JP2021026041A - Welding skill level determination system, storage medium, learnt model generation method, neural network, and learnt model - Google Patents

Abstract

To provide a welding skill level determination system etc., capable of detecting a welding skill level of an operator which performs welding operation with a welder which has a heating instrument for heating a welding wire based upon a temporal transition in current value of the welder, a temporal transition in voltage value of the welder, a temporal transition in feed quantity, in every unit time, of the welding wire of the welder, and a temporal transition in acceleration of the heating instrument.SOLUTION: A welding skill level determination system comprises model generation means 52, and arithmetic means 56 which inputs to a learnt model 54 generated by the model generation means 52 a temporal transition in current value of a welder 10 that reception means 55 receives, a temporal transition in voltage value of the welder 10, a temporal transition in feed quantity, in every unit time, of a welding wire of the welder 10, and a temporal transition in acceleration of a heating instrument 15 so as to compute a welding skill level.SELECTED DRAWING: Figure 5

Description

The present invention relates to a welding skill level determination system for detecting the welding skill level of a worker performing welding work by a welding machine, a storage medium, a trained model generation method, a neural network, and a trained model.

Conventionally, various types of semi-automatic welding machines have been known for joining two members by melting a welding wire composed of molten metal or the like with a heating tool (torch) (see, for example, Patent Document 1 and the like). ). In such a semi-automatic welding machine, a welding wire is supplied from a wire supply device, and the supplied welding wire is heated by a heating device and then automatically sent out from the tip of the heating device. In such a welding method, it is possible to save the trouble of replacing the welding rod as compared with the case where welding is performed using the welding rod, and it is possible to improve the work efficiency. As described above, in the semi-automatic welding machine, the welding wire is automatically supplied, but the welding work itself is manually performed by the operator.

JP-A-2015-205332

However, the conventional semi-automatic welding machine has a problem that the welding quality changes greatly depending on the skill level of the operator. For this reason, it is required to standardize the welding work and manage it numerically, instead of relying on the intuition of an expert.

The present invention has been made in consideration of such points, and the transition of the current value of the welding machine having a heating device for heating the welding wire with time, the transition of the voltage value of the welding machine with time, and the welding machine. Welding skill level determination that can detect the welding skill level of workers who perform welding work with a welding machine based on the time course of the feed amount per unit time of the welding wire and the time course of the acceleration of the heating equipment. It is an object of the present invention to provide a system, a storage medium, a trained model generation method, a neural network, and a trained model.

In the welding skill level determination system of the present invention, the current value of a welding machine having a heating device for heating the welding wire changes with time, the voltage value of the welding machine changes with time, and the welding wire in the welding machine per unit time. Over time, the current value of the welding machine, using teacher data including the time course of the feed amount, the time course of the acceleration of the heating appliance, and the welding skill level of the worker who performs the welding work by the welding machine. , The time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, and the time course of the acceleration of the heating appliance are input, and the welding work is performed by the welding machine. A model generation means for generating a learned model that outputs a worker's welding skill level by machine learning, a time-dependent transition of the current value of the welding machine, a time-dependent transition of the voltage value of the welding machine, and the welding. The receiving means for receiving the time course of the feed amount of the welding wire per unit time in the machine and the time course of the acceleration of the heating appliance, and the learned model generated by the model generating means, the receiving means. The time course of the current value of the welding machine, the time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, and the acceleration of the heating appliance. It is characterized by being provided with a calculation means for calculating the welding skill level by inputting a transition over time.

The welding skill level determination system of the present invention measures the ammeter that detects the current value of the welding machine, the voltmeter that detects the voltage value of the welding machine, and the feed amount of the welding wire in the welding machine per unit time. A welding wire feed amount detecting mechanism for detecting may be further provided.

In this case, the welding wire feed amount detecting mechanism may have a feed roller for feeding the welding wire in the welding machine and an encoder for detecting the rotation amount of the feed roller.

Further, the welding machine has an acceleration sensor for measuring the acceleration of the heating appliance, and the receiving means receives the transition of the acceleration of the heating appliance measured by the acceleration sensor over time. You may.

The welding skill level determination system of the present invention may further include a notification unit for notifying when the welding skill level calculated by the calculation means is lower than a predetermined threshold value.

The welding skill level determination system of the present invention may further include a display unit that displays the welding skill level calculated by the calculation means.

In this case, if the welding skill level calculated by the calculation means is lower than a predetermined threshold value, the guidance screen for welding work by the welding machine may be displayed on the display unit.

Further, a microphone is attached to the heating device, and the teacher data includes the sound of the welded portion detected by the microphone, and the learned data generated by the model generation means. The model also includes the sound of the welded portion as an input, the receiving means also receives the sound of the welded portion detected by the microphone, and the calculation means is the model generating means. In the trained model generated by the above, the change in the current value of the welding machine received by the receiving means over time, the change in the voltage value of the welding machine over time, and the time per unit time of the welding wire in the welding machine. The welding skill level may be calculated by inputting the sound of the welded portion in addition to the time course of the feed amount and the time course of the acceleration of the heating appliance.

According to the present invention, the current value of a welding machine having a heating device for heating the welding wire changes with time, the voltage value of the welding machine changes with time, and the feed amount of the welding wire in the welding machine per unit time changes with time. Accepting the transition and the temporal transition of the acceleration of the heating appliance, the temporal transition of the current value of the welding machine, the temporal transition of the voltage value of the welding machine, and the feed amount per unit time of the welding wire in the welding machine. The current value of the welding machine generated by machine learning using teacher data including the time course, the time course of the acceleration of the heating appliance, and the welding skill level of the worker performing the welding work by the welding machine. Input the time course, the time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, and the time course of the acceleration of the heating appliance, and weld by the welding machine. In the trained model that outputs the welding skill level of the worker who performs the work, the current value of the welding machine received over time, the voltage value of the welding machine over time, and the welding wire in the welding machine It is a storage medium in which a program for causing a computer to execute a process of calculating a welding skill level by inputting a time course of a feed amount per unit time and a time course of an acceleration of the heating appliance is stored.

The learned model generation method of the present invention includes a time course of a current value of a welding machine having a heating device for heating a welding wire, a time course of a voltage value of the welding machine, and a unit time of the welding wire in the welding machine. Using the process of receiving teacher data including the time course of the feed amount, the time course of the acceleration of the heating appliance, and the welding skill level of the worker performing the welding work by the welding machine, and the received teacher data. The time course of the current value of the welding machine, the time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, and the time course of the acceleration of the heating appliance. It includes a process of inputting a transition and generating a learned model by machine learning that outputs the welding skill level of a worker who performs welding work by the welding machine.

According to the present invention, the current value of a welding machine having a heating device for heating the welding wire changes over time, the voltage value of the welding machine changes over time, and the feed amount of the welding wire in the welding machine per unit time changes over time. An input layer for inputting teacher data including a transition, a transition of the acceleration of the heating appliance over time, and a welding skill level of a worker performing welding work by the welding machine, an output layer for outputting a trained model, and the above. It includes an intermediate layer in which parameters are trained using the teacher data, and when the teacher data is input to the input layer, the intermediate layer calculates and outputs the trained model by the output layer. It is a neural network.

According to the present invention, the current value of a welding machine having a heating device for heating the welding wire changes with time, the voltage value of the welding machine changes with time, and the feed amount of the welding wire in the welding machine per unit time changes with time. Over time of the current value of the welding machine generated by machine learning using teacher data including the transition, the transition of the acceleration of the heating appliance over time, and the welding skill level of the worker performing the welding work by the welding machine. Input the transition, the temporal transition of the voltage value of the welding machine, the temporal transition of the feed amount per unit time of the welding wire in the welding machine, and the temporal transition of the acceleration of the heating appliance, and the welding work by the welding machine. This is a trained model that outputs the welding skill level of the worker who performs the operation.

According to the welding skill level determination system, the storage medium, the trained model generation method, the neural network, and the trained model of the present invention, the change over time of the current value of the welding machine having a heating device for heating the welding wire, and the welding machine Welding skill level of workers who perform welding work with a welding machine based on the time course of the voltage value, the time course of the feed amount of the welding wire per unit time in the welding machine, and the time course of the acceleration of the heater. Can be detected.

It is a schematic block diagram which shows schematic structure of the welding system by embodiment of this invention. It is a side view which shows the structure of the wire feeding mechanism of the wire feeding device in the welding system shown in FIG. It is a block diagram which shows the mode of connection of each component in the welding unit of the welding system shown in FIG. It is a table which shows the time-dependent transition of the value detected by various sensors and encoders in the welding system shown in FIG. It is a figure which shows roughly the structure which calculates the welding skill level by the welding system shown in FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. 1 to 5 are diagrams showing a welding system according to the present embodiment.

First, the welding system according to the present embodiment will be described with reference to FIGS. 1 to 4. As shown in FIG. 1, in the welding system according to the present embodiment, various data are sent from the welding machine 10 for the worker to perform the welding work, the fitting 30 worn by the worker, and the welding machine 10. The data collection terminal 40 and the management terminal 50 are provided. Here, the welding machine 10, the fitting 30, and the data collecting terminal 40 are arranged at the welding work site 2. On the other hand, the management terminal 50 is installed in the management room 4 which is a place different from the work site 2. The management terminal 50 is not limited to being installed in the management room 4, and the management terminal 50 may be installed in the work site 2. However, when the management terminal 50 is installed in the management room 4, one management terminal 50 can manage the welding machines 10 and the data collection terminals 40, which are arranged at the plurality of work sites 2, respectively.

The welding machine 10 has a power supply device 12, a wire supply device 14, and a heating device (torch) 15. The wire supply device 14 supplies a welding wire made of molten metal or the like to the heating device 15. The heating device 15 is gripped by an operator, and the welding wire sent from the wire supply device 14 is heated by the heating device 15 to be melted, so that the two members 24 are made of molten metal (FIG. 3) can be joined. The power supply device 12 supplies electric power to the wire supply device 14 and the heating appliance 15.

As shown in FIG. 2, the wire feeding device 14 has a first feed roller 16 and a second feed roller 17, and a welding wire (FIG. 2) is provided between the first feed roller 16 and the second feed roller 17. (Represented by the reference code W in 2) is sent out. Further, the first feed roller 16 is provided with an encoder 18 for detecting the amount of rotation of the first feed roller 16. The feed amount of the welding wire is detected based on the rotation amount of the first feed roller 16 detected by the encoder 18. The encoder 18 is not limited to the one that detects the amount of rotation of the first feed roller 16. The encoder 18 may be adapted to detect the rotation amount of rollers other than the first feed roller 16 as long as the feed amount of the welding wire can be finally detected. Further, the feed amount of the welding wire may be detected by means other than the encoder 18. In the present embodiment, the first feed roller 16, the second feed roller 17, and the encoder 18 constitute a welding wire feed amount detection mechanism that detects the feed amount of the welding wire.

The connection mode of each component in such a welding machine 10 will be described with reference to FIG. As shown in FIG. 3, the power supply device 12 includes a first gas inlet 12a, a second gas inlet 12b, a torch connection terminal 12c, a remote outlet 12d, a torch switch outlet 12e, a manual connection terminal (+) 12f, and a manual connection terminal. (-) 12 g is provided. Here, the gas hose 20a of the argon gas cylinder 20 is connected to the first gas inlet 12a. As a result, the argon gas can be supplied to the power supply device 12 from the argon gas cylinder 20. In addition to the argon gas, the power supply device 12 may be supplied with another type of gas that serves as a shield gas. Further, the cord 22a of the remote control device 22 is connected to the remote outlet 12d. As a result, the welding machine 10 can be remotely controlled by the remote control device 22. Further, the heating appliance 15 is provided with a first cord 15a, a second cord 15b, and a third cord 15c, respectively. Here, the first cord 15a is connected to the torch switch outlet 12e. Further, the second cord 15b is connected to the negative pole of the torch connection terminal 12c. Further, the third cord 15c is connected to the second gas inlet 12b. Further, the heating appliance 15 is provided with a switch 15d, and when the operator holding the heating appliance 15 presses the switch 15d, the welding wire sent from the wire supply device 14 is heated and the heated welding is performed. The wire is sent out from the heating device 15.

Further, a fourth cord 24a is connected to one of the two members 24 welded by the welding machine 10, so that the fourth cord 24a is connected to the + pole of the torch connection terminal 12c. It has become.

Further, as shown in FIG. 3, an ammeter 25 is provided in the second code 15b, and the ammeter 25 detects the current value flowing through the second code 15b. Further, the voltmeter 26 is adapted to measure the potential difference between the second code 15b and the fourth code 24a. Specifically, by pressing the first probe 26a of the voltmeter 26 against the fourth cord 24a and pressing the second probe 26b against the second cord 15b, the potential difference between the second cord 15b and the fourth cord 24a is increased. It will be measured. Further, a 3-axis acceleration sensor 27 is attached to the heating appliance 15. With such an acceleration sensor 27, it becomes possible to detect the acceleration of the heating appliance 15 in three directions including the x direction, the y direction, and the z direction. It should be noted that what is attached to the heating appliance 15 is not limited to the 3-axis accelerometer 27. A 6-axis accelerometer that can detect the rotational angular velocity of the heating device 15 in addition to the acceleration of the heating device 15 in the three directions consisting of the x direction, the y direction, and the z direction may be attached to the heating device 15. As yet another example, a 9-axis accelerometer that can detect the acceleration and rotational angular velocity of the heating appliance 15 in three directions as well as the orientation (compass) of the heating appliance 15 may be attached to the heating appliance 15. Further, a temperature sensor 28 and a humidity sensor 29 are installed at the work site 2 where the welding machine 10 is arranged.

The fitting 30 has a smart glass (registered trademark) 32 equipped with a camera and a projection type or retinal projection type display, and a microphone 34, and an operator wearing the fitting 30 can wear the smart glass 32. The welding work can be performed while looking at the screen displayed on the display mounted on the retina, and the voice spoken by the worker can be picked up by the microphone 34 and sent to the data collecting terminal 40 described later. Further, the field of view of the worker wearing the wearing tool 30 can also be captured by the camera mounted on the smart glasses 32, and the captured image is sent to the data collecting terminal 40 described later. Further, the earphone may be attached to the attachment 30. In this case, it becomes possible to give a voice instruction to the operator wearing the wearing tool 30 by earphones.

The power supply device 12 and the wire supply device 14 of the welding machine 10 are connected to the data collection terminal 40 installed at the work site 2, respectively. Then, information related to the detection results by the encoder 18, the ammeter 25, the voltmeter 26, the acceleration sensor 27, the temperature sensor 28, the humidity sensor 29, and the like of the welding machine 10 is sent over time. The table of FIG. 4 shows various types of information sent to the data collection terminal 40 (specifically, changes over time in values detected by each sensor or the like). In the table of FIG. 4, the “feed amount” is the feed amount of the welding wire, and such a feed amount of the welding wire is calculated based on the rotation amount of the first feed roller 16 detected by the encoder 18. ..

Further, the attachment 30 is connected to the data collection terminal 40. Then, as described above, the voice spoken by the worker is picked up by the microphone 34 and sent to the data collection terminal 40, and the image captured by the camera mounted on the smart glasses 32 is sent to the data collection terminal 40. It has become.

The management terminal 50 provided in the management room 4, which is a place different from the work site 2, is composed of, for example, a personal computer or the like. Specifically, the management terminal 50 includes a display unit 50a such as a monitor, an operation unit 50b such as a keyboard, a storage means 50c composed of a memory such as a ROM or RAM, and a processor 50d such as a CPU. A portable tablet terminal may be used instead of a personal computer having a display unit 50a such as a monitor and an operation unit 50b such as a keyboard. The touch panel provided on such a tablet terminal also serves as a display unit 50a and an operation unit 50b. A wearer 30 and a data collection terminal 40 are communicably connected to such a management terminal 50. Here, the information sent from the power supply device 12 and the wire supply device 14 of the welding machine 10 and the fitting 30 to the data collection terminal 40 is sent from the data collection terminal 40 to the management terminal 50. As a result, the field of view of the operator captured by the camera mounted on the smart glasses 32 can be displayed on the display unit 50a of the management terminal 50. Further, by sending information related to the guidance screen from the management terminal 50 to the fitting 30, the smart glass 32 of the fitting 30 can display the guidance screen for welding work. Further, a microphone may be connected to the management terminal 50. In this case, the administrator who operates the management terminal 50 can transmit the instruction to the operator by voice from the earphone attached to the wearing tool 30 by sending the instruction to the microphone by voice.

In this way, by communicably connecting the fitting 30 attached to the worker who performs the welding work by the welding machine 10 at the work site 2 and the management terminal 50 provided in the management room 4, for example, experience. When a small number of workers perform welding work with the welding machine 10 at the work site 2, an experienced manager can manage the information displayed on the display unit 50a of the management terminal 50 (specifically, manage from the data collection terminal 40). The operation unit 50b can input an instruction by referring to various information of the welding machine 10 sent to the terminal 50 and an image captured by a camera mounted on the smart glass 32. Then, by sending the instruction input to the operation unit 50b to the wearing tool 30, the manager's instruction can be displayed on the screen displayed on the smart glasses 32 of the wearing tool 30. As a result, even an inexperienced worker can perform more accurate welding work by the welding machine 10 at the work site 2 by following the instructions given by the experienced manager.

Further, the welding system of the present embodiment functions as a welding skill level determination system for calculating the welding skill level of an operator. Here, as the welding skill level of the worker, for example, a numerical parameter in the range of 0 to 100 is used. The configuration of such a welding skill level determination system will be described with reference to FIG. The operation of detecting the welding skill level of the operator as shown below is executed by the processor 50d of the management terminal 50.

As shown in FIG. 5, the welding skill level determination system includes a model generation means 52 that creates a trained model 54 based on teacher data 51, a transition of the current value of the welding machine 10 over time, and a voltage value of the welding machine 10. The time-dependent transition of the welding wire in the welding machine 10, the time-dependent transition of the feed amount per unit time of the welding machine 10, and the time-dependent transition of the acceleration of the heating device 15 and the learned means generated by the model generation means 52. The time course of the current value of the welding machine 10 received by the receiving means 55 in the model 54, the time course of the voltage value of the welding machine 10, and the time course of the feed amount of the welding wire in the welding machine 10 per unit time. , And a calculation means 56 for calculating the welding skill level by inputting the transition of the acceleration of the heating device 15 over time, and an output means 58 for outputting the welding skill level output by the calculation means 56.

In such a welding skill level determination system, the trained model 54 is generated in advance by the model generation means 52. Here, the model generating means 52 includes a time course of the current value of the welding machine 10, a time course of the voltage value of the welding machine 10, a time course of the feed amount of the welding wire in the welding machine 10 per unit time, and heating. Using the teacher data 51 including the time course of the acceleration of the appliance 15 and the welding skill level of the worker performing the welding work by the welding machine 10, the time course of the current value of the welding machine 10 and the voltage value of the welding machine 10 Input the time course, the time course of the feed amount of the welding wire in the welding machine 10 per unit time, and the time course of the acceleration of the heating device 15, and enter the welding skill level of the worker who performs the welding work by the welding machine 10. The trained model 54 to be output is generated by machine learning. The time course of the current value of the welding machine 10, the time course of the voltage value of the welding machine 10, the time course of the feed amount of the welding wire in the welding machine 10 per unit time, and the time course of the feed amount per unit time of the welding machine 10 used in the teacher data 51. The time course of the acceleration of the heating device 15 is determined based on the information sent from the data collecting terminal 40 to the management terminal 50 and the information input by the display unit 50a of the management terminal 50. Further, the welding skill level of the worker who performs the welding work by the welding machine 10 used in the teacher data 51 is determined based on the information input by the display unit 50a of the management terminal 50.

The model generation means 52 is composed of, for example, a neural network. Such a neural network includes a time course of the current value of the welding machine 10, a time course of the voltage value of the welding machine 10, a time course of the feed amount of the welding wire per unit time in the welding machine 10, and a heating device 15. The input layer into which the teacher data 51 including the transition of the acceleration of the above and the welding skill level of the worker performing the welding work by the welding machine 10 is input, the output layer by which the trained model 54 is output, and the teacher data 51. It has an intermediate layer whose parameters have been trained using it. Then, in such a neural network, when the teacher data 51 is input to the input layer, it is calculated in the intermediate layer and the trained model 54 is output by the output layer. A known neural network can be used.

Here, when the current value of the welding machine 10, the voltage value of the welding machine 10, or the feed amount of the welding wire in the welding machine 10 per unit time suddenly changes, the welding work is performed by a less skilled worker. It is inferred that it is being done. Further, even when the acceleration of the heating appliance 15 changes suddenly, it is presumed that the heating appliance 15 is operated by an operator with a low skill level. Therefore, in such a case, the welding skill level of the worker performing the welding work can be underestimated. The model generation means 52 comes to generate the trained model 54 regarding such a correlation.

The receiving means 55 per unit time of the welding wire calculated based on the time course of the current value measured by the ammeter 25, the time course of the voltage value measured by the voltmeter 26, and the measurement result by the encoder 18. The time course of the feed amount of the electric current and the time course of the acceleration of the heating device 15 measured by the acceleration sensor 27 are accepted. Then, the calculation means 56 changes the current value of the welding machine 10 over time and the voltage value of the welding machine 10 over time, which is received by the receiving means 55 in the learned model 54 generated by the model generating means 52. , The welding skill level is calculated by inputting the time course of the feed amount of the welding wire per unit time in the welding machine 10 and the time course of the acceleration of the heating device 15. The welding skill level calculated by the calculation means 56 is output by the output means 58. Specifically, the welding skill level output by the output means 58 is displayed on the operation unit 50b of the management terminal 50 or on the smart glasses 32 of the fitting 30.

If the welding skill level output by the output means 58 is lower than a predetermined threshold value, it is considered that the worker who performed the welding work by the welding machine 10 is not an expert (experienced) but a person with little experience. .. In this case, it is notified by voice by a speaker or the like (not shown) that the welding skill level output by the output means 58 is lower than a predetermined threshold value. At this time, the speaker functions as a notification unit for notifying when the welding skill level calculated by the calculation means 56 is lower than a predetermined threshold value. In addition, the notification unit may be configured to perform notification by means other than voice. When the welding skill level output by the output means 58 is lower than a predetermined threshold value, the management terminal 50 sends information related to the guidance screen to the fitting 30. As a result, the welding work guidance screen can be displayed on the smart glass 32 of the fitting 30, so that even an inexperienced worker can perform the welding work while looking at the guidance screen displayed on the smart glass 32. You will be able to do it. Further, regarding the welding work actually performed in this way, information detected by the power supply device 12 of the welding machine 10, the wire supply device 14, the sensors provided in the heating device 15, and the like, and input to the operation unit 50b. Information such as the welding work level of the worker is sent to the data collection terminal 40, and is transmitted from the data collection terminal 40 to the management terminal 50. As a result, new teacher data 51 is input to the model generation means 52 in the management terminal 50, and thus the accuracy of the trained model 54 is further improved.

Further, the welding skill level output by the output means 58 includes information detected by the power supply device 12 of the welding machine 10, the wire supply device 14, the sensors provided in the heating appliance 15, and the smart glasses of the fitting 30. It is stored in the storage means 50c together with the moving image captured by the camera mounted on the 32. By doing this, the actual welding work contents by a skilled worker with a high welding skill level can be compared with the actual welding work contents by a worker with a low welding skill level, and the actual welding work by a skilled worker with a high welding skill level can be compared. You will be able to standardize the content. In addition, since the welding skill level of a certain worker is calculated after the welding work is performed by the welding machine 10, the worker strives to improve the welding skill aiming at a higher score welding skill level. You will be able to study.

In the storage means 50c composed of memories such as ROM and RAM in the management terminal 50, first, the transition of the current value of the welding machine 10 with time, the transition of the voltage value of the welding machine 10 with time, and the welding wire in the welding machine 10 The time course of the feed amount per unit time and the time course of the acceleration of the heating device 15 are accepted, and then the received welding is applied to the trained model 54 generated by machine learning using the teacher data 51. The time course of the current value of the machine 10, the time course of the voltage value of the welding machine 10, the time course of the feed amount per unit time of the welding wire in the welding machine 10, and the time course of the acceleration of the heating device 15. A program for causing the processor 50d to execute a process of calculating the welding skill level by inputting is stored.

According to the welding skill level determination system of the present embodiment having the above configuration, the transition of the current value of the welding machine 10 with time, the transition of the voltage value of the welding machine 10 with time, and the welding wire in the welding machine 10 By machine learning, a trained model 54 is input by inputting the temporal transition of the feed amount per unit time and the temporal transition of the acceleration of the heating device 15, and outputting the welding skill level of the worker performing the welding work by the welding machine 10. Generated and, in the generated learned model 54, the time course of the current value of the welding machine 10 received by the receiving means 55, the time course of the voltage value of the welding machine 10, and the welding wire in the welding machine 10. In order to calculate the welding skill level by inputting the time course of the feed amount per unit time and the time course of the acceleration of the heating device 15, it depends on the expert's intuition about the welding work using the welding machine 10. Instead, welding work can be standardized and managed numerically. In particular, by managing the welding work at the plurality of work sites 2 with one management terminal 50, the number of teacher data 51 input to the model generation means 52 can be increased, and thus the accuracy of the trained model 54 can be improved. It can be further improved. Further, when the welding work at the plurality of work sites 2 is managed by one management terminal 50, the management terminal 50 is compared with the case where the management terminal 50 is installed corresponding to each of the plurality of welding machines 10. Since the number can be reduced, the cost can be reduced.

The welding system and the welding skill level determination system according to the present invention are not limited to the above-described aspects, and various changes can be made.

For example, in the above-mentioned welding skill level determination system, the model generating means 52 uses the current value of the welding machine 10 over time, the voltage value of the welding machine 10 over time, and the welding wire in the welding machine 10 per unit time. The current value of the welding machine 10 is used by using the teacher data 51 including five elements: the time course of the feed amount, the time course of the acceleration of the heating device 15, and the welding skill level of the worker who performs the welding work by the welding machine 10. Input four elements: the time course of the welding machine 10, the time course of the voltage value of the welding machine 10, the time course of the feed amount per unit time of the welding wire in the welding machine 10, and the time course of the acceleration of the heating appliance 15. , The trained model 54 that outputs the welding skill level of the worker who performs the welding work by the welding machine 10 is generated by machine learning. However, the present invention is not limited to such aspects. As another example of the welding skill level determination system according to the present invention, the teacher data 51 includes the time course of the current value of the welding machine 10, the time course of the voltage value of the welding machine 10, and the unit time of the welding wire in the welding machine 10. In addition to the five factors of the change in the amount of feed per feed over time, the change in the acceleration of the heating device 15 over time, and the welding skill level of the worker performing the welding work by the welding machine 10, other factors are further included. May be good. Other factors include, for example, the sound of the weld, the moving image of the pool, and the temperature of the weld. Specifically, when the sound of the welded portion is used as another element, a microphone for detecting the sound of the welded portion is attached to the heating device 15. Further, when the temperature of the welded portion is used as another factor, a thermometer for detecting the temperature of the welded portion is attached to the heating device 15.

Further, in this case, the trained model 54 determines the time course of the current value of the welding machine 10, the time course of the voltage value of the welding machine 10, and the time course of the feed amount of the welding wire in the welding machine 10 per unit time. Welding by a worker who inputs five or more elements including other added elements in addition to the four elements of the transition and the transition of the acceleration of the heating device 15 over time, and performs the welding work by the welding machine 10. The skill level is used as the output. Further, the receiving means 55 includes a time course of the current value of the welding machine 10, a time course of the voltage value of the welding machine 10, a time course of the feed amount of the welding wire per unit time in the welding machine 10, and a heating device. In addition to the four elements of the transition of the acceleration of 15 over time, information on other elements will be accepted. Then, the calculation means 56 applies the trained model 54 generated by the model generation means 52 to the time course of the current value of the welding machine 10 received by the reception means 55, the change of the voltage value of the welding machine 10 with time, and so on. The welding skill level is calculated by inputting the time course of the feed amount of the welding wire per unit time in the welding machine 10, the time course of the acceleration of the heating device 15, and other factors.

The model generation means 52 is not limited to the one that generates the trained model 54 from the teacher data 51 by the neural network. As another example of the welding skill level determination system according to the present invention, the model generation means 52 uses other types of machine learning such as decision trees learning, correlation rule learning, gene programming, and clustering to obtain a trained model 54. It may be generated.

Further, as yet another example of the welding skill level determination system according to the present invention, the determination of the welding skill level of the worker is not limited to the management terminal 50. The model generation means 52, the reception means 55, the calculation means 56, and the output means 58 described above are provided not in the management terminal 50 but in the data collection terminal 40 of each work site 2, and the data collection terminal 40 works at each work site 2. The welding skill level of the person may be determined.

2 Work site 4 Control room 10 Welder 12 Power supply 12a 1st gas inlet 12b 2nd gas inlet 12c Torch connection terminal 12d Remote outlet 12e Torch switch outlet 14 Wire supply device 15 Heating equipment 15a 1st cord 15b 2nd cord 15c 3 Code 15d Switch 16, 17 Roller 18 Encoder 20 Argon Gas Bomb 20a Gas Hose 22 Remote Control Device 22a Code 24 Member 24a 4th Code 25 Current Meter 26 Voltmeter 26a 1st Probe 26b 2nd Probe 27 Acceleration Sensor 28 Temperature Sensor 29 Humidity Sensor 30 Wearing device 32 Smart glass 34 Microphone 40 Data collection terminal 50 Management terminal 50a Display unit 50b Operation unit 50c Storage means 50d Processor 51 Teacher data 52 Model generation means 54 Learned model 55 Reception means 56 Calculation means 58 Output means

Claims (12)

The time course of the current value of the welding machine having a heating device for heating the welding wire, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the heating. Using teacher data including the time course of the acceleration of the equipment and the welding skill level of the worker performing the welding work by the welding machine, the time course of the current value of the welding machine and the time course of the voltage value of the welding machine. The transition, the temporal transition of the feed amount of the welding wire per unit time in the welding machine, and the temporal transition of the acceleration of the heating appliance are input, and the welding skill level of the worker who performs the welding work by the welding machine is output. A model generation means that generates a trained model by machine learning,
The time course of the current value of the welding machine, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the time course of the acceleration of the heating appliance. Reception means for accepting transitions and
In the trained model generated by the model generating means, the current value of the welding machine received by the receiving means over time, the voltage value of the welding machine over time, and the welding wire in the welding machine. A calculation means for calculating the welding skill level by inputting the time course of the feed amount per unit time and the time course of the acceleration of the heating appliance.
Welding skill level judgment system equipped with. An ammeter that detects the current value of the welding machine and
A voltmeter that detects the voltage value of the welding machine and
A welding wire feed amount detection mechanism that detects the feed amount of the welding wire per unit time in the welding machine, and
The welding skill level determination system according to claim 1, further comprising. The welding skill level determination system according to claim 2, wherein the welding wire feed amount detecting mechanism includes a feed roller for feeding the welding wire in the welding machine and an encoder for detecting the rotation amount of the feed roller. The welding machine has an acceleration sensor that measures the acceleration of the heating appliance.
The welding skill level determination system according to any one of claims 1 to 3, wherein the receiving means receives a transition of the acceleration of the heating appliance with time measured by the acceleration sensor. The welding skill level determination system according to any one of claims 1 to 4, further comprising a notification unit for notifying when the welding skill level calculated by the calculation means is lower than a predetermined threshold value. The welding skill level determination system according to any one of claims 1 to 4, further comprising a display unit for displaying the welding skill level calculated by the calculation means. The welding skill level determination system according to claim 6, wherein when the welding skill level calculated by the calculation means is lower than a predetermined threshold value, a guidance screen for welding work by the welding machine is displayed on the display unit. A microphone is attached to the heating appliance.
The teacher data also includes the sound of the welded portion detected by the microphone.
The trained model generated by the model generation means also includes the sound of the welded portion as an input.
The receiving means also receives the sound of the welded portion detected by the microphone.
The calculation means has the trained model generated by the model generation means, the current value of the welding machine received by the receiving means over time, the voltage value of the welding machine over time, and the welding. Claims 1 to 7, wherein the welding skill level is calculated by inputting the sound of the welded portion in addition to the time course of the feed amount of the welding wire per unit time in the machine and the time course of the acceleration of the heating device. The welding skill level determination system according to any one item. The time course of the current value of a welding machine having a heating device for heating the welding wire, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the heating. Accepting changes in the acceleration of the equipment over time,
The time course of the current value of the welding machine, the time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, the time course of the acceleration of the heating appliance, And the time course of the current value of the welding machine and the time course of the voltage value of the welding machine generated by machine learning using the teacher data including the welding skill level of the worker who performs the welding work by the welding machine. , Learning to input the time course of the feed amount of the welding wire per unit time in the welding machine and the time course of the acceleration of the heating appliance, and output the welding skill level of the worker who performs the welding work by the welding machine. The current value of the welding machine received over time, the voltage value of the welding machine over time, the feed amount of the welding wire in the welding machine over time, and the heating appliance A storage medium that stores a program that causes a computer to execute a process to calculate the welding skill level by inputting the transition of the acceleration of. The time course of the current value of the welding machine having a heating device for heating the welding wire, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the heating. The process of accepting teacher data including the transition of the acceleration of the equipment over time and the welding skill level of the worker performing the welding work by the welding machine, and
Using the received teacher data, the time course of the current value of the welding machine, the time course of the voltage value of the welding machine, the time course of the feed amount per unit time of the welding wire in the welding machine, and the time course of the feed amount per unit time in the welding machine. A process of generating a learned model by machine learning, in which the transition of the acceleration of the heating appliance over time is input and the welding skill level of the worker performing the welding work by the welding machine is output.
A trained model generation method with. The time course of the current value of the welding machine having a heating device for heating the welding wire, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the heating. An input layer into which teacher data including the transition of the acceleration of the equipment over time and the welding skill level of the worker performing the welding work by the welding machine is input.
The output layer that outputs the trained model and
An intermediate layer whose parameters have been learned using the teacher data, and
With
A neural network that, when the teacher data is input to the input layer, calculates in the intermediate layer and outputs the trained model by the output layer. The time course of the current value of the welding machine having a heating device for heating the welding wire, the time course of the voltage value of the welding machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the heating. The time course of the current value of the welding machine generated by machine learning using the teacher data including the time course of the acceleration of the appliance and the welding skill level of the worker performing the welding work by the welding machine, the welding. A worker who inputs the time course of the voltage value of the machine, the time course of the feed amount of the welding wire per unit time in the welding machine, and the time course of the acceleration of the heating appliance, and performs the welding work by the welding machine. A trained model that outputs the welding skill level of.

Images