Automatic ash loading method and system for ash silo tank car

Abstract

The invention relates to the technical field of industrial automation and safety monitoring, in particular to an ash silo tank car automatic ash loading method and system, and the method comprises the following steps: obtaining image data of a tank car loading port and a surrounding area; based on a pre-established ash material optical model, processing the image data to identify a suspected ash material dust area in the image data; an airflow compensation algorithm is applied to analyze the diffusion form of the suspected ash material dust area so as to reversely calculate the position of a leakage source; comprehensively judging the severity level of the leakage event according to the position of the leakage source and the leakage amount and the leakage development trend obtained by continuously analyzing the image data; and generating and outputting a corresponding alarm signal based on the severity level. According to the invention, real ash and dust can be effectively distinguished from common on-site interferents such as water mist, steam and light and shadow changes, the false alarm rate and the missing report rate are greatly reduced, and high-sensitivity capture can be realized even for early and trace leakage.

Description

Technical Field

[0001] This application relates to the fields of industrial automation and safety monitoring technology, and more specifically, to an automatic ash loading method and system for ash storage tank trucks. Background Technology

[0002] In industries such as thermal power plants and cement production, the storage and transportation of powdery materials such as fly ash and cement are critical links. These materials are typically stored in large ash silos and loaded into specialized tank trucks using automated or semi-automated loading systems. During this process, dust leaks are highly likely to occur due to factors such as inadequate sealing of equipment interfaces, operational errors, equipment aging, or pipeline damage. These leaks not only cause material losses and severe environmental pollution, threatening the occupational health of on-site workers, but high-concentration dust clouds can also trigger major safety accidents such as dust explosions. Therefore, effective monitoring of dust leaks during the loading process is crucial.

[0003] Currently, the industry's technical solutions to this problem mainly focus on two directions. The first is physical containment and pneumatic control, which suppresses dust escape at the source, and the technology is relatively mature. However, its equipment structure is large, costly, and complex to maintain. Moreover, its design logic is passive defense, that is, to contain as much as possible, but it lacks the ability to detect, locate, and assess the severity of unexpected leaks caused by sudden events such as seal aging or accidental collisions. The second is indirect state sensing and interlocking control, which mainly uses indirect physical quantities to determine the loading status and achieve automated control of overflow. However, it is essentially an overflow detection rather than a universal leak detection; it cannot identify dust leaks that occur around the loading port during the loading process due to reasons such as misalignment of interface flanges or damage to flexible connections.

[0004] In summary, existing technologies, whether employing complex mechanical structures for physical containment or using single indirect sensors for status assessment, all share a common critical deficiency: they lack highly sensitive and accurate detection methods for complex industrial environments, especially for low-concentration, early-stage dust leaks. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the purpose of this application is to provide an automatic ash loading method for ash storage tank trucks, the specific solution of which is as follows.

[0006] An automated ash loading method for ash storage tank trucks, applied to detect ash leakage during loading, is characterized by the following steps: acquiring image data of the loading port and surrounding area of ​​the tank truck; processing the image data based on a pre-established optical model of ash to identify suspected ash dust areas in the image data; applying an airflow compensation algorithm to analyze the diffusion pattern of the suspected ash dust areas to reversely calculate the location of the leakage source; comprehensively judging the severity level of the leakage event based on the location of the leakage source and the leakage amount and leakage development trend obtained through continuous analysis of the image data; and generating and outputting a corresponding alarm signal based on the severity level to achieve real-time detection and graded early warning of ash leakage.

[0007] Optionally, acquiring image data of the tanker loading port and surrounding area specifically includes: acquiring multi-band images covering visible and non-visible light bands simultaneously or in time-division using at least one image acquisition device configured with different spectral channels; performing data fusion processing on the multi-band images to enhance the distinction between ash material and environmental interference, and improve the accuracy of identification.

[0008] Optionally, the method further includes a self-cleaning monitoring step, specifically including: continuously monitoring the contamination status of the lens surface of the image acquisition device or the signal-to-noise ratio of the acquired signal; when the contamination status or signal-to-noise ratio reaches a preset threshold, automatically triggering a cleaning mechanism linked to the image acquisition device to perform a cleaning operation, and recalibrating the detection sensitivity of the system after cleaning is completed.

[0009] Optionally, the gray material optical model is a mathematical model based on the Mie scattering, absorption and reflection characteristics of gray material within a specific electromagnetic spectrum range, used to extract the characteristic spectral or texture features of gray material from complex background images.

[0010] Optionally, the image data is two-dimensional image data. Before the reverse calculation of the location of the leak source, the method further includes: performing distortion correction and spatial mapping on the acquired two-dimensional image data based on a pre-set three-dimensional geometric model of the tanker body, loading arm, or hopper.

[0011] Optionally, the application of the airflow compensation algorithm to analyze the diffusion pattern of the suspected ash dust area in order to reversely deduce the location of the leakage source includes: combining the positive or negative pressure airflow model generated by the loading process near the tanker loading port and the external airflow field model generated by the ambient wind force to perform hydrodynamic inversion calculation on the identified dust diffusion trajectory, thereby accurately deducing the location of the leakage source from the observed center of gravity of the dust cloud.

[0012] Optionally, generating and outputting corresponding alarm signals specifically includes: classifying leakage events into at least three levels, including a prompt level for trace or instantaneous leakage, a warning level for continuous small-to-medium scale leakage, and a severe level for large-scale or rapidly expanding leakage; configuring different alarm modes for different levels of leakage events, wherein the alarm modes include, but are not limited to, audible and visual prompts, sending shutdown commands to the central control system, or triggering emergency response procedures.

[0013] Optionally, the alarm signal includes precise location information about the specific component of the leak source on the tanker or loading equipment, obtained after distortion correction and spatial mapping.

[0014] Optionally, the method further includes an ash storage environment calibration step, specifically including: real-time or periodic monitoring of environmental parameters in the loading operation area, the environmental parameters including at least ambient light intensity, temperature and air humidity; and dynamically adjusting the key parameters of the ash material optical model or the trigger threshold of the alarm signal according to changes in the environmental parameters, so as to adapt to different weather and time period detection conditions and maintain the stability of the system detection performance.

[0015] The second objective of this application is to provide an automatic ash loading system for ash storage tank trucks, the specific solution of which is as follows.

[0016] An automated ash loading system for ash storage tank trucks includes: an image acquisition unit configured to acquire multi-band image data of the tank truck loading port and surrounding area; a data processing unit connected to the image acquisition unit, configured to process the image data based on a pre-established optical model of ash material to identify suspected ash dust areas; a leak location unit configured to analyze the suspected ash dust areas using an airflow compensation algorithm to determine the true physical location of the leak source; a comprehensive judgment unit configured to comprehensively judge the severity level of the leak event based on the location of the leak source and the leakage amount and leakage development trend obtained through continuous analysis of the image data; and an intelligent alarm unit configured to generate and output a corresponding alarm signal based on the severity level to achieve real-time detection and graded early warning of ash material leaks. Compared with existing technologies, this application, by establishing a specific optical model for the target ash material and fusing multi-band image data of visible and non-visible light, can effectively distinguish between real ash dust and common on-site interferences such as water mist, steam, and changes in light and shadow, greatly reducing the false alarm rate and missed alarm rate, and achieving high-sensitivity detection even for early, minute leaks. This application innovatively introduces an airflow compensation algorithm, combining it with an on-site airflow model to perform fluid dynamics inversion calculations on the observed dust cloud diffusion morphology. It overcomes the problem of traditional visual inspection being able to see but not accurately pinpoint the source, enabling the precise tracing of passively observed, wind-borne dust clouds to their physical leak points, such as a specific flange, valve, or flexible connection. This application not only determines the presence or absence of a leak but also, through continuous monitoring and analysis of the leak volume and development trend, intelligently classifies leak events into alert, warning, and severe levels. This progressive alarm mechanism provides richer decision-making information for back-end operators and automated control systems, avoiding simplistic, one-size-fits-all shutdowns and achieving refined and differentiated management of on-site conditions, thus improving the balance between production efficiency and safety management. Attached Figure Description

[0017] Figure 1 This is a block diagram of the automatic ash loading system for ash storage tank trucks in this application.

[0018] Figure 2 This is a schematic diagram of the automatic ash loading method for ash storage tank trucks in this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be described in detail below with reference to specific embodiments. It should be understood that the specific embodiments herein are only used to explain this application and are not intended to limit the scope of protection of this application.

[0020] This application provides an automatic ash loading method and system for ash storage tank trucks. Through non-contact optical vision, it simulates an experienced safety officer, enabling the system not only to "see" dust leaks but also to "understand" the source and severity of the leak, and to make appropriate judgments and responses accordingly. Below, we will use a system deployed at a fly ash loading station in a thermal power plant as an example to elaborate on the various technical aspects of this application.

[0021] like Figure 1 As shown in the embodiment of this application, the automatic ash loading system for ash storage tank trucks is deployed in the tank truck loading area below the ash storage. The system includes an image acquisition unit, a data processing unit, a leak location unit, a comprehensive judgment unit, and an intelligent alarm unit. These units are logically separated, but can be physically integrated into one or more industrial computers.

[0022] The system's workflow follows the steps of the automated ash loading method for ash storage tank trucks, forming a continuous, closed-loop monitoring system. The image acquisition unit continuously captures high-frame-rate, multi-band video streams of the loading area while simultaneously monitoring environmental parameters in real time. The data processing unit performs preprocessing on the raw images, including noise reduction and enhancement. The leak location unit performs distortion correction and spatial mapping on the acquired two-dimensional image data based on a pre-defined three-dimensional geometric model of the tank truck body, loading arm, or hopper to eliminate visual errors caused by the observation angle and the curvature of the measured surface, ensuring location accuracy. The data processing unit analyzes the corrected images using an ash material optical model to accurately identify and segment suspected ash material dust areas in the image. The leak location unit acquires the morphology, location, and motion information of the suspected ash material dust areas and uses an airflow compensation algorithm to determine the precise location of the leak source in three-dimensional space. The comprehensive judgment unit performs time-series analysis on the analysis results of consecutive frames, calculating indicators such as leakage amount, duration, and diffusion rate to comprehensively assess the severity of the leak event. Based on the assessment level, the intelligent alarm unit generates and issues corresponding control commands or alarm signals, such as audible and visual alarms, HMI (Human Machine Interface) prompts, and sends shutdown commands to the factory's DCS (Distributed Control System) or PLC (Programmable Logic Controller).

[0023] like Figure 2 As shown, the automatic ash loading method for ash storage tank trucks in this application includes the following steps.

[0024] S01: Acquire image data of the tanker loading port and surrounding area.

[0025] Specifically, the image acquisition unit typically consists of several image acquisition devices deployed around the loading area. To achieve 360-degree monitoring of the loading port without blind spots, at least two to three image acquisition devices are usually deployed. They are mounted on specially designed columns or frames to cover the entire working area from a top-down angle, including the loading port on the top of the tanker, the loading arm, the telescopic hopper, and the connecting parts between them.

[0026] To enhance the ability to identify target ash and dust, multi-band image acquisition can be employed. Specifically, each image acquisition device can integrate two or more sensors.

[0027] Specifically, CCD (Charge-coupled Device) or CMOS (Complementary Metal-Oxide-Semiconductor) visible light cameras are used to acquire images in the visible light band (approximately 400-700 nanometers). They can provide high-resolution color or black-and-white texture information, conform to the observation habits of the human eye, and can capture the shape, color, and dynamics of dust clouds.

[0028] Non-visible light cameras, such as SWIR (Short-Wave Infrared) cameras, typically use InGaAs (Indium Gallium Arsenide) sensors, operating in the 900-1700 nm wavelength range. In this range, moisture exhibits strong absorption peaks, for example, around 1450 nm, while most minerals have relatively flat reflectance; for example, fly ash is primarily composed of SiO2 and Al2O3. This physical characteristic allows SWIR cameras to distinguish ash dust from common environmental contaminants such as water mist and vapor. In SWIR images, water mist appears deep black, while ash dust appears grayish-white, resulting in extremely high contrast.

[0029] For example, LWIR (Long-Wave Infrared) cameras, also known as thermal imagers, typically use uncooled microbolometer sensors that operate in the 8-14 micrometer wavelength range. They detect the thermal radiation emitted by the object itself. Although the temperature difference between dust and the environment may not be large, thermal imaging can serve as an auxiliary detection method in certain situations (such as when the temperature of the loaded ash material is higher or lower than the ambient temperature).

[0030] These cameras are precisely coaxially calibrated to ensure they have the same field of view and can acquire each frame synchronously, for example, by synchronizing trigger signals via hardware, thus obtaining visible and non-visible light image data of the same scene at the same time. The frame rate is typically set at 15-30 fps (Frames Per Second) to ensure that the dynamic process of dust diffusion can be captured.

[0031] After acquiring multi-band images, the data processing unit will perform data fusion processing on them. The purpose of fusion is to combine the advantageous information from different band images to generate a fused image that is richer in information and makes the target more prominent, thereby enhancing the distinction between the target dust and environmental interferences such as water mist and shadows.

[0032] Specifically, even with coaxial camera calibration, minute installation errors and optical distortions from different sensors can still cause pixel-level misalignments between images. Using registration algorithms based on feature points or mutual information, for example, can precisely align these images, ensuring that the same spatial point from different images corresponds to the same pixel coordinates.

[0033] Understandably, after registration, a pixel-level fusion algorithm is used to generate the fused image. Many algorithms are available, such as: Weighted averaging: simple and intuitive, but may lead to blurred details. IHS (Intensity-Hue-Saturation) transform: converts the color visible light image to IHS space, replaces its intensity (I) component with a high-resolution SWIR image, and then converts it back to RGB space. This preserves visible light color information while injecting detail and contrast from the SWIR image. Wavelet transform: a more advanced multi-scale analysis method. It decomposes each source image into sub-bands of different frequencies and directions. The fusion rule can be designed to average on the low-frequency sub-bands (representing contour information) and select the coefficient with the larger absolute value on the high-frequency sub-bands (representing detail and edge information). This method effectively preserves the salient features of each source image, resulting in a clear and high-contrast fused image.

[0034] For example, in this embodiment, the visible light image shows a grayish-white cloud of smoke, making it difficult to determine whether it is water vapor or dust. Simultaneously, in the synchronously acquired SWIR image, this area appears as a bright white. The data processing unit fuses these two images using wavelet transform. In the resulting fused image, this area is highlighted and retains the fine texture of the visible light image. Based on this, the system can determine with high confidence that it is gray dust, rather than water vapor, because water vapor would appear black in the SWIR image, and the fused area would appear darker.

[0035] In this way, multi-band fusion technology greatly improves the input quality of subsequent recognition algorithms, which is the first key guarantee for achieving high-accuracy detection.

[0036] S02: Based on a pre-established optical model of ash material, process the image data to identify suspected ash material dust areas in the image data.

[0037] Understandably, the optical model for fly ash is not a universal model, but rather a customized model for fly ash from a specific source. For example, fly ash from a specific boiler in a power plant may have different particle size distribution, shape, color, and chemical composition due to different compositions and combustion processes, resulting in different optical properties.

[0038] Specifically, this optical model of the ash material is a mathematical model based on the Mie scattering, absorption, and reflection characteristics of the target ash material within a specific electromagnetic spectrum range. Its construction process is typically completed in a controlled laboratory and includes the following steps.

[0039] First, representative ash samples were collected from the actual production line. The particle size distribution was precisely measured using equipment such as a laser particle size analyzer; the particle morphology was observed using a scanning electron microscope, which generally showed spherical or irregular shapes; and the chemical composition, such as the content of SiO2, Al2O3, and Fe2O3, was analyzed using methods such as X-ray fluorescence spectroscopy.

[0040] Next, the complex refractive index n+ik of the ash material is measured in the wavelength range of interest, such as 400-1700 nm, where the real part n represents the scattering ability and the imaginary part k represents the absorption ability. This is the most critical input parameter in the Mie scattering theory calculation.

[0041] Understandably, fly ash particles typically have diameters between 1 and 100 micrometers, a scale comparable to the wavelengths of visible and short-wave infrared light. Therefore, the physical laws describing the interaction between light and these particles are based on Mie scattering theory. This theory can accurately calculate the scattered light intensity, extinction cross section, and absorption cross section in various directions when light of a given wavelength is incident on a single spherical particle. By substituting the measured particle size distribution and complex refractive index into the Mie scattering theory formulas, the macroscopic optical properties of a dust cloud composed of numerous ash particles can be calculated, such as its backscattering coefficient and extinction coefficient at different wavelengths. These calculations form the core of the model: the spectral characteristics.

[0042] In addition to spectral features, the grayscale optical model also includes a description of the texture features of the dust cloud. Real dust clouds typically exhibit soft, diffuse edges and dynamic, swirling, flocculent textures within. These texture features can be extracted by analyzing a large number of images of real-world leakage scenes. Commonly used texture descriptors include gray-level co-occurrence matrix (GLCM) and local binary patterns (LOBs). These features are quantized and stored in the model.

[0043] Ultimately, the optical model of the grey material is a multi-dimensional, multi-level database or parameter set that contains the unique spectral fingerprint and texture morphology of the target grey material.

[0044] In actual monitoring, the data processing unit applies this pre-established model to the fused image data acquired in real time. The recognition process is as follows.

[0045] First, a dynamic background model is established, representing a normal scene without leakage. Then, using algorithms such as Gaussian mixture model or ViBe (Visual Background Extractor), the real-time image is compared with the background model to initially identify foreground regions that have changed.

[0046] For each foreground region, calculate the spectral and texture features of its internal pixels; for example, the spectral features include the average intensity ratio of the visible light and SWIR channels, and the texture features include the LBP histogram.

[0047] The extracted feature vectors are input into a pre-trained classifier, which can be, for example, a support vector machine or a convolutional neural network in deep learning. This classifier has been trained using a dataset containing a large number of positive and negative samples. Positive samples include real gray dust images, while negative samples include images of interference such as water mist, shadows, birds, and moving people. The classifier's task is to determine whether the features of the foreground region match the features defined by the gray dust optical model.

[0048] Once an area is classified as ash dust with high confidence, an image segmentation algorithm such as region growing or GrabCut is used to accurately outline the dust cloud and generate a binary mask image, in which the dust area is white and the background is black; this binary mask image is the accurate representation of the suspected ash dust area.

[0049] Through the above processing, the actual ash dust can be accurately identified from the complex dynamic background, laying a solid foundation for subsequent positioning and analysis.

[0050] Because the camera observes from a fixed angle, while the tanker's surface is curved, the original image suffers from perspective distortion and surface distortion. Therefore, it is necessary to solve the mapping problem between the two-dimensional image and the three-dimensional physical world.

[0051] Therefore, before reverse-engineering the location of the leak source, it is necessary to perform distortion correction and spatial mapping on the acquired two-dimensional image data based on the pre-set three-dimensional geometric structure model of the tanker body, loading arm or hopper, in order to eliminate visual errors caused by the observation angle and the curvature of the measured surface and ensure positioning accuracy.

[0052] Specifically, the first step is to create accurate 3D geometric models of key fixed facilities and moving objects within the monitored area. Key fixed facilities include loading arms and hoppers, while moving objects include standard tank trucks. These models can be obtained using specialized 3D scanning equipment such as laser scanners, or directly converted from the equipment's CAD drawings. These models are stored in the form of point clouds or triangular meshes.

[0053] After installing the camera, a one-time camera calibration is required. The calibration process aims to determine the camera's intrinsic and extrinsic parameters. Intrinsic parameters include focal length, principal point coordinates, and distortion coefficients, while extrinsic parameters include the camera's position and orientation in the world coordinate system. Once calibration is complete, a precise mathematical transformation relationship is established from three-dimensional world coordinates to two-dimensional image pixel coordinates, i.e., the projection matrix.

[0054] In real-time monitoring, for each pixel in an image identified as dust, if it happens to coincide with a surface in a 3D model—for example, dust just emerging from a tank surface—its coordinates in the 3D world can be calculated via backprojection using calibrated camera parameters and the 3D model. For dust clouds suspended in the air, their line-of-sight vector in 3D space can be calculated. In this way, the entire two-dimensional dust region is transformed into 3D space, eliminating visual errors and providing accurate geometric input for subsequent fluid dynamics calculations.

[0055] S03: Apply an airflow compensation algorithm to analyze the diffusion pattern of the suspected ash dust area in order to reverse-calculate the location of the leakage source.

[0056] It is understandable that the location of the dispersed dust cloud is not the same as the location of the leak source. Wind and loading airflow can cause the dust cloud to deviate from its source. Therefore, this application introduces an airflow compensation algorithm to analyze the diffusion pattern of the dust in order to reverse-calculate the true physical location of the leak source.

[0057] Specifically, the airflow compensation algorithm includes the following steps.

[0058] First, a mathematical model describing the airflow distribution in the loading area is established, namely the airflow field model. This airflow field model is dynamic and consists of the superposition of the internal airflow field and the external airflow field.

[0059] The internal airflow field is generated by the loading process itself. In positive pressure conveying, an outward blowing airflow forms near the loading port; in negative pressure suction, an inward suction airflow forms. The shape and intensity of the internal airflow field can be pre-simulated using CFD (Computational Fluid Dynamics) based on process parameters of the loading system, such as fan pressure and pipe diameter, and simplified into a parameterized model. The external airflow field is generated by the ambient wind, and the speed and direction of the ambient wind are acquired in real time using a three-dimensional ultrasonic anemometer installed on-site. By vector superimposing the internal and external airflow fields, the comprehensive airflow velocity vector V(x, y, z, t) at any time and any spatial point can be obtained.

[0060] Then, fluid dynamics inversion calculations are performed. Given the three-dimensional dust cloud C(t) observed at time t, with its centroid position, shape, and size known, and the airflow field model V(x, y, z, t), the most probable leak source location S(x0, y0, z0) and leak initiation time t are solved. This is a typical reverse source tracing problem, and an accurate solution is very complex. However, in engineering practice, several simplified and effective methods can be used, such as the centroid reverse trajectory method, Gaussian plume model inversion, or probabilistic source tracing methods.

[0061] Specifically, the centroid-inverse trajectory method calculates the centroid position P(t) of the current dust cloud. Then, based on the airflow field model, starting from P(t), a reverse integration is performed in time along the opposite direction of the airflow velocity (-V), i.e., P(t-Δt) = P(t) - V(P(t)) * Δt. This process is iterated until the trajectory intersects the surface of the 3D model of the tanker or loading equipment. This intersection point is the estimated leak source location S.

[0062] For persistent small leaks, the resulting dust cloud can be approximated by a Gaussian plume model, which links the source strength, meteorological conditions, and concentration at any point downwind. The model parameters are solved in reverse by fitting the observed dust cloud concentration distribution, thereby determining the source location and source strength.

[0063] The probabilistic source tracing method constructs a probabilistic graphical model in which each potential leak point in the scene is regarded as a possible source. For each possible source, the dust cloud morphology that it would produce under the current airflow conditions is simulated in a forward manner. Then, the simulated morphology is compared with the actual observed morphology. For example, using an image similarity index, the potential leak point with the highest matching degree is selected as the final source tracing result.

[0064] With the addition of airflow compensation, the alarm information output by the system will have unprecedented accuracy. Moreover, the alarm signal can also include the precise location information of the leak source on the specific component of the tanker or loading equipment, obtained after distortion correction and spatial mapping.

[0065] For example, alarm messages are no longer vague statements about leak detection, but rather structured, machine-readable, and human-understandable detailed reports, such as: { "Event ID": "L20250801-001", "Severity Level": "Warning Level", "3D Coordinates of Leak Source": "[10.5, 3.2, 4.8]", "Leak Source Description": "Loading position 2, west side of the manhole cover seal on the top of the tank truck", "Confidence Level": "92%"}. This type of alarm message can be directly highlighted in the factory's digital system or 3D monitoring screen, guiding maintenance personnel to the accurate location for immediate action, greatly improving the efficiency and effectiveness of emergency response.

[0066] S04: Based on the location of the leak source and the amount and trend of leakage obtained by continuous analysis of the image data, the severity level of the leak event is comprehensively judged.

[0067] After detection and location, it is also necessary to assess the severity of the problem. This application utilizes a comprehensive judgment unit to comprehensively determine the severity level of the leakage event and generate and output corresponding alarm signals, thus achieving intelligent decision-making.

[0068] Understandably, the severity level is not determined by a single-frame snapshot, but rather by a time-series analysis of continuous image data. The system continuously tracks and calculates key indicators such as leakage amount, leakage duration, leakage development trend, and leakage source location.

[0069] Specifically, the leakage amount can be estimated using the area identified as dust, the average grayscale value of pixels or the blended intensity value, and its volume in three-dimensional space. This value can be calibrated to correlate with the actual dust concentration (mg / m³). The leakage duration refers to the time from when the dust is first detected until it disappears. The leakage development trend is the first and second derivatives of the leakage amount with respect to time, i.e., the growth rate and acceleration, reflecting whether the leakage is worsening, stabilizing, or decreasing. The risk of leakage at certain critical locations, such as main conveyor pipelines, is much higher than at other locations; therefore, the location of the leakage source is also important.

[0070] S05: Based on the severity level, generate and output a corresponding alarm signal to achieve real-time detection and graded early warning of ash leakage.

[0071] Specifically, the intelligent alarm unit maintains a state machine or rule engine internally, which determines the severity level based on the combination of these indicators, classifies leakage events into at least three levels, and configures different alarm modes; leakage events include the alert level for minor or instantaneous leakage, the warning level for continuous small to medium-sized leakage, and the severe level for large-scale or rapidly expanding leakage.

[0072] Specifically, the trigger condition for the alert level is the detection of a minor or transient leak. For example, a dust cloud area smaller than threshold A1 and a duration less than T1, or a momentary, small amount of powdered material being shaken off. The alarm mode is a silent alarm; a pop-up message appears on the operator's HMI interface, or a log entry is recorded in the event list. Its purpose is to alert the operator without interrupting production.

[0073] The warning level is triggered by the detection of a persistent small-to-medium-scale leak. For example, the dust cloud area is consistently larger than threshold A1 but smaller than A2, and the duration exceeds T1; or the rate of increase in the leak exceeds threshold R1. This typically corresponds to issues requiring attention and handling, such as aging seals or loose connections. The alarm mode is an audible and visual alarm; it triggers a flashing yellow warning light and a buzzer in the control room, sending a warning signal to the DCS system. The system can automatically reduce the loading rate by 50%, and the HMI will display detailed leak location and information in a prominent color; simultaneously, it can be configured to send alarm SMS messages to the mobile phones of relevant management personnel.

[0074] The severe level alarm is triggered by the detection of a large-scale or rapidly expanding leak. For example, the dust cloud area instantaneously or continuously exceeds a very high threshold A2, or the rate of increase in the leak exceeds the danger threshold R2. This typically corresponds to major malfunctions such as pipe rupture or valve jamming. The alarm mode is the highest level emergency alarm, triggering a piercing alarm sound and a red flashing warning light; it immediately sends an emergency shutdown command to the PLC or DCS, forcibly closing all relevant valves, stopping the fan operation, and cutting off material transport. Simultaneously, it automatically initiates emergency response procedures, such as a dust suppression spray system.

[0075] It is understandable that the dust-filled environment of the ash storage facility will inevitably contaminate the camera lenses, leading to a decrease in image quality and affecting detection performance. Therefore, this application includes a self-cleaning monitoring step.

[0076] The self-cleaning monitoring unit continuously monitors the contamination status of the lens surface using either a reference area-based method or a signal-to-noise ratio-based method.

[0077] Specifically, within the camera's field of view, a stationary object with clear texture is selected as a reference area, such as a mark on a wall or the head of a bolt. The system continuously calculates the image sharpness and contrast of this reference area, for example, using the Laplacian operator or gradient energy. When these metrics drop below a preset threshold (e.g., 20%), the lens is deemed contaminated.

[0078] It can also analyze the signal statistical characteristics of the entire image. Lens contamination usually introduces a uniform or non-uniform "haze" effect, which leads to a decrease in the overall signal-to-noise ratio of the image. When the signal-to-noise ratio is lower than a preset threshold, it can also be judged as contamination.

[0079] Once lens contamination is detected, a cleaning mechanism linked to the device can be automatically triggered to perform a cleaning operation. This cleaning mechanism can be a miniature electric wiper installed outside the camera housing's window, which automatically wipes the window surface upon receiving a command; or it can be a compressed air nozzle, commonly known as an air knife, aimed at the window. Upon receiving a command, a solenoid valve opens, spraying out a strong, clean burst of compressed air to blow away the attached dust.

[0080] After the cleaning process is complete, the detection sensitivity can be recalibrated. This involves re-analyzing the sharpness of the reference area to confirm that it has returned to normal levels. Then, the background model is updated, and certain parameters in the gray material optical model are fine-tuned to accommodate the more transparent imaging conditions after cleaning.

[0081] Besides lens contamination, changes in the external environment can also have a significant impact on optical inspection systems. This application proactively adapts to these changes through a gray-box environment calibration procedure, specifically including the following steps.

[0082] The environmental parameters of the loading operation area are monitored in real time or periodically by connected external sensors. These parameters include at least ambient light intensity, temperature, and air humidity. Ambient light intensity is measured by a illuminance meter installed on-site; temperature and air humidity are measured by temperature and humidity sensors. Based on changes in these environmental parameters, the key parameters of the ash material optical model or the trigger threshold of the alarm signal are dynamically adjusted.

[0083] When there are drastic changes in light intensity, such as from sunny to cloudy, or from daytime to nighttime artificial lighting, the camera's exposure time, gain, and white balance are automatically adjusted. More importantly, it may call upon subsets of spectral features stored in the grayscale optical model for different lighting conditions. For example, in low light conditions, image noise increases, and the system can appropriately relax the matching requirements for texture features.

[0084] When air humidity increases, especially to conditions conducive to water mist formation, the system enters a high-humidity mode. This significantly increases the reliance on SWIR non-visible light channel data, as SWIR best distinguishes between water mist and dust. Simultaneously, it may slightly raise the alarm trigger threshold to further suppress potential false alarms caused by residual water mist interference. Temperature changes affect air density, thus subtly influencing airflow field calculations. Measured temperature values ​​can be used as one of the input parameters for the CFD model for more accurate compensation.

[0085] In summary, this application fundamentally solves the problem that existing technologies cannot provide real-time, accurate location and categorized early warning of dust leakage during the loading process of ash storage tank trucks. This method and system can significantly improve the safety, environmental protection level, and automated management capabilities of bulk powder loading operations, and has extremely high industrial application value.

[0086] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional modules is used as an example. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device, and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0087] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0088] 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0089] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0090] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0091] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. An automatic ash loading method for ash storage tank trucks, used to detect ash leakage during loading, characterized in that, Includes the following steps: Acquire image data of the tanker loading port and surrounding area; Based on a pre-established optical model of ash particles, the image data is processed to identify suspected ash particle dust areas in the image data; An airflow compensation algorithm is applied to analyze the diffusion pattern of the suspected ash dust area in order to reverse-calculate the location of the leakage source. The severity level of the leakage event is comprehensively determined based on the location of the leakage source, the leakage amount obtained through continuous analysis of the image data, and the leakage development trend. Based on the severity level, a corresponding alarm signal is generated and output to achieve real-time detection and graded early warning of ash leakage.

2. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The acquisition of image data of the tanker loading port and surrounding area specifically includes: Multi-band images covering visible and non-visible light bands are acquired synchronously or in time-division using at least one image acquisition device configured with different spectral channels. The multi-band images are subjected to data fusion processing to enhance the distinction between gray material and environmental interference, thereby improving the accuracy of identification.

3. The automatic ash loading method for ash storage tank trucks according to claim 2, characterized in that, The method also includes a self-cleaning monitoring step, specifically including: Continuously monitor the surface contamination of the lens of the image acquisition device or the signal-to-noise ratio of the acquired signal; When the pollution level or signal-to-noise ratio reaches a preset threshold, the cleaning mechanism linked to the image acquisition device is automatically triggered to perform a cleaning operation, and the detection sensitivity of the system is recalibrated after the cleaning is completed.

4. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The optical model of the gray material is a mathematical model based on the Mie scattering, absorption and reflection characteristics of the gray material in a specific electromagnetic spectrum range. It is used to extract the characteristic spectral or texture features of the gray material from complex background images.

5. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The image data is two-dimensional image data, and before the reverse calculation of the location of the leak source, it also includes: Based on a pre-defined three-dimensional geometric model of the tanker body, loading arm, or hopper, the acquired two-dimensional image data is subjected to distortion correction and spatial mapping.

6. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The application of the airflow compensation algorithm analyzes the diffusion pattern of the suspected ash dust area to reverse-calculate the location of the leak source, specifically including: By combining the positive or negative pressure airflow model generated by the loading process near the loading port of the tanker truck, and the external airflow field model generated by the ambient wind, the identified dust diffusion trajectory is subjected to hydrodynamic inversion calculation, thereby accurately deducing the location of the leakage source from the observed center of gravity of the dust cloud.

7. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The generation and output of the corresponding alarm signal specifically includes: Leakage incidents are classified into at least three levels, including an alert level for minor or transient leaks, a warning level for persistent small to medium-sized leaks, and a severe level for large-scale or rapidly escalating leaks. Different alarm modes are configured for different levels of leakage events. The alarm modes include, but are not limited to, audible and visual alerts, sending shutdown commands to the central control system, or triggering emergency response procedures.

8. The automatic ash loading method for ash storage tank trucks according to claim 7, characterized in that: The alarm signal includes precise location information about the specific component of the leak source on the tanker or loading equipment, obtained after distortion correction and spatial mapping.

9. The automatic ash loading method for ash storage tank trucks according to claim 1, characterized in that, The method also includes a gray-market environment calibration step, specifically including: The environmental parameters of the loading operation area are monitored in real time or periodically, and the environmental parameters include at least ambient light intensity, temperature and air humidity; Based on changes in the environmental parameters, the key parameters of the ash optical model or the trigger threshold of the alarm signal are dynamically adjusted to adapt to different weather and time periods and maintain the stability of the system's detection performance.

10. An automatic ash loading system for ash storage tank trucks, characterized in that, include: The image acquisition unit is configured to acquire multi-band image data of the tanker loading port and surrounding area; A data processing unit, connected to the image acquisition unit, is configured to process the image data based on a pre-established optical model of ash particles in order to identify suspected ash particle dust areas. The leak location unit is configured to apply an airflow compensation algorithm to analyze the suspected ash and dust area in order to determine the true physical location of the leak source. The comprehensive judgment unit is configured to comprehensively judge the severity level of the leakage event based on the location of the leakage source and the leakage amount and leakage development trend obtained by continuously analyzing the image data; The intelligent alarm unit is configured to generate and output a corresponding alarm signal based on the severity level, so as to realize real-time detection and graded early warning of ash leakage.

Images