Online real-time visual monitoring system and method for blockage of air pre-heater

Abstract

The invention belongs to the field of energy conservation and emission reduction of coal-fired boilers, and particularly relates to an online real-time visual monitoring system and method for blockage of an air pre-heater, the system comprises an industrial personal computer, a PLC, an industrial camera, a shield, a light supplementing lamp and a cross rod, the cross rod is installed at the cold end of a secondary air bin of the air pre-heater, and the industrial camera, the light supplementing lamp and the shield are all installed on the cross rod. The industrial camera is located in the shield, the light supplementing lamp is located beside the shield, the PLC is in signal connection with the industrial camera and the light supplementing lamp, along with rotation of the rotor, the industrial camera shoots images of all positions of the rotor of the air pre-heater and transmits collected image information of a cold end to the industrial personal computer, the industrial personal computer preprocesses and recognizes the images, and the air pre-heater is started. And the acquired image is processed and analyzed to obtain the blockage condition of the air pre-heater. The cost of main equipment of the system is low. Compared with an overall pressure difference monitoring method, visual monitoring is more visual; compared with a visual monitoring system based on an infrared image, the system is less interfered by the environment, and the system is more stable and reliable.

Description

Technical Field

[0001] The present invention belongs to the field of energy conservation and emission reduction of coal-fired boilers, and specifically relates to an online real-time visualization monitoring system and method for air preheater blockage. Background Art

[0002] An air preheater is a heat exchange device installed in the tail flue of a boiler, which uses the waste heat of flue gas to preheat the required air. It can make full use of the waste heat in the flue gas, reduce the heat loss of flue gas exhaust, and thus improve the thermal efficiency of the boiler. At the same time, preheating the air helps to lower the ignition temperature, improve the ignition and combustion conditions of pulverized coal air flow, and reduce the heat loss caused by incomplete combustion. In order to improve the heat exchange efficiency and facilitate the layout of the flue, the heat storage elements inside the air preheater adopt corrugated plates with a compact layout, and the flow channels are relatively narrow, making them prone to blockage. When the SCR system over-injects ammonia, the ammonia content in the flue gas increases, and it may react with sulfur trioxide in the flue gas to form highly viscous ammonium bisulfate. When the temperature drops, these substances will condense on the surface of the corrugated plates of the air preheater and adsorb fly ash in the flue gas, resulting in blockage of the heat storage elements. This not only reduces the heat exchange efficiency of the air preheater but also increases the resistance of the flow channel, leading to an increase in fan energy consumption, and may cause phenomena such as fan surge and stall. In severe cases, even shutdown for cleaning is required. Therefore, analyzing the causes of air preheater blockage, accurately monitoring the degree of ash accumulation, and timely purging to prevent ash hardening and excessive blockage are the keys to ensuring the safe and economic operation of the boiler.

[0003] At present, the monitoring of the ash deposition situation of air preheaters in power plants mainly focuses on obtaining their internal operating status. Existing research and practice mainly rely on methods such as numerical simulation and flow resistance monitoring. Through numerical simulation, the temperature distribution inside the air preheater can be obtained, thereby predicting the easily blocked parts. However, existing numerical simulations are mostly limited to the clean state, and the evolution law after blockage is not clear, and the working conditions are prone to deviate from the actual state, making it difficult to judge the real-time blockage situation. The flow resistance monitoring method is based on existing operating monitoring data such as flue gas pressure difference and flow rate, and analyzes the development of the overall blockage degree of the air preheater, with good real-time performance. In the specific implementation process, aiming at the problems of inaccurate pressure measurement and large deviation, most existing technologies eliminate measurement errors through data preprocessing methods. There are mainly two ways to quantify the blockage degree: the first is to use the load and excess air coefficient instead of the flow rate, and use the ratio of the measured pressure difference to the corresponding standard state pressure difference to characterize the blockage degree; the second is to calculate the real-time heat transfer coefficient using the inlet and outlet parameters of the air preheater, and reflect the internal ash deposition degree through the ratio of the real-time heat transfer coefficient to the design value. These methods are limited by the small number of measuring points such as temperature and pressure, and the monitoring data has large errors, making it difficult to accurately characterize the blockage state, and can only give the overall blockage degree, and the spatial distribution of the internal blockage of the air preheater cannot be obtained. Therefore, it is necessary to enrich the monitoring means to accurately monitor the blockage growth rate and blockage distribution of the air preheater to guide targeted soot blowing and cleaning, and improve the operating safety of the air preheater. At present, the visual monitoring method of air preheaters mostly uses an infrared video monitoring system to obtain internal infrared videos using infrared supplementary lights and CCD image sensors. However, due to the presence of a large amount of dust and flue gas inside the air preheater during hot-state operation, the infrared images have defects such as high noise, low contrast, strong non-uniformity, and poor spatial resolution. Even after enhancing the low illuminance of the image through the Retinex algorithm, the infrared imaging pictures still have many defects compared with ordinary images, and the blockage situation cannot be accurately quantitatively described.

[0004] The prior art is as follows: A Chinese invention patent application with the patent publication number CN109442469A and the title "A Visual State Monitoring Device and Method for Air Preheaters in Thermal Power Plants" discloses a method for monitoring the thermal state and predicting corrosion of high-temperature heating surfaces in waste incinerators, belonging to the technical field of monitoring the thermal state of high-temperature heating surfaces in waste incinerators. The method includes the following processes: calculating the working medium temperature based on the thermal calculation of heat transfer characteristics by partition, obtaining the flue gas temperature and flow rate through three-dimensional numerical simulation, coupling the above two results to obtain the wall temperature of the pipe wall, calculating for typical working conditions, storing the calculation results in a database, and calling the database according to the XGBoost model and calculating to obtain the real-time flue gas temperature, wall temperature, fly ash velocity and concentration, realizing the monitoring of the thermal condition of the high-temperature heating surface and implementing over-temperature alarm. This patent detects the ash accumulation situation of the air preheater based on infrared image analysis, and determines the coordinates of the characteristic abnormal area according to the proportional relationship between the infrared image and the actual operating state, and uses median filtering preprocessing and an improved region growing algorithm for analysis. However, infrared imaging is easily affected by the environment, and the dust and flue gas during the hot operation of the air preheater will cause the infrared image to have large noise, low contrast, strong non-uniformity and poor spatial resolution, affecting the detection reliability.

[0005] And a Chinese invention patent application with the patent publication number CN115219732A and the title "A Visual Localization Method and System for a Rotary Air Preheater" discloses a visual localization method and system for a rotary air preheater. The method includes attaching a marker to the rotary air preheater, and using a monitoring device to take a working image of the rotating rotary air preheater, calculating the real-time rotation speed ω_real of the rotary air preheater according to the working image with the marker, using the latest shooting time in the working image with the marker as the positioning reference time t1, using the time when the temperature of the rotary air preheater is measured as the measurement time t2, and positioning the temperature measurement position of the rotary air preheater according to the positioning reference time t1, the measurement time t2 and the real-time rotation speed ω_real. This patent uses an industrial endoscope lens to insert into the monitoring hole of the air preheater to obtain internal images, and evaluates the degree of blockage in combination with the temperature data measured by an infrared temperature sensor. However, this method mainly relies on single-point or local monitoring, and does not provide a quantitative description method for the degree of blockage. Summary of the Invention

[0006] Aiming at the deficiencies of the existing monitoring means, an online real-time visual monitoring system and method for air preheater blockage are proposed by using an industrial camera to monitor the blockage situation of the air preheater.

[0007] To achieve the above effects, the technical solutions adopted by the present invention are as follows:

[0008] An online real-time visual monitoring system for air preheater blockage, including a data storage module, a human-machine interaction interface, an alarm system, an industrial control computer, a PLC, an industrial camera, a protective cover, a supplementary light, and a crossbar. The crossbar is installed at the cold end of the secondary air chamber of the air preheater. The industrial camera, the supplementary light, and the protective cover are all installed on the crossbar. The industrial camera is located inside the protective cover, and the supplementary light is located beside the protective cover. The PLC is respectively connected to the industrial camera and the supplementary light in signal. As the rotor rotates, the industrial camera takes images of each position of the air preheater rotor, and transmits the collected image information of the cold end to the industrial control computer. The industrial control computer preprocesses and identifies the images, and processes and analyzes the obtained images to obtain the blockage situation of the air preheater;

[0009] The online real-time visual monitoring system for air preheater blockage also includes a data storage module: used to store the original images, processing results, and historical data to support long-term trend analysis and model optimization;

[0010] The human-machine interaction interface: provides a visual interface, displays the monitoring results, and allows users to manually view images, adjust parameters, or receive alarm information;

[0011] The alarm system: when blockage is detected, triggers an audible and visual alarm or sends an alarm signal to the control center.

[0012] Furthermore, the industrial camera shoots the images of the cold end vertically upward. The angular velocity, exposure time t, and shooting frequency f should meet the following requirements:

[0013] First, obtain the following parameters: rotor speed: ω, unit: r / min; rotor radius: R, unit: meter; industrial camera resolution: N, unit: pixel / meter; allowable maximum smear: Δp, unit: pixel;

[0014] Then calculate the angular velocity as:

[0015]

[0016] The exposure time t should meet:

[0017]

[0018] The shooting frequency f should meet:

[0019]

[0020] Furthermore, each industrial camera is equipped with two supplementary lights, and the two supplementary lights are both placed obliquely relative to each other.

[0021] Furthermore, the dust-proof protective cover is installed vertically upward towards the cold end of the air preheater. A window and a wiper are provided at the upper part of the dust-proof protective cover. The lens of the industrial camera is installed closely against the inner side of the window, and the wiper is installed on the outer side of the window. The wiper is connected to the PLC.

[0022] An online real-time visualization monitoring method for air preheater blockage, comprising the following steps:

[0023] Step 1. Control the shield wiper to work through the PLC to clean the shield window;

[0024] Step 2. Control the wiper to stop working through the PLC, the fill light starts to illuminate the shooting object, and the industrial camera collects images of the cold end of the air preheater rotor;

[0025] Further, in Step 2, the illumination intensity range of the fill light is 5000–15000 lux, and the following parameters are obtained: rotor speed: ω, unit: r / min; rotor radius: R, unit: m; industrial camera resolution: N, unit: pixel / m; allowable maximum smear: Δp, unit: pixel;

[0026] Then calculate the angular velocity as:

[0027]

[0028] The exposure time t should satisfy:

[0029]

[0030] The shooting frequency f should satisfy:

[0031]

[0032] Step 3. The industrial camera images are transmitted at high speed and stored in the industrial control computer;

[0033] Step 4. Enhance the images and perform preprocessing to screen the ROI area;

[0034] Still further, in Step 4, the Retinex algorithm is used to enhance the images, and the images are decomposed into a reflection component and an illumination component, including the following steps:

[0035] Step 4.1.1. Image decomposition;

[0036] I(x,y) = R(x,y) × L(x,y);

[0037] Where I(x,y) is the intensity value of the original image at the pixel point (x,y); R(x,y): reflection component, representing the essential color of the object; L(x,y): illumination component, representing the illumination condition;

[0038] Step 4.1.2. Logarithmic transformation;

[0039] log(I(x,y)) = log(R(x,y)) + log(L(x,y));

[0040] Step 4.1.3. Estimate the illumination component;

[0041] log(L(x,y)) = log(I(x,y)) * G(x,y);

[0042] where G(x,y) is a Gaussian filter used to smooth the image. The formula for the Gaussian filter is as follows,

[0043]

[0044] σ represents the standard deviation of the Gaussian distribution;

[0045] Step 4.1.4. Calculate the reflection component;

[0046] log(R(x,y)) = log(I(x,y)) - log(L(x,y));

[0047] R(x,y) = exp(log(I(x,y)) - log(L(x,y)));

[0048] Step 4.1.5. Reconstruct the enhanced image;

[0049] I'(x,y) = R'(x,y) × L(x,y);

[0050] where the calculation process of R'(x,y) is as follows:

[0051] R′(x, y) = α·R(x, y) + β

[0052] where α is the adjustment factor and β is the offset;

[0053] where I'(x,y) is the enhanced image and R'(x,y) is the adjusted reflection component.

[0054] Furthermore, the specific process of preprocessing to screen the ROI region is as follows:

[0055] Adopt the method of manual marking and neural network training. The steps are as follows:

[0056] Step 4.2.1. Manually annotate the ROI training set;

[0057] Manually mark the ROI region on some training images to generate a binary mask M(x,y), which is defined as follows:

[0058]

[0059] Step 4.2.2. Neural network model training;

[0060] Use a convolutional neural network or a segmentation model based on the Transformer structure, and train with the enhanced image R'(x, y) as the input and the ROI mask M(x, y) as the supervision signal;

[0061] Training objective: Learn the mapping relationship for automatically predicting the ROI region from the enhanced image:

[0062]

[0063] where f θ is the neural network model with parameters θ; is the ROI result predicted by the model;

[0064] Step 4.2.3. ROI screening and post-processing;

[0065] The ROI prediction results obtained through neural network inference may contain noise and are optimized through morphological operations:

[0066]

[0067] After the final ROI region is extracted, it can be used for subsequent object detection or analysis.

[0068] Step 5. Segment the image through the region growing algorithm, and perform contour extraction and regional area statistics on the segmented image to obtain the blocked area and the flowing area.

[0069] Furthermore, in Step 5, the steps of image segmentation by the region growing algorithm are as follows:

[0070] Step 5.1.1. Select seed points;

[0071] Set one or more seed points S;

[0072] Step 5.1.2. Set growth conditions;

[0073] Calculate the similarity between the pixel to be expanded and the pixels in the selected region, and the growth condition is the gray difference or color difference:

[0074] D(x, y) = |I(x, y) - I(s x , s y )|

[0075] where D(x, y) is the gray or color difference between the pixel to be expanded and the seed point; I(x, y) is the gray value or color value of the pixel to be expanded; I(s x , s y ) is the gray value or color value of the seed point;

[0076] If D(x, y) < T, where T is a preset threshold, then this pixel is classified into the current region; otherwise, stop the expansion.

[0077] Step 5.1.3. Region expansion;

[0078] Perform similarity judgment on all adjacent pixels, and add the pixels that meet the conditions to region R. Repeat this process until all pixels are classified or there are no eligible pixels to add.

[0079] Step 5.1.4. Termination conditions;

[0080] The conditions for terminating region growth are: 1) no new pixels meet the growth criterion; 2) the preset maximum region size is reached; 3) the preset similarity threshold is reached.

[0081] Finally, the entire image is divided into multiple regions, and each region corresponds to a different feature part.

[0082] Furthermore, in Step 5, the contour extraction is specifically as follows:

[0083] The contour of the segmented image is extracted through edge detection or morphological processing:

[0084]

[0085] where G(x, y) is the gradient magnitude at the pixel point; G x and G y are the gradients in the horizontal and vertical directions respectively.

[0086] After that, the final contour is obtained through non-maximum suppression, double-threshold processing, and edge tracking.

[0087] Furthermore, in Step 5, the calculation of the region area is specifically as follows:

[0088] The region area is calculated by counting the number of pixels inside the contour:

[0089]

[0090] where A is the region area; R is the target region; (x, y) are the pixel coordinates within the region.

[0091] Furthermore, in Step 5, the calculation of the blocked region area and the flowing region area is specifically as follows:

[0092] The two regions inside and outside the contour in the segmented image represent the blocked region and the flowing region respectively, and the area calculation is as follows:

[0093] (1) Blocked region area

[0094]

[0095] Among them, A block is the area of the blocked area; R block is the set of pixels in the blocked area;

[0096] Area of the flow-through area

[0097]

[0098] Among them, A flow is the area of the flow-through area; R flow is the set of pixels in the flow-through area.

[0099] Step 6. Calculate the proportion of the blocked area in the total area to obtain a quantitative index characterizing the blockage degree of the air preheater rotor.

[0100] Furthermore, the specific content of the said Step 6 is:

[0101]

[0102] A total = A block + A flow ;

[0103] Among them, P block is the blockage ratio; A total is the total area of the image; A block is the area of the blocked area.

[0104] The advantages of this application are as follows:

[0105] In this application, a plurality of dust-proof industrial cameras and corresponding supplementary lighting devices are arranged radially in the secondary air chamber at the cold end of the air preheater to capture the blockage morphology at the cold end of the air preheater; furthermore, the image segmentation algorithm is used to obtain the blocked area within the image area, that is, the blockage degree at each radial position. The system hardware layout method and information transmission structure are as shown in the attached Figure 1 and Figure 2 figures. The main equipment cost of this system is relatively low. Compared with the overall differential pressure monitoring method, the visual monitoring is more intuitive; compared with the visual monitoring system based on infrared images, it is less affected by the environment, and the system is more stable and reliable.

[0106] The present invention collects visible light images of each position of the air preheater through industrial cameras, avoids the inherent limitations of infrared imaging, and uses image processing and recognition algorithms to achieve accurate quantification of the blockage situation, which has higher comprehensiveness and accuracy compared with traditional methods. Description of the Drawings

[0107] Figure 1 is the layout diagram of the visual monitoring system

[0108] Figure 2This is a schematic diagram of the visualization monitoring method process. Detailed implementation mode

[0109] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. The components of the embodiments of the present application usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0110] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts fall within the scope of protection of the present application.

[0111] It should be noted that similar reference numerals and letters indicate similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0112] In the description of the present application, it should be noted that the orientation or positional relationship indicated by the terms "upper", "vertical", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this application is usually placed during use, or the orientation or positional relationship commonly understood by those skilled in the art. It is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present application. In addition, the terms "first", "second", etc. are only used for descriptive distinction and cannot be understood as indicating or implying relative importance.

[0113] In the description of the present application, it should also be noted that unless otherwise clearly specified and limited, the terms "set", "installed", "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific circumstances.

[0114] Embodiment 1

[0115] As Figure 1As shown in the figure, an online real-time visual monitoring system for air preheater blockage includes a data storage module, a human-machine interaction interface, an alarm system, an industrial computer, a PLC, an industrial camera, a protective cover, a supplementary light, and a crossbar. The crossbar is installed at the cold end of the secondary air chamber of the air preheater. The industrial camera, the supplementary light, and the protective cover are all installed on the crossbar. The industrial camera is located inside the protective cover, and the supplementary light is located beside the protective cover. The PLC is respectively connected to the industrial camera and the supplementary light by signals. As the rotor rotates, the industrial camera takes images of various positions of the air preheater rotor, and transmits the collected image information of the cold end to the industrial computer. The industrial computer preprocesses and identifies the images, and processes and analyzes the obtained images to obtain the blockage condition of the air preheater; it is more intuitive than the existing flow resistance monitoring method, and has less disturbance and error than the method based on infrared images. Image acquisition is completed by the industrial camera, and the collected images are transmitted to the industrial computer or edge computing device for preprocessing and identification. The industrial computer runs an identification program based on deep learning or traditional image processing algorithms to judge the blockage condition of the air preheater. Image processing and analysis are completed by a Python program on the industrial computer. This program integrates image preprocessing, feature extraction, and deep learning models to identify the blockage condition of the air preheater.

[0116] The monitoring system also includes a data storage and analysis module and a human-machine interaction interface (HMI) to achieve long-term data recording and visual monitoring functions.

[0117] The online real-time visual monitoring system for air preheater blockage also includes a data storage module: used to store original images, processing results, and historical data to support long-term trend analysis and model optimization;

[0118] The human-machine interaction interface: provides a visual interface, displays the monitoring results, and allows users to manually view images, adjust parameters, or receive alarm information;

[0119] The alarm system: when serious blockage is detected, it triggers an audible and visual alarm or sends an alarm signal to the control center.

[0120] The image recognition part includes an industrial computer: responsible for image processing and recognition, running a Python program to execute image preprocessing, feature extraction, deep learning model inference, or traditional image analysis algorithms, and finally judging the blockage condition of the air preheater. The industrial computer can be equipped with a GPU to improve the deep learning inference speed, receive and store the rotor images taken by the industrial camera for processing and output the blockage rate.

[0121] The industrial camera takes images of the cold end vertically upward. To prevent dust from falling on the camera lens and affecting the shooting effect, the industrial camera needs to be installed in a dust-proof protective cover;

[0122] When shooting a moving rotor, there should be no trailing shadow. The angular velocity, exposure time t, and shooting frequency f should meet the following requirements:

[0123] First, obtain the following parameters: rotor speed: ω, unit: r / min; rotor radius: R, unit: meter; industrial camera resolution: N, unit: pixel / meter; maximum allowable smear: Δp, unit: pixel, usually set to 1 pixel.

[0124] Then calculate the angular velocity as:

[0125]

[0126] The exposure time t should satisfy:

[0127]

[0128] The shooting frequency f should satisfy:

[0129]

[0130] In addition, it is also necessary to select appropriate focal length and resolution of the industrial camera through laboratory verification, where

[0131] 1. Resolution: It should be ensured that within the working distance range, the blockage details on the surface of the air preheater can be distinguished. Generally speaking, 5 million pixels (or higher, such as 8 million pixels) can provide sufficient image details, but the influence of different resolutions on the recognition accuracy needs to be tested in the experimental environment.

[0132] 2. Focal length: It needs to be selected according to the installation position and monitoring range of the industrial camera, so that the entire monitoring area can enter the field of view completely and ensure clear images. For example, a short focal length is suitable for close-range shooting, while a longer focal length is suitable for long-distance monitoring. The optimal focal length can be determined through experiments to make the details on the surface of the air preheater clearly visible, while avoiding defocusing or insufficient field of view.

[0133] 3. Lens selection: The FA lens (industrial fixed-focus lens) is suitable for high-precision imaging environments and has good anti-vibration and anti-light change performance. The manual aperture needs to be adjusted during the experiment to adapt to the lighting conditions under different working conditions and optimize the imaging quality.

[0134] The supplementary lights are installed on the crossbar. Each industrial camera is equipped with two supplementary lights, and the two supplementary lights are placed obliquely relative to each other. The light is focused around the rotor of the air preheater photographed by the industrial camera to provide lighting for the industrial camera shooting.

[0135] The dust cover is installed on the crossbar, and an industrial camera is installed inside. It is also installed vertically upward toward the cold end of the air preheater. The industrial camera needs to be verified to be able to work normally when the temperature inside the air preheater is high. The industrial camera is placed in an oven and the operating temperature is set to 65°C. The test time is 5 hours. The industrial camera, laser displacement sensor, and cover can all work normally under this condition and pass the verification. The upper part of the dust cover is provided with a window and a wiper. The lens of the industrial camera is installed close to the inside of the window, and the wiper is installed on the outside of the window. The window will be cleaned during the working interval of the industrial camera to ensure that the industrial camera has a clear and clean field of view. The wiper is connected to the PLC, and the start and stop of the wiper is controlled by the PLC program. When the wiper is started, it sweeps back and forth to clean the dust on the window. The window of the cover needs to be large enough to avoid affecting the normal operation of the industrial camera.

[0136] In order to prevent dust from falling on the industrial camera lens and affecting the shooting accuracy, the industrial camera is installed in a corresponding dust cover, which is equipped with a wiper, and the industrial camera lens is facing the cold end of the air preheater. The industrial camera is connected to the PLC. Before the industrial camera is ready to collect images, the PLC program controls the start of the wiper to clean the dust, providing a good working environment for the industrial camera.

[0137] The control part PLC is connected to the camera, fill light and shield wiper respectively to control the working logic of each device. The wiper cleans the dust on the surface of the shield before the camera works. When the wiper works for sufficient time (the shield surface is clean), the camera receives the shooting command and the fill light illuminates the subject.

[0138] Example 2

[0139] like Figure 2 As shown, a real-time visual monitoring method for online air preheater blockage includes the following steps:

[0140] Step 1. Control the shield wiper through PLC to clean the shield window and provide a clean field of view for industrial camera shooting.

[0141] Step 2. The PLC controls the wipers to stop working, the fill light starts to illuminate the subject, and the industrial camera collects images of the cold end of the air preheater rotor; under the control of the PLC program, the wipers of the shield will clean the shield window every few hours, and the industrial camera and fill light start working after the cleaning is completed. The fill light provides lighting for the industrial camera shooting area, and the industrial camera starts to obtain images of the cold end of the air preheater.

[0142] In step 2, the fill light needs to provide enough light to improve the image contrast while avoiding overexposure. The light intensity range is generally recommended to be 5000–15000 lux, and the specific value should be optimized through experiments based on the working scenario. Excessive light may cause reflection or overexposure, affecting recognition accuracy. Glare can be reduced by using a diffuser or polarizing filter.

[0143] Obtain the following parameters: rotor speed: ω, unit: r / min; rotor radius: R, unit: meter; industrial camera resolution: N, unit: pixel / meter; maximum allowable smear: Δp, unit: pixel, usually set to 1 pixel;

[0144] Then calculate the angular velocity as:

[0145]

[0146] The exposure time t should satisfy:

[0147]

[0148] The shooting frequency f should satisfy:

[0149]

[0150] Step 3. The industrial camera image is transmitted at high speed through the GigE Vision protocol and stored in the industrial control computer; the industrial control computer mainly consists of a CPU (executing image processing algorithms), a GPU (accelerating deep learning inference), a RAM (caching data), an SSD / HDD (storing real-time and historical data), a network card (supporting GigE Vision high-speed transmission), an I / O interface (connecting to PLC, display devices, storage devices), and an operating system (Windows / Linux), and is responsible for image processing, recognition, and monitoring the system operation.

[0151] The high-speed transmission protocol can adopt GigE Vision. GigE Vision has the advantages of high speed, long-distance transmission, and strong anti-interference ability. According to application requirements, USB3Vision (short-distance high-speed transmission), Camera Link (low latency and high bandwidth), CoaXPress (ultra-high speed and high resolution), etc. can also be adopted. The specific selection needs to comprehensively consider the transmission distance, bandwidth, latency, and cost.

[0152] Step 4. Enhance the image and perform preprocessing to screen the ROI area;

[0153] In Step 4, the Retinex algorithm is used to enhance the image. Retinex is an image enhancement algorithm based on the human visual system, mainly used to improve the brightness and color performance of the image. Its core idea is to decompose the image into a reflection component (the essential color of the object) and an illumination component (the illumination condition), and enhance the image by adjusting the illumination component. It mainly includes the following steps:

[0154] Step 4.1.1. Image decomposition;

[0155] I(x,y) = R(x,y) × L(x,y);

[0156] Where I(x,y) is the intensity value of the original image at the pixel point (x,y); R(x,y): the reflection component, representing the intrinsic color of the object; L(x,y): the illumination component, representing the illumination condition;

[0157] Step 4.1.2. Logarithmic transformation;

[0158] log(I(x,y)) = log(R(x,y)) + log(L(x,y));

[0159] Step 4.1.3. Estimate the illumination component;

[0160] log(L(x,y)) = log(I(x,y)) * G(x,y);

[0161] Where G(x,y) is a Gaussian filter used to smooth the image. The formula for the Gaussian filter is as follows,

[0162]

[0163] σ represents the standard deviation of the Gaussian distribution;

[0164] Step 4.1.4. Calculate the reflection component;

[0165] log(R(x,y)) = log(I(x,y)) - log(L(x,y));

[0166] R(x,y) = exp(log(I(x,y)) - log(L(x,y)));

[0167] Step 4.1.5. Reconstruct the enhanced image;

[0168] I'(x,y) = R'(x,y) × L(x,y);

[0169] Where the calculation of R'(x,y) can be adjusted according to specific requirements and the technologies adopted. Common methods include contrast stretching, Gamma correction, etc. Taking contrast stretching as an example, the specific calculation process is as follows:

[0170] R′(x, y) = α · R(x, y) + β

[0171] Where α is an adjustment factor used to control the degree of enhancement; β is an offset, which can usually be set to 0 or a small positive value to prevent the reflection component from becoming negative.

[0172] Where I'(x,y) is the enhanced image and R'(x,y) is the adjusted reflection component.

[0173] The specific steps for preprocessing to screen the ROI region are as follows:

[0174] To achieve automated ROI screening, the method of manual marking and neural network training needs to be adopted. The specific steps are as follows:

[0175] Step 4.2.1. Manually annotate the ROI training set;

[0176] Manually mark the ROI region on some training images to generate a binary mask M(x, y), which is defined as follows:

[0177]

[0178] If the data volume is insufficient, data augmentation (such as rotation, flipping, scaling, etc.) can be used to expand the training samples;

[0179] Step 4.2.2. Neural network model training;

[0180] Adopt a convolutional neural network (CNN) or a segmentation model based on the Transformer structure (such as U-Net, DeepLabV3+), and use the enhanced image R'(x, y) as the input and the ROI mask M(x, y) as the supervision signal for training;

[0181] Training objective: Learn the mapping relationship to automatically predict the ROI region from the enhanced image:

[0182]

[0183] where f θ is the neural network model with parameters θ; is the ROI result predicted by the model;

[0184] Step 4.2.3. ROI screening and post-processing;

[0185] The ROI prediction result obtained through neural network inference may contain noise and is optimized through morphological operations (opening operation, closing operation):

[0186]

[0187] After the final ROI region is extracted, it can be used for subsequent object detection or analysis.

[0188] Step 5. Segment the image through the region growing algorithm, and perform contour extraction and region area statistics on the segmented image to obtain the blocked area and the flowing area.

[0189] The image segmentation using the region growing algorithm is an image segmentation method based on pixel clustering. It mainly expands the region step by step according to the similarity between pixels to form the segmentation result. The basic steps are as follows:

[0190] Step 5.1.1. Select seed points;

[0191] Set one or more seed points S, which are usually obtained by manual selection, threshold method or automatic selection.

[0192] Step 5.1.2. Set the growth conditions;

[0193] Set the growth criterion and calculate the similarity between the pixels to be expanded and the pixels in the selected region. Common growth conditions can be gray level difference or color difference:

[0194] D(x, y) = |I(x, y) - I(s x , s y )|

[0195] where D(x, y) is the gray level or color difference between the pixel to be expanded and the seed point; I(x, y) is the gray level value or color value of the pixel to be expanded; I(s x , s y ) is the gray level value or color value of the seed point;

[0196] If D(x, y) < T, where T is the set threshold, then this pixel is classified into the current region, otherwise the expansion stops;

[0197] Step 5.1.3. Region expansion;

[0198] Judge the similarity of all adjacent pixels and add the pixels that meet the conditions to the region R. Repeat this process until all pixels are classified or there are no eligible pixels to add;

[0199] Step 5.1.4. Termination conditions;

[0200] The conditions for the termination of region growth can be: 1) no new pixels meet the growth criterion; 2) the preset maximum region size is reached; 3) the preset similarity threshold is reached;

[0201] Finally, the entire image is divided into multiple regions, and each region corresponds to a different characteristic part.

[0202] Specifically, contour extraction

[0203] The contour of the segmented image can be extracted through edge detection or morphological processing. Taking Canny edge detection as an example:

[0204]

[0205] Among them, G(x, y) is the gradient amplitude at the pixel point; G x and G y are the gradients in the horizontal and vertical directions respectively;

[0206] After that, the final contour is obtained through non-maximum suppression, double-threshold processing, and edge tracking.

[0207] The specific calculation of the area of the region is as follows:

[0208] The area calculation of the region is achieved by calculating the number of pixels inside the contour:

[0209]

[0210] Among them, A is the area of the region; R is the target region; (x, y) are the pixel coordinates in the region;

[0211] During the calculation, cv2.contourArea(contour) in OpenCV can be used to perform the area statistics of the contour region.

[0212] The specific calculation of the area of the blocked region and the flowing region is as follows:

[0213] After segmentation, the two regions inside and outside the contour in the image represent the blocked region and the flowing region respectively, and the area calculation is as follows:

[0214] (1) The area of the blocked region

[0215]

[0216] Among them, A block is the area of the blocked region; R block is the pixel set of the blocked region;

[0217] The area of the flowing region

[0218]

[0219] Among them, A flow is the area of the flowing region; R flow is the pixel set of the flowing region.

[0220] Step 6. Calculate the proportion of the blocked region in the total area to obtain a quantitative index characterizing the blockage degree of the air preheater rotor.

[0221] Specifically:

[0222]

[0223] A total = A block + A flow

[0224] Among them, P block is the blockage ratio; A total is the total area of the image; A block is the area of the blocked area.

[0225] Embodiment 3

[0226] As Figure 1 shown, the present invention mainly includes an industrial camera 1, a shield 2, a fill light 3, and a cross bar 4. Under the control of the PLC program, the wiper of the shield 2 will clean the shield window every few hours. After the cleaning is completed, the industrial camera 1 and the fill light 3 start to work. The fill light 3 provides illumination for the shooting area of the industrial camera, and the industrial camera 1 starts to obtain the image of the cold end of the air preheater. Among them, the industrial camera 1 is installed in the shield 2, and the lens is close to the shield window. The fill light 3 and the shield are both installed on the cross bar 4. After the image is collected, it is transmitted and stored in the computer. Further, the area of the blocked area is obtained through image preprocessing, threshold segmentation, contour extraction, and regional area statistics. The ratio of the area of the blocked area to the overall area is used as an index to quantify the degree of blockage of the air preheater.

[0227] The above-mentioned industrial camera 1 is a GigE interface industrial area array camera that uses Gigabit Ethernet (GigE) to quickly and real-time transmit images. After the image sensor receives the image data of the cold end of the air preheater, it completes the image data processing through various built-in ISP image processing algorithms, and finally completes the high-speed transmission of the image data through the GigE Vision protocol. To meet the image acquisition requirements of the industrial camera, a lens with a focal length of 16 mm and eight million pixels is selected considering factors such as focal length, interface, resolution, and working distance, and high-definition pictures can be obtained when shooting the moving air preheater.

[0228] For the said shield 2, the industrial camera is closely installed in the shield against the window. The shield window is aligned with the cold end of the air preheater and is equipped with a wiper that can be controlled to start and stop by the PLC. Since the temperature in the secondary air chamber of the air preheater is relatively high, the shield has passed the temperature resistance test and can work normally in this environment, extending the service life of the industrial camera and providing a relatively clean field of view for the shooting of the industrial camera.

[0229] For the said fill light 3, the lens angle is 15 degrees, the irradiation distance is 1 m, and the power is 40 W. Since the exposure time of the industrial camera is very short, overexposure will not occur. The fill light needs to be installed at a certain angle to ensure that its surface will not accumulate too much dust and affect the light intensity. The beam of the fill light hits the center of the shooting range of the industrial camera, providing a good field of view for the industrial camera shooting and improving the shooting quality.

[0230] The measurement method of the present invention includes the following steps:

[0231] (1) The wiper of the protective cover is controlled by a PLC to clean the viewing window of the protective cover, providing a clean field of view for the industrial camera to take pictures.

[0232] (2) The PLC controls the wiper to stop working, and then the fill light and the industrial camera start to work.

[0233] (3) The fill light illuminates the object to be photographed, and the industrial camera collects images of the cold end of the air preheater rotor.

[0234] (4) The images of the industrial camera are transmitted at high speed through the GigE Vision protocol and stored in the industrial control computer.

[0235] (5) The images are enhanced by the Retinex algorithm, and the ROI region is preprocessed and screened.

[0236] (6) The images are segmented by the region growing algorithm, and the contour of the segmented image is extracted and the area of the region is statistically analyzed to obtain the area of the blocked region and the area of the flow region.

[0237] (7) Calculate the proportion of the blocked area in the total area to obtain a quantitative index characterizing the degree of blockage of the air preheater rotor.

[0238] In addition, it should be noted that the above content described in this specification is only an example of the structure of the present invention. Any equivalent changes made according to the structure, features and principles described in the inventive concept of the present invention are included in the protection scope of the present invention. Those skilled in the art of the present invention can make various modifications, supplements or use similar methods to replace the specific embodiments described, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should fall within the protection scope of the present invention.

Claims

1. A real-time visual monitoring system for online air preheater blockage, characterized by: It includes a data storage module, a human-computer interaction interface, an alarm system, an industrial computer, a PLC, an industrial camera, a shield, a fill light and a crossbar. The crossbar is installed at the cold end of the secondary air bin of the air preheater. The industrial camera, the fill light and the shield are all installed on the crossbar. The industrial camera is located in the shield. The fill light is located next to the shield. The PLC is connected to the industrial camera and the fill light signal respectively. As the rotor rotates, the industrial camera captures images of various positions of the air preheater rotor, and transmits the image information collected at the cold end to the industrial computer. The industrial computer preprocesses and recognizes the image, and processes and analyzes the acquired image to obtain the blockage condition of the air preheater. The online air preheater blockage real-time visual monitoring system also includes a data storage module: used to store original images, processing results and historical data; Human-computer interaction interface: provides a visual interface to display monitoring results; Alarm system: When a blockage is detected, it triggers an audible and visual alarm or sends an alarm signal to the control center.

2. The real-time visual monitoring system for online air preheater blockage according to claim 1 is characterized by: The industrial camera shoots the image of the cold end vertically upwards, and each industrial camera is equipped with two fill lights, which are both tilted and placed opposite to each other.

3. The real-time visual monitoring system for online air preheater blockage according to claim 1 is characterized by: The dust cover is installed vertically upward toward the cold end of the air preheater. A window and a wiper are provided on the upper part of the dust cover. The lens of the industrial camera is installed close to the inside of the window, and the wiper is installed on the outside of the window. The wiper is connected to the PLC.

4. A real-time visual monitoring method for online air preheater blockage, characterized in that: The steps include: Step 1. Control the shield wiper through PLC to clean the shield window; Step 2. The wiper is controlled to stop working through the PLC, the fill light starts to illuminate the subject, and the image of the cold end of the air preheater rotor is collected through the industrial camera; Step 3. The industrial camera image is transmitted at high speed and stored in the industrial computer; Step 4. Enhance the image and perform preprocessing to select the ROI area; Step 5. Segment the image using a region growing algorithm, and perform contour extraction and regional area statistics on the segmented image to obtain the blocked area and the flow area; Step 6. Calculate the proportion of the blocked area to the total area to obtain a quantitative index characterizing the blockage degree of the air preheater rotor.

5. The method for real-time visual monitoring of online air preheater blockage according to claim 4 is characterized in that: In step 2, the fill light intensity range is 5000–15000 lux, and the following parameters are obtained: rotor speed: ω, unit: r / min; rotor radius: R, unit: meter; industrial camera resolution: N, unit: pixel / meter; maximum allowed smear: Δp, unit: pixel; Then calculate the angular velocity: The exposure time t should satisfy: The shooting frequency f should satisfy:

6. The method for real-time visual monitoring of online air preheater blockage according to claim 4 is characterized in that: In step 4, the image is enhanced by using the Retinex algorithm to decompose the image into a reflection component and an illumination component, including the following steps: Step 4.1.

1. Image decomposition; I(x,y)=R(x,y)×L(x,y); Where I(x,y) is the intensity value of the original image at the pixel point (x,y); R(x,y): reflection component, indicating the essential color of the object; L(x,y): illumination component, indicating the illumination conditions; Step 4.1.

2. Logarithmic transformation; log(I(x,y))=log(R(x,y))+log(L(x,y)); Step 4.1.

3. Estimate the illumination component; log(L(x,y)) = log(I(x,y)) * G(x,y); where G(x,y) is a Gaussian filter used to smooth the image. The formula of the Gaussian filter is as follows, σ represents the standard deviation of the Gaussian distribution; Step 4.1.

4. Calculate the reflection component; log(R(x,y)) = log(I(x,y)) - log(L(x,y)); R(x,y) = exp(log(I(x,y)) - log(L(x,y))); Step 4.1.

5. Reconstruct the enhanced image; I'(x,y) = R'(x,y) × L(x,y); where the calculation process of R'(x,y) is: R′(x,y) = α·R(x,y) + β where α is the adjustment factor and β is the offset; where I'(x,y) is the enhanced image and R'(x,y) is the adjusted reflection component.

7. The method for real-time visual monitoring of online air preheater blockage according to claim 4 is characterized by: The specific process of preprocessing to screen the ROI region is as follows: Adopt the method of manual marking and neural network training. The steps are as follows: Step 4.2.

1. Manually annotate the ROI training set; Manually mark the ROI region on some training images to generate a binary mask M(x,y), which is defined as follows: Step 4.2.

2. Train the neural network model; Adopt a convolutional neural network or a segmentation model based on the Transformer structure, and use the enhanced image R'(x,y) as the input and the ROI mask M(x,y) as the supervision signal for training; Training objective: Learn the mapping relationship to automatically predict the ROI region from the enhanced image: Among them, f θ is a neural network model with parameter θ; ROI results predicted by the model; Step 4.2.

3. ROI screening and post-processing; ROI prediction results obtained through neural network inference There may be noise, which can be optimized through morphological operations: After the final ROI region is extracted, it can be used for subsequent object detection or analysis.

8. The method for real-time visual monitoring of online air preheater blockage according to claim 4 is characterized by: In Step 5, the steps of image segmentation by the region growing algorithm are as follows: Step 5.1.

1. Select seed points; Set one or more seed points S; Step 5.1.

2. Set the growth conditions; Calculate the similarity between the pixels to be expanded and the pixels in the selected region. The growth condition is the gray difference or color difference: D(x,y)=|I(x,y)-I(s x ,s y )|; Where D(x,y) is the grayscale or color difference between the pixel to be expanded and the seed point; I(x,y) is the grayscale value or color value of the pixel to be expanded; I(s x ,s y ) is the grayscale value or color value of the seed point; If D(x,y) < T, where T is the set threshold, then this pixel is included in the current region, otherwise stop expanding; Step 5.1.

3. Region expansion; Judge the similarity of all adjacent pixels and add the pixels that meet the conditions to the region R. Repeat this process until all pixels are classified or there are no eligible pixels to add; Step 5.1.

4. Termination conditions; The conditions for terminating the region growth are: 1) No new pixels meet the growth criterion; 2) The preset maximum region size is reached; 3) The preset similarity threshold is reached; Finally, the entire image is divided into multiple regions, and each region corresponds to different feature parts; The specific process of contour extraction is: The contour of the segmented image is extracted through edge detection or morphological processing: Among them, G(x,y) is the gradient amplitude at the pixel point; G x and G y are the gradients in the horizontal and vertical directions respectively; After that, the final contour is obtained through non-maximum suppression, double-threshold processing, and edge tracking.

9. The method for real-time visual monitoring of online air preheater blockage according to claim 4, characterized in that: In Step 5, the specific process of calculating the region area is: The region area is calculated by calculating the number of pixels inside the contour: where A is the region area; R is the target region; (x,y) are the pixel coordinates in the region; The specific process of calculating the blocked region area and the flowing region area is: The two areas inside and outside the contour in the segmented image represent the blocked area and the flow area respectively, and the area calculation is as follows: (1 Blocked area Among them, A block is the area of ​​the blocked region; R block is the pixel set of the blocked area; Circulation area Among them, A flow is the circulation area; R flow To circulate the area pixel collection.

10. The method for real-time visual monitoring of online air preheater blockage according to claim 4, characterized in that: The step 6 is specifically as follows: A total =A block +A flow ; Among them, P block is the blocking ratio; A total is the total area of ​​the image; A block is the area of ​​the blocked region.

Images