Virtual welding system, processing method for virtual welding system, and program

Abstract

To display a realistic welding state in a virtual welding using a virtual welding system.SOLUTION: The virtual welding system includes: prediction means for predicting a welding state in a virtual welding by using a prediction model for predicting the future welding state; simulation means for estimating the current welding state in the virtual welding by a physics calculation with reference to a predicted welding state predicted by the prediction means; and display means for displaying the welding state in the virtual welding in a screen on the basis of the result of estimation by the simulation means.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a virtual welding system, a method for processing a virtual welding system, and a program.

Background Art

[0002] As training for acquiring welding technology, using an actual device has been conventionally performed, and in recent years, training utilizing VR (Virtual Reality) and AR (Augmented Reality) has been increasing. Patent Document 1 discloses a technique in which a user undergoing training wears a helmet incorporating a small display device and operates a simulated welding object (welding coupon) and a mock welding tool, and a state of being pseudo-welded on the display device is displayed in a virtual reality space. By using such a mechanism, it becomes possible to perform training for improving welding technology in a safe environment.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In welding work, it is important to grasp the state of the molten pool formed during welding and proceed with the work. In the conventional training of welding technology utilizing VR, although the virtually formed molten pool and weld marks are displayed based on physically calculated information, there has been a problem that they lack reality. An object of the present invention is to enable display of a realistic welding state in virtual welding using a virtual welding system.

Means for Solving the Problems

[0005] The virtual welding system according to the present invention The setting conditions of virtual welding, the pseudo torch operation information during virtual welding execution, prediction means for predicting the welding state in virtual welding using a prediction model for predicting the future welding state; simulation means for estimating the current welding state in virtual welding by physical calculation with reference to the welding state predicted by the prediction means; and display means for displaying an image of the welding state in virtual welding based on the estimation result by the simulation means. and the prediction model is a prediction model that estimates the welding state at time (t + 1) by machine learning using the information of a plurality of welding states up to time t and the feature amount data at time t as teacher data. It is characterized by this.

Effect of the Invention

[0006] According to the present invention, it becomes possible to display a realistic welding state in virtual welding using a virtual welding system.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Mode for Carrying Out the Invention

[0008] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, "on-site welding" refers to actual welding performed in the on-site environment, and "virtual welding" refers to virtual welding performed in a simulated virtual space (virtual reality environment).

[0009] FIG. 1 is a diagram showing an example of the hardware configuration of a virtual welding system 100 according to an embodiment of the present invention. The virtual welding system 100 in the present embodiment can execute virtual welding (virtual welding) in a simulated virtual space (virtual reality environment), and can be used, for example, as a training support system for performing training on welding techniques based on the movement of the torch and the posture of the operator during welding work.

[0010] The virtual welding system 100 includes a CPU 101, a main storage device 102, an auxiliary storage device 103, an output device 104, an HMD 105, a controller 106, a motion sensor 107, a camera 108, and a torch motion detection device 109. These CPU 101, main storage device 102, auxiliary storage device 103, output device 104, HMD 105, controller 106, motion sensor 107, camera 108, and torch motion detection device 109 are communicably connected.

[0011] The CPU (Central Processing Unit) 101 is a central processing unit that controls the virtual welding system 100. The main memory device 102 is a memory device that functions as a work area and a temporary storage location for data of the CPU 101. The main memory device 102 is implemented using, for example, RAM (Random Access Memory) or the like. The auxiliary storage device 103 is a storage device that stores various data, various programs, and the like. The auxiliary storage device 103 is implemented using, for example, ROM (Read Only Memory), a hard disk drive (HDD), a solid state drive (SSD), or the like. Each processing operation and the like in the virtual welding system 100 are realized by the CPU 101 executing processing based on, for example, a program stored in the auxiliary storage device 103. The output device 104 is a device that outputs various information such as the state of virtual welding (virtual welding) performed in the virtual space and the evaluation results of the welding operation, and is used to present various information to the user and the like.

[0012] The HMD (Head Mounted Display) 105 is a head-mounted display device that displays images (VR images) of the virtual space and the like. The HMD 105 is not limited to being a non-transmissive HMD that cannot confirm the external situation when worn, or a transmissive (see-through type) HMD that can confirm the external situation when necessary, such as by reflecting the external situation even when worn. The controller 106 is a controller associated with the HMD 105 and is used to perform operations and the like in the virtual space and the like displayed on the HMD 105. For example, in virtual welding (virtual welding) performed in the virtual space, a VR image showing the state of welding, such as various information and images during welding, is displayed on the HMD 105, and virtual welding is performed using the controller 106. At this time, the controller 106 is used, for example, to detect a pseudo torch operation.

[0013] The motion sensor 107 is a sensor for three-dimensionally measuring (motion-capturing) the user's posture (movement). The motion sensor 107 only needs to be able to measure the user's posture or the like, and is not limited to the method of motion capture. The motion sensor 107 may be, for example, of the motion capture suit type that incorporates an IMU sensor (inertial measurement unit), also called smart apparel, that is worn, or may be, for example, of the worn suit type with markers detected by a tracker. Further, the motion sensor 107 may be, for example, of the imaging type (camera type) that detects the distance, position, etc. to each part of the subject by imaging the subject.

[0014] The camera 108 is a camera that captures the welding phenomenon in actual welding (on-site welding) performed in the on-site environment. The camera 108 captures the welding state in actual welding (on-site welding) performed in the on-site environment, and for example, captures the molten pool, weld marks, etc. formed during the welding operation. The torch motion detection device 109 detects the torch motion in actual welding (on-site welding) performed in the on-site environment. The torch motion detection device 109 detects the movement of the torch, for example, based on markers attached to the torch. Note that the torch motion detection device 109 only needs to be able to detect the motion of the torch during welding, and is not limited to the detection method.

[0015] The virtual welding system 100 can acquire various data during welding in actual welding (on-site welding) performed in the on-site environment, for example, using the motion sensor 107, the camera 108, and the torch motion detection device 109. Further, the virtual welding system 100 can perform image display related to virtual welding in the virtual space, for example, using the HMD 105, and can acquire various data during welding in virtual welding performed in the virtual space, using the motion sensor 107 and the controller 106.

[0016] In addition, when data regarding the operator's posture during welding work is not acquired, the virtual welding system 100 does not necessarily need to include the motion sensor 107. Further, when the welding state prediction model described later is created using the photographed images of the welding state and the torch operation information, etc. in actual welding (on-site welding) performed in the on-site environment and acquired and recorded in advance, the virtual welding system 100 does not necessarily need to include the camera 108 and the torch operation detection device 109.

[0017] FIG. 2 is a diagram showing an example of the functional configuration of the virtual welding system 100 in the present embodiment. The virtual welding system 100 includes an acquisition unit 201, a storage unit 202, a display unit 203, an operation unit 204, an analysis unit 205, an evaluation unit 206, a learning processing unit 207, a simulation unit 208, an image generation unit 209, a display control unit 210, and a control unit 211.

[0018] The acquisition unit 201 acquires various data used for each process and the like in the virtual welding system 100. For example, the acquisition unit 201 acquires data regarding the operation of the torch and the posture of the user (operator) in on-site welding, photographed images of the welding state, and the like. Further, for example, the acquisition unit 201 acquires various data regarding the operation of the torch and the posture of the user (operator) in virtual welding.

[0019] The storage unit 202 stores various data acquired by the acquisition unit 201 and various models generated by the learning processing unit 207 including the welding state prediction model. Further, the storage unit 202 may store data such as the processing results obtained by the analysis unit 205 and the evaluation unit 206. The display unit 203 displays various information and the like presented to the user and the like according to the control by the display control unit 209. For example, the display unit 203 displays various information when performing virtual welding, VR images showing the state of welding, or a result screen showing the evaluation result of the welding work by the evaluation unit 206. The operation unit 204 is used to perform operations and the like in the virtual space. The operation unit 204 is used, for example, when performing various operations in virtual welding.

[0020] Based on the various data acquired by the acquisition unit 201, the analysis unit 205 analyzes the operation of the torch in welding, the posture of the user (operator), and the welding phenomenon, etc. For example, the analysis unit 205 analyzes the operation (speed and position / angle) of the torch during welding based on the data obtained from the torch operation detection device 109 in on-site welding and the data obtained from the controller 106 in virtual welding. Also, the analysis unit 205 analyzes the posture (movement) of the user (operator) during welding based on the data obtained from the motion sensor 107 in on-site welding or virtual welding.

[0021] Based on the movement of the torch, the posture of the user (operator), and the welding state during the virtual welding operation, the evaluation unit 206 evaluates the welding operation. For example, the evaluation unit 206 compares the movement of the torch during the virtual welding operation with the pre-generated torch model regarding the movement of the torch of a skilled worker to evaluate the movement of the torch of the user (operator) during virtual welding. Also, the evaluation unit 206 compares the posture of the user (operator) during the virtual welding operation with the pre-generated welding posture model regarding the posture of a skilled worker during work to evaluate the posture of the user (operator) during virtual welding. Further, the evaluation unit 206 compares the simulated welding state based on the movement of the torch, etc. during the virtual welding operation with the welding state model (to be described later) created in advance regarding the welding state in the welding by a skilled worker to evaluate the welding state during virtual welding.

[0022] The learning processing unit 207 performs machine learning (e.g., deep learning) with the photographed image of the welding state taken in on-site welding and the feature quantity data regarding the welding operation as inputs, and generates a prediction model of the welding state for predicting the future welding state. Also, for example, the learning processing unit 207 performs machine learning with various data obtained in on-site welding or virtual welding operations by skilled workers as inputs, and generates models for evaluating welding operations such as a torch model regarding the movement of the torch, a welding posture model regarding the posture during welding operations, and a welding state model (e.g., a molten pool model regarding the molten pool) regarding the welding state.

[0023] The simulation unit 208 performs a simulation of the welding state in virtual welding. The simulation unit 208 simulates the welding state during the welding operation in virtual welding based on various data obtained in virtual welding. The simulation unit 208 combines a simulation based on a physical calculation model and a prediction model of the welding state generated by machine learning by the learning processing unit 207 to perform a simulation of the welding state in virtual welding. The simulation unit 208 is an example of prediction means and simulation means.

[0024] The image generation unit 209 generates an image to be displayed in the virtual welding system 100. The image generation unit 209 generates, for example, a VR image to be displayed when performing virtual welding, an image showing the evaluation result of the welding operation, and the like. The display control unit 210 performs control related to the display on the display unit 203. The display control unit 210 performs control, for example, of the display position, size, etc. of the image generated by the image generation unit 209 on the display unit 203. The control unit 211 controls each functional unit included in the virtual welding system 100.

[0025] FIG. 3 is a diagram for explaining virtual welding performed in the virtual space in the present embodiment. FIG. 3(a) shows an example of the VR image 301 displayed on the HMD 105 during the execution of virtual welding. In the VR image 301, as elements related to the welding operation itself, a torch (welding rod) 302, a weld bead 303, and a molten pool 304 formed by the generated arc, which are modeled after actual welding (on-site welding) performed in the on-site environment, are displayed. In addition, the VR image 301 can also display an evaluation 306 of the welding operation (for example, four-level display, etc.), operation information 307 of the torch (welding rod) 302, posture information 308 during the welding operation, and the like. Further, it is also possible to output sounds during welding in virtual welding, sounds that mark the rhythm, and the like together with the VR image 301. Also, during virtual welding, cautions (warning matters) regarding the operation and posture of the torch may be displayed in an image or output as sound.

[0026] Settings such as the display (output) of each element in virtual welding can be set to determine whether to perform the display (output) via a setting screen as shown in, for example, Fig. 3(b). In Fig. 3(b), for the elements to be displayed in virtual welding, it is possible to set whether to display the arc light, spatter, fume, and molten pool respectively. As guides to assist the welding operation, it is possible to set whether to display (output) the ruler, metronome (rhythm sound), molten pool frame, and movement angle / operation angle of the torch respectively, and an example of setting whether to display real-time evaluation is shown. Note that the setting items shown in Fig. 3(b) are just an example and are not limited to this.

[0027] The virtual welding system 100 in this embodiment executes a simulation based on a physical calculation model using a welding state prediction model for predicting the future welding state in virtual welding, and performs the display of the VR image in virtual welding based on the simulation result. Here, the welding state includes both the molten pool formed during welding and / or both the molten pool and the weld bead. The virtual welding system 100 predicts the welding state in virtual welding using the welding state prediction model, and by referring to the prediction result and executing the simulation based on the physical calculation model, estimates the welding state regarding the molten pool, weld bead, etc. virtually formed in virtual welding. The virtual welding system 100, for example, simulates the melting phenomenon of the welded base material, the molten pool, fluid, rust, etc., which are phenomena occurring to the welded base material after the arc is generated in welding, and displays an image according to the simulation result.

[0028] The operation of the virtual welding system 100 related to image display in virtual welding will be described below. Note that below, mainly the display of the welding state regarding the molten pool, weld bead, etc. among the image displays in virtual welding will be described. For the image display of a pseudo-torch or the like operated by the user (operator) in virtual welding, the image display can be performed by a general method, and the description thereof will be omitted.

[0029] First, a welding state prediction model for predicting the welding state in virtual welding will be described. FIG. 4 is a diagram for explaining the generation of the welding state prediction model. The learning processing unit 207 performs machine learning with the time-series captured images 401 of the welding state captured by the camera 108 during on-site welding and the feature amount data 402 indicating the feature amounts in the welding operation as inputs, and generates a welding state prediction model 403.

[0030] Machine learning is performed, for example, using a set of a plurality of (for example, n sheets where n is an integer of 2 or more) captured images of the welding state up to time t (captured images at times t, (t - 1), ···, (t - n - 1)) 401, the feature amount data 402 at time t, and the captured image 401 of the welding state as one piece of teacher data. The learning processing unit 207 performs machine learning (for example, deep learning using a DNN: Deep Neural Network) using a plurality of pieces of teacher data with different times t, and generates a welding state prediction model 403. In this way, a welding state prediction model 403 for estimating the welding state at time (t + 1) is generated based on the n captured images of the welding state up to time t and the feature amount data at time t. The generated welding state prediction model 403 is stored in, for example, the storage unit 202. In the initial stage of the welding operation when there are no captured images of the welding state of a predetermined number, the captured images of the welding state from the start of the welding operation may be used in the machine learning.

[0031] Next, the display operation by the virtual welding system 100 during the execution of virtual welding will be described. FIG. 5 is a flowchart for explaining an example of the display process of the welding state in virtual welding by the virtual welding system 100. Note that the data obtained in each step of the process shown in FIG. 5 and the process results such as the simulation results are appropriately stored in the storage unit 202.

[0032] When an instruction to start the execution of virtual welding by displaying a VR image on the HMD105 is given, the virtual welding system 100 estimates and determines the output in virtual welding based on the setting conditions of virtual welding and the like. The setting conditions of virtual welding and the like include the diameter of the wire as the welding material, the wire feed speed, the maximum value of the wire length, the minimum value, standard value, and maximum value of the arc length, the standard value of the voltage, the minimum value, standard value, and maximum value of the current, and the like. Further, the output in virtual welding includes the estimated values of the wire length, arc length, current, voltage, and penetration amount, respectively.

[0033] After estimating and determining the output in virtual welding, virtual welding is started. In step S501, the virtual welding system 100 acquires various data related to the ongoing virtual welding. In this step S501, the acquisition unit 201 of the virtual welding system 100 acquires data for acquiring information regarding the pseudo torch operation (speed and position / angle) in virtual welding detected by, for example, the controller 106 or the like. For example, by analyzing the data acquired by the acquisition unit 201 in the analysis unit 205, torch operation data indicating the torch operation (speed and position / angle) in virtual welding can be obtained.

[0034] Next, in step S502, the virtual welding system 100 performs a simulation of the welding state in virtual welding based on the output in virtual welding estimated and determined at the start of virtual welding and the various data acquired in step S501. In the simulation of the welding state in step S502, the simulation unit 208 of the virtual welding system 100 estimates the molten pool, the size and shape of the weld bead, etc. according to the prediction model of the welding state based on the characteristic quantities in the welding operation and the past welding state, and performs a simulation of the welding state based on the physical calculation model using the estimated molten pool, the size and shape of the weld bead, etc.

[0035] Next, in step S503, the virtual welding system 100 displays the welding state in virtual welding on the HMD 105 based on the simulation results in step S502 and the like. At this time, the virtual welding system 100 not only displays the welding state such as the molten pool and the weld mark, but also displays the arc light, spatter (spark), fume, etc. in virtual welding, and outputs the welding sound. In this step S503, for example, the image generation unit 209 of the virtual welding system 100 generates a VR image showing the state of welding in virtual welding, in which the display color according to the temperature is set based on the simulation results in step S502 and the like, and the display control unit 210 causes the VR image generated by the image generation unit 209 to be displayed on the display unit 203.

[0036] Next, in step S504, the virtual welding system 100 determines whether to end the virtual welding. The virtual welding system 100 determines whether to end the virtual welding based on, for example, the operation of the user (operator) on the controller 106. If it is determined to end the virtual welding, the process shown in FIG. 5 is ended, and if it is determined not to end the virtual welding, that is, to continue the virtual welding, the process returns to step S501.

[0037] FIG. 6 is a flowchart for explaining the flow of the simulation of the welding state in step S502 of FIG. 5. In the simulation of the welding state shown in FIG. 6, the simulation based on the physical calculation model divides the simulation target area into a plurality of meshes, for example, as shown in FIG. 7, and is performed using known physical calculations (finite volume method, finite element method, boundary tracking method, etc.). FIG. 7 shows an example in which the target area shown in FIG. 7(a) is divided into 10 as shown in FIG. 7(b), but the size and number of divisions of the mesh are arbitrary and may be appropriately determined according to the accuracy of the simulation results, the calculation load (computation cost), etc.

[0038] When starting the simulation of the welding state, at step S601, the simulation unit 208 determines the melting position due to the generation of an arc in the virtual welding. The simulation unit 208 calculates the start position and end position (the position where the arc drops) of the generated arc based on the estimated arc length in the virtual welding and the information regarding the pseudo torch operation (speed and position / angle) in the virtual welding obtained in step S501 of FIG. 5, thereby determining the melting position in the current simulation.

[0039] At step S602, the simulation unit 208 acquires the prediction result of the welding state from the storage unit 202. The prediction result of the welding state acquired at step S602 was predicted using the welding state prediction model at step S608 described later in the previous simulation and stored in the storage unit 202. In the first simulation, for example, the welding state obtained by physical calculation may be used as the prediction result, or the welding state based on the photographed image in on-site welding may be used as the prediction result.

[0040] At step S603, the simulation unit 208 stacks beads on the mesh-divided target area. The simulation unit 208 stacks beads according to the penetration amount and the like within a predetermined range based on the end position (the position where the arc drops) of the arc obtained at step S601 in the mesh-divided target area. Here, for example, a cone model in which the distribution of bead stacking is in the shape of a cone is used to calculate the bead stacking. FIG. 8 shows an example of the processing result at step S603. FIG. 8 shows an example in which bead stacking is performed as shown in FIG. 8(b) using a cone model in the mesh-divided target area shown in FIG. 8(a). Note that the predetermined range for bead stacking may be determined based on the end position (the position where the arc drops) of the arc obtained at step S601 and the prediction result of the welding state acquired at step S602.

[0041] In step S604, the simulation unit 208 calculates the heat input (input of temperature parameters) for the mesh-divided target region. The simulation unit 208 applies a predetermined heat input model to calculate the heat input for the range of the molten pool having the shape and size indicated by the predicted result of the welding state acquired in step S602 and formed at the end position of the arc (the position where the arc drops) obtained in step S601. At this time, if the range of the molten pool indicated by the predicted result of the welding state is different from the heat input range in the heat input model, the heat input model is applied only within the range of the molten pool indicated by the predicted result of the welding state. For example, if the range of the molten pool indicated by the predicted result of the welding state is the region 901 shown in FIG. 9(a) and the heat input range in the heat input model is the region 902 shown in FIG. 9(b), the heat input model is applied only to the region 901 as shown in FIG. 9(c) excluding the range outside the molten pool to calculate the heat input.

[0042] In step S605, the simulation unit 208 calculates the fluid for the mesh-divided target region. The simulation unit 208 moves the mesh indicating a temperature equal to or higher than a predetermined temperature as a fluid based on physical calculations, and deforms each mesh due to the flow phenomenon. The simulation unit 208 performs the fluid calculation, for example, by approximating the diffusion equation by the central difference method and solving it by the explicit method on a two-dimensional mesh. In this method, basically, the value changes so as to approach the average value of the values of the upper, lower, left, and right meshes. Therefore, for the components in the adjacent directions, the same amount of increase and decrease with opposite signs is performed, and a result corresponding to the movement of the volume is obtained.

[0043] In step S606, the simulation unit 208 calculates the heat dissipation (heat dissipation of temperature parameters) for the mesh-divided target region. The simulation unit 208 calculates the heat dissipation for each mesh by applying a predetermined heat dissipation model.

[0044] In step S607, the simulation unit 208 calculates rust for the mesh-divided target area. The simulation unit 208 defines a rust generation range with the outer edge of the range where bead stacking was performed in step S603 as the start position of rust generation and a predetermined edge far from the range where bead stacking was performed as the end position of rust generation, and calculates the rust to be generated.

[0045] As described above, the simulation by physical calculation is performed by the simulation unit 208, and the welding state in virtual welding at a certain time is estimated. The simulation result (estimation result) of the welding state obtained by the above processing is stored in, for example, the storage unit 202.

[0046] Subsequently, in step S608, the simulation unit 208 predicts the welding state to be used in the next simulation using the welding state prediction model. The simulation unit 208 predicts the next welding state using the welding state prediction model based on the time-series welding state images in the virtual welding so far and the feature amount data indicating the feature amounts in the virtual welding work, based on the simulation result obtained as described above. When the processing of step S608 is completed, the simulation of the welding state shown in FIG. 6 is terminated, and the process returns to the flowchart processing shown in FIG. 5.

[0047] FIG. 10 is a diagram for explaining the prediction of the welding state, which is executed in step S608. FIG. 10 shows an example of predicting the welding state at time (t + 1) based on data up to time t. As shown in FIG. 10, the simulation unit 208 receives an image 1001 of the welding state based on simulation results of a plurality (for example, n) of welding states up to time t (simulation results of the welding states at times t, (t - 1),..., (t - n - 1)), and feature amount data 1002 at time t. The simulation unit 208 uses the welding state prediction model 403 generated as shown in FIG. 4 described above based on the welding state image 1001 up to time t and the feature amount data 1002 at time t, and obtains a prediction result of the welding state at time (t + 1).

[0048] The virtual welding system 100 in the present embodiment predicts the welding state in virtual welding using a welding state prediction model, executes a simulation based on a physical calculation model with reference to the prediction result of the welding state, and displays the welding state by a VR image in virtual welding based on the simulation result. Thereby, in virtual welding, it is possible to perform a realistic display of the welding state as compared with the display of the welding state based on physically calculated information. Further, by predicting and processing the welding state in virtual welding using the welding state prediction model, the welding state can be displayed and reproduced in real time according to the operation of the user (operator) of the virtual welding.

[0049] Here, as the feature amounts in the welding operation used for generating the welding state prediction model and predicting the welding state using the prediction model, for example, there are the welding speed (torch speed), the distance between the welding base material and the torch (tip), and the operation angle and movement angle of the torch during the welding operation. These feature amounts can be obtained based on data indicating the operation of the torch during the welding operation, which is detected by, for example, the torch operation detection device 109 or the controller 106. Note that the feature amounts in the welding operation are not limited to these, and other feature amounts may be used.

[0050] In the above-described embodiments, the heat input in the simulation of the welding state is calculated by applying a predetermined heat input model. As an example of the heat input model, there is a conical heat input model in which heat is added based on the shape distribution of a cone 1101 as shown in Fig. 11(a). When the heat input model to be applied is the conical heat input model and the range of the molten pool indicated by the prediction result of the welding state is different from the heat input range in the heat input model, for example, as shown in Fig. 11(b), the heat input model 1103 may be applied only within the range 1102 of the molten pool indicated by the prediction result of the welding state in the same manner as the above description. As another example of the heat input model, there is also a cylindrical heat input model in which heat is added based on the shape distribution of a cylinder having the same bottom surface as the cone 1101 shown in Fig. 11(a).

[0051] Also, in the simulation of the welding state, instead of simulating each phenomenon by physical calculation, a surrogate model that predicts the phenomenon with a machine-learned model may be applied. By performing prediction with the surrogate model, the calculation load can be reduced and the calculation cost can be reduced.

[0052] Note that each of the above embodiments merely shows an example of the implementation of the present invention, and the technical scope of the present invention should not be construed in a limited manner by these. That is, the present invention can be implemented in various forms without departing from its technical idea or its main features.

Explanation of Reference Numerals

[0053] 100 Virtual Welding System 201 Acquisition Unit 202 Storage Unit 203 Display Unit 204 Operation Unit 205 Analysis Unit 206 Evaluation Unit 207 Learning Processing Unit 208 Simulation Unit 209 Image Generation Unit 210 Display Control Unit 211 Control Unit

Claims

1. Prediction means for predicting the welding state in virtual welding using the set conditions of virtual welding, the pseudo torch operation information during virtual welding execution, and a prediction model for predicting the future welding state; Simulation means for estimating the current welding state in virtual welding by physical calculation with reference to the welding state predicted by the prediction means; Display means for displaying the welding state in virtual welding as an image based on the estimation result by the simulation means, The virtual welding system is characterized in that the prediction model is a prediction model that estimates the welding state at time (t + 1) by machine learning using information on a plurality of welding states up to time t and feature quantity data at time t as teacher data.

2. The virtual welding system according to claim 1, wherein the welding state includes the state of a molten pool or a weld mark formed in welding.

3. The virtual welding system according to claim 1 or 2, further comprising image generation means for generating an image showing the welding state in virtual welding, in which a display color according to temperature is set, based on the estimation result by the simulation means.

4. The virtual welding system according to claim 1 or 2, wherein the prediction means predicts at least one of the shape and size related to the welding state.

5. The virtual welding system according to claim 1 or 2, wherein the prediction means predicts the future welding state based on the prediction model, the feature quantity in the welding operation, and the welding state in the virtual welding already predicted.

6. The virtual welding system according to claim 5, wherein the feature quantity includes at least one of the welding speed, the distance between the welding base material and the torch, and the operation angle and movement angle of the torch in virtual welding.

7. The virtual welding system according to claim 1 or 2, further comprising learning means for generating the prediction model by machine learning using a photographed image of the welding state taken in actual welding performed in the field environment and the feature quantity in the actual welding operation as inputs.

8. A prediction step of predicting the welding state in virtual welding using the set conditions of virtual welding, the pseudo torch operation information during virtual welding execution, and a prediction model for predicting the future welding state; A simulation step of estimating the current welding state in virtual welding by physical calculation with reference to the welding state predicted in the prediction step; A display step of displaying an image of the welding state in virtual welding on a display means based on the estimation result in the simulation step; and The prediction model is a prediction model that estimates the welding state at time (t + 1) by machine learning using information on a plurality of welding states up to time t and feature amount data at time t as teacher data. A processing method of a virtual welding system, characterized in that.

9. On the computer of the virtual welding system, A prediction step of predicting the welding state in virtual welding using the setting conditions of virtual welding, the operation information of a pseudo torch during virtual welding execution, and a prediction model for predicting the future welding state; A simulation step of estimating the current welding state in virtual welding by physical calculation with reference to the welding state predicted in the prediction step; Based on the estimation result in the simulation step, a display step of displaying an image of the welding state in virtual welding on a display means is executed. The prediction model is a program that is a prediction model that estimates the welding state at time (t + 1) by machine learning using information on a plurality of welding states up to time t and feature amount data at time t as teacher data.

Images