Corrosion defect infrared experiment device and corrosion defect infrared experiment process
By designing the corrosion defect infrared experimental device, simulating the corrosion defect status of the refining equipment, collecting infrared data, and establishing a self-identification model, it solves the accuracy and data acquisition problems of corrosion monitoring in the existing technology, and achieves efficient corrosion defect identification and management.
Patent Information
- Application Number
- CN202310826343.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-07
- Publication Date
- 2025-01-07
AI Technical Summary
The existing technology is difficult to accurately monitor the corrosion types, sizes and locations of refining equipment through effective means, and infrared thermal imaging technology is difficult to systematically collect corrosion data at multiple categories, multiple sizes and multiple temperatures, and it is impossible to establish a targeted infrared defect database. .
An infrared experimental device for corrosion defects was designed, including a corrosion defect model box and a heating device. Through precise temperature control and infrared monitoring, typical corrosion defect states are simulated, infrared data is collected, infrared defect database is established, and a self-identification model algorithm is built.
Accurate monitoring of corrosion defects is achieved, ensuring that the identification accuracy exceeds 90%, and providing petroleum and petrochemical enterprises with efficient equipment management and analysis tools.
Smart Images
Figure CN119269579A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of corrosion defect experimental devices for petroleum refining equipment, and in particular to a corrosion defect infrared experimental device and a corrosion defect infrared experimental process. Background Art
[0002] At present, the corrosion problems of refining and chemical equipment mainly include corrosion thinning, pitting, cracking, perforation, coking, etc., and there is a lack of targeted real-time monitoring methods. Traditional monitoring methods require the equipment to be shut down, and then conduct corrosion analysis through manual photography and sampling, or conduct online monitoring of in-service pipelines and other equipment through pulse eddy current scanning. The operation is relatively complicated and the process is cumbersome. In addition, due to factors such as different locations, different scales, different types, and different temperature media of equipment corrosion, it is impossible to accurately determine the specific internal corrosion location, area, and specific corrosion problem. There is a method of corrosion monitoring using infrared thermal imaging principles in the prior art. When corrosion problems occur in equipment, the surface temperature of the equipment changes due to material thinning or coking thickening. Infrared technology can accurately monitor changes in surface temperature, thereby intuitively reflecting the internal corrosion condition. However, due to differences in corrosion types, sizes and other parameters, the infrared thermal images fed back to the surface are all different. Therefore, it is necessary to establish a mapping relationship between typical corrosion problem types and infrared thermal images. By collecting infrared thermal images of corrosion problems at different scales, types, and medium temperatures, and using machine learning algorithms, an infrared database of typical corrosion problems is established to realize the machine self-diagnosis function of infrared image data.
[0003] Since infrared is used for temperature monitoring, one side of the experimental workpiece needs to be heated and the infrared thermal state of the other side needs to be monitored. In addition, the scale, type, medium temperature, etc. need to be changed. The experimental requirements are relatively stringent. Currently, there is no similar infrared corrosion data acquisition experimental device, so there is no infrared database for corrosion defects, and it is impossible to systematically and standardizedly collect infrared corrosion defect data, as well as accurately classify infrared images of corrosion defects. Summary of the invention
[0004] The purpose of the present invention is to solve the problem in the above-mentioned existing background technology that it is difficult to accurately know the type and size of corrosion through effective means during the operation of the equipment for corrosion defect monitoring; and it is difficult to simultaneously collect corrosion data of multiple categories, multiple sizes and multiple temperatures when using infrared thermal imaging for corrosion monitoring, and it is difficult to systematically establish an infrared defect database. A corrosion defect infrared experimental device and a corrosion defect infrared experimental process are provided, which can simulate the typical corrosion defect state of refining equipment, collect external surface infrared data through precise temperature control and infrared monitoring, and establish a corrosion defect infrared database, providing data support for the construction of an infrared defect self-identification model algorithm.
[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a corrosion defect infrared experimental device, including a corrosion defect sample box and a heating device, the corrosion defect sample box includes a top opening, a crack sample plate, a thinning sample plate, a pitting sample plate, a blank rough sample plate and a bottom plate, which are assembled and fixed to each other to form a box structure, wherein the crack sample plate, the thinning sample plate, the pitting sample plate, and the blank rough sample plate constitute the surrounding surfaces of the box body, the bottom plate constitutes the bottom of the box body, and the top opening constitutes the top of the box body; the corrosion defect surfaces of the crack sample plate, the thinning sample plate, the pitting sample plate and the blank rough sample plate face the inside of the box body; the heating device is threadedly connected to the top opening of the corrosion defect sample box, including a top cover, a temperature monitoring rod and a heating rod, the temperature monitoring rod and the heating rod pass through the top cover and are fixed, and the outer ends of the temperature monitoring rod and the heating rod are connected to the heating switch and the temperature controller through cables.
[0006] Furthermore, an infrared thermal imager is installed beside the corrosion defect pattern box to capture infrared images of the outer sides of the corrosion defect pattern box to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line to build an infrared database.
[0007] Furthermore, water or oil medium is injected into the corrosion defect sample box, and the liquid level is close to the top opening but does not exceed it.
[0008] Furthermore, a plurality of crack patterns are evenly distributed on the crack pattern plate, each column of crack patterns has the same depth and different widths, and each row of crack patterns has the same width and different depths.
[0009] Furthermore, the thinned pattern plate has a plurality of pattern areas with different thinning degrees distributed in a stepped manner, and each area has the same area.
[0010] Furthermore, a plurality of pitting patterns are evenly distributed on the pitting pattern plate, the circular pits of each column of pitting patterns have the same diameter and different depths, and the circular pits of each row of pitting patterns have the same depth and different diameters.
[0011] Furthermore, the inner side of the blank rough sample plate is frosted.
[0012] Furthermore, a calcium carbonate layer is evenly coated on the rough inner surface of the blank rough sample plate to simulate a coking state.
[0013] The technical solution of the present invention also provides an experimental process based on the above-mentioned corrosion defect infrared experimental device, which specifically includes:
[0014] First, add water or oil medium into the corrosion defect sample box;
[0015] Secondly, assemble the corrosion defect sample box with the heating device, and set the initial temperature through the heating switch and the temperature controller;
[0016] Secondly, the infrared thermal imager is set up beside the corrosion defect sample box, and the focus is adjusted and calibrated. The center of the infrared thermal imager lens is aligned with the center of the side of the corrosion defect sample box, and the distance from the corrosion defect sample box is less than 1 meter. The infrared thermal imager is moved to take infrared images of the outside of the crack sample plate, the thinned sample plate, the pitted sample plate, and the blank rough sample plate, and the corresponding infrared image data is formed. The data is transmitted to the data algorithm processing computer through the communication line to build an infrared database.
[0017] Then, train the model. First, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set. Perform model building training. After more than 10 iterations of training, the model building is completed when the error function meets the requirements.
[0018] Finally, after the training is completed, after inputting the infrared image of the unknown corrosion defect, it can accurately determine what kind of defect problem the image data belongs to.
[0019] Furthermore, the model training process is as follows: using the collected standard infrared image data as a model training set, using MATLAB for deep learning training, using the k-means clustering analysis method to establish cluster centers of different corrosion defects, identifying typical corrosion defects according to the eigenvalue state, and establishing an infrared corrosion defect recognition model.
[0020] Compared with the prior art, the present invention has the following beneficial effects: the infrared experimental device for corrosion defects described in the present invention can simulate four typical corrosion defect states, namely, corrosion thinning, pitting, cracks and coking, inside the device by using infrared thermal imaging technology. Through precise temperature control and multi-type and multi-size setting of defects, infrared images are collected from the outside of the experimental device in a standardized manner to form an infrared database of corrosion defects, and an infrared corrosion defect self-identification model algorithm is established to achieve precise monitoring of corrosion defects and ensure that the defect identification accuracy is greater than 90%. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 This is the overall structure diagram of the corrosion defect infrared experimental device of the present invention;
[0022] Figure 2 It is a structural schematic diagram of the corrosion defect sample box in the present invention;
[0023] Figure 3 The explosion view of the corrosion defect sample box in the present invention is Figure 1 ;
[0024] Figure 4 The explosion view of the corrosion defect sample box in the present invention is Figure 2 ;
[0025] Figure 5 It is a structural view of the heating device in the present invention;
[0026] Figure 6 It is the experimental flow chart of the present invention;
[0027] In the figure: 1. top opening, 2. crack sample, 3. thinning sample, 4. pitting sample, 5. blank rough sample, 6. bottom plate, 7. top cover, 8. temperature monitoring rod, 9. heating rod, 10. heating device, 11. corrosion defect sample box, 12. infrared thermal imager, 13. cable, 14. communication line. DETAILED DESCRIPTION
[0028] It should be noted that, in the description of the present invention, terms such as "upper", "lower", "left", "right", "front", "back", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings. They are relational words determined only for the convenience of describing the structural relationships of the various components in the present invention, and do not specifically mean that any component in the present invention must have a specific direction, be constructed and operated in a specific direction, and should not be understood as a limitation on the present invention.
[0029] It should also be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "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 directly connected or indirectly connected through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0030] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings:
[0031] Embodiment 1,
[0032] like Figures 1 to 5As shown, a corrosion defect infrared experimental device includes a corrosion defect sample box 11 and a heating device 10, wherein the corrosion defect sample box 11 includes a top opening 1, a crack sample plate 2, a thinning sample plate 3, a pitting sample plate 4, a blank rough sample plate 5 and a bottom plate 6, which are spliced and assembled with each other, and the side welding parts thereof are all chamfered at 45° to facilitate better welding and fixing into a box body structure, wherein the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4, and the blank rough sample plate 5 constitute the surrounding surfaces of the box body, the bottom plate 6 constitutes the bottom of the box body, the top opening 1 constitutes the top of the box body, and the top opening 1 is provided with an opening protruding upward, and the outer wall of the opening is provided with an external thread. Since the corrosion defect exists on the inner side, when welding and assembling, the corrosion defect surfaces of the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4 and the blank rough sample plate 5 face the inside of the box body. The heating device 10 is screwed to the top opening 1 of the corrosion defect sample box 11 through a thread, and includes a top cover 7, a temperature monitoring rod 8 and a heating rod 9. The temperature monitoring rod 8 and the heating rod 9 pass through the top cover 7 and are sealed and fixed. The outer ends of the temperature monitoring rod 8 and the heating rod 9 are connected to the heating switch and the temperature controller through a cable 13 to accurately control the temperature and control the opening and closing of the heating rod 9.
[0033] An infrared thermal imager 12 is installed beside the corrosion defect pattern box 11 to take infrared images of the outer sides of the corrosion defect pattern box 11 to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line 14 to build an infrared database.
[0034] The dimensions of each of the above corrosion type specimens are as follows:
[0035] 1) Cracked sample 2,
[0036] The size of the entire crack pattern plate 2 is 250mm×250mm×5mm, with three rows and four columns evenly distributed thereon, with a total of 12 crack patterns. The crack patterns in each column have the same depth and different widths, and the crack patterns in each row have the same width and different depths. The widths of the crack patterns are 1mm, 3mm, and 5mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0037] 2) Thinning sample plate 3,
[0038] The size of the entire thinned sample plate 3 is 250mm×250mm×5mm, and there are 4 sample areas with different thinning degrees distributed in a stepped manner. The area of each area is the same, which is 200mm×55mm, and the thinning depths are: 1mm, 2mm, 3mm, and 4mm respectively.
[0039] 3) Pit-etched sample 4,
[0040] The size of the entire pitting sample 4 is 250mm×250mm×5mm, with four rows and five columns evenly distributed thereon, with a total of 20 pitting patterns. The circular pits in each column of pitting patterns have the same diameter and different depths, and the depths of the pitting patterns in each row have the same depth and different circular pit diameters. The diameters of the circular pits of the pitting patterns are 5mm, 7mm, 9mm, 11mm, and 13mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0041] 4) Blank rough template 5,
[0042] The size of the entire blank rough sample plate 5 is 250 mm×250 mm×5 mm, and the inner side thereof is frosted, serving as a control group.
[0043] This embodiment also provides an experimental process based on the above corrosion defect infrared experimental device, combined with Figure 6 As shown, specifically including:
[0044] First, water or oil medium is added to the corrosion defect sample box 11, and its liquid level is close to the top port 1 but not higher, leaving a safety margin to prevent overflow. Here, the choice of oil or water as the medium depends on the specific experimental requirements; if the experiment requires the medium temperature to be within 0-80°C, water or oil can be selected as the medium according to the requirements; if the experimental temperature is required to be 80°C or above and not more than 250°C, oil needs to be selected as the medium.
[0045] Secondly, the corrosion defect sample box 11 is assembled with the heating device 10. The outer side wall of the top opening 1 of the corrosion defect sample box 11 has an external thread, and the top cover 7 in the heating device 10 has an internal thread matching it. The two are screwed together to fix the heating device 10 and the corrosion defect sample box 11 together; the temperature monitoring rod 8 extending outward from the heating device 10 and the heating rod 9 are connected to the heating switch and the temperature controller through the cable 13. The heating switch and the temperature controller are controlled by PLC programming. The initial temperature is set, for example, set to 53°C, the heating switch of the heating rod 9 is turned on, the heating rod 9 starts to heat, and the heating switch is automatically turned off when the temperature reaches 53°C; when the temperature of the heating rod 9 is less than 10% of the set temperature, the heating switch is automatically turned on again to heat the medium again.
[0046] Again, the infrared thermal imager 12 is set up beside the corrosion defect pattern box 11, and the focus is calibrated. The reflectivity of the monitoring material is set on the infrared thermal imager 12. Generally, the reflectivity of carbon steel is set to 0.98. The center of the lens of the infrared thermal imager 12 is aligned with the center of the side of the corrosion defect pattern box 11, and the distance from the corrosion defect pattern box 11 is less than 1 meter, so that it can clearly focus and shoot the infrared image of the outside of the corrosion defect pattern box 11. By moving the infrared thermal imager 12, infrared images of the outside of the cracked pattern plate 2, the thinned pattern plate 3, the pitting pattern plate 4, and the blank rough pattern plate 5 are taken respectively, and the corresponding infrared image data are formed respectively. The data is transmitted to the data algorithm processing computer through the communication line 14 to build an infrared database. According to the different infrared image data of each corrosion pattern collected, it is input into the infrared intelligent recognition analysis algorithm model software, and the image is deeply analyzed to train the accuracy of the model algorithm. Here, after the infrared image data is transmitted to the computer, a folder is established according to the defect type for storage.
[0047] Then, train the model; first, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set, and conduct model building training. After more than 10 iterations of training, when the error function meets the requirements, the model building is completed, and a random infrared image is selected to input the model for testing, and its corrosion defect classification results are observed. Here, the collected standard infrared image data is used as the model training set, and deep learning training is performed using matlab (a commercial mathematical software with advanced technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical calculation), and the k-means cluster analysis method is used to establish the cluster centers of different corrosion defects, and typical corrosion defects are identified according to the eigenvalue state, and an infrared corrosion defect recognition model is established.
[0048] Finally, the trained infrared defect intelligent recognition and analysis algorithm model can accurately judge the probability of what kind of defect problem the image data represents after inputting an infrared image of an unknown corrosion defect, providing a theoretical reference for equipment managers of petroleum and petrochemical enterprises, improving personnel inspection efficiency and analysis accuracy, and has broad market prospects.
[0049] Embodiment 2,
[0050] like Figures 1 to 5As shown, a corrosion defect infrared experimental device includes a corrosion defect sample box 11 and a heating device 10, wherein the corrosion defect sample box 11 includes a top opening 1, a crack sample plate 2, a thinning sample plate 3, a pitting sample plate 4, a blank rough sample plate 5 and a bottom plate 6, which are spliced and assembled with each other, and the side welding parts thereof are all chamfered at 45° to facilitate better welding and fixing into a box body structure, wherein the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4, and the blank rough sample plate 5 constitute the surrounding surfaces of the box body, the bottom plate 6 constitutes the bottom of the box body, the top opening 1 constitutes the top of the box body, and the top opening 1 is provided with an opening protruding upward, and the outer wall of the opening is provided with an external thread. Since the corrosion defect exists on the inner side, when welding and assembling, the corrosion defect surfaces of the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4 and the blank rough sample plate 5 face the inside of the box body. The heating device 10 is screwed to the top opening 1 of the corrosion defect sample box 11 through a thread, and includes a top cover 7, a temperature monitoring rod 8 and a heating rod 9. The temperature monitoring rod 8 and the heating rod 9 pass through the top cover 7 and are sealed and fixed. The outer ends of the temperature monitoring rod 8 and the heating rod 9 are connected to the heating switch and the temperature controller through a cable 13 to accurately control the temperature and control the opening and closing of the heating rod 9.
[0051] An infrared thermal imager 12 is installed beside the corrosion defect pattern box 11 to take infrared images of the outer sides of the corrosion defect pattern box 11 to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line 14 to build an infrared database.
[0052] The dimensions of each of the above corrosion type specimens are as follows:
[0053] 1) Cracked sample 2,
[0054] The size of the entire crack pattern plate 2 is 250mm×250mm×5mm, with three rows and four columns evenly distributed thereon, with a total of 12 crack patterns. The crack patterns in each column have the same depth and different widths, and the crack patterns in each row have the same width and different depths. The widths of the crack patterns are 1mm, 3mm, and 5mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0055] 2) Thinning sample plate 3,
[0056] The size of the entire thinned sample plate 3 is 250mm×250mm×5mm, and there are 4 sample areas with different thinning degrees distributed in a stepped manner. The area of each area is the same, which is 200mm×55mm, and the thinning depths are: 1mm, 2mm, 3mm, and 4mm respectively.
[0057] 3) Pit-etched sample 4,
[0058] The size of the entire pitting sample 4 is 250mm×250mm×5mm, with four rows and five columns evenly distributed thereon, with a total of 20 pitting patterns. The circular pits in each column of pitting patterns have the same diameter and different depths, and the depths of the pitting patterns in each row have the same depth and different circular pit diameters. The diameters of the circular pits of the pitting patterns are 5mm, 7mm, 9mm, 11mm, and 13mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0059] 4) Blank rough template 5,
[0060] The size of the entire blank rough sample plate 5 is 250mm×250mm×5mm, the inside of which is frosted, and the inner surface is evenly coated with a calcium carbonate coating to simulate the coking state for the experiment. The area size of the calcium carbonate layer applied here is 200mm×200mm, and it is applied in the center. The thickness can be selected in the range of: 1mm, 2mm, 3mm, 4mm, and 5mm.
[0061] This embodiment also provides an experimental process based on the above corrosion defect infrared experimental device, combined with Figure 6 As shown, specifically including:
[0062] First, water or oil medium is added to the corrosion defect sample box 11, and its liquid level is close to the top port 1 but not higher, leaving a safety margin to prevent overflow. Here, the choice of oil or water as the medium depends on the specific experimental requirements; if the experiment requires the medium temperature to be within 0-80°C, water or oil can be selected as the medium according to the requirements; if the experimental temperature is required to be 80°C or above and not more than 250°C, oil needs to be selected as the medium.
[0063] Secondly, the corrosion defect sample box 11 is assembled with the heating device 10. The outer side wall of the top opening 1 of the corrosion defect sample box 11 has an external thread, and the top cover 7 in the heating device 10 has an internal thread matching it. The two are screwed together to fix the heating device 10 and the corrosion defect sample box 11 together; the temperature monitoring rod 8 extending outward from the heating device 10 and the heating rod 9 are connected to the heating switch and the temperature controller through the cable 13. The heating switch and the temperature controller are controlled by PLC programming. The initial temperature is set, for example, set to 53°C, the heating switch of the heating rod 9 is turned on, the heating rod 9 starts to heat, and the heating switch is automatically turned off when the temperature reaches 53°C; when the temperature of the heating rod 9 is less than 10% of the set temperature, the heating switch is automatically turned on again to heat the medium again.
[0064] Again, the infrared thermal imager 12 is set up beside the corrosion defect pattern box 11, and the focus is calibrated. The reflectivity of the monitoring material is set on the infrared thermal imager 12. Generally, the reflectivity of carbon steel is set to 0.98. The center of the lens of the infrared thermal imager 12 is aligned with the center of the side of the corrosion defect pattern box 11, and the distance from the corrosion defect pattern box 11 is less than 1 meter, so that it can clearly focus and shoot the infrared image of the outside of the corrosion defect pattern box 11. By moving the infrared thermal imager 12, infrared images of the outside of the cracked pattern plate 2, the thinned pattern plate 3, the pitting pattern plate 4, and the blank rough pattern plate 5 are taken respectively, and the corresponding infrared image data are formed respectively. The data is transmitted to the data algorithm processing computer through the communication line 14 to build an infrared database. According to the different infrared image data of each corrosion pattern collected, it is input into the infrared intelligent recognition analysis algorithm model software, and the image is deeply analyzed to train the accuracy of the model algorithm. Here, after the infrared image data is transmitted to the computer, a folder is established according to the defect type for storage.
[0065] Then, train the model; first, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set, and conduct model building training. After more than 10 iterations of training, when the error function meets the requirements, the model building is completed, and a random infrared image is selected to input the model for testing, and its corrosion defect classification results are observed. Here, the collected standard infrared image data is used as the model training set, and deep learning training is performed using matlab (a commercial mathematical software with advanced technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical calculation), and the k-means cluster analysis method is used to establish the cluster centers of different corrosion defects, and typical corrosion defects are identified according to the eigenvalue state, and an infrared corrosion defect recognition model is established.
[0066] Finally, the trained infrared defect intelligent recognition and analysis algorithm model can accurately judge the probability of what kind of defect problem the image data represents after inputting an infrared image of an unknown corrosion defect, providing a theoretical reference for equipment managers of petroleum and petrochemical enterprises, improving personnel inspection efficiency and analysis accuracy, and has broad market prospects.
[0067] Embodiment 3,
[0068] like Figures 1 to 5As shown, a corrosion defect infrared experimental device includes a corrosion defect sample box 11 and a heating device 10, wherein the corrosion defect sample box 11 includes a top opening 1, a crack sample plate 2, a thinning sample plate 3, a pitting sample plate 4, a blank rough sample plate 5 and a bottom plate 6, which are spliced and assembled with each other, and the side welding parts thereof are all chamfered at 45° to facilitate better welding and fixing into a box body structure, wherein the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4, and the blank rough sample plate 5 constitute the surrounding surfaces of the box body, the bottom plate 6 constitutes the bottom of the box body, the top opening 1 constitutes the top of the box body, and the top opening 1 is provided with an opening protruding upward, and the outer wall of the opening is provided with an external thread. Since the corrosion defect exists on the inner side, when welding and assembling, the corrosion defect surfaces of the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4 and the blank rough sample plate 5 face the inside of the box body. The heating device 10 is screwed to the top opening 1 of the corrosion defect sample box 11 through a thread, and includes a top cover 7, a temperature monitoring rod 8 and a heating rod 9. The temperature monitoring rod 8 and the heating rod 9 pass through the top cover 7 and are sealed and fixed. The outer ends of the temperature monitoring rod 8 and the heating rod 9 are connected to the heating switch and the temperature controller through a cable 13 to accurately control the temperature and control the opening and closing of the heating rod 9.
[0069] An infrared thermal imager 12 is installed beside the corrosion defect pattern box 11 to take infrared images of the outer sides of the corrosion defect pattern box 11 to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line 14 to build an infrared database.
[0070] The dimensions of each of the above corrosion type specimens are as follows:
[0071] 1) Cracked sample 2,
[0072] The size of the entire crack pattern plate 2 is 250mm×250mm×5mm, with three rows and four columns evenly distributed thereon, with a total of 12 crack patterns. The crack patterns in each column have the same depth and different widths, and the crack patterns in each row have the same width and different depths. The widths of the crack patterns are 1mm, 3mm, and 5mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0073] 2) Thinning sample plate 3,
[0074] The size of the entire thinned sample plate 3 is 250mm×250mm×5mm, and there are 4 sample areas with different thinning degrees distributed in a stepped manner. The area of each area is the same, which is 200mm×55mm, and the thinning depths are: 1mm, 2mm, 3mm, and 4mm respectively.
[0075] 3) Pit-etched sample 4,
[0076] The overall size of the pitting corrosion sample 4 is 250mm × 250mm × 5mm. There are four rows and five columns evenly distributed on it, with a total of 20 pitting corrosion patterns. The round pits of each column of pitting corrosion patterns have the same diameter but different depths, and the depths of each row of pitting corrosion patterns are the same while the diameters of the round pits are different. The diameters of the round pits of the pitting corrosion patterns are: 5mm, 7mm, 9mm, 11mm, 13mm, and the depths are: 1mm, 2mm, 3mm, 4mm respectively.
[0077] 4) Blank rough sample 5,
[0078] The overall size of the blank rough sample 5 is 250mm × 250mm × 5mm. Its inner side is frosted. The inner surface is evenly divided into four "field"-shaped areas. Calcium carbonate coating is evenly applied in three adjacent areas to simulate the coking state for experiments. The area size of each area where the calcium carbonate layer is applied here is 100mm × 100mm, and it is applied in the center. The thicknesses from the upper left area, upper right area to the lower left area are: 2mm, 3mm, 4mm respectively. The bottom rightmost area is not coated with calcium carbonate.
[0079] This embodiment also provides an experimental procedure based on the above infrared experimental device for corrosion defects, combined with Figure 6 as shown, specifically including:
[0080] First, add water or oil medium into the corrosion defect sample box 11, and its liquid level is close to the top opening 1 but does not exceed it, leaving a safety margin to prevent overflow. Here, when choosing oil or water as the medium, it depends on the specific experimental requirements; if the experimental requirement for the medium temperature is within 0 - 80°C, water or oil can be selected according to the need; if the experimental temperature requirement is 80°C and above and not more than 250°C, then oil needs to be selected as the medium.
[0081] Secondly, assemble the corrosion defect sample box 11 with the heating device 10. The outer side wall of the opening of the top opening 1 of the corrosion defect sample box 11 has an external thread, and the top cover 7 of the heating device 10 has a matching internal thread. The two are fixed together by screwing; the temperature monitoring rod 8 and the heating rod 9 extending outward from the heating device 10 are connected to the heating switch and the temperature controller through the cable 13. The heating switch and the temperature controller are controlled by PLC programming. Set the initial temperature. For example, set it to 40°C, turn on the heating switch of the heating rod 9, and the heating rod 9 starts to heat. When the temperature reaches 40°C, the heating switch is automatically turned off; when the temperature of the heating rod 9 is less than 10% of the set temperature, the heating switch will be automatically turned on again to heat the medium again.
[0082] Again, the infrared thermal imager 12 is set up beside the corrosion defect pattern box 11, and the focus is calibrated. The reflectivity of the monitoring material is set on the infrared thermal imager 12. Generally, the reflectivity of carbon steel is set to 0.98. The center of the lens of the infrared thermal imager 12 is aligned with the center of the side of the corrosion defect pattern box 11, and the distance from the corrosion defect pattern box 11 is less than 1 meter, so that it can clearly focus and shoot the infrared image of the outside of the corrosion defect pattern box 11. By moving the infrared thermal imager 12, infrared images of the outside of the cracked pattern plate 2, the thinned pattern plate 3, the pitting pattern plate 4, and the blank rough pattern plate 5 are taken respectively, and the corresponding infrared image data are formed respectively. The data is transmitted to the data algorithm processing computer through the communication line 14 to build an infrared database. According to the different infrared image data of each corrosion pattern collected, it is input into the infrared intelligent recognition analysis algorithm model software, and the image is deeply analyzed to train the accuracy of the model algorithm. Here, after the infrared image data is transmitted to the computer, a folder is established according to the defect type for storage.
[0083] Then, train the model; first, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set, and conduct model building training. After more than 10 iterations of training, when the error function meets the requirements, the model building is completed, and a random infrared image is selected to input the model for testing, and its corrosion defect classification results are observed. Here, the collected standard infrared image data is used as the model training set, and deep learning training is performed using matlab (a commercial mathematical software with advanced technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical calculation), and the k-means cluster analysis method is used to establish the cluster centers of different corrosion defects, and typical corrosion defects are identified according to the eigenvalue state, and an infrared corrosion defect recognition model is established.
[0084] Finally, the trained infrared defect intelligent recognition and analysis algorithm model can accurately judge the probability of what kind of defect problem the image data represents after inputting an infrared image of an unknown corrosion defect, providing a theoretical reference for equipment managers of petroleum and petrochemical enterprises, improving personnel inspection efficiency and analysis accuracy, and has broad market prospects.
[0085] Embodiment 4,
[0086] like Figures 1 to 5As shown, a corrosion defect infrared experimental device includes a corrosion defect sample box 11 and a heating device 10, wherein the corrosion defect sample box 11 includes a top opening 1, a crack sample plate 2, a thinning sample plate 3, a pitting sample plate 4, a blank rough sample plate 5 and a bottom plate 6, which are spliced and assembled with each other, and the side welding parts thereof are all chamfered at 45° to facilitate better welding and fixing into a box body structure, wherein the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4, and the blank rough sample plate 5 constitute the surrounding surfaces of the box body, the bottom plate 6 constitutes the bottom of the box body, the top opening 1 constitutes the top of the box body, and the top opening 1 is provided with an opening protruding upward, and the outer wall of the opening is provided with an external thread. Since the corrosion defect exists on the inner side, when welding and assembling, the corrosion defect surfaces of the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4 and the blank rough sample plate 5 face the inside of the box body. The heating device 10 is screwed to the top opening 1 of the corrosion defect sample box 11 through a thread, and includes a top cover 7, a temperature monitoring rod 8 and a heating rod 9. The temperature monitoring rod 8 and the heating rod 9 pass through the top cover 7 and are sealed and fixed. The outer ends of the temperature monitoring rod 8 and the heating rod 9 are connected to the heating switch and the temperature controller through a cable 13 to accurately control the temperature and control the opening and closing of the heating rod 9.
[0087] An infrared thermal imager 12 is installed beside the corrosion defect pattern box 11 to take infrared images of the outer sides of the corrosion defect pattern box 11 to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line 14 to build an infrared database.
[0088] The dimensions of each of the above corrosion type specimens are as follows:
[0089] 1) Cracked sample 2,
[0090] The size of the entire crack pattern plate 2 is 250mm×250mm×5mm, with three rows and four columns evenly distributed thereon, with a total of 12 crack patterns. The crack patterns in each column have the same depth and different widths, and the crack patterns in each row have the same width and different depths. The widths of the crack patterns are 1mm, 3mm, and 5mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0091] 2) Thinning sample plate 3,
[0092] The size of the entire thinned sample plate 3 is 250mm×250mm×5mm, and there are 4 sample areas with different thinning degrees distributed in a stepped manner. The area of each area is the same, which is 200mm×55mm, and the thinning depths are: 1mm, 2mm, 3mm, and 4mm respectively.
[0093] 3) Pit-etched sample 4,
[0094] The size of the entire pitting sample 4 is 250mm×250mm×5mm, with four rows and five columns evenly distributed thereon, with a total of 20 pitting patterns. The circular pits in each column of pitting patterns have the same diameter and different depths, and the depths of the pitting patterns in each row have the same depth and different circular pit diameters. The diameters of the circular pits of the pitting patterns are 5mm, 7mm, 9mm, 11mm, and 13mm, respectively, and the depths are 1mm, 2mm, 3mm, and 4mm, respectively.
[0095] 4) Blank rough template 5,
[0096] The size of the entire blank rough sample plate 5 is 250mm×250mm×5mm, and the inner side is frosted. The inner surface is divided into two half areas, left and right. Calcium carbonate coating is evenly applied in one of the half areas to simulate the coking state for experiments. The area size of each area where the calcium carbonate layer is applied is 100mm×200mm, and it is applied in the center with a thickness of 3mm. The other half area is not coated with calcium carbonate coating.
[0097] This embodiment also provides an experimental process based on the above corrosion defect infrared experimental device, combined with Figure 6 As shown, specifically including:
[0098] First, water or oil medium is added to the corrosion defect sample box 11, and its liquid level is close to the top port 1 but not higher, leaving a safety margin to prevent overflow. Here, the choice of oil or water as the medium depends on the specific experimental requirements; if the experiment requires the medium temperature to be within 0-80°C, water or oil can be selected as the medium according to the requirements; if the experimental temperature is required to be 80°C or above and not more than 250°C, oil needs to be selected as the medium.
[0099] Secondly, the corrosion defect sample box 11 is assembled with the heating device 10. The outer side wall of the top opening 1 of the corrosion defect sample box 11 has an external thread, and the top cover 7 in the heating device 10 has an internal thread matching it. The two are screwed together to fix the heating device 10 and the corrosion defect sample box 11 together; the temperature monitoring rod 8 extending outward from the heating device 10 and the heating rod 9 are connected to the heating switch and the temperature controller through the cable 13. The heating switch and the temperature controller are controlled by PLC programming. The initial temperature is set, for example, set to 75°C, the heating switch of the heating rod 9 is turned on, the heating rod 9 starts to heat, and the heating switch is automatically turned off when the temperature reaches 75°C; when the temperature of the heating rod 9 is less than 10% of the set temperature, the heating switch is automatically turned on again to heat the medium again.
[0100] Again, the infrared thermal imager 12 is set up beside the corrosion defect pattern box 11, and the focus is calibrated. The reflectivity of the monitoring material is set on the infrared thermal imager 12. Generally, the reflectivity of carbon steel is set to 0.98. The center of the lens of the infrared thermal imager 12 is aligned with the center of the side of the corrosion defect pattern box 11, and the distance from the corrosion defect pattern box 11 is less than 1 meter, so that it can clearly focus and shoot the infrared image of the outside of the corrosion defect pattern box 11. By moving the infrared thermal imager 12, infrared images of the outside of the cracked pattern plate 2, the thinned pattern plate 3, the pitting pattern plate 4, and the blank rough pattern plate 5 are taken respectively, and the corresponding infrared image data are formed respectively. The data is transmitted to the data algorithm processing computer through the communication line 14 to build an infrared database. According to the different infrared image data of each corrosion pattern collected, it is input into the infrared intelligent recognition analysis algorithm model software, and the image is deeply analyzed to train the accuracy of the model algorithm. Here, after the infrared image data is transmitted to the computer, a folder is established according to the defect type for storage.
[0101] Then, train the model; first, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set, and conduct model building training. After more than 10 iterations of training, when the error function meets the requirements, the model building is completed, and a random infrared image is selected to input the model for testing, and its corrosion defect classification results are observed. Here, the collected standard infrared image data is used as the model training set, and deep learning training is performed using matlab (a commercial mathematical software with advanced technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical calculation), and the k-means cluster analysis method is used to establish the cluster centers of different corrosion defects, and typical corrosion defects are identified according to the eigenvalue state, and an infrared corrosion defect recognition model is established.
[0102] Finally, the trained infrared defect intelligent recognition and analysis algorithm model can accurately judge the probability of what kind of defect problem the image data represents after inputting an infrared image of an unknown corrosion defect, providing a theoretical reference for equipment managers of petroleum and petrochemical enterprises, improving personnel inspection efficiency and analysis accuracy, and has broad market prospects.
[0103] Embodiment 5,
[0104] like Figures 1 to 5As shown, a corrosion defect infrared experimental device includes a corrosion defect sample box 11 and a heating device 10, wherein the corrosion defect sample box 11 includes a top opening 1, a crack sample plate 2, a thinning sample plate 3, a pitting sample plate 4, a blank rough sample plate 5 and a bottom plate 6, which are spliced and assembled with each other, and the side welding parts thereof are all chamfered at 45° to facilitate better welding and fixing into a box body structure, wherein the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4, and the blank rough sample plate 5 constitute the surrounding surfaces of the box body, the bottom plate 6 constitutes the bottom of the box body, the top opening 1 constitutes the top of the box body, and the top opening 1 is provided with an opening protruding upward, and the outer wall of the opening is provided with an external thread. Since the corrosion defect exists on the inner side, when welding and assembling, the corrosion defect surfaces of the crack sample plate 2, the thinning sample plate 3, the pitting sample plate 4 and the blank rough sample plate 5 face the inside of the box body. The heating device 10 is screwed to the top opening 1 of the corrosion defect sample box 11 through a thread, and includes a top cover 7, a temperature monitoring rod 8 and a heating rod 9. The temperature monitoring rod 8 and the heating rod 9 pass through the top cover 7 and are sealed and fixed. The outer ends of the temperature monitoring rod 8 and the heating rod 9 are connected to the heating switch and the temperature controller through a cable 13 to accurately control the temperature and control the opening and closing of the heating rod 9.
[0105] An infrared thermal imager 12 is installed beside the corrosion defect pattern box 11 to take infrared images of the outer sides of the corrosion defect pattern box 11 to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line 14 to build an infrared database.
[0106] The dimensions of each of the above corrosion type specimens are as follows:
[0107] 1) Cracked sample 2,
[0108] The size of the entire crack pattern plate 2 is 250mm×250mm×5mm, with three rows and four columns evenly distributed thereon, with a total of 12 crack patterns. The crack patterns in each column have the same depth and different widths, and the crack patterns in each row have the same width and different depths. The widths of the crack patterns are 1mm, 2mm, and 3mm, respectively, and the depths are 1mm, 3mm, 5mm, and 7mm, respectively.
[0109] 2) Thinning sample plate 3,
[0110] The size of the entire thinned sample plate 3 is 250mm×250mm×5mm, and there are 4 sample areas with different thinning degrees distributed in a stepped manner. The area of each area is the same, which is 200mm×55mm, and the thinning depths are: 1mm, 2mm, 3mm, and 4mm respectively.
[0111] 3) Pit-etched sample 4,
[0112] The size of the entire pitting sample 4 is 250mm×250mm×5mm, with four rows and five columns evenly distributed thereon, with a total of 20 pitting patterns. The circular pits of each column of pitting patterns have the same diameter and different depths, and the depths of each row of pitting patterns have the same depth and different circular pit diameters. The diameters of the circular pits of the pitting patterns are 1mm, 2mm, 3mm, 4mm, and 5mm, respectively, and the depths are 1mm, 3mm, 5mm, and 7mm, respectively.
[0113] 4) Blank rough template 5,
[0114] The size of the entire blank rough sample plate 5 is 250mm×250mm×5mm, and the inside is frosted. The inner surface is divided into two half areas, left and right. Calcium carbonate coating is evenly applied in one of the half areas to simulate the coking state for experiments. The area size of each area where the calcium carbonate layer is applied is 100mm×200mm, and it is applied in the center with a thickness of 5mm. The other half area is not coated with calcium carbonate coating.
[0115] This embodiment also provides an experimental process based on the above corrosion defect infrared experimental device, combined with Figure 6 As shown, specifically including:
[0116] First, water or oil medium is added to the corrosion defect sample box 11, and its liquid level is close to the top port 1 but not higher, leaving a safety margin to prevent overflow. Here, the choice of oil or water as the medium depends on the specific experimental requirements; if the experiment requires the medium temperature to be within 0-80°C, water or oil can be selected as the medium according to the requirements; if the experimental temperature is required to be 80°C or above and not more than 250°C, oil needs to be selected as the medium.
[0117] Secondly, the corrosion defect sample box 11 is assembled with the heating device 10. The outer side wall of the top opening 1 of the corrosion defect sample box 11 has an external thread, and the top cover 7 in the heating device 10 has an internal thread matching it. The two are screwed together to fix the heating device 10 and the corrosion defect sample box 11 together; the temperature monitoring rod 8 extending outward from the heating device 10 and the heating rod 9 are connected to the heating switch and the temperature controller through the cable 13. The heating switch and the temperature controller are controlled by PLC programming. The initial temperature is set, for example, set to 90°C, the heating switch of the heating rod 9 is turned on, the heating rod 9 starts to heat, and when the temperature reaches 90°C, the heating switch is automatically turned off; when the temperature of the heating rod 9 is less than 10% of the set temperature, the heating switch is automatically turned on again to heat the medium again.
[0118] Again, the infrared thermal imager 12 is set up beside the corrosion defect pattern box 11, and the focus is calibrated. The reflectivity of the monitoring material is set on the infrared thermal imager 12. Generally, the reflectivity of carbon steel is set to 0.98. The center of the lens of the infrared thermal imager 12 is aligned with the center of the side of the corrosion defect pattern box 11, and the distance from the corrosion defect pattern box 11 is less than 1 meter, so that it can clearly focus and shoot the infrared image of the outside of the corrosion defect pattern box 11. By moving the infrared thermal imager 12, infrared images of the outside of the cracked pattern plate 2, the thinned pattern plate 3, the pitting pattern plate 4, and the blank rough pattern plate 5 are taken respectively, and the corresponding infrared image data are formed respectively. The data is transmitted to the data algorithm processing computer through the communication line 14 to build an infrared database. According to the different infrared image data of each corrosion pattern collected, it is input into the infrared intelligent recognition analysis algorithm model software, and the image is deeply analyzed to train the accuracy of the model algorithm. Here, after the infrared image data is transmitted to the computer, a folder is established according to the defect type for storage.
[0119] Then, train the model; first, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 3:1 between the test set and the validation set, and conduct model building training. After more than 20 iterations of training, when the error function meets the requirements, the model building is completed, and a random infrared image is selected to input the model for testing, and its corrosion defect classification results are observed. Here, the collected standard infrared image data is used as the model training set, and deep learning training is performed using matlab (a commercial mathematical software with advanced technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical calculation), and the k-means cluster analysis method is used to establish the cluster centers of different corrosion defects, and typical corrosion defects are identified according to the eigenvalue state, and an infrared corrosion defect recognition model is established.
[0120] Finally, the trained infrared defect intelligent recognition and analysis algorithm model can accurately judge the probability of what kind of defect problem the image data represents after inputting an infrared image of an unknown corrosion defect, providing a theoretical reference for equipment managers of petroleum and petrochemical enterprises, improving personnel inspection efficiency and analysis accuracy, and has broad market prospects.
[0121] In addition, in addition to training the corrosion defect recognition and classification model algorithm, the collected infrared data can also be used to collect infrared images of defects of different sizes to observe the changes in the temperature distribution morphology on the outside. This allows on-site front-line operators to understand the temperature morphology changes on the outside of the corrosion defect sample under different temperatures and sizes, and has the function of personnel training.
[0122] The data recorded in the specific experimental process can be recorded in the manner or format shown in the following table, and the name is actually changed according to a certain side of the corrosion defect pattern box 11 that the infrared thermal imager 12 is facing.
[0123]
[0124] Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by technicians in this technical field within the essential scope of the present invention should also fall within the protection scope of the present invention.
Claims
1. A corrosion defect infrared experimental device, characterized by: It includes a corrosion defect pattern box and a heating device, wherein the corrosion defect pattern box includes a top opening, a crack pattern plate, a thinned pattern plate, a pitted pattern plate, a blank rough pattern plate and a bottom plate, which are assembled and fixed to each other to form a box structure, wherein the crack pattern plate, the thinned pattern plate, the pitted pattern plate, and the blank rough pattern plate constitute the surrounding surfaces of the box body, the bottom plate constitutes the bottom of the box body, and the top opening constitutes the top of the box body; the corrosion defect surfaces of the crack pattern plate, the thinned pattern plate, the pitted pattern plate and the blank rough pattern plate face the inside of the box body; the heating device is threadedly connected to the top opening of the corrosion defect pattern box, and includes a top cover, a temperature monitoring rod and a heating rod, the temperature monitoring rod and the heating rod pass through the top cover and are fixed, and the outer ends of the temperature monitoring rod and the heating rod are connected to the heating switch and the temperature controller through cables.
2. The corrosion defect infrared experimental device according to claim 1, characterized in that: An infrared thermal imager is installed beside the corrosion defect pattern box to take infrared images of the outer sides of the corrosion defect pattern box to form infrared image data, and the data is transmitted to a data algorithm processing computer through a communication line to build an infrared database.
3. The corrosion defect infrared experimental device according to claim 2, characterized in that: Water or oil medium is injected into the corrosion defect sample box, and the liquid level is close to the top opening but does not exceed it.
4. The corrosion defect infrared experimental device according to claim 3 is characterized by: A plurality of crack patterns are evenly distributed on the crack pattern plate, each column of crack patterns has the same depth and different widths, and each row of crack patterns has the same width and different depths.
5. The corrosion defect infrared experimental device according to claim 3, characterized in that: The thinned pattern plate is provided with a plurality of pattern areas with different thinning degrees distributed in a stepped manner, and each area has the same area.
6. The corrosion defect infrared experimental device according to claim 3, characterized in that: A plurality of pitting patterns are evenly distributed on the pitting pattern plate, the circular pits of each column of pitting patterns have the same diameter and different depths, and the circular pits of each row of pitting patterns have the same depth and different diameters.
7. The corrosion defect infrared experimental device according to claim 3, characterized in that: The inner side of the blank rough style plate is frosted.
8. The corrosion defect infrared experimental device according to claim 3 is characterized by: The rough inner surface of the blank rough sample plate is evenly coated with a calcium carbonate layer to simulate a coking state.
9. An experimental process based on the corrosion defect infrared experimental device according to any one of claims 1 to 8, characterized in that: First, add water or oil medium into the corrosion defect sample box; Secondly, assemble the corrosion defect sample box with the heating device, and set the initial temperature through the heating switch and the temperature controller; Secondly, the infrared thermal imager is set up beside the corrosion defect sample box, and the focus is adjusted and calibrated. The center of the infrared thermal imager lens is aligned with the center of the side of the corrosion defect sample box, and the distance from the corrosion defect sample box is less than 1 meter. The infrared thermal imager is moved to take infrared images of the outside of the crack sample plate, the thinned sample plate, the pitted sample plate, and the blank rough sample plate, and the corresponding infrared image data is formed. The data is transmitted to the data algorithm processing computer through the communication line to build an infrared database. Then, train the model. First, randomly divide the collected infrared image data into a test set folder and a validation set folder according to the ratio of 5:1 between the test set and the validation set. Perform model building training. After more than 10 iterations of training, the model building is completed when the error function meets the requirements. Finally, after the training is completed, after inputting the infrared image of the unknown corrosion defect, it can accurately determine what kind of defect problem the image data belongs to.
10. The experimental process based on the corrosion defect infrared experimental device according to claim 9 is characterized in that: The training model process is as follows: using the collected standard infrared image data as the model training set, using MATLAB for deep learning training, using the k-means clustering analysis method to establish cluster centers of different corrosion defects, identifying typical corrosion defects according to the eigenvalue state, and establishing an infrared corrosion defect recognition model.