Symmetrical welding three-dimensional complex welding seam tracking method based on laser vision sensing

Through laser visual sensing and non-uniform rational B-spline fitting methods, the problem of insufficient tracking of MAG welding equipment in three-dimensional complex weld scenarios is solved, efficient welding automation and accuracy improvement is achieved, and production costs are reduced.

CN118926653APending Publication Date: 2024-11-12UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410936670.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing MAG welding equipment is insufficiently used in three-dimensional complex weld scenarios, mainly focusing on two-dimensional plane welding, and cannot effectively track and process three-dimensional complex welds.

Method used

Using a laser visual sensing method, three-dimensional welded groove shape data is obtained through segmented scanning, filtering and feature point extraction are performed, and the weld path of the robot is planned using a non-uniform rational B-spline fitting method.

Benefits of technology

It significantly improves automation and accuracy during welding, reduces production defects, improves production efficiency, and reduces dependence on high-skilled welding operators and production costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118926653A_ABST
    Figure CN118926653A_ABST
Patent Text Reader

Abstract

The invention discloses a symmetric welding three-dimensional complex welding seam tracking method based on laser vision sensing, and belongs to the technical field of robot welding, and the method comprises the steps that a laser sensor is adopted for conducting segmented scanning along a welding seam, and three-dimensional weldment groove shape data are obtained; filtering processing is conducted on the three-dimensional weldment groove shape data, and weld joint feature points of the three-dimensional weldment groove shape data are extracted; and according to the welding seam feature points, a non-uniform rational B spline fitting method is adopted to plan a welding seam path of the robot. Three-dimensional complex welding seam tracking is achieved through the methods of segmented scanning, combined filtering treatment, feature point extraction and welding path planning, the automation and precision in the welding process can be remarkably improved, production defects can be reduced by tracking and adjusting the welding seam path in real time, the production efficiency is improved, and the production cost is reduced. And the dependence on high-skill welding operators is reduced, and the production cost is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of robot welding, and in particular relates to a method for tracking a symmetrical three-dimensional complex weld based on laser vision sensing. Background Art

[0002] In modern manufacturing, welding is one of the most common manufacturing processes, especially in automotive, aerospace and heavy industrial production. Traditional welding technology relies on highly skilled operators to perform complex welding tasks, which often involve complex three-dimensional welds whose shapes and positions may vary due to irregularities of the workpiece. In order to improve welding quality and production efficiency, robotic welding systems are widely adopted.

[0003] MAG welding is a symmetrical welding process that achieves metal fusion by heating the continuous consumable filler metal electrode and the workpiece through the welding arc. In the traditional robotic MAG welding process, the commonly used working mode is "teach-playback" or offline programming. These traditional practices are prone to errors and cumbersome to operate. Therefore, more and more research has begun to focus on the advancement of weld detection and tracking technology.

[0004] In order to achieve the required welding accuracy, a seam tracking algorithm was developed to enable the welding robot to plan its welding path and weld along the actual weld seam. Automatic seam tracking ensures that the welding robot's welding torch can follow the weld seam throughout the robotic welding process. Currently, most robotic MAG welding equipment uses a teaching method to achieve seam tracking before welding. This teaching method has attracted the attention of many researchers in the past few decades and has also achieved fruitful results in this field. Despite many improvements in seam tracking performance, there are still several limitations: current research mainly focuses on two-dimensional (2D) planar welding. Some spatial welding only targets relatively regular curves, such as spirals and continuous lines, while insufficient attention is paid to three-dimensional (3D) complex welds, although these welds are widely used in many scenarios such as shipbuilding and automotive stampings. Summary of the invention

[0005] The purpose of the present invention is to provide a symmetrical welding three-dimensional complex weld tracking method based on laser vision sensing to address the above-mentioned deficiencies in the prior art, so as to solve the problem that the existing MAG welding equipment is lacking in use in three-dimensional complex weld scenes.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for tracking symmetrical three-dimensional complex welds based on laser vision sensing comprises the following steps:

[0008] S1. Use a laser sensor to scan along the weld in sections to obtain three-dimensional weldment groove shape data;

[0009] S2, filtering the three-dimensional weldment slot shape data to extract weld feature points of the three-dimensional weldment slot shape data;

[0010] S3. According to the weld feature points, the robot's weld path is planned using a non-uniform rational B-spline fitting method.

[0011] Furthermore, step S2 includes the following sub-steps:

[0012] S2.1. Smoothing the weld bevel data in the three-dimensional weldment groove shape data by using local weighted regression smoothing filtering, and deriving the first-order derivative of the weld bevel profile;

[0013] S2.2, use extreme filtering and Gaussian filtering to smooth the first-order derivative of the weld bevel profile;

[0014] S2.3, derive the second derivative of the weld bevel profile;

[0015] S2.4, use extreme filtering and Gaussian filtering to smooth the second-order derivative of the weld bevel profile;

[0016] S2.5. Based on the second-order derivative after smoothing, calculate the global maximum and minimum values ​​to obtain the coordinates of the weld feature points in the sensor coordinate system {S}.

[0017] Furthermore, in step S2.1, local weighted regression smoothing filtering is used to smooth the weld bevel data in the three-dimensional weldment groove shape data, including:

[0018] f ∧ (x) = Ave(y i |x i ∈N k (x i ))

[0019]

[0020] f ∧ (x0)=∑ i ω i (x0)y i

[0021] Among them, f ∧ (x) is the average value of the y coordinates of all points near the given point x, that is, the smoothed value of the weld bevel data; f ∧ (x0) is the average value of the y-axis coordinates of all points on the slot scanned by the laser sensor; i is the y-axis coordinate of the point on the slot scanned by the laser sensor;i N is the x-axis coordinate of the point on the slot scanned by the laser sensor; k (x i ) is the set of all points on a certain slot segment; x0 is the horizontal coordinate of the initial scanning point; N is the number of scanning points; ω i (x0) is the weight; K λ (x0,x i )=D(|x0-x i | / λ),D(t)=0.75(1-t 2 ), t is a parameter, and λ is the window width.

[0022] Furthermore, the filtering process is limited to:

[0023]

[0024] Among them, y, y n and n-1 is the sampling value, and ΔT is the threshold value.

[0025] Further, the Gaussian filtering process is:

[0026]

[0027] Where f(x) is a one-dimensional Gaussian function; μ is the mean value and σ is the standard deviation.

[0028] Further, the method comprises the following sub-steps:

[0029] S3.1. Calibrate the welding robot TCP, combine the compatibility conditions, control statics and dynamic equations to obtain the transformation matrix of the welding robot, and use the six-point method to obtain the transformation matrix from the welding torch coordinate system {T} to the end flange coordinate system {E}

[0030] S3.2. Use the three-point method to calibrate the laser sensor and obtain the transformation matrix between the sensor coordinate system {S} and the end flange coordinate system {E}

[0031] S3.3. Select three fixed points in the base coordinate system {B} as calibration reference points. The laser sensor collects the coordinate data of the calibration points, records the posture data of the end flange coordinate system {E}, and obtains the transformation matrix from the base coordinate system {B} to the end flange coordinate system

[0032] S3.4, according to the transformation matrix from the torch coordinate system {T} to the end flange coordinate system {E} Transformation matrix between sensor coordinate system {S} and end flange coordinate system {E} Transformation matrix from base coordinate system {B} to end flange coordinate system {E} Convert the two-dimensional coordinate experimental data in the sensor coordinate system {S} into three-dimensional coordinates in the base coordinate system {B};

[0033] S3.5. Based on the three-dimensional coordinates of the base coordinate system {B}, the non-uniform rational B-spline fitting method is used to plan the robot's welding path.

[0034] Furthermore, the transformation matrix and the transformation matrix They are:

[0035]

[0036] Furthermore, the coordinate transformation relationship of the calibration point between the base coordinate system {B} and the sensor coordinate system {S} is:

[0037]

[0038] in, B P is the coordinate of the point in the base coordinate system {B}, expressed as B P=(x B ,y B .z B ,1); S P is the coordinate in the sensor coordinate system {S}, expressed as S P=(x S ,0,z S ,1).

[0039] Furthermore, the non-uniform rational B-spline fitting method is used to plan the robot's welding path, and the curve fitting process is as follows:

[0040]

[0041] Where, C(u) is; ω i is the weight factor, i = 0, 1, ···, n, respectively corresponding to the control point d i , i=0, 1, ···, n is related, n is the number of control points; M i,p (u) is the node vector V = {v0,v0,…,v n+p+1}Determine the pth th B-spline basis functions.

[0042] The symmetrical welding three-dimensional complex weld tracking method based on laser vision sensing provided by the present invention has the following beneficial effects:

[0043] The present invention realizes three-dimensional complex weld tracking by using segmented scanning, combined filtering processing, feature point extraction and welding path planning methods, which can significantly improve the automation and accuracy in the welding process. By real-time tracking and adjusting the weld path, production defects can be reduced, production efficiency can be improved, and dependence on highly skilled welding operators can be reduced, thereby reducing production costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 The present invention is a flow chart of the method for tracking symmetrical three-dimensional complex welds based on laser vision sensing.

[0045] Figure 2 The invention discloses the principle and effect of segmented scanning of a two-dimensional symmetrical S-shaped weldment.

[0046] Figure 3 This is the filtering effect of different combinations of the present invention.

[0047] Figure 4 This is the feature point extraction algorithm of the present invention.

[0048] Figure 5 These are the coordinate systems for robot welding of the present invention. DETAILED DESCRIPTION

[0049] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0050] Example 1

[0051] This embodiment provides a symmetrical welding three-dimensional complex weld tracking method based on laser vision sensing, which can improve the automation level, accuracy and efficiency of the welding process. Figure 1 , which specifically includes the following steps

[0052] Step S1, use a laser sensor to perform segmented scanning along the weld to obtain three-dimensional weld groove shape data; to collect original weld bevel data, the segmentation method of this embodiment helps to capture the 3D shape of the weld more comprehensively, especially when the weld is long or has a complex shape, the three-dimensional weld groove shape data can be obtained completely and accurately.

[0053] The span of the three-dimensional weld in the vertical direction is large. During the scanning process, when the distance to the laser sensor is relatively close, part of the weld may exceed the measurement range and information cannot be collected, resulting in the loss of information on part of the top weld groove. It is usually difficult to obtain the overall shape data of the spatial weld groove. Therefore, this embodiment adopts a segmented scanning method.

[0054] Specifically, refer to Figure 2 The experimental objects of this embodiment are two-dimensional symmetrical S-shaped welds and three-dimensional space welds. The three-dimensional welds are composed of linear welds and symmetrical S-shaped welds. Since the three-dimensional welds used in this embodiment are two-segment overlapping types, intensive acquisition is performed at the overlap (i.e., the junction of the third and fourth segments of the three-dimensional welds), and the acquisition frequency is increased to prevent data distortion at the overlap of the welds.

[0055] Step S2, filtering the three-dimensional weldment slot shape data to extract weld feature points of the three-dimensional weldment slot shape data;

[0056] The data acquired by the laser sensor is affected by natural light, reflection from the weld surface, and burrs, which cause the acquired data to be distorted. Therefore, this embodiment uses three methods of combined filtering: Lowess filtering, limit filtering, and Gaussian filtering to smooth and correct the data graph. For the corrected data graph, feature extraction is performed. Figure 4 , which specifically includes the following sub-steps:

[0057] Step S2.1, using local weighted regression smoothing filtering to smooth the weld bevel data in the three-dimensional weldment groove shape data, and deriving the first-order derivative of the weld bevel profile;

[0058] Step S2.2, using limit filtering and Gaussian filtering to smooth the first-order derivative of the weld bevel profile;

[0059] Step S2.3, deriving the second-order derivative of the weld bevel profile;

[0060] Step S2.4, using limit filtering and Gaussian filtering to smooth the second-order derivative of the weld bevel profile;

[0061] Step S2.5: Based on the second-order derivative after smoothing, calculate the global maximum and minimum values ​​to obtain the coordinates of the weld feature points in the sensor coordinate system {S}.

[0062] Among them, the local weighted regression smoothing filter in this embodiment, that is, Lowess filtering, is a weighted linear least squares method combined with a first-order polynomial model, which can better smooth the volatile data and is used to process distortions such as burrs on the weld surface; first, the weld bevel data is averaged and regressed, then weighted smoothing is performed, and finally the Lowess estimation equation is derived. The process is as follows:

[0063] f ∧ (x) = Ave(y i |x i ∈N k (x i ))

[0064]

[0065] f ∧ (x0)=∑ i ω i (x0)y i

[0066] Among them, f ∧ (x) is the average value of the y coordinates of all points near the given point x, which is used to represent the smoothed value of the weld bevel data; f ∧ (x0) is the average value of the y-axis coordinates of all points on the slot scanned by the laser sensor; i is the y-axis coordinate of the point on the slot scanned by the laser sensor; i N is the x-axis coordinate of the point on the slot scanned by the laser sensor; k (x i ) is the set of all points on a certain slot segment; x0 is the horizontal coordinate of the initial scanning point; N is the number of scanning points; ω i (x0) is the weight.

[0067] In addition, K λ (x0,x i )=D(|x0-x i | / λ),D(t)=0.75(1-t 2 ). Where t is just a parameter, and |t|<1, and λ is the window width.

[0068] The limiting filtering principle is used to deal with pulse interference caused by accidental factors, which is specifically:

[0069]

[0070] Among them, y n and n-1 is the sampling value, and ΔT is the threshold value.

[0071] Gaussian filtering can effectively suppress normally distributed interference signals and prevent local peaks from appearing in the detection results. Specifically:

[0072]

[0073] Here, μ is the mean, which determines the location of the Gaussian function, and σ is the standard deviation, which determines the amplitude of the distribution.

[0074] In the process of processing the three-dimensional weld bevel, this embodiment uses low-value filtering to smooth the original shape of the weld bevel. The first-order and second-order derivatives are processed by limiting filtering and Gaussian filtering to make the local maximum minimum more significant. The effects of different combined filtering are as follows: Figure 3 As shown, x, y and z represent the values ​​on the x-axis, y-axis and z-axis respectively.

[0075] refer to Figure 4 , Lowess filtering is first used to smooth the weld bevel, and then the first-order derivative of the weld bevel profile is derived; the first-order derivative is smoothed using limit filtering and Gaussian filtering, and the second-order derivative is smoothed; finally, the global minimum and maximum values ​​are calculated to obtain the coordinates required for the weld feature points.

[0076] Step S3, planning the robot's welding path using a non-uniform rational B-spline fitting method according to the welding feature points;

[0077] Since the weld data measured by the laser sensor is based on its own coordinate system, it is necessary to convert the feature points into the base coordinate system {B} of the welding robot through posture calibration.

[0078] The coordinate transformation of experimental data can be used for path planning, that is, the experimental data of two-dimensional coordinates in the sensor coordinate system {S} is transformed into three-dimensional coordinates in the base coordinate system {B}. Figure 5 , this conversion process includes the end flange coordinate system and the robot welding torch coordinate system. S , Y S , Z S They represent the coordinate axes of the sensor coordinate system {S}. B , Y B , Z B They represent the coordinate axes of the base coordinate system {B}. E , Y E , Z E They represent the coordinate axes of the end flange coordinate system {E}. T , Y T , Z T They respectively represent the coordinate axes of the robot welding torch coordinate system {T}.

[0079] The coordinate conversion of this embodiment specifically includes the following sub-steps:

[0080] Step S3.1, calibrate the welding robot TCP (tool center point) of the welding torch, combine the compatibility conditions and the control statics and dynamics equations to obtain the transformation matrix of the robot; use the six-point method to obtain the transformation matrix from the welding torch coordinate system {T} to the end flange coordinate system By selecting a suitable fixed corner point on the calibration block as the calibration point, the movement of the welding robot is controlled to ensure the obvious change of the welding gun posture to the greatest extent and conform to the spherical uniform distribution, thereby effectively improving the accuracy and stability of the calibration.

[0081] Step S3.2: Use the three-point method to calibrate the laser sensor and obtain the transformation matrix between the sensor coordinate system {S} and the end flange coordinate system {E}

[0082] Step S3.3, select three fixed points in space as calibration reference points, collect the coordinate data of the calibration points with the laser sensor, record the posture data of the end flange coordinate system {E}, and obtain the transformation matrix from the base coordinate system {B} to the end flange coordinate system

[0083] Step S3.4: Transformation matrix from torch coordinate system {T} to end flange coordinate system {E} Transformation matrix between sensor coordinate system {S} and end flange coordinate system {E} Transformation matrix from base coordinate system {B} to end flange coordinate system {E} Convert the two-dimensional coordinate experimental data in the sensor coordinate system {S} into three-dimensional coordinates in the base coordinate system {B};

[0084] Step S3.5: Based on the three-dimensional coordinates in the base coordinate system {B}, the welding path of the robot is planned using a non-uniform rational B-spline fitting method.

[0085] In order to improve the reliability and accuracy of the calibration results, the welding gun calibration experiment was repeated three times to obtain three sets of welding robot posture data, and then the results were averaged; the transformation matrix and The calculation result is:

[0086]

[0087] The coordinate transformation relationship of the calibration point between the base coordinate system {B} and the sensor coordinate system {S} is as follows:

[0088]

[0089] in, B P=(x B ,y B .z B ,1), S P=(x S ,0,z S ,1).

[0090] The non-uniform rational B-spline fitting method is used to plan the robot's welding path. The trivial NURBS fitting method is more convenient when fitting complex shapes. The curve fitting process is as follows:

[0091]

[0092] Where, C(u) is; ω i is the weight factor, i = 0, 1, ···, n, respectively corresponding to the control point d i , i=0, 1, ···, n is related, n is the number of control points; M i,p (u) is the node vector V = {v0,v0,…,v n+p+1}Determine the pth th B-spline basis functions.

[0093] Although the specific implementation of the invention is described in detail in conjunction with the drawings, it should not be understood as limiting the scope of protection of this patent. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still fall within the scope of protection of this patent.

Claims

1. A method for tracking symmetrical three-dimensional complex welds based on laser vision sensing, characterized in that: The following steps are involved: S1. Use a laser sensor to scan along the weld in sections to obtain three-dimensional weldment groove shape data; S2, filtering the three-dimensional weldment slot shape data to extract weld feature points of the three-dimensional weldment slot shape data; S3. According to the weld feature points, the robot's weld path is planned using a non-uniform rational B-spline fitting method.

2. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 1 is characterized in that: The step S2 comprises the following sub-steps: S2.

1. Smoothing the weld bevel data in the three-dimensional weldment groove shape data by using local weighted regression smoothing filtering, and deriving the first-order derivative of the weld bevel profile; S2.2, use extreme filtering and Gaussian filtering to smooth the first-order derivative of the weld bevel profile; S2.3, derive the second derivative of the weld bevel profile; S2.4, use extreme filtering and Gaussian filtering to smooth the second-order derivative of the weld bevel profile; S2.

5. Based on the second-order derivative after smoothing, calculate the global maximum and minimum values ​​to obtain the coordinates of the weld feature points in the sensor coordinate system {S}.

3. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 1 is characterized in that: In the step S2.1, the local weighted regression smoothing filter is used to smooth the weld bevel data in the three-dimensional weldment groove shape data, including: F ∧ (x)=Bird(y) i |x i ∈N k (x i )) Among them, f ∧ (x) is the average value of the y coordinates of all points near the given point x, that is, the smoothed value of the weld bevel data; f ∧ (x0) is the average value of the y-axis coordinates of all points on the slot scanned by the laser sensor; i is the y-axis coordinate of the point on the slot scanned by the laser sensor; i N is the x-axis coordinate of the point on the slot scanned by the laser sensor; k (x i ) is the set of all points on a certain slot segment; x0 is the horizontal coordinate of the initial scanning point; N is the number of scanning points; ω i (x0) is the weight; K λ (x0,x i )=D(|x0-x i | / λ),D(t)=0.75(1-t 2 ), t is a parameter, and λ is the window width.

4. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 2 is characterized in that: The limiting filter process is: Among them, y, y n and n-1 is the sampling value, and ΔT is the threshold value.

5. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 2 is characterized in that: Gaussian filtering is done as follows: Where f(x) is a one-dimensional Gaussian function; μ is the mean value and σ is the standard deviation.

6. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 2 is characterized in that: The following steps are included: S3.

1. Calibrate the welding robot TCP, combine the compatibility conditions, control statics and dynamic equations to obtain the transformation matrix of the welding robot, and use the six-point method to obtain the transformation matrix from the welding torch coordinate system {T} to the end flange coordinate system {E} S3.

2. Use the three-point method to calibrate the laser sensor and obtain the transformation matrix between the sensor coordinate system {S} and the end flange coordinate system {E} S3.

3. Select three fixed points in the base coordinate system {B} as calibration reference points. The laser sensor collects the coordinate data of the calibration points, records the posture data of the end flange coordinate system {E}, and obtains the transformation matrix from the base coordinate system {B} to the end flange coordinate system S3.4, according to the transformation matrix from the torch coordinate system {T} to the end flange coordinate system {E} Transformation matrix between sensor coordinate system {S} and end flange coordinate system {E} Transformation matrix from base coordinate system {B} to end flange coordinate system {E} Convert the two-dimensional coordinate experimental data in the sensor coordinate system {S} into three-dimensional coordinates in the base coordinate system {B}; S3.

5. Based on the three-dimensional coordinates of the base coordinate system {B}, the non-uniform rational B-spline fitting method is used to plan the robot's welding path.

7. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 6 is characterized in that: Transformation Matrix and the transformation matrix They are:

8. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 7 is characterized in that: The coordinate transformation relationship of the calibration point between the base coordinate system {B} and the sensor coordinate system {S} is: in B P is the coordinate of the point in the base coordinate system {B}, expressed as B P=(x B ,y B .z B ,1); S P is the coordinate in the sensor coordinate system {S}, expressed as S P=(x S ,0,z S ,1).

9. The method for tracking symmetrical three-dimensional complex welds based on laser vision sensing according to claim 7 is characterized in that: The non-uniform rational B-spline fitting method is used to plan the robot's welding path, and the curve fitting process is as follows: Where C(u) is; ω i is the weight factor, i = 0, 1, ···, n, respectively corresponding to the control point d i , i=0, 1, ···, n is related, n is the number of control points; M i,p (u) is the node vector V = {v0,v0,…,v n+p+1 }Determine the pth th B-spline basis functions.

Citation Information

Cited By

  • Groove path and welding speed planning method and system

    CN119347794A

  • Multi-sensor dynamic path planning method and system of teaching-free welding robot for large iron tower component

    CN120962657A