JPH0630802B2 - Welding start point detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention detects a welding start point by an image pickup means mounted on an industrial robot and matches a control point at the tip of a robot end effector with the welding start point. The present invention relates to a welding start point detection method for controlling each axis of a robot.

[Conventional technology]

An example of applying the image pickup means as a sensor of an industrial robot is old, and is widespread. Examples applied to welding work robots have become popular in recent years, but from the comprehensive viewpoint of seam finding, seam tracking, and adaptive function, the invention disclosed in Japanese Patent Laid-Open No. 61-279481 reaches a practical level. This is one of the technologies we have done. The above-mentioned application discloses technical matters in various fields. Among them, regarding seam finding, the electrode of the tip of the welding torch attached to the wrist part is determined by the image information of the imaging means attached to the robot wrist. The image of the wire and the image of the electrode wire projected on the lower plate side of the work to be welded by the irradiation of the light source mounted on the wrist are matched in the welding line image formed on the joint surface of the work to be welded. It is a unique method to control each axis of the robot, which is a two-dimensional detection sensor and enables three-dimensional trajectory control.

[Problems to be solved by the invention]

Generally, the control point of the welding robot is the tip of the electrode wire supplied from the tip of the welding torch attached to the wrist. Then, trajectory control and trajectory correction control are performed so that the tip of the electrode wire having a desired length follows a desired trajectory.

However, at the end of welding, the unevenness of the electrode wire length caused by the burning of the wire, the coasting of the wire, the deposition of wire droplets, etc., which depends on the welding conditions and welding power source characteristics,
It is inevitable. The variation of the electrode wire length causes the variation of the actual control point. The method of positioning the fluctuated actual control point at the welding start point is effective as such, but fluctuations in the distance between the feed tip and the base metal may adversely affect the quality of the welding result. Therefore, it is necessary to make the electrode wire length match the desired value.

However, the wire feeding device that feeds out the electrode wire does not have a position control loop, is an open loop for position control, and has wire-cable friction, slack, rattling, etc. in the wire feeding cable. Due to a factor, it is not always easy to match the electrode wire length to a desired value in a short time.

The present invention has been made under such a background.

That is, the present invention is based on the invention disclosed in the above-mentioned application, in particular, since the electrode wire length varies due to many factors at the end of welding, it is corrected, and even when the electrode wire length does not match the desired value. The object of the present invention is to provide a method for producing the same effect as the coincidence, and to improve and stabilize the positioning accuracy at the welding start point and to perform the welding start point detection operation quickly.

[Means for solving problems]

In order to achieve this object, the welding start point detecting method of the present invention comprises a welding torch, a light source, and an image pickup means mounted on the wrist of an industrial robot, and the wire tip portion of the welding torch and the light source are attached by the image pickup means. Each axis of the robot is controlled so that the shadow of the wire projected onto the workpiece to be welded and the welding line of the actual workpiece when irradiated with the laser beam are coincident with each other on the welding line on the screen imaged by the imaging means. On the occasion of
The deviation amount between the wire tip position on the screen and the desired value is detected, and wire inching is performed so that the wire tip position matches the desired value according to the deviation amount, and the wire protrusion length generated at the end of welding is detected. It is characterized in that the variation is corrected to prepare for the start of the next welding.

[Action]

FIG. 1 shows a schematic diagram of a robot to which the present invention is applied. In FIG. 1, a wire feeding device 1 installed on a robot arm 2 feeds a welding wire 4 via a welding torch 3 fixed to a robot wrist 6.

A shadow 10 of the wire 4 is formed on the lower plate 13 of the work to be welded by the illumination of the light source 5 attached to the camera drive shaft 14 at the tip of the robot wrist 6.

The joint 11 between the upper plate 12 and the lower plate 13 of the work to be welded is the welding line.

The image of the welding wire 4, the image of the shadow 10 and the image of the welding line 11 are optically transmitted to the camera box 9 by the lens system 7 and the fiberscope 8 mounted on the camera drive shaft 14 at the tip of the robot wrist 6. It is converted into an image signal by the image pickup means 15 (FIG. 2) in the camera box.

The camera drive axis 14 is an additional axis of the wrist axis of the robot,
It is a shaft that is treated in the same manner as other wrist shafts and that allows the camera angle of the welding torch 3 to be freely changed.

FIG. 2 shows a schematic block diagram of the system.

The image signal from the image pickup means 15 is subjected to image processing by the image processor 16 and the required information is transmitted to the robot controller 17.

The image processor 16 is a means for converting at least an image signal from an analog amount into a digital amount, a frame memory,
It has a microprocessor, program memory D / A conversion means, and the like.

The robot controller 17 is a controller of a teaching / playback type robot, and drives and controls each drive axis of the robot 18 including the camera drive axis 14 and issues a command to the welding machine 19.

The welding machine 19 drives a motor (not shown) of the wire feeding device 1 according to a command from the robot controller 17.

The image processor 16 outputs a wire inching command according to the result of image processing. Robot controller 17
Sends a desired command to the welding machine 19 based on this information.

Now, the relationship between the real space and the screen will be described with reference to FIG.

3 (a) shows the actual space, and FIG. 3 (b) shows the state of the screen captured by the imaging means 15. Upper plate 12 and lower plate 1 in actual space
3, with respect to the welding line 11, the moving direction of the wire 4 is 2
It is defined as 0 and vertical direction 21.

The definition is made by the operator when in the teach mode of the robot.

The state on the screen corresponding to FIG. 3 (a) is shown by adding a "dash (')" to FIG. 3 (b). The contour edge in FIG. 3 (a) having a circular shape is due to the laminated structure of the fiber bundles forming the fiber scope.

Welding line image 11 ', wire image 4'and shadow image of wire 10'
Shows clearly different gray levels with respect to the background such as the upper plate image 12 'and the lower plate image 13' of the work to be welded. Among the welding line image 11 ', the wire image 4', and the wire image 10 ', the positions of the welding line image 11' and the wire image 10 'on the screen may change due to variations in the work to be welded and variations in setting. The position of the wire image does not change geometrically as long as it is the welding torch posture according to the specific robot wrist posture and the specific camera posture, that is, as long as the relative angular relationship between the wire and the camera is maintained.

Therefore, as long as the wire length from the feeding chip 29 is constant, the wire tip 30 is fixed on the screen coordinates as the tip point 30 'of the wire image.

However, the actual wire length varies due to many fluctuation factors.

In the example shown in FIG. 4, the tip 30 'of the actual wire image 4'is shorter than the tip 31 having a desired wire length. In the present invention, the case where the trajectory control point is the actual tip 30 ′, the case where the desired wire tip 31 and the actual wire tip 30 ′ are inching, and the case where the actual wire tip 30 ′ is the desired wire tip are used. If there is some excess or deficiency with respect to 31 and if it is within the desired range, it includes three virtual point methods for controlling so that the desired wire tip 31 coincides with the point 35 of the welding line 11 '.

Incidentally, 37 is an actual wire central axis, and 38 is an actual central axis of the image 10 '.

The left-right direction 20 'shown in FIG. 3 (b) is equivalent to the direction connecting the actual wire tip 30' and the point 36 in FIG. 4 and the direction connecting the desired wire tip 31 and the point 35 in FIG. Similarly FIG. 3 (b)
The vertical direction 21 'shown in FIG.
It is equivalent to the direction connecting 0'and 33. These are predefined.

The actual wire tip 30 'when the control point, the wire tip 30' distance D _{1 ',} the wire tip 30' left-right direction defined from the wire tip on the screen between the actual shadow tip 33 A line is drawn and the distance L ₁ ′ on the screen between the intersection 36 with the welding line image 11 ′ is calculated.

If the desired wire tip 31 and the actual wire tip 30 'were coincident, then the actual shadow tip would be 34 instead of 33.

The vertical distance D ₁ is calculated as the distance between 31 and 34, and the horizontal distance L ₁ is calculated as the distance between 31 and 35.

In the case of the virtual point method, although the actual wire tip 30 ′ is separated from the desired wire tip 31, the vertical distance D ₁ and the horizontal distance with respect to the desired wire tip 31 are used. Calculate L ₁ .

The actual shadow when moved by D _{1 in the} vertical direction is 10 ″, and the extension line of the central axis 38 ′ passing through the tip 32 intersects at the point 31. Similarly, the point 35 or the point 36 is also moved in the vertical direction by D 1. Move ₁ (not shown).

〔Example〕

 Hereinafter, the present invention will be specifically described based on Examples.

FIG. 5 is a schematic flow of the actual seam finding operation, which will be described with reference to this figure.

Once in the pre-programmed seam finding mode, the image processor 16 performs wire recognition and wire inching processing (block 90). as a result,
Judgment Circular 91 is used to judge whether there is a wire abnormality,
If it is abnormal, a warning is given by a display (not shown) or the like.

On the other hand, when normal, the wire tip coordinates, the wire center line, the wire length, etc. are calculated. Details of calculating these will be described later.

In block 92, image processing for recognition of the wire shadow is performed to calculate the center line, tip coordinates, etc. of the image, and in block 93, image processing for welding line recognition is executed to calculate the direction line of the welding line.

In block 94, the eccentric position (ΔX, ΔY, ΔZ) of the robot Cartesian coordinates is calculated based on the calculated values in block 90 and block 92.
It is calculated by equation (1) or equation (1) '.

Equation (1) represents the virtual point method, and equation (1) ′ represents the actual wire tip method.

In the equation (1), D ₁ is the vertical distance at the desired wire tip on the screen, and L ₁ is the horizontal distance.
K ₁ and k ₁ are conversion coefficients from the distance on the screen to the distance on the actual robot Cartesian coordinates, and are calculated or defined in advance.

(A, b, c) ^T is a vertical unit vector on the robot Cartesian coordinates, and (e, f, g) ^T is a horizontal unit vector on the robot Cartesian coordinates. It is defined.

The equation (1) 'is a case where the actual wire tip on the screen is used as the control point, D ₁ ′ is the distance between the actual wire tip and the shadow, and L ₁ ′ is the welding from the actual wire tip. The horizontal distance to the line is shown.

The method to match the actual wire length to the desired wire length is
Equation (1) 'is used.

Which of the three control point schemes is selected is arbitrary and programmable.

The image processor 16 is involved in the blocks 90 to 93, but the robot controller 17 is in charge of the blocks 94 to 96. After calculating (ΔX, ΔY, ΔZ) in block 94, the robot controller 17 adds this to the current value in block 95 to calculate the end point of seam finding.

In block 96, an actual movement operation starting from the current value is executed, and the control point is reached at the end point.

Fig. 6 (a) shows the relationship between actual image processing and wire inching.
Shown in. FIG. 7A is a concrete example of the block 90 shown in FIG.

Perform a wire search operation within the window specified in block 100. Details of this block 100 are shown in Fig. 6 (b).
Shown in.

In FIG. 6B, first, in block 1000, the wire edge point coordinates (X _k , Y _k ) are initialized, and image processing is sequentially performed on the screen axes (i, j) shown in FIG. 3B. . Coordinate
(I _o , J _o ), coordinates (I _max , J _max ) indicate the specified window, and (i, J _o ).
j) is a region of I _o ≦ i ≦ I _max , J _o ≦ j ≦ J _max )
It is defined in advance.

In the circular 1002, P _ij indicates the gray level of the (i, j) coordinate, and P _o indicates the threshold value.

When P _ij ≧ P _o , it is regarded as a white level, and when P _ij <P _o , it is regarded as a black level.

Block 1001 ZONE is a status flag, ZO
NE = 0 indicates the background part, and ZONE = 1 indicates the wire part. In the background part, P _ij ≧ P _o , and in the wire part, P _ij <
_Po .

X _L indicates the i coordinate value of the left edge of the wire part, and (X _k , Y
_k ) means the center coordinate between the left edge coordinate value and the right edge coordinate value of the wire portion. k is the number of times the wire portion is detected. 1015 when the final value N (1012 blocks) is greater than the designated circuit n _{1 and} 1014 when it is smaller.
Go to the block. WIRE is a status flag and WI
RE = 1 indicates that the wire has been recognized. This completes block 100 and proceeds to circular 101 in FIG. 6 (a).
Circular 101, when WIRE = 0, Circular 10
Go to 6. Reference numeral 106 is a circular that tests whether the number of times wire inching has been executed has exceeded a specified number. If it has, the wire error flag is set and the process returns. If the specified number of times has not been exceeded, block
At 109, the wire inching command W and the command time T are calculated. In the experimental value table, for example, (W _m , T _m , Δl _m ), (W _m ,
It is represented by T _p , Δl _p ).

That is, the command value W _m, the extension of the wire amount when the inching time T _m becomes .DELTA.l _m, command value W _m, the elongation of the wire amount when the inching time T _p is a table such that .DELTA.l _p. Now, we extracted the _{Δl m ≦ k 3 Δl ≦ Δl} p relative deviation .DELTA.l on the screen _{(W m, T m, Δl} m), (W m, T p, Δl p) a, Is calculated as (W, T). Note that k ₃ is a conversion coefficient between the wire length on the screen and the actual wire length.

Thus, the process proceeds to block 110, and W of the second equation is output for T time. When complete, proceed to block 100.

On the other hand, if WIRE = 1 in the circular 101, the process proceeds to block 102 to calculate the wire length l. Block 102
The details are shown in Fig. 6 (c). In block 1020 of FIG. 6 (c), (X _K , Y _K ), k = 0 to 0 calculated in block 100.
Least square approximation is performed on N to obtain the central axis of the wire. Based on this central axis, the block 1021 obtains the wire tip coordinates (x ₁ , y ₁ ) and the block 1022 obtains the wire feed-out side coordinates (x ₂ , y ₂ ).

The reason why the least squares approximation is performed in block 1020 is to prepare for the case where the droplet is bent and the droplet portion is attached to the tip portion of the wire.

In block 1023, the wire length 1 is calculated, the process returns and the process proceeds to block 103. In block 103, the wire length l and the desired wire length l _o previously taught
Calculate the difference Δl between and, and proceed to circular 104. When the method for matching the actual wire length with the desired wire length is executed, the process goes to the circular 108. Circular 10
In the case of 8, the allowable amount l for which Δl can be regarded as being in agreement
Compare with _min and decide whether to inching or consider as completed.

On the other hand, if it is not the matching method, the circular 105
At, a test is made whether Δl is in the desired range l _zone to determine whether to wire inch or complete wire inching.

The detailed description of blocks 92 and 93 is basically omitted because it can be explained by using all or part of the flowcharts of FIGS. 6 (b) and 6 (c).

〔The invention's effect〕

As described above, in the present invention, the variation of the wire length is measured based on the captured image, and the wire length is managed by performing the wire inching based on the result. As a result, even if the wire length varies due to many factors at the end of welding, it is corrected to improve the positioning accuracy at the welding start point and to
The start can be stabilized and the welding start point detection operation can be performed quickly.

[Brief description of drawings]

FIG. 1 is a schematic diagram of a robot to which the present invention is applied, FIG. 2 is a schematic block diagram of the system of the present invention, FIG. 3 is an explanatory diagram showing the relationship between real space and a screen, and FIG. 4 is FIG. FIG. 5 is a detailed view of (b), FIG. 5 is a flowchart of the seam finding operation, and FIG. 6 is a flowchart of actual image processing and wire inching.

Claims (4)

[Claims] 1. A welding torch, a light source and an image pickup means are attached to a wrist portion of an industrial robot, and the tip of the wire of the welding torch and the work to be welded of the wire when irradiated with the light source by the image pickup means. When controlling each axis of the robot such that the shadow projected on the screen and the welding line of the actual work match on the welding line on the screen imaged by the imaging means, the wire tip position on the screen and the desired value Deviation is detected, wire inching is performed according to the deviation so that the wire tip position matches the desired value, and the variation in wire protrusion length that occurs at the end of welding is corrected to start the next welding. And a welding start point detecting method. 2. The welding start point detecting method according to claim 1, wherein when the wire does not appear on the screen despite the execution of the wire inching, it is determined that the wire is abnormal or broken. . 3. If the tip of the wire is set in a desired range, it is possible to control each axis of the robot so that the tip of the wire and the tip of the shadow of the wire coincide with each other on a welding line in the screen. The method for detecting a welding start point according to claim 1, which is characterized in that. 4. If the wire tip portion is set in a desired range, the desired wire tip position and the shadowed tip portion of the desired wire will be the in-screen welding line regardless of the position of the wire tip portion. The welding start point detecting method according to claim 1, wherein each axis of the robot is controlled so that the actual wire tip and the actual shadow tip move so as to coincide with each other.