Method for generate welding process model

Abstract

To generate a welding process model capable of predicting welding results.SOLUTION: A method for generating a welding process model acquires welding parameters including a plurality of parameters, welding results of laser welding in which at least one displacement parameter of the welding parameters is displaced, and characteristic information of the welding results corresponding to the displacement of the displacement parameter; generates welding prediction results which predict the welding results of the laser welding based on the characteristic information of the welding results; and generates the welding process model which predicts the welding results of the laser welding based on learning using the welding prediction results.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a method for generating a welding processing model. [Background technology]

[0002] Patent Document 1 discloses a computer-implemented method for operating a laser material processing machine, which identifies estimated results that characterize how good an actual result of laser material processing will be depending on specified process parameters, and varies the process parameters by Bayesian optimization using a data-based model until the actual result of laser material processing is sufficiently good. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2021-178362 A Summary of the Invention [Problem to be solved by the invention]

[0004] The present disclosure provides a method for generating a welding model that generates a welding model capable of predicting welding results. [Means for solving the problem]

[0005] The present disclosure provides a method for generating a welding processing model performed by a computer, which acquires welding parameters including a plurality of parameters, a welding result of laser welding in which at least one displacement parameter among the welding parameters is displaced, and characteristic information of the welding result corresponding to the displacement of the displacement parameter, generates a welding prediction result that predicts the welding result of the laser welding based on the characteristic information of the welding result, and generates a welding processing model that predicts the welding result of the laser welding based on learning using the welding prediction result. Effect of the Invention

[0006] According to the present disclosure, a welding process model capable of predicting welding results can be generated. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram showing an example of a system configuration of a terminal device according to a first embodiment; [Diagram 2] A flowchart showing an example of an operation procedure of a terminal device according to the first embodiment. [Diagram 3] A diagram showing an example of a welding result [Figure 4] FIG. 13 is a diagram showing a comparison between a welding result and a welding prediction result. [Diagram 5] A diagram showing an example of the relationship between the number of learning times for a welding processing model and the coefficient of determination [Figure 6] An example of retraining a learning model [Figure 7] A diagram showing an example of the relationship between the number of times of re-learning for a welding processing model and the coefficient of determination [Figure 8] A block diagram showing an example of a system configuration of a welding processing model generation system according to a second embodiment. [Figure 9] A sequence diagram showing an example of an operation procedure of the welding processing model generation system according to the second embodiment. [Figure 10] A diagram showing an example of a welding prediction result screen. [Figure 11] A diagram explaining an example of inputting welding results [Figure 12] A diagram explaining an example of generating learning data [Figure 13] A diagram showing an example of a prediction result DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] (Background to this disclosure) Conventionally, there is a method for optimizing process parameters of laser welding (hereinafter, referred to as "welding parameters") using deep learning, as in Patent Document 1. However, since Patent Document 1 optimizes welding parameters using Bayesian optimization, the obtained welding results are sometimes not suitable for analyzing welding parameters having nonlinear and complex configurations.

[0009] In addition, conventionally, a large amount of training data is required to generate a model. However, in order to prepare a large amount of training data, a large number of actual laser processes and inspection results of a large number of workpieces that are the results of the laser processing are required, which is costly and time-consuming. Therefore, there has been a demand for a technology that generates a model for optimizing welding parameters using less training data.

[0010] Hereinafter, with reference to the drawings as appropriate, each embodiment specifically disclosing a method for generating a welding processing model according to the present disclosure will be described in detail. However, more detailed explanation than necessary may be omitted. For example, detailed explanation of already well-known matters or duplicate explanation of substantially the same configuration may be omitted. This is to avoid the following explanation becoming unnecessarily redundant and to facilitate understanding by those skilled in the art. Note that the attached drawings and the following explanation are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the subject matter described in the claims.

[0011] <Terminology> In the following description, the term "laser welding" is intended to have the broadest possible meaning, unless expressly stated otherwise, and includes welding, soldering, melting, joining, annealing, softening, tackifying, resurfacing, peening, heat treating, fusing, sealing, and stacking.

[0012] (Embodiment 1) First, a description will be given of the configuration of the terminal device P1 according to embodiment 1. Fig. 1 is a block diagram showing an example of the internal configuration of the terminal device P1 according to embodiment 1.

[0013] The terminal device P1 is realized by, for example, a personal computer (hereinafter, referred to as "PC"), a notebook PC, a tablet terminal, etc., and receives input of various data used to generate a welding model, and generates a welding model capable of predicting a laser welding result. The terminal device P1 includes a communication unit 10, a processor 11, a memory 12, an input unit 13, and a monitor 14.

[0014] The welding model here is a model for predicting a welding result when a predetermined laser welding based on welding parameters is performed. In the first embodiment, an example of generating a welding model for predicting the penetration depth of a workpiece during laser welding will be described, but it goes without saying that the present invention is not limited to this.

[0015] The communication unit 10 is a communication interface, and executes reading or writing of data between a recording medium such as an SD card (registered trademark), a microSD (registered trademark), or a Universal Serial Bus (USB (registered trademark)), or between the communication unit 10 and another computer (e.g., a PC, a tablet terminal, a smartphone, etc.). Note that the communication unit 10 is not an essential component and may be omitted.

[0016] The processor 11 is configured using, for example, a Central Processing Unit (hereinafter, referred to as "CPU") or a Field Programmable Gate Array (hereinafter, referred to as "FPGA"), and performs various processes and controls in cooperation with the memory 12. Specifically, the processor 11 refers to a program stored in the memory 12 and executes the program to realize various functions of the terminal device P1.

[0017] The memory 12 has, for example, a random access memory (hereinafter, referred to as "RAM") as a work memory used when executing the processing of the processor 11, and a read only memory (hereinafter, referred to as "ROM") that stores a program that defines the processing of the processor 11. Data generated or acquired by the processor 11 is temporarily stored in the RAM. The program that defines the processing of the processor 11 is written in the ROM.

[0018] The input unit 13 can accept an input operation by a worker who performs the work of generating a welding processing model, converts the input operation into an electrical signal, and outputs the electrical signal to the processor 11. The input unit 13 is a user interface, and may be configured using, for example, a mouse, a keyboard, a touch panel, or the like.

[0019] The monitor 14 is configured using, for example, a Liquid Crystal Display (hereinafter, referred to as "LCD") or an organic electroluminescence (hereinafter, referred to as "EL"). The monitor 14 displays various screens (for example, a screen including a welding result, a screen including a welding prediction result, etc.) output from the processor 11.

[0020] The welding prediction result here refers to the welding result of laser welding predicted based on the characteristic information.

[0021] The learning data DB1 is configured using, for example, a hard disk drive (hereinafter, referred to as "HDD") or a solid state drive (hereinafter, referred to as "SSD"). The learning data DB1 stores welding results used for learning the welding model in association with welding parameters corresponding to the welding results.

[0022] The welding model DB2 is configured using, for example, an HDD or SSD, and stores the welding model generated by the processor 11.

[0023] Next, an example of a procedure for generating a welding model will be described with reference to each of Figs. 2 to 7. Fig. 2 is a flowchart showing an example of an operation procedure of terminal device P1 in embodiment 1. Fig. 3 is a diagram showing an example of a welding result. Fig. 4 is a diagram showing an example of fitting between a welding result and a welding prediction result. Fig. 5 is a diagram showing an example of a relationship between the number of times of learning of a welding model and a coefficient of determination. Fig. 6 is a diagram showing an example of fitting between a welding prediction result and a welding result by a re-learned welding model. Fig. 7 is a diagram showing a relationship between the number of times of re-learning of a welding model and a coefficient of determination.

[0024] The terminal device P1 accepts input of welding parameters (that is, welding conditions for laser welding) and a welding result Dt0 (see, for example, FIG. 3) by an operator (St11).

[0025] 3 is data showing the penetration depth of the workpiece Wk obtained when performing laser welding with varying head position and laser welding speed. The welding parameters include, for example, the laser output, the thickness of the workpiece Wk, the laser beam diameter, the laser spread angle, the gas pressure, the gas flow rate, the head position, and the laser welding speed.

[0026] Terminal device P1 accepts input of each of the welding parameters from which welding result Dt0 was obtained, the parameters whose values ​​change (hereinafter referred to as "displacement parameters"), and characteristic information (e.g., a function, etc.) indicating the characteristics of the change in the welding result caused by the displacement of the displacement parameters (St12). Here, terminal device P1 accepts input of characteristic information corresponding to each of the displacement parameters included in the welding parameters from which welding result Dt0 was obtained.

[0027] The horizontal axis of the welding result Dt0 indicates the laser welding speed, and the vertical axis indicates the head position. The welding result Dt0 includes, as an example, information on the penetration depth of the workpiece Wk by laser welding performed with 58 different combinations of displacement parameters, in which the head position and welding speed are displacement parameters among the welding parameters, and the head position is displaced between -2.0 and +2.0 mm, and the welding speed is displaced between 30 and 700 mm / sec.

[0028] Terminal device P1 accepts input of first characteristic information (e.g., a peak function having a peak at a head position of 0 (zero) mm) indicating the characteristic of welding result Dt0 with respect to a change in head position, and second characteristic information (e.g., a decreasing function) indicating the characteristic of welding result Dt0 with respect to a change in welding speed. Note that, although an example in which two of the welding parameters are set as displacement parameters is described here, the number of displacement parameters is not limited to two and may be at least one.

[0029] Based on the input characteristic information, the terminal device P1 predicts the welding result of laser welding when each of the displacement parameters included in the welding parameters is displaced within the displacement range, and generates a welding prediction result Dt1 (St13). Note that if deep learning is directly performed using a small number of data points, optimization of the learning parameters does not go well due to a lack of learning data, and the prediction accuracy of the welding result does not improve. Therefore, the terminal device P1 in the present disclosure optimizes the learning parameters in advance by fitting using the characteristic information.

[0030] In the example shown in Fig. 3, the terminal device P1 receives input of a function f(x1) shown in (Formula 1) as the first characteristic information, and a function g(x2) shown in (Formula 2) as the second characteristic information. Note that the parameters "x1" and "x2" are displacement parameters. In this case, the parameters to be adjusted to obtain an appropriate fitting are limited to three, μ, σ, and λ2. Therefore, the terminal device P1 determines three parameters μ, σ, and λ2 other than variables x1 and x2 in the input functions f(x1) and g(x2), and varies variable x1 (i.e., head position) between -2.0 and +2.0 mm and variable x2 (i.e., welding speed) between 30 and 700 mm / sec, thereby making it possible to more easily obtain a welding prediction result Dt1 that is closer to the data contained in the welding result Dt0 and includes data for parameters (e.g., head position = 1.5 mm, welding speed 400 mm / sec, etc.) for which no data (welding result) has been obtained from the welding result Dt0.

[0031] Generally, the number of parameters required to generate a welding model by deep learning ranges from several hundred to several thousand. However, since the terminal device P1 generates a welding model by fitting, it is possible to generate a welding model with a small number of parameters (e.g., several) as described above.

[0032]

number

[0033]

number

[0034] The terminal device P1 performs fitting between the predicted welding prediction result Dt1 and the input welding result Dt0 (raw data) by the least squares method (St14).

[0035] Here, the fitting result in which the welding prediction result Dt1 is superimposed on the welding result Dt0 is shown in Fig. 4. The welding result shown in Fig. 4 shows an example in which the welding result Dt0 input by the operator is superimposed on the predicted welding prediction result Dt1 in order to make the fitting between the welding result Dt0 and the welding prediction result Dt1 easier to understand.

[0036] The terminal device P1 executes machine learning or deep learning using the data included in the welding prediction result Dt1 as learning data (St15). The machine learning or deep learning for generating the welding model may be executed repeatedly multiple times. The terminal device P1 generates a welding model capable of predicting the penetration depth of the workpiece (St15). This process enables the terminal device P1 to further optimize a large number of parameters required for machine learning or deep learning. The terminal device P1 also calculates a coefficient of determination R2 based on the results of the machine learning or deep learning using the fitted data.

[0037] Here, the number of learning times (epochs) using the data included in the welding prediction results (learning data) and the coefficient of determination R 2 The relationship between the fitting and the welding prediction results is shown in Figure 5. As shown in Figure 5, the welding prediction results obtained by fitting contain a large amount of data, so the learning converges with high accuracy (score 0.98). The coefficient of determination R 2 is an index showing the accuracy rate of the prediction results by the generated welding processing model, and the closer the score is to 1, the higher the prediction accuracy of the welding processing model is, and the closer the score is to 0 (zero), the lower the prediction accuracy of the welding processing model is. As shown in Figure 5, the coefficient of determination R 2 The score of is closer to 1.

[0038] The terminal device P1 executes re-learning for the generated welding model using the data included in the welding result Dt0 (raw data) as learning data (St16). The re-learning of the welding model may be executed repeatedly multiple times. The terminal device P1 generates a welding model capable of predicting the penetration depth of the workpiece through the re-learning (St16).

[0039] Here, the fitting result in which the welding prediction result Dt2 predicted by the re-learned welding processing model is superimposed on the welding result Dt0 is shown in Fig. 6. Compared with the fitting result shown in Fig. 4, the fitting result shown in Fig. 6 has an increased agreement rate between the welding prediction result Dt2 and the welding result Dt0, and is a prediction result that reflects the magnitude of the values ​​of each data point (here, the penetration depth) that are distributed more nonlinearly.

[0040] In addition, the number of re-learning times (epochs) using the data included in the welding results (learning data) and the coefficient of determination R 2 The relationship between and is shown in Figure 7. If the fitting process is not performed, the welding model optimizes a large number of learning parameters (e.g., hundreds to tens of thousands), making it difficult to improve the prediction accuracy of the welding result. However, by performing fitting using a simple function (step St14), the welding model can obtain learning parameters that are optimized to a certain extent in advance, and the coefficient of determination R is reduced compared to when the fitting process is not performed. 2 The value of R is improved. 2 is an index showing the accuracy rate of the prediction results by the generated welding processing model. As shown in Fig. 7, the coefficient of determination R 2 The score of is closer to 1.

[0041] The terminal device P1 stores the welding prediction result or the welding result used to generate the welding model in the learning data DB1, and stores the welding model generated by re-learning in the welding model DB2. The terminal device P1 stores the coefficient of determination R 2is determined to be equal to or greater than a predetermined value (for example, 0.8, 0.85, etc.), the welding model may be stored, or only the welding model selected by the operator may be stored.

[0042] As described above, even when terminal device P1 according to the first embodiment uses less data (welding result Dt0) than the data (welding result) required for machine learning or deep learning, it is possible to increase the amount of data required for machine learning or deep learning (i.e., data included in the welding prediction result) by using the characteristic information (function). For example, terminal device P1 can efficiently increase the amount of data included in the welding prediction result Dt1 by making the displacement increment of the displacement parameter included in the welding parameters smaller.

[0043] Therefore, terminal device P1 can generate a welding model capable of predicting the welding result (here, the penetration depth) with higher accuracy with a smaller number of data. Furthermore, terminal device P1 can generate a welding model capable of predicting the welding result with higher accuracy by executing re-learning on the generated welding model using data included in the welding result Dt0 (raw data).

[0044] (Embodiment 2) The terminal device P1 according to the first embodiment has been described as an example of generating a welding model capable of predicting welding results with higher accuracy. The terminal device P1 according to the second embodiment will be described as an example of using the generated welding model. In the following description of the second embodiment, the same components as those of the terminal device P1 according to the first embodiment will be given the same reference numerals and will not be described.

[0045] An example of the system configuration of the welding model generation system 100 will be described with reference to Fig. 8. Fig. 8 is a block diagram showing an example of the system configuration of the welding model generation system 100 according to the first embodiment.

[0046] The welding model generation system 100 uses a server S1 to manage a welding model generated by a terminal device P1, and welding results and welding prediction results used in generating the welding model. When welding parameters input by a user are acquired, the welding model generation system 100 generates prediction results that predict welding results corresponding to the welding parameters using the welding model and transmits the prediction results to a user device P2. The welding model generation system 100 includes a terminal device P1, a server S1, a user device P2, and a network NW.

[0047] Since terminal device P1 in the second embodiment transmits the generated welding model and the welding result and welding prediction result used in generating the welding model to server S1, learning data DB1 or welding model DB2 may be omitted. Terminal device P1 may be capable of re-learning the welding model managed by server S1.

[0048] The server S1 includes a communication unit 20, a processor 21, a memory 22, a learning data DB1, and a welding model DB2. The learning data DB1 and the welding model DB2 may be configured separately, or may be an external recording device or an external recording medium connected to the server S1 so as to be capable of data communication therebetween.

[0049] The communication unit 20 is connected to the terminal device P1 and the user device P2 via the network NW so as to be capable of wired or wireless communication with them, and transmits and receives data. The communication unit 20 transmits various data (e.g., prediction results, etc.) output from the processor 21 to the user device P2. The communication unit 20 also outputs various data transmitted from the terminal device P1 or the user device P2 to the processor 21.

[0050] The processor 21 is configured using, for example, a CPU or an FPGA, and performs various processes and controls in cooperation with the memory 22. Specifically, the processor 21 refers to a program stored in the memory 22 and executes the program to realize various functions of the server S1.

[0051] The memory 22 has, for example, a RAM as a work memory used when executing the processing of the processor 21, and a ROM for storing a program that defines the processing of the processor 21. The RAM temporarily stores data generated or acquired by the processor 21. The ROM has written therein the program that defines the processing of the processor 21.

[0052] The user device P2 is a device capable of accepting user operations by a user, and is realized by, for example, a PC, a notebook PC, a tablet terminal, etc. The user device P2 includes a communication unit 30, a processor 31, a memory 32, an input unit 33, and a monitor 34.

[0053] The communication unit 30 is connected to the server S1 via the network NW so as to be able to perform wired or wireless communication with the server S1, and transmits and receives data. The communication unit 30 transmits various data (e.g., welding parameters, etc.) output from the processor 31 to the server S1. The communication unit 30 also outputs various data (e.g., prediction results corresponding to the welding parameters) transmitted from the user device P2 to the processor 31.

[0054] The processor 31 is configured using, for example, a CPU or an FPGA, and performs various processes and controls in cooperation with the memory 32. Specifically, the processor 31 refers to a program stored in the memory 32 and executes the program to realize various functions of the user device P2.

[0055] The memory 32 includes, for example, a RAM as a work memory used when executing the processing of the processor 31, and a ROM for storing a program that defines the processing of the processor 31. The RAM temporarily stores data generated or acquired by the processor 31. The ROM has written therein the program that defines the processing of the processor 31.

[0056] The input unit 33 can accept an input operation by a user, converts the input operation into an electrical signal, and outputs it to the processor 31. The input unit 33 is a user interface, and may be configured using, for example, a mouse, a keyboard, or a touch panel.

[0057] The monitor 34 is configured using, for example, an LCD or an organic EL. The monitor 34 displays various screens output from the processor 31 (for example, a welding parameter input screen, a screen including a predicted result (welding result) of laser welding corresponding to the welding parameters, etc.).

[0058] The network NW connects the terminal device P1 and the server S1, and the server S1 and the user device P2, so that data can be transmitted and received between them.

[0059] Next, a procedure for generating a predicted result (welding result) of laser welding using a welding model will be described with reference to Fig. 9. Fig. 9 is a sequence diagram showing an example of an operation procedure of welding model generation system 100 according to the second embodiment. Note that the example of the operation procedure shown in Fig. 9 is executed after a welding model is generated by the operation procedure shown in Fig. 2. Also, in the example of the operation procedure shown in Fig. 9, the processes of steps St25 to St27 are not essential and may be omitted.

[0060] The user device P2 accepts an input operation for designating welding parameters including a plurality of parameters (e.g., the penetration depth of the workpiece, etc.) (St21), and transmits the welding parameters to the server S1 (St22). The user device P2 may accept a designation operation for a parameter to be predicted (e.g., the penetration depth of the workpiece Wk, etc.) based on the input welding parameters.

[0061] Server S1 acquires the welding parameters transmitted from user device P2, and predicts a welding result obtained when laser welding is performed with the acquired welding parameters using a welding processing model (St23). Server S1 generates a prediction result and transmits it to user device P2 (St24). Server S1 may refer to welding processing model DB2 based on parameters included in the welding parameters or parameters the user wishes to predict, and select a welding processing model capable of predicting parameters included in the welding parameters or parameters the user wishes to predict from among the parameters included in the welding parameters.

[0062] The user device P2 displays the prediction result transmitted from the server S1 (for example, the welding prediction result screen SC1 shown in FIG. 10) on the monitor 34. In addition, the user device P2 accepts input of the welding result obtained when laser welding is actually performed using the parameters included in the welding parameters in response to the prediction result (hereinafter, referred to as "feedback") (St25), and transmits the same to the server S1 (St26).

[0063] Server S1 executes relearning of the welding model using the feedback transmitted from user device P2, i.e., data (raw data) included in the welding result (St27). Note that the relearning of the welding model executed here may be the generation of the welding prediction result executed by terminal device P1 in embodiment 1 and learning using the welding prediction result, or it may be relearning using the welding result. Server S1 executes the same process as terminal device P1 shown in embodiment 1 to execute the relearning of the welding model.

[0064] Next, a prediction result (welding result) screen of the second welding parameter displayed on user device P2 will be described with reference to Fig. 10. Fig. 10 is a diagram showing an example of a welding prediction result screen SC1. It should be noted that the welding prediction result screen SC1 shown in Fig. 10 is merely an example and is not limited thereto.

[0065] 10 includes, as an example, input fields Cd1 and Cd2 for welding parameters, and a display area AR1 that displays a predicted result Rs0 when laser welding is performed with the welding parameters specified in the input fields Cd1 and Cd2. Note that the input fields Cd1 and Cd2 for welding parameters may be omitted from the welding prediction result screen SC1.

[0066] Each of the input fields Cd1 and Cd2 accepts input of welding parameters. For example, the input field Cd1 shown in Fig. 10 accepts input of a welding parameter related to the thickness (mm) of the workpiece Wk. The input field Cd2 accepts input of a welding parameter related to the laser output (W). Although not shown in Fig. 10, the user device P2 also accepts input of welding parameters such as a head position of -1 mm and a welding speed of 130 mm / sec in addition to the input fields Cd1 and Cd2.

[0067] When the user device P2 receives a selection (press) operation of the button BT by the user, the user device P2 generates the welding parameters input in the input fields Cd1 and Cd2 and a control command requesting a prediction of the welding result, and transmits them to the server S1. In addition, the user device P2 displays and visualizes the prediction result Rs0 transmitted from the server S1 in the display area AR1.

[0068] The server S1 selects a predictable welding processing model based on the welding parameters transmitted from the user device P2. The server S1 predicts the welding result of the laser welding using the selected welding processing model and a plurality of parameters included in the welding parameters. The server S1 generates prediction results Rs0 and Rs1 and transmits them to the terminal device P1.

[0069] The predicted result Rs0 is a predicted result of laser welding when the head position and the welding speed, which are displacement parameters among the parameters included in the welding parameters, are changed. Note that the predicted result Rs0 shown in Fig. 10 shows, as an example, a predicted result of the penetration depth of the workpiece Wk. The predicted result Rs1 is a predicted result corresponding to the welding parameters, and shows the penetration depth of the workpiece Wk when the head position is -1 mm and the welding speed is 130 mm / sec.

[0070] The welding prediction result screen SC1 is displayed on the monitor 34 by the processor 31. On the welding prediction result screen SC1, a prediction result Rs1 corresponding to a displacement parameter among the input welding parameters is superimposed on a prediction result Rs0 indicated by the welding model.

[0071] As a result, the welding processing model generation system 100 can predict the welding result when laser welding is performed with the welding parameters, and can support the user in optimizing the welding parameters. The user can know more suitable welding parameters based on the position of the predicted result Rs1 relative to the predicted result Rs0. In addition, the user can obtain suitable welding parameters for obtaining the desired welding result by changing the displacement parameters.

[0072] Further, with reference to Fig. 11 to Fig. 13, another example of generating a welding model will be described. Fig. 11 is a diagram for explaining an example of input of a welding result. Fig. 12 is a diagram for explaining an example of generating learning data. Fig. 13 is a diagram showing an example of a prediction result. Figs. 11 to 13 explain, as an example, the generation of a welding model capable of predicting the penetration width and penetration depth of a workpiece Wk by laser welding.

[0073] The welding result Dt51 shown in Fig. 11 is image data showing a cross section of the workpiece Wk after laser welding. The welding result Dt52 is data in which a boundary line BDL between an area of ​​the workpiece Wk that has been welded by laser welding and an area of ​​the workpiece Wk that has not been welded by laser welding is drawn on the cross section of the workpiece Wk shown in the welding result Dt51. The boundary line BDL may be drawn by an operator or automatically by image processing.

[0074] The workpiece Wk shown in the welding results Dt51 and Dt52 has a penetration width W of the workpiece Wk on the surface SF of the workpiece Wk, and a depth DP of the distance (= penetration depth) between the point with the deepest penetration depth and the surface SF of the workpiece Wk as a reference.

[0075] Terminal device P1 accepts input of welding result Dt52 and the welding parameters from which this welding result Dt52 was obtained. Terminal device P1 generates data Dt53 by extracting a boundary line BDL from the welding result Dt52, and generates data included in the welding result (i.e., learning data) by converting the boundary line BDL into X and Y coordinates based on a reference point Pt0 (0,0) of the data Dt53.

[0076] Terminal device P1 performs machine learning or deep learning using data included in the generated welding result, and generates a welding processing model for predicting the penetration width and penetration depth of the workpiece Wk when laser welding is performed with the input welding parameters. Note that machine learning or deep learning may be performed repeatedly multiple times. Terminal device P1 stores the generated welding processing model in welding processing model DB2, and stores data included in the welding result used to generate the welding processing model in learning data DB1.

[0077] The prediction result Dt61 is the boundary line BDL1 predicted using the welding processing model, and is data predicting the boundary line BDL1 showing the penetration result of the workpiece Wk when laser welding is performed with other welding parameters. Although not shown in Fig. 13, the prediction result Dt61 may be generated including information (e.g., numerical values) on the penetration width of the workpiece Wk on the surface SF of the workpiece Wk and the penetration depth of the workpiece Wk.

[0078] (Additional Note) The above description of each embodiment discloses the following techniques.

[0079] (Technology 1) A method for generating a welding processing model performed by a computer (terminal device P1), Acquiring welding parameters including a plurality of parameters, a welding result of laser welding in which at least one displacement parameter among the welding parameters is displaced, and characteristic information of the welding result corresponding to the displacement of the displacement parameter; generating a welding prediction result that predicts a welding result of the laser welding based on characteristic information of the welding result; generating a welding processing model for predicting a welding result of the laser welding based on learning using the welding prediction result; How to generate a welding model. With this configuration, the terminal device P1 can perform machine learning or deep learning without being limited to the actually measured welding results by generating welding prediction results that predict the welding results, thereby generating a welding processing model that can predict welding parameters (e.g., penetration depth) with higher accuracy.

[0080] (Technology 2) generating a welding model by re-learning the welding model using the welding result; A method for generating a welding processing model as described in (Technology 1). With this configuration, terminal device P1 can generate a welding model capable of predicting the welding result with higher accuracy by executing re-learning of the generated welding model using the welding result Dt0 (raw data) which is the correct data.

[0081] (Technology 3) The characteristic information is a function indicating a characteristic of the welding result. A method for generating a welding processing model according to (Technology 1) or (Technology 2). With this configuration, the terminal device P1 can more easily predict the welding prediction result by using the function.

[0082] (Technology 4) The characteristic information indicates a characteristic of the welding result based on a change in the parameter. A method for generating a welding processing model according to (Technology 1) or (Technology 2). With this configuration, the terminal device P1 can increase the amount of data used for machine learning or deep learning, and can generate a welding processing model that can predict welding parameters (e.g., penetration depth) with higher accuracy.

[0083] (Technology 5) Obtaining a maximum value and a minimum value of the displacement parameter among the welding parameters; generating a welding prediction result of the laser welding in which the displacement parameter is displaced between a maximum value and a minimum value based on the characteristic information of the welding result; The method for generating a welding processing model according to any one of (Technical Technique 1) to (Technical Technique 4). With this configuration, the terminal device P1 can more easily execute prediction of the welding prediction result when each displacement parameter is changed in all patterns by using the function.

[0084] Although various embodiments have been described above with reference to the drawings, it goes without saying that the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications, corrections, substitutions, additions, deletions, and equivalents within the scope of the claims, and it is understood that these also naturally belong to the technical scope of the present disclosure. In addition, the components in the various embodiments described above may be arbitrarily combined within the scope of the invention. [Industrial Applicability]

[0085] The present disclosure is useful as a method for generating a welding model that generates a welding model capable of predicting welding results. [Explanation of symbols]

[0086] 10,20,30 Communications Department 11,21,31 Processors 12,22,32 memory 13,33 Input section 14,34 monitor 100 Welding processing model generation system DB1 training data DB2 Welding Model NW Network P1 terminal equipment P2 User Device S1 Server Wk Work

Claims

1. A method for generating a welding processing model by a computer, comprising the steps of: Acquiring welding parameters including a plurality of parameters, a welding result of laser welding in which at least one displacement parameter among the welding parameters is displaced, and characteristic information of the welding result corresponding to the displacement of the displacement parameter; generating a welding prediction result that predicts a welding result of the laser welding based on characteristic information of the welding result; generating a welding processing model for predicting a welding result of the laser welding based on learning using the welding prediction result; How to generate a welding model.

2. generating a welding model by re-learning the welding model using the welding result; The method for generating a welding processing model according to claim 1 .

3. The characteristic information is a function indicating a characteristic of the welding result. The method for generating a welding processing model according to claim 1 .

4. The characteristic information indicates a characteristic of the welding result based on a change in the parameter. The method for generating a welding processing model according to claim 1 .

5. Obtaining a maximum value and a minimum value of the displacement parameter among the welding parameters; generating a welding prediction result of the laser welding in which the displacement parameter is displaced between a maximum value and a minimum value based on the characteristic information of the welding result; The method for generating a welding processing model according to claim 1 .

Images