Floating roof storage tank sealing online monitoring and early warning method and system

Abstract

The invention discloses a floating roof storage tank sealing online monitoring and early warning method and system, and belongs to the technical field of storage and transportation equipment safety monitoring, and the method comprises the steps: obtaining air pressure difference data, deformation data and temperature field data, and generating multi-source sensing data; generating a dynamic threshold parameter through time sequence decomposition and sliding statistics; the parameter is used for carrying out cooperative verification on the air pressure difference data and the deformation data, and a leakage suspected signal is generated when synchronization is abnormal; acquiring an infrared image in response to the signal, analyzing detailed temperature field data to identify a low-temperature abnormal area, and performing spatial correlation analysis with a deformation abnormal position to generate a leakage confirmation signal; and performing coordinate fusion based on the leakage confirmation signal and the deformation data, generating a leakage point coordinate, compiling early warning information according to the leakage point coordinate, and pushing the early warning information to a user terminal. The technical scheme of multi-source sensing data cooperative verification and time-space correlation analysis is adopted, accurate confirmation and accurate positioning of leakage events can be achieved, the false alarm rate is remarkably reduced, and the early warning reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for storage and transportation equipment, and in particular to a method and system for online monitoring and early warning of the sealing of floating roof storage tanks. Background Technology

[0002] Floating roof tanks are crucial equipment in the petroleum and chemical industries for storing volatile liquids, and the integrity of their sealing devices is paramount. These sealing devices are installed in an annular space along the inner wall of the tank and the edge of the floating roof, designed to prevent leakage of the evaporating medium inside the tank, protecting the environment and reducing safety risks. Continuous online monitoring of this sealing device is a key technical aspect of ensuring the safe operation of the tank and preventing leakage accidents.

[0003] In existing technologies, monitoring leaks in floating roof tank seals typically relies on methods that detect a single physical quantity. For example, some systems monitor pressure changes by placing pressure sensors within the sealed cavity, triggering an alarm when the pressure falls below a set value. Other solutions employ gas detectors arranged around the top of the tank to determine the seal condition by detecting the concentration of leaked volatile gases. Additionally, some systems utilize infrared thermal imagers for periodic inspections, manually interpreting thermal images to locate low-temperature areas caused by leaks.

[0004] However, the aforementioned existing technical solutions have significant limitations. Methods based on single pressure or gas concentration monitoring are highly susceptible to interference from factors such as ambient temperature, wind speed, and normal fluctuations in tank levels, leading to a high false alarm rate. Relying on manual infrared inspection is not only labor-intensive and inefficient, but also cannot achieve continuous monitoring around the clock, easily missing the optimal opportunity to detect early, minor leaks. Furthermore, these methods mostly only determine the presence or absence of a leak, lacking precise location capabilities for the leak point, making it difficult to quickly and effectively guide maintenance work. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for online monitoring and early warning of floating roof tank seals. Employing a multi-source sensor data collaborative verification and spatiotemporal correlation analysis technique, it enables accurate confirmation and precise location of leakage events, significantly reducing false alarm rates and improving the reliability of early warnings.

[0006] The above objectives can be achieved through the following approach:

[0007] A method for online monitoring and early warning of the sealing of a floating roof tank includes acquiring differential pressure data of the sealing cavity, deformation data of the sealing strip, and temperature field data of the sealing surface to generate multi-source sensor data; performing time series decomposition and sliding statistics on the multi-source sensor data to obtain a first dynamic threshold and a second dynamic threshold, which are combined into a dynamic threshold parameter; using the dynamic threshold parameter to perform collaborative verification of the differential pressure data and the deformation data, and generating a suspected leak signal when the differential pressure data and the deformation data simultaneously show anomalies; acquiring a high-resolution infrared image in response to the suspected leak signal to generate detailed temperature field data; analyzing the detailed temperature field data to identify low-temperature anomaly areas, and performing spatial correlation analysis between the low-temperature anomaly areas and the abnormal positions of the deformation data to generate a leak confirmation signal; performing coordinate fusion based on the leak confirmation signal and the deformation data to generate leak point coordinates; compiling information based on the leak point coordinates and the leak confirmation signal to generate early warning information and push it to the user terminal.

[0008] Optionally, the combination of dynamic threshold parameters includes: acquiring air pressure difference data, deformation data, and temperature field data to form multi-source sensing data; performing time series analysis and fluctuation feature extraction on the multi-source sensing data to generate fluctuation features; calculating the fluctuation features based on the moving mean variance and confidence interval to obtain a first dynamic threshold for verifying the air pressure difference data and a second dynamic threshold for verifying the deformation data, which are then combined into dynamic threshold parameters.

[0009] Optionally, generating a suspected leak signal includes: comparing the real-time acquired pressure difference data with the first dynamic threshold to determine whether there is a pressure difference anomaly; comparing the real-time acquired deformation data with the second dynamic threshold to determine whether there is a deformation anomaly; and generating a suspected leak signal when both the pressure difference anomaly and the deformation anomaly are determined to exist simultaneously.

[0010] Optionally, generating detailed temperature field data includes: determining the target area and field of view in response to the suspected leakage signal, setting exposure parameters and emissivity parameters and acquiring high-resolution multi-frame infrared images to generate multi-frame raw data; performing non-uniformity correction and radiation temperature conversion on the multi-frame raw data to generate temperature sequence data; and performing denoising and sub-pixel registration reconstruction on the temperature sequence data to generate detailed temperature field data.

[0011] Optionally, the step of analyzing the detailed temperature field data to identify low-temperature anomaly regions includes: performing image processing on the detailed temperature field data to generate a visualized temperature distribution map; identifying low-temperature regions in the temperature distribution map using an image segmentation algorithm and calculating the center coordinates of the low-temperature regions; and matching the center coordinates of the low-temperature regions with the anomaly locations of the deformation data in a unified coordinate system to generate a matching result.

[0012] Optionally, generating a leak confirmation signal includes: aligning the coordinates and scales of the center coordinates of the overlapping low-temperature anomaly regions in the matching results with the anomaly positions in the deformation data to generate aligned data; calculating the positional deviation and overlap ratio based on the aligned data to generate a spatial correlation index; performing a time-series comparison of the matching results within a sliding time window to generate a temporal consistency index; and determining a threshold based on the spatial correlation index and the temporal consistency index to generate a leak confirmation signal.

[0013] Optionally, generating the leak point coordinates includes: identifying the deformation peak point in the deformation data and generating peak coordinates; extracting the center coordinates of the corresponding low-temperature anomaly region based on the leak confirmation signal and generating temperature coordinates; and performing coordinate fusion of the peak coordinates and the temperature coordinates to generate the leak point coordinates.

[0014] Optionally, generating early warning information and pushing it to the user terminal includes: assessing the severity based on the strength of the leakage confirmation signal and the confidence level of the leakage point coordinates, generating an early warning level; compiling the early warning level and the leakage point coordinates into information, generating early warning information, and pushing it to the user terminal.

[0015] Optionally, generating the warning level includes: quantifying the strength of the leakage confirmation signal and the confidence level of the leakage point coordinates to generate an intensity index and a confidence index; performing a weighted operation on the intensity index and the confidence index to generate a warning score; and mapping the warning score to a level according to a threshold range to generate a warning level.

[0016] Based on the same inventive concept, this invention also provides an online monitoring and early warning system for the sealing of floating roof tanks. The system includes: a data acquisition and fusion module for acquiring pressure difference data of the sealing cavity, deformation data of the sealing strip, and temperature field data of the sealing surface, generating multi-source sensor data; a dynamic threshold calculation module for performing time-series decomposition and sliding statistics on the multi-source sensor data to obtain a first dynamic threshold and a second dynamic threshold, which are combined into a dynamic threshold parameter; and a leakage signal detection module for using the dynamic threshold parameter to perform collaborative verification of the pressure difference data and the deformation data. When the pressure difference data and the deformation data simultaneously show abnormalities, a leak is detected. The system generates a suspected leak signal; an image acquisition module, used to acquire high-resolution infrared images in response to the suspected leak signal and generate detailed temperature field data; a leak confirmation analysis module, used to analyze the detailed temperature field data to identify low-temperature anomaly areas, and to perform spatial correlation analysis between the low-temperature anomaly areas and the abnormal positions of the deformation data to generate a leak confirmation signal; a leak point location module, used to perform coordinate fusion based on the leak confirmation signal and the deformation data to generate leak point coordinates; and a warning information push module, used to compile information based on the leak point coordinates and the leak confirmation signal, generate warning information, and push it to the user terminal.

[0017] Compared with the prior art, the present invention has the following advantages: This invention constructs a collaborative verification mechanism for multi-source sensor data, which synchronously correlates and judges global pressure differences with local sealing strip deformation anomalies. This effectively eliminates false alarms caused by environmental interference or the sensor's own malfunction due to single sensor information, and significantly improves the accuracy and reliability of identifying suspected leak signals.

[0018] This invention adopts a phased and progressive monitoring strategy. It first performs macroscopic screening using conventional sensors, and only activates high-resolution infrared imaging equipment for detailed confirmation when highly suspicious events are identified. This on-demand activation mode optimizes the allocation of system resources, reduces energy consumption and data processing burden while ensuring high sensitivity, and improves the overall operating efficiency of the monitoring system.

[0019] This invention performs spatial correlation analysis and coordinate fusion of abnormal information from different physical modes in a unified coordinate system, accurately matching and locating one-dimensional deformation anomaly locations with two-dimensional temperature field anomaly regions. This enables the confirmation of leak points and high-precision coordinate calculation, providing intuitive and reliable location guidance for subsequent maintenance decisions and emergency responses, and improving the accuracy and timeliness of fault handling.

[0020] This invention introduces dynamic threshold technology and a multi-indicator severity assessment model, enabling the early warning system to adapt to changes in tank operating conditions and classify confirmed leakage events into risk levels. This not only improves the system's ability to detect early, minor leaks but also makes the early warning information more hierarchical and instructive, providing a scientific basis for managers to implement differentiated emergency response strategies.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an online monitoring and early warning method for the sealing of a floating roof tank according to an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of the dynamic threshold parameters in an embodiment of the present invention.

[0025] Figure 3 This is a leakage confirmation signal determination diagram according to an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the fusion of leak point coordinates according to an embodiment of the present invention.

[0027] Figure 5 This is a schematic diagram of the structure of an online monitoring and early warning system for the sealing of a floating roof tank according to an embodiment of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1One embodiment of the present invention proposes an online monitoring and early warning method for the sealing of floating roof tanks. It adopts a technical solution of multi-source sensor data collaborative verification and spatiotemporal correlation analysis, which can realize accurate confirmation and precise location of leakage events, significantly reduce false alarm rate and improve the reliability of early warning.

[0030] The method described in this embodiment specifically includes: Acquire the air pressure difference data of the sealed cavity, the deformation data of the sealing strip, and the temperature field data of the sealing surface to generate multi-source sensing data; Time series decomposition and sliding statistics are performed on multi-source sensor data to obtain a first dynamic threshold and a second dynamic threshold, which are then combined into a dynamic threshold parameter. The dynamic threshold parameter is used to perform joint verification of the pressure difference data and the deformation data. When the pressure difference data and the deformation data show abnormality simultaneously, a suspected leak signal is generated. In response to the suspected leakage signal, a high-resolution infrared image is acquired, generating detailed temperature field data; The detailed temperature field data is analyzed to identify low-temperature anomaly areas, and the low-temperature anomaly areas are spatially correlated with the abnormal locations of the deformation data to generate a leak confirmation signal. Based on the leakage confirmation signal and the deformation data, coordinate fusion is performed to generate the coordinates of the leakage point;

[0031] Information is compiled based on the coordinates of the leak point and the leak confirmation signal, and an early warning message is generated and pushed to the user terminal.

[0032] Specifically, the system first performs long-term monitoring of two macroscopic physical quantities: the pressure difference in the sealed cavity and the deformation of the sealing strip. Dynamic threshold technology is then used for adaptive judgment to achieve preliminary collaborative screening of abnormal states. Once both pressure and deformation data show anomalies simultaneously, it is classified as a highly suspicious event, triggering a high-precision, small-scale detailed investigation mechanism. This mechanism acquires high-resolution infrared thermal imaging data to capture localized low-temperature characteristics caused by media leakage. Subsequently, in-depth spatiotemporal correlation analysis is performed to match and verify the macroscopic deformation anomaly location with the microscopic low-temperature anomaly region in a unified spatial coordinate system. Only when the physical deformation and thermal field anomalies highly overlap in spatial location is the leakage event finally confirmed. Finally, by fusing the coordinate information from deformation sensors and infrared images, a precise leak point coordinate is generated, and an early warning message including the location and confirmation status is pushed to the user. By combining spatial correlation analysis and coordinate fusion of physical deformation information with high-resolution thermal imaging information, a complete closed loop for monitoring leak events, from suspected to confirmed to precise location, is achieved. This not only greatly improves the accuracy of leak diagnosis but also provides precise location guidance for subsequent maintenance and emergency response, thereby enabling comprehensive, efficient, and reliable online monitoring and early warning of the sealing status of floating roof tanks.

[0033] Optionally, the combination of dynamic threshold parameters includes: Acquire air pressure difference data, deformation data, and temperature field data to form multi-source sensing data; Time series analysis and fluctuation feature extraction are performed on the multi-source sensor data to generate fluctuation features;

[0034] The fluctuation characteristics are calculated based on the moving mean variance and confidence interval to obtain a first dynamic threshold for verifying the pressure difference data and a second dynamic threshold for verifying the deformation data, which are combined into a dynamic threshold parameter.

[0035] Specifically, firstly, three key physical quantities are acquired in real time using distributed sensors: the pressure difference between the sealed cavity and the external environment, the deformation data of the sealing strip along its circumference, and the temperature field data of the sealing surface. These three sets of time-series data constitute multi-source sensor data. Next, to generate dynamic thresholds that can adapt to changes in the tank's operating conditions, time-series analysis and fluctuation feature extraction are performed on the multi-source sensor data. The core is to quantify the fluctuation range of each sensor data point under normal operating conditions. Taking the pressure difference data as an example, within a sliding time window, the standard deviation of all pressure difference data points within that window is calculated. This standard deviation characterizes the recent volatility of the pressure signal and serves as its fluctuation feature. Similarly, the deformation data undergoes the same processing to extract its fluctuation features within the same time window. After obtaining the fluctuation features of each data source, based on the statistical principles of the sliding mean variance and confidence interval, dynamic anomaly judgment boundaries are calculated for the pressure difference data and deformation data respectively. This process is achieved through the following formula: , , Here, and Represents respectively in The upper and lower limits of the dynamic threshold at any given time; It is the mean of the sensor data within the current sliding time window, which reflects the baseline level of the signal; The standard deviation of the sensor data within the current sliding time window is the extracted fluctuation feature, which quantifies the normal fluctuation amplitude of the signal. The confidence coefficient is a constant pre-set based on historical operational data analysis or expert experience. It is used to adjust the sensitivity of the early warning system, and its value determines the probability that normal data falls within the threshold range. For example... Figure 2 As shown, the moving average of the differential pressure data at different time points and the upper and lower thresholds dynamically calculated based on its moving standard deviation are illustrated. Using this method, a first dynamic threshold for verifying the differential pressure data and a second dynamic threshold for verifying the deformation data are calculated. These two dynamic thresholds together constitute a combined dynamic threshold parameter. Compared to traditional methods using fixed thresholds, the dynamic thresholds generated by this method can dynamically adjust according to the operating conditions of the storage tank, changes in ambient temperature, and normal aging of the sealing device, effectively filtering out signal disturbances caused by fluctuations in normal operating conditions.

[0036] Optionally, generating a suspected leak signal includes: The real-time acquired air pressure difference data is compared with the first dynamic threshold to determine whether there is an abnormal air pressure difference. The deformation data acquired in real time is compared with the second dynamic threshold to determine whether there is any deformation anomaly; When both the pressure difference and the deformation anomaly are determined to exist simultaneously, a suspected leak signal is generated.

[0037] Specifically, after acquiring the first dynamic threshold for verifying the pressure difference data and the second dynamic threshold for verifying the deformation data, the real-time monitoring and anomaly detection stage begins. In this stage, two streams of real-time sensor data are processed in parallel. The first stream is for anomaly detection of the pressure difference. The real-time acquired pressure difference data of the sealed cavity is set as... It is continuously compared with a first dynamic threshold. The first dynamic threshold includes an upper limit. and a lower limit When real-time air pressure difference data Exceeding this dynamic range, especially when it falls below the lower limit. When this occurs, it indicates a decrease in the pressure maintenance capacity within the sealing cavity, and this event is judged as an abnormality in air pressure. The second approach is for abnormal judgment of sealing strip deformation. Simultaneously, the deformation data acquired in real time along the circumferential distribution of the sealing strip is set as... ,in The sensor's location index is compared to a second dynamic threshold. Similar to the pressure difference threshold, the second dynamic threshold also includes an upper limit. and a lower limit When the deformation data at any location When the dynamic threshold range is exceeded, for example, when local deformation increases due to media leakage impact and exceeds the upper limit. If the deformation anomaly is detected at that location, it is determined that an anomaly exists. The key to generating a suspected leak signal lies in verifying the synchronicity of these two anomalies. A logical gate is established; a suspected leak signal is only generated when the judgment results of both the pressure difference anomaly and the deformation anomaly are true simultaneously at the same monitoring time point or within a very short time window. This logical relationship can be expressed as: , in, Representative at Constantly generated suspected leak signals; For the Boolean determination of abnormal air pressure difference, when The time is true; The Boolean determination result for deformation anomalies is given when at least one location exists. Make When is true; symbol This represents a logical AND operation. The suspected leak signal serves as an internal trigger, initiating a subsequent high-precision verification process. This multimodal information fusion verification method significantly improves the confidence level of the initial warning, ensuring that only highly suspicious events trigger more complex subsequent verification analyses.

[0038] Optionally, generating detailed temperature field data includes: In response to the suspected leakage signal, the target area and field of view are determined, exposure parameters and emissivity parameters are set, and high-resolution multi-frame infrared images are acquired to generate multi-frame raw data. Non-uniformity correction and radiation temperature conversion are performed on the multiple frames of raw data to generate temperature sequence data; The temperature sequence data is denoised and reconstructed using sub-pixel registration to generate detailed temperature field data.

[0039] Specifically, firstly, based on the deformation anomaly location information associated with the generation of the suspected leak signal, the pan-tilt platform of the infrared camera is controlled to precisely point the camera's field of view towards the target area. After target locking is completed, imaging parameters are optimized. This includes setting precise emissivity parameters based on the material properties of the tank sealing strip and environmental conditions, which is fundamental to ensuring the accuracy of temperature measurements. Simultaneously, the camera's exposure parameters, i.e., integration time, are adjusted to obtain the optimal image dynamic range, avoiding thermal information loss due to overexposure or underexposure. After parameter settings are completed, the camera continuously acquires a sequence of high-resolution multi-frame infrared images of the target area; these unprocessed image sequences constitute multi-frame raw data. Next, a series of processes are performed on the acquired multi-frame raw data to purify and enhance thermal information. The first step is to perform non-uniformity correction on the multi-frame raw data. Non-uniformity correction refers to eliminating fixed-pattern noise caused by inconsistent responses of individual pixels in the infrared detector array, ensuring that changes in image brightness truly reflect differences in target radiation. After calibration, the grayscale value of each pixel is converted into an absolute radiation temperature value using preset emissivity parameters and internal calibration curves. This process is called radiation temperature conversion. The preset emissivity parameters are precisely set during the imaging parameter optimization stage, based on the material properties of the tank sealing strip and environmental conditions. The internal calibration curves are functions established and fixed within the infrared camera during manufacturing or periodic calibration, used to convert the detector's original grayscale values ​​into radiation temperature. After this step, multiple frames of raw data are converted into a time series containing precise temperature information, i.e., temperature sequence data. Finally, to obtain the final detailed temperature field data, advanced image processing is performed on the temperature sequence data. First, denoising is performed by averaging multiple frames of the time series or using more complex temporal filtering algorithms to effectively suppress random thermal noise and improve the image's signal-to-noise ratio. Then, subpixel registration and reconstruction are performed. Subpixel registration is an image processing technique that can accurately calculate subpixel-level displacements between multiple frames caused by factors such as minute vibrations. After precisely aligning each frame of the image, the reconstruction algorithm utilizes this complementary information to generate a super-resolution temperature image with a spatial resolution exceeding that of a single original frame. This final high-quality, high-resolution temperature image serves as detailed temperature field data, providing precise evidence for subsequent leak confirmation. This method, through a complete workflow from target localization and optimized acquisition to multi-level image processing, transforms a vague, suspected signal into high-definition, high-precision thermal imaging evidence. Compared to single-capture infrared images, the detailed temperature field data generated by this method, through non-uniformity correction and precise radiation temperature conversion, ensures the accuracy of temperature readings.

[0040] Optionally, analyzing the detailed temperature field data to identify low-temperature anomaly regions includes: The detailed temperature field data is processed to generate a visualized temperature distribution map; The low-temperature region is identified in the temperature distribution map using an image segmentation algorithm, and the center coordinates of the low-temperature region are calculated. The coordinates of the center of the low-temperature region are matched with the abnormal locations of the deformation data in a unified coordinate system to generate a matching result.

[0041] Specifically, firstly, detailed temperature field data undergoes image processing, transforming it from a numerical matrix into an intuitive, visualized temperature distribution map. This is achieved through pseudo-color encoding, where different colors or grayscale levels represent different temperature values, making temperature changes, especially in low-temperature regions, clearly observable. Next, automated low-temperature region identification is performed on this visualized temperature distribution map. This step employs image segmentation algorithms, such as threshold-based or region-growing-based segmentation methods. In threshold-based segmentation, a temperature threshold is automatically calculated or set based on the overall temperature statistics of the image. Regions with pixel temperatures below this threshold are segmented, forming one or more connected components, which are the candidate low-temperature regions. Subsequently, the geometric center of each identified low-temperature region is calculated, typically its centroid coordinates. These coordinates are the center coordinates of the low-temperature region, pinpointing the precise location of the thermal anomaly in the image coordinate system. The final step is spatial correlation matching. To achieve this, a unified coordinate system must be pre-established, capable of describing the physical space of the floating roof tank sealing strip. The deformation sensor itself has a defined physical installation location, and its anomaly location can be directly mapped to this unified coordinate system. Simultaneously, the geometric transformation relationship between the imaging field of view of the infrared camera and this unified coordinate system needs to be determined in advance through calibration. Using this transformation relationship, the center coordinates of the low-temperature region in the image coordinate system can be converted into physical coordinates in the unified coordinate system. Then, the transformed center coordinates of the low-temperature region are compared with the anomaly location of the deformation data that triggered this analysis in the unified coordinate system. If the spatial positions of the two are close or coincident within the tolerance range, a matching result is generated, which records the pair of interrelated anomalies and their location information in the unified coordinate system. This method effectively fuses a non-localized air pressure signal and a one-dimensional deformation signal with a two-dimensional temperature field signal, providing strong spatial consistency evidence for subsequent leak confirmation and greatly improving the accuracy and reliability of leak diagnosis.

[0042] Optionally, generating a leak confirmation signal includes: The coordinates of the center coordinates of the low-temperature anomaly regions that overlap in the matching results are aligned with the anomaly positions of the deformation data, and the scale is normalized to generate aligned data. Based on the alignment data, the positional deviation and overlap ratio are calculated to generate spatial correlation indicators; The matching results are compared temporally within a sliding time window to generate a time consistency index. A leakage confirmation signal is generated by determining a threshold based on the spatial correlation index and the temporal consistency index.

[0043] Specifically, the first step is to align and normalize the coordinates of the centers of the spatially overlapping low-temperature anomaly regions in the matching results with the anomaly locations in the deformation data. Coordinate alignment aims to eliminate minor errors that may exist in coordinate system transformation, while scale normalization ensures that the point-like deformation sensor location information and the area-like low-temperature region information are placed within a comparable framework. After processing, aligned data is generated. Based on this aligned data, a series of quantitative indicators are calculated to assess the strength of their correlation. The first is the spatial correlation index. This index consists of two components: one is the positional deviation, i.e., the Euclidean distance between the center of the aligned low-temperature region and the deformation anomaly location; the other is the overlap ratio, which assesses the degree of overlap between the extent of the low-temperature anomaly region and the influence range of the deformation sensor. By weighted combination of these two components, a comprehensive spatial correlation index is generated, which can quantitatively describe the spatial coupling strength between thermal anomalies and physical deformation. Next, a time dimension is introduced for verification. Within a sliding time window, the continuously generated matching results are compared temporally. By analyzing whether the matching phenomenon at a specific location occurs consistently and stably within this time window, a temporal consistency index is calculated. For example, this indicator can be defined as the proportion of matching occurrences to the total number of monitoring occurrences. A real leak event should have physical persistence; therefore, high temporal consistency is an important characteristic. Finally, the two key indicators mentioned above are comprehensively judged based on pre-set thresholds. These thresholds refer to the specific judgment limits set for the spatial correlation indicator and the temporal consistency indicator during the calibration phase, based on safety standards, engineering experience, historical data statistical analysis, etc. Only when the spatial correlation indicator exceeds its set threshold, and the temporal consistency indicator also exceeds its threshold, will a leak confirmation signal be finally generated. Figure 3 As shown, this illustrates how spatial correlation indicators and temporal consistency indicators jointly determine the generation of a leak confirmation signal. This logical judgment ensures that only anomalies that are highly overlapping spatially and stable temporally are confirmed as genuine leaks. The leak confirmation signal is a clear Boolean signal or a signal with a confidence score, used to trigger subsequent location and early warning processes. This spatiotemporal joint deep analysis method constructs a rigorous logical verification chain, resulting in a leak confirmation signal with extremely high confidence, providing a solid and reliable decision-making basis for subsequent accurate early warning and emergency response.

[0044] Optionally, the generation of leak point coordinates includes: Identify the peak deformation points in the deformation data and generate peak coordinates; Based on the leak confirmation signal, the center coordinates of the corresponding low-temperature anomaly region are extracted to generate temperature coordinates; The peak coordinates and the temperature coordinates are fused to generate the coordinates of the leak point.

[0045] Specifically, firstly, the deformation data associated with the leak confirmation signal is analyzed. Deformation sensors are discretely deployed circumferentially along the sealing strip. The readings of all deformation sensors at the time of leak confirmation are retrieved, and the point with the largest deformation is identified as the deformation peak point. Since the physical installation position of each deformation sensor is precisely calibrated beforehand, the coordinates of this deformation peak point in a unified coordinate system can be directly obtained and generated as peak coordinates. These peak coordinates provide a preliminary, but coarse, indication of the leak location. Simultaneously, based on the same leak confirmation signal, the center coordinates of the corresponding low-temperature anomaly region are extracted. During the generation of the leak confirmation signal, a specific low-temperature anomaly region has been identified and verified, and its center coordinates in a unified coordinate system have been calculated. This step directly uses the existing calculation results as temperature coordinates. Compared to peak coordinates, temperature coordinates calculated from high-resolution infrared images have higher spatial resolution and can more precisely depict the center of the leak's influence. Finally, the two types of coordinates are fused to generate the final leak point coordinates. This fusion process employs a weighted averaging strategy, aiming to combine the advantages of two measurement methods to obtain a more accurate and reliable result than a single measurement. Its calculation method can be expressed as: , in, These are the final coordinates of the leak point; These are the peak coordinates generated from the deformation data. These are temperature coordinates generated from detailed temperature field data; and These are the weighting coefficients assigned to the peak coordinate and the temperature coordinate, respectively, and their sum is 1. These two weighting coefficients are preset or dynamically adjusted based on factors such as the inherent accuracy of the two sensors and the confidence level of the signal. For example, since the spatial resolution of infrared imaging is usually much higher than that of discretely deployed deformation sensors, the weight of the temperature coordinate is... The values ​​are typically set higher. This weighted fusion calculation yields a high-precision coordinate system for the leak point, integrating information on macroscopic physical deformation and fine thermal field distribution. For example... Figure 4The diagram illustrates the process of generating the final leak point coordinates by weighted fusion of the deformation peak coordinates and the center coordinates of the low-temperature anomaly region. This multimodal data fusion positioning strategy significantly improves the final accuracy and reliability of the leak point coordinates, providing precise target guidance for subsequent maintenance and emergency response.

[0046] Optionally, generating the warning information and pushing it to the user terminal includes: The severity is assessed based on the strength of the leak confirmation signal and the confidence level of the leak point coordinates, and an early warning level is generated. The warning level and the coordinates of the leak point are combined to generate warning information and push it to the user terminal.

[0047] Specifically, the severity of the leak event is first assessed online. This assessment is based on two core inputs: the strength of the leak confirmation signal and the confidence level of the leak point coordinates. The strength of the leak confirmation signal is a comprehensive quantification of various indicators during the leak confirmation process, such as spatial correlation indicators and temporal consistency indicators. A strong signal indicates that the evidence of the leak is very conclusive and significant. The confidence level of the leak point coordinates is a measure of the reliability of the coordinate fusion result, which depends on the quality and consistency of the sensor data involved in the fusion. Using a predefined assessment model, these two inputs are analyzed and calculated to generate a clear warning level. This level is typically divided into multiple levels, such as "Level 1 Alert," "Level 2 Alert," and "Level 3 Alert," corresponding to different levels of leak risk and emergency response requirements. After determining the warning level, information is compiled. This step integrates multiple key information elements into a structured warning information package. The core content of this information package includes the previously calculated warning level and high-precision leak point coordinates. In addition, it may include other auxiliary information, such as the timestamp of the leak, a summary of the associated raw sensor data, and a snapshot of a visualized temperature distribution map. The goal of information compilation is to enable users to quickly and comprehensively understand the nature, location, and severity of a leak. The final step is the push notification of the warning information. The compiled warning information is proactively pushed to one or more designated user terminals via communication links, such as wireless networks, mobile communication networks, or industrial Ethernet. These user terminals can be monitoring screens in the field control room, handheld devices of maintenance personnel, or workstations in the remote management center. The push mechanism ensures that critical information reaches the relevant responsible persons as soon as possible, enabling them to promptly activate emergency plans. By binding precise leak point coordinates with warning levels and proactively pushing the information, this method ensures the integrity and timeliness of the information, significantly shortening the response time from incident discovery to personnel response, and providing crucial technical support for quickly controlling the situation and minimizing losses.

[0048] Optionally, the generation of early warning levels includes: The intensity of the leakage confirmation signal and the confidence level of the leakage point coordinates are quantified to generate intensity and confidence indices. The intensity index and the confidence index are weighted and calculated to generate a warning score; The warning scores are mapped to different levels based on the threshold range to generate warning levels.

[0049] Specifically, the strength of the leak confirmation signal is first converted into a numerical strength index. The strength of the leak confirmation signal itself comes from the quantitative assessment of its supporting evidence, such as spatial correlation indicators and temporal consistency indicators. These composite index values ​​are mapped to a standardized range, such as 0 to 1, using a normalization function, thus generating the strength index. A higher strength index indicates a more significant and longer-lasting leak, making the leak event more certain. Simultaneously, the confidence level of the leak point coordinates is quantified to generate a confidence index. The leak point coordinates are obtained by fusing deformation peak coordinates and temperature coordinates; their confidence level depends on the consistency between these two source coordinates. The spatial distance or other statistical deviation between these two coordinates before fusion can be calculated, and this deviation can be normalized using an inverse function; the smaller the deviation, the higher the confidence level. The resulting confidence index also falls within a standardized range, reflecting the reliability of the location results. After obtaining these two independent quantitative indicators, they are weighted to generate a comprehensive warning score. This calculation is performed using the following formula: , in, This is the final calculated warning score. It is the aforementioned strength index, and It is a confidence level indicator. and These are the weighting coefficients assigned to the strength index and the confidence index, respectively. They are constants pre-set according to the risk management strategy and satisfy the following conditions: and The sum of these weights is 1. These weights reflect the different emphases on the certainty of event occurrence and the precision of location when assessing overall risk. Finally, based on the defined threshold ranges, the calculated warning scores are mapped to levels to generate the final warning level. Internally, a series of score thresholds divide the entire range of warning scores into several non-overlapping intervals, each interval corresponding to a warning level, such as "Level 1 Alert," "Level 2 Alert," and "Level 3 Alert." The calculated warning score is compared with these thresholds to determine its interval, and the corresponding warning level is output as the final result. This method, through a structured quantitative assessment model, condenses complex, multi-dimensional leakage event information into a single, intuitive warning level, thereby achieving standardized and hierarchical risk management.

[0050] Based on the same inventive concept, such as Figure 5 As shown, the present invention also provides an online monitoring and early warning system for the sealing of floating roof tanks, the system comprising: The data acquisition and fusion module is used to acquire the air pressure difference data of the sealed cavity, the deformation data of the sealing strip, and the temperature field data of the sealing surface, and generate multi-source sensor data. The dynamic threshold calculation module is used to perform time series decomposition and sliding statistics on multi-source sensor data to obtain the first dynamic threshold and the second dynamic threshold, which are combined into dynamic threshold parameters. The leakage signal detection module is used to perform collaborative verification of the pressure difference data and the deformation data using the dynamic threshold parameter. When the pressure difference data and the deformation data show abnormality simultaneously, a suspected leakage signal is generated. The image acquisition module is used to acquire high-resolution infrared images in response to the suspected leakage signal and generate detailed temperature field data; The leak confirmation analysis module is used to analyze the detailed temperature field data to identify low-temperature anomaly areas, and to perform spatial correlation analysis between the low-temperature anomaly areas and the abnormal locations of the deformation data to generate a leak confirmation signal. The leak point location module is used to perform coordinate fusion based on the leak confirmation signal and the deformation data to generate leak point coordinates;

[0051] The early warning information push module is used to compile information based on the coordinates of the leak point and the leak confirmation signal, generate early warning information, and push it to the user terminal.

[0052] To verify the feasibility of this invention in practice, it was applied to a large petrochemical storage and transportation base. This base possesses a series of large crude oil floating roof storage tanks, and the safe operation of their sealing devices is crucial. Traditional tank sealing monitoring mainly relies on manual inspections, which is not only inefficient but also makes it difficult to detect early, minor leaks in real time, posing safety hazards. This base aims to adopt the method and system of this invention for 24 / 7 online monitoring and automated early warning of the tank sealing status.

[0053] In this embodiment, the online monitoring and early warning system described in this invention was deployed on a crude oil storage tank, T-501, at the base. The system continuously acquires multi-source sensor data, including pressure difference, deformation, and temperature field, through pressure sensors installed within the sealed cavity, deformation sensors evenly distributed circumferentially along the sealing strip, and an infrared thermal imager monitoring the sealing surface. The system's backend includes a dynamic threshold calculation module, a leak signal detection module, a leak confirmation and analysis module, a leak point location module, and an early warning information push module.

[0054] To verify the effectiveness of the invention, the system underwent a six-month continuous operation test, during which all monitoring data of the storage tank during normal operation, operating condition fluctuations, and a real minor leak event were recorded. The following is the monitoring, analysis, and early warning process for this minor leak event.

[0055] At 10:30 AM on January 15, 2025, the system's data acquisition and fusion module normally acquired various sensor data. The dynamic threshold calculation module performed sliding statistics based on data from the past 24 hours, generating a first dynamic threshold for the pressure difference data and a second dynamic threshold for the deformation data at each location. At 10:32 AM, the leak signal detection module detected that the pressure difference data in the sealed cavity rapidly decreased from the normal -250 Pa to -350 Pa, falling below the lower limit of the first dynamic threshold (-300 Pa), thus identifying it as an abnormal pressure difference. Almost simultaneously, the reading of deformation sensor No. 27, located at a 75° circumferential position on the storage tank, increased from the normal fluctuation range of 2 mm to 8 mm, exceeding its second dynamic threshold upper limit (5 mm), thus identifying it as an abnormal deformation. Because the abnormal pressure difference and abnormal deformation occurred synchronously, the system generated a suspected leak signal.

[0056] In response to the suspected leak signal, the image acquisition module immediately activated, controlling the high-resolution infrared thermal imager to aim at the target area where deformation sensor No. 27 was located, and acquiring multiple frames of raw data. After non-uniformity correction, radiation temperature conversion, noise reduction, and sub-pixel registration reconstruction, detailed temperature field data was generated. The leak confirmation analysis module analyzed this data and identified a low-temperature anomaly area with a center temperature approximately 3.2°C lower than the surrounding area using an image segmentation algorithm. Using a pre-calibrated coordinate system, the system performed spatial correlation analysis between the center coordinates of this low-temperature anomaly area and the anomaly position of deformation sensor No. 27. The calculation showed that the spatial positional deviation between the two was less than 5 cm, with an overlap ratio as high as 92%, indicating a significant spatial correlation index. Furthermore, this matching phenomenon continued to occur stably within the next 5-minute sliding time window, and the time consistency index exceeded the threshold. Therefore, the system ultimately generated a leak confirmation signal.

[0057] Subsequently, the leak location module was activated. The system extracted the peak coordinates of deformation sensor #27 and the center coordinates of the confirmed low-temperature anomaly area. This was achieved through weighted coordinate fusion (temperature coordinate weighting). Set to 0.7, deformation coordinate weight (Set to 0.3), the final leak point coordinates were generated with centimeter-level accuracy.

[0058] Finally, the early warning information push module assesses the severity based on the strength of the leak confirmation signal and the confidence level of the leak point coordinates. The system calculates an intensity index of 0.85 and a confidence index of 0.92, resulting in a weighted early warning score of 0.88, which is mapped to a "Level 2 Alert" according to the threshold range. The system immediately pushes the early warning information, including the early warning level, precise leak point coordinates, and incident timestamp, to the monitoring screen in the central control room and the handheld terminal of the on-site maintenance engineer. The entire process, from detecting the anomaly to pushing the early warning information, takes less than 2 minutes.

[0059] Table 1. Data on routine monitoring and abnormal detection of the sealing status of storage tanks

[0060] Table 2. Leakage Incident Confirmation and Location Analysis Data Table

[0061] Table 3. Early Warning Information Generation and Push Response Data Table

[0062] As can be seen from the data recorded in Tables 1 to 3 above, the practical application effect of the present invention in the monitoring of floating roof tank seals is significant.

[0063] Table 1 clearly demonstrates how the system successfully captured synchronous anomalies caused by minor leaks using dynamic thresholds. Compared to fixed thresholds, dynamic thresholds can adapt to normal fluctuations in operating conditions, accurately identifying the dual breakthroughs in pressure difference and deformation at the moment the anomaly occurred (10:32), effectively avoiding false alarms and missed alarms.

[0064] The data in Table 2 demonstrate that the leak confirmation analysis logic of this invention is rigorous and quantitative. Through in-depth analysis of thermal imaging data, a high spatial correlation index (0.90) and a high temporal consistency index (0.95) were calculated, providing strong evidence for the confirmation of leak events. Its coordinate fusion positioning method combines the advantages of different sensors, ultimately outputting centimeter-level precise leak point coordinates, providing accurate guidance for subsequent maintenance.

[0065] Table 3 highlights the significant advantage of this invention in response efficiency. The entire automated process, from detection, confirmation, location to early warning, is completed within 2 minutes, while the response time for traditional manual inspection can be as long as 24 hours. Through intelligent severity assessment, the system-generated "Level 2 Alarm" provides managers with clear decision-making basis, enabling differentiated risk management. These data results fully demonstrate the superior performance of this invention in improving the accuracy, timeliness, and reliability of floating roof tank sealing monitoring.

[0066] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0067] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for online monitoring and early warning of the sealing of floating roof tanks, characterized in that, The method includes: Acquire the air pressure difference data of the sealed cavity, the deformation data of the sealing strip, and the temperature field data of the sealing surface to generate multi-source sensing data; Time series decomposition and sliding statistics are performed on multi-source sensor data to obtain a first dynamic threshold and a second dynamic threshold, which are then combined into a dynamic threshold parameter. The dynamic threshold parameter is used to perform joint verification of the pressure difference data and the deformation data. When the pressure difference data and the deformation data show abnormality simultaneously, a suspected leak signal is generated. In response to the suspected leakage signal, a high-resolution infrared image is acquired, generating detailed temperature field data; The detailed temperature field data is analyzed to identify low-temperature anomaly areas, and the low-temperature anomaly areas are spatially correlated with the abnormal locations of the deformation data to generate a leak confirmation signal. Based on the leakage confirmation signal and the deformation data, coordinate fusion is performed to generate the coordinates of the leakage point; Information is compiled based on the coordinates of the leak point and the leak confirmation signal, and an early warning message is generated and pushed to the user terminal.

2. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 1, characterized in that, The combination of dynamic threshold parameters includes: Acquire air pressure difference data, deformation data, and temperature field data to form multi-source sensing data; Time series analysis and fluctuation feature extraction are performed on the multi-source sensor data to generate fluctuation features; The fluctuation characteristics are calculated based on the moving mean variance and confidence interval to obtain a first dynamic threshold for verifying the pressure difference data and a second dynamic threshold for verifying the deformation data, which are combined into a dynamic threshold parameter.

3. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 2, characterized in that, The generation of suspected leak signals includes: The real-time acquired air pressure difference data is compared with the first dynamic threshold to determine whether there is an abnormal air pressure difference. The deformation data acquired in real time is compared with the second dynamic threshold to determine whether there is any deformation anomaly; When both the pressure difference and the deformation anomaly are determined to exist simultaneously, a suspected leak signal is generated.

4. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 3, characterized in that, The generation of detailed temperature field data includes: In response to the suspected leakage signal, the target area and field of view are determined, exposure parameters and emissivity parameters are set, and high-resolution multi-frame infrared images are acquired to generate multi-frame raw data. Non-uniformity correction and radiation temperature conversion are performed on the multiple frames of raw data to generate temperature sequence data; The temperature sequence data is denoised and reconstructed using sub-pixel registration to generate detailed temperature field data.

5. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 4, characterized in that, The analysis of the detailed temperature field data to identify low-temperature anomaly regions includes: The detailed temperature field data is processed to generate a visualized temperature distribution map; The low-temperature region is identified in the temperature distribution map using an image segmentation algorithm, and the center coordinates of the low-temperature region are calculated. The coordinates of the center of the low-temperature region are matched with the abnormal locations of the deformation data in a unified coordinate system to generate a matching result.

6. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 5, characterized in that, The generation of the leak confirmation signal includes: The coordinates of the center coordinates of the low-temperature anomaly regions that overlap in the matching results are aligned with the anomaly positions of the deformation data, and the scale is normalized to generate aligned data. Based on the alignment data, the positional deviation and overlap ratio are calculated to generate spatial correlation indicators; The matching results are compared temporally within a sliding time window to generate a time consistency index. A leakage confirmation signal is generated by determining a threshold based on the spatial correlation index and the temporal consistency index.

7. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 6, characterized in that, The generated leak point coordinates include: Identify the peak deformation points in the deformation data and generate peak coordinates; Based on the leak confirmation signal, the center coordinates of the corresponding low-temperature anomaly region are extracted to generate temperature coordinates; The peak coordinates and the temperature coordinates are fused to generate the coordinates of the leak point.

8. The method for online monitoring and early warning of the sealing of a floating roof tank according to claim 7, characterized in that, The process of generating and pushing early warning information to the user terminal includes: The severity is assessed based on the strength of the leak confirmation signal and the confidence level of the leak point coordinates, and an early warning level is generated. The warning level and the coordinates of the leak point are combined to generate warning information and push it to the user terminal.

9. A method for online monitoring and early warning of the sealing of a floating roof tank according to claim 8, characterized in that, The generated early warning levels include: The intensity of the leakage confirmation signal and the confidence level of the leakage point coordinates are quantified to generate intensity and confidence indices. The intensity index and the confidence index are weighted and calculated to generate a warning score; The warning scores are mapped to different levels based on the threshold range to generate warning levels.

10. A floating roof tank sealing online monitoring and early warning system, applied to the floating roof tank sealing online monitoring and early warning method as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition and fusion module is used to acquire the air pressure difference data of the sealed cavity, the deformation data of the sealing strip, and the temperature field data of the sealing surface, and generate multi-source sensor data. The dynamic threshold calculation module is used to perform time series decomposition and sliding statistics on multi-source sensor data to obtain the first dynamic threshold and the second dynamic threshold, which are combined into dynamic threshold parameters. The leakage signal detection module is used to perform collaborative verification of the pressure difference data and the deformation data using the dynamic threshold parameter. When the pressure difference data and the deformation data show abnormality simultaneously, a suspected leakage signal is generated. The image acquisition module is used to acquire high-resolution infrared images in response to the suspected leakage signal and generate detailed temperature field data; The leak confirmation analysis module is used to analyze the detailed temperature field data to identify low-temperature anomaly areas, and to perform spatial correlation analysis between the low-temperature anomaly areas and the abnormal locations of the deformation data to generate a leak confirmation signal. The leak point location module is used to perform coordinate fusion based on the leak confirmation signal and the deformation data to generate leak point coordinates; The early warning information push module is used to compile information based on the coordinates of the leak point and the leak confirmation signal, generate early warning information, and push it to the user terminal.

Images