Photovoltaic support online quality inspection system and detection method based on machine vision

Abstract

The invention discloses a photovoltaic support online quality inspection system and method based on machine vision. The photovoltaic support online quality inspection system comprises a welding seam detection module, a punching welding seam joint detection module and a forming quality detection module. The welding seam detection module is installed in the feeding machine frame and used for collecting image information of a welding seam on a material belt, the punching and welding seam combined detection module is used for collecting the size between a punched hole and the welding seam, and the punching and welding seam combined detection module is connected with the cold bending and cutting unit, so that the punched material belt can be bent in the width direction, and the punching and welding seam combined detection module is connected with the cold bending and cutting unit. And whether the surface quality and the bending shape of the photovoltaic support beam meet the requirements or not is detected through the forming quality detection module. During use, the artificial intelligence model judges whether products reach the standard or not according to image information collected by the welding seam detection module, the punching welding seam joint detection module and the forming quality detection module, and after a judgment result is obtained, unqualified products are removed in a centralized mode, and the artificial intelligence model is optimized in a reinforcement learning mode.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial online detection equipment, in particular to a photovoltaic support online quality inspection system and method based on machine vision. BACKGROUND

[0002] With the continuous development of the photovoltaic industry, the demand for supporting photovoltaic racks is also increasing. Although the production technology of photovoltaic racks is not difficult, the market demand is wide, and a large number of qualified products are needed to match the photovoltaic industry.

[0003] Therefore, how to quickly and efficiently produce photovoltaic racks is particularly important. At present, the existing technology still uses traditional stamping process to produce support beams. Although the production speed is fast and the working hours are less, the product quality is difficult to guarantee, and the products often have inconsistent precision, poor assembly precision, large shape deviation, and many surface defects.

[0004] Therefore, the applicant proposes a new invention, which introduces machine vision technology to transform the production line, so that the product can be inspected online. Thus, the quality of photovoltaic support products can be significantly improved without increasing working hours, and the scrap rate can be reduced. SUMMARY

[0005] Therefore, the present application provides a photovoltaic support online quality inspection system and method based on machine vision to solve the problem of unstable product quality control caused by backward stamping production process in the prior art.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] According to the first aspect of the present application;

[0008] The present application discloses a photovoltaic support online quality inspection system based on machine vision, which continuously detects the machining precision online during the processing of the material belt to process photovoltaic supports, comprising:

[0009] A weld detection module is installed at the bottom of the material running frame, and a pair of leveling devices is arranged on the front and rear sides of the weld detection module, and a displacement encoder sensor is installed on the leveling device;

[0010] A combined punching and weld detection module is connected to the weld detection module through a punching unit at the first end, and the combined punching and weld detection module is connected to the cold bending and cutting unit at the first end at the tail end.

[0011] A forming quality detection module is arranged at the tail end of the cold bending and cutting unit, and the forming quality detection module and the combined punching and weld detection module are respectively installed on two material running buffer devices.

[0012] The punching unit and the cold bending cutting unit are connected with the lower computer circuit, the welding seam detection module, the punching and welding seam joint detection module, the forming quality detection module and the displacement encoding sensor are connected with the upper computer signal, and the upper computer is connected with the lower computer in network communication.

[0013] Further, the welding seam detection module, the punching and welding seam joint detection module and the forming quality detection module have the same structure.

[0014] The welding seam detection module comprises an industrial camera and a strip-shaped light source, the strip-shaped light source is arranged on the outside of the industrial camera, and the industrial camera is connected with the upper computer signal.

[0015] Further, the leveling device comprises:

[0016] A pair of leveling bases are slidably arranged on the guide rail, the guide rail is arranged in the material feeding rack, and the pair of leveling bases and the adjusting screw form a screw pair and move towards each other along the guide rail under the action of the adjusting screw.

[0017] A material feeding roller group is rotatably arranged at the front end of the leveling base, and the end of the material feeding roller group is mechanically connected with the displacement encoding sensor, and the front end of the leveling base is provided with a leveling roller group.

[0018] Further, the punching unit comprises a punch, a material guide frame, an anvil plate and a centering component, the material feeding rack is connected with the material guide frame, the material guide frame is provided with the punch at the top and the anvil plate at the bottom, the anvil plate is aligned with the punch and punches holes on the material belt by the punch, and the anvil plate is provided with the centering component on both sides, and the centering component is adapted to guide the material belt to pass through the anvil plate.

[0019] Further, the material feeding buffer device comprises a base frame, guide rollers, a detection support and dampers, one of the punching and welding seam joint detection module or the forming quality detection module is installed on the detection support, the detection support is located at the top of the base frame, a plurality of guide rollers are horizontally arranged on the base frame, any two adjacent guide rollers are connected by synchronous belts, and any guide roller is frictionally connected with a damper, and the damper is arranged on the base frame.

[0020] Further, the cold bending cutting unit comprises:

[0021] A follow-up cutter is fixed at the top of a protective cover, and the bottom of the protective cover is connected with the machine table, wherein the follow-up cutter is connected with the lower computer circuit.

[0022] The bottom of the progressive bending machine is bolted with an anvil, a plurality of pressure rollers are rotatably arranged in the progressive bending machine, and the plurality of pressure rollers are adapted to gradually extrude the material belt on the anvil.

[0023] The present application has the following advantages:

[0024] The position of the weld is detected by the weld detection module, and the obtained data information is input into the lower computer for calculation to obtain data for punching the material belt by the punching unit, and then the quality is detected again by the punching weld joint detection module, and then the obtained data is sent to the cold bending cutting unit for bending and displacement cutting processing, and finally the surface quality of the product is checked by the forming quality detection module, and the product is classified. Compared with the prior art, the technical scheme disclosed by the present application can detect data in real time on the product production line and correct the production process, thereby significantly improving the product quality without reducing the production capacity and reducing the labor cost.

[0025] According to a second aspect of the application;

[0026] The present application discloses a detection method applied to the machine vision-based photovoltaic support online quality inspection system as described above, comprising:

[0027] After the production line starts production, the first and last ends of the multiple steel material belts need to be welded at the welding station and sent to the weld detection module through the first material feeding buffer device for detection;

[0028] The leveling device flattens the material belt and drives the material belt to feed, and in the feeding process, the punching weld joint detection module is used for image acquisition and weld position recognition, and the displacement encoding sensor is used to obtain the material belt feeding speed;

[0029] The material belt is sent to the punching unit for punching processing, and after the punching is completed, it is sent to the second material feeding buffer device to keep the material belt feeding in a uniform motion state and enter the cold bending cutting unit;

[0030] The material belt is gradually extruded and plastically formed by the cold bending cutting unit and completes the displacement cutting, forming a photovoltaic support beam, and the photovoltaic support beam is sent to the forming quality detection module;

[0031] The artificial intelligence model is used as the upper computer, and the PLC controller is used as the lower computer, and according to the image information collected by the weld detection module, the punching weld joint detection module and the forming quality detection module, it is judged whether the product meets the standard;

[0032] The qualified products are stacked and packaged, the unqualified products are removed and recycled, and image processing, cold bending forming size measurement and surface defect recognition are performed to optimize the artificial intelligence model through reinforcement learning.

[0033] Further, the process of detecting the weld in the weld detection module comprises:

[0034] The displacement encoder sensor determines the movement status of the steel strip and provides a trigger signal to the industrial camera, which identifies the welding position of the strip in real time; this information is then transmitted to the PLC control system of the punching station.

[0035] The host computer combines spatial location information, communication duration, and material advance distance to calculate the relative position data of the weld seam in front of the punching station from the punching station, and transmits the data to the slave computer.

[0036] After receiving the weld position data, the programmable logic controller calculates the tail gauge size using the follow-up encoder and distance data. Since the delayed punching distance is equal to the material tail gauge distance, the displacement punching process is automatically completed.

[0037] Furthermore, the displacement punching process includes:

[0038] The industrial camera identifies the weld seam information and transmits the location to the programmable logic controller system of the lower-level machine, while adding a 10mm safety margin for punching and cutting;

[0039] The programmable logic controller calculates the distance between the weld position and the starting end of the preset punching position, and calculates the ratio M and remainder N of the distance to the profile length. If M is less than 1, the lower computer controls the punching unit to stop punching. If M is greater than 1, after punching M profiles normally, the punching starts from the new preset profile starting end with 10mm added to the weld position as the reference.

[0040] Furthermore, the molding quality inspection module's inspection process includes:

[0041] The vision camera detects surface cracks and defects in the finished product from the left, right and bottom directions to determine whether the coating quality is up to standard. If it is not up to standard, the waste material is rejected.

[0042] After inspecting the material strip bending, determine whether the cold bending position is qualified based on the inward fold length. If it is not qualified, discard the waste material.

[0043] The quality of cold bending is judged by the degree of symmetry of the short side of the bend. If the small deviation exceeds three times or a major deviation occurs, the line should be stopped for inspection in time.

[0044] The present invention has the following advantages:

[0045] This invention discloses a detection method that uses an artificial intelligence model as the host computer and a PLC controller as the slave computer. Based on the image information collected by the weld detection module, the punched weld joint detection module, and the forming quality detection module, the method determines whether the product meets the standards. At the same time, it collects image information of rejected defective products and uses reinforcement learning to correct the artificial intelligence model, thereby improving the efficiency and accuracy of quality inspection and ensuring that the equipment can operate continuously and stably. Attached Figure Description

[0046] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0047] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0048] Figure 1 A three-dimensional view of the online quality inspection system for photovoltaic brackets based on machine vision provided by the present invention;

[0049] Figure 2 A perspective view of the molding quality inspection module provided by the present invention;

[0050] Figure 3 A perspective view of the cold bending and cutting unit provided by the present invention;

[0051] Figure 4 A perspective view of the follower cutter provided by the present invention;

[0052] Figure 5 A perspective view of the material feeding buffer device provided by the present invention;

[0053] Figure 6 A perspective view of the punching unit provided by the present invention;

[0054] Figure 7 A perspective view of the leveling device provided by the present invention;

[0055] Figure 8 A perspective view of the displacement encoding sensor provided by the present invention;

[0056] Figure 9 A perspective view of the leveling roller assembly provided by the present invention;

[0057] Figure 10 The following is a flowchart of the detection method steps provided by the present invention;

[0058] In the figure: 1 welding seam detection module; 11 industrial camera; 12 bar light source; 2 leveling device; 21 leveling base; 22 guide rail; 23 adjusting screw; 24 leveling roller group; 25 feeding roller group; 3 combined detection module of punching and welding seam; 4 punching unit; 41 puncher; 42 material guide frame; 43 anvil table; 44 centering component; 5 forming quality detection module; 6 cold bending and cutting unit; 61 follow-up cutter; 62 protective cover; 63 machine table; 64 anvil iron; 65 progressive bender; 66 pressure roller; 7 feeding buffer device; 71 underframe; 72 guide roller; 73 detection support; 74 damper; 8 displacement encoding sensor; 9 feeding frame. DETAILED DESCRIPTION

[0059] The embodiments of the present application will be described in detail with specific embodiments, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0060] Reference should also be made to Figures 1-9The application discloses a photovoltaic support online quality inspection system based on machine vision, which can continuously detect the machining precision when a material belt is used to process a photovoltaic support, and main components of the system include a welding seam detection module 1, a punching and welding seam joint detection module 3 and a forming quality detection module 5. The welding seam detection module 1 is installed in a material walking rack 9, a pair of leveling devices 2 are symmetrically arranged in the material walking rack 9, and the welding seam detection module 1 is located between the pair of leveling devices 2 and used for collecting image information of welding seams on the material belt. A displacement encoding sensor 8 is further installed on each leveling device 2 and used for detecting the moving distance of the material belt and feeding the material belt into a punching unit 4 to punch holes, and the material belt is fed into the punching and welding seam joint detection module 3 after being punched, and the punching and welding seam joint detection module 3 is mainly used for detecting whether the distance between the welding seams and the punching holes and the shape and position of the punching holes are out of tolerance. Specifically, a first end of the punching and welding seam joint detection module 3 is connected with the welding seam detection module 1 through the punching unit 4, and a tail end of the punching and welding seam joint detection module 3 is connected with a first end of a cold bending and cutting unit 6. The cold bending and cutting unit 6 can bend the material belt along the width direction after the punching treatment, so that the photovoltaic support beam is basically formed and cut according to needs, and a tail end of the cold bending and cutting unit 6 is installed with the forming quality detection module 5, which is used for detecting whether the surface quality and the bending shape of the processed photovoltaic support beam meet the requirements. In the embodiment, the forming quality detection module 5 and the punching and welding seam joint detection module 3 are respectively installed on two material walking buffer devices 7, and the material walking buffer devices 7 are mainly used for realizing smooth transition of the material belt in each machining module and detection module, and realizing that the material walking speed of the material belt is kept balanced, so that deviation in the machining or detection process caused by unstable material walking speed is reduced.

[0061] In some embodiments, the software part of the photovoltaic support online quality inspection system mainly realizes closed-loop control by issuing control instructions through an upper computer, controlling the hardware part through a lower computer, and collecting machining data and feeding back to the upper computer. Specifically, the punching unit 4 and the cold bending and cutting unit 6 are connected with the lower computer circuit, the welding seam detection module 1, the punching and welding seam joint detection module 3, the forming quality detection module 5 and the displacement encoding sensor 8 are connected with the upper computer signal, and the upper computer and the lower computer are connected through network communication.

[0062] In some embodiments, the machining data is mainly collected through the welding seam detection module 1, the punching and welding seam joint detection module 3 and the forming quality detection module 5, and the welding seam detection module 1, the punching and welding seam joint detection module 3 and the forming quality detection module 5 all realize product quality control in each process through machine vision detection. Specifically, the welding seam detection module 1 includes an industrial camera 11 and a strip light source 12, the strip light source 12 is arranged outside the industrial camera 11 and irradiates on the material belt to improve the definition of image collection, and the industrial camera 11 is connected with the upper computer signal and can directly input detection data to the upper computer for processing.

[0063] In some embodiments, the flattening device 2 comprises a pair of flattening bases 21, a material feeding roller set 25 and a flattening roller set 24, wherein a guide rail 22 is arranged on the material feeding frame 9, and the two flattening bases 21 are slidingly arranged on the guide rail 22, and it is to be noted that the two flattening bases 21 are symmetrically arranged on the guide rail 22 and are threadedly connected with the same adjusting screw 23, thereby forming a screw pair. In this embodiment, the adjusting screw 23 is symmetrically provided with two threads with opposite rotation directions for connecting the pair of flattening bases 21, respectively, so that when the adjusting screw 23 is rotated, the pair of flattening bases 21 can be moved towards each other along the guide rail 22. On the other hand, the material feeding roller set 25 is rotatably arranged at the front end of the flattening base 21, and the end of the material feeding roller set 25 is mechanically connected with the displacement encoder sensor 8, and the material feeding roller set 25 is used to drive the material belt to feed. The front end of the flattening base 21 is provided with the flattening roller set 24, which is used to flatten the material belt for subsequent punching and cutting processing.

[0064] In a specific embodiment of the present disclosure, the punching unit 4 comprises a punch 41, a guide frame 42, an anvil table 43 and a centering member 44, the material feeding frame 9 is connected with the guide frame 42 to access the material belt. The punch 41 is arranged at the top of the guide frame 42, and the anvil table 43 is arranged at the bottom of the guide frame 42, and the anvil table 43 is aligned with the punch 41 and punches holes in the material belt by using the punch 41. Optionally, the punch 41 can punch holes in the anvil table 43 by using a stamping process or a laser cutting process. The centering members 44 are arranged on both sides of the anvil table 43, and the centering members 44 can guide the material belt to pass through the anvil table 43, thereby preventing the material belt from deviating.

[0065] In a specific embodiment of the present disclosure, as Figure 5 , the material feeding buffer device 7 comprises a base frame 71, a guide roller 72, a detection bracket 73 and a damper 74, one of the punch and weld joint detection module 3 or the forming quality detection module 5 is installed on the detection bracket 73, which is used to detect the position and shape deviation of the punch and the distance between the hole and the weld. The detection bracket 73 is located at the top of the base frame 71, and a plurality of guide rollers 72 are horizontally arranged on the base frame 71, and two adjacent guide rollers 72 are connected by a synchronous belt transmission, so that the guide rollers 72 can rotate synchronously. At the same time, the single guide roller 72 is frictionally connected with the damper 74, and the damper 74 is arranged on the base frame 71, thereby maximizing the uniform speed of the material belt, thereby reducing the error of the punch position.

[0066] In one specific embodiment of the present disclosure, the cold bending and cutting unit 6 comprises a servo cutter 61 and a progressive bender 65, wherein the top of the servo cutter 61 is fixed on a protective cover 62 which can play a certain sound insulation role, and the servo cutter 61 is connected with the lower machine circuit to coordinate the online cooperation of the processing part and the quality inspection part. The bottom of the progressive bender 65 is bolted with an anvil 64, and a plurality of pressure rollers 66 are rotatably arranged inside the progressive bender 65, such as Figure 3 and Figure 4 In this embodiment, the anvil 64 is long strip-shaped, and the material strip is gradually extruded and formed on the anvil 64 under the extrusion of the plurality of pressure rollers 66. On this basis, optionally, the anvil 64 is adapted to be connected to high-frequency current, which is rapidly heated under the action of the current, thereby softening the material strip during the extrusion and forming of the material strip, so as to reduce the cracks generated on the surface of the product.

[0067] Based on the same inventive concept, the present disclosure discloses a detection method, which applies the machine vision-based photovoltaic support online quality inspection system as above, such as Figure 10 , comprising the following steps: in step S1, the production line starts production, and in the welding station, a plurality of steel material strips are first welded end to end, and then sent to the welding seam detection module 1 through the first material feeding buffer device 7 for detection. The welding seam detection module 1 is mainly used for identifying the welding seam and determining the position of the welding seam. In this process, step S2 is performed, the material strip is flattened by the flattening device 2 and is driven to move, and in the process of moving, image acquisition and welding seam position identification are performed by the punching and welding seam joint detection module 3, and at the same time, the displacement encoding sensor 8 is used to obtain the material strip moving speed. After the welding seam is identified and the position needing to be punched is calculated, the displacement encoding sensor 8 synchronously records the relative position, and then step S3 is performed, the material strip is sent to the punching unit 4 for punching processing. After the punching is completed, the material strip is sent to the second material feeding buffer device 7, and the material strip moving at a uniform speed is kept in a state of uniform motion and enters the cold bending and cutting unit 6. Step S4 is performed, so that the material strip is gradually extruded and plastically formed by the cold bending and cutting unit 6, and the position cutting is completed, forming the photovoltaic support beam, and the photovoltaic support beam is sent to the forming quality detection module 5. Step S5 is performed to control the hardware of the whole system. Among them, the artificial intelligence model is the upper machine, and the PLC controller is the lower machine. According to the image information collected by the welding seam detection module 1, the punching and welding seam joint detection module 3 and the forming quality detection module 5, it is judged whether the product meets the standard. Step S6 is performed, after the judgment result is obtained, the products meeting the standard are stacked and packaged, the products not meeting the standard are removed and recycled, and at the same time, image processing, cold bending forming size measurement and surface defect identification are performed, so as to optimize the artificial intelligence model by means of reinforcement learning.

[0068] In one specific embodiment disclosed in the present application, the process of detecting the weld in the weld detection module 1 includes the following steps: first, the displacement encoding sensor 8 judges the running state of the steel strip and provides a trigger signal for the industrial camera 11, which identifies the welding position of the material strip in real time. Then, the information is transmitted to the punching station PLC control system, combined with the spatial position information, communication time length and material advancing distance, the relative position data of the weld in front of the punching station is calculated, and the data is transmitted to the lower computer. Finally, after the programmable logic controller receives the weld position data, the tail size is calculated through the follow-up encoder and distance data, and since the delay punching distance is equal to the tail distance of the material, the variable position punching processing is automatically completed.

[0069] In this embodiment, the variable position punching processing process includes: first, the industrial camera 11 identifies the weld information and transmits the position to the programmable logic controller system of the lower computer, and adds a 10mm punching and cutting safety margin. Then, the programmable logic controller calculates the distance between the weld position and the starting end of the preset punching position, and calculates the ratio M and the remainder N of the distance and the profile length. If M is less than 1, the lower computer controls the punching unit 4 to stop punching processing, if M is greater than 1, the punching process of M profiles is carried out normally, and then 10mm is added as a new preset profile starting end based on the weld position to start punching.

[0070] In some embodiments, the forming quality detection module 5 detection process includes: first, the visual camera detects the surface cracks and defects of the product in the left, right and bottom directions, judges whether the plating quality is qualified, and if not, the waste is removed. On this basis, after detecting the bending of the material strip, it is judged whether the cold bending position is qualified according to the inward flange length, and if not, the waste is removed. In this embodiment, the artificial intelligence model mainly combines AI YOLO algorithm and traditional algorithm, first detects the first hole position by YOLO model, and then carries out shearing. Then, the traditional algorithm is used for edge detection, so as to calibrate the hole edge position. Finally, the detection accuracy requirement can be met without affecting the production efficiency. Based on the dynamic image grabbing of product models, more than 10 test items such as total length, hole number, hole to edge distance, height, width, weld, dezincification and the like of the product are tested. Automatic identification of weld is realized, and the cold bending quality is judged according to the symmetry degree of the bent short side. If the small deviation exceeds three times or there is a major deviation, the line can be stopped for inspection in time.

[0071] Although the present application has been described in detail in the foregoing description with general principles and specific embodiments, some modifications or improvements can be made on the basis of the present application, which is obvious to those skilled in the art. Therefore, these modifications or improvements made on the basis of not deviating from the spirit of the present application, all belong to the scope of protection required by the present application.

Claims

1. A machine vision-based photovoltaic bracket online quality inspection system, which is used for processing photovoltaic brackets by using a material belt and continuously detecting processing precision online, characterized in that, include: The weld inspection module (1) is installed at the bottom inside the material feeding frame (9), and a pair of leveling devices (2) are provided on the front and rear sides of the weld inspection module (1). A displacement coding sensor (8) is installed on the leveling device (2). The punched weld joint inspection module (3) is connected at its first end to the weld inspection module (1) via a punching unit (4), and at its tail end to the first end of the cold bending and cutting unit (6). The forming quality inspection module (5) is set at the tail end of the cold bending cutting unit (6). The forming quality inspection module (5) and the punching weld joint inspection module (3) are respectively installed on two material feeding buffer devices (7). The punching unit (4) and the cold bending and cutting unit (6) are connected to the lower-level machine circuit, the weld detection module (1), the punching weld joint detection module (3), the forming quality detection module (5) and the displacement coding sensor (8) are connected to the upper-level machine signal, and the upper-level machine is connected to the lower-level machine network communication. 2.The machine vision based photovoltaic bracket online quality inspection system according to claim 1, characterized in that, The weld inspection module (1), the punch weld joint inspection module (3), and the forming quality inspection module (5) have the same structure; The weld detection module (1) includes an industrial camera (11) and a bar light source (12). The bar light source (12) is located outside the industrial camera (11), and the industrial camera (11) is connected to the host computer via signal. 3.The machine vision based photovoltaic bracket online quality inspection system according to claim 2, characterized in that, The leveling device (2) includes: A pair of leveling bases (21) are slidably mounted on guide rails (22) at their bottoms. The guide rails (22) are located inside the material feeding frame (9). The pair of leveling bases (21) and the adjusting screw (23) form a helical pair and move towards each other along the guide rails (22) under the action of the adjusting screw (23). The feeding roller assembly (25) is rotatably disposed at the front end of the leveling base (21), and the end of the feeding roller assembly (25) is mechanically connected to the displacement encoding sensor (8). The leveling base (21) is equipped with a leveling roller assembly (24). 4.The machine vision based photovoltaic bracket online quality inspection system according to claim 3, characterized in that, The punching unit (4) includes a punching machine (41), a guide frame (42), an anvil (43), and a centering component (44). The feeding frame (9) is connected to the guide frame (42). The guide frame (42) has a punching machine (41) at the top and an anvil (43) at the bottom. The anvil (43) is aligned with the punching machine (41) and punches holes in the material strip using the punching machine (41). The centering component (44) is provided on both sides of the anvil (43). The centering component (44) is adapted to guide the material strip through the anvil (43). 5.The machine vision based photovoltaic bracket online quality inspection system according to claim 4, characterized in that, The material feeding buffer device (7) comprises a chassis (71), guide rollers (72), a detection support (73) and a damper (74), one of the punch-weld joint detection module (3) or the forming quality detection module (5) is installed on the detection support (73), the detection support (73) is located on the top of the chassis (71), a plurality of guide rollers (72) are horizontally arranged on the chassis (71), any two adjacent guide rollers (72) are connected through a synchronous belt transmission, and any guide roller (72) is frictionally connected with the damper (74), and the damper (74) is arranged on the chassis (71). 6.The machine vision based photovoltaic bracket online quality inspection system according to claim 5, characterized in that, The cold bending cutting unit (6) comprises: A follow-up cutter (61) is fixed at the top of a protective cover (62), and the bottom of the protective cover (62) is connected with a machine table (63), wherein the follow-up cutter (61) is connected with the lower machine circuit; A progressive bending machine (65) is bolted with an anvil (64) at the bottom, and a plurality of pressing rollers (66) are rotatably arranged in the progressive bending machine (65), and the pressing rollers (66) are adapted to gradually extrude the material belt on the anvil (64).

7. A detection method applied to the machine vision-based photovoltaic bracket online quality inspection system according to claim 6, characterized in that, It comprises: After the production line starts production, a plurality of steel material belts are welded at the tail end in the welding station, and are sent into the weld detection module (1) through the first material feeding buffer device (7) for detection; The leveling device (2) flattens the material belt and drives the material belt to feed, in the process of feeding, image acquisition is performed through the punch-weld joint detection module (3), and the weld position is identified, at the same time, the displacement encoding sensor (8) is used to obtain the feeding speed of the material belt; The material belt is sent into the punching unit (4) for punching, after the punching is completed, the material belt is sent into the second material feeding buffer device (7), and the material belt is kept in a uniform motion state and enters the cold bending cutting unit (6); The material belt is gradually extruded and formed in the cold bending cutting unit (6), and the position cutting is completed, the photovoltaic support beam is formed, and the photovoltaic support beam is sent into the forming quality detection module (5); The artificial intelligence model is used as the upper computer, and the PLC controller is used as the lower computer, according to the image information collected by the weld detection module (1), the punch-weld joint detection module (3) and the forming quality detection module (5), whether the product meets the standard is judged; The qualified products are stacked and packaged, the unqualified products are removed and recycled, image processing, cold bending forming size measurement and surface defect identification are performed, and the artificial intelligence model is optimized through reinforcement learning.

8. The detection method according to claim 7, characterized in that, The process of detecting the weld in the weld detection module (1) comprises: The displacement encoding sensor (8) is used to judge the running state of the steel belt, and a trigger signal is provided for the industrial camera (11), the industrial camera (11) identifies the welding position of the material belt in real time; the information is transmitted to the punching station PLC control system; The upper computer combines the spatial position information, the communication time length and the material advancing distance, calculates the relative position data of the punch-weld joint before the punching station, and transmits the data to the lower computer; The programmable logic controller receives the weld position data, calculates the tail size through the follow-up encoder and distance data, and automatically completes the displacement punching process because the delay punching distance is equal to the tail size of the material.

9. The detection method according to claim 8, characterized in that, The displacement punching process comprises the following steps: The visual camera identifies the weld position information and transmits the weld position information to the programmable logic controller system of the lower computer, and adds a 10mm punching and cutting safety margin; The programmable logic controller calculates the distance between the weld position and the starting end of the preset punching position, and calculates the ratio M and the remainder N of the distance and the profile length. If M is less than 1, the lower computer controls the punching unit (4) to stop punching; if M is greater than 1, the punching process of M profiles is normally performed, and then 10mm is added as a new preset profile starting end based on the weld position to start punching.

10. The method of claim 8, wherein the step of detecting is characterized by, The forming quality detection module (5) detection process comprises the following steps: The visual camera detects the surface cracks and defects of the product in the left, right and lower directions, judges whether the plating quality is qualified, and if not, rejects the waste; After the detection of the material belt bending, it is judged whether the cold bending position is qualified according to the length of the inward bending edge, and if not, the waste is rejected; The cold bending quality is judged according to the symmetry degree of the bending short side. If the small deviation exceeds three times or a major deviation occurs, the line is stopped for inspection in time.

Images