JP7385410B2 - arc welding system - Google Patents

Description

The present invention relates to an arc welding system for performing arc welding such as TIG welding.

BACKGROUND ART Conventionally, there are systems that automatically perform arc welding using a robot or the like (see, for example, Patent Document 1).

JP 2017-30014 Publication

In general, improving the quality of arc welding largely depends on the knowledge of welding techniques of the workers who set and adjust welding conditions, and training of operators skilled in arc welding is also important for arc welding systems. be.

The present invention has been made to solve the above problems, and an object of the present invention is to provide an arc welding system that can train operators skilled in arc welding.

In order to achieve the above object, an arc welding system according to an aspect of the present invention is a welding device capable of performing weaving welding by moving a welding torch in the welding direction while swinging it in a direction crossing the welding direction. an operating device for an operator to operate the welding device; and a controller for controlling the welding device to execute the weaving welding based on a predetermined welding program, and for controlling the welding device to execute the weaving welding by the operator's operation during the weaving welding. When a change operation command consisting of a command for changing the swing center position of the welding torch or the swing width of the welding torch is input from the welding device, the welding process is performed so that the change according to the change operation command is reflected. a controller that controls the device to continue the weaving welding; a camera that takes a video of the welding area including the molten pool during the weaving welding; a display device that displays images taken by the camera in real time; a memory device that stores time-series data of photographed images taken by a camera, an input time of the change operation command input to the controller, and change contents according to the command; Learning is performed using the time-series data of the photographed images, the input time of the change operation command, and the change contents according to the command, the input is the time-series data of the photographed images, and the output is the predicted change contents. The apparatus includes a machine learning device that generates a predictive model and outputs changes predicted from time-series data of captured images captured by the camera using the predictive model during subsequent weaving welding.

According to this configuration, for example, when a skilled operator operates an operating device to perform weaving welding, the input time of a change operation command by a skilled operator and the change contents according to the command are stored in the memory device. A predictive model is generated by learning using the time-series data of the images taken with the camera. During subsequent weaving welding, the predicted changes can be output using the predictive model. Therefore, even if an operator is not skilled in arc welding, he or she can become an expert by operating the operating device based on the changes output while viewing the photographed image of the welding site displayed in real time on the display device. Since it is possible to learn operator operations, it is possible to develop skilled operators.

Further, the controller may be arranged such that the changing operation command inputted from the operating device includes a command for changing a swinging center position of the welding torch or a swinging width of the welding torch. It may be configured to include a command for changing any one of the cycle, the stopping time at the swing end of the welding torch, the distance between the welding torch and the workpiece, the welding current, and the welding speed.

Further, the vehicle may include a notification device that notifies changes output from the machine learning device when in the recommendation operation mode.

The machine learning device is configured to input changes predicted from time-series data of captured images captured by the camera to the controller using the generated prediction model during the automatic driving mode. In the automatic operation mode, the controller controls the welding device to execute the weaving welding based on the predetermined welding program, and also controls the welding device to perform the weaving welding from the machine learning device during the weaving welding. The welding device may be configured to continue the weaving welding by controlling the welding device to reflect the changed content when the changed content is input.

According to this configuration, in automatic operation mode, the changes predicted by the prediction model are input to the controller, and the controller controls the welding equipment to reflect the changes, so the operator can Welding quality can be improved without changing welding conditions such as the position of the center of swing of the welding torch.

The present invention has the above-described configuration and has the advantage of being able to provide an arc welding system that can train operators skilled in arc welding.

FIG. 1 is a block diagram showing a schematic configuration of an arc welding system according to an example of this embodiment. FIG. 2 is a perspective view showing an example of an object to be welded. FIGS. 3A to 3C are schematic diagrams each showing an example of the locus of the tip of the welding torch when performing weaving welding on the groove of the workpiece in the first operation mode. FIGS. 4A to 4C show an example of the shape of a molten pool photographed by a camera for photographing a welding site and displayed on the screen of a display device, and the cross-sectional shape of a weld bead formed in that case, respectively. FIG.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. In addition, below, the same reference numerals are given to the same or equivalent element throughout all the drawings, and the overlapping explanation will be omitted. Furthermore, the present invention is not limited to the following embodiments.

(Embodiment)
FIG. 1 is a block diagram showing a schematic configuration of an arc welding system according to an example of this embodiment.
The arc welding system WS shown in FIG. 1 includes a welding device A, a control device 2 for controlling the welding device A, an operating device 3 for an operator to perform various operations such as starting and stopping the operation of the welding device A, and the like. We are prepared.

The welding apparatus A includes a welding robot 1 that performs TIG welding, and also includes a moving device (not shown) that can move the welding robot 1 and change the three-dimensional position of the welding robot 1. The welding robot 1 is an articulated robot and includes a head support section 10 at the tip of the arm. An oscillator device 11 is attached to the head support portion 10, and a welding torch 12 is attached to the oscillator device 11.

The welding torch 12 is removably fitted with an electrode rod 13 made of tungsten, with the tip of the electrode rod 13 protruding. The welding torch 12 electrically connects the electrode rod 13 to the negative power terminal of the welding power supply device 6 . The welding power supply device 6 has its positive power terminal electrically connected to the workpiece 50 and applies voltage between the electrode rod 13 and the workpiece 50 to connect the tip of the electrode rod 13 and the workpiece 50. An arc is generated between the workpiece 50 and the workpiece 50 to be welded. In addition, the welding torch 12 guides the shielding gas supplied from the gas supply source and sprays it onto the arc generation area, thereby preventing the effects of atmospheric oxygen, nitrogen, etc., and preventing oxidation and nitridation of the molten part. It is configured as follows.

The oscillator device 11 is configured to swing the welding torch 12 in a direction intersecting the welding direction x. Control commands such as the swing width (weaving width) and the swing period (weaving cycle) at this time are output from the oscillator control device 4 to the oscillator device 11 based on commands from the control device 2 (controller 21). The welding direction x is the direction in which the groove of the object to be welded 50 extends, and is the direction in which welding proceeds.

While the head support part 10 is moved in the welding direction x by the arm movement of the welding robot 1, the welding torch 12 is oscillated by the oscillator device 11, so that the welding torch 12 is moved in the welding direction x. moves in the welding direction x while swinging in the intersecting direction (that is, performs a weaving operation). Other known configurations may be used as the configuration for performing such a weaving operation. In this welding apparatus A, arc welding (weaving welding) can be performed while the welding torch 12 is in a weaving operation.

Furthermore, a wire feeding device 14 for feeding a wire-shaped welding wire (filler metal) 15 is attached to the welding robot 1. The wire feeding device 14 pays out the welding wire 15 from a wire reel (not shown) and feeds it to the molten pool via the contact tip 16 in response to a feeding command given from the controller 21 .

Hot wire heating device 5 is a power source that applies a predetermined voltage between contact tip 16 and workpiece 50 . This allows current (hot wire current) to flow through the welding wire 15 to perform Joule heating, thereby increasing the wire feeding speed.

In this example, the oscillator control device 4 is provided separately from the controller 21 that controls the welding device A, but the oscillator control device 4 and the controller 21 constitute the controller that controls the welding device A. be done.

Further, the welding device A is equipped with an AVC (arc voltage control) device 7. The AVC device 7 is configured to detect the arc voltage and adjust the distance between the welding torch 12 and the workpiece 50 so that the arc voltage becomes the arc reference voltage in order to keep the arc length constant. There is. As a configuration for adjusting this interval, for example, an electric slide shaft (not shown) for sliding the welding torch 12 up and down (in a direction to enlarge or reduce the above-mentioned interval) is attached to the welding torch 12, and the electric slide A configuration may be adopted in which the axis is operated, or a configuration in which the above-mentioned interval is adjusted by operation of the arm of the welding robot 1.

Here, the distance between the welding torch 12 and the workpiece 50 is determined according to the arc reference voltage set in the AVC device 7, and the distance can be changed by changing the arc reference voltage. The arc reference voltage can be changed by the controller 21 outputting a signal for changing the arc reference voltage to the AVC device 7 based on a change operation command for changing the arc reference voltage from the operating device 3. can.

The arc welding system WS includes a camera 17 for photographing a welding site, a camera 18 for photographing a bead, and a 3D laser sensor 19, which are attached to the head support section 10 of the welding robot 1. The welding site photographing camera 17 photographs the welding site including the molten pool from an oblique direction (obliquely upward in FIG. 1) in front of the welding torch 12 in the welding direction x during welding. The bead photographing camera 18 photographs the welded bead from an oblique direction (obliquely upward in FIG. 1) behind the welding torch 12 in the welding direction x. Further, the 3D laser sensor 19 images the groove region of the workpiece 50 in front of the welding torch 12 in the welding direction x.

Further, the arc welding system WS includes an image processing device 9 and a display device 8. The image processing device 9 inputs the image (video) data taken by the welding site photographing camera 17 and the bead photographing camera 18, performs appropriate processing, and displays the data as a video on the display device 8 in real time. is configured to allow. The image processing device 9 also processes the information indicating the groove area imaged by the 3D laser sensor 19 to obtain groove information (position of the inner wall surface of the groove, center position in the width direction of the groove, groove area). The groove depth, etc.) can be determined, and the cross-sectional shape of the groove, etc., based on the determined groove information can be displayed on the display device 8. Further, the groove information is output to the controller 21. Note that the display device 8 and the image processing device 9 may be provided for each of the welding site photographing camera 17, bead photographing camera 18, and 3D laser sensor 19.

The control device 2 includes a controller 21 that controls the welding device A described above, a machine learning device 22, and a memory 23. The controller 21, machine learning device 22, and storage device 23 are configured to be able to send and receive information to and from each other. Note that the machine learning device 22 and the storage device 23 may be provided outside the control device 2, separated from the controller 21.

The controller 21 is composed of, for example, a microcontroller, MPU, PLC, etc., and includes an arithmetic processing section such as a CPU, and a storage section having nonvolatile and volatile memories. This storage unit stores a predetermined welding program (including teaching information for the workpiece 50 of the welding robot 1) for controlling the welding device A, and the CPU reads and executes the welding program. By doing so, arc welding is performed on the workpiece 50 by the welding apparatus A.

The machine learning device 22 includes an arithmetic processing unit such as a CPU, and a storage unit including nonvolatile and volatile memories. This storage unit stores a predetermined learning program for performing machine learning, and when the CPU reads and executes the learning program, machine learning is performed using the information stored in the storage device 23. A predictive model, which will be described later, is constructed (generated) as a model. Note that the storage device 23 has a storage medium that stores information used to construct a machine learning model (prediction model).

The operating device 3 is a device that receives operating instructions from an operator, and has an operating section operated by the operator. The operation unit can be configured using a touch panel, a toggle switch, etc., and is not particularly limited. When the operating device 3 receives an operating instruction from an operator, it outputs a corresponding operating instruction to the control device 2 (controller 21).

FIG. 2 is a perspective view showing an example of the object to be welded 50. As shown in FIG. The object to be welded 50 is composed of two members to be welded 50a and 50b. The object to be welded 50 is, for example, a part of a liquefied natural gas storage tank. In a state where the two members to be welded 50a and 50b are butted against each other, a groove (groove portion 51) is formed in the butted portion. The joint portion where the groove portion 51 is formed is arc welded by the welding device A. In this example, weaving welding is performed by moving the welding torch 12 in the welding direction x while swinging it in a direction y that intersects with the welding direction x (performing a weaving operation).

Next, the operation of the arc welding system WS of this embodiment will be explained. This arc welding system WS has three operation modes, first to third, as operation modes for performing arc welding. An operator can operate the operating device 3 to select one of the operating modes. The arc welding performed in this embodiment is weaving welding as described above, and is multilayer welding in which multilayer welding is performed. Alternatively, pulse welding may be used in which a peak current and a base current are alternately repeated as the welding current.

First, the first operation mode will be explained. In the first operation mode, the controller 21 controls the welding apparatus A based on the welding program, and performs weaving welding on the groove portion 51 of the workpiece 50. During this welding, for example, a skilled operator looks at the image (video) of the welding site taken by the welding site photographing camera 17 and displayed on the display device 8, and performs operations to ensure high-quality welding. By operating the device 3, welding conditions (welding parameters) set based on a welding program can be changed.

For example, the workpieces 50a and 50b, which are the workpieces 50, have processing errors (manufacturing errors) and setting errors during manufacturing, and variations in the groove state due to thermal deformation during welding. . If the controller 21 only controls the welding apparatus A based on the welding program, a deterioration in welding quality is unavoidable, and welding defects may occur. Note that by executing the welding program, the controller 21 adjusts (corrects) the movement path of the welding torch 12 from the planned path based on the groove information obtained from the output of the 3D laser sensor 19. However, even then, it may be difficult to deal with variations in the groove condition due to thermal deformation during welding, for example. Therefore, the operator looks at the image taken by the welding part photographing camera 17 (video of the welding part) and changes the welding conditions as necessary.

The images taken by the welding site photographing camera 17 include images of the tip of the welding torch 12, the electrode rod 13 protruding from the tip, the generated arc, the molten pool, the welding wire 15, and the like.

In this first operation mode and the second operation mode described later, welding conditions that can be changed by the operator by operating the operating device 3 during welding include, for example, weaving center position (swinging center position of the welding torch), weaving Width (oscillation width of the welding torch), weaving period (oscillation period of the welding torch), stopping time at the left and right ends of weaving of the welding torch 12 (stopping time at the left and right ends of the welding torch oscillation), arc standard These include voltage (distance between welding torch and workpiece), welding current, welding speed (speed in welding direction x), hot wire current, wire feeding speed, wire vertical position, wire horizontal position, etc. In order to change these welding conditions, the operating device 3 includes, for example, an operating section in which the amount of change by one operation (for example, touch operation or button press operation) is determined in advance for each welding condition.

An example in which the operator operates the operating device 3 to change welding conditions such as the weaving center position and weaving width in the first operation mode will be described with reference to FIGS. 3 and 4.

3(A) to (C) respectively show a trajectory S of the tip of the welding torch 12 (the tip of the electrode rod 13) when performing weaving welding on the groove 51 of the workpiece 50 in the first operation mode. It is a schematic diagram showing an example. 3A to 3C show cases in which the welding width of the groove portion 51 is different. Here, the welding width of the groove 51 is the distance between the welding left end line 52a and the welding right end line 52b in the figure, and in the case of the first layer of multilayer welding, the welding width of the groove 51 or the opening (gap ), and in the case of second and subsequent layers, it is the width of the bead of the previous layer.

During welding, the operator monitors the images taken by the welding part photographing camera 17 displayed on the screen of the display device 8, and, for example, determines the positions of the left and right ends of the weaving indicated by the tip trajectory S of the welding torch 12, and the position of the welding part. It is determined whether or not the left end and right end lines 52a, 52b match, and if they do not match, an operation is performed to make them match. Note that the operator knows the positions of the left end and right end of weaving by looking at the tip portion (electrode rod 13) of the welding torch 12 that is performing the weaving operation.

FIG. 3A shows an ideal case in which the welding width of the groove portion 51 is constant. In this case, since the positions of the left and right ends of the weaving continue to match the left and right welding end lines 52a and 52b, no operation is performed by the operator, and a constant weaving width is maintained.

FIG. 3(B) shows a case where the welding width of the groove portion 51 widens from side to side (y direction) when welding progresses in the welding direction x. In this case, the operator determines to increase the weaving width by looking at the positions of the left and right ends of the weaving and the welding left and right end lines 52a and 52b from time t1 to t6, and the welding torch 12, for example, While passing through region T1, an operation is performed to increase the weaving width. Here, for example, the operator wants to move the right end position of the weaving to the right from the state at times t1, t3, and t5, and also wants to move the left end position of the weaving from the state at times t2, t4, and t6. We consider that we want to move the position to the left and, as a result, decide to increase the weaving width.

The operation unit for changing the weaving width is composed of, for example, a weaving width increase switch and a weaving width decrease switch, and when the weaving width increase switch is operated once, the weaving width increases by a predetermined amount, and when the weaving width decrease switch is operated once, the weaving width decrease switch is operated once. The oscillator device 11 attached to the welding robot 1 is controlled so that the weaving width is reduced by a predetermined amount when the welding robot 1 is operated once.

Note that although FIG. 3(B) illustrates a case where the weaving width is increased, when the weaving width is decreased, a phenomenon opposite to that in FIG. 3(B) occurs.

FIG. 3C shows a case where the welding right end line 52b is shifted to the right side when welding progresses in the welding direction x. In this case, the operator looks at the right end position of the weaving from time t11 to time t13 and the welding right end line 52b, and considers that he wants to move the right end position of the weaving to the right, and at the same time, the operator wants to move the right end position of the weaving to the right. Looking at the position and the welding left end line 52a, we consider that we want to maintain the current left end position of the weaving, and as a result, we decide to move the weaving center position to the right and increase the weaving width. Then, while the welding torch 12 passes through, for example, region T2, an operation for moving the weaving center position to the right and an operation for increasing the weaving width are performed.

Here, in this example, the operation for moving the weaving center position to the left or right is performed by moving the oscillator device 11, on which the welding torch 12 is attached, to the left or right along with the head support 10 by the arm movement of the welding robot 1. This is a moving operation. This operation section is composed of, for example, a right movement switch and a left movement switch, and when the right movement switch is operated once, the head support part 10 moves to the right by a predetermined amount (for example, 0.2 mm), and the left movement switch When operated once, the welding robot 1 is controlled so that the head support section 10 moves to the left by a predetermined amount (for example, 0.2 mm).

In addition, in FIG. 3(C), the case where the weaving center position is moved to the right and the weaving width is increased is illustrated, but when the weaving center position is moved to the left and the weaving width is increased, This is a case where a left-right reverse phenomenon occurs compared to the case of FIG. 3(C).

As described above, the operator looks at the positions of the left and right ends of the weaving, the left welding end line 52a and the right welding end line 52b, and performs operations to change the weaving width and the weaving center position. I have to.

The operator also changes the welding conditions (welding parameters) other than the weaving center position and weaving width by looking at the state of the molten pool in the image (video) taken by the welding site photographing camera 17 displayed on the display device 8. be able to.

FIGS. 4A to 4C show the shapes of molten pools P1 to P3 photographed by the welding part photographing camera 17 and displayed on the screen of the display device 8, and the weld bead B1 formed in that case, respectively. 3 is a schematic diagram showing an example of the cross-sectional shape of B3. FIG.

As shown in FIG. 4(A), when the molten pool P1 appears approximately elliptical, a well-shaped bead B1 is formed. In this case, the operator does not perform any operation to change the welding conditions.

Further, as shown in FIG. 4(B), when the molten pool P2 appears gourd-shaped, a bead B2 having a cross-sectional shape with a concave center is formed. Therefore, when the molten pool P2 is about to take on a gourd shape or a shape approaching it, the operator performs an operation to change the welding conditions. In this case, for example, an operation such as lengthening the weaving cycle is performed.

Further, as shown in FIG. 4(C), when the molten pool P3 appears to have a protruding center portion, a bead B3 having a cross-sectional shape with a raised center is formed. Therefore, the operator performs an operation to change the welding conditions when the molten pool P3 has a shape in which the central portion of the molten pool P3 is about to protrude or become close to it. In this case, for example, an operation such as lengthening the stopping time at the left end and right end of the weaving is performed.

Each operating section for changing the weaving period, the stopping time at the left end of weaving, and the stopping time at the right end of weaving is composed of, for example, an increase switch and a decrease switch, and the increase switch is operated once. Then, the oscillator device 11 is controlled so that the time increases by a predetermined time, and when the decrease switch is operated once, the time decreases by a predetermined time.

In addition, each operating section for changing each of the arc reference voltage (distance between the welding torch and the workpiece), welding current, and welding speed (speed in the welding direction It is composed of an increase switch that increases the amount and a decrease switch that decreases the amount by a predetermined amount, and the welding device A is controlled according to the operation. Each operating section for changing other welding conditions (hot wire current, wire feeding speed, wire vertical position, wire horizontal position, etc.) can also be configured in the same manner.

In the first operation mode, the operator is preferably a skilled person, and when the skilled operator performs an operation on the operator 3 to change the welding conditions, the operator 3 transmits the changed content (changed information) to the controller 21. A change operation command indicating the welding conditions and the amount of change thereof is output. The storage device 23 stores time-series data of photographed images of the welding part taken by the welding part photographing camera 17 during welding, as well as the input time and date of the change operation command input from the operating device 3 to the controller 21. The contents of the change corresponding to the change operation command are stored.

Further, the arc welding system WS has a learning mode in which the machine learning device 22 generates a prediction model (machine learning model). Note that the second and third operation modes can be implemented after at least one learning mode has been performed and a predictive model has been generated. The learning mode can be selected by the operator by operating the controller 3.

When the operator operates the controller 3 to select one of the first to third operation modes and learning mode, a signal indicating that the mode has been selected is sent to the controller 3. 3 to the controller 21. Further, when one of the second and third operating modes and learning mode is selected, a signal indicating that the mode has been selected is further output from the controller 21 to the machine learning device 22.

In the learning mode, the machine learning device 22 uses time-series data of captured images of the welding site imaging camera 17 stored in the storage device 23 when the first operation mode was performed, and the input time and input time of the change operation command. Machine learning, such as deep learning, is performed using the changes according to the command, and a prediction model is generated using time-series data of captured images as input and the predicted changes as output. Note that, as is well known, deep learning has a function of automatically extracting feature quantities, and there is no need to design feature quantities in advance. As the prediction model, a model that can handle time-series data, such as a long short term memory (LSTM) network, which is a type of recurrent neural network (RNN), can be used. Since the LSTM network is a known technology, detailed explanation thereof will be omitted.

For example, as described with reference to FIGS. 3(B) and 3(C), a skilled operator can determine the position of the left and right ends of weaving and the welding left and right end lines 52a and 52b during a plurality of weaving cycles. By looking at the positional relationship, it is determined whether or not it is necessary to change the weaving width and/or the weaving center position, and if it is determined that it is necessary, the changing operation is performed. Therefore, the prediction model determines the time-series data of captured images during a period of multiple weaving cycles immediately before the welding condition change operation is performed as data that begins to show signs of the change operation being performed, and sets a label. Constructed.

Next, the second operation mode (recommendation operation mode) will be explained. This second operation mode is an operation mode in which recommendations are made to the operator. In the second operation mode, the controller 21 controls the welding apparatus A based on the welding program, and performs weaving welding on the groove portion 51 of the workpiece 50. During this welding, the images taken by the welding part imaging camera 17 are displayed on the display device 8 in real time, and the time series data of the images are input to the machine learning device 22 in real time. The machine learning device 22 inputs the time-series data of the photographed images to the above-mentioned prediction model in real time, and when changes are obtained as output from the prediction model, the changes are output to the display device 8. The display device 8 displays the entered changes on the screen. The operator looks at the content of the change and operates the operating device 3 according to the content of the change, whereby a change operation command is output from the operating device 3 to the controller 21. The controller 21 controls the welding apparatus A so as to reflect the change contents according to the change operation command from the operation device 3.

Here, the display device 8 is used as an example of a notification device for notifying the operator of the content of the change, but the content of the change may be outputted in audio using a speaker. For example, the notification of the change may be continued until the operator operates the operating device 3 according to the change.

In this second operation mode, the expected changes are notified on the display device 8, etc., so that even an operator who is not skilled in arc welding can view the photographed image of the welding part displayed on the display device 8. However, by operating the operating device 3 based on the notified change details, the user can learn the operations of a skilled operator, and thus it is possible to train skilled operators. Further, the controller 21 controls the welding apparatus A so as to reflect changes input from the operating device 3, thereby improving welding quality.

In addition, in order to verify the prediction model, when we tested the prediction model by inputting time series data of photographed images taken by the welding site photographing camera 17 in advance, the output (displayed on the display device 8 which is an alarm) was tested. ), the predicted change content and its predicted timing matched with the change content and operation timing by the skilled operator with a high probability of 80% or more. Here, we have only verified the leftward and rightward movements of the weaving center position as changes, but by accumulating a larger amount of learning data in the storage device 23 and having it learn, other changes such as the weaving width can be made. It is thought that it will also be possible to match the output content of the prediction model with the change content and operation timing by a skilled operator with a high probability.

Also in the second operation mode, the input time of the change operation command input from the operating device 3 to the controller 21 together with the time series data of the photographed images of the welding site taken by the welding site photographing camera 17 during welding. and the contents of the change according to the change operation command may be stored in the storage device 23. The data stored in the storage device 23 in the second operation mode may also be used in the learning mode that will be performed again later.

Next, the third operation mode (automatic operation mode) will be explained. In the third operation mode, the controller 21 controls the welding apparatus A based on the welding program, and performs weaving welding on the groove portion 51 of the workpiece 50. During this welding, time-series data of images taken by the welding part photographing camera 17 is input to the machine learning device 22 in real time. The machine learning device 22 inputs the time-series data of the photographed images to the prediction model described above in real time, and when changes are obtained as output from the prediction model, the changes are output to the controller 21. The controller 21 controls the welding apparatus A to reflect the changed contents. Thereby, it is possible to improve welding quality without the operator changing the welding conditions.

In this embodiment, a device that performs TIG welding is illustrated as the welding device A, but it may be a device that performs MIG welding or the like that sends out a welding wire from a welding torch.

From the above description, many modifications and other embodiments of the invention will be apparent to those skilled in the art. Accordingly, the above description is to be construed as illustrative only, and is provided for the purpose of teaching those skilled in the art the best mode of carrying out the invention. Substantial changes may be made in the structural and/or functional details thereof without departing from the spirit of the invention.

INDUSTRIAL APPLICATION This invention is useful as an arc welding system etc. which can aim at training of operators skilled in arc welding.

WS Arc welding system A Welding device 21 Controller 22 Machine learning device 23 Memory device 3 Operating device 8 Display device 12 Welding torch 17 Camera for photographing the welding site

Claims (3)

A welding device capable of performing weaving welding by moving a welding torch in the welding direction while swinging it in a direction crossing the welding direction;
an operating device for an operator to operate the welding device;
a controller that controls the welding device to execute the weaving welding based on a predetermined welding program;
a camera that takes a video of the welding area including the molten pool during the weaving welding;
a display device that displays images taken by the camera in real time;
memory device,
a machine learning device that performs machine learning based on a predetermined learning program ;
Equipped with an alarm,
having a first operating mode and a second operating mode;
The controller is configured to change the swing center position of the welding torch or the swing width of the welding torch from the operating device by an operator's operation during the weaving welding in the first operation mode and the second operation mode. When a change operation command consisting of a command for contents to be changed is input, the welding device is controlled to reflect the change contents according to the change operation command to continue the weaving welding,
The storage device stores, in the first operation mode, time-series data of captured images taken by the camera, an input time of the change operation command input to the controller, and change contents according to the command. remember,
The machine learning device performs learning using the time-series data of the photographed images stored in the storage device, the input time of the change operation command, and the change contents according to the command, and inputs the input time of the photographed images. Generate a prediction model with time series data and the output as the predicted change content,
In the second operation mode, the machine learning device outputs changes predicted from time-series data of captured images captured by the camera using the prediction model during the weaving welding to the alarm. death,
The notification device notifies an operator of the change content output from the machine learning device.
Arc welding system. The changing operation command inputted from the operating device to the controller includes , in addition to a command for changing the swing center position of the welding torch or the swing width of the welding torch, a swing period of the welding torch. , including a command to change any one of the stopping time of the welding torch at the swinging end, the distance between the welding torch and the workpiece, the welding current, and the welding speed ,
The arc welding system according to claim 1. It has an automatic operation mode which is a third operation mode,
The machine learning device is
When in the automatic operation mode, the prediction model is used to input changes predicted from time series data of photographed images taken by the camera during the weaving welding to the controller,
The controller is
In the automatic operation mode , when the change content is input from the machine learning device during the weaving welding, the welding apparatus is controlled to reflect the change content to continue the weaving welding. Ta,
The arc welding system according to claim 1 or 2 .

Images