Thermal crossing pipe expansion detection method and device based on AI image recognition technology

Abstract

The invention discloses an AI image recognition technology-based thermal crossing tube expansion detection method and device, and belongs to the field of production quality detection of heat exchange devices. An image acquisition device is arranged at one end of a heat exchanger tube plate, a first light source with adjustable brightness, color and position is arranged at the other end, and a second light source is additionally arranged to assist in detecting leakage and expansion tubes. Image data are collected and input into an AI heat crossing expansion tube detection model, the model calculates category similarity coefficients by integrating frame data of streaming media images, and heat exchange tubes with the coefficients exceeding a set threshold value and the abnormal category detection frequency exceeding a set category affiliation coefficient threshold value are marked through a display device to remind reinspection. Six types of material training models are provided, and a first light source adopts an area light source to improve detection efficiency. The detection device comprises a light source device, a camera shooting device, a display device and a control center, and the light source device and the camera shooting device can be driven by the plane moving device to move for accurate detection. According to the technology, the accuracy and efficiency of thermal crossing tube expansion detection are improved.

Description

Technical Field

[0001] The present invention relates to the field of production quality inspection of heat exchange devices, and particularly to a method and device for detecting the expansion of heat exchange tubes based on AI image recognition technology. Background Art

[0002] The basic structure of a heat exchanger is that there is a matrix-type heat transfer tube, also known as a heat exchange tube, inside the shell. The heat transfer tube is generally made of a metal with a high thermal conductivity such as copper, stainless steel, or titanium. The channel inside the heat transfer tube is called the tube side, and the channel between the heat transfer tube and the shell is called the shell side. The tube side and the shell side are in an isolated state. Its working principle is to make the cold and hot media flow through the tube side and the shell side respectively, and the heat exchange is achieved through the heat transfer tube. Hereinafter, the heat exchanger will be simply referred to as "heat exchange".

[0003] Inside the heat exchange, the matrix-arranged heat transfer tubes do not penetrate the shell one by one. Instead, a tube sheet is provided inside the shell, and installation holes adapted to the heat transfer tubes are preset on the tube sheet. The end of each heat transfer tube penetrates into the installation hole, and then through the tube expanding process, the end of the heat transfer tube is expanded to achieve a sealed connection with the tube sheet.

[0004] During the heat exchange assembly process, after the heat transfer tubes are inserted and expanded, it is necessary to check all the tube holes at the tube sheet end to prevent problems such as missed insertion, missed expansion, and insufficient expansion. Due to the large number and density of the tube rows on the tube sheet, traditional manual inspection is prone to visual fatigue, high work intensity, and extremely easy to miss inspections. Even through processes such as counters and helium inspections, it is still difficult to avoid omission problems. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a method and device for detecting the expansion of heat exchange tubes based on AI image recognition technology.

[0006] The specific technical solutions for the present invention to solve the above technical problems are as follows: A method for detecting the expansion of heat exchange tubes based on AI image recognition technology. An image acquisition device is provided at one end of the tube sheet to be inspected of the heat exchanger, and a first light source is provided at the other end of the heat exchanger. The first light source is used to provide illumination inside the heat exchange tube. The image acquisition device acquires the image data of the end of the tube sheet to be inspected and inputs the image data into a trained AI heat exchange tube expansion detection model. The AI heat exchange tube expansion detection model judges the expansion quality of each heat exchange tube in the image data. For the heat exchange tubes whose expansion quality may be abnormal, they are prominently marked through a display device to remind the staff to conduct a re-inspection.

[0007] Compared with the prior art, the present invention has the following beneficial effects: The AI thermal exchange tube expansion detection model determines whether the similarity between each thermal exchange tube in the image data and different category situations exceeds a set similarity threshold. If it exceeds, it is labeled as the corresponding category. For the same pipe orifice, for any category, if the frequency of detecting a similarity exceeding the set similarity threshold exceeds the set category attribution threshold, the pipe orifice is marked as belonging to the corresponding category and prominently marked by a display device. This solution uses AI image automatic detection technology. Based on a large number of manually marked normal and abnormal expansion tube materials, the AI thermal exchange tube expansion detection model is trained, enabling the AI thermal exchange tube expansion detection model to have the ability to identify abnormal expansion tubes, liberating the staff, assisting manual detection, reducing the requirements for operators, and the AI has stable and high-speed characteristics, ensuring the quality of the detection process and improving the operation speed.

[0008] Further, the AI thermal exchange tube expansion detection model is based on streaming media image data. For the thermal exchange tubes at the same position, considering the data of each frame in the streaming media image for comprehensive calculation to determine the final category attribution.

[0009] The beneficial effect of adopting the above further solution is that as the imaging device and the light source move, for the same thermal exchange tube, multiple photos with different lighting angles can be obtained in the streaming media. Based on the comprehensive judgment of the photos from multiple angles, the expansion tube situation can be judged more accurately, preventing misjudgment in the case of too weak or too strong light.

[0010] Further, the brightness and color of the first light source are adjustable to form a distinct contrast effect with the color of the thermal exchange tube itself.

[0011] The beneficial effect of adopting the above further solution is that adjusting the brightness and color of the first light source is to obtain an image with a distinct contrast effect at the observation end, improving the success rate of identification by the AI large model.

[0012] Further, the color of the first light source is selected as the complementary color of the color of the thermal exchange tube itself.

[0013] The beneficial effect of adopting the above further solution is that complementary colors can obtain a very good distinct contrast effect.

[0014] Further, the position of the image acquisition device or the position of the first light source is adjustable.

[0015] The beneficial effect of adopting the above further solution is that by adjusting the position of the image acquisition device or the position of the first light source, photos under different lighting angles can be obtained, so as to obtain image data more suitable for analysis by the AI large model and improve the judgment accuracy.

[0016] Further, a second light source is also provided on the side of the image acquisition device. The brightness of the second light source is lower than that of the first light source. The AI thermal cross-expansion tube detection model is based on whether there is a gap between the heat exchange tube and the tube sheet to assist in detecting the leaking expanded tubes.

[0017] The beneficial effect of adopting the above further solution is that another feature of the leaking expanded tube is that there is a relatively wide gap between the heat exchange tube and the tube sheet (this gap is occupied and disappears after tube expansion), but this feature is not as obvious as the feature inside the tube, so it can be used as a condition for auxiliary judgment and secondary verification.

[0018] Further, when providing training materials for the AI thermal cross-expansion tube detection model, six types of materials are included: The first type is the material of the tube inserted but not expanded. It is manually judged whether there is an arc segment or a circular contour line in the inner cavity of the tube end, and whether there is an obvious gap between the tube orifice and the tube hole; if the similarity between the tube end image and the first type of material is greater than X (X is the abnormal similarity threshold, which is a percentage, generally set between [50%, 60%]), it is marked as suspected of not being expanded, that is, it is considered that there is a probability of X that this tube hole is in the situation of the tube inserted but not expanded. The second type is the material of the leaking through tube. It is manually judged that there is no bright tube end image at the tube sheet end. If the similarity between the tube end image and the second type of material is greater than X (the empirical value of X is between [50%, 60%]), it is marked as suspected of leaking through. The third type is the material of insufficient tube expansion. It is manually judged that the roundness of the tube orifice is poor or there is still an obvious gap between the tube orifice and the tube hole. If the similarity between the tube end image and the third type of material is greater than X (the empirical value of X is between [50%, 60%]), it is marked as suspected of insufficient tube expansion. The fourth type is the material of the tube inserted and normal. The similarity between the tube end image and the normal image is judged; if the similarity is greater than Y (Y is the normal similarity threshold, the empirical value is between [60%, 70%], which is higher than X), and there is no situation of the first, second, and third types, it is marked as a normal tube orifice; otherwise, no marking is made, and this tube orifice is processed according to the sixth type of material. The fifth type is the material of tubes of other materials, which provides generalization learning training for the model, so that the model has the ability to identify the above situations for tubes of different materials. The sixth type is the untested situation where the tube orifice exists but the state cannot be identified. Generally, it is the outer tube orifice where the first light source cannot be detected. The position of this type of tube orifice can be used as a position reference during automatic detection to plan the moving path of the light source and the image acquisition device during the automatic detection process. Further, the first light source is a surface light source.

[0019] The beneficial effects of adopting the above further solution are that the surface light source has soft, uniform, and parallel light, a large light spreading area, and can inspect a large number of heat exchange tubes at one time, improving the detection efficiency.

[0020] The present invention also relates to a device for implementing the foregoing heat exchange penetration and expansion tube detection method based on AI image recognition technology, including a light source device, a camera device, a display device, and a control center. The light source device is used to provide illumination, the camera device is used to collect images, and the heat exchange tube to be detected is arranged between the light source device and the camera device. The images collected by the camera device are transmitted to the control center. The control center is built-in with a trained AI heat exchange penetration and expansion tube detection model, which is used to analyze and judge whether there are phenomena such as leaking penetration tubes, leaking expansion tubes, or insufficient expansion tubes at the ends of the heat exchange tubes in the images, and mark the positions with a relatively high probability of leaking penetration tubes, leaking expansion tubes, or insufficient expansion tubes, and display them to the staff through the display device.

[0021] Further, the light source device and the camera device are respectively arranged on independent planar moving devices. The planar moving device includes a base, a vertical moving device fixedly arranged on the base, a horizontal moving device driven by the vertical moving device to lift, and a carrying part driven by the horizontal moving device to move horizontally. The carrying part is used to carry the light source device or the camera device; the vertical moving device and the horizontal moving device are both electrically connected to the control center, and the control center is also used to control the planar moving device to carry the light source or the camera device to move in the plane.

[0022] The beneficial effects of the present invention are that the control center can separately control the driving of the camera device to move, separately control the driving of the light source device to move, or move both at the same time. When a certain heat exchange tube is not detected based on the AI heat exchange penetration and expansion tube detection model, the control center can, based on the coordinates of the target heat exchange tube, control the light source of the light source device and the camera of the camera device to reach the target position for special detection. If it is detected that there are situations such as leaking penetration tubes or leaking expansion tubes, the control center controls the camera and the light source to stop operating, sends a warning message through the display device, and waits for the staff to review and confirm before continuing. Description of the Drawings

[0023] Figure 1 It is a process flow chart of the partial assembly process of the heat exchanger; Figure 2 It is a schematic diagram of the heat exchange penetration and expansion tube detection method based on AI image recognition technology of the present invention; Figure 3 It is a schematic diagram of the structure of the first moving control part and the second moving control part in the heat exchange penetration and expansion tube detection device based on AI image recognition technology of the present invention.

[0024] Figure 4Schematic diagram of the original photo and the photo after annotation in the thermal cross-expansion tube detection method based on AI image recognition technology of the present invention; Figure 5 Schematic diagram of the original photo and the annotated photo in the thermal cross-expansion tube detection method based on AI image recognition technology of the present invention for detecting copper thermal exchange tubes; Figure 6 Schematic diagram of the situation where the tube expansion at the tube end is insufficient; Figure 7 Schematic diagram of the situation where a leak-through tube is detected at the tube end.

[0025] In the attached drawings, the list of component names represented by each reference numeral is as follows: 100, First light source; 200, Image acquisition device; 300, Display device; 400, Heat exchanger; 500, Vertical motion servo motor; 501, Horizontal motion servo motor. Detailed implementation manners

[0026] The principles and features of the present invention will be described below with reference to the attached drawings. The examples given are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0027] As Figure 1 shown, it is the thermal exchange tube assembly process. After the preliminary expansion is completed, an AI image detection device (embedded with a trained AI thermal cross-expansion tube detection model) automatically detects whether there are leak-through tubes, leak-expanded tubes, or insufficient tube expansion. Subsequently, significant prompts are given for the tube orifices that may be abnormal detected by the AI thermal cross-expansion tube detection model. Manual recheck is carried out on the positions of the tube holes that may have problems. If it is determined that all the tube penetrations or expansions are normal, the next assembly process is entered. Otherwise, the positions of the problematic tube holes are processed, and the above process is repeated. This process checks all the tube holes to ensure that all the tube hole operations are good and prevent problems such as leak-through tubes, leak-expanded tubes, and insufficient tube expansion. The AI thermal cross-expansion tube detection model automatically identifies the positions of the tube holes that may have problems. The operator refers to the position prompts and conducts a recheck. If there are problems, they are repaired.

[0028] As Figure 2 shown, a schematic diagram of a thermal cross-expansion tube detection method based on AI image recognition technology. An image acquisition device 200 is arranged at one end of the tube sheet to be inspected of the heat exchanger 400, and a first light source 100 is arranged at the other end of the heat exchanger 400. The first light source 100 is used to provide illumination for the inside of the thermal exchange tube. The image acquisition device 200 acquires the image data of the end of the tube sheet to be inspected and inputs the image data into the trained AI thermal cross-expansion tube detection model.

[0029] When training the AI thermal cross-expansion tube detection model, when providing training materials for the AI thermal cross-expansion tube detection model, six types of materials are included: The first type is the material of the pipe inserted but not expanded. It is judged whether there are arc segments or circular contour lines in the inner cavity of the pipe end of the heat exchange pipe, and whether there is an obvious gap between the pipe orifice of the heat exchange pipe and the pipe hole of the tube sheet. If the similarity between the pipe end image and the first type of material is greater than 50%, it is marked as suspected of not being expanded.

[0030] The second type is the material of the pipe with a leakage through the pipe. When there is a leakage through the pipe, there is no bright pipe end image at the tube sheet end, as Figure 7 shown. If the similarity between the pipe end image and the second type of material is greater than 50%, it is marked as suspected of having a leakage through the pipe.

[0031] The third type is the material of the pipe with insufficient expansion. When the expansion is insufficient, the roundness of the pipe orifice is poor or there is still an obvious gap between the pipe orifice and the pipe hole, as Figure 6 shown. If the similarity between the pipe end image and the third type of material is greater than 50%, it is marked as NG, and it is judged as suspected of insufficient expansion.

[0032] The fourth type is the material of the pipe inserted and normal. The similarity between the pipe end image and the normal image is judged. If the similarity is greater than 60% and there are no situations of the first, second, and third types, it is marked as OK, otherwise it is marked as NG.

[0033] The fifth type is the material of pipes made of other materials, which provides generalization learning training for the model, enabling the model to have the ability to identify the above situations for pipes made of different materials.

[0034] The sixth type is the situation where the pipe orifice is detected, but the state cannot be recognized. Mostly, due to the position of the first light source 100, the image acquisition device 200 cannot detect the pipe orifice of the light rays of the first light source 100. The position of such pipe orifices can be used to plan the movement paths of the first light source 100 and the image acquisition device 200 during the full-automatic detection process, so as to ensure that all pipe orifices are detected.

[0035] The above-mentioned traces of pipe expansion and gaps inside the pipe, etc., are features that can be relatively intuitively observed by the human eye, and are the judgment and compilation during the manual annotation of training materials. However, during the training process of the AI model, the features for judging whether the material is OK or NG are automatically extracted by the AI model, not limited to the features described above, and more features are automatically extracted and realized by the AI model.

[0036] Among them, the first light source 100 includes a light source part and a first movement control part. The first movement control part can control the first light source 100 to move freely in the vertical plane. According to different types of heat exchange pipes, the first light source 100 can be configured with different brightness and colors; the image acquisition device 200 includes a camera device and a second movement control part. The second movement control part can carry the camera device to move freely in the vertical plane; The first movement control part and the second movement control part are the same in terms of hardware, asFigure 3 As shown, both include a vertical motion servo motor 500 and a horizontal motion servo motor 501. The vertical motion servo motor 500 drives the horizontal motion servo motor 501 to move in the vertical direction, and the horizontal motion servo motor 501 drives the carried light source part or camera device to move in the horizontal direction. The two cooperate to achieve the free movement of the light source part or camera device in the vertical plane. The first movement control part and the second movement control part are communicatively connected to each other to carry the light source part and the camera device respectively to achieve synchronous matching movement of positions, ensuring that the camera device can collect good image data like a camera. Among them, the automatic detection path is jointly determined by two parts of logic. The first part is a preset S-shaped folding main path covering the entire tube panel surface, and the second part is a secondary path for supplementing local position fluctuations near the main path according to the positions of the locally undetected pipe orifices detected by the sixth type of material. The two path algorithms cooperate to ensure that every pipe orifice is detected.

[0037] The AI thermal exchange tube expansion and penetration detection model gives the similarity between each heat exchange tube in the image data and various situations, that is, the category similarity. The value of the category similarity ranges from 0 to 1. 0 represents that it does not belong to this category situation at all, and 1 means 100% belongs to this category situation. It is judged whether the category similarity exceeds the set similarity threshold. If it exceeds, the corresponding pipe orifice is marked as the corresponding category. If the frequency of a same pipe orifice being detected and classified into a certain category exceeds the set category attribution threshold, a prominent category identifier is added at the position of the corresponding pipe orifice through the display device 300. The abnormal similarity threshold corresponding to the abnormal situation is set between X = [50%, 60%]. In this example, it is specifically set to 50%, that is, for a single photo, when the AI thermal exchange tube expansion and penetration detection model believes that a certain pipe has a similarity exceeding 50% in the case of leakage of expansion, leakage of penetration or insufficient tube expansion, it will be judged as the corresponding abnormality; the normal similarity threshold corresponding to the normal feature is set between Y = [60%, 70%]. In this example, it is specifically set to 60%, that is, for a single photo, when the AI thermal exchange tube expansion and penetration detection model believes that a certain pipe has a similarity exceeding 60% in the normal situation, this pipe orifice will be judged as normal; when the AI thermal exchange tube expansion and penetration detection model is actually detecting, the recognition of multiple features is synchronized and parallel, and the priority of belonging to the abnormal classification is higher than that of belonging to the normal classification, that is: the detection priorities are the same, but only when finally judging which classification to belong to, the priority of belonging to the abnormal classification is higher than that of belonging to the normal classification. For example: for the same photo, assuming that there is any abnormal feature recognition similarity exceeding 50% and at the same time the normal feature similarity is also judged to exceed 60%, the whole picture will be preferentially judged as abnormal.

[0038] The AI heat exchange expansion tube detection model is based on streaming media image data. It collects 30 or more photos per second for detection. For the heat exchange tubes at the same position, it comprehensively calculates considering the data of each frame in the streaming media image. As the camera device and the light source move, for the same heat exchange tube, multiple photos with different lighting angles can be obtained in the streaming media. Based on the comprehensive judgment of the photos from multiple angles, when n = {4, 5, 6} (n is a parameter that can be set) of the same tube hole in 100 consecutive images are identified as belonging to a certain abnormal category, the system will suspend further detection and mark the corresponding tube hole to remind manual re-inspection. In this example, n takes 4. If in 100 consecutive images, two are judged as insufficient expansion tubes, three are judged as unperforated tubes, and four are judged as leaking expansion tubes, then the system will classify this pipe orifice as "suspected leaking expansion tube", suspend detection, and mark it at the pipe orifice to remind the staff to check. For the situation where less than 3 are judged as abnormal in 100 images, it is defaulted to be an accidental misidentification.

[0039] The detection accuracy of which category the pipe orifice belongs to in the AI heat exchange expansion tube detection model, that is, the detection accuracy of classification attribution, is obtained by accumulating two parts of probabilities: The first part is the probability automatically identified by the AI model, that is, the probability of a single pipe orifice being in a certain situation, which is determined by the similarity threshold. For example, specifically in this example, the similarity threshold for judging a certain pipe orifice in the figure to have a certain abnormal feature is X, and the similarity threshold for a normal feature is Y, to control the accuracy of grasping a certain category of situation; The second part is the manual experience probability, which controls the sensitivity of stopping detection and giving an alarm, and is determined by the category attribution threshold, that is, if a certain pipe orifice is judged to be in a certain abnormal category situation n times in 100 images, then stop detection and give an alarm; In summary, the final abnormal category attribution coefficient is , where , where x is the similarity degree (percentage) of the AI heat exchange expansion tube detection model in judging the pipe orifice and a specific abnormal category, X is the corresponding set similarity threshold, and the final alarm coefficient p for a certain situation is p = n / 100; Taking the leaking expansion situation as an example, set X = 60%, when n = 5, then p = 0.05, that is, when g(x) > 0.05, the logical judgment is that there is a leaking expansion situation at the pipe orifice; During actual detection, if the number of times of detecting the leaking expansion situation n = 6 times, that is, the probability of leaking expansion is 0.06, then it is considered a leaking expansion situation, suspend detection, and give a significant prompt for the abnormal pipe orifice. Since the recognition of abnormal features and normal features is carried out simultaneously, the system will take the situation of detecting 6 abnormal times first as the abnormal category to which the pipe orifice finally belongs.

[0040] On the premise of the foregoing settings, the threshold value Y for determining that the pipe orifice is a normal pipe orifice should be set to a value greater than X. Y is the corresponding set normal similarity threshold. For example, Y = 70%. That is, for a single photo, only when the similarity with the normal pipe orifice exceeds 70% will it be determined that the pipe orifice is a normal pipe orifice.

[0041] As Figure 4 shown, it is a comparison diagram with and without markings. Figure 4 In the right part, the heat exchange pipes that may be abnormal are framed in red, and the normal pipe orifice parts are framed in green.

[0042] As Figure 5 shown, it is a detection schematic diagram of the copper heat exchange pipe. In this figure, due to the problem of color contrast, the recognition accuracy of the original model decreases. Therefore, the color of the first light source 100 is preferably selected as the complementary color of the color of the heat exchange pipe itself. The complementary color can obtain better distinct contrast image data at the pipe orifice. Then, increase the collection of samples of this type of situation to enhance the generalization ability of the model.

[0043] A second light source is also provided on the side of the image acquisition device 200. The brightness of the second light source is lower than that of the first light source 100. Generally, natural light can be used. When the natural light is too strong, certain shielding is also required to ensure that the image acquisition device 200 can collect image data with clear features. The setting of the second light source enables the AI heat exchange pipe expansion detection model to assist in detecting the situation of leaking expanded pipes based on whether there is a gap between the heat exchange pipe and the tube sheet; in addition, for the black hole pipe orifice that cannot be irradiated by the first light source 100, the second light source can also enable the AI heat exchange pipe expansion detection model to detect the pipe orifice in an unrecognized state and assist in the planning of the movement path during automatic detection.

[0044] The control principle of the entire device is as follows: The imaging device continuously acquires images of the tube sheet end and sends them to the AI heat exchange pipe expansion detection model. After being recognized and processed by the AI heat exchange pipe expansion detection model, if a pipe orifice with possible problems is found, the vertical movement device and the horizontal movement device will stop synchronously, mark and display the position of the pipe orifice with possible problems on the display device 300 to remind the operator to perform further inspection and repair operations. After manual confirmation, the movement and detection can continue.

[0045] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for detecting thermal cross-expansion tubes based on AI image recognition technology, characterized in that, An image acquisition device is set at one end of the tube sheet to be inspected of the heat exchanger, and a first light source is set at the other end of the heat exchanger. The first light source is used to provide illumination inside the heat exchange tubes. The image acquisition device acquires the image data of the end of the tube sheet to be inspected and inputs the image data into the trained AI heat exchange tube expanding detection model. The AI heat exchange tube expanding detection model judges the expanding quality of each heat exchange tube in the image data. For the heat exchange tubes with possible abnormal expanding quality, they are prominently marked by a display device to remind the staff to conduct a re-inspection.

2. The method for detecting thermal cross-through expansion tubes based on AI image recognition technology according to claim 1, characterized in that Based on the streaming media image data, the AI heat exchange tube expanding detection model comprehensively calculates the data of each frame in the streaming media image for the heat exchange tubes at the same position to determine the final category attribution of each pipe orifice.

3. The method for detecting thermal cross-through expansion tubes based on AI image recognition technology according to claim 2, wherein The brightness and color of the first light source are adjustable to form a distinct contrast effect with the color of the heat exchange tubes themselves.

4. The method for detecting thermal cross-through expansion tubes based on AI image recognition technology according to claim 3, wherein The color of the first light source is selected as the complementary color of the color of the heat exchange tubes themselves.

5. The thermal cross-through expansion tube detection method based on AI image recognition technology according to any one of claims 1-4, characterized in that The position of the image acquisition device and / or the position of the first light source are adjustable.

6. The method for detecting heat exchange expansion tubes based on AI image recognition technology according to any one of claims 1-4, characterized in that, A second light source is also provided on the side of the image acquisition device. The brightness of the second light source is lower than that of the first light source. The AI heat exchange tube expanding detection model assists in detecting the situation of leaking expanded tubes based on whether there is a gap between the heat exchange tubes and the tube sheet.

7. The method for detecting heat exchange expansion tubes based on AI image recognition technology according to any one of claims 1-4, characterized in that, When providing training materials for the AI heat exchange tube expanding detection model, six types of materials are included: The first type is the material of tubes inserted but not expanded. It is judged whether there are arc line segments or circular contour lines in the inner cavity of the tube end, and whether there is an obvious gap between the pipe orifice and the pipe hole. If the similarity between the tube end image and the first type of material is greater than the abnormal similarity threshold X, it is marked as NG. The second type is the material of tubes not inserted. When the tube is not inserted, there is no bright tube end image at the tube sheet end. If the similarity between the tube end image and the second type of material is greater than the abnormal similarity threshold X, it is marked as NG. The third type is the material of insufficiently expanded tubes. When the expansion is insufficient, the roundness of the pipe orifice is poor or there is still an obvious gap between the pipe orifice and the pipe hole. If the similarity between the tube end image and the third type of material is greater than the abnormal similarity threshold X, it is marked as NG. The fourth type is the material of inserted and normal tubes. The similarity between the tube end image and the normal image is judged. If the similarity is greater than the normal similarity threshold Y, and Y is higher than X, and there are no situations of the first, second, and third types, it is marked as OK. The fifth type is the material of tubes made of other materials, which provides generalization learning training for the model to enable the model to have the ability to identify the above situations for tubes made of different materials. The sixth type is the uninspected situation where there are pipe orifices detected but the status cannot be identified, mostly the outer pipe orifices where the first light source cannot be detected, which is used to provide path guidance supplement during automatic detection.

8. The hot cross-through expansion tube detection method based on AI image recognition technology according to any one of claims 1-3, characterized in that The first light source is a surface light source.

9. An apparatus for implementing the method for detecting thermal cross-expansion tubes based on AI image recognition technology according to any one of claims 1-8, characterized in that, It includes a light source device, a camera device, a display device and a control center. The light source device is used to provide light, and the camera device is used to collect images. The heat exchange tube plate to be detected is arranged between the light source device and the camera device. The images collected by the camera device are transmitted to the control center. The control center is built-in with a trained AI heat exchange tube expansion detection model, which is used to analyze and judge whether there are phenomena such as leaking through tubes, leaking expanding tubes or insufficiently expanding tubes at the ends of the heat exchange tubes in the images, and mark the positions with a relatively high probability of leaking through tubes, leaking expanding tubes or insufficiently expanding tubes, and display them to the staff through the display device.

10. The device according to claim 9, characterized in that, The light source device and the camera device are respectively arranged on independent planar moving devices. The planar moving device includes a base, a vertical moving device fixedly arranged on the base, a horizontal moving device driven by the vertical moving device to lift, and a bearing part driven by the horizontal moving device to move horizontally. The bearing part is used to bear the light source device or the camera device; both the vertical moving device and the horizontal moving device are electrically connected to the control center, and the control center is also used to control the planar moving device to carry the light source device or the camera device to move in the plane.

Images