A full-position MAG welding robot swing width dynamic adjustment method

Abstract

This invention relates to the field of robot control technology and discloses a method for dynamically adjusting the swing width of an all-position MAG welding robot. This method acquires welding signals and bevel images using an arc sensor and a vision sensor, extracts the geometric features of the bevel, and calculates the bevel filling coefficient. Identify the current welding position type; establish a multi-factor coupling model for swing width. The invention calculates the target swing width using a gain scheduling algorithm and achieves real-time correction using closed-loop control. It can precisely control the swing width based on the bevel depth-to-width ratio and welding position differences, solving the problems of insufficient filling in deep and narrow bevels and unstable welding quality at different positions, significantly improving the forming quality and consistency of all-position welding.

Description

A method for dynamically adjusting the oscillation width of an all-position MAG welding robot Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method for dynamically adjusting the swing width of an all-position MAG welding robot. Background Technology

[0002] MAG (Metal Electrode Gas) welding is widely used in shipbuilding, pressure vessels, construction machinery, automobiles, and energy equipment due to its high production efficiency, wide applicability, and stable weld quality. With the increasing level of industrial automation, welding robots are being used more and more frequently in the welding of complex structures and in multi-position welding. However, during all-position welding, the weld formation characteristics change significantly due to factors such as gravity direction, molten pool shape, groove structure, and heat input fluctuations, making traditional fixed-parameter control methods difficult to adapt to complex working conditions.

[0003] Existing welding robots mostly use preset fixed oscillation widths and frequencies for welding control. Although some systems can make simple corrections based on the average values ​​of welding current and voltage, they fail to comprehensively consider the coupling relationship between the weld spatial position, groove geometry, and dynamic feedback signals during the welding process. Therefore, under conditions of significant posture changes or uneven grooves, defects such as uneven weld width, fluctuating reinforcement height, insufficient fusion, or weld misalignment are prone to occur, affecting weld formation quality and welding stability.

[0004] In existing technologies, for example, Chinese patent CN117066647A discloses an all-position swing width adaptive MAG welding method, which uses laser sensors and arc sensors to collect weld information in real time, calculates the optimal welding parameters based on spatial position information, and makes adaptive adjustments. However, this solution has the following shortcomings:

[0005] (1) No quantitative mapping relationship has been established between the geometric features of the bevel and the swing width. The existing scheme only makes simple adjustments based on the bevel width without considering the ratio between the bevel depth and width. This leads to the use of similar swing strategies for deep and narrow bevels and wide and shallow bevels, which can easily result in insufficient filling of deep and narrow bevels or failure to fuse the edges of wide and shallow bevels.

[0006] (2) The position correction uses linear interpolation, which does not fully consider the nonlinear effect of different welding positions on the fluidity of the molten pool. For example, in the overhead welding position, the molten pool is prone to sag due to gravity, and the swing width needs to be significantly reduced, but linear adjustment cannot meet the requirements of this nonlinear change;

[0007] (3) Lack of a multi-factor coordinated control mechanism. The existing scheme cannot adjust the swing width synchronously when the welding speed changes, and has not established a coupling model of multiple factors such as position, bevel, and speed, resulting in poor adaptability under complex working conditions.

[0008] Therefore, it is necessary to provide a more precise and comprehensive method for dynamically adjusting the swing width of an all-position MAG welding robot. Summary of the Invention

[0009] To overcome the above deficiencies, this invention provides a method for dynamically adjusting the swing width of an all-position MAG welding robot. By integrating arc sensing and visual sensing information, and combining welding position recognition and gain scheduling adaptive control, the method achieves real-time dynamic adjustment of the welding swing width, thereby improving the weld formation quality and welding stability under complex posture and irregular bevel conditions.

[0010] This invention provides the following technical solution: a method for dynamically adjusting the swing width of an all-position MAG welding robot, comprising:

[0011] Welding electrical signals are collected by an arc sensor, and the bevel shape image is obtained by a vision sensor. The current welding spatial position is determined based on the robot trajectory coordinates.

[0012] The image of the bevel shape is processed to extract the bevel geometric features, including the gap width. Bevel depth and bevel angle And calculate the bevel filling coefficient. ,when When it is determined to be a deep and narrow bevel, When it is determined to be a wide and shallow bevel, when The weld bevel is determined to be a standard bevel. The welding electrical signal is filtered and processed to extract the electrical characteristics of the left and right sides during the oscillation process. The deviation of the weld centerline is calculated based on the difference in electrical characteristics, including the direction and amount of deviation.

[0013] Based on the trajectory coordinates of the robot's end effector, the angle of the welding torch relative to the direction of gravity is calculated through coordinate transformation to identify the current welding position type.

[0014] Establish a multi-factor coupled model for swing width and calculate the target swing width. : ;

[0015] in, The reference swing width; This is a position correction factor, based on the welding torch tilt angle. The nonlinear mapping table is obtained by searching a preset nonlinear mapping table, which is established based on the influence of different welding positions on the flowability of the molten pool. This is the bevel filling correction factor, based on the bevel filling factor. Sure; This is a welding speed correction factor, determined based on the current welding speed v;

[0016] Based on the groove geometry, welding position type and welding electrical signal, the target swing width under the current working condition is calculated by the gain scheduling algorithm. The gain scheduling algorithm is implemented by piecewise linear interpolation. Different gain matrices are set for flat welding position, vertical welding position, overhead welding position and horizontal welding position respectively. At the same time, the correction amount of the swing trajectory center position is calculated according to the deviation of the weld center line.

[0017] Based on the target swing width and trajectory center position correction amount, the robot welding process parameters are adjusted collaboratively, and the adjustment instructions are sent to the robot controller for execution;

[0018] During the welding process, at preset time intervals Collect weld deviation ,when At that time, according to Fine-tuning of the swing width is performed, among which This is a proportionality coefficient, with a value ranging from 0.5 to 2.0. The deviation threshold is used; the changes in welding electrical signals and weld formation quality are continuously monitored, and the parameters of the gain scheduling algorithm are dynamically corrected based on the feedback information to achieve closed-loop control.

[0019] Preferably, the step of extracting the geometric features of the bevel includes:

[0020] The bevel shape image is preprocessed by grayscale conversion and filtering to remove welding arc light interference and noise;

[0021] An edge detection algorithm is used to extract the bevel edge contour, and a fitting algorithm is used to determine the left and right bevel boundary lines and feature points at the bottom of the bevel.

[0022] Calculate the geometric characteristic parameters between the left and right bevel boundary lines, including the gap width W, bevel angle α, and bevel depth D.

[0023] Preferably, the step of calculating the deviation of the weld centerline includes:

[0024] The welding arc current signal and voltage signal are acquired synchronously, and a filtering algorithm is used to remove high-frequency noise;

[0025] Based on the welding torch oscillation cycle, the electrical signal is divided into a left oscillation segment and a right oscillation segment. The voltage and current values ​​of the left and right oscillation segments are collected to obtain the arc energy characteristics on both sides.

[0026] The offset direction and amount of the weld centerline are determined based on the voltage or current difference between the left and right sides. According to the formula Calculation, where , These are the current values ​​for the left and right swing segments, respectively. , These are the voltage values ​​for the left and right swing segments, respectively. , These are calibration coefficients.

[0027] Preferably, the step of identifying the current welding position type includes:

[0028] Calculate the angle between the welding torch axis and the direction of gravity based on the real-time trajectory coordinates and attitude information of the robot's end effector.

[0029] The included angle is compared with a preset welding position angle threshold to determine the type of welding position, which includes flat welding position, vertical welding position, overhead welding position and horizontal welding position.

[0030] Preferably, the position correction coefficient The rules for determining the value are as follows:

[0031] When identified as a flat weld position The value range is 1.0-1.1;

[0032] When identified as a vertical welding position The value range is 0.8-0.9;

[0033] When identified as a horizontal welding position The value range is 0.9-1.0;

[0034] When identified as an overhead welding position The value range is 0.7-0.8.

[0035] Preferably, the bevel filling correction coefficient The rules for determining the value are as follows:

[0036] When the bevel filling coefficient When it is determined to be a wide and shallow bevel The value range is 1.1-1.3;

[0037] When the bevel filling coefficient When it is determined to be a standard bevel The value is 1.0;

[0038] When the bevel filling coefficient When it is determined to be a deep and narrow bevel, The value range is 0.7-0.9.

[0039] Preferably, the step of calculating the correction amount for the center position of the swing trajectory includes:

[0040] Using the deviation of the weld centerline as input, a nonlinear mapping relationship between the center position of the welding torch oscillation trajectory and the deviation is established;

[0041] Based on the direction and amount of deviation of the weld centerline, calculate the correction amount of the center position of the welding torch oscillation trajectory in the lateral direction.

[0042] The calculated correction value is filtered and limited, and the corrected swing trajectory center position parameter is sent to the robot controller.

[0043] Preferably, the steps for adjusting the robot welding process parameters include:

[0044] Based on the correction amount of the target swing width and the center position of the swing trajectory, the motion compensation amount of the welding torch tip is calculated in real time.

[0045] The welding trajectory instructions in the robot controller are corrected, and the lateral swing amplitude, swing frequency and swing center position of the welding torch are dynamically adjusted.

[0046] Preferably, the step of dynamically correcting the gain scheduling algorithm parameters includes:

[0047] The weld formation quality is continuously monitored by visual sensors or arc sensors, and welding quality features are extracted, including weld width, reinforcement height and fusion state.

[0048] Establish a correlation between weld formation quality and swing width control effect; when the quality deviation exceeds the preset threshold, trigger algorithm parameter correction.

[0049] Based on the direction and magnitude of the deviation, the control gain coefficient in the gain scheduling algorithm is adjusted using the gradient descent method, and the corrected parameters are updated in real time.

[0050] The present invention also provides a dynamic adjustment system for the swing width of an all-position MAG welding robot, comprising:

[0051] The data acquisition module is used to acquire welding electrical signals through an arc sensor, obtain bevel shape images through a vision sensor, and determine the current welding spatial position based on the robot trajectory coordinates.

[0052] The feature extraction module is used to process the bevel shape image and extract the geometric features of the bevel, including the gap width. Bevel depth and bevel angle And calculate the bevel filling coefficient. The welding electrical signal is filtered to extract the electrical characteristics of the left and right sides during the oscillation process, and the deviation of the weld centerline is calculated based on the difference in electrical characteristics, including the direction and amount of deviation.

[0053] The welding position type identification module is used to calculate the angle of the welding torch relative to the direction of gravity through coordinate transformation based on the trajectory coordinates of the robot's end effector, and to identify the current welding position type.

[0054] The swing trajectory correction module is used for multi-factor coupling models based on swing width. The target swing width under the current working condition is calculated by the gain scheduling algorithm, and the correction amount of the swing trajectory center position is calculated according to the deviation of the weld center line. The gain scheduling algorithm adopts a piecewise linear interpolation method and sets different gain matrices for different welding positions.

[0055] The welding process parameter adjustment module is used to coordinately adjust the robot welding process parameters according to the target swing width and trajectory center position correction amount, and send the adjustment command to the robot controller for execution;

[0056] Closed-loop control module, used at preset time intervals Collect weld deviation ,when According to Fine-tuning of the swing width is performed, among which The value range is 0.5-2.0; the welding electrical signal changes and weld formation quality are continuously monitored, and the parameters of the gain scheduling algorithm are dynamically corrected based on the feedback information to achieve closed-loop control.

[0057] The present invention has the following beneficial effects:

[0058] 1. This invention establishes a quantitative mapping relationship between the groove depth-to-width ratio and the swing width by introducing a groove filling coefficient. Different swing strategies are adopted for deep / narrow grooves and wide / shallow grooves, solving the problems of insufficient filling in deep / narrow grooves or lack of edge fusion in wide / shallow grooves caused by simple adjustments based solely on groove width in existing technologies. Experimental verification shows that, under vertical welding conditions with deep / narrow grooves, the method of this invention significantly improves the groove filling rate, significantly reduces the weld reinforcement, substantially decreases the edge lack of fusion rate, and significantly improves the weld formation coefficient compared to the traditional fixed swing width method.

[0059] 2. This invention establishes a multi-factor coupled model that comprehensively considers the synergistic effects of welding position, groove geometry, and welding speed on the oscillation width. To address the nonlinear influence of different welding positions on the molten pool's fluidity, a pre-set gain scheduling table is used for precise correction. Specifically, the position correction coefficient for the overhead welding position significantly reduces the oscillation width, effectively preventing molten pool sagging. This model overcomes the shortcomings of existing technologies that use linear interpolation methods, which cannot adapt to nonlinear changes, and significantly improves weld width consistency in all-position welding.

[0060] 3. This invention achieves real-time adaptive adjustment of the swing width through a closed-loop control mechanism. The system periodically collects weld deviations, and triggers fine-tuning of the swing width when the deviation exceeds a preset threshold. The gain scheduling algorithm parameters are dynamically corrected using a gradient descent method. This method achieves coordinated control of multiple factors such as position, bevel, and speed. Compared to existing technologies that lack a multi-factor coordination mechanism, it significantly reduces the incidence of defects such as weld misalignment and incomplete fusion, and improves the intelligence and long-term stability of the welding process. Attached Figure Description

[0061] Figure 1 is a flowchart of a dynamic adjustment method for the swing width of an all-position MAG welding robot proposed in this invention.

[0062] Figure 2 is a flowchart of a dynamic adjustment system for the swing width of an all-position MAG welding robot proposed in an embodiment of the present invention. Detailed Implementation

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0064] Example 1

[0065] In a first embodiment of the present invention, the present invention provides a method for dynamically adjusting the swing width of an all-position MAG welding robot, as shown in Figure 1, including:

[0066] S1. Acquire welding electrical signals through an arc sensor, obtain bevel shape images through a vision sensor, and determine the current welding spatial position based on the robot trajectory coordinates;

[0067] Specifically, an arc sensor is installed at the end of the welding torch or on the welding power source of the MAG welding robot to collect the current and voltage signals generated during the welding process. The current signal output by the sensor is converted into a digital signal by a sampling circuit, and the voltage signal is sampled by a voltage sampling module and input to the data acquisition system. The sampling frequency can be set from 1 to 10 kHz.

[0068] Industrial cameras or laser profiling sensors are installed in the welding area to acquire images of the weld bevel shape. Industrial cameras can be equipped with filters and narrow-band light sources to reduce interference from welding arc light and ensure image clarity.

[0069] By reading the trajectory coordinates (X, Y, Z and attitude angles) of the robot's end effector and combining them with camera calibration information, the precise position of the welding torch in three-dimensional space is calculated. The weld bevel image is then mapped to the spatial position of the welding torch, and the image features are mapped to the welding coordinate system, achieving precise spatial positioning of the weld.

[0070] S2. Process the bevel shape image to extract the bevel geometric features; filter the welding electrical signal to extract the electrical features on the left and right sides during the oscillation process, and calculate the weld centerline deviation based on the difference in electrical features, including the deviation direction and deviation amount.

[0071] Preferably, the step of extracting the geometric features of the bevel includes:

[0072] The bevel shape image is preprocessed by grayscale conversion and filtering to remove welding arc light interference and noise;

[0073] An edge detection algorithm is used to extract the bevel edge contour, and a fitting algorithm is used to determine the left and right bevel boundary lines and feature points at the bottom of the bevel.

[0074] Calculate the geometric characteristic parameters between the left and right bevel boundary lines, including the gap width, bevel angle, and bevel depth.

[0075] Furthermore, the system also calculates the bevel filling coefficient. Defined as bevel depth With gap width The ratio, i.e. This coefficient reflects the depth-to-width ratio of the bevel and is used for precise control of the subsequent swing width. Based on the bevel filling coefficient... Based on the size, the system classifies bevels into three categories:

[0076] when When the bevel is wide and shallow, it is judged to be a wide and shallow bevel. Such a bevel requires a large swing width during welding to ensure that the bevel edge is fully fused.

[0077] when When the groove is deemed to be a standard bevel with a suitable depth-to-width ratio, welding can be performed using the standard oscillation width.

[0078] when When the groove is deep and narrow, it is judged to be a deep and narrow groove. When welding this type of groove, the oscillation width needs to be reduced to ensure that the welding wire can fully fill the bottom of the groove and avoid defects such as incomplete filling or depression in the middle of the weld.

[0079] The aforementioned thresholds are derived from statistical analysis of a large amount of MAG welding experimental data. In practical applications, when the groove filling coefficient approaches the threshold, the system can use linear interpolation to smoothly transition the control parameters, avoiding welding instability caused by abrupt changes in the control strategy.

[0080] Specifically, the bevel images captured by the industrial camera at the tip of the robotic welding torch or in the welding area are converted to grayscale, transforming the color images into single-channel grayscale images to simplify subsequent image processing calculations. Gaussian filtering or median filtering is then applied to the grayscale images to remove welding arc light, spatter, and high-frequency noise, improving the continuity and clarity of the bevel edges.

[0081] Edge detection algorithms, such as the Canny operator, Sobel operator, or Laplacian operator, are used to extract the bevel edge contours from the filtered grayscale image. The resulting set of edge points is used to describe the left and right boundaries and bottom shape of the bevel. The extracted edge points are then fitted to the left and right bevel boundary lines, using least-squares linear fitting, curve fitting, or polynomial fitting methods to reduce the influence of edge noise. The feature points at the bottom of the bevel, i.e., the intersection or approximate intersection points of the two bevel boundary lines, are determined from the fitting results and used to calculate the bevel depth.

[0082] Based on the fitted left and right bevel boundary lines and bottom feature points, calculate the following geometric feature parameters: gap width, which is the horizontal distance between the left and right boundary lines at the bevel opening; bevel angle, which is the angle between the left and right boundary lines and the horizontal line or weld center line; and bevel depth, which is the vertical distance from the bevel opening to the bottom feature point.

[0083] By performing grayscale conversion, filtering preprocessing, edge detection, and boundary fitting on the bevel image, the geometric feature parameters of the bevel, including gap width, bevel angle, and bevel depth, can be accurately extracted, providing reliable geometric information for subsequent calculation of weld centerline deviation and adaptive swing width adjustment.

[0084] Preferably, the step of calculating the deviation of the weld centerline includes:

[0085] The welding arc current signal and voltage signal are acquired synchronously, and a filtering algorithm is used to remove high-frequency noise;

[0086] Based on the welding torch oscillation cycle, the electrical signal is divided into a left oscillation segment and a right oscillation segment. The voltage and current values ​​of the left and right oscillation segments are collected to obtain the arc energy characteristics on both sides.

[0087] The offset direction and amount of the weld centerline are determined based on the voltage or current difference between the left and right sides.

[0088] Specifically, current and voltage sensors are installed at the end of the welding torch or the welding power source of the MAG welding robot to collect welding current signals in real time. and voltage signal The sensor signal is converted into a digital signal by the sampling module and synchronized with a timestamp to ensure that the current and voltage signals correspond under the same time base. The acquired current and voltage signals undergo high-frequency noise filtering, which can be achieved using a low-pass filter or a moving average filter to remove transient arc interference and sensor noise.

[0089] Based on the period or period marker of the welding torch oscillation, the electrical signal is divided into left and right oscillation segments. The start and end points of each oscillation cycle can be determined by changes in the robot trajectory coordinates or the lateral displacement information of the welding torch. When the welding torch oscillates to its left limit position, the welding current at that moment is collected. and voltage When the swing reaches the right limit position, the current is collected. and voltage .

[0090] The offset Δd of the weld centerline is calculated using the comprehensive evaluation method of arc energy difference, and the calculation formula is as follows: ;

[0091] in: , The values ​​are the average current values ​​(in A) for the left and right swing segments, respectively. , These are the average voltage values ​​(in V) for the left and right swing segments, respectively. , These are calibration coefficients, determined through offline calibration experiments.

[0092] Method for determining calibration coefficients: Welding is performed on a standard specimen with known weld position deviations, current and voltage data are collected, and the coefficients are obtained through least squares fitting. and For typical MAG welding processes, The value range of is generally 0.01-0.05 mm / A, and the value range of K2 is generally 0.1-0.3 mm / V.

[0093] offset A positive value indicates that the weld centerline has shifted to the left; a negative value indicates that it has shifted to the right. According to... The system calculates the lateral correction amount of the center position of the swing trajectory based on the size of the swing trajectory and performs real-time compensation for the robot's motion trajectory.

[0094] This formula comprehensively considers the influence of two electrical parameters, current and voltage, on weld deviation. Current differences primarily reflect the asymmetry in penetration depth and heat input, while voltage differences primarily reflect the asymmetry in arc length and the distance between the welding torch and the workpiece. By fusing these two parameters, the actual position of the weld centerline can be determined more accurately.

[0095] By synchronously acquiring current and voltage signals, and combining this with the analysis of the swing segment division and the energy characteristics of the left and right arcs, the offset direction and amount of the weld centerline can be accurately calculated. The offset amount is calculated as follows:

[0096] ;

[0097] in , The calibration coefficient. When When, it indicates that the welding torch is deflected to the right; when When the welding torch is tilted to the left, it indicates that the welding torch is deflected to the left.

[0098] S3. Based on the trajectory coordinates of the robot's end effector, calculate the angle of the welding torch relative to the direction of gravity through coordinate transformation, and identify the current welding position type;

[0099] Preferably, the step of identifying the current welding position type includes:

[0100] Calculate the angle between the welding torch axis and the direction of gravity based on the real-time trajectory coordinates and attitude information of the robot's end effector.

[0101] The included angle is compared with a preset welding position angle threshold to determine the type of welding position, which includes flat welding position, vertical welding position, overhead welding position and horizontal welding position.

[0102] Specifically, the robot controller acquires the real-time trajectory coordinates and attitude information (roll, pitch, and yaw angles represented by Euler angles or quaternions) of the end effector. This attitude information is used to determine the direction vector of the welding torch axis in three-dimensional space. Calculation of the angle between the welding torch and the direction of gravity.

[0103] Define the direction of gravity as a unit vector. The angle between the welding torch axis and the direction of gravity is... It can be calculated using the vector dot product: Based on the preset welding position angle threshold, the included angle is... Compared to a range of different welding positions, in one embodiment, the flat weld angle... Vertical welding angle ; Horizontal welding angle Overhead welding angle .

[0104] By acquiring the trajectory coordinates and attitude information of the welding torch tip in real time, calculating the angle between the welding torch axis and the direction of gravity, and combining it with a preset threshold to determine the welding position type, welding positions such as flat welding, vertical welding, overhead welding, and horizontal welding can be accurately identified, providing reliable input for adaptive gain scheduling and target swing width calculation.

[0105] S4. Based on the groove geometry, welding position type and welding electrical signal, calculate the target swing width under the current working condition using a gain scheduling algorithm, and at the same time calculate the correction amount of the swing trajectory center position based on the weld centerline deviation.

[0106] The core of this invention lies in establishing a multi-factor coupled model for the swing width. This model comprehensively considers the influence of welding position, groove geometry, and welding speed on the swing width. The calculation formula is as follows: ;

[0107] in:

[0108] The reference swing width is determined based on the bevel gap width W, and is generally taken as... ;

[0109] This is a position correction factor, reflecting the influence of different welding positions on the fluidity of the molten pool. Since the effect of gravity on the molten pool varies at different spatial positions, the correction factor needs to be obtained by looking up a pre-defined nonlinear mapping table based on the welding torch tilt angle θ. Table 1 shows the range of correction factor values ​​for typical welding positions:

[0110] Table 1: Position Correction Factors Value Table

[0111] Welding position tilt angle θ range k1(θ) value range adjustment basis Flat welding (1G) 0°-45° 1.0-1.1 Stable molten pool, can appropriately increase the swing width Vertical welding (3G) 45°-90° 0.8-0.9 Lateral gravity effect, swing width needs to be reduced Horizontal welding (2G) 90°-135° 0.9-1.0 Fine adjustment according to uphill / downhill direction Overhead welding (4G) 135°-180° 0.7-0.8 Molten pool is prone to sagging, swing width needs to be significantly reduced surface

[0112] This is the bevel filling correction factor, based on the bevel filling factor. Confirmed. The bevel filling coefficient reflects the depth-to-width ratio of the bevel; different depth-to-width ratios require different sizing strategies.

[0113] Table 2: Correction Factors for Beveling Fill Value Table

[0114] Explanation of the adjustment strategy for the filling coefficient ηk2(η) of the bevel type: Wide and shallow bevel: η < 0.4 1.1-1.3 Small depth-to-width ratio, require increasing the bevel width to ensure edge fusion. Standard bevel: 0.4 ≤ η < 0.6 1.0 Use standard bevel width, no correction required. Deep and narrow bevel: η ≥ 0.6 0.7-0.9 Large depth-to-width ratio, require decreasing the bevel width to prevent incomplete filling. surface

[0115] by and The threshold value was derived from a large amount of experimental data:

[0116] when At this time, the bevel exhibits obvious wide and shallow characteristics, and too small an oscillation range will lead to poor fusion at the edge of the bevel;

[0117] when At this time, the bevel is deep and narrow, and an excessively wide oscillation width will prevent the welding wire from fully filling the bottom of the bevel.

[0118] This is a welding speed correction factor, determined based on the current welding speed v. The faster the welding speed, the less molten metal is deposited per unit time, requiring a corresponding reduction in the oscillation width to ensure proper filling. Generally, it is taken as... ,in This is the standard welding speed.

[0119] The above multi-factor coupling model can achieve accurate calculation of swing width, which can significantly improve the weld formation quality compared with the traditional fixed swing width method or single-factor adjustment method.

[0120] This invention employs a gain scheduling algorithm to dynamically calculate each correction coefficient. The gain scheduling algorithm is implemented using piecewise linear interpolation, specifically including:

[0121] Different gain matrices are set for flat welding, vertical welding, horizontal welding, and overhead welding positions. Position correction coefficients are used. Taking the calculation as an example, for the vertical welding position, the following piecewise linear interpolation is used:

[0122] when hour: ;

[0123] when hour: .

[0124] Piecewise linear interpolation can achieve a balance between computational efficiency and accuracy. It is smoother than the lookup table method and has less computational cost than nonlinear fitting, making it suitable for real-time control applications.

[0125] Bevel filling correction factor The calculation also employs a similar piecewise linear interpolation method. Furthermore, the system includes a boundary protection mechanism; when the calculated value... , , When the amplitude exceeds a reasonable range, automatic amplitude limiting is applied to ensure system stability.

[0126] Preferably, the step of calculating the target swing width under the current operating condition includes:

[0127] The bevel geometry, welding position type, and welding electrical signals are used as input variables, and the corresponding initial values ​​of the control gain are set.

[0128] Based on the real-time detected changes in input variables, the control gain coefficients are dynamically adjusted to form a gain scheduling table that changes with the operating conditions.

[0129] Within each control cycle, the target swing width is calculated using the gain scheduling table to match the current welding posture, bevel gap, and arc state.

[0130] Specifically, before welding begins, the system collects input variables in real time through sensors and a robot controller, including bevel geometry parameters. Welding position type and welding electrical signals The system sets the initial control gain parameters based on process experience. , and , which serves as the initial value for the gain scheduling algorithm.

[0131] A hierarchical gain scheduling algorithm is used to map different influencing factors into adjustment factors. ,in This represents the overall control gain coefficient. During the welding process, the system dynamically calculates the changes in each input variable based on sensor update data for each control cycle (e.g., 100ms).

[0132] ;

[0133] in, The dynamic gain scheduling table is determined by the rate of change of the input variables and the sensitivity coefficient. The updated combinations of control gains form the dynamic gain scheduling table. It stores the gain distribution under different operating conditions.

[0134] Calculate the target swing width based on the gain scheduling table. :

[0135] ;

[0136] in, As the reference swing width, , These are the initial operating conditions. This is the welding position correction function. The calculation result is smoothed and filtered before being output to the controller for real-time adjustment of the welding torch oscillation amplitude.

[0137] By introducing a gain scheduling algorithm based on multi-parameter input, the oscillation width of the welding robot can be adaptively adjusted under different welding positions and bevel shapes. This method can dynamically match the spatial posture of the weld with the arc characteristics, maintaining weld formation consistency and penetration uniformity.

[0138] Preferably, the step of calculating the correction amount for the center position of the swing trajectory includes:

[0139] Using the deviation of the weld centerline as input, a nonlinear mapping relationship between the center position of the welding torch oscillation trajectory and the deviation is established;

[0140] Based on the direction and amount of deviation of the weld centerline, calculate the correction amount of the center position of the welding torch oscillation trajectory in the lateral direction.

[0141] The calculated correction value is filtered and limited, and the corrected swing trajectory center position parameter is sent to the robot controller.

[0142] Specifically, the welding robot calculates the deviation and direction of the current weld centerline using an arc sensing module. The system establishes a nonlinear mapping model between the center position of the welding torch oscillation trajectory and the deviation of the weld centerline, for example, through empirical data fitting, fuzzy logic, or neural network algorithms. When the deviation is small, the system makes minor adjustments proportionally; when the deviation is large, segmented control or amplitude limiting is used to prevent excessive adjustment of the welding torch trajectory.

[0143] During the calculation process, the system determines the correction direction based on the deviation direction: when the weld seam is detected to be deviating to the left, the system controls the welding torch's swing center to correct to the right; when the weld seam is detected to be deviating to the right, the system controls the welding torch's swing center to correct to the left; if the deviation is zero, the welding torch's current position remains unchanged. To avoid welding torch jitter caused by arc fluctuations or signal noise, the system filters the calculated correction amount to remove short-term abrupt changes.

[0144] The corrected swing trajectory center position parameters are sent to the robot controller in real time. The controller adjusts the swing path of the welding torch according to the new trajectory center coordinates, so that the weld center is re-aligned with the target welding position. The entire process is continuously executed in a loop during the welding process, thereby achieving real-time tracking and automatic centering control of the weld centerline.

[0145] By establishing a nonlinear mapping relationship between the weld centerline deviation and the center position of the welding torch oscillation trajectory, and combining filtering and amplitude limiting processing, accurate adaptive correction of the welding torch trajectory center can be achieved.

[0146] S5. Based on the target swing width and trajectory center position correction amount, coordinately adjust the robot welding process parameters and send the adjustment command to the robot controller for execution;

[0147] Preferably, the steps for adjusting the robot welding process parameters include:

[0148] Based on the correction amount of the target swing width and the center position of the swing trajectory, the motion compensation amount of the welding torch tip is calculated in real time.

[0149] The welding trajectory instructions in the robot controller are corrected, and the lateral swing amplitude, swing frequency and swing center position of the welding torch are dynamically adjusted.

[0150] Specifically, the motion compensation amount at the end of the welding torch is obtained by comprehensively calculating the target swing width and the correction result of the weld centerline deviation. This compensation amount is used to correct the range of motion and center position of the welding torch during lateral swing, thereby ensuring that the weld always remains on the correct welding path. The motion compensation calculation module takes the welding state parameters acquired in real time as input and calculates the required lateral offset value and swing amplitude correction value at the end of the welding torch through a control algorithm.

[0151] The calculated compensation parameters are input into the robot controller to correct the original welding trajectory commands in real time. Based on the corrected parameters, the controller dynamically adjusts the amplitude, frequency, and center position of the welding torch's lateral oscillation. When a change in the bevel gap or welding posture is detected, the system automatically updates the control parameters to ensure the welding torch trajectory matches the actual weld shape. This allows for automatic matching of appropriate oscillation modes and welding parameters at different welding positions.

[0152] Using the above method, the system can automatically adjust the welding torch's trajectory and oscillation parameters based on real-time weld condition monitoring, achieving adaptive control of the welding process. This method significantly improves the consistency and stability of weld formation and reduces manual intervention.

[0153] S6. Continuously monitor changes in welding electrical signals and weld formation quality, and dynamically adjust the parameters of the gain scheduling algorithm based on feedback information to achieve closed-loop control.

[0154] This invention achieves real-time correction of the oscillation width through closed-loop control. Specifically, during the welding process, at preset time intervals... Collect weld deviation This deviation is obtained through a visual sensor or an arc sensor. When the absolute value of the deviation... Exceeding the preset deviation threshold At that time, the swing width fine-tuning mechanism is triggered.

[0155] The correction amount for the swing width is calculated according to the proportional control law:

[0156] ;

[0157] in This is a proportionality coefficient, with a value ranging from 0.5 to 2.0. The specific value is determined based on the welding conditions and material properties: for welding thick plates with good rigidity, a larger value (such as 1.5-2.0) can be taken to achieve a fast response; for thin plates or workpieces that are sensitive to deformation, a smaller value (such as 0.5-1.0) should be taken to avoid excessive adjustment that could lead to weld fluctuations.

[0158] The corrected swing width is: The system will Conduct a reasonableness check to ensure it is within the allowable range (e.g., not less than 0.8 times the bevel width, and not more than 1.5 times).

[0159] In addition, the system continuously monitors changes in welding electrical signals and weld formation quality, extracting welding quality features (including weld width, reinforcement height, fusion state, etc.). When the weld formation quality deviation exceeds a preset threshold, the gain scheduling algorithm parameters are corrected. Parameter correction uses the gradient descent method, adjusting the control gain coefficient according to the direction and magnitude of the deviation to achieve adaptive optimization. The step size of gradient descent is generally set to 0.01-0.05, and the iteration period is every 10-20 weld points.

[0160] Preferably, the step of dynamically correcting the gain scheduling algorithm parameters includes:

[0161] The weld formation quality is continuously monitored by visual sensors or arc sensors, and welding quality features are extracted, including weld width, reinforcement height and fusion state.

[0162] Establish a correlation between weld formation quality and swing width control effect; when the quality deviation exceeds the preset threshold, trigger algorithm parameter correction.

[0163] Based on the direction and magnitude of the deviation, the control gain coefficient in the gain scheduling algorithm is adjusted using the gradient descent method, and the corrected parameters are updated in real time.

[0164] Specifically, visual sensors or arc sensors are used to continuously monitor the weld formation status. Visual sensors acquire image data of the weld formation area, and image processing algorithms extract geometric features such as weld width, reinforcement height, and weld edge morphology; arc sensors, on the other hand, can indirectly reflect the fusion state and formation uniformity of the weld by analyzing current and voltage fluctuation characteristics.

[0165] A correlation model is established between weld formation quality and the effectiveness of oscillation width control to determine whether the current welding state is under ideal conditions. When the monitored weld width, reinforcement height, or fusion depth deviates from the preset target range, the system considers the welding quality to have deviated. The deviation value can be calculated by the difference between the target value and the actual measured value.

[0166] When the deviation exceeds a preset threshold, the system automatically triggers the algorithm correction process. The control module adjusts the control gain coefficient in the gain scheduling algorithm using the gradient descent method based on the direction and magnitude of the deviation. Specifically, when the weld width is too small or the fusion is insufficient, the system gradually increases the gain coefficient related to the oscillation width; when the weld width is too large or the fusion depth is too deep, the system appropriately reduces the control gain. The gradient descent algorithm ensures that the system tends towards the optimal control state after multiple iterations by gradually correcting the parameters.

[0167] Through the aforementioned dynamic correction mechanism, the stability and consistency of weld formation quality can be continuously improved, and the risk of welding defects caused by posture changes, heat input fluctuations, or uneven bevels can be significantly reduced.

[0168] Example 2

[0169] In the manufacturing process of large pressure vessels, circumferential welding at vertical welding positions is often required. Due to gravity, the molten metal in the vertical welding state is prone to sag, resulting in uneven weld formation. At the same time, the arc stability is poor and the groove gap distribution is uneven during vertical welding, making it difficult to balance the traditional welding control method with fixed swing width to achieve both penetration and weld formation quality. This can easily lead to problems such as excessive weld reinforcement, uneven fusion on both sides, and weld deviation.

[0170] To address the aforementioned problems, this invention provides a dynamic adjustment system for the swing width of an all-position MAG welding robot, the structure of which is shown in Figure 2. The system includes:

[0171] The data acquisition module is used to acquire welding electrical signals through an arc sensor, obtain bevel shape images through a vision sensor, and determine the current welding spatial position based on the robot trajectory coordinates.

[0172] The feature extraction module is used to process the bevel shape image and extract the bevel geometric features; to filter the welding electrical signal, extract the electrical features on the left and right sides during the oscillation process, and calculate the weld centerline deviation based on the difference in electrical features, including the deviation direction and deviation amount.

[0173] The welding position type identification module is used to calculate the angle of the welding torch relative to the direction of gravity through coordinate transformation based on the trajectory coordinates of the robot's end effector, and to identify the current welding position type.

[0174] The swing trajectory correction module is used to calculate the target swing width under the current working condition based on the groove geometry, welding position type and welding electrical signal, and to calculate the swing trajectory center position correction amount according to the weld centerline deviation.

[0175] The welding process parameter adjustment module is used to coordinately adjust the robot welding process parameters according to the target swing width and trajectory center position correction amount, and send the adjustment command to the robot controller for execution;

[0176] The closed-loop control module is used to continuously monitor changes in welding electrical signals and weld formation quality, and dynamically corrects the parameters of the gain scheduling algorithm based on feedback information to achieve closed-loop control.

[0177] Example 3

[0178] To verify the technical effect of the present invention, a comparative experiment was conducted on the welding of V-groove of Q235 steel plate using the method of the present invention and the traditional fixed swing width method.

[0179]

Experimental Conditions

[0180] [Method Comparison] Traditional method: using a fixed swing width of 10mm and a swing frequency of 1.2Hz.

[0181] The method of this invention: calculates the swing width based on a multi-factor coupling model.

[0182] Reference swing width: ;

[0183] Position correction factor: (Vertical welding, refer to Table 1);

[0184] Bevel filling correction factor: (For deep and narrow bevels, refer to Table 2);

[0185] Speed ​​correction factor: (Standard speed);

[0186] The calculation yields: .

[0187] Actual control: real-time fine-tuning based on weld deviation, ranging from 10.8 to 11.8 mm.

[0188]

Experimental Results

[0189] Table 3: Comparison of Welding Quality Data (n=10, Average Value)

[0190] Evaluation Indicators: Traditional Fixed Swing Width, Improved Effect of Invention Method: Average Swing Width 10.0mm (Fixed) 11.2mm (Dynamic) +12%; Weld Reinforcement Height 3.2±0.4mm 2.1±0.2mm -34.4%; Weld Width 11.5±0.8mm 12.3±0.3mm +7.0%; Edge Lack of Fusion Rate 15% 3% -80%; Bevel Filling Rate 82% 96% +14%; Weld Forming Coefficient 0.68 0.89 +30.9%. surface

[0191] Note: Weld formation coefficient = weld width / (2 × reinforcement height), the closer to 1, the better the formation.

[0192] As can be seen from Table 3, the method of the present invention has significant advantages over the traditional fixed swing width method:

[0193] (1) By introducing the bevel filling coefficient And appropriately reduce the filling correction factor based on the characteristics of deep and narrow bevels. This makes the swing width more compatible with the groove geometry, reduces the weld reinforcement by 34.4%, and increases the groove filling rate from 82% to 96%.

[0194] (2) For vertical welding positions, a position correction factor is used. This effectively compensates for the lateral effect of gravity on the molten pool, reducing the edge non-fusion rate from 15% to 3%;

[0195] (3) Real-time fine adjustment of the swing width is achieved through closed-loop control, and the standard deviation of the weld width is reduced from 0.8 mm to 0.3 mm, significantly improving consistency;

[0196] (4) The weld formation coefficient increased from 0.68 to 0.89, which is closer to the ideal value of 1.0, indicating that the appearance quality of the weld has been significantly improved.

[0197] This experiment fully demonstrates the effectiveness and advancement of the method of the present invention.

[0198] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for dynamically adjusting the swing width of an all-position MAG welding robot, characterized in that, include: Welding electrical signals are collected by an arc sensor, and the bevel shape image is obtained by a vision sensor. The current welding spatial position is determined based on the robot trajectory coordinates. The image of the bevel shape is processed to extract the bevel geometric features, including the gap width. Bevel depth and bevel angle And calculate the bevel filling coefficient. ,when When it is determined to be a deep and narrow bevel, When it is determined to be a wide and shallow bevel, when The process involves several steps: first, determining the weld bevel as a standard bevel; second, filtering the welding electrical signal to extract the electrical characteristics of the left and right sides during the oscillation, and calculating the weld centerline deviation based on the differences in these characteristics, including the direction and amount of deviation; third, calculating the angle of the welding torch relative to the direction of gravity through coordinate transformation based on the trajectory coordinates of the robot's end effector to identify the current welding position type; and finally, establishing a multi-factor coupling model for the oscillation width to calculate the target oscillation width. ： ;in, The reference swing width; This is a position correction factor, based on the welding torch tilt angle. The nonlinear mapping table is obtained by searching a preset nonlinear mapping table, which is established based on the influence of different welding positions on the flowability of the molten pool. This is the bevel filling correction factor, based on the bevel filling factor. Sure; The welding speed correction coefficient is determined based on the current welding speed v. Based on the groove geometry, welding position type, and welding electrical signals, the target swing width under the current working condition is calculated using a gain scheduling algorithm. This gain scheduling algorithm employs piecewise linear interpolation, setting different gain matrices for flat, vertical, overhead, and horizontal welding positions. Simultaneously, the correction amount for the swing trajectory center position is calculated based on the weld centerline deviation. Based on the target swing width and the trajectory center position correction amount, the robot welding process parameters are adjusted collaboratively, and the adjustment command is sent to the robot controller for execution. During the welding process, at preset time intervals… Collect weld deviation ,when At that time, according to Fine-tuning of the swing width is performed, among which This is a proportionality coefficient, with a value ranging from 0.5 to 2.

0. This is the deviation threshold; Continuous monitoring of changes in welding electrical signals and weld formation quality, and dynamic adjustment of the parameters of the gain scheduling algorithm based on feedback information to achieve closed-loop control.

2. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for extracting the geometric features of the bevel include: performing grayscale conversion and filtering preprocessing on the bevel shape image to remove welding arc light interference and noise; using an edge detection algorithm to extract the bevel edge contour, and determining the left and right bevel boundary lines and feature points at the bottom of the bevel based on a fitting algorithm; and calculating the geometric feature parameters between the left and right bevel boundary lines, including the gap width W, the bevel angle α, and the bevel depth D.

3. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for calculating the weld centerline deviation include: synchronously acquiring welding arc current and voltage signals, and using a filtering algorithm to remove high-frequency noise; dividing the electrical signal into a left and right swing segment according to the welding torch oscillation period, acquiring the voltage and current values ​​of the left and right swing segments to obtain the arc energy characteristics on both sides; determining the offset direction and amount of the weld centerline based on the voltage or current difference between the left and right sides, wherein the offset amount... According to the formula Calculation, where 、 These are the current values ​​for the left and right swing segments, respectively. 、 These are the voltage values ​​for the left and right swing segments, respectively. 、 These are calibration coefficients.

4. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for identifying the current welding position type include: calculating the angle between the welding torch axis and the direction of gravity based on the real-time trajectory coordinates and attitude information of the robot end effector; comparing the angle with a preset welding position angle threshold to determine the type of welding position, which includes flat welding position, vertical welding position, overhead welding position and horizontal welding position.

5. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The position correction coefficient The rule for determining the value is: when it is identified as a flat weld position, The value range is 1.0-1.1; when identified as a vertical welding position, The value range is 0.8-0.9; when identified as a horizontal weld position, The value range is 0.9-1.0; when identified as an overhead welding position, The value range is 0.7-0.

8.

6. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The bevel filling correction coefficient The rule for determining the value is: when the bevel filling coefficient When it is determined to be a wide and shallow bevel The value range is 1.1-1.3; when the bevel filling coefficient When it is determined to be a standard bevel The value is 1.0; when the bevel filling coefficient When it is determined to be a deep and narrow bevel, The value range is 0.7-0.

9.

7. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for calculating the correction amount of the center position of the swing trajectory include: establishing a nonlinear mapping relationship between the center position of the welding torch swing trajectory and the deviation amount, using the deviation amount of the weld centerline as input; calculating the correction amount of the center position of the welding torch swing trajectory in the lateral direction based on the direction and amount of the weld centerline deviation; filtering and limiting the calculated correction amount; and sending the corrected swing trajectory center position parameters to the robot controller.

8. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for adjusting the robot welding process parameters include: calculating the motion compensation amount at the end of the welding torch in real time based on the correction amount of the target swing width and the center position of the swing trajectory; correcting the welding trajectory command in the robot controller; and dynamically adjusting the lateral swing amplitude, swing frequency, and swing center position of the welding torch.

9. The method for dynamically adjusting the swing width of an all-position MAG welding robot according to claim 1, characterized in that, The steps for dynamically correcting the gain scheduling algorithm parameters include: continuously monitoring the weld formation quality through a visual sensor or an arc sensor, extracting welding quality features, including weld width, reinforcement height, and fusion state; establishing the correlation between weld formation quality and the oscillation width control effect, and triggering algorithm parameter correction when the quality deviation exceeds a preset threshold; adjusting the control gain coefficient in the gain scheduling algorithm using the gradient descent method according to the direction and magnitude of the deviation, and updating the corrected parameters in real time.

10. A dynamic adjustment system for the swing width of an all-position MAG welding robot, characterized in that, include: The data acquisition module is used to acquire welding electrical signals through an arc sensor, obtain bevel shape images through a vision sensor, and determine the current welding spatial position based on the robot trajectory coordinates. The feature extraction module is used to process the bevel shape image and extract the geometric features of the bevel, including the gap width. Bevel depth and bevel angle And calculate the bevel filling coefficient. The welding electrical signal is filtered to extract the electrical features of the left and right sides during the swing process, and the deviation of the weld centerline is calculated based on the difference in electrical features, including the deviation direction and the deviation amount; the welding position type identification module is used to calculate the angle of the welding torch relative to the direction of gravity through coordinate transformation based on the trajectory coordinates of the robot end effector, and to identify the current welding position type. The swing trajectory correction module is used for multi-factor coupling models based on swing width. The robot employs a gain scheduling algorithm to calculate the target swing width under the current working condition, and simultaneously calculates the correction amount for the center position of the swing trajectory based on the deviation of the weld centerline. The gain scheduling algorithm uses piecewise linear interpolation, setting different gain matrices for different welding positions. A welding process parameter adjustment module is used to collaboratively adjust the robot's welding process parameters based on the target swing width and the trajectory center position correction amount, and sends the adjustment instructions to the robot controller for execution. A closed-loop control module is used to control the robot at preset time intervals. Collect weld deviation ,when According to Fine-tuning of the swing width is performed, among which The value range is 0.5-2.0; the welding electrical signal changes and weld formation quality are continuously monitored, and the parameters of the gain scheduling algorithm are dynamically corrected based on the feedback information to achieve closed-loop control.

Images