Intelligent detection system and method for catalyst structure based on multi-dimensional AI vision

Abstract

The present invention relates to the field of catalyst carrier detection technology, and in particular to a catalyst structure intelligent detection system and method based on multi-dimensional AI vision, including upper and lower detection of the catalyst carrier; determining the detection method of the lower detection according to the carrier detection evaluation value; for partially blocked channels and completely blocked channels, determining whether the corresponding single channel has structural deformation according to the grayscale value distribution trend of the corresponding area in the lower projection image; for normal channels, adjusting the illumination intensity of the parallel light source, and determining whether the normal channel is a slightly dusty channel according to the average grayscale value and infrared grayscale of the lower projection image; determining the grid where the partially blocked channels and completely blocked channels that require secondary judgment of the channel type are located according to the grayscale change amplitude in the mobile lower detection, and adjusting the detection standard for the mobile lower detection. The present invention improves the accuracy of the detection results by performing upper and lower detection on the catalyst carrier.

Description

Technical Field

[0001] The present invention relates to the field of catalyst carrier detection technology, and in particular to a catalyst structure intelligent detection system and method based on multi-dimensional AI vision. Background Art

[0002] In the environmental protection sector, particularly for industrial flue gas treatment, SCR (Selective Catalytic Reduction) technology is widely used in the denitrification process. Corrugated plate SCR catalysts are a core component. Their performance directly impacts nitrogen oxide (NOx) removal efficiency, system operating costs, and environmental compliance. Testing corrugated plate SCR catalysts is crucial. For example, to ensure denitrification efficiency and environmental compliance, activity decay is often monitored. Long-term exposure to high temperatures (300-400°C), high dust levels, sulfur (SO2 / SO3), and heavy metal (As, Pb)-laden flue gases can deactivate active components (such as V2O5-WO3 / TiO2) through poisoning, sintering, or clogging. Regular testing assesses the catalyst's residual activity and prevents excessive NOx emissions (e.g., the national ultra-low emission standard requires NOx <50 mg / Nm³) due to efficiency degradation. At the same time, the structural integrity is tested. The multi-layer corrugated structure of the corrugated plate catalyst is prone to cracks and collapse due to mechanical stress or uneven thermal expansion. The test can detect physical damage in time and prevent local airflow short-circuiting (Bypass) from causing a sudden drop in denitrification efficiency.

[0003] Chinese Patent Publication No.: CN113008897A discloses a method, device and equipment for detecting defects in ceramic catalyst carriers. The detection method includes: obtaining a carrier end face image by utilizing machine vision, determining a target carrier image based on the carrier end face image, and determining the contour of the target carrier image, and calculating the mesh side length of the target carrier based on the contour of the target carrier image; then, according to the mesh side length of the target carrier, establishing a judgment model for the connected domain type of a single mesh of the target carrier, and judging the connected domain type of each mesh in the target carrier according to the judgment model for the connected domain type of a single mesh, and generating a mapping table for the judgment result; finally, judging the defect types of meshes of closed connected domains and meshes of non-closed connected domains respectively according to the mapping table. It can be seen that the ceramic catalyst carrier defect detection method, device and equipment have the following problems:

[0004] When penetrating irradiation is performed on the bottom of the catalyst carrier, the shading strips at the bottom edge are difficult to achieve the ideal complete shading condition, resulting in misjudgment and missed detection results, affecting the accuracy of the detection results. Summary of the Invention

[0005] To this end, the present invention provides a catalyst structure intelligent detection system and method based on multi-dimensional AI vision, which is used to overcome the problem in the prior art that when the bottom of the catalyst carrier is subjected to penetrating irradiation, the shading strip at the bottom edge is difficult to achieve the ideal complete shading condition, resulting in misjudgment and missed judgment in the detection results, affecting the accuracy of the detection results.

[0006] To achieve the above objectives, the present invention provides a catalyst structure intelligent detection method based on multi-dimensional AI vision, comprising:

[0007] Acquiring an image of an upper portion of a catalyst carrier in a closed space, and dividing the image into a plurality of grids according to the outer contours of a plurality of channels of the catalyst carrier;

[0008] Determining surface cracking, straightness deviation, and angular deviation of the catalyst carrier to determine the detection status of the upper portion of the catalyst carrier;

[0009] Determining a carrier detection evaluation value based on the pore characteristics of the carrier upper image, obtaining a lower projection image of the catalyst carrier based on a first comparison result obtained by comparing the carrier detection evaluation value with a standard preset value, and generating an overall qualified evaluation value of the catalyst carrier based on the grayscale value of the lower projection image, the carrier detection evaluation value, and the actual weight of the catalyst carrier;

[0010] Calculating the transparency of each single channel in the catalyst carrier based on the grayscale value of the lower projection image, determining the channel type of each single channel, and distinguishing between dust accumulation and structural deformation based on the grayscale value distribution trend of the corresponding area of ​​the blocked channel in the lower projection image;

[0011] Adjust the position and illumination intensity of the parallel light source, and verify the normal channel in the channel type according to the average gray value of the lower projection image and the increase of the infrared average gray value;

[0012] After adjusting the position of the parallel light source, the grid positions of partially blocked channels and completely blocked channels are determined based on the grayscale variation of the lower projected image, and a secondary judgment is made on the channel types at the grid positions;

[0013] Whether to adjust the inspection standard for the mobile lower part inspection is determined according to whether the channel type has changed, and the overall qualified evaluation value is corrected according to the distribution and degree of change of several single channels whose channel types have changed.

[0014] Further, the process of determining the straightness deviation and angular deviation conditions includes;

[0015] According to the side length of a single channel as the extraction length, the maximum value of a plurality of vertical distances within the extraction length is extracted as the straightness deviation;

[0016] When the straightness deviation is greater than a first standard value, the straightness of the catalyst carrier is determined to be unqualified; when the straightness deviation is less than or equal to the first standard value, the straightness of the catalyst carrier is determined to be qualified;

[0017] The included angle is calculated as the actual angle value using the vector dot product formula, and the absolute difference between the actual angle value and 90° is calculated as the angle deviation;

[0018] When the angle deviation is less than or equal to the second standard value, the angle deviation of the catalyst carrier is determined to be qualified; when the angle deviation is greater than the second standard value, the angle deviation of the catalyst carrier is determined to be unqualified.

[0019] Furthermore, the process of determining the detection status of the upper portion of the catalyst carrier includes:

[0020] When the straightness and angle deviation are qualified and there is no surface cracking of the catalyst carrier, the upper portion of the catalyst carrier is determined to be in the first detection state, and the pore characteristic detection is performed to determine the carrier detection evaluation value;

[0021] If either the straightness or the angle deviation is unqualified, or the catalyst carrier has surface cracks, the upper portion of the catalyst carrier is judged to be in the second detection state, and the fault type is output to alarm.

[0022] Furthermore, the process of determining the carrier detection evaluation value includes:

[0023] Detecting the number of through holes in the image on the carrier, calculating the effective through hole rate, detecting the hole area and hole perimeter of a plurality of holes in the image on the carrier, and calculating the shape factor of the holes;

[0024] Calculating a carrier detection evaluation value according to the effective through-hole ratio and the shape factor of the hole, and comparing the carrier detection evaluation value with a first preset value and a second preset value;

[0025] When the carrier detection evaluation value is less than the first preset value and greater than the second preset value, the comparison result is used as the first comparison result, and the catalyst carrier is subjected to a mobile lower part detection;

[0026] When the carrier detection evaluation value is less than or equal to a second preset value, outputting the catalyst carrier as an unqualified carrier, issuing an alarm signal and marking the fault type as porosity deviation;

[0027] In the process of performing mobile lower inspection on the catalyst carrier, an overall qualified evaluation value of the catalyst carrier is generated according to the grayscale value of the lower projection image, the carrier inspection evaluation value and the actual weight of the inspected catalyst carrier.

[0028] Furthermore, the process of determining the channel type of each single channel includes:

[0029] detecting the grayscale value of the lower projection image and calculating the single-channel transmittance of a plurality of single channels in the catalyst carrier;

[0030] If the permeability of the single channel is greater than or equal to the first permeability, then the corresponding single channel in the catalyst carrier is judged to be a normal channel;

[0031] If the permeability of the single channel is less than the first permeability and greater than or equal to the second permeability, then the corresponding single channel in the catalyst carrier is determined to be a partially blocked channel;

[0032] If the single-channel permeability is less than the second permeability, it is determined that the corresponding single channel in the catalyst carrier is a completely blocked channel.

[0033] Further, extracting the grayscale value curves of the corresponding areas of the partially blocked and completely blocked single channels in the lower projection image along the axial direction, and calculating the grayscale change rate;

[0034] When the grayscale change rate within a unit length exceeds the preset range, it is judged that the grayscale value drops drastically. The blockage of a partially blocked single channel is caused by dust accumulation, and no physical deformation occurs.

[0035] When the grayscale change rate within the unit length is lower than the preset range, it is determined that the grayscale value is gradually reduced, and the cause of the blockage of the partially blocked single channel is deformation, and the corresponding single channel undergoes physical deformation.

[0036] Furthermore, the process of determining whether the normal channel is a slightly dusty channel includes:

[0037] Increase the illumination intensity of the bottom parallel light source according to the initial increase amplitude, and calculate the increase amplitude of the average grayscale value and the infrared average grayscale;

[0038] After increasing the light intensity, if the difference between the increase in the average grayscale value and the increase in the infrared average grayscale is greater than the difference evaluation value, it is judged that there is slight dust accumulation in the normal channel of the catalyst carrier, and it is a slightly dusty channel.

[0039] Furthermore, the process of adjusting the position of the parallel light source includes:

[0040] The bottom parallel light source moves laterally by an initial distance, and the resulting lower projection image is obtained. The grayscale variation amplitudes of the partially blocked channel and the completely blocked channel at the corresponding grid positions before and after the movement are calculated.

[0041] Determine the corresponding grid positions of the lower projection images generated after the partial blocking channel and the completely blocking channel with a grayscale change amplitude greater than or equal to the change value are moved.

[0042] Furthermore, a secondary single-channel transparency calculation is performed at the corresponding grid position to verify the channel type of the single channel;

[0043] If the channel types of the determined plurality of single channels of the catalyst carrier change, increasing the inspection standard for the mobile lower portion inspection, the inspection standard being the first preset value compared with the carrier inspection evaluation value;

[0044] The process of correcting the overall qualified evaluation value includes coordinateizing a number of single channels whose channel types have changed to obtain corresponding change coordinates, and dividing a change area corresponding to any single channel whose channel type has changed in the lower projection image;

[0045] The distribution of several single channels whose channel types have changed is determined according to the change coordinates, the change ratio of the change area in the lower projection image is calculated, and the overall qualified evaluation value is corrected.

[0046] A catalyst structure intelligent detection system based on multi-dimensional AI vision, including:

[0047] The supporting frame has several supporting components inside, and the frame is equipped with a top plate, a sheet metal bracket, a visual camera, a fill light, a glass panel, a catalyst carrier, a glass base, a U-shaped scale, a parallel light source and a moving component from top to bottom;

[0048] The visual camera is hung on the bottom of the top plate by a sheet metal bracket, and is used to obtain the upper image and lower projection image of the catalyst carrier. The sheet metal bracket can adjust the height and lens swing angle of the visual camera and control its four-way lateral movement;

[0049] The sheet metal bracket is a multi-angle adjustable bracket that supports four-way horizontal movement of the camera, front and back, left and right, and supports 100mm height adjustment of the camera up and down, and supports 180° free adjustment of the camera lens tilt angle;

[0050] The glass panel is arranged on the top of the supporting assembly located on the top of the supporting frame, and is used to carry the lower projection image;

[0051] The glass base is located at the bottom of the catalyst carrier and is used to support the catalyst carrier and limit the relative position relationship between the catalyst carrier and the U-shaped scale;

[0052] The U-shaped scale is fixedly connected to the inner side of the support assembly at the bottom of the carrier body, is located at the bottom of the glass base, and is used to detect the weight of the catalyst carrier;

[0053] A rolling door is provided on the side of the supporting frame, and when the rolling door is lowered, it forms a closed space with the supporting frame;

[0054] The parallel light source is plate-shaped and located at the bottom of the U-shaped scale. The moving component is arranged at the bottom of the parallel light source. The moving component includes a bracket, a first slide rail and a second slide rail. The first slide rail and the second slide rail are arranged in parallel on both sides of the top of the bracket.

[0055] Compared to existing technologies, the present invention offers significant advantages in that its corrugated plate catalyst, a unique structure, is commonly used in environmental protection applications such as flue gas denitrification (SCR systems). The catalyst substrate of this device is pressed into a honeycomb-shaped channel pattern, creating a regularly arranged flow path. This increases the contact area between the gas and the catalyst while reducing airflow resistance. The substrate is coated with a porous carrier (such as TiO2 or Al2O3) and loaded with active components (such as V2O5-WO3 or MnO3). The specific composition is adjusted based on application requirements (e.g., vanadium-based catalysts are commonly used in SCR denitrification). The corrugated channels optimize airflow distribution, significantly reducing energy consumption compared to traditional pellet catalysts. The catalyst is suitable for use in dusty or high-particulate matter flue gases, is resistant to clogging, and offers advantages such as low pressure drop and high clogging resistance. Furthermore, due to the different brightness of the upper and lower projection images of the carrier, the visual camera of this device uses different exposure compensation preset parameters to capture the upper and lower projection images of the carrier, increasing the adaptability and accuracy of upper and lower catalyst carrier inspections.

[0056] Furthermore, although the catalyst carrier has advantages such as low pressure drop and high anti-clogging performance in the SCR (selective catalytic reduction) denitrification system, the start-up and shutdown of the unit may cause the surface of the catalyst carrier to be distorted or cracked during actual application. The method divides the upper image of the carrier into several grids, extracts the grid edges through edge detection, and performs a preliminary inspection of the catalyst carrier for straightness, angular deviation and crack detection on the upper part of the catalyst carrier. The state of the catalyst carrier is determined based on the preliminary inspection of the upper image of the carrier, and whether the edge contour of the carrier is intact or cracked is detected. The output fault type is further determined based on the operating state of the catalyst carrier, or the carrier inspection evaluation value is calculated based on the effective through-hole rate and the shape factor of the hole. According to the carrier inspection evaluation value, the inspection method of the catalyst carrier is penetrating lower inspection or mobile lower inspection, which saves the inspection process, improves the inspection efficiency, and the adaptability and flexibility of the inspection of the catalyst carrier.

[0057] Furthermore, the structure of the corrugated plate catalyst is easily clogged by fly ash in the flue gas. After long-term operation, the fly ash will accumulate and form a solid surface or blockage on the catalyst surface and in the corrugated channel, resulting in a reduction in the effective reaction area and a decrease in the denitrification efficiency. At the same time, ammonium bisulfate (ABS) may also adhere to the catalyst surface in liquid form, exacerbating the blockage. This device realizes deep non-destructive testing and all-round quality control of the carrier's "surface-interior" and "geometry-function" by performing projection deformation analysis through upper and lower detection. At the same time, structural deformation and dust accumulation in the catalyst carrier channel will cause changes in the channel's transparency. This method obtains a lower projection image by reversely irradiating the catalyst carrier, calculates the single-channel transparency of several single channels in the catalyst carrier based on the grayscale value of the lower projection image, and determines whether the channel type of the several single channels is a normal channel, a partially blocked channel, or a completely blocked channel. Based on the experimental results that the pore deformation of a single channel will affect adjacent pores and cause the transparency of several adjacent pores to decrease step by step from near to far, for partially blocked channels and completely blocked channels, the grayscale value distribution trend of the corresponding area in the lower projection image is used to determine whether the corresponding single channel blockage is caused by structural deformation, thereby improving the accuracy of detecting fault types and avoiding the aggravation of faults caused by taking incorrect adjustment measures.

[0058] Furthermore, due to the diffuse reflection of fly ash and the scattering angle of the light source, the actual dust accumulation in the edge area of ​​the catalyst carrier may be mistakenly judged as normal (insufficient illumination masks the grayscale change), and the slight discoloration in the central area may be magnified as a serious defect; for the normal channel, this method determines whether the normal channel is a slightly dusty channel based on the average grayscale value and infrared grayscale of the lower projection image by adjusting the illumination intensity of the parallel light source; for partially blocked channels and completely blocked channels, the channel type is secondary judged based on the grayscale change amplitude by horizontally moving the bottom parallel light source; this scheme performs optical dynamic reconstruction by horizontally moving the bottom parallel light source, and uses multi-angle projection differences to analyze the three-dimensional transparency characteristics of the catalyst carrier, which can significantly improve the accuracy and depth information acquisition capabilities of traditional static detection, and by moving the bottom parallel light source, reduces the risk of detection error caused by illumination attenuation in the edge area according to the changes in the lower projection image, and improves the edge detection sensitivity to the same level as the center area. At the same time, after determining that a misjudgment has occurred, the detection standard for mobile lower detection is increased, thereby increasing the accuracy and flexibility of detection.

[0059] Furthermore, in this embodiment, since the shading strips set along the periphery of the catalyst carrier cannot achieve the purpose of complete shielding, light leakage occurs, which will affect the lower projection image and cause illumination differences in different areas, thereby masking grayscale changes or amplifying defects. For example, the middle and lower areas of the lower projection image in the attached figure have obvious illumination differences compared to the edge areas. In response to this technical problem, the present invention verifies the channel type according to the mobile lower detection, and then calculates the clustering response distribution of several single channels whose channel types have changed. It is determined whether the several single channels whose channel types have changed are concentratedly distributed or discretely distributed in the lower projection image, and whether the cause of the change in channel type is caused by the illumination difference can be analyzed; and the degree of change is reflected by the number of changes and the proportion of changes. The degree of influence of the illumination difference on the detection accuracy can be analyzed, and the calculated overall qualified evaluation value is corrected according to the distribution and degree of change of several single channels whose channel types have changed, so as to improve the detection accuracy of each single channel in the catalyst carrier, and solve the problem of misjudgment and missed judgment of the detection results due to the existence of illumination differences, which affects the accuracy of the detection results.

[0060] Furthermore, the glass base is a highly transparent glass base, and the light source can directly penetrate the base (the frosted glass can be used as an optical diffuser to achieve a uniform light effect and reduce the illumination attenuation in the edge area). The U-shaped scale can measure the weight of the catalyst carrier and the position of the clamped glass base and avoid blocking the penetrating light of the bottom parallel light source. The parallel light source can fully illuminate the catalyst carrier through the sliding rail of the moving mechanism. After the rolling shutter door is started, a flexible rolling shutter can be lowered to provide a closed, light-proof detection space. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 Flowchart of a catalyst structure intelligent detection method based on multi-dimensional AI vision in an embodiment of the present invention;

[0062] Figure 2 This is a schematic diagram of the main structure of the catalyst structure intelligent detection system in an embodiment of the present invention;

[0063] Figure 3 This is a front view structural diagram of a catalyst carrier placed inside a carrier body in an embodiment of the present invention;

[0064] Figure 4 This is a schematic diagram of the main structure of the supporting frame in an embodiment of the present invention;

[0065] Figure 5 This is a schematic structural diagram of a sheet metal bracket in an embodiment of the present invention;

[0066] Figure 6 is the lower projection image in the embodiment of the present invention;

[0067] In the figure: 2-catalyst carrier, 3-U-shaped scale, 4-parallel light source, 5-rolling door, 6-visual camera, sheet metal bracket 61, 611-hanging hole, 612-hanging plate, 613-side plate, 614-side hole, 7-fill light, 8-top plate, 9-support assembly, 11-glass base, 41-bracket, 42-first slide rail, 43-second slide rail. DETAILED DESCRIPTION

[0068] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0069] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0070] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0071] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0072] See also Figure 1-Figure 5 As shown, Figure 1 Flowchart of a catalyst structure intelligent detection method based on multi-dimensional AI vision in an embodiment of the present invention; Figure 2 This is a schematic diagram of the main structure of the catalyst structure intelligent detection system in an embodiment of the present invention; Figure 3 This is a front view structural diagram of a catalyst carrier placed inside a carrier body in an embodiment of the present invention; Figure 4 This is a schematic diagram of the main structure of the supporting frame in an embodiment of the present invention; Figure 5 This is a schematic structural diagram of a sheet metal bracket in an embodiment of the present invention; Figure 6 This is the lower projection image in the embodiment of the present invention.

[0073] The present invention provides a catalyst structure intelligent detection method based on multi-dimensional AI vision, comprising:

[0074] Step S1, placing the catalyst carrier in a detection system for upper detection and lower detection, wherein the upper detection obtains an image of the upper portion of the carrier, and the lower detection obtains a projection image of the lower portion;

[0075] Step S2, dividing the upper image of the catalyst carrier into a plurality of grids according to the outer contours of the plurality of channels of the catalyst carrier, performing image detection on the upper image of the catalyst carrier to obtain grid edges and grid corners of the plurality of grids, and determining surface cracking, straightness deviation, and angular deviation of the catalyst carrier;

[0076] Step S3, determining a detection state of the upper portion of the catalyst carrier based on surface cracking, straightness, and angular deviation, obtaining a carrier detection evaluation value when the upper portion of the catalyst carrier is in a first detection state, and outputting a fault type when the upper portion of the catalyst carrier is in a second detection state;

[0077] Step S4, determining a carrier detection evaluation value based on the pore characteristics of the carrier upper image, comparing the carrier detection evaluation value with a first preset value and a second preset value, respectively, obtaining a lower projection image of the catalyst carrier based on a first comparison result in the comparison results, and then determining whether the detection method of the lower detection is a penetrating lower detection or a moving lower detection, and correspondingly obtaining a lower projection image of the catalyst carrier or obtaining a lower projection image of the catalyst carrier before and after the parallel light source moves;

[0078] Step S5, calculating the transparency of each single channel in the catalyst carrier according to the grayscale value of the lower projection image, and determining the channel type of the plurality of single channels as a normal channel, a partially blocked channel, or a completely blocked channel;

[0079] Step S6: For partially blocked channels and completely blocked channels, determine the cause of blockage of the corresponding single channel based on the grayscale value distribution trend of the corresponding area in the lower projection image, distinguish between dust accumulation and structural deformation, and generate an overall qualified evaluation value of the catalyst carrier based on the grayscale value of the lower projection image, the carrier detection evaluation value, and the actual weight of the catalyst carrier;

[0080] Step S7: For the normal channel in the channel type, adjust the position and illumination intensity of the parallel light source, and determine whether the normal channel is a slightly dusty channel based on the average grayscale value of the lower projection image and the increase in the infrared average grayscale;

[0081] Step S8, after adjusting the position of the parallel light source, determining the grid positions of partially blocked channels and completely blocked channels based on the grayscale variation of the lower projection image in the movable lower detection, and performing a secondary determination of the channel type at the grid positions;

[0082] Step S9, adjusting the inspection standard for the mobile lower part inspection according to whether the channel type has changed, and revising the overall qualified evaluation value according to the distribution and degree of change of several single channels whose channel types have changed.

[0083] The application scenario of this embodiment is SCR denitrification: NOx emission reduction in industrial boilers such as power plants and steel mills (temperature range 300-400°C), and the catalyst structure is a honeycomb structure.

[0084] Upper inspection: After the box is closed, press the layer identification button on the box, and the upper strip fill light 7 lights up for 1 second. At the same time, the visual camera 6 takes a photo according to the first exposure compensation preset parameters to obtain the image of the upper part of the carrier. The computer reads the reading of the U-shaped scale 3 through the RS485 interface and displays the analyzed carrier size, number of layers, weight, and qualification information based on the captured image, together with the image of the upper part of the carrier, on the 15.6-inch screen of the box.

[0085] For lower inspection, raise the electric rolling door 5, place the translucent glass panel 8, lower the electric rolling door 5, and then press the hole blocking recognition button on the box. The bottom parallel light source lights up for 5 seconds, and the light beam penetrates the carrier and projects it onto the translucent glass panel 8. At the same time, the visual camera 6 takes a photo according to the second exposure compensation preset parameters to obtain the lower projection image. The computer displays the analyzed carrier transparency and qualification information together with the lower projection image on the 15.6-inch screen of the box based on the captured lower projection image.

[0086] In implementation, the difference between the penetration detection and the mobile detection is that the mobile detection includes the penetration detection, and the mobile detection is to move a parallel light source to perform a secondary penetration detection after the penetration detection.

[0087] Specifically, the corrugated plate catalyst, a unique catalyst structure, is commonly used in environmental protection applications such as flue gas denitrification (SCR systems). The catalyst substrate in this device is pressed into a honeycomb-like pattern, creating a regularly arranged flow path. This increases the contact area between the gas and the catalyst while reducing airflow resistance. The substrate is coated with a porous carrier (such as TiO2 or Al2O3) and loaded with active components (such as V2O5-WO3 or MnO3). The specific composition is adjusted according to application requirements (for example, vanadium-based catalysts are commonly used in SCR denitrification). The corrugated channels optimize airflow distribution, significantly reducing energy consumption compared to traditional pellet catalysts. They are suitable for use in dusty or high-particulate matter flue gases, are resistant to clogging, and offer advantages such as low pressure drop and high clogging resistance. Furthermore, due to the different brightness of the upper and lower projection images of the carrier, the visual camera in this device uses different exposure compensation preset parameters to capture the upper and lower projection images of the carrier, increasing the adaptability and accuracy of upper and lower catalyst carrier inspections.

[0088] During the upper inspection process, a visual camera is used to obtain an image of the upper portion of the carrier, and the image of the upper portion of the carrier is analyzed to perform a preliminary inspection of the catalyst carrier. The state of the catalyst carrier is determined based on the analysis results of the upper portion of the carrier image, and whether the edge contour of the carrier is intact or cracked is detected;

[0089] Performing edge contour detection on a plurality of single channels in the image on the carrier, extracting the channel outer contours of the plurality of single channels as grid edge detection edge straightness, and extracting the maximum value of the vertical distance within the plurality of extraction lengths as the straightness deviation according to the side length of the single channel as the extraction length;

[0090] During implementation, the upper image of the carrier is divided into several grids according to the external contours of several single-channel channels. The continuous point set of the grid edge is obtained through sub-pixel edge positioning. The RANSAC (random sampling consensus) algorithm is used to fit the ideal straight line, and the vertical distance from each edge point to the fitted straight line is calculated.

[0091] When the straightness deviation is greater than a first standard value, the straightness of the catalyst carrier is determined to be unqualified; when the straightness deviation is less than or equal to the first standard value, the straightness of the catalyst carrier is determined to be qualified;

[0092] Harris corner detection is used to locate grid corners, three adjacent corner points are selected to form the included angle, and the actual angle value is calculated using the vector dot product method;

[0093] Specifically, the Harris algorithm is used to detect sub-pixel corner points in the image above the carrier, a valid corner point group is screened using a distance constraint, three adjacent corner points are selected to form two edge vectors, and the included angle is calculated as the actual angle value using the vector dot product formula θ = arccos[(a·b) / (|a||b|)];

[0094] Calculating the absolute difference between the actual angle value and 90° as the angle deviation, determining that the angle deviation of the catalyst carrier is qualified when the angle deviation is less than or equal to a second standard value, and determining that the angle deviation of the catalyst carrier is unqualified when the angle deviation is greater than the second standard value;

[0095] The first standard value is 0.1 mm, and the second standard value is 0.5°.

[0096] Inputting the upper portion image of the catalyst carrier into the trained YOLOv8 model to perform surface crack detection, and determining that the catalyst carrier has surface cracks when cracks in the upper portion image of the carrier are selected in the recognition box;

[0097] When the straightness and angle deviation are qualified and there is no surface cracking of the catalyst carrier, the upper portion of the catalyst carrier is determined to be in the first detection state, and the carrier detection evaluation value is determined according to the pore characteristics;

[0098] If any of the straightness or angle deviation is unqualified, or if there is surface cracking on the catalyst carrier, the upper part of the catalyst carrier is judged to be in the second detection state, and the fault type is output to alarm;

[0099] Specifically, the fault types include straightness fault, angle deviation fault and surface cracking fault;

[0100] Performing pore characteristic detection on the catalyst carrier in the first operating state, and determining a carrier detection evaluation value according to the pore characteristic;

[0101] The pore characteristics include effective through-porosity Φe and shape factor SF.

[0102] Detect the number of through holes on the upper image of the carrier and calculate the effective through hole rate Φe = number of through holes / total number of holes × 100%;

[0103] Detect the hole area and hole perimeter of several holes in the image on the carrier, and calculate the hole shape factor SF, SF = (4π × hole area) / (hole perimeter) 2 ×100%;

[0104] The carrier detection evaluation value = 50 × (effective through-hole ratio Φe + shape factor SF), and the carrier detection evaluation value is compared with the first preset value and the second preset value;

[0105] When the carrier detection evaluation value is greater than or equal to a first preset value, obtaining a lower projection image of the catalyst carrier to perform a penetrating lower detection on the catalyst carrier;

[0106] When the carrier detection evaluation value is less than the first preset value and greater than the second preset value, performing a mobile lower detection on the catalyst carrier using the comparison result as the first comparison result;

[0107] When the carrier detection evaluation value is less than or equal to the second preset value, the catalyst carrier is output as an unqualified carrier, an alarm signal is issued, and the fault type is marked as pore characteristic deviation.

[0108] Among them, the first preset value is 85, and the second preset value is 70.

[0109] Specifically, although the catalyst carrier has advantages such as low pressure drop and high anti-clogging performance in the SCR (selective catalytic reduction) denitrification system, the start-up and shutdown of the unit may cause the surface of the catalyst carrier to be distorted or cracked during actual application. The method divides the upper image of the carrier into several grids, extracts the grid edges through edge detection, and performs a preliminary inspection of the catalyst carrier for straightness, angular deviation and crack detection on the upper part of the catalyst carrier. The state of the catalyst carrier is determined based on the preliminary inspection of the upper image of the carrier, and whether the edge contour of the carrier is intact or cracked is detected. The output fault type is further determined based on the operating state of the catalyst carrier, or the carrier inspection evaluation value is calculated based on the effective through-hole rate and the shape factor of the hole. According to the carrier inspection evaluation value, the inspection method of the catalyst carrier is penetrating lower inspection or mobile lower inspection, which saves the inspection process, improves the inspection efficiency, and the adaptability and flexibility of the inspection of the catalyst carrier.

[0110] The lower part detection obtains the lower part projection image and analyzes the carrier transparency of the catalyst carrier (honeycomb structure) based on the lower projection image;

[0111] Normal carrier: When the cellular channel is transparent, the projection is a uniform dot matrix (high brightness).

[0112] Blocked carrier: The brightness of the local area is reduced (light intensity attenuation) or there is no light spot at all (complete blockage).

[0113] Structural deformation: The light spot arrangement is distorted or the spacing is abnormal.

[0114] In this embodiment, the catalyst carrier is divided into a plurality of single channels according to the internal structure of the catalyst carrier.

[0115] Transparency calculation, detecting the gray value of the lower projection image and calculating the single-channel transmittance of several single channels in the catalyst carrier;

[0116] The permeability threshold is preset to classify several channels in the catalyst carrier into three categories: "normal / partially blocked / completely blocked";

[0117] Single channel transparency Ti=(Ii-Imin) / (Imax-Imin), where Ii is the average grayscale value of any single channel after excluding edge pixels, and Imax and Imin are the maximum and minimum values ​​of Ii;

[0118] If the single-channel permeability Ti is greater than or equal to the first permeability, the corresponding single channel in the catalyst carrier is judged to be a normal channel;

[0119] If the single channel permeability Ti is less than the first permeability and greater than or equal to the second permeability, the corresponding single channel in the catalyst carrier is judged to be a partially blocked channel, and the abnormality degree of the single channel is 0.2-0.8;

[0120] If the single channel permeability Ti is less than the second permeability, the corresponding single channel in the catalyst carrier is judged to be a completely blocked channel, and the abnormality degree of the single channel is 0.8-1.0;

[0121] For partially blocked and completely blocked catalyst carriers, the grayscale value distribution trend of the corresponding area of ​​the partially blocked channel or the completely blocked single channel in the lower projection image is used to determine whether the honeycomb structure inside the catalyst carrier is physically deformed in the blocked single channel, thereby distinguishing between dust accumulation and structural deformation.

[0122] In implementation, grayscale value curves of the corresponding areas of the partially blocked and completely blocked single channels in the lower projection image are extracted along the axial direction, and the grayscale change rate is calculated;

[0123] When the grayscale change rate within a unit length exceeds the preset range, it is judged that the grayscale value drops drastically. The blockage of a partially blocked single channel is caused by dust accumulation, and no physical deformation occurs.

[0124] When the grayscale change rate within the unit length is lower than the preset range, it is determined that the grayscale value is gradually reduced, and the cause of the blockage of the partially blocked single channel is deformation, and the corresponding single channel undergoes physical deformation.

[0125] The first transparency is 0.6, the second transparency is 0.3, the preset range is (10, 15), and the unit length is mm.

[0126] Specifically, the structure of the corrugated plate catalyst is easily clogged by fly ash in the flue gas. After long-term operation, the fly ash will accumulate and form blocks on the catalyst surface and in the corrugated channel, resulting in a reduction in the effective reaction area and a decrease in denitrification efficiency. At the same time, ammonium bisulfate (ABS) may also adhere to the catalyst surface in liquid form, exacerbating the blockage. This device realizes deep non-destructive testing and all-round quality control of the carrier's "surface-interior" and "geometry-function" by performing projection deformation analysis through upper and lower detection. At the same time, structural deformation and dust accumulation in the catalyst carrier channel will cause changes in the channel's transparency. This method obtains a lower projection image by reversely irradiating the catalyst carrier, calculates the single-channel transparency of several single channels in the catalyst carrier based on the grayscale value of the lower projection image, and determines whether the channel type of the several single channels is a normal channel, a partially blocked channel, or a completely blocked channel. Based on the experimental results that the pore deformation of a single channel will affect adjacent pores and cause the transparency of several adjacent pores to decrease step by step from near to far, for partially blocked channels and completely blocked channels, the grayscale value distribution trend of the corresponding area in the lower projection image is used to determine whether the corresponding single channel blockage is caused by structural deformation, thereby improving the accuracy of detecting fault types and avoiding the aggravation of faults caused by taking incorrect adjustment measures.

[0127] Generate an overall qualified evaluation value for any catalyst carrier detected based on the average grayscale value Ii of any single channel after excluding edge pixels, combined with the carrier detection evaluation value determined based on the image of the upper part of the carrier, and the actual weight of the catalyst carrier detected by the U-shaped scale;

[0128] Overall qualified evaluation value = 0.6× +0.3×support test evaluation value+0.1×weight deviation, where n is the total number of single channels in the catalyst support;

[0129] The weight deviation is the absolute value of the difference between the actual weight and the preset standard weight.

[0130] Furthermore, for the normal channels of the catalyst carrier, the normal channel type is determined by combining the infrared grayscale, and the normal channels are divided into slightly dust-accumulated channels, and the abnormality degree of a single channel is 0-0.2;

[0131] Fly ash particles (1-50 μm in diameter) form a micro-rough structure on the catalyst surface, causing diffuse reflection of incident light. Compared to a smooth catalyst surface (primarily specular reflection), the reflected light in the deposited area is stronger, resulting in an increased grayscale value captured by the camera.

[0132] When the fly ash layer accumulates to a critical thickness (usually >100μm), part of the light is absorbed, and the grayscale value changes from a peak to a decrease, forming a nonlinear response of "first rising and then falling".

[0133] Increase the illumination intensity of the bottom parallel light source according to the initial increase amplitude, use a dual-band (visible light + infrared) camera for differentiation, calculate the increase amplitude of the average grayscale value based on the change in the average grayscale value of the normal channel of the catalyst carrier before and after the change in illumination intensity, and detect the change in the infrared average grayscale to calculate the increase amplitude of the infrared average grayscale;

[0134] After increasing the light intensity, if the difference between the increase in the average gray value and the increase in the infrared average gray value is greater than the difference evaluation value, it is judged that there is slight dust accumulation in the normal channel of the catalyst carrier, and it is a slightly dusty channel;

[0135] During implementation, those skilled in the art can judge the dust accumulation tendency of the catalyst carrier based on the slight dust accumulation in the normal channels.

[0136] Among them, the difference evaluation value is 10%.

[0137] In this embodiment, the catalyst carrier has a cubic structure, and the bottom parallel light source will experience illumination attenuation in the edge area due to the optical cosine law.

[0138] like Figure 6 As shown, there is an obvious illumination difference between the middle and lower areas of the lower projected image and the edge areas. At the same time, the shading strips set along the periphery of the catalyst carrier cannot achieve the purpose of complete shielding. There is light leakage, which will also affect the lower projected image, resulting in illumination differences in different areas, thereby covering up grayscale changes or amplifying defects.

[0139] Specifically, even if the light source is ideally parallel, there is still a divergence angle (usually <0.5°). According to the fourth power cosine law, the effective light intensity in the edge area is Ie=I0 cos 4 θ, I0 is the central light intensity, θ is the incident angle of the light beam, and the incident angle of the light beam is the largest at the edge of the square bottom of the catalyst support;

[0140] Real dust accumulation in edge areas may be misjudged as normal (inadequate illumination masks grayscale changes), and slight discoloration in the center area may be magnified as a serious defect;

[0141] The process of the mobile lower part detection includes: obtaining a lower part projection image after moving the bottom parallel light source, analyzing the change of the lower part projection image after the movement, and verifying the original lower part detection result;

[0142] During implementation, the bottom parallel light source moves laterally by an initial moving distance, which is 1 / 2 of the side length of a single channel.

[0143] For the edge and center areas, verify the original bottom detection results based on the grayscale change amplitude (increase or decrease) and the difference in the projected image before and after the light source displacement;

[0144] Specifically, a lower projection image generated by the horizontal movement of the bottom parallel light source according to the initial movement distance is obtained, and grayscale detection is performed on the corresponding grid positions of the lower projection image generated after the movement of the partially blocked channel and the completely blocked channel determined before the movement, and the grayscale change amplitude of the corresponding grid positions before and after the movement is calculated;

[0145] For the corresponding grid positions of the lower projection images generated by the movement of the partially blocked channels and the completely blocked channels whose grayscale change amplitude is greater than or equal to the change value,

[0146] Performing a secondary single-channel permeability calculation on the partially blocked channel and the completely blocked channel at corresponding grid positions to determine whether the channel type of the single channel is a normal channel, a partially blocked channel, or a completely blocked channel;

[0147] Specifically, if the channel types of several single channels of the determined catalyst carrier change, the detection standard for the mobile lower part detection is increased, that is, the first preset value compared with the carrier detection evaluation value is increased.

[0148] The change value is 15%.

[0149] Specifically, due to the diffuse reflection of fly ash and the scattering angle of the light source, the actual dust accumulation in the edge area of ​​the catalyst carrier may be mistakenly judged as normal (insufficient illumination masks the grayscale change), and the slight discoloration in the central area may be magnified as a serious defect; for the normal channel, this method determines whether the normal channel is a slightly dusty channel based on the average grayscale value and infrared grayscale of the lower projection image by adjusting the illumination intensity of the parallel light source; for partially blocked channels and completely blocked channels, the channel type is secondary judged based on the grayscale change amplitude by horizontally moving the bottom parallel light source; this scheme performs optical dynamic reconstruction by horizontally moving the bottom parallel light source, and uses multi-angle projection differences to analyze the three-dimensional transparency characteristics of the catalyst carrier, which can significantly improve the accuracy and depth information acquisition capabilities of traditional static detection, and by moving the bottom parallel light source, reduces the risk of detection error caused by illumination attenuation in the edge area according to the changes in the lower projection image, and improves the edge detection sensitivity to the same level as the center area. At the same time, after determining that a misjudgment has occurred, the detection standard for mobile lower detection is increased, thereby increasing the accuracy and flexibility of detection.

[0150] Detecting a number n1 of changes in a number of single channels of the catalyst carrier whose channel types have changed, coordinate-izing the number of single channels whose channel types have changed, and dividing a change region corresponding to any single channel whose channel types have changed in the lower projection image;

[0151] Specifically, the center of the lower projection image is used as the origin of the coordinate system, and the lower projection image is divided into an Nx×Ny grid, where Nx and Ny are the number of rows and columns of a single channel;

[0152] Use (row, col) two-tuple positioning to locate and transform the coordinates of several single channels whose channel types have changed, and generate corresponding change coordinates (such as (3,5) represents the single channel in the 3rd row and 5th column).

[0153] According to the distribution of several single channels whose channel types are changed as reflected by the change coordinates, the clustering degree is calculated. The clustering degree = / [n1×(n1-1) / 2];

[0154] The clustering degree Di is the coordinate distance between any two single channels whose channel types have changed, and i represents any value in the number of changes n1;

[0155] When the clustering degree is less than or equal to a critical value, a plurality of single channels whose channel types are determined to have changed are concentratedly distributed in the lower projection image;

[0156] When the clustering degree is greater than a critical value, a plurality of single channels whose channel types are determined to have changed are discretely distributed in the lower projection image;

[0157] The critical value is equal to twice the side length of a single channel.

[0158] Calculate the change ratio of the changed region in the lower projection image, where the change ratio = the area of ​​the changed region / the area of ​​the lower projection image;

[0159] The degree of change includes the percentage of the change and the amount of the change;

[0160] The overall qualified evaluation value is corrected according to the distribution and the degree of change, and the corrected overall qualified evaluation value = the original overall qualified evaluation value - the basic correction item - the spatial distribution correction item × (1 + the change ratio);

[0161] Specifically, the basic correction item = / n1, spatial distribution correction term = β × clustering degree, where Wi is the channel position weight, 1.2 for edge channels and 1.1 for center channels, and Si is the degree of single channel anomaly;

[0162] Where β is the distribution factor, β=0.25 in concentrated distribution and β=0.15 in discrete distribution.

[0163] Specifically, in this embodiment, since the shading strips set along the periphery of the catalyst carrier cannot achieve the purpose of complete shielding, light leakage occurs, which will affect the lower projection image and cause illumination differences in different areas, thereby masking grayscale changes or amplifying defects. For example, the middle and lower areas of the lower projection image in the attached figure have obvious illumination differences compared to the edge areas. To address this technical problem, the present invention verifies the channel type according to the mobile lower detection, and then calculates the clustering response distribution of several single channels whose channel types have changed. It determines whether the several single channels whose channel types have changed are concentratedly distributed or discretely distributed in the lower projection image, and can analyze whether the cause of the change in channel type is caused by the illumination difference; and reflects the degree of change by the number of changes and the proportion of changes, and can analyze the degree of influence of the illumination difference on the detection accuracy. The calculated overall qualified evaluation value is corrected according to the distribution and degree of change of several single channels whose channel types have changed, thereby improving the detection accuracy of each single channel in the catalyst carrier, and solving the problem of misjudgment and missed judgment of the detection results due to the existence of illumination differences, which affects the accuracy of the detection results.

[0164] like Figure 2-5 As shown, a catalyst structure intelligent detection system based on multi-dimensional AI vision includes:

[0165] The carrier body has several support components 9 installed inside. The carrier body is equipped with a top plate 8, a sheet metal bracket 61, a visual camera 6, a fill light 7, a glass panel (not shown), a catalyst carrier 21, a glass base 11, a U-shaped scale 3, a parallel light source 4 and a moving component from top to bottom.

[0166] The visual camera 6 is suspended from the bottom of the top plate 8 by a sheet metal bracket 61, and is used to obtain the upper image and lower projection image of the catalyst carrier 2. The sheet metal bracket can adjust the height and lens swing angle of the visual camera and control its four-way lateral movement;

[0167] The sheet metal bracket includes a hanging plate 612 and a side plate 613. The hanging plate is provided with a plurality of hanging holes 611 for hanging the sheet metal bracket 61 on the top of the carrier body. The side plate is provided with a plurality of side holes 614 in the vertical direction. The side holes 614 are used to hang the industrial camera 6 so that it can slide up and down and adjust the angle.

[0168] The sheet metal bracket is a multi-angle adjustable bracket that supports four-way horizontal movement of the camera, front and back, left and right, and supports 100mm height adjustment of the camera up and down, and supports 180° free adjustment of the camera lens tilt angle;

[0169] The glass panel is arranged on the top of the support assembly 9 located on the top of the supporting frame, and is used to carry the lower projection image;

[0170] The glass base 11 is located at the bottom of the catalyst carrier 2 and is used to support the catalyst carrier 2 and limit the relative position relationship between the catalyst carrier 2 and the U-shaped scale 3;

[0171] The U-shaped scale 3 is fixedly connected to the inner side of the support assembly 9 at the bottom of the carrier body and is located at the bottom of the glass base 11, and is used to detect the actual weight of the catalyst carrier 2;

[0172] A rolling door 5 is provided on the side of the carrier body, and when the rolling door 5 is lowered, it forms a closed space with the carrier body;

[0173] The parallel light source 5 is plate-shaped and located at the bottom of the U-shaped scale 4. The moving component is arranged at the bottom of the parallel light source 5. The moving component includes a bracket 41, a first slide rail 42 and a second slide rail 43. The first slide rail 42 and the second slide rail 43 are arranged in parallel on both sides of the top of the bracket 41.

[0174] Specifically, the glass base is a highly transparent glass base, and the light source can directly penetrate the base (the frosted glass can be used as an optical diffuser to achieve a uniform light effect and reduce the illumination attenuation in the edge area). The U-shaped scale can measure the weight of the catalyst carrier and the position of the clamped glass base and avoid blocking the penetrating light of the bottom parallel light source. The parallel light source can fully illuminate the catalyst carrier through the sliding rail of the moving mechanism. After the rolling shutter door is started, a flexible rolling shutter can be lowered to provide a closed, light-proof detection space.

[0175] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0176] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A catalyst structure intelligent detection method based on multi-dimensional AI vision, characterized in that: include: Acquiring an image of an upper portion of a catalyst carrier in a closed space, and dividing the image into a plurality of grids according to the outer contours of a plurality of channels of the catalyst carrier; Determining surface cracking, straightness deviation, and angular deviation of the catalyst carrier to determine the detection status of the upper portion of the catalyst carrier; Determining a carrier detection evaluation value based on the pore characteristics of the carrier upper image, obtaining a lower projection image of the catalyst carrier based on a first comparison result obtained by comparing the carrier detection evaluation value with a standard preset value, and generating an overall qualified evaluation value of the catalyst carrier based on the grayscale value of the lower projection image, the carrier detection evaluation value, and the actual weight of the catalyst carrier; Calculating the transparency of each single channel in the catalyst carrier based on the grayscale value of the lower projection image, determining the channel type of each single channel, and distinguishing between dust accumulation and structural deformation based on the grayscale value distribution trend of the corresponding area of ​​the blocked channel in the lower projection image; Adjust the position and illumination intensity of the parallel light source, and verify the normal channel in the channel type according to the average gray value of the lower projection image and the increase of the infrared average gray value; After adjusting the position of the parallel light source, the grid positions of partially blocked channels and completely blocked channels are determined based on the grayscale variation of the lower projected image, and a secondary judgment is made on the channel types at the grid positions; Whether to adjust the inspection standard for the mobile lower part inspection is determined according to whether the channel type has changed, and the overall qualified evaluation value is corrected according to the distribution and degree of change of several single channels whose channel types have changed.

2. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 1 is characterized in that: The process of determining the conditions of straightness deviation and angular deviation includes, According to the side length of a single channel as the extraction length, the maximum value of a plurality of vertical distances within the extraction length is extracted as the straightness deviation; When the straightness deviation is greater than a first standard value, the straightness of the catalyst carrier is determined to be unqualified; when the straightness deviation is less than or equal to the first standard value, the straightness of the catalyst carrier is determined to be qualified; The included angle is calculated as the actual angle value using the vector dot product formula, and the absolute difference between the actual angle value and 90° is calculated as the angle deviation; When the angle deviation is less than or equal to the second standard value, the angle deviation of the catalyst carrier is determined to be qualified; when the angle deviation is greater than the second standard value, the angle deviation of the catalyst carrier is determined to be unqualified.

3. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 2 is characterized in that: The process of determining the detection status of the upper portion of the catalyst carrier includes: When the straightness and angle deviation are qualified and there is no surface cracking of the catalyst carrier, the upper portion of the catalyst carrier is determined to be in the first detection state, and the pore characteristic detection is performed to determine the carrier detection evaluation value; If either the straightness or the angle deviation is unqualified, or the catalyst carrier has surface cracks, the upper portion of the catalyst carrier is judged to be in the second detection state, and the fault type is output to alarm.

4. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 3 is characterized in that: The process of determining the carrier detection evaluation value includes, Detecting the number of through holes in the image on the carrier, calculating the effective through hole rate, detecting the hole area and hole perimeter of a plurality of holes in the image on the carrier, and calculating the shape factor of the holes; Calculating a carrier detection evaluation value according to the effective through-hole ratio and the shape factor of the hole, and comparing the carrier detection evaluation value with a first preset value and a second preset value; When the carrier detection evaluation value is less than the first preset value and greater than the second preset value, the comparison result is used as the first comparison result, and the catalyst carrier is subjected to a mobile lower part detection; When the carrier detection evaluation value is less than or equal to a second preset value, outputting the catalyst carrier as an unqualified carrier, issuing an alarm signal and marking the fault type as porosity deviation; In the process of performing mobile lower inspection on the catalyst carrier, an overall qualified evaluation value of the catalyst carrier is generated according to the grayscale value of the lower projection image, the carrier inspection evaluation value and the actual weight of the inspected catalyst carrier.

5. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 4 is characterized in that: The process of determining the channel type of each single channel includes, detecting the grayscale value of the lower projection image and calculating the single-channel transmittance of a plurality of single channels in the catalyst carrier; If the permeability of the single channel is greater than or equal to the first permeability, then the corresponding single channel in the catalyst carrier is judged to be a normal channel; If the permeability of the single channel is less than the first permeability and greater than or equal to the second permeability, then the corresponding single channel in the catalyst carrier is determined to be a partially blocked channel; If the single-channel permeability is less than the second permeability, it is determined that the corresponding single channel in the catalyst carrier is a completely blocked channel.

6. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 5 is characterized in that: Extracting grayscale value curves of the corresponding areas of the partially blocked and completely blocked single channels in the lower projection image along the axial direction, and calculating the grayscale change rate; When the grayscale change rate within a unit length exceeds the preset range, it is judged that the grayscale value drops drastically. The blockage of a partially blocked single channel is caused by dust accumulation, and no physical deformation occurs. When the grayscale change rate within the unit length is lower than the preset range, it is determined that the grayscale value is gradually decreasing, and the cause of the blockage of the partially blocked single channel is deformation, and the corresponding single channel undergoes physical deformation.

7. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 6, characterized in that: The process of determining whether a normal channel is a slightly dusty channel includes: Increase the illumination intensity of the bottom parallel light source according to the initial increase amplitude, and calculate the increase amplitude of the average grayscale value and the infrared average grayscale; After increasing the light intensity, if the difference between the increase in the average grayscale value and the increase in the infrared average grayscale is greater than the difference evaluation value, it is judged that there is slight dust accumulation in the normal channel of the catalyst carrier, and it is a slightly dusty channel.

8. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 7 is characterized in that: The process of adjusting the position of the parallel light source includes: The bottom parallel light source moves laterally by an initial distance, and the resulting lower projection image is obtained. The grayscale variation amplitudes of the partially blocked channel and the completely blocked channel at the corresponding grid positions before and after the movement are calculated. Determine the corresponding grid positions of the lower projection images generated after the partial blocking channel and the completely blocking channel with a grayscale change amplitude greater than or equal to the change value are moved.

9. The catalyst structure intelligent detection method based on multi-dimensional AI vision according to claim 8, characterized in that: Perform secondary single-channel transparency calculation at the corresponding grid position to verify the channel type of the single channel; If the channel types of the determined plurality of single channels of the catalyst carrier change, increasing the inspection standard for the mobile lower portion inspection, the inspection standard being the first preset value compared with the carrier inspection evaluation value; The process of correcting the overall qualified evaluation value includes coordinateizing a number of single channels whose channel types have changed to obtain corresponding change coordinates, and dividing a change area corresponding to any single channel whose channel type has changed in the lower projection image; The distribution of several single channels whose channel types have changed is determined according to the change coordinates, the change ratio of the change area in the lower projection image is calculated, and the overall qualified evaluation value is corrected.

10. A catalyst structure intelligent detection system based on multi-dimensional AI vision using the detection method according to any one of claims 1 to 9, characterized in that: include: The supporting frame has several supporting components inside, and the frame is equipped with a top plate, a sheet metal bracket, a visual camera, a fill light, a glass panel, a catalyst carrier, a glass base, a U-shaped scale, a parallel light source and a moving component from top to bottom; The visual camera is hung on the bottom of the top plate by a sheet metal bracket, and is used to obtain the upper image and lower projection image of the catalyst carrier. The sheet metal bracket can adjust the height and lens swing angle of the visual camera and control its lateral and longitudinal movement; The glass panel is arranged on the top of the supporting assembly located on the top of the supporting frame, and is used to support the lower projection image; The glass base is located at the bottom of the catalyst carrier and is used to support the catalyst carrier and limit the relative position relationship between the catalyst carrier and the U-shaped scale; The U-shaped scale is fixedly connected to the inner side of the support assembly at the bottom of the carrier body and is located at the bottom of the glass base, and is used to detect the actual weight of the catalyst carrier; A rolling door is provided on the side of the supporting frame, and when the rolling door is lowered, it forms a closed space with the supporting frame; The parallel light source is plate-shaped and located at the bottom of the U-shaped scale. The moving component is arranged at the bottom of the parallel light source. The moving component includes a bracket, a first slide rail and a second slide rail. The first slide rail and the second slide rail are arranged in parallel on both sides of the top of the bracket.

Images