Twin wire submerged arc welding tracking system based on twin wire arc coordination sensing and machine learning

Abstract

Abstract: The present application relates to a kind of double-wire submerged-arc welding tracking systems based on double-wire arc collaborative sensing and machine learning.For single-wire welding low efficiency and in the process of submerged arc welding, it is difficult to accurately control the relative position of arc and weld by intuitive means due to the arc covered by flux. Utilize the swing double-wire submerged-arc welding arc signal filtering method to filter out the interference signal in the arc signal; Utilize the arc signal recognition method based on machine learning to identify the arc signal in the current swing period; Utilize the weld deviation calculation method based on double-wire arc signal collaborative processing to accurately calculate the weld deviation.

Description

Dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning Technical Field

[0001] This invention relates to the field of dual-wire submerged arc welding tracking, and is a dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning. Background Technology

[0002] Twin-wire submerged arc welding (TWA) is a widely used welding method for medium / thick plates in major engineering projects such as boiler manufacturing, large pressure vessel construction, nuclear power engineering, and offshore ultra-large wind power facilities. Currently, TWA heavily relies on skilled workers' experience to control the welding trajectory, making it difficult to accurately control the relative position of the arc and the weld seam through intuitive means. To address these challenges, this invention discloses a TWA tracking system based on dual-wire arc collaborative sensing and machine learning. Utilizing a dual-welding-torch alternating oscillation mechanism and a weld seam deviation calculation method based on alternating oscillation arc sensing, the system achieves automated tracking of the TWA welding process. Summary of the Invention

[0003] A dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is used for automated tracking of the weld seam during dual-wire submerged arc welding. Its system schematic diagram is shown in Figure 1. The system is characterized by comprising a dual-wire submerged arc welding robot system, an industrial computer, a robot controller, a data acquisition card, a Hall voltage sensor, a submerged arc welding power supply, and a wire feeder. The dual-wire submerged arc welding robot system controls the welding torch oscillation to adapt to the oscillating welding process, utilizing the drive of the robot joints to position the welding torch along the X, Y, and Z axes. The robot can be freely adjusted in the dimensional direction to achieve precise positioning, path planning, and process control; the industrial computer is used to acquire arc signals and realize arc tracking using a submerged arc welding seam tracking method based on dual-wire arc collaborative sensing; the robot controller is used to drive the dual-wire submerged arc welding robot and control the industrial computer; the data acquisition card uses a Hall voltage sensor to acquire DC and AC arc signals; the Hall voltage sensor is used to measure the welding arc signal in real time during the dual-wire submerged arc welding process; the submerged arc welding power supply is used to provide the energy required for welding; and the wire feeder is used to transfer the welding wire.

[0004] A dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is characterized by: utilizing a dual-welding-torch oscillation mechanism to achieve alternating oscillation of the two welding torches, thereby realizing weld seam tracking; the dual-welding-torch oscillation mechanism includes dual welding torches, a main controller, a servo motor, a coupling, a telescopic drive shaft, an oscillating connecting rod, and a clamping device, as shown in Figure 2; the main controller outputs pulse signals connected to the servo motor to drive the telescopic drive shaft; one end of the telescopic drive shaft is connected to the servo motor via a coupling, and the other end is connected to the oscillating connecting rod via bolts; one end of the oscillating connecting rod is connected to the clamping device to fix the welding torch, and the movement of the telescopic drive shaft... The system achieves periodic mechanical oscillation of the swing linkage, thereby enabling periodic oscillation of the welding torch. The main controller outputs pulse signals by sending a reference pulse signal to the first servo motor within one oscillation cycle, and simultaneously sending an inverted pulse signal to the second servo motor via an inverter. An encoder monitors the welding torch position in real time and synchronously feeds it back to the main controller, ensuring a 180° phase difference between the two welding torches. The mechanical structure enables the two welding torches to oscillate alternately. The dual welding torch oscillation mechanism also includes a servo motor-driven lateral and longitudinal adjustment screw device for adjusting the welding torch. Before welding, this screw device adjusts the position and height of the welding torch assembly according to the welding conditions and workpiece condition to quickly approach the ideal reference.

[0005] The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is characterized by: using a support vector machine-based arc signal interference pattern recognition method to identify the acquired arc signal and improve weld tracking accuracy; the support vector machine-based arc signal interference pattern recognition method is achieved by real-time acquisition of arc signal features during welding and calculation of feature vectors, which are then input into a trained model to identify the degree of interference to the arc signal; the support vector machine method extracts feature vectors by filtering, denoising, and normalizing the voltage signals acquired during welding, and then uses the support vector machine algorithm to establish a model of the degree of interference to the arc signal; the support vector machine algorithm solves the hyperplane through convex optimization to maximize the classification margin; the model of the degree of interference to the arc signal includes three modes: no interference mode, slight interference mode, and severe interference mode.

[0006] The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is characterized by: calculating weld deviation using a weld deviation calculation method based on alternating oscillating arc sensing to achieve weld tracking; the weld deviation calculation method based on alternating oscillating arc sensing includes weld centering bevel deviation calculation, weld left bevel deviation calculation, and weld right bevel deviation calculation, and the deviation discrimination of these three weld deviation calculation methods is determined by the arc signal of the welding torch group;

[0007] The calculation of weld alignment and bevel deviation involves a model combination of interference levels in the arc signal of the welding torch assembly, including: no interference mode + no interference mode, slightly interfered mode + slightly interfered mode, and no interference mode + slightly interfered mode. This is achieved by calculating the difference between the integral values ​​of the arc signal in the first and second halves of the oscillation cycle. And it is achieved through comparative analysis, when At that time, the welding torch assembly deviated to the right. If the welding torch group deviates to the left, it is considered that there is no deviation; The integral value of the left deviation arc signal during the first half-cycle's leftward swing of the welding torch; This is the integral value of the right deviation arc signal due to the rightward swing of the first welding torch during the second half of the cycle; the aforementioned This is the integral value of the right deviation arc signal due to the rightward swing of the second welding torch during the first half of the cycle; the aforementioned This is the integral value of the left deviation arc signal due to the leftward swing of the second welding torch during the second half of the cycle;

[0008] The calculation of the left and right bevel deviations of the weld seam involves considering that if the arc signals of both welding torches are in an interference-free or slightly interfered mode, both arc signals are considered valid. However, due to the alternating oscillation of the dual welding torches, the alternating arc oscillations cause signal confusion, making it difficult to determine which arc signal belongs to the left or right bevel. Therefore, the calculation of the left bevel deviation involves a model combination of interference levels for the arc signals of the welding torch group, including: interference-free mode + heavily interfered mode, and slightly interfered mode + heavily interfered mode. After removing the arc signals from the heavily interfered mode, the calculation of the left bevel deviation is determined only by the arc signals from the interference-free or slightly interfered modes. This is achieved by calculating the integral value of the arc signal as the welding torch oscillates to the right during one oscillation cycle in this mode. Integral value of the arc signal swinging to the left The difference is compared and analyzed with a preset left and right deviation threshold. At that time, it was believed that the welding torch was deflected to the left. If the welding torch is deflected to the right, it is considered to be deflected to the right; otherwise, it is considered not to have any deviation. and The threshold values ​​for left and right deviation of the welding torch when welding the left bevel of the weld seam;

[0009] The calculation of the right-side bevel deviation of the weld seam is as follows: the interference level model combination of the arc signal of the welding torch assembly includes: no interference mode + heavy interference mode, and light interference mode + heavy interference mode. After removing the arc signal of the welding torch in the heavy interference mode, the calculation of the right-side bevel deviation of the weld seam is determined only by the arc signal of the welding torch in the no interference mode or the light interference mode. This is achieved by calculating the integral value of the arc signal of the welding torch swinging to the left in one oscillation cycle under this mode. Integral value of the rightward swinging arc signal The difference is compared and analyzed with the preset left and right deviation thresholds. At that time, it was believed that the welding torch was deflected to the left. If the welding torch is deflected to the right, it is considered to be deflected to the right; otherwise, it is considered not to have any deviation. and The threshold values ​​for left and right deviation of the welding torch when welding the right bevel of the weld seam;

[0010] In the mild interference mode, part of the arc signal of the welding torch is interfered with during the oscillation cycle, and the undisturbed arc signal needs to be extracted for deviation calculation. In the severe interference mode, the arc signal of the welding torch is severely interfered with during the oscillation cycle and cannot reflect the bevel information, so it needs to be discarded. The oscillation cycle refers to the time taken for the first welding torch to swing to the left from the oscillation center to a certain amplitude, then swing to the right, and finally return to the oscillation center. At the same time, the second welding torch swings to the right from the oscillation center to a certain amplitude, then swings to the left, and finally returns to the oscillation center. The amplitude of the welding torch swinging to the left and to the right with the oscillation center as the reference is equal. The flowchart of the weld deviation calculation method based on alternating oscillation arc sensing is shown in Figure 3.

[0011] The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is characterized by: utilizing a laser ranging-based adaptive control method for dual-wire weld curvature to achieve adaptive weld curvature, the flowchart of which is shown in Figure 4; the laser ranging-based adaptive control method for dual-wire weld curvature is based on the real-time distance between the laser ranging sensor and the workpiece during the welding process. With the preset ideal height This was achieved through comparative analysis, utilizing a servo motor to drive the welding torch clamping device to move along the groove rail in the welding torch axial direction in real time, thus mitigating the deviation. Adaptive weld curvature; the laser rangefinder is rigidly connected to the end where the welding torch tip connects to the straight tube, and requires a high-temperature resistant, spatter-proof protective cover; the real-time distance of the laser rangefinder relative to the workpiece. That is, the height of the welding torch tip from the workpiece during welding plus the height of the welding torch tip itself; the preset ideal height. It is obtained by real-time acquisition of workpiece curvature changes by a laser rangefinder during the welding process, combined with the welding process, and must meet the following requirements. ,in This refers to the distance between the tips of the welding torches in the welding torch assembly. The ideal height difference for the welding torch assembly. The curvature of the welded workpiece.

[0012] A dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning is characterized by: utilizing a laser ranging-based closed-loop control method for the dual-wire angle to achieve closed-loop control; the laser ranging-based closed-loop control method monitors the real-time angle of the welding torch assembly. The angle with the preset ideal This was achieved through comparative analysis, using a PID controller to adjust the included angle. Calculations are performed, and the drive device simultaneously rotates the welding torch. Achieve closed-loop control; the real-time included angle of the welding torch assembly ,in The distance from the tip of the first welding torch to the workpiece. This is the distance from the tip of the second welding torch to the workpiece. The distance between the welding torch tips in the welding torch assembly; the distance from the welding torch tip to the workpiece. That is, the real-time distance between the laser rangefinder and the workpiece. Subtract the height of the welding torch tip itself; the distance between the welding torch tips of the welding torch assembly. The distance is calculated by installing a laser rangefinder sensor on one welding torch tip to emit a laser beam to another welding torch tip, using the time-of-flight method and the radius of the welding torch's conductive tip. The laser rangefinder sensor, located at the end where the welding torch tip connects to the straight tube and is tangential to the tip, measures the distance between the welding torch tips. It requires a high-temperature resistant, splash-proof protective cover. The preset ideal angle... This refers to the ideal double-wire angle obtained by combining welding process and welding path; the driving device includes a servo motor, a telescopic drive shaft, and a swing link; one end of the telescopic drive shaft is connected to the swing link by bolts; the swing link is rigidly connected to the welding torch clamping device, and the PID controller outputs a signal to the servo motor to drive the telescopic drive shaft to rotate, which in turn drives the swing link to rotate, thereby realizing the rotation of the welding torch. Real-time angle of welding torch assembly The schematic diagram is shown in Figure 5.

[0013] Inventions have beneficial results

[0014] This invention relates to a tracking system for dual-wire submerged arc welding based on dual-wire arc collaborative sensing and machine learning. It addresses the challenges of low efficiency in single-wire welding and the difficulty in accurately controlling the relative position of the arc and weld seam through visual means during submerged arc welding, as the arc is covered by flux. The system utilizes an oscillating dual-wire submerged arc welding arc signal filtering method to remove interference signals from the arc signal; it employs a machine learning-based arc signal recognition method to identify the arc signal within the current oscillation cycle; and it uses a weld seam deviation calculation method based on dual-wire arc signal collaborative processing to accurately calculate the weld seam deviation. Attached Figure Description

[0015] Figure 1 is a schematic diagram of a dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning.

[0016] Figure 2 is a schematic diagram of the double welding gun swing mechanism.

[0017] In the figure, 1 is the welding torch, 2 is the clamping device, 3 is the swing linkage, 4 is the telescopic drive shaft, 5 is the servo motor, 6 is the PID controller, 7 is the main controller, 8 is the coupling, 9 is the longitudinal drive screw device, 10 is the transverse drive screw device, 11 is the servo motor, and 12 is the laser rangefinder sensor.

[0018] Figure 3 is a flowchart of the weld deviation calculation method based on alternating oscillating arc sensing.

[0019] Figure 4 shows the flowcharts of the adaptive control method for the curvature of the double-wire weld and the closed-loop control method for the included angle of the double-wire weld based on laser ranging.

[0020] Figure 5 shows the real-time included angle of the welding torch assembly. Schematic diagram. Detailed Implementation

[0021] To better illustrate the technical solution and beneficial effects of the invention, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. The embodiments of the invention are not limited thereto.

[0022] Step 1: Alternating oscillation control of dual welding torches

[0023] In the process of dual-wire submerged arc welding, it is necessary to control the periodic alternating oscillation of the two welding torches to achieve precise weld seam tracking and improve welding efficiency. To address this challenge, this invention discloses a dual-welding torch alternating oscillation mechanism, as shown in Figure 2. By activating the main controller and setting two servo motors to output drive signals with a phase difference of 180°, the control telescopic drive shaft drives the oscillating linkage to achieve the periodic alternating mechanical oscillation of the two welding torches. Furthermore, an encoder provides real-time feedback on the welding torch position to ensure the synchronization and stability of the welding torch position.

[0024] Step 2: Arc signal interference pattern identification

[0025] In the process of dual-wire submerged arc welding, it is necessary to extract the undisturbed arc signal for deviation calculation and weld tracking. This is because the welding arc is easily affected by factors such as contaminants, iron oxide chips, and weld assembly errors, making it impossible to accurately reflect the deviation information of the weld bevel. To address this problem, this invention discloses an arc signal interference pattern recognition method based on support vector machine. During the welding process, the voltage signal is collected in real time, filtered, denoised, and normalized to extract feature vectors. A support vector machine algorithm is used to solve the hyperplane through convex optimization to maximize the classification margin and establish a model of the interference level of the arc signal. The interference level model includes three types: undisturbed mode, slightly interfered mode, and heavily interfered mode.

[0026] Step 3: Calculation of weld deviation

[0027] In dual-wire submerged arc welding, it is necessary to calculate deviations by combining different combinations of interference patterns affecting the arc signal of the welding torch assembly to improve weld tracking accuracy. To address this challenge, this invention discloses a weld deviation calculation method based on alternating oscillating arc sensing. This method includes calculating weld centering bevel deviation, left-side weld bevel deviation, and right-side weld bevel deviation.

[0028] In the calculation of weld bevel deviation, the welding robot will calculate... The size of the value, where The integral value of the left deviation arc signal during the first half of the cycle, representing the leftward swing of the welding torch. The integral value of the right deviation arc signal due to the rightward swing of the first welding torch during the second half of the cycle. The integral value of the right deviation arc signal due to the rightward swing of the second welding torch during the first half of the cycle. The integral value of the left deviation arc signal due to the leftward swing of the second welding torch during the second half of the cycle, when When the welding torch assembly deviates to the right, control the welding torch assembly to move to the left; when If the welding torch group deviates to the left, control the welding torch group to move to the right; otherwise, it is assumed that there is no deviation, and the welding torch swing position remains unchanged.

[0029] In the calculation of the bevel deviation on the left side of the weld, the welding robot calculates the integral value of the arc signal swinging to the right within the oscillation cycle after the arc signal in the heavily interfered mode has been removed. Integral value of the arc signal swinging to the left difference ,when When the welding torch is considered to be deviating to the left, the welding torch is moved to the right; when If the welding torch is considered to be deviating to the right, the torch is moved to the left; otherwise, if no deviation is considered, the torch's oscillation position remains unchanged. and This is the preset threshold for the left and right deviation of the welding torch when welding the left bevel.

[0030] In the calculation of the bevel deviation on the right side of the weld, the welding robot calculates the integral value of the arc signal swinging to the left within the oscillation cycle after the arc signal in the heavily interfered mode has been removed. Integral value of the arc signal swinging to the right difference ,when When the welding torch is considered to be deviating to the left, the welding torch is moved to the right; when If the welding torch is considered to be deviating to the right, the torch is moved to the left; otherwise, if no deviation is considered, the torch's oscillation position remains unchanged. and This is the preset threshold for the left and right deviation of the welding torch when welding the right bevel.

[0031] Step 4: Adaptive Control of Weld Curvature

[0032] In dual-wire submerged arc welding, the height of the welding torch needs to be adjusted in real time based on deviations to adapt to changes in the curvature of the workpiece. To address this challenge, this invention discloses a laser ranging-based adaptive control method for the curvature of dual-wire welds. This method achieves adaptive weld curvature control by having the welding robot calculate the real-time distance between the laser ranging sensor, rigidly connected to the end of the welding torch tip and the straight pipe, and the workpiece. With the preset ideal height The difference is determined by using a servo motor to drive the welding torch clamping device to move along the groove rail in real time along the welding torch axis. This allows the welding torch tip to reach the ideal welding height, thereby achieving adaptive weld curvature.

[0033] Step 5: Closed-loop control of the dual-wire angle

[0034] In the process of dual-wire submerged arc welding, the angle between the welding torches in the dual-torch oscillation mechanism needs to be dynamically corrected. An angle that is too large or too small will affect weld formation and easily lead to welding defects such as burn-through. To address this problem, this invention discloses a closed-loop control method for the dual-wire angle based on laser ranging. This method achieves closed-loop control of the dual-wire angle by using a laser ranging sensor to measure the distance between the welding torch tips in real time. and the distance from the welding torch tip to the workpiece The actual included angle of the welding torch assembly is obtained by utilizing the geometric relationship between its data, and the PID controller will calculate the actual included angle. Angle with ideal Angle adjustment amount This forms a closed-loop feedback loop, using a servo motor to drive the telescopic drive shaft and the connected swing linkage to rotate, thereby causing the welding torch to rotate. Achieving the ideal angle enables closed-loop control of the double-wire angle, thereby improving welding quality.

Claims

1. A dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning, used for automated tracking of weld seams during dual-wire submerged arc welding, characterized in that: The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning consists of a dual-wire submerged arc welding robot system, an industrial computer, a robot controller, a data acquisition card, a Hall voltage sensor, a submerged arc welding power supply, and a wire feeder. The dual-wire submerged arc welding robot system controls the welding torch oscillation to adapt to the oscillating welding process. The robot joint drive allows the welding torch position to be freely adjusted in three dimensions (X, Y, and Z axes) to achieve precise positioning, path planning, and process control. The dual-welding-torch oscillation mechanism enables alternating oscillation of the two welding torches, thereby achieving weld seam tracking. An encoder is used... The welding torch position is monitored in real time and synchronously fed back to the main controller to ensure a 180° phase difference between the two welding torches; the industrial computer is used to acquire the arc signal and achieve arc tracking using a submerged arc welding seam tracking method based on dual-wire arc collaborative sensing; the robot controller is used to drive the dual-wire submerged arc welding robot and control the industrial computer; the data acquisition card uses a Hall voltage sensor to acquire DC and AC arc signals; the Hall voltage sensor is used to measure the welding arc signal in real time during the dual-wire submerged arc welding process; the submerged arc welding power supply is used to provide the energy required for welding; the wire feeder is used for... The welding wire is transferred; the arc signal interference pattern recognition method based on support vector machine is used to identify the acquired arc signal and improve the weld tracking accuracy; the arc signal interference pattern recognition method based on support vector machine is achieved by collecting arc signal features in real time during the welding process and calculating feature vectors, which are then input into a trained model to identify the degree of interference to the arc signal; the machine learning method based on support vector machine is used to filter, reduce noise and normalize the voltage signal collected during the welding process to extract feature vectors, and then use the support vector machine algorithm to establish the interference pattern of the arc signal. Interference level model; the support vector machine algorithm solves the hyperplane through convex optimization to maximize the classification margin; the interference level model of the arc signal includes three types: no interference mode, slight interference mode, and heavy interference mode; the weld deviation is calculated using the weld deviation calculation method based on alternating oscillating arc sensing to achieve weld tracking; the weld deviation calculation method based on alternating oscillating arc sensing includes weld centering bevel deviation calculation, weld left bevel deviation calculation, and weld right bevel deviation calculation, and the deviation discrimination of these three weld deviation calculation methods is determined by the arc signal of the welding torch group; The calculation of weld bevel deviation is as follows: the interference level model combination of the arc signal of the welding torch group includes: no interference mode + no interference mode, slightly interfered mode + slightly interfered mode, and no interference mode + slightly interfered mode. This is achieved by comparing the difference in the integral values ​​of the arc signals of the first and second welding torches during one oscillation cycle. When the difference in integral values ​​is greater than zero, the welding torch group deviates to the right; when the difference in integral values ​​is less than zero, the welding torch group deviates to the left; otherwise, no deviation is considered. The calculation of the left and right bevel deviations of the weld is also performed if the arc signals of both welding torches are in the no interference mode or slightly interfered mode. In both interference modes and arc signals, the signal is considered valid. However, due to the alternating oscillation of the dual welding torches, the alternating arc oscillation leads to signal confusion, making it difficult to determine which arc signal belongs to the left or right bevel. Therefore, the calculation of the left bevel deviation of the weld seam is based on the interference level model combination of the arc signal of the welding torch group, which includes: no interference mode + heavily interfered mode and lightly interfered mode + heavily interfered mode. After removing the arc signal of the welding torch in the heavily interfered mode, the calculation of the left bevel deviation of the weld seam is determined only by the arc signal of the welding torch in the no interference mode or the lightly interfered mode. This is achieved by calculating the integral value of the arc signal of the welding torch oscillating to the right in one oscillation cycle under this mode. Integral value of the arc signal swinging to the left The difference is compared and analyzed with a preset left and right deviation threshold. At that time, it was believed that the welding torch was deflected to the left. If the welding torch is deflected to the right, it is considered to be deflected to the right; otherwise, it is considered not to have any deviation. and The threshold values ​​for left and right deviation of the welding torch are defined when welding the left bevel of the weld seam. The calculation of the right bevel deviation is as follows: the interference level model combination of the arc signal of the welding torch group includes: no interference mode + heavily interfered mode, and lightly interfered mode + heavily interfered mode. After removing the arc signal of the welding torch in the heavily interfered mode, the right bevel deviation calculation is determined only by the arc signal of the welding torch in the no interference mode or the lightly interfered mode. This is achieved by calculating the integral value of the arc signal of the welding torch swinging to the left in one oscillation cycle under this mode. Integral value of the rightward swinging arc signal The difference is compared and analyzed with the preset left and right deviation thresholds. At that time, it was believed that the welding torch was deflected to the left. If the welding torch is deflected to the right, it is considered to be deflected to the right; otherwise, it is considered not to have any deviation. and The threshold values ​​for left and right deviation of the welding torch are defined when welding the right bevel of the weld seam. In the mild interference mode, part of the arc signal of the welding torch is interfered with during the oscillation cycle, and the undisturbed arc signal needs to be extracted for deviation calculation. In the severe interference mode, the arc signal of the welding torch is severely interfered with during the oscillation cycle and cannot reflect the bevel information, so it needs to be discarded. The oscillation cycle refers to the time taken for the first welding torch to swing to the left from the oscillation center to a certain amplitude, then swing to the right, and finally return to the oscillation center. At the same time, the second welding torch swings to the right from the oscillation center to a certain amplitude, then swings to the left, and finally returns to the oscillation center. The amplitude of the welding torch swinging to the left and to the right with the oscillation center as the reference is equal.

2. The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning according to claim 1, characterized in that: The dual welding torch oscillation mechanism includes dual welding torches, a main controller, a servo motor, a coupling, a telescopic drive shaft, an oscillating link, and a clamping device. The main controller outputs pulse signals connected to the servo motor to drive the telescopic drive shaft. One end of the telescopic drive shaft is connected to the servo motor via a coupling, and the other end is connected to the oscillating link via bolts. One end of the oscillating link is connected to the clamping device to fix the welding torch. The periodic mechanical oscillation of the oscillating link is achieved through the movement of the telescopic drive shaft, thereby realizing the periodic oscillation of the welding torch. The pulse signal output by the main controller refers to sending a reference pulse signal to the first servo motor within one oscillation cycle, and simultaneously sending an inverted pulse signal to the second servo motor via an inverter. The mechanical structure enables the dual welding torches to oscillate alternately. The dual welding torch oscillation mechanism also includes a servo motor-driven lateral and longitudinal adjustment screw device for adjusting the welding torches. This screw device adjusts the position and height of the welding torch assembly according to the welding conditions and workpiece condition before welding to quickly approach the ideal reference.

3. The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning according to claim 1, characterized in that: An adaptive weld curvature control method based on laser ranging is used to achieve adaptive weld curvature. This method utilizes the real-time distance between the laser ranging sensor and the workpiece during the welding process. With the preset ideal height This was achieved through comparative analysis, utilizing a servo motor to drive the welding torch clamping device to move along the groove rail in real-time along the welding torch axis to correct the deviation. Adaptive weld curvature; the laser rangefinder is rigidly connected to the end where the welding torch tip connects to the straight tube, and requires a high-temperature resistant, spatter-proof protective cover; the real-time distance of the laser rangefinder relative to the workpiece. That is, the height of the welding torch tip from the workpiece during welding plus the height of the welding torch tip itself; the preset ideal height. It is obtained by real-time acquisition of workpiece curvature changes by a laser rangefinder during the welding process, combined with the welding process, and must meet the following requirements. ,in This refers to the distance between the tips of the welding torches in the welding torch assembly. The ideal height difference for the welding torch assembly. The curvature of the welded workpiece.

4. The dual-wire submerged arc welding tracking system based on dual-wire arc collaborative sensing and machine learning according to claim 1, characterized in that: A closed-loop control method for the dual-wire angle is implemented using laser ranging; this method monitors the real-time angle of the welding torch assembly. The angle with the preset ideal This was achieved through comparative analysis, using a PID controller to adjust the included angle. Calculations are performed, and the drive device simultaneously rotates the welding torch. Achieve closed-loop control; the real-time included angle of the welding torch assembly ,in The distance from the tip of the first welding torch to the workpiece. This is the distance from the tip of the second welding torch to the workpiece. The distance between the welding torch tips in the welding torch assembly; the distance from the welding torch tip to the workpiece. That is, the real-time distance between the laser rangefinder and the workpiece. Subtract the height of the welding torch tip itself; the distance between the welding torch tips of the welding torch assembly. The distance is calculated by installing a laser rangefinder sensor on one welding torch tip to emit a laser beam to another welding torch tip, using the time-of-flight method and the radius of the welding torch's conductive tip. The laser rangefinder sensor, located at the end where the welding torch tip connects to the straight tube and is tangential to the tip, measures the distance between the welding torch tips. It requires a high-temperature resistant, splash-proof protective cover. The preset ideal angle... This refers to the ideal double-wire angle obtained by combining welding process and welding path; the driving device includes a servo motor, a telescopic drive shaft, and a swing link; one end of the telescopic drive shaft is connected to the swing link by bolts; the swing link is rigidly connected to the welding torch clamping device, and the PID controller outputs a signal to the servo motor to drive the telescopic drive shaft to rotate, which in turn drives the swing link to rotate, thereby realizing the rotation of the welding torch. 。

Images