JP7719979B2 - Automatic welding method for large rebar mesh based on AI vision - Google Patents

Description

The present invention relates to the technical field of rebar mesh welding, and in particular to an automated welding method for large rebar meshes based on AI vision.

Prefabricated rebar mesh structures are widely used in the construction industry. Currently, prefabricated rebar mesh structures are typically hand-tied or welded at the intersections of the rebar mesh. However, manual welding has limitations, including unstable welding quality, low construction efficiency, high welding costs, and an inability to work under extreme conditions such as high temperatures and pressures. Mechanized automatic welding technology is currently widely used in various fields, including automotive, aerospace, and industrial manufacturing. However, meeting the requirements of the welding process typically relies on high-precision positioning of the workpiece, combined with multiple path teaching of the robot welding torch. However, the steel bars themselves are prone to bending, and errors occur in the bent steel bars due to their elastic modulus characteristics. This inevitably leads to irregularities in the joined steel bar mesh, uneven gaps between the steel bars, and unevenness between the steel bars. These factors can prevent the arc from starting, cause arc interruptions, collisions, and ultimately lead to the failure of mechanized automatic welding.

The present invention provides an AI vision-based automatic welding method for large rebar meshes to solve the problem of irregularities and large gaps in large rebar meshes, making automatic welding difficult.

The present invention provides an automatic welding method for large-scale rebar meshes based on AI vision, comprising: a work platform; and an industrial computer; the work platform comprises a guide rail and a base that moves along the guide rail; a support arm is rotatably mounted on the base; a robot is slidably mounted on the support arm; a welding torch and a second vision device are mounted on the robot; a first vision device is mounted on the support arm; the work platform further comprises an operation table mounted under the robot; and the method also includes the following steps:
3D template construction step: construct a 3D model template of the working platform, the 3D model template including: a coordinate position of the robot, a coordinate position of the operation table, a coordinate position of the guide rail, a coordinate position of the base, a coordinate position of the support arm, and a coordinate position of the first visual device; a first reinforcing bar mesh model and a second reinforcing bar mesh model pre-constructed on the operation table, the second reinforcing bar mesh model being a 180-degree flip of the first reinforcing bar mesh model; and the first reinforcing bar mesh model and the second reinforcing bar mesh model both including coordinate positions of a plurality of joint sections; the industrial computer respectively rearranges the joint sections of the first reinforcing bar mesh model and the second reinforcing bar mesh model, the joint sections having weld seams, the length of the weld seams being shorter than the length of the joint sections, and the joint sections and the weld seams have a one-to-one correspondence;
Rebar mesh laying step: Lay multiple steel bars on the operating table to form a rebar mesh, and secure adjacent steel bars with clips on the operating table.
a first rough positioning step: the base moves along the guide rail to move the support arm from one end of the rebar mesh to the other end; while the support arm is moving, the first vision device takes continuous images of the rebar mesh to generate a first real 3D model; and compares the joint section of the first rebar mesh model with the joint section of the first real 3D model to obtain an offset amount of the joint section of the first real 3D model in the first rebar mesh model;
[0023] Precision positioning step: according to the offset amount and the 3D model template, the robot moves to the first joint section, the second vision device collects images of the first joint section to obtain a 3D model of the first joint section, the industrial computer obtains the position coordinates of the weld seam A in the first joint section through the 3D model of the first joint section, and then adjusts the starting position of the weld seam A through the preset gap value, the welding torch starts an arc on one steel bar at the starting point of the weld seam A, and then moves to the middle position D between the two steel bars while swinging, and completes the welding of the first joint section from D as the starting point along the extension direction of the steel bars while swinging;
By repeating the precision positioning steps, the robot completes the welding of the remaining joint sections.
Flip step: Loosen the clips and flip the rebar mesh 180 degrees so that the back side of the rebar mesh is facing up.
a second rough positioning step: the base moves along the guide rail to move the support arm from one end of the rebar mesh to the other end; while the support arm is moving, the first vision device takes continuous images of the rebar mesh to generate a second actual 3D model; and compares the joint section of the second rebar mesh model with the joint section of the second actual 3D model to obtain an offset amount of the joint section of the second rebar mesh model in the second actual 3D model;
The welding of the back side of the rebar mesh is completed through a precision positioning step.

Preferably, the specific steps for adjusting the start position of weld seam A through a preset gap value are as follows: if the gap at the start of weld seam A is smaller than the preset value, the start of weld seam A becomes the current position; if the gap at the start of weld seam A is larger than the preset value, weld seam A moves L mm in the narrower direction from the start position; if the gap at the start of weld seam A after the movement is smaller than the preset value, the start of weld seam A becomes the current position; if it is larger than the preset value, the above process is repeated until the gap at the start of weld seam A becomes smaller than the preset value.

Preferably, in the step of starting the arc on the single steel bar at the start of the weld seam A, the gap W mm at the start of the weld seam A is obtained through the 3D model of the first joint section, and the welding torch at the midpoint of the start of the weld seam A is offset by half W mm, and then further offset by H mm, after which the welding wire on the welding torch comes into contact with the single steel bar, where H represents the depth of the thread on the steel bar.

Preferably, in the precision positioning process, the welding torch adjusts its posture according to the coordinate positions of the two steel bars within the 3D model of the first joint section so that the welding torch is always perpendicular to the plane on which the two steel bars are located.

Preferably, in the precision positioning process, when the gap between the weld seams is small, the welding torch movement speed is fast and the wire feed speed is slow; when the gap between the weld seams is large, the welding torch movement speed is slow and the wire feed speed is fast.

Preferably, in the precision positioning step, the industrial computer calculates spatial position information of the joint section of the first rebar mesh model based on the offset amount and the 3D model template, and then the robot moves to the first joint section based on the spatial position information of the joint section of the first rebar mesh model; or, the industrial computer calculates spatial position information of the joint section of the second rebar mesh model based on the offset amount and the 3D model template, and then the robot moves to the first joint section based on the spatial position information of the joint section of the second rebar mesh model.

Preferably, there are two operation tables, each located on either side of the guide rail, and while the robot is welding rebar mesh on one operation table, it is laying or flipping the rebar mesh on the other operation table.

Preferably, during the first coarse positioning step, while the support arm is moving, the first vision device collects images of the rebar mesh and also collects images of the operation table, and the images collected by the first vision device are processed to obtain a first real 3D model on the operation table model.

Preferably, the reinforcing steel mesh comprises a plurality of second steel bars and a plurality of straight first steel bars, with one or both ends of the second steel bars being curved.

Preferably, in the flipping step, the support arm is rotated so that it does not rest on the operating table, and the rebar mesh is flipped 180 degrees through the hanging device.

Compared to the prior art, in this invention, the first visual device takes continuous photographs along the rebar mesh from start to finish to obtain the rebar mesh coordinate position. Although the accuracy of the rebar mesh is low, the speed and efficiency of obtaining the rebar mesh position are high. Then, using the obtained rebar mesh coordinate positions, the device moves to the vicinity of each joint section and takes photographs with the second visual device to obtain a 3D model of each joint section. Using the 3D model of the joint section, the start position of the weld seam is adjusted, the welding torch is offset so that the welding wire of the welding torch contacts the steel bar, and two undulation points are set to complete the welding of the joint section. The welding torch posture is adjusted so that the welding torch is always perpendicular to the plane on which the two steel bars are located during the welding process. This ensures a smooth arc start, prevents arc interruption during the welding process, ensures the stability of the welding wire extension length parameters, ensures the effectiveness of the weld pool, and ensures the smooth continuation of the welding process.

Below, we will briefly explain the accompanying drawings that need to be used to depict the embodiments of the present invention or the technical means in the prior art in order to clearly explain the embodiments or the technical means in the prior art. The accompanying drawings depicted below are only some embodiments of the present invention, and those skilled in the art can derive other accompanying drawings based on these accompanying drawings without any creative effort.

FIG. 1 is a top view of the present invention. FIG. 1 is a front view of the present invention. FIG. 2 is a left side view of the present invention. FIG. 10 is a schematic diagram illustrating a state in which the attitude of the welding torch of the present invention is adjusted. FIG. 2 is a schematic diagram illustrating the adjustment of the welding seam start point according to the present invention. 2 is a schematic diagram of the movement path of the welding torch of the present invention.

In order to clarify the objectives, technical means, and advantages of the present invention, the technical means of the present invention will be clearly and completely described below with reference to the accompanying drawings. It goes without saying that the described embodiments are only some of the embodiments of the present invention, and do not represent all of the embodiments. All other implementations that can be obtained by those skilled in the art based on the embodiments of the present invention without any creative effort fall within the scope of protection of the present invention.

Referring to Figures 1 to 3, this embodiment provides an automatic welding method for large rebar meshes based on AI vision, and includes a work platform 100 and an industrial computer. The work platform 100 includes a guide rail 1 and a base 2 that moves along the guide rail 1. A support arm 3 is rotatably mounted on the base 2. A robot 4 is slidably mounted on the support arm 3. A welding torch and a second vision device are mounted on the robot 4. A first vision device is mounted on the support arm 3. The work platform 100 further includes an operation table 5 mounted below the robot 4. A plurality of clips 6 are mounted on the operation table 5. The method also includes the following steps:

3D template construction step: A 3D model template of the work platform 100 is constructed. The 3D model template includes the coordinate positions of the robot 4, the coordinate positions of the operation table 5, the coordinate positions of the guide rail 1, the coordinate positions of the base 2, the coordinate positions of the support arm 3, and the coordinate positions of the first visual device, as well as a first rebar mesh model and a second rebar mesh model pre-constructed on the operation table 5, where the second rebar mesh model is a 180-degree flip of the first rebar mesh model, and the first rebar mesh model and the second rebar mesh model Each of the models includes the coordinate positions of multiple joint sections 200. The industrial computer rearranges the joint sections 200 of the first rebar mesh model and the second rebar mesh model, respectively. The joint sections 200 have weld seams 300, the lengths of the weld seams 300 are shorter than the lengths of the joint sections 200. The joint sections 200 and the weld seams 300 correspond one-to-one, and each joint section 200 includes a weld seam 300. Positioning a joint section 200 is equivalent to positioning a weld seam 300. Because one side (front side) of the rebar mesh 7 must be welded and then flipped over to weld the other side (back side), two rebar mesh 7 models must be constructed on the manipulation table 5. The joint section 200 is a closed section where two steel bars 71 are joined within the rebar mesh 7. The two steel bars 71 within the joint section 200 have some contacting and some not-contacting areas, but the gap is relatively small (as shown in the figure).

Reinforcing bar mesh 7 laying step: Multiple steel bars 71 are laid on the operating table 5 to form the reinforcing bar mesh 7, and adjacent steel bars 71 are secured with clips 6 on the operating table 5. The reinforcing bar mesh 7 must be joined to the steel bars 71. During the joining process, many errors occur, such as the straightness of the steel bars 71 themselves and the elastic modulus characteristics of the steel bars 71. These errors result in differences between the joined reinforcing bar mesh 7 and the reinforcing bar mesh 7 in the 3D model template. These differences can cause arc start failures, arc interruptions during the welding process, changes in wire extension length, etc., which may affect the automatic welding of the reinforcing bar mesh 7.

First rough positioning step: The base 2 moves along the starting point of the guide rail 1 to the end point of the guide rail 1, moving the support arm 3 from one end of the rebar mesh 7 to the other. While the support arm 3 is moving, the first visual device takes continuous images of the rebar mesh 7 to generate a first real 3D model. During this process, the height position of the support arm 3, i.e., the height position of the first visual device, is determined. The first visual device collects images of the rebar mesh 7 and also collects images of the operation table 5. The images collected by the first visual device are processed to obtain a first real 3D model on the operation table 5 model. The first real 3D model is a rebar mesh The first reinforcing steel mesh model includes a model of the reinforcing steel mesh 7 and a model of the manipulation table 5, and the manipulation table 5 remains constant, thereby determining the coordinate position of the manipulation table 5. When comparing the joint section 200 of the first reinforcing steel mesh model with the joint section 200 of the first actual 3D model, the manipulation table 5 of the 3D model template is superimposed on the manipulation table 5 of the first actual 3D model using the manipulation table 5 as a common reference point, and then an offset amount of the joint section 200 of the first actual 3D model in the first reinforcing steel mesh model is obtained, and then the coordinate position of the joint section 200 of the reinforcing steel mesh 7 is obtained through the offset amount. In this step, the coordinate position of the joint section 200 of the reinforcing steel mesh 7 (actual) is obtained by moving the first visual device to collect and process images, which is fast and efficient. However, because the first visual device is installed high in order to capture the entire rebar mesh 7 in the width direction, the imaging accuracy is insufficient, and as a result, there is a certain error in the coordinate position of the rebar mesh joint section 200 obtained, making it difficult for this accuracy to meet the welding needs of the robot 4.

Precise positioning step: Based on the offset amount and the 3D model template, the robot 4 moves to the first joint section 200, the second vision device collects images of the first joint section 200 to obtain a 3D model of the first joint section 200, and through the 3D model of the first joint section 200, the industrial computer obtains the position coordinates of the weld seam 300A in the first joint section 200, and then adjusts the starting position of the weld seam 300A through the preset gap value. Referring to Figure 6, the welding torch 8 starts an arc at point C on one steel bar 71 at the starting point of the weld seam 300A, then moves to the middle position D between the two steel bars 71 while swinging, and then moves from D as the starting point to point E along the extension direction of the steel bars 71 while swinging to complete the welding of the first joint section 200. In this step, the industrial computer obtains the position coordinates of the weld seam 300A in the first joint section 200. By obtaining the position coordinates of the weld seam 300A, the difference between the actual rebar mesh 7 and the rebar mesh 7 in the 3D model template is eliminated. However, if the gap at the start of the weld seam 300A is too large, it may result in an inability to start the arc and welding. Therefore, the start position of the weld seam 300A is adjusted via a preset gap value so that the gap at the start of the weld seam 300A meets the arc start requirements of the welding torch 8, preventing the inability to start the arc and welding and ensuring the smooth progress of the welding process. Next, the welding torch 8 starts an arc on one steel bar 71 at the start of the weld seam 300A, which is the first arc start point. During the first arc start, the welding wire 9 of the welding torch 8 comes into contact with the steel bar 71 (ensuring a smooth arc start), and then moves to position D, midway between the two steel bars 71, during oscillation. D is the second arc start point, and from D as the starting point, it moves along the extension direction of the steel bar 71 during oscillation to complete the welding of the first joint section 200. The welded metal strip from the first arc start point to D ensures that the welding wire 9 of the welding torch 8 comes into contact with the metal strip during the second arc start (ensuring a smooth arc start). By providing two arc start points, electric welding enables 100% arc start, ensuring the welding process proceeds smoothly.

By repeating the precision positioning step, the robot 4 completes the welding of the remaining joint sections 200. For example, the robot 4 moves to the second joint section 200 according to the offset amount and the 3D model template, the second vision device collects images of the second joint section 200 to obtain a 3D model of the second joint section 200, the industrial computer obtains the position coordinates of the weld seam 300B within the second joint section 200 through the 3D model of the second joint section 200, and then obtains the position of the weld seam 300 by adjusting the starting position of the weld seam 300B according to a preset gap value. The welding torch 8 starts an arc on one steel bar 71 at the starting point of the weld seam 300B, then moves to position D between the two steel bars 71 while swinging, and then moves from D as the starting point along the extension direction of the steel bars 71 while swinging to complete the welding of the second joint section 200.

Turning step: Loosen the clips 6 and turn the reinforcing steel mesh 7 180 degrees so that the back side of the reinforcing steel mesh 7 faces up.
a second rough positioning step: the base 2 moves along the guide rail 1 to move the support arm 3 from one end of the reinforcing steel mesh 7 to the other end; while the support arm 3 is moving, the first visual device takes continuous images of the reinforcing steel mesh 7 to generate a second real 3D model; and compares the joint section 200 of the second reinforcing steel mesh model with the joint section 200 of the second real 3D model to obtain an offset amount of the joint section 200 of the second real 3D model in the second reinforcing steel mesh model;
The welding of the back side of the reinforcing steel mesh 7 is completed through a precision positioning step.

The specific steps for adjusting the start position of the weld seam 300A using the preset gap value are as follows: If the gap at the start of the weld seam 300A is smaller than the preset value, for example, if the gap at the start of the weld seam 300A is less than 10 mm, there is no need to offset the weld seam 300A. Referring to FIG. 5, if the gap at the start O of the weld seam 300A is greater than 10 mm, the weld seam 300A is shifted 2 mm in the narrower direction from the start O to P (if the weld seam 300A extends to both ends from the start O, one end must be narrow and the other end must be wide; otherwise, the rebar mesh 7 must be re-fixed). After the shift, it is determined whether the gap at the start P of the weld seam 300A is greater than 10 mm. If it is less than 10 mm, the start of the weld seam 300A is P; if it is greater than 10 mm, the above process is repeated until the gap at the start of the weld seam 300A is less than 10 mm.

In the step where the welding torch 8 starts an arc on one steel bar at the start of the weld seam 300A, the gap W mm at the start of the weld seam 300A is obtained through the 3D model of the first joint section 200. For example, if the gap at the start of the weld seam 300A is 10 mm, the welding torch 8 is offset 5 mm from the midpoint of the start of the weld seam 300A, and after a further offset of H mm, the welding wire 9 on the welding torch 8 comes into contact with one steel bar 71. There are two steel bars 71 within one joint section 200, and the welding wire 9 on the welding torch 8 typically comes into contact with the steel bar 71 inside the rebar mesh 7. The lowest point determined after scanning with the second vision device is the midpoint of the start of the weld seam 300A. However, due to errors in the rebar mesh 7 joining process, starting the arc from the midpoint of the start of the weld seam 300A is likely to fail. To ensure a successful arc start, the conventional arc start at the intermediate position must be offset to ensure that the welding wire 9 makes contact with the steel bars 71 and has a stable arc starting point. Because the gap between the steel bars 71 is dynamic, an offset of 5 mm is first performed to eliminate any gap error in the steel bars 71. However, because the steel bars 71 have threads and the thread height H is generally set to 5 mm, an additional 5 mm offset is required to allow the welding wire 9 to contact the steel bars 71 inside the reinforcing steel mesh 7.

In the precision positioning step, the industrial computer calculates the spatial position information of the joint section 200 of the first rebar mesh model based on the offset amount and the 3D model template, and then the robot 4 moves to the first joint section 200 based on the spatial position information of the joint section 200 of the first rebar mesh model; alternatively, the industrial computer calculates the spatial position information of the joint section 200 of the second rebar mesh model based on the offset amount and the 3D model template, and then the robot 4 moves to the first joint section 200 based on the spatial position information of the joint section 200 of the second rebar mesh model.

4, in the precision positioning process, the welding torch 8 adjusts its posture according to the coordinate positions of the two steel bars 71 in the 3D model of the first joint section 200 so that it is always perpendicular to the plane on which the two steel bars 71 are located. After the rebar mesh 7 is joined, there will be height variations between the steel bars 71. If the height difference between the two bars exceeds a certain range, the welding torch 8 will hit the taller steel bar 71, causing the welding torch 8 to interrupt. The second visual device obtains a 3D model of each joint section 200 through 3D modeling. This 3D model not only provides the accurate trajectory of the weld seam 300 but also the coordinate positions of the steel bars 71. An algorithm then guides the welding gun 8 to adjust the angle offset in the z direction so that the welding torch 8 is always perpendicular to the plane on which the two steel bars 71 are located. This prevents collisions and ensures that the distances between the end of the welding torch 8 and the two steel bars 71 are consistent, ensuring the stability of the welding wire extension length parameters and the effectiveness of the weld pool.

During the precision positioning process, the movement speed of the welding torch 8 and the wire feed speed are adjusted according to the size of the gap in the weld seam 300. If the gap in the weld seam 300 is small, the movement speed of the welding torch 8 is fast and the wire feed speed is slow; if the gap in the weld seam 300 is large, the movement speed of the welding torch 8 is slow and the wire feed speed is fast. This step ensures that no arc interruption occurs and the fill amount is consistent. The wire feed speed can be adjusted by the magnitude of the current; the higher the current, the faster the wire feed speed, and the lower the current, the slower the wire feed speed.

The amplitude of the welding torch 8 is set to a fixed value rather than automatically adjusted according to the width of the weld seam 300. This setting ensures stability of the welding wire extension length parameters and also guarantees the effectiveness of the welding molten pool. The welding torch 8 movement speed and wire feed speed are adjusted to ensure that the welding arc does not break and the fill amount is guaranteed.

In another embodiment of the present invention, there are two operation tables 5, each located on either side of the guide rail 1. When the robot 4 is welding the rebar mesh 7 on one operation table 5, the support arm 3 rotates above that operation table 5, and the rebar mesh 7 is laid or flipped over on the other operation table 5. At this time, there is no interference between the support arm 39 and the robot 4, allowing these processes to proceed smoothly. Rotating the support arm 3 helps improve the welding efficiency of the robot 4 and makes it easier to lay and flip the rebar mesh 7.

The reinforcing steel mesh 7 comprises a plurality of straight first steel bars 71 and a plurality of second steel bars 71, with one or both ends of the second steel bars 71 being curved. Due to the characteristics of the elastic modulus of the steel bars 71, errors always occur after the second steel bars 71 are manufactured, and the laid reinforcing steel mesh 7 does not match the first reinforcing steel mesh model.

In the flipping step, the support arm 3 is rotated so that it does not remain on the operation table 5, and the reinforcing steel mesh 7 is flipped 180 degrees using the lifting device so that the back side of the reinforcing steel mesh 7 faces upward, and the steel bar 71 can of course be carried onto the operation table 5 using the lifting device.

In the flipping step, one side of the support platform is attached to the operation table 5, and the reinforcing steel mesh 7 is then inverted onto the support platform using a lifting device. The support platform is then moved onto the operation table 5. Specifically, multiple positioning square pipes are fixed to the operation table 5, the support platform is equipped with multiple crossbars (the crossbars are also square pipes), one end of the crossbars is inserted into the positioning square pipes, and the lifting device inverts the reinforcing steel mesh 7 onto the support platform. The crossbars move forward along the positioning square pipes until they straddle the operation table 5. At this time, the reinforcing steel mesh 7 is placed on the operation table 5 via the crossbars. The positioning square pipes are chamfered to allow the reinforcing steel mesh 7 to pass through them easily, and the height of the crossbars is greater than that of the clips 6 so that they do not affect the reinforcing steel mesh 7.

In this invention, the first visual device takes continuous photographs along the rebar mesh 7 from start to finish to obtain the coordinate position of the rebar mesh 7. While the accuracy of the rebar mesh 7 is low, the speed and efficiency of obtaining the position of the rebar mesh 7 are high. Then, using the obtained coordinate positions of the rebar mesh 7, the device moves to the vicinity of each joint section 200, takes photographs with the second visual device, and obtains a 3D model of each joint section 200. Using the 3D model of the joint section 200, the start position of the weld seam 300 is adjusted, the welding torch 8 is offset so that the welding wire 9 of the welding torch 8 contacts the steel bars 71, and the posture of the welding torch 8 is adjusted so that the welding torch 8 is always perpendicular to the plane on which the two steel bars 71 lie. This ensures a smooth arc start, prevents arc interruption during the welding process, ensures the stability of the welding wire extension length parameters, ensures the effectiveness of the welding molten pool, and ensures the smooth continuation of the welding process.

Finally, please note that the above embodiments are merely intended to clarify the technical means of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art will understand that the technical means described in each of the above embodiments can be modified or some of the technical features can be replaced with equivalents. Such modifications or substitutions do not deviate from the essence of the corresponding technical means and the spirit and scope of the technical means of each embodiment of the present invention.

100 Work platform 1 Guide rail 2 Base 3 Support arm 4 Robot 5 Operation table 6 Clip 7 Reinforced iron mesh

Claims (10)

An automatic welding method for large rebar meshes based on AI vision, comprising: a work platform; and an industrial computer; the work platform comprises a guide rail and a base that moves along the guide rail; a support arm is rotatably mounted on the base; a robot is slidably mounted on the support arm; a welding torch and a second visual device are mounted on the robot; a first visual device is mounted on the support arm; and the work platform further comprises an operation table mounted below the robot;
a 3D template construction step: constructing a 3D model template of the working platform, the 3D model template including the coordinate positions of the robot, the coordinate positions of the operation table, the coordinate positions of the guide rail, the coordinate positions of the base, the coordinate positions of the support arm, and the coordinate positions of the first visual device, a first reinforcing bar mesh model and a second reinforcing bar mesh model pre-constructed on the operation table, the second reinforcing bar mesh model being a 180-degree flip of the first reinforcing bar mesh model, and the first reinforcing bar mesh model and the second reinforcing bar mesh model both including coordinate positions of a plurality of joint sections; the industrial computer respectively rearranges the joint sections of the first reinforcing bar mesh model and the second reinforcing bar mesh model, the joint sections having weld seams, the length of the weld seams being shorter than the length of the joint sections, and the joint sections and the weld seams having a one-to-one correspondence;
A reinforcing bar mesh laying step: laying a plurality of steel bars on the operation table to form the reinforcing bar mesh, and fixing the adjacent steel bars with clips on the operation table;
a first rough positioning step: the base moves along the guide rail to move the support arm from one end of the reinforcing bar mesh to the other end, and while the support arm is moving, the first vision device takes continuous images of the reinforcing bar mesh to generate a first real 3D model, and compares the joint section of the first reinforcing bar mesh model with the joint section of the first real 3D model to obtain an offset amount of the joint section of the first reinforcing bar mesh model in the first real 3D model;
precision positioning step: based on the offset amount and the 3D model template, the robot moves to the first joint section, the second vision device collects images of the first joint section to obtain a 3D model of the first joint section, the industrial computer obtains the position coordinates of the weld seam A in the first joint section through the 3D model of the first joint section, and then adjusts the start position of the weld seam A through a preset gap value, the welding torch starts an arc on one of the steel bars at the start of the weld seam A, and then moves to a middle position D between the two steel bars while swinging, and completes the welding of the first joint section from D as the starting point along the extension direction of the steel bars while swinging;
repeating the fine positioning step, the robot completes welding of the remaining joint sections;
Flip step: loosen the clips and flip the rebar mesh 180 degrees so that the back side of the rebar mesh faces up;
a second rough positioning step: the base moves along the guide rail to move the support arm from one end of the reinforcing bar mesh to the other end, and while the support arm is moving, the first vision device takes continuous images of the reinforcing bar mesh to generate a second actual 3D model, and compares the joint section of the second reinforcing bar mesh model with the joint section of the second actual 3D model to obtain an offset amount of the joint section of the second reinforcing bar mesh model in the second actual 3D model;
and completing the welding of the back side of the reinforcing steel mesh through the precise positioning step. The specific steps of adjusting the start position of the weld seam A using the preset gap value are as follows: if the gap at the start of the weld seam A is smaller than the preset value, the start of the weld seam A becomes the current position; if the gap at the start of the weld seam A is larger than the preset value, the weld seam A moves L mm in the narrower direction from the start position; if the gap at the start of the weld seam A after the movement is smaller than the preset value, the start of the weld seam A becomes the current position; if it is larger than the preset value, the above process is repeated until the gap at the start of the weld seam A becomes smaller than the preset value. The AI vision-based automatic welding method for large rebar meshes described in claim 1 is characterized in that The AI vision-based automatic welding method for large rebar meshes described in claim 1, characterized in that in the step of starting the arc on one of the steel bars at the start of the weld seam A, the gap W mm at the start of the weld seam A is obtained through the 3D model of the first joint section, and the welding torch at the midpoint of the start of the weld seam A is offset by half W mm, and then further offset by H mm, after which the welding wire on the welding torch comes into contact with one of the steel bars, where H represents the height of the thread on the steel bar. The AI vision-based automatic welding method for large rebar meshes described in claim 1, characterized in that in the precision positioning process, the welding torch adjusts its posture according to the coordinate positions of the two steel bars within the 3D model of the first joint section so that the two steel bars are always perpendicular to a plane. The AI vision-based automatic welding method for large rebar meshes described in claim 4, characterized in that in the precision positioning process, when the gap of the weld seam is small, the welding torch movement speed is fast and the wire feed speed is slow, and when the gap of the weld seam is large, the welding torch movement speed is slow and the wire feed speed is fast. The AI vision-based automatic welding method for large rebar meshes according to claim 1, characterized in that in the precision positioning step, the industrial computer calculates spatial position information of the joint section of the first rebar mesh model based on the offset amount and the 3D model template, and then the robot moves to the first joint section based on the spatial position information of the joint section of the first rebar mesh model, or the industrial computer calculates spatial position information of the joint section of the second rebar mesh model based on the offset amount and the 3D model template, and then the robot moves to the first joint section based on the spatial position information of the joint section of the second rebar mesh model. The AI vision-based automatic welding method for large rebar meshes described in claim 1 is characterized in that there are two operation tables, each located on either side of the guide rail, and while the robot is welding the rebar mesh on one operation table, it lays or flips the rebar mesh on the other operation table. The AI vision-based automatic welding method for large rebar meshes described in claim 1, characterized in that during the first rough positioning step, while the support arm is moving, the first vision device collects images of the rebar mesh and the operation table, and processes the images collected by the first vision device to obtain the first real-life 3D model on the model of the operation table. The AI vision-based automatic welding method for large rebar meshes described in claim 1 is characterized in that the rebar mesh comprises a plurality of second steel bars and a plurality of straight first steel bars, with one or both ends of the second steel bars being curved. The automatic welding method for large reinforcing steel meshes based on AI vision according to claim 7, characterized in that in the flipping step, the support arm is rotated so that the support arm does not remain on the operation table, and the reinforcing steel mesh is flipped 180 degrees through a hanging device.