CN116252289A

CN116252289A - Robot self-adaptive teaching method for thin-wall edge machining

Info

Publication number: CN116252289A
Application number: CN202310234857.9A
Authority: CN
Inventors: 高永卓; 李明洋; 董为
Original assignee: Harbin Institute of Technology
Current assignee: Harbin Institute of Technology
Priority date: 2023-03-13
Filing date: 2023-03-13
Publication date: 2023-06-13

Abstract

The invention provides a self-adaptive teaching method of a robot for thin-wall edge processing, and belongs to the technical field of industrial robots. The method aims to solve the problems that the existing self-adaptive teaching method of the robot mostly adopts an online compensation method, and when the thin-wall edge of the workpiece with weak rigidity and large rigidity change is processed, the quick multi-degree-of-freedom compensation and the damage to the surface of the workpiece are difficult to realize. According to the invention, a plurality of path points are selected from the visual extraction path, and the multi-degree-of-freedom initial attachment of the processing tool to the workpiece is realized by controlling the contact force aiming at each path point, so that the original attachment pose of the workpiece and the attachment pose of the deformed workpiece are obtained; correcting the measurement deviation of the visual extraction path according to the original pose of the workpiece, and correcting the processing deformation error of the path according to the attaching pose after the workpiece is deformed, so as to finally obtain a corrected processing path. According to the invention, the robot compensation quantity is obtained in an off-line state, the robot deviation is comprehensively corrected, and the machining precision is greatly improved.

Description

Robot self-adaptive teaching method for thin-wall edge machining

Technical Field

The invention relates to the technical field of industrial robots, in particular to a robot self-adaptive teaching method for thin-wall edge processing.

Background

In the fields of aerospace, national defense industry and the like, due to special requirements on weight reduction and the like of products, a large number of light thin-wall structures are adopted for design so as to be used as component parts of various integral workpieces. For the general definition of thin-walled workpieces, the workpiece is considered to have a wall thickness to length ratio of less than 1/5 to 1/8. Thin-walled workpieces meeting such definitions are quite widely used, for example, storage box workpieces having diameters of about 3.5 meters, and generally have wall thicknesses within 2cm, and generally, the entire workpiece is formed by welding a plurality of thin-walled parts. Before and after the welding of the parts, the parts are required to be subjected to treatment processing technologies such as chamfering of the edges before welding and removing of the welding seams after welding. And because the process has high labor intensity and high pollution, the robot automatic processing is of great significance.

After the robot completes visual on-line measurement and analysis of the thin-wall workpiece, the processing track of the robot can be calculated according to the edge point cloud of the workpiece. However, due to certain errors in the robotic measurement system, the weak stiffness and variable stiffness characteristics of the thin-walled workpiece can cause relatively large and uneven deformation of the workpiece when subjected to processing forces. Therefore, there is a large deviation of the ideal processing trajectory from the visual extraction path. If the visual track is directly adopted, on one hand, inaccurate processing positions can make technological parameters such as contact force, processing angle and the like difficult to ensure in the processing process, and further, the processing effect is difficult to realize. On the other hand, using only visual trajectories will likely result in the workpiece colliding with the robot, risking. However, if the on-line compensation method based on the visual track is adopted, the above problems can be avoided in principle. On the one hand, however, for machining requiring simultaneous compensation of multiple degrees of freedom, it is generally difficult to achieve rapid response by using a robot, and the surface of a workpiece is damaged by tool rotation in the compensation process, so that the machining quality is reduced. On the other hand, chatter of the thin-walled workpiece with weak rigidity during processing is difficult to avoid, which makes fluctuation of online force feedback or position feedback of the workpiece very large, and it is difficult to realize an ideal online control effect. Therefore, in actual production, the technical difficulty and cost are considered, and the manual operation mode is still the main mode.

Disclosure of Invention

The invention aims to solve the technical problems that:

the existing self-adaptive teaching method of the robot mostly adopts an online compensation method, and has the problems that the robot is difficult to realize quick multi-degree-of-freedom compensation and the surface of the workpiece is damaged due to the compensation rotation when the thin-wall edge of the workpiece with weak rigidity and large rigidity change is processed.

The invention adopts the technical scheme for solving the technical problems:

the invention provides a robot self-adaptive teaching method for thin-wall edge processing, which comprises the following steps:

s1, acquiring a visual extraction path of a processing tool, selecting a plurality of path points in the visual extraction path, wherein the path points are selected relatively uniformly and cover the whole processing path as far as possible, and obtaining a discrete visual extraction path { T } _i }；

S2, aiming at each path point, the multi-degree-of-freedom initial attachment of the processing tool to the workpiece is realized by controlling the contact force, so that the original attachment pose of the workpiece and the attachment pose Tt of the workpiece after deformation are obtained;

s3, correcting vision measurement deviation of the vision extraction path according to the original pose of the workpiece to obtain an original path of the workpiece, and correcting processing deformation errors of the original path of the workpiece according to the attached pose of the deformed workpiece to obtain a corrected processing path.

Further, in S2, the contact force is controlled by a damping control method.

Further, S3 includes the following steps:

s31, acquiring a vision measurement deviation change pose T by an ICP method according to the obtained original attaching pose of the workpiece ₀₁ ；

S32, correcting the visual extraction path according to the obtained visual measurement deviation change pose to obtain a workpiece original path { T f } _i }；

S33, according to the attaching pose Tt of the deformed workpiece and the original path { T f of the workpiece _i Calculation of deformation T d of each Path Point _i Amount of deformation Td _i Comprising a translational component and a rotational component;

s34, calculating a translation component and a rotation component by adopting a polynomial fitting method, and carrying out { T f on the original path of the workpiece _i All path points of the correctionA corrected machining path is obtained.

Further, the specific calculation process of S3 is:

acquiring vision measurement deviation change pose T by ICP method ₀₁ ：

Wherein pv _j Is the position of the waypoint, pt _j The position component of the pose is attached to the path point originally, R is a rotation pose matrix, and t is a translation pose matrix;

according to the change position T ₀₁ Correcting the vision extraction path to obtain a workpiece original path { T f } _i }：

T f _i ＝T _i T ₀₁ (2)

For k _i Path point, calculate its deformation

The method comprises the following steps:

in Tt _ki Is k _i Fitting pose after point deformation, calculating result Td _ki The translational and rotational components of (a), i.e. the displacement deformation d _ki And a rotational deformation a _ki ；

All path points { T f of original path _i The method for correcting is as follows:

wherein D is _yz And R is _yi Is based on displacement deformation d _i And a rotational deformation a _i Calculated translation vector and rotation matrix, te _ki And calculating a result for correcting the pose of the path.

Compared with the prior art, the invention has the beneficial effects that:

the invention discloses a self-adaptive teaching method of a robot for thin-wall edge processing, which adopts an offline state of unopened processing of the robot, selects a plurality of path points based on visual extraction paths, and performs multi-degree-of-freedom intelligent laminating compensation on each path point; according to the invention, the real-time stress information of the tool sensor is used as feedback control contact force to realize autonomous compensation of the robot, and autonomous teaching of the robot is realized; finally, calculating a final processing track according to the fusion of the compensation quantity and the initial track in the self-adaptive teaching process so as to be directly used for robot processing; the control difficulty brought by the thin-wall characteristic of the workpiece by the online real-time compensation method in the processing process is avoided.

The invention comprehensively considers the robot vision measurement deviation and the stress deformation deviation of the thin-wall workpiece, corrects the processing path of the edge of the thin-wall workpiece, and greatly improves the processing precision.

Drawings

FIG. 1 is a flow chart of a robot adaptive teaching method for thin-wall edge processing in an embodiment of the invention;

FIG. 2 illustrates a doctor blade chamfering tool according to an embodiment of the invention;

FIG. 3 is a diagram illustrating path extraction and intelligent fit compensation point selection according to an embodiment of the present invention;

FIG. 4 is a graph showing the force applied by a sensor during a Z-axis approach of a tool in accordance with an embodiment of the present invention;

FIG. 5 is a graph showing force applied to a sensor during a process of tool X-axis proximity in accordance with one embodiment of the present invention;

FIG. 6 is a graph showing experimental results of a tool Z-axis approach process in accordance with an embodiment of the present invention;

FIG. 7 is a graph showing force applied to a sensor during a Z-axis approach of a tool in accordance with an embodiment of the present invention;

FIG. 8 is a graph showing the result of visually extracted pose, original pose of a workpiece and pose of the workpiece after deformation and fitting in an embodiment of the invention; wherein, 1, 2, 3 marked are respectively in turn: and measuring the extracted pose, the original attaching pose of the workpiece and the attaching pose of the deformed workpiece.

FIG. 9 is a diagram showing the result of fusing the intelligent joint compensation result and the measured extraction path to the original path of the workpiece in the embodiment of the invention;

FIG. 10 is a graph showing the result of the deformation compensation amount fitting calculation according to the embodiment of the present invention;

fig. 11 is a diagram showing a comparison of the compensated robot processing deformation path and the initial measurement extraction path in the embodiment of the present invention.

Detailed Description

In the description of the present invention, it should be noted that the terms "first," "second," and "third" mentioned in the embodiments of the present invention are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first", "a second", or a third "may explicitly or implicitly include one or more such feature.

In order that the above objects, features and advantages of the invention will be readily understood, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.

As shown in fig. 1, the invention provides a self-adaptive teaching method of a robot for processing a thin-wall edge, which can be used for chamfering a thin-wall workpiece before welding, a robot chamfering tool and a coordinate direction are adopted, wherein the origin of the coordinate system is positioned at the midpoint of contact points of two contact wheels below and the workpiece, the coordinate axis is parallel to the tail end coordinate axis at a flange plate of the robot, a robot connecting plate above the tool is fixedly connected with the tail end flange of the robot through screw installation, the chamfering tool comprises a scraper assembly and a force feedback assembly, the scraper assembly comprises a cylinder and a scraper, the output force of the cylinder is realized through control air pressure, when the scraper contacts the workpiece, the processing force and the output force can be considered to be approximately equal, the scraper has a certain length and can float within a certain range to realize passive compliance, and the chamfering tool can be suitable for the range of the workpiece shown in fig. 2. The force feedback assembly comprises four one-dimensional force sensors and contact wheels connected with the force sensors, and the contact wheels can enable the relative movement of the tool and the workpiece to be smoother in the machining process.

The method comprises the following steps:

s1, acquiring a visual extraction path of a processing tool, and selecting from the visual extraction pathThe plurality of path points are selected relatively uniformly and cover the whole processing path as far as possible, and a discrete visual extraction path { T is obtained _i }；

The selection of the plurality of path points should be relatively uniform and cover the entire path as much as possible. The specific compensation method of each selected point is the same, the initial position of the selected point is calculated according to the measurement extraction path, and the reasonable processing pose of the point is obtained after the intelligent lamination of the robot; compared with a real-time dynamic compensation method of the robot when the whole track runs, the single-point compensation method can ensure that the robot is stable in an ideal processing pose.

and S2, controlling the contact force by adopting a damping control method.

S21, ensuring that the tool is not contacted with the workpiece in the initial state for ensuring safety. Translation is performed along the measurement extraction path pose of the selected point to the positive direction of the X axis and the negative direction of the Z axis (namely, the direction far away from the workpiece) of the tool coordinate system so as to calculate and obtain the initial pose, which can be expressed as:

wherein T1 is the initial pose, A is the pose of the measurement extraction path, and dx and dz are translation setting values.

S22, the robot transversely translates along the Z direction and performs rotation compensation motion around the Y direction. Damping control is adopted for the motion of two degrees of freedom, and the damping control can be expressed as formulas (2) and (3);

F ₁ +F ₂ -F _d ＝b ₁ v (2)

(F ₁ -F ₂ )l ₁₂ -T _d ＝b ₂ ω (3)

wherein F is ₁ And F ₂ Is the stress of the two sensors, F _d To expect the combination of force, T _d To be a desired moment, b ₁ And b ₂ V and ω are calculated robot speed and angular velocity for the set damping coefficient.

The end condition of the fitting movement is |F ₁ +F ₂ -F _d |<F _t And |F ₁ -F ₂ |<w _t Representing that the tool sensor is stressed within an ideal range.

After S22, the tool is bonded to the surface of the workpiece in the Z direction.

S23, carrying out X-direction translation compensation and Z-direction rotation compensation by the robot; the same damping control as in the formulas (2) and (3) is adopted, and the damping coefficients are b3 and b4.

At the end of S23, the tool achieves an initial fit with the workpiece in four degrees of freedom.

Because the workpiece is stiffer in the X-direction, small positional deviations can result in excessive contact forces, damaging the tool contact wheel, and should therefore be kept a distance in that direction to accommodate the deviations.

The S24 tool makes a certain amount of offset in the X direction.

After the initial lamination and adjustment are completed, the intelligent lamination result can be obtained. In order to facilitate the fusion calculation of the subsequent correction paths, the single-point intelligent fitting outputs two types of results, including the original fitting pose of the workpiece when the workpiece is hardly deformed and the fitting pose of the workpiece after deformation.

S25, acquiring the original pose of the workpiece: since the translation of S24 may have some effect on the fit of the tool in the Z direction, the fit action of S22 is repeated in this step to fine tune, and the desired force Fd is set close to 0 to obtain the pose where the workpiece is hardly deformed.

S26, acquiring the pose of the deformed workpiece: the expected force is set to be larger than the actual processing force, and the fitting operation of S26 is repeated to obtain the post-deformation pose T t.

Because the cylinder air pressure of the tool is certain during processing, the contact force of the tool and the workpiece can be controlled within a certain range, therefore, if the scraper is started to process under the fitting position, when the workpiece is subjected to small deflection caused by the contact force of the scraper, the stress of the contact wheel with high rigidity is greatly reduced, and the workpiece is in stress balance. The method can be understood as applying a certain pre-pressure to the workpiece so as to enable the deformation amount in the processing process to be more controllable and enable the fitting pose to be directly used for processing.

And because vision measurement deviation and machining deformation deviation exist between the robot vision extraction path and the final reasonable machining path. The visual measurement deviation can be approximately described as pose transformation between the real edge point cloud and the measured edge point cloud of the workpiece. For doctor blade machining, the displacement variation of the stressed deformation of the workpiece can be considered to be along the normal direction of the workpiece, and the gesture variation can be considered to be tangential around the workpiece, but the variation is unevenly distributed due to the rigidity variation of the workpiece.

S3 comprises the following steps:

s31, acquiring a vision measurement deviation change pose T through a method (ICP method) based on nearest point iteration according to the obtained original attaching pose of the workpiece ₀₁ ；

Wherein pv _j Is the position of the waypoint, pt _j The position component of the original pose;

T f _i ＝T _i T ₀₁ (5)

In the original path { T f of the workpiece _i In the case of the Y-axis, the transformation of the selected point may deviate from the corresponding original pose, typically by rotation about the Y-axis, due to the distance between the workpiece's joint and the edge, which is caused by the small curvature of the workpiece. But the partial deviation is usually small, and can be compensated and adjusted in a linear interpolation way on the whole path。

S33, fitting the workpiece according to the T position after deformation _t And workpiece original path { T f _i Calculation of deformation T d of each Path Point _i Calculate the deformation T d _i A translational component and a rotational component of (a);

for k _i The calculation formula of the deformation of the path point is as follows:

in Tt _ki Is k _i Post-deformation fitting pose of point, td _ki Comprising a translational component and a rotational component, i.e. displacement deformation d _ki And a rotational deformation a _ki ；

The deformation prediction of the whole path is carried out according to the joint pose after the workpiece with the selected point is deformed, and the deformation of the whole path comprises two parts, namely displacement deformation dz along the Z axis of a tool coordinate system _ki Rotational deformation ay about Y axis _ki 。

S34, calculating a translation component and a rotation component by adopting a polynomial fitting method, and carrying out { T f on the original path of the workpiece _i All path points are corrected to obtain a corrected processing path.

When the thin-wall workpiece is subjected to the same force in the state of partial clamping, the deformation of the edge of the thin-wall workpiece can be approximately considered to be polynomial change. Sequence number k for each selected point _i And its deformation dz _ki Calculating deformation dz of all path points by using polynomial fitting method _i . For the rotational deformation ay _ki Because the upper contact wheel and the lower contact wheel are attached to the workpiece when the workpiece attachment compensation is finished, the rotation angle value can be approximately considered to be consistent with the polynomial change by the fact that the distance between the two contact wheels is unchanged and the rotation angle is smaller, and the deformation ay is calculated by adopting a polynomial fitting method _i 。

S34 all path points { T f of original path _i The method for correcting is as follows:

wherein D is _yz And R is _yi Based on the deformation dz _i And ay _i Calculated translation vector and rotation matrix, te _ki And calculating a result for correcting the pose of the path.

Example 1

Chamfering is carried out on the edges of the thin-wall workpiece by adopting the method of the invention, so as to verify the method of the invention.

The experimental platform comprises a 6-axis industrial robot, a camera, a workpiece, a clamp and a chamfering tool; the robot is of the type of Epstein ER-50A, the camera is of the type of DS Ensenso3D, the type of N35-602-16-BL, the cylinder is of the type of MGPL20-50Z-M9BL, when the air pressure is 0.2-0.3MPa, the cylinder output is about 63-94N, the chamfering tool is of a HSS high-speed steel small triangular scraper, the type of the chamfering tool is up to 93461, the four force sensors are all of the Rayleigh force sensor T317-10kg, the measuring range is 100N, and the precision is 0.3%. The path correction algorithm runs on the grinding and warfare industrial personal computer UNO-3000G, the industrial personal computer reads force sensor data through the grinding and warfare PCI-1710L data acquisition card, and the industrial robot and the industrial personal computer receive movement instructions through Modbus communication. The thin-wall workpiece is a part of storage box melon petal workpiece, the position to be chamfered is the upper edge, the clamp is a vice, and the bottom of the workpiece can be clamped and clamped.

S1, as shown in FIG. 3, acquiring a visual extraction path of a workpiece, selecting 100 path points in the visual extraction path, and dispersing the visual extraction path into 100 equidistant poses; for clarity, the robot paths are shown one at 4 intervals, and the points on the two sides of the workpiece cannot be corrected by the contact wheel attachment in the tool, so 6 path points as shown in fig. 3b are selected as an example, and attachment compensation is sequentially performed on the 6 path points.

S2, aiming at each path point, the multi-degree-of-freedom initial attachment of the processing tool to the workpiece is realized by controlling the contact force, so as to obtain the original attachment pose of the workpiece and the attachment pose T t of the deformed workpiece;

s21, for each path point, the tool is not contacted with the workpiece, and the initial pose of the robot is obtained.

S22, as shown in FIG. 4, the measured stress value is smaller than a set threshold value 2N, the robot transversely translates along the Z direction, the processing tool is in contact with a workpiece, the stress of the sensor 2 is increased, the stress of the sensor 1 is almost unchanged, the tool and the workpiece have angle deviation, the workpiece rotates around the Y negative direction to compensate movement, the stress of the sensor 1 is increased, and meanwhile, the stress of the sensor 2 is reduced, and the stress of the sensor 2 and the stress of the workpiece are approximately equal to each other to achieve balance. Finally, the tool is attached to the surface of the workpiece in the Z direction. Damping control is used for both degrees of freedom motions according to equations (2) and (3).

On a specific parameter, the expected force is fd=10n, and the expected moment is 0, i.e. the two sensors are expected to be stressed equally. For Z-axis translational motion, damping parameters are set in a segmented mode, and b is set when the actual stress is smaller than the expected force ₁ =3.3, b when the force is greater than the desired force ₂ =100. For rotational movement about the Y-axis, the damping parameter is set to 0.26. In order to reduce the influence of measurement noise, the stress threshold of the sensor for opening translational motion is set to be 2N, the stress difference threshold of the sensor for opening rotational motion is 1.67N, and two compensation motions are started when the stress difference threshold is higher than the threshold. The robot compensation motion termination condition is that the difference between the sum of the stress of the two sensors and the expected force is smaller than 5N, and the difference between the stress of the two sensors is smaller than 2.1N. In the lamination compensation of this step, the robot moves about 35mm in the Z direction, rotates about 3.5 degrees in the negative Y direction, and finally is laminated with the workpiece.

S23, on the basis of completing lamination in the Z direction, the robot carries out X-direction translation compensation in the same mode, and damping control similar to formulas (2) and (3) is adopted, wherein the damping coefficients are b3 and b4. As shown in fig. 5, after the tool contacts the workpiece, the tool performs rotation compensation around the Z direction by the difference between the two sensors, and after the tool contacts the workpiece, the force of the sensors rises faster due to the greater rigidity of the workpiece in the direction, so that the force is greater than the resultant force threshold, and the tool is sprung back faster due to the greater difference between the forces of the two sensors, and the tool performs angular rotation simultaneously. After the process is performed for several times, the pose of the robot is adjusted to be attached to the workpiece, and the equilibrium position is reached.

In a specific parameter, the desired force is fd=30n and the desired torque is 0. Damping parameters for X-axis translational motionThe number adopts a sectional setting, and b is when the actual stress is smaller than the expected force ₃ =14.3, b when the force is greater than the desired force ₄ =100. For rotational motion about the Z axis, the damping parameter is set to 1.75. The threshold value of the stress opening of the sensor is the same as the setting of the S22, and the robot compensation motion termination condition is that the difference between the sum of the stress of the two sensors and the expected force is smaller than 10N, and the difference between the stress of the two sensors is smaller than 3.3N. In the lamination compensation of this step, the robot moves about 25mm in the X direction and rotates about 4 degrees in the Z direction, and at this time, the tool and the workpiece realize initial lamination in four degrees of freedom.

S24, due to the fact that the rigidity of the workpiece in the X direction is high, the contact force is too high due to small position deviation, and the tool contact wheel is damaged, so that a certain distance is kept in the direction to adapt to the deviation, and therefore the tool is translated upwards by 10mm along the X direction.

S25, because the translation of S24 possibly changes the fit of the tool in the Z direction, the Z direction fit is finely adjusted again, and the expected force F is set on parameters _d Because the angle change is smaller, the rotation damping parameter is properly increased and set to 0.79, the other parameters are the same as S22, as shown in FIG. 6, the sum of the stress of the two sensors after adjustment is 5N, the difference of the stress is within 2N, the robot almost does not change the angle, the tool moves 0.8mm along the negative Z direction, and the attaching compensation result is more approximate to the undeformed condition, so that the original pose of the workpiece almost undeformed at the position is obtained.

S26, setting the expected force to be larger than the actual processing force, and repeating the attaching action of S22 to obtain the deformed pose. As shown in fig. 7, the stress of the sensor 2 is increased first and the stress variation of the sensor 1 is smaller in the initial stage, because a certain angle exists between the workpiece and the tool after deformation, the stress of the sensor 2 is gradually reduced and the stress of the sensor 1 is gradually increased along with the adjustment of the robot angle, and finally, the stress of the sensor and the tool are both about 50N.

On a specific parameter, the desired force is fd=100deg.N. Since contact of the tool with the workpiece has been ensured, the damping is greater than the parameters in S22 to achieve a stable fit, b for the translational damping parameters when the actual force is less than the desired force ₁ When the stress is greater than the period =10B when looking at force ₂ =100. For rotational movement about the Y axis, the damping parameter is set to 0.79, the remaining parameters being the same as the S22 setting. The robot moved about 13mm in the Z direction in the same direction and rotated about 3.5 degrees in the Y direction during this process. The method shows that the edge position of the workpiece is greatly deformed after being stressed, and the ideal machining position and angle are greatly changed.

The visually extracted pose, the original pose of the workpiece and the pose of the workpiece after being deformed and attached are shown in table 1 and fig. 8:

TABLE 1

The relative transformation values of the pose results are calculated as shown in table 2:

TABLE 2

The results show that the deviation of the original pose of the workpiece relative to the pose of visual extraction is relatively large, wherein the error along the Z axis of the tool coordinate system is about 14mm at maximum, the error along the X axis is about 3mm at maximum, and the angle error is about 1 degree at maximum. In the deviation of the deformation pose of the workpiece relative to the original pose of the workpiece, the displacement deviation is about 18mm at maximum, and the angle deviation is about 4 degrees at maximum. Therefore, larger deviation can be generated without adopting a path intelligent correction method, the position deviation is too large, the edge of the workpiece exceeds the tool application range to cause danger, and the angle deviation is too large, so that the chamfering angle value does not meet the requirement.

S3, correcting vision measurement deviation of the vision extraction path according to the original pose of the workpiece to obtain the original path of the workpiece, and correcting processing deformation errors of the original path of the workpiece according to the attaching pose of the deformed workpiece to obtain a corrected processing path.

The vision measurement deviation obtained according to the formula (4) changes the pose T ₀₁ The method comprises the following steps:

correcting the visual measurement deviation according to formula (5); as shown in fig. 9, the result shows that the transformed measurement extraction path selection point substantially coincides with the fitting compensation result. The relative pose errors of the 6 transformed selected points and the corresponding intelligent bonding compensation points are shown in a table 3, and the positions in the table are relative to each intelligent bonding point coordinate system; the attitude differences are shown in Table 4, and the results were obtained using Euler angles with a rotation order of Z-Y-X.

TABLE 3 Table 3

As can be seen from the results of Table 3, the Z-axis error after transformation was reduced from about 10mm to about 1mm, and the X-axis error was reduced from a maximum of 3mm to within 0.3mm, but the Y-axis was offset by about 3 mm. Because the chamfering tool can only realize the lamination of the X and Z axes, the lamination can not be carried out on the Y axis along the edge direction of the workpiece, and smaller sliding errors along the Y axis occur between the corrected position and the true selected point. The analog "slip" condition compensates for this error.

TABLE 4 Table 4

From the results of Table 4, it can be seen that the r-angle and y-angle errors are small and the p-angle changes are relatively large. Compensating the angle error by adopting a linear interpolation method, and performing interpolation calculation by taking the serial number of the coordinate point as an independent variable; and finally obtaining the original path of the tool.

K in S33 _i The calculation formula of the deformation of the path point is as follows:

According to the results in table 2, the Z-axis displacement and the p-angle around the Y-axis of the tool coordinate system are subjected to deformation compensation, the serial numbers of pose points are used as independent variables, polynomial fitting is carried out to calculate the compensation, the fitting effect is good, the maximum deviation between the Z-axis displacement fitting result and the correction point is 0.49mm, the maximum deviation between the Y-axis rotation angle is 0.31 degrees, and the deviation can meet the processing requirement.

The pair of the compensated robot processing deformation path and the initial measurement extraction path is shown in fig. 11, and the result shows that the two paths have obvious differences in position and posture, which indicates that the multipoint intelligent fitting compensation and path fusion method provided herein performs larger adjustment on the initial measurement extraction path.

Although the present disclosure is disclosed above, the scope of the present disclosure is not limited thereto. Various changes and modifications may be made by one skilled in the art without departing from the spirit and scope of the disclosure, and such changes and modifications would be within the scope of the disclosure.

Claims

1. A robot self-adaptive teaching method for thin-wall edge processing is characterized by comprising the following steps:

2. The adaptive teaching method for a thin-wall edge processing robot according to claim 1, wherein the contact force is controlled by a damping control method in S2.

3. The robot adaptive teaching method for thin-wall edge processing according to claim 2, wherein S3 comprises the steps of:

S33, according to the attaching pose Tt of the deformed workpiece and the original path { T f of the workpiece _i Calculating a deformation Td of each path point, the deformation Td including a translational component and a rotational component;

4. The adaptive teaching method for a thin-wall edge processing robot according to claim 3, wherein the specific calculation process of S3 is as follows:

acquiring vision measurement deviation change pose T by ICP method ₀₁ ：

Wherein, the liquid crystal display device comprises a liquid crystal display device,pv _j is the position of the waypoint, pt _j The position component of the pose is attached to the path point originally, R is a rotation pose matrix, and t is a translation pose matrix;

Tf _i ＝T _i T ₀₁ (2)

For the ki path point, calculating the deformation amount

The method comprises the following steps: