried out on the multispectral video data, and transient steam plumes are separated out; then constructing a spatial-temporal variation template; using an AI video analysis model to match the temporal and spatial change template with the steam leakage behavior pedigree to obtain a matching result; then introducing a real-time thermal hydraulic state as an environmental constraint condition, and dynamically correcting a matching result; and finally, when the corrected matching result exceeds the confidence boundary of the corresponding leakage form in the behavior pedigree, generating a pipeline leakage identification signal. According to the technical scheme provided by the invention, the accuracy and early warning capability of leakage detection are improved, and high-precision identification and reliable early warning of tiny leakage of the petrochemical pipeline in a complex industrial environment are also realized.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence video analysis technology, and in particular to an AI-based method and system for dynamic identification of leaking steam in petrochemical pipelines. Background Technology

[0002] In the operation of petrochemical pipelines, early identification of steam leaks under high temperature and high pressure environments is crucial for safe production. These leaks often start weakly, have varied diffusion patterns, and are often located in complex industrial environments with numerous sources of thermal interference and visual noise. This necessitates that the monitoring system possess the ability to keenly capture the characteristics of weak leaks, resist interference from complex environments, and fuse and analyze multi-physics data.

[0003] Existing technical solutions mainly employ video analysis systems based on fixed thresholds. These systems deploy visible light or infrared cameras to collect video data of the pipeline area, use background subtraction algorithms to identify abnormal moving targets, and combine this with simple temperature or grayscale thresholds to determine whether a leak has occurred. This system relies on preset static parameters and triggers an alarm by comparing the difference between the current frame and the background model.

[0004] However, this scheme has significant drawbacks. Its static threshold is difficult to adapt to the dynamic changes in leakage patterns caused by changes in pipeline pressure and temperature, and it is prone to false alarms in rainy, foggy weather or under changing lighting conditions. At the same time, it lacks correlation analysis between steam diffusion characteristics and pipeline operating parameters, and cannot distinguish between real leaks and normal steam emissions. For minor initial leaks, due to the lack of multi-feature fusion analysis and dynamic learning capabilities, it often results in missed alarms or delayed alarms, and cannot meet the needs of early and accurate early warning. Summary of the Invention

[0005] This application provides an AI-based method and system for dynamic identification of leaking steam in petrochemical pipelines, which addresses the problems of high false alarm rate, high risk of missed alarm, and insensitivity to minor initial leaks caused by the lack of dynamic fusion of multi-source data and environmental adaptability in existing static threshold monitoring schemes.

[0006] Firstly, this application provides an AI-based method for dynamic identification of leaking steam in petrochemical pipelines, including: Acquire multispectral video data simultaneously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines; Simultaneously, the operating parameters of the petrochemical pipeline are acquired in real time interactively, including the pressure and temperature of the medium inside the pipeline; The multispectral video data is subjected to thermodynamic visualization processing to separate the transient steam plume driven by the micro-positive pressure at the leak point in the complex background of the petrochemical pipeline; The dynamic texture of the transient steam plume is coupled with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak. The spatiotemporal variation template is matched with a preset steam leakage behavior spectrum using a pre-optimized AI video analysis model that has undergone transfer learning, and the matching results are obtained. During the matching process, the real-time thermal-hydraulic state of the petrochemical pipeline is introduced as an environmental constraint condition to dynamically correct the matching results and obtain the corrected matching results. When the corrected matching result exceeds the confidence boundary of the corresponding leakage pattern in the behavior spectrum, a pipeline leakage identification signal is generated.

[0007] Optionally, the operating parameters of the petrochemical pipeline can be acquired simultaneously and interactively in real time, wherein the operating parameters include the pressure and temperature of the medium inside the pipeline, including: Pressure and temperature readings are acquired synchronously by pressure and temperature sensing units deployed on the pipe surface. A dynamic filtering mechanism for the pressure and temperature readings is established, and pressure and temperature readings that match the video frame timestamps are selected based on the acquisition timing of the multispectral video data. The selected pressure and temperature readings are evaluated for consistency, and instantaneous disturbance data caused by pipeline vibration are removed to obtain the evaluated pressure and temperature readings. The determined pressure and temperature readings are integrated into a set of operating parameters with time synchronization characteristics.

[0008] Optionally, the multispectral video data is subjected to thermodynamic visualization processing to separate the transient steam plume driven by the micro-positive pressure at the leak point from the complex background of the petrochemical pipeline, including: The RGB video stream acquired by the visible light sensor and the thermal radiation video stream acquired by the infrared imaging sensor are fused at the pixel level in the multispectral video data to generate a fused video stream. A background model based on pixel thermodynamic properties is established in the fused video stream; Extract pixel regions from the fused video stream of the current frame that have thermodynamic properties different from the background model; Motion continuity detection is performed on the pixel region to identify a coherent set of pixels that have continuous motion characteristics and thermodynamic properties that conform to the vapor volatilization characteristics. The motion trajectories of the coherent pixel set in consecutive video frames are integrated to obtain a transient steam plume.

[0009] Optionally, the dynamic texture of the transient steam plume is coupled with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template reflecting the specificity of the leak, including: Extract the motion trajectory of the transient steam plume in consecutive video frames to form a trajectory point sequence; Analyze the changing patterns of the motion direction of the trajectory point sequence to establish a directional distribution map of steam diffusion; The directional distribution map is correlated with the real-time pressure readings, and the weight distribution of each direction in the directional distribution map is adjusted according to the pressure value. Simultaneously, the motion speed change of the trajectory point sequence is correlated with the real-time temperature reading, and the sensitivity parameter of the motion speed change is adjusted according to the temperature value; By combining the adjusted directional weight allocation and the adjusted velocity sensitivity parameters, a spatiotemporal variation template with pressure and temperature adaptive characteristics is obtained.

[0010] Optionally, the spatiotemporal variation template is matched with a pre-set steam leakage behavior spectrum using an AI video analysis model optimized through transfer learning to obtain matching results, including: Multi-scale structural analysis was performed on the spatiotemporal variation template to extract the feature elements and structural relationships of the feature elements related to the vapor diffusion morphology in the spatiotemporal variation template. Access the storage of a steam leakage behavior spectrum with multiple reference modes, each of which contains standard steam diffusion morphology characteristics corresponding to different leakage orifice diameters under specific operating parameters; The feature elements and their structural relationships are matched step by step with each reference pattern in the steam leakage behavior spectrum. First, the overall morphology is matched, and then the local feature is matched. During the matching process, the consistency index of each matching stage is recorded. Based on the consistency index, the comprehensive matching degree between the spatiotemporal change template and each reference mode is calculated to obtain a set of matching metric values; Select the reference pattern with the highest matching metric value from a set of matching metrics as the matching result.

[0011] Optionally, during the matching process, the real-time thermal-hydraulic state of the petrochemical pipeline is introduced as an environmental constraint to dynamically correct the matching results, resulting in corrected matching results, including: Based on the pressure and temperature readings in the operating parameters, calculate the medium density and flow velocity parameters under the operating conditions of the petrochemical pipeline; Establish a corrected relationship table for the influence of the medium density and flow velocity parameters on the vapor diffusion morphology; Based on the current medium density and flow velocity parameters, the corresponding morphological correction coefficient is obtained by querying the correction relationship table. The morphological correction coefficient is applied to the matching metric obtained during the matching process to dynamically adjust the matching degree of each reference pattern; The reference pattern with the highest matching metric value after dynamic adjustment is selected as the corrected matching result.

[0012] Optionally, when the corrected matching result exceeds the confidence boundary of the corresponding leakage pattern in the behavioral spectrum, a pipeline leakage identification signal is generated, including: Confidence boundary parameters of the reference mode corresponding to the corrected matching result are obtained from the steam leakage behavior spectrum; The matching metric in the corrected matching result is compared with the confidence boundary parameter; When the matching metric value continuously exceeds the confidence boundary parameter for a preset time length, the leakage confirmation mechanism is triggered; In the leakage confirmation mechanism, the persistence and stability characteristics of the matching metric exceeding the confidence boundary parameter are detected; When the persistence and stability characteristics simultaneously meet preset conditions, a pipeline leak identification signal containing the leak location and leak level is generated.

[0013] Secondly, this application provides an AI-based dynamic identification system for leaking steam in petrochemical pipelines, comprising: The first acquisition module is used to acquire multispectral video data synchronously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines. The second acquisition module is used to simultaneously and interactively acquire the operating parameters of the petrochemical pipeline in real time, wherein the operating parameters include the pressure and temperature of the medium inside the pipeline; The processing module is used to perform thermodynamic visualization processing on the multispectral video data in order to separate the transient steam plume driven by the micro-positive pressure at the leak point in the complex background of the petrochemical pipeline. The coupling module is used to couple the dynamic texture of the transient steam plume with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak. The matching module is used to match the spatiotemporal change template with a preset steam leakage behavior spectrum using a pre-optimized AI video analysis model that has undergone transfer learning, and obtain the matching result. The correction module is used to introduce the real-time thermal-hydraulic state of the petrochemical pipeline as an environmental constraint during the matching process, and to dynamically correct the matching result to obtain the corrected matching result. The generation module is used to generate a pipeline leak identification signal when the corrected matching result exceeds the confidence boundary of the corresponding leak pattern in the behavior spectrum.

[0014] Thirdly, this application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are to be invoked and executed by the processing component to implement an AI-based dynamic identification method for leaking steam in petrochemical pipelines as described in the first aspect above.

[0015] Fourthly, this application provides a computer storage medium storing a computer program, which, when executed by a computer, implements an AI-based dynamic identification method for leaking steam in petrochemical pipelines as described in the first aspect.

[0016] This application constructs a spatiotemporal variation template that accurately reflects the specificity of leaks by deeply fusing multispectral video data with pipeline operating parameters, and uses an AI video analysis model optimized through transfer learning for intelligent matching. This method effectively overcomes the technical challenges of variable background interference and the difficulty in capturing weak leak characteristics in complex industrial environments, achieving highly sensitive identification of micro-positive pressure driven steam leaks, and significantly improving the accuracy of leak detection and early warning capabilities.

[0017] Furthermore, by designing a leakage confirmation mechanism with dual criteria of time persistence and state stability, the system ensures that an alarm is only triggered when the matching result continuously exceeds the confidence boundary and exhibits stable abnormal characteristics. This mechanism effectively avoids false alarms caused by transient environmental interference or system noise, significantly improving the reliability of the alarm signal. At the same time, the output leakage location and level information provides clear guidance for subsequent emergency handling.

[0018] These or other aspects of this application will become more apparent from the description of the following embodiments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of an AI-based dynamic identification method for leaking steam in petrochemical pipelines, provided in this application, is shown. Figure 2 A schematic diagram of the structure of an AI-based dynamic identification system for leaking steam in petrochemical pipelines provided in this application is shown. Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present application, the technical solution of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0022] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0023] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Figure 1 This application provides a flowchart of an AI-based dynamic identification method for leaking steam in petrochemical pipelines, as shown in the following figure. Figure 1 As shown, the method includes: Step 101: Acquire multispectral video data synchronously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines.

[0025] In the above scheme, the high-risk area of ​​petrochemical pipelines refers to the critical parts of the pipeline system that are prone to leakage or failure, including stress concentration or corrosion-prone areas such as welded joints, valve connections, and bends; a visible light sensor is a device that can capture optical images within the visible spectrum of the human eye, used to obtain visual appearance information of the pipeline surface and its surrounding environment; an infrared imaging sensor is a device that generates thermal distribution images by detecting infrared radiation emitted by objects, used to capture the temperature distribution characteristics of the pipeline surface; multispectral video data is a composite data stream that is simultaneously acquired by the visible light sensor and the infrared imaging sensor, containing both visual appearance information and thermodynamic characteristic information of the pipeline area.

[0026] In this scheme, firstly, visible light sensors and infrared imaging sensors deployed in high-risk areas of the pipeline are synchronously triggered to ensure that the two sensors begin data acquisition at the same timestamp. Secondly, a precision time synchronization module performs frame-level alignment processing on the raw video streams acquired by the two sensors, so that each frame of the visible light image has a strict correspondence with the infrared thermal image at the same time. Next, a multi-source data fusion module integrates the aligned visible light video stream and infrared video stream into a unified multispectral video data stream, in which each pixel contains both RGB color information and thermal radiation intensity value. Finally, a data encapsulation module adds a timestamp and spatial location tag to each synchronized frame, forming a multispectral video data packet with spatiotemporal correlation characteristics.

[0027] For example, in the monitoring system of a distillation unit in a petrochemical plant, a visible light camera (model A) and an infrared thermal imager (model B) are deployed at the weld seams of high-temperature, high-pressure pipeline sections. When the system starts monitoring, both sensors simultaneously begin acquiring data via GPS synchronization signals. The visible light camera captures steam condensation on the pipeline surface, while the thermal imager records the temperature distribution in the corresponding area. The data acquisition system transmits both video streams to the processing center at a rate of 30 frames per second, and uses a timestamp alignment algorithm to ensure that each frame of the visible light image is precisely matched with the corresponding infrared image at that moment, ultimately generating multispectral video data containing both visual and thermodynamic information.

[0028] This solution uses synchronous acquisition and multi-source data fusion technology to obtain multispectral video data that simultaneously contains visual and thermodynamic features, providing a rich information foundation for subsequent leak identification. It effectively solves the problem of insufficient perception capability of a single sensor in complex industrial environments and significantly improves the ability of subsequent processing stages to perceive subtle leak features.

[0029] Step 102: Simultaneously, the operating parameters of the petrochemical pipeline are acquired interactively in real time, including the pressure and temperature of the medium inside the pipeline.

[0030] Optionally, step 102 may specifically include the following steps: Step 1021: Pressure and temperature readings are simultaneously acquired by pressure sensing units and temperature sensing units deployed on the pipe surface; Step 1022: Establish a dynamic filtering mechanism for the pressure readings and temperature readings, and select pressure readings and temperature readings that match the video frame timestamps based on the acquisition time sequence of the multispectral video data. Step 1023: Perform state consistency judgment on the selected pressure and temperature readings, remove instantaneous disturbance data caused by pipeline vibration, and obtain the judged pressure and temperature readings. Step 1024: Integrate the determined pressure and temperature readings into a set of operating parameters with time synchronization characteristics.

[0031] In the above scheme, the operating parameters of the petrochemical pipeline refer to the set of physical quantity parameters that reflect the working state of the pipeline during operation; the pressure and temperature of the medium inside the pipeline refer to the pressure intensity and heat level of the flowing substance inside the pipeline, respectively; the pressure sensing unit is a measuring device installed on the outer wall of the pipeline to detect changes in the internal medium pressure; the temperature sensing unit is a measuring device fixed on the surface of the pipeline to detect changes in the medium temperature; the pressure reading and temperature reading are the raw pressure and temperature values ​​collected by the sensors in real time; the dynamic filtering mechanism is a processing rule that automatically filters data according to specific conditions; the video frame timestamp is the time information marked for each frame of the image in the video acquisition system; the state consistency judgment is the process of judging whether the data conforms to the normal operating state; the instantaneous disturbance data is the short-term abnormal data caused by external interference; the judged pressure reading and temperature reading are the data that have been verified and confirmed to be valid; the operating parameters with time synchronization characteristics are the set of parameters that have the same time reference as the video data.

[0032] In this scheme, firstly, step 1021 involves synchronously acquiring pressure and temperature readings at a fixed frequency using pressure and temperature sensing units deployed on the pipe surface to ensure time consistency of data acquisition. Secondly, step 1022 uses a dynamic filtering mechanism to select pressure and temperature readings that perfectly match the video frame timestamps from the continuously acquired data streams of the pressure and temperature sensors, based on the acquisition sequence of the multispectral video data. Next, step 1023 performs a state consistency judgment on the selected pressure and temperature readings, identifying and eliminating instantaneous disturbance data caused by pipe vibration or other interference by analyzing the data's changing trends and fluctuation characteristics, thus obtaining the judged pressure and temperature readings. Finally, step 1024 integrates the judged pressure and temperature readings in chronological order to form a set of operating parameters that have a strict time synchronization relationship with the video data.

[0033] Following on from the previous specific implementation, in the monitoring system of a distillation unit area in a petrochemical plant, to obtain the operating parameters of the petrochemical pipeline in real time interactively, the system synchronously acquires pressure and temperature readings from pressure sensing units (model C) and temperature sensing units (model D) deployed on the same section of the pipeline. The system initiates a dynamic filtering mechanism: based on the timestamp (e.g., 13:05:25.045) carried by the 305th frame of video data being processed, the processing module automatically selects pressure readings (5.2 MPa) and temperature readings (285°C) with completely consistent timestamps from the continuously transmitted data stream from the sensors. Subsequently, the system performs a consistency judgment on these selected readings, checking whether their values ​​and fluctuation trends are within the normal historical range, thereby eliminating instantaneous disturbance data caused by the start-up and shutdown of nearby pumps and valves, and obtaining stable and reliable pressure and temperature readings that have passed the judgment. Finally, the system binds these valid data with the corresponding video frames, integrating them into a set of operating parameters that are completely aligned in time and have time synchronization characteristics, and transmits them to the next analysis module.

[0034] This solution obtains operating parameters that are strictly synchronized with video data and have undergone quality control through multi-sensor synchronous acquisition, dynamic time-series matching, and data validity verification. This provides reliable process status data for subsequent analysis and ensures the time consistency and data accuracy of multi-source data collaborative analysis.

[0035] Step 103: Perform thermodynamic visualization processing on the multispectral video data to separate the transient steam plume driven by the micro-positive pressure at the leak point from the complex background of the petrochemical pipeline.

[0036] Optionally, step 103 may specifically include the following steps: Step 1031: Perform pixel-level fusion of the RGB video stream acquired by the visible light sensor and the thermal radiation video stream acquired by the infrared imaging sensor in the multispectral video data to generate a fused video stream; Step 1032: Establish a background model based on pixel thermodynamic properties in the fused video stream; Step 1033: Extract pixel regions from the fused video stream of the current frame that have different thermodynamic properties from the background model; Step 1034: Perform motion continuity detection on the pixel region to identify a coherent set of pixels that have continuous motion characteristics and thermodynamic characteristics that conform to the vapor volatilization characteristics. Step 1035: Integrate the motion trajectories of the coherent pixel set in consecutive video frames to obtain a transient steam plume.

[0037] In the above scheme, thermodynamic visualization is a technical method that intuitively presents invisible thermodynamic processes through images; micro-positive pressure at the leak point refers to the small pressure difference formed at the leak point of the pipeline due to the internal pressure being higher than that of the external environment; transient steam plume is the steam flow pattern ejected from the leak point and changing rapidly over time; RGB video stream is a color video sequence containing information of the three primary colors (red, green, and blue) collected by a visible light sensor; thermal radiation video stream is a video sequence reflecting the temperature distribution of an object's surface collected by an infrared imaging sensor; fused video stream is composite video data formed by combining visible light video and thermal radiation video information; the background model of pixel thermodynamic characteristics is a reference model describing the thermodynamic characteristics of each pixel point under normal pipeline conditions; pixel regions with thermodynamic characteristic differences refer to regions in the current frame whose thermodynamic characteristics have changed significantly compared to the background model; steam volatilization characteristics are the specific thermodynamic and behavioral characteristics exhibited by steam during the leak process; coherent pixel set is a group of pixels that maintains spatial and temporal correlation across multiple consecutive frames; motion trajectory integration is the process of connecting and reconstructing the motion paths of pixel sets across multiple frames.

[0038] In this scheme, firstly, step 1031 uses a pixel-level fusion algorithm to deeply fuse the color information in the RGB video stream with the temperature information in the thermal radiation video stream, generating a fused video stream that simultaneously contains visual and thermodynamic features. Secondly, step 1032 uses the fused video stream to establish a background model based on pixel thermodynamic characteristics, which can characterize the heat distribution pattern under normal pipeline operation. Next, step 1033 extracts pixel regions with significant differences in thermodynamic characteristics by comparing the current frame with the background model in real time. Then, step 1034 performs motion continuity detection on these different pixel regions, analyzes their motion patterns and thermodynamic features in consecutive frames, and identifies coherent pixel sets that conform to the characteristics of steam volatilization. Finally, step 1035 integrates and reconstructs the motion trajectories of these coherent pixel sets in the time dimension to form a complete transient steam plume representation.

[0039] Following on from the previous specific implementation, in the monitoring system of the distillation unit area of ​​a petrochemical plant, specifically in the monitoring system of the distillation unit area of ​​Petrochemical Plant A, the system receives synchronous multispectral video data and operating parameters transmitted in steps 101 and 102. The processing module first performs pixel-level fusion of the RGB video stream acquired by the visible light sensor and the thermal radiation video stream from the infrared sensor, generating a fused video stream in which each pixel contains color and temperature information. The system uses the first 30 seconds of normal operating data to build a background model, which accurately reflects the thermal distribution characteristics of the pipeline weld in a leak-free state. When processing frame 420, the system detects a pixel area with an abnormally high temperature in the upper right corner of the background model. This area continues to move diagonally upwards in the following 5 frames, and its temperature characteristics conform to the steam volatilization law. The system connects these continuous pixels and successfully separates a transient steam plume escaping from the weld.

[0040] This solution utilizes multispectral data fusion and dynamic background modeling techniques to accurately extract subtle steam leak characteristics in complex industrial environments. It effectively overcomes the difficulties in identification caused by background interference and environmental changes, providing an accurate and reliable visual representation basis for subsequent leak analysis and identification.

[0041] Step 104: Couple the dynamic texture of the transient steam plume with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak.

[0042] Optionally, step 104 may specifically include the following steps: Step 1041: Extract the motion trajectory of the transient steam plume in consecutive video frames to form a trajectory point sequence; Step 1042: Analyze the changing pattern of the motion direction of the trajectory point sequence and establish a directional distribution map of steam diffusion; Step 1043: Associate the directional distribution map with the real-time pressure readings, and adjust the weight distribution of each direction in the directional distribution map according to the pressure value. Step 1044: Simultaneously, the change in the motion speed of the trajectory point sequence is correlated with the real-time temperature reading, and the sensitivity parameter of the change in motion speed is adjusted according to the temperature value. Step 1045: Combining the adjusted directional weight allocation and the adjusted velocity sensitivity parameters, a spatiotemporal variation template with pressure and temperature adaptive characteristics is obtained.

[0043] In the above scheme, the spatiotemporal variation template is a digital model that can simultaneously characterize the spatial distribution and temporal evolution of the target; the motion trajectory in continuous video frames refers to the record of the target object's movement path in multiple consecutive video frames; the trajectory point sequence is a set of trajectory spatial coordinate data arranged in chronological order; the motion direction change law describes the pattern and trend of the target's movement direction changing over time; the direction distribution map is a graphical representation of the statistical distribution of the target's movement probability in various directions; the weight allocation of each direction is an importance coefficient assigned to different movement directions; the motion speed change is the dynamic characteristic of the target's movement rate changing over time; the sensitivity parameter is a control coefficient used to adjust the system's response to speed changes; the adjusted direction weight is a direction importance coefficient optimized according to actual working conditions; and the adjusted speed sensitivity parameter is a speed response coefficient calibrated according to environmental conditions.

[0044] In this scheme, firstly, step 1041 uses a motion trajectory extraction algorithm to obtain the spatial position sequence of transient steam plumes from continuous video frames, forming a trajectory point sequence containing timestamps and coordinate information; secondly, step 1042 performs kinematic analysis on the trajectory point sequence, statistically analyzes its motion frequency and intensity in different directions, and establishes a directional distribution map reflecting the steam diffusion characteristics; next, step 1043 performs correlation analysis between the real-time pressure readings and the directional distribution map, dynamically adjusting the weight distribution of each direction in the directional distribution map according to the current pressure value, increasing the weight of the main diffusion direction under high pressure; simultaneously, step 1044 correlates the motion velocity change characteristics of the trajectory point sequence with real-time temperature readings, adjusting the system's sensitivity parameter to velocity changes according to the temperature value, improving the detection sensitivity to small velocity changes under high temperature conditions; finally, step 1045 integrates the adjusted directional weight distribution and velocity sensitivity parameters to generate a spatiotemporal change template that can adapt to pressure and temperature changes.

[0045] Following on from the previous specific implementation, in the monitoring system of a distillation unit area in a petrochemical plant, based on the transient steam plume identified in step 103, the system first extracts the motion trajectory of the plume in 15 consecutive video frames, forming a trajectory point sequence containing 45 spatial coordinate points. Analysis reveals that the sequence mainly moves upwards and to the upper right, thus establishing an eight-directional distribution map. At this time, the real-time pressure reading acquired by the system is 5.2 MPa, which is in a relatively high pressure range; therefore, the weight of the upward direction is adjusted from the default value of 0.3 to 0.5. Simultaneously, the real-time temperature reading is 285℃, and the system accordingly increases the velocity sensitivity parameter from 1.0 to 1.5 to enhance the ability to capture subtle velocity changes under high-temperature conditions. The resulting spatiotemporal variation template accurately reflects the diffusion characteristics of steam leakage under the current pressure and temperature conditions.

[0046] This solution deeply integrates visual motion features with pipeline operating parameters to construct a spatiotemporal feature template that can adapt to changes in operating conditions. This effectively improves the accuracy and environmental adaptability of leak feature characterization, providing a more accurate and reliable basis for subsequent leak identification.

[0047] Step 105: Using a pre-optimized AI video analysis model based on transfer learning, the spatiotemporal change template is matched with a preset steam leakage behavior spectrum to obtain the matching result.

[0048] Optionally, step 105 may specifically include the following steps: Step 1051: Perform multi-scale structural analysis on the spatiotemporal variation template to extract the feature elements and structural relationships of the feature elements related to the vapor diffusion morphology in the spatiotemporal variation template. Step 1052: Access the steam leakage behavior spectrum that stores multiple reference modes, each reference mode containing standard steam diffusion morphology characteristics corresponding to different leakage orifice diameters under specific operating parameters; Step 1053: Match the feature elements and their structural relationships with each reference pattern in the steam leakage behavior spectrum step by step, first perform overall morphological matching and then local feature matching, and record the consistency index of each matching stage during the matching process. Step 1054: Calculate the comprehensive matching degree between the spatiotemporal change template and each reference mode based on the consistency index to obtain a set of matching metric values; Step 1055: Select the reference pattern with the highest matching metric value from a set of matching metrics as the matching result.

[0049] In the appeal proposal, the AI ​​video analysis model is a computational model developed based on artificial intelligence technology for analyzing video content; the pre-set steam leakage behavior spectrum is a pre-established dataset containing standard features under various leakage scenarios; the matching result is the reference pattern identifier with the highest similarity given by the analysis model; multi-scale structural analysis is a method for analyzing the structural features of the target from different levels of detail; the feature elements related to steam diffusion morphology are the basic units describing the key morphological features in the steam diffusion process; the structural relationship of feature elements is the spatial and temporal association between different feature elements; the steam leakage behavior spectrum of multiple reference patterns contains a feature template library under various standard leakage scenarios; the standard steam diffusion morphological features are the typical morphological manifestations of steam leakage under specific conditions; the feature elements and their structural relationships are the collective term for feature elements and the association between them; overall morphological matching is a comparison of the similarity of the overall shape and motion pattern of the target; local feature matching is a fine comparison of the similarity of the detailed features of the target; the consistency index of each matching stage is a quantitative value of the degree of similarity in different matching links; a set of matching metrics is a set of similarity scores of multiple reference patterns; the reference pattern with the highest matching metric is the standard pattern with the best similarity score.

[0050] In this scheme, firstly, step 1051 uses a multi-scale structural analysis method to perform layered processing on the spatiotemporal variation template, extracting feature elements and structural relationships related to steam diffusion morphology from multiple scales, from macroscopic motion patterns to microscopic texture features. Secondly, step 1052 accesses a steam leakage behavior spectrum database storing multiple reference patterns to obtain standard steam diffusion morphological features corresponding to different leakage orifice diameters under various operating parameters. Next, step 1053 matches the extracted feature elements and their structural relationships with each reference pattern in the steam leakage behavior spectrum step by step, first comparing the overall morphological similarity and then matching the local detail features, and recording the corresponding consistency index at each matching stage. Then, step 1054 calculates the comprehensive matching degree between the spatiotemporal variation template and each reference pattern based on the recorded consistency index at each stage using a weighted fusion algorithm, obtaining a complete set of matching metrics. Finally, step 1055 selects the reference pattern with the highest value from this set of matching metrics and determines it as the final matching result.

[0051] Following on from the previous specific implementation, in the monitoring system of a distillation unit area in a petrochemical plant, the system inputs the spatiotemporal variation template generated in step 104 into an AI video analysis model optimized through transfer learning. The model first performs multi-scale structural analysis on the template, extracting the macroscopic motion features of upward steam diffusion and the microscopic texture features of edge fluctuations. Then, the model accesses a pre-set steam leakage behavior spectrum, which includes standard diffusion patterns with different leakage orifice diameters from 1 mm to 10 mm under various pressure and temperature conditions. The model first performs overall shape matching between the current template and all reference patterns in the spectrum, recording preliminary similarity scores; then, it performs fine-grained matching of local features on patterns with high matching degrees. After calculation, the current template achieves the highest overall matching degree of 0.92 with the reference pattern for a 3 mm leakage orifice diameter in the 5.0-5.5 MPa pressure range, and the system determines this as the matching result.

[0052] This solution achieves accurate identification of steam leakage modes through multi-scale feature analysis and hierarchical matching strategies, effectively improving the accuracy of leakage aperture and type discrimination, and providing a reliable technical basis for subsequent leakage risk assessment and handling decisions.

[0053] Step 106: During the matching process, the real-time thermal-hydraulic state of the petrochemical pipeline is introduced as an environmental constraint condition to dynamically correct the matching result and obtain the corrected matching result.

[0054] Optionally, step 106 may specifically include the following steps: Step 1061: Calculate the medium density and flow velocity parameters under the operating conditions of the petrochemical pipeline based on the pressure and temperature readings in the operating parameters. Step 1062: Establish a corrected relationship table for the influence of the medium density and flow velocity parameters on the vapor diffusion morphology; Step 1063: Based on the current medium density and flow velocity parameters, query the correction relationship table to obtain the corresponding morphological correction coefficient; Step 1064: Apply the morphological correction coefficient to the matching metric value obtained during the matching process to dynamically adjust the matching degree of each reference pattern; Step 1065: Select the reference pattern with the highest matching metric value after dynamic adjustment as the corrected matching result.

[0055] In the appealed scheme, the real-time thermo-hydraulic state of the petrochemical pipeline refers to the thermodynamic and hydrodynamic state conditions of the medium inside the pipeline at the current moment; environmental constraints are the restrictive requirements imposed on the system operation by external environmental factors; the corrected matching result is the optimized matching conclusion obtained after adjusting the environmental conditions; the medium density and flow velocity parameters are physical quantity parameters describing the compactness and flow velocity of the material inside the pipeline; the correction relationship table is a data table recording the corresponding relationship of adjustment coefficients under different operating conditions; the morphological correction coefficient is an adjustment coefficient that corrects the morphological feature similarity according to environmental conditions; the matching metric value is a quantitative value that measures the similarity between the target and the reference model; the reference model with the highest matching metric value after dynamic adjustment is the standard model with the best similarity score after environmental condition optimization.

[0056] In this scheme, firstly, step 1061 calculates the medium density and flow velocity parameters under the operating conditions of the petrochemical pipeline in real time using a physical property parameter calculation model based on the pressure and temperature readings in the operating parameters, thus obtaining fluid characteristic data under the current operating conditions; secondly, step 1062 establishes a correction relationship table for the influence of medium density and flow velocity parameters on steam diffusion morphology, which contains adjustment coefficients for the matching degree of each leakage mode under different density and flow velocity combinations; next, step 1063, based on the currently calculated medium density and flow velocity parameter values, queries the correction relationship table to obtain the corresponding morphology correction coefficients; then, step 1064 applies the morphology correction coefficients to the original matching metric values ​​obtained during the matching process, dynamically adjusting and optimizing the matching degree of each reference mode; finally, step 1065 selects the reference mode with the highest value from the dynamically adjusted matching metric value set, and uses it as the corrected matching result after environmental condition optimization.

[0057] Following on from the previous specific implementation, in the monitoring system of a distillation unit area in a petrochemical plant, the system, based on the preliminary matching result obtained in step 105 (matching degree of 0.92 for a 3mm leakage orifice diameter), further incorporates real-time thermal-hydraulic conditions for correction. The system calculates the medium density as 15.6 kg / m³ and the flow velocity as 25 m / s based on the current pressure of 5.2 MPa and temperature of 285°C. After consulting the preset correction relationship table, the morphological correction coefficient under the current density and flow velocity conditions is found to be 0.95. Applying this coefficient to the original matching metric, the matching degree for the 3mm leakage orifice diameter is adjusted from 0.92 to 0.874, while the matching degree for the 2.5mm orifice diameter, after correction, becomes 0.891, becoming the new highest value. The system ultimately determines the 2.5mm leakage orifice diameter as the corrected matching result.

[0058] This solution introduces real-time thermal-hydraulic conditions as environmental constraints, enabling dynamic optimization and correction of the matching results. This effectively improves the adaptability and accuracy of leak identification results to environmental changes, ensuring the reliability and practicality of leak diagnosis conclusions under actual operating conditions.

[0059] Step 107: When the corrected matching result exceeds the confidence boundary of the corresponding leakage pattern in the behavior spectrum, a pipeline leakage identification signal is generated.

[0060] Optionally, step 107 may specifically include the following steps: Step 1071: Obtain the confidence boundary parameters of the reference mode corresponding to the corrected matching result from the steam leakage behavior spectrum; Step 1072: Compare the matching metric in the corrected matching result with the confidence boundary parameter; Step 1073: When the matching metric value continues to exceed the confidence boundary parameter for a preset time length, the leakage confirmation mechanism is triggered; Step 1074: In the leakage confirmation mechanism, the persistence and stability characteristics of the matching metric exceeding the confidence boundary parameter are detected. Step 1075: When the persistence and stability characteristics simultaneously meet preset conditions, a pipeline leak identification signal containing the leak location and leak level is generated.

[0061] In the above scheme, the pipeline leak identification signal is the warning information that the system finally outputs, indicating that a pipeline leak has been detected; the confidence boundary parameter is a pre-set similarity threshold standard used to determine whether a leak has occurred; the leak confirmation mechanism is the verification process used by the system to finally confirm whether a leak event has occurred; the persistence and stability characteristics of the confidence boundary parameter refer to the time continuity and numerical fluctuation characteristics of the matching metric value exceeding the threshold.

[0062] In this scheme, firstly, step 1071 obtains the confidence boundary parameter of the reference mode corresponding to the corrected matching result from the steam leakage behavior spectrum. This parameter defines the minimum confirmation threshold of the matching metric under this leakage mode. Secondly, step 1072 compares the matching metric value in the corrected matching result with the obtained confidence boundary parameter in real time to determine whether the current matching degree exceeds the threshold standard. Next, step 1073, when the matching metric value continuously exceeds the confidence boundary parameter for a preset time length, the system triggers the leakage confirmation mechanism and enters a more stringent verification stage. Then, in step 1074, in the leakage confirmation mechanism, the system detects the persistence and stability characteristics of the matching metric value exceeding the confidence boundary parameter and analyzes its temporal continuity and numerical fluctuation. Finally, step 1075, when the persistence and stability characteristics simultaneously meet preset conditions, the system generates a pipeline leakage identification signal containing specific leakage location information and leakage level assessment.

[0063] Following on from the previous specific implementation, in the monitoring system of a distillation unit area in a petrochemical plant, based on the corrected matching result obtained in step 106 (2.5mm leakage orifice diameter, matching degree 0.891), the system queries the steam leakage behavior spectrum and finds the confidence boundary parameter corresponding to this orifice diameter to be 0.85. The system detects that the current matching degree of 0.891 exceeds the threshold for 18 seconds, and then triggers the leakage confirmation mechanism. In the confirmation mechanism, the system detects that the matching degree value maintains stable fluctuations (fluctuation range 0.885-0.895) within 18 seconds, and continuously exceeds the threshold without interruption. After all confirmation conditions are met, the system generates a pipeline leakage identification signal, clearly indicating that the leak is located at the weld on the northeast side of the distillation unit area, and the leakage level is level two.

[0064] This solution, by setting confidence boundaries and multi-dimensional verification mechanisms, achieves accurate identification and reliable alarm of leakage events, effectively avoiding false alarms and missed alarms, providing accurate location and level information for on-site emergency response, and ensuring pipeline operation safety.

[0065] Figure 2 This application provides a schematic diagram of the structure of an AI-based dynamic identification system for leaking steam in petrochemical pipelines, as shown in the figure. Figure 2 As shown, the system includes: The first acquisition module 21 is used to acquire multispectral video data synchronously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines; The second acquisition module 22 is used to simultaneously and interactively acquire the operating parameters of the petrochemical pipeline in real time, wherein the operating parameters include the pressure and temperature of the medium inside the pipeline; Processing module 23 is used to perform thermodynamic visualization processing on the multispectral video data in order to separate the transient steam plume driven by the micro-positive pressure at the leak point in the complex background of the petrochemical pipeline. The coupling module 24 is used to couple the dynamic texture of the transient steam plume with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak. The matching module 25 is used to match the spatiotemporal change template with a preset steam leakage behavior spectrum using a pre-optimized AI video analysis model that has undergone transfer learning, and obtain the matching result. The correction module 26 is used to introduce the real-time thermal-hydraulic state of the petrochemical pipeline as an environmental constraint during the matching process, and to dynamically correct the matching result to obtain the corrected matching result. The generation module 27 is used to generate a pipeline leak identification signal when the corrected matching result exceeds the confidence boundary of the corresponding leak pattern in the behavior spectrum.

[0066] Figure 2The aforementioned AI-based dynamic identification system for leaking steam in petrochemical pipelines can perform... Figure 1 The implementation principle and technical effects of the AI-based dynamic identification method for leaking steam in petrochemical pipelines described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the AI-based dynamic identification system for leaking steam in petrochemical pipelines performs its operations have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0067] In one possible design, Figure 2 The AI-based dynamic identification system for leaking steam in petrochemical pipelines, as shown in the embodiment, can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0068] The processing component 32 is used for the above Figure 1 The embodiment describes an AI-based method for dynamic identification of leaking steam in petrochemical pipelines.

[0069] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0070] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0071] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0072] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0073] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0074] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0075] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is an AI-based method for dynamic identification of leaking steam in petrochemical pipelines.

[0076] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0077] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0078] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0079] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for dynamic identification of leaking steam in petrochemical pipelines based on AI, characterized in that, include: Acquire multispectral video data simultaneously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines; Simultaneously, the operating parameters of the petrochemical pipeline are acquired in real time interactively, including the pressure and temperature of the medium inside the pipeline; The multispectral video data is subjected to thermodynamic visualization processing to separate the transient steam plume driven by the micro-positive pressure at the leak point in the complex background of the petrochemical pipeline; The dynamic texture of the transient steam plume is coupled with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak. The spatiotemporal variation template is matched with a preset steam leakage behavior spectrum using a pre-optimized AI video analysis model that has undergone transfer learning, and the matching results are obtained. During the matching process, the real-time thermal-hydraulic state of the petrochemical pipeline is introduced as an environmental constraint condition to dynamically correct the matching results and obtain the corrected matching results. When the corrected matching result exceeds the confidence boundary of the corresponding leakage pattern in the behavior spectrum, a pipeline leakage identification signal is generated.

2. The method according to claim 1, characterized in that, Simultaneously, the operating parameters of the petrochemical pipeline are acquired in real time interactively. These operating parameters include the pressure and temperature of the medium within the pipeline, including: Pressure and temperature readings are acquired synchronously by pressure and temperature sensing units deployed on the pipe surface. A dynamic filtering mechanism for the pressure and temperature readings is established, and pressure and temperature readings that match the video frame timestamps are selected based on the acquisition timing of the multispectral video data. The selected pressure and temperature readings are evaluated for consistency, and instantaneous disturbance data caused by pipeline vibration are removed to obtain the evaluated pressure and temperature readings. The determined pressure and temperature readings are integrated into a set of operating parameters with time synchronization characteristics.

3. The method according to claim 1, characterized in that, Thermodynamic visualization processing is performed on the multispectral video data to separate the transient steam plume driven by micro-positive pressure at the leak point from the complex background of the petrochemical pipeline, including: The RGB video stream acquired by the visible light sensor and the thermal radiation video stream acquired by the infrared imaging sensor are fused at the pixel level in the multispectral video data to generate a fused video stream. A background model based on pixel thermodynamic properties is established in the fused video stream; Extract pixel regions from the fused video stream of the current frame that have thermodynamic properties different from the background model; Motion continuity detection is performed on the pixel region to identify a coherent set of pixels that have continuous motion characteristics and thermodynamic properties that conform to the vapor volatilization characteristics. The motion trajectories of the coherent pixel set in consecutive video frames are integrated to obtain a transient steam plume.

4. The method according to claim 1, characterized in that, The dynamic texture of the transient steam plume is coupled with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template reflecting the specificity of the leak, including: Extract the motion trajectory of the transient steam plume in consecutive video frames to form a trajectory point sequence; Analyze the changing patterns of the motion direction of the trajectory point sequence to establish a directional distribution map of steam diffusion; The directional distribution map is correlated with the real-time pressure readings, and the weight distribution of each direction in the directional distribution map is adjusted according to the pressure value. Simultaneously, the motion speed change of the trajectory point sequence is correlated with the real-time temperature reading, and the sensitivity parameter of the motion speed change is adjusted according to the temperature value; By combining the adjusted directional weight allocation and the adjusted velocity sensitivity parameters, a spatiotemporal variation template with pressure and temperature adaptive characteristics is obtained.

5. The method according to claim 1, characterized in that, Using a pre-optimized AI video analysis model based on transfer learning, the spatiotemporal variation template is matched with a pre-set steam leakage behavior spectrum to obtain matching results, including: Multi-scale structural analysis was performed on the spatiotemporal variation template to extract the feature elements and structural relationships of the feature elements related to the vapor diffusion morphology in the spatiotemporal variation template. Access the storage of a steam leakage behavior spectrum with multiple reference modes, each of which contains standard steam diffusion morphology characteristics corresponding to different leakage orifice diameters under specific operating parameters; The feature elements and their structural relationships are matched step by step with each reference pattern in the steam leakage behavior spectrum. First, the overall morphology is matched, and then the local feature is matched. During the matching process, the consistency index of each matching stage is recorded. Based on the consistency index, the comprehensive matching degree between the spatiotemporal change template and each reference mode is calculated to obtain a set of matching metric values; Select the reference pattern with the highest matching metric value from a set of matching metrics as the matching result.

6. The method according to claim 1, characterized in that, During the matching process, the real-time thermal-hydraulic state of the petrochemical pipeline is introduced as an environmental constraint to dynamically correct the matching results, resulting in corrected matching results, including: Based on the pressure and temperature readings in the operating parameters, calculate the medium density and flow velocity parameters under the operating conditions of the petrochemical pipeline; Establish a corrected relationship table for the influence of the medium density and flow velocity parameters on the vapor diffusion morphology; Based on the current medium density and flow velocity parameters, the corresponding morphological correction coefficient is obtained by querying the correction relationship table. The morphological correction coefficient is applied to the matching metric obtained during the matching process to dynamically adjust the matching degree of each reference pattern; The reference pattern with the highest matching metric value after dynamic adjustment is selected as the corrected matching result.

7. The method according to claim 1, characterized in that, When the corrected matching result exceeds the confidence boundary of the corresponding leakage pattern in the behavioral spectrum, a pipeline leakage identification signal is generated, including: Confidence boundary parameters of the reference mode corresponding to the corrected matching result are obtained from the steam leakage behavior spectrum; The matching metric in the corrected matching result is compared with the confidence boundary parameter; When the matching metric value continuously exceeds the confidence boundary parameter for a preset time length, the leakage confirmation mechanism is triggered; In the leakage confirmation mechanism, the persistence and stability characteristics of the matching metric exceeding the confidence boundary parameter are detected. When the persistence and stability characteristics simultaneously meet preset conditions, a pipeline leak identification signal containing the leak location and leak level is generated.

8. An AI-based dynamic identification system for leaking steam in petrochemical pipelines, characterized in that, include: The first acquisition module is used to acquire multispectral video data synchronously collected by visible light sensors and infrared imaging sensors pre-deployed in high-risk areas of petrochemical pipelines. The second acquisition module is used to simultaneously and interactively acquire the operating parameters of the petrochemical pipeline in real time, wherein the operating parameters include the pressure and temperature of the medium inside the pipeline; The processing module is used to perform thermodynamic visualization processing on the multispectral video data in order to separate the transient steam plume driven by the micro-positive pressure at the leak point in the complex background of the petrochemical pipeline. The coupling module is used to couple the dynamic texture of the transient steam plume with the operating parameters of the petrochemical pipeline to construct a spatiotemporal variation template that reflects the specificity of the leak. The matching module is used to match the spatiotemporal change template with a preset steam leakage behavior spectrum using a pre-optimized AI video analysis model that has undergone transfer learning, and obtain the matching result. The correction module is used to introduce the real-time thermal-hydraulic state of the petrochemical pipeline as an environmental constraint during the matching process, and to dynamically correct the matching result to obtain the corrected matching result. The generation module is used to generate a pipeline leak identification signal when the corrected matching result exceeds the confidence boundary of the corresponding leak pattern in the behavior spectrum.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the AI-based dynamic identification method for leaking steam in petrochemical pipelines as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements an AI-based dynamic identification method for leaking steam in petrochemical pipelines as described in any one of claims 1 to 7.

Images