A weld quality on-line detection weld mark device and method

Abstract

The application discloses a kind of welding quality online detection weld mark device and method, when welding actual structural member, the welding current of same time of welding torch, the visual information of weld pool that camera is collected to welding voltage signal and real-time transmission to the welding defect online detection neural network carried on control system, if the certainty percentage of the result of existence welding defect in predicted output is higher than set threshold, then record the mark signal including picture of welding defect, the number of welding defect, corresponding welding parameter of welding defect, defect type of welding defect and the certainty percentage of the result of existence welding defect;Control system starts laser and laser head, and laser head carries out laser marking at the position of welding defect.The application carries out accurate marking according to mark signal at the position of welding defect, effectively improves the applicability of welding quality online detection system in actual production and the precision and work efficiency of welding inspection link.

Description

Technical Field

[0001] This invention relates to a weld marking device and method for online welding quality inspection, belonging to the field of online welding quality inspection technology. Background Technology

[0002] Welding inspection is the last line of defense in ensuring welding quality. Because inspectors often don't know the exact location of defects before conducting non-destructive testing, they must explore and inspect over a large area, leading to inefficiency. With the development of welding automation and intelligence, utilizing signals such as arc light, molten pool, welding electrical parameters, and arc acoustic spectrum during the welding process to detect and assess various defects such as porosity, lack of fusion, undercut, and slag inclusions has become an inevitable trend in the industry, resulting in the rapid development of online welding quality inspection methods. However, when online welding quality inspection applications, primarily based on arc light and molten pool visual signals, move from laboratory welding test plates to structural components in actual production, the inefficiency caused by the large number of welds required for defect verification needs to be considered.

[0003] In addition to online welding defect detection methods, automated non-destructive testing methods are also emerging. The application method is to mount welding defect detection equipment on automated equipment to automate the entire detection process.

[0004] After welding is completed, according to the welding inspection plan, welding inspectors use inspection equipment to inspect each weld individually. Taking ultrasonic testing, the most commonly used method in actual production, as an example, inspectors need to hold a probe and gradually inspect each part of the weld in a zigzag path around the weld, which is very inefficient. Although the efficiency of inspection has improved, the amount of weld inspection has not decreased.

[0005] By collecting and processing various signals during the welding process, the system provides real-time detection results for welding defects. Since the entire process lacks the judgment of welding inspectors, the detection results for welding defects need to be verified post-weld. This verification mainly includes the following two methods:

[0006] (1) After the defect is detected, the machine is temporarily stopped and the results are verified and repaired before welding continues;

[0007] (2) After welding is completed, the location of welding defects is determined by reverse calculation based on the inspection timeline, and then verification and rework are carried out.

[0008] Manual welding inspection offers high precision but is extremely inefficient, directly impacting the overall production cycle. Automated welding inspection, performed post-weld, is more efficient than manual inspection but is more expensive and has a narrower range of applications.

[0009] The following problems exist in online welding quality inspection:

[0010] (1) In actual production, it is completely impractical for the online welding quality inspection system to stop welding as soon as it detects a possible welding defect in the weld. On the one hand, the welding defect already exists, and the process of "stopping - confirming the result - grinding - restarting the arc" will disrupt the continuity of welding production and seriously affect production efficiency. If it is just a system misjudgment, restarting the arc will increase the risk of welding defects.

[0011] (2) As can be seen from the above analysis, the detected welding defects should be handled uniformly after welding. For single test plate welding, inspectors can use the time axis provided by the detection system to find the location of the defect in the weld. However, this method is not feasible for actual structural parts. The main reason is that there are many welds on actual structural parts, the welding time is long, and welding inspectors cannot keep a close eye on the welding production process of a structural part. Therefore, it is difficult to determine the location of the defect through a time parameter.

[0012] (3) The method of uniformly treating welding defects after welding makes it difficult to verify the correctness of the inspection results. The results obtained from the inspection cannot be applied to the progressive training of the neural network. Therefore, the accuracy of online welding inspection will remain stagnant and cannot be effectively improved. Summary of the Invention

[0013] The technical problem to be solved by the present invention is to overcome the defects of the prior art and provide a welding quality online detection weld marking device and method.

[0014] To achieve the above objectives, the present invention provides an online welding quality detection and weld marking device, comprising a welding torch, a connecting block, a laser head, a camera, a Hall effect device, a laser, and a control system. The welding torch, laser head, and camera are all mounted on the connecting block. The laser head and laser are connected via optical fiber, and the camera and laser are electrically connected to the control system.

[0015] The welding torch is electrically connected to a Hall effect device, which in turn is electrically connected to the control system.

[0016] Preferred, a circular hole for the welding gun is vertically opened at one end of the connecting block, and a cross-shaped slot for the laser head is opened at the other end of the connecting block.

[0017] Preferably, the online welding quality inspection and weld marking device also includes a connecting rod, and the camera is fixedly connected to the connecting block through the connecting rod.

[0018] Preferably, the online welding quality inspection and weld marking device further includes a scanning galvanometer, which is disposed inside the laser head.

[0019] A method for online inspection and marking of weld seams to ensure welding quality, using the device described in any one of the above-mentioned embodiments as the executing entity, comprises the following steps:

[0020] When welding actual structural components, the welding current and voltage signals of the welding torch and the visual information of the molten pool collected by the camera are transmitted in real time to the online welding defect detection neural network mounted on the control system. The online welding defect detection neural network predicts and outputs the percentage of certainty of the existence of welding defects and the type of welding defects.

[0021] If the percentage of certainty in the online detection neural network for welding defects is higher than a set threshold, a marker signal is recorded. The marker signal includes the number of the welding defect, the type of the welding defect, and the percentage of certainty in the result that the welding defect exists.

[0022] The control system activates the laser and laser head, and uses the laser head to mark the location of the welding defect with laser.

[0023] Prior to this, after the actual structural component is welded, the control system obtains a list of marking signals, locates the laser marking on the actual structural component according to the number of the welding defect, and verifies and inspects the welding defect.

[0024] Prioritize the control system to start the laser and laser head. If the marking position of the laser head is aligned with the location of the welding defect, then the laser head is used to make a laser mark at the location of the welding defect.

[0025] If the marking position of the laser head is inconsistent with the location of the welding defect, that is, there is a certain distance between the marking position of the laser head and the location of the welding defect in the welding direction, the control system sets the delay of the laser head according to the distance, and uses the laser head to mark the welding defect after moving to the location of the welding defect.

[0026] Prioritize obtaining marking information including arrows indicating the location of welding defects, the number of welding defects, the type of welding defects, the percentage of certainty of the presence of welding defects, and recommended detection methods.

[0027] Preferredly, the defect types of welding defects and recommended detection methods are represented by designated symbols;

[0028] If the welding defect is a continuous defect, it is represented by a line with an arrow.

[0029] Prior to this, after the online welding defect detection neural network predicts and outputs the certainty percentage of the result of the existence of welding defects and the type of welding defects, it verifies whether the location of the corresponding welding defect actually has a welding defect and the corresponding actual welding defect type to obtain the actual inspection result.

[0030] The actual inspection results include whether welding defects actually exist, whether the actual welding defect type corresponding to the welding defect actually exists, and the defect type of the welding defect predicted by the online detection neural network.

[0031] The actual test results are fed back to the control system;

[0032] Based on actual inspection results, the control system periodically updates and trains the online detection neural network for welding defects.

[0033] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.

[0034] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0035] The beneficial effects achieved by this invention are as follows:

[0036] (1) This invention fully leverages the advantages of existing advanced laser marking technology, and accurately marks the location of welding defects based on the results of online welding quality detection, skipping the large-scale post-weld inspection process, effectively improving the applicability of the online welding quality detection system in actual production as well as the accuracy and efficiency of the welding inspection process;

[0037] (2) Since the entire inspection process is recorded, the actual results can be fed back to the online welding quality inspection system after the welding defects on the actual structural parts are inspected, which promotes the progressive training of the online welding defect inspection neural network and continuously improves the accuracy of online welding quality inspection.

[0038] (3) The key point of this invention is to use a deep learning-based neural network to detect welding defects in the welding process, and then to mark the precise location of the welding defects by a laser head set near the welding torch, thereby improving the welding detection efficiency;

[0039] (4) The welding quality online detection weld marking device and method provided by the present invention improves the accuracy and efficiency of the welding inspection process, enhances the accuracy of online welding quality detection, and strengthens the applicability of the online welding quality detection system in actual production. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the device of the present invention;

[0041] Figure 2 This is a structural diagram of the connecting block.

[0042] The meanings of the labels in the attached diagram are as follows: 1-Welding torch; 2-Connecting block; 3-Connecting rod; 4-Laser head; 5-Camera; 6-Hall element; 7-Welding power source; 8-Laser; 9-Control system; 2-1-Cross slot; 2-2-Round hole. Detailed Implementation

[0043] The following embodiments are only used to illustrate the technical solutions of the present invention more clearly, and should not be used to limit the scope of protection of the present invention.

[0044] It should be noted that if there are directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention, they are only used to explain the relative positional relationship and movement of the components in a certain specific posture. If the specific posture changes, the directional indicator will also change accordingly.

[0045] A welding quality online inspection and weld marking device includes components such as a welding torch 1, a connecting block 2, a connecting rod 3, a laser head 4, a camera 5, a Hall effect device 6, a welding power source 7, a laser 8, and a control system 9.

[0046] The welding torch 1 passes through the circular hole 2-2 in the connecting block 2 and can be fixed with bolts. A cross-shaped slot 2-1 is opened on the other side of the connecting block 2. The laser head 4 is fixed to the connecting block 2 with bolts passing through the cross-shaped slot 2-1, and its position and orientation can be adjusted arbitrarily through the cross-shaped slot 2-1. The camera 5 is fixedly connected to the connecting block 2 via a connecting rod 3. The Hall element 6 is located in the welding circuit, and the laser 8 is connected to the laser head 4 via optical fiber. The laser 8, Hall element 6, welding power supply 7, and camera 5 are all connected to the control system 9. The welding torch is electrically connected to the Hall element, and the Hall element is electrically connected to the control system.

[0047] A vertical through-hole 2-2 is opened at the right end of the connecting block 2, and a cross-shaped groove 2-1 is opened on the front side of the left end of the connecting block 2. The connecting rod 3 is a cylindrical rod. The welding gun 1, laser head 4, camera 5, Hall element 6, welding power supply 7, laser 8 and control system 9 are all of the above components. There are many models that can be used in the prior art. Those skilled in the art can select the appropriate model according to actual needs. This embodiment will not list them all.

[0048] First, the online welding defect detection neural network is trained based on the visual and electrical signals corresponding to artificially created welding defects and detected defects including porosity, lack of fusion, undercut, and slag inclusions. The electrical signals are welding current and welding voltage. When the present invention is applied to actual welding production, the control system 9 transmits the collected visual information of the molten pool and the welding current and voltage signals to the trained online welding defect detection neural network in real time. The online welding defect detection neural network predicts and outputs the percentage of certainty of the presence of welding defects and the type of welding defects.

[0049] If the percentage of certainty in the online detection neural network for welding defects is higher than a set threshold, a marker signal is recorded. The marker signal includes the number of the welding defect, the type of the welding defect, and the percentage of certainty in the result that the welding defect exists.

[0050] Furthermore, in this embodiment, the control system 9 activates the laser 8 and the laser head. If the marking position of the laser head 4 is aligned with the location of the welding defect, the laser head 4 is used to make a laser mark at the location of the welding defect.

[0051] If the marking position of the laser head 4 is inconsistent with the location of the welding defect, that is, there is a certain distance between the marking position of the laser head 4 and the location of the welding defect in the welding direction, the control system 9 sets the delay of the laser head 4 according to the distance, and performs laser marking after the laser head 4 moves to the location of the welding defect.

[0052] The laser head is equipped with a scanning galvanometer, such as a dual-dimensional magnetic levitation scanning galvanometer, which facilitates the marking of complex information.

[0053] Furthermore, this embodiment acquires marking information including the arrow indicating the location of the welding defect in the controller, the welding defect number, the welding defect type, the certainty percentage of the welding defect's presence, and the recommended detection method. The recommended detection method refers to post-weld non-destructive testing methods, such as manual methods like ultrasonic testing or magnetic particle testing.

[0054] Furthermore, in this embodiment, the defect types of welding defects and the recommended detection methods are represented by defined symbols;

[0055] If the welding defect is a continuous defect, it is represented by a line with an arrow.

[0056] Furthermore, in this embodiment, after the online detection neural network for welding defects predicts and outputs the percentage of certainty of the existence of welding defects and the type of welding defects, it verifies whether the location of the corresponding welding defect actually contains a welding defect and the corresponding actual welding defect type to obtain the actual inspection result.

[0057] The actual inspection results include whether welding defects actually exist, whether the actual welding defect type corresponding to the welding defect actually exists, and the defect type of the welding defect predicted by the online detection neural network.

[0058] The actual test results are fed back to the control system 9;

[0059] Based on actual inspection results, the control system 9 periodically updates and trains the online detection neural network for welding defects.

[0060] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the methods described above.

[0061] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0062] In this embodiment, the internal architecture of the online detection neural network for welding defects is existing technology. The existing feedforward neural network can be directly used to train the current and voltage parameter neural network, and the YOLO framework is used to train the visual image.

[0063] The training process for the online detection neural network for welding defects is achieved through the following steps:

[0064] Taking images captured by a camera as an example, first, the locations of defects are marked in the images to generate training material. Then, parameters (category, number of training epochs, and number of batches, etc.) are modified based on the existing training framework. Finally, the training script is run to obtain the weight file, thus completing the training process. When applying the training, the weight file is simply added to the detection script for normal use.

[0065] This invention provides a weld marking device and method for online inspection of welding quality, which has the following beneficial effects:

[0066] (1) This invention fully leverages the advantages of existing advanced laser marking technology, and accurately marks the location of welding defects based on the results of online welding quality detection, skipping the large-scale post-weld inspection process, effectively improving the applicability of the online welding quality detection system in actual production as well as the accuracy and efficiency of the welding inspection process;

[0067] (2) Since the entire inspection process is recorded, welding inspectors can feed back the real results to the online welding quality inspection system after the inspection, which promotes the progressive training of the online welding defect detection neural network and continuously improves the accuracy of online welding quality inspection.

[0068] The key point of this invention is to use a deep learning-based neural network to detect defects in the welding process, and then use a laser head placed near the welding torch to mark the precise location of the welding defects and related information, thereby improving welding inspection efficiency.

[0069] The detection device shown in the technical disclosure is only an example. Other connection forms or connection devices, as well as variations of the marking method in this example, are all within the protection scope of this invention.

[0070] In this embodiment, the welding torch passes through the round hole 2-2 in the connecting block and can be fixed by bolts. The other side of the connecting block 2 has a cross slot hole 2-1. The laser head 4 can be fixed to the connecting block 2 by bolts passing through the cross slot hole 2-1. The position and posture can be adjusted arbitrarily through the cross slot hole 2-1. The camera 5 is fixedly connected to the connecting block 2 by the connecting rod 3.

[0071] The cross slot 2-1 used to adjust the position and attitude of the laser head 4 can be replaced by a mechanical device such as a servo motor or a walking mechanism. The control system 9 (which can be a microcontroller) can control the laser head 4 through these mechanical devices, thereby realizing the adjustment of the position and attitude of the laser head 4, its movement, and the change of its focal length.

[0072] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method of marking a weld bead for online detection of weld quality, characterized by, The application discloses a welding quality online detection welding seam marking device which comprises a welding gun (1), a connecting block (2), a laser head (4), a camera (5), a Hall element (6), a laser (8) and a control system (9), wherein the welding gun (1), the laser head (4) and the camera (5) are all installed on the connecting block (2), the laser head (4) is connected with the laser (8) through an optical fiber, and the camera (5) and the laser (8) are electrically connected with the control system (9); the welding gun (1) is electrically connected with the Hall element (6), and the Hall element (6) is electrically connected with the control system (9); The method comprises the following steps: Based on the artificial manufacturing welding defects and the corresponding visual and electrical signals of the detected defects including pores, incomplete fusion, undercut and slag inclusion, the welding defect online detection neural network is trained, and the electrical signals are welding current and welding voltage; When welding an actual structure, the welding current of the welding gun (1), the welding voltage signal of the welding gun (1) and the visual information of the molten pool collected by the camera (5) at the same time are transmitted to the welding defect online detection neural network loaded on the control system (9) in real time, the welding defect online detection neural network predicts the certainty percentage of the result that welding defects exist and the defect type of the welding defects; If the certainty percentage of the result that welding defects exist predicted by the welding defect online detection neural network is higher than a set threshold, a marking signal is recorded, and the marking signal comprises the number of the welding defects, the defect type of the welding defects and the certainty percentage of the result that welding defects exist; The control system (9) starts the laser (8) and the laser head (4), and the laser head (4) is used for laser marking at the position of the welding defects; if the marking position of the laser head (4) is aligned with the position of the welding defects, the laser head (4) is used for laser marking at the position of the welding defects; if the marking position of the laser head (4) is not aligned with the position of the welding defects, that is, there is a certain distance between the marking position of the laser head (4) and the position of the welding defects in the welding direction, the control system (9) sets a delay of the laser head (4) according to the distance, and the laser head (4) is moved to the position of the welding defects and then laser marking is performed; Marking information including an arrow indicating the position of the welding defects, the number of the welding defects, the defect type of the welding defects, the certainty percentage of the result that welding defects exist and a recommended detection method is obtained; After the welding defect online detection neural network predicts the certainty percentage of the result that welding defects exist and the defect type of the welding defects, it is verified whether the welding defects exist at the position of the welding defects corresponding to the welding defects and the actual welding defect type corresponding to the welding defects, and an actual verification result is obtained; The actual verification result comprises whether the welding defects exist, whether the actual welding defect type corresponding to the welding defects exists and the defect type of the welding defects predicted by the welding defect online detection neural network; The actual verification result is fed back to the control system (9); According to the actual verification result, the control system (9) regularly updates the welding defect online detection neural network; After the actual structure is welded, the control system (9) obtains a list of mark signals, finds the laser mark on the actual structure according to the number of the welding defect, and checks the welding defect. The defect type of the welding defect and the recommended detection method are represented by a set symbol; if the defect type of the welding defect is a continuous defect, an arrow line is used to represent it.

2. The welding seam mark method for online welding quality detection according to claim 1, characterized in that, One end of the connecting block (2) is vertically provided with a round hole (2-2) matched with the welding gun (1), and the other end of the connecting block (2) is provided with a cross slot hole (2-1) matched with the laser head (4).

3. The welding seam mark method for online welding quality detection according to claim 1, characterized in that, The welding seam mark device for online welding quality detection further comprises a connecting rod (3), and the camera (5) is fixedly connected with the connecting block (2) through the connecting rod (3).

4. The welding seam mark method for online welding quality detection according to claim 1, characterized in that, The welding seam mark device for online welding quality detection further comprises a scanning galvanometer, which is arranged in the laser head (4).

5. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the steps of the welding seam mark method for online welding quality detection according to any one of claims 1 to 4.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the welding seam mark method for online welding quality detection according to any one of claims 1 to 4.

Images