Machine learning device, stacking modeling system, machine learning method for welding conditions, method for determining welding conditions, and storage medium

Abstract

A machine learning device performs machine learning on welding conditions when shaping a laminated object by depositing a filler material and laminating weld beads. The machine learning device includes a learning processing mechanism that performs learning processing for generating a learned model using the welding conditions of the weld beads and the block pattern formed by the weld beads as input data and the shape data of the weld beads as output data.

Description

Technical Field

[0001] The present invention relates to a machine learning device, a stacking modeling system, a machine learning method for welding conditions, a method for determining welding conditions, and a storage medium storing a program. More specifically, it relates to a technique for determining modeling conditions when modeling a stacked modeled object using stacked weld beads. Background Art

[0002] Conventionally, stacked weld beads are used to create stacked objects. To improve the accuracy of stacked welds, various welding conditions must be considered and controlled. These welding conditions present numerous combinations, making manual extraction of appropriate welding conditions extremely complex and cumbersome.

[0003] Regarding the above-mentioned situation, Patent Document 1, for example, discloses a learning device for automatically determining optimal welding conditions in a welding device without instruction from a skilled operator. In this case, the learning information includes the appearance of the weld bead, its height, width, and penetration.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2017-30014 Summary of the Invention

[0007] Problems to be solved by the invention

[0008] As described above, when adjusting welding conditions during stacking, many combinations of conditions must be considered to understand the changing trends in weld bead shape (width, height, etc.), making it difficult to determine the appropriate combination. Furthermore, when forming a weld bead, the shape of the weld bead can vary even under the same welding conditions due to the relationship between the weld bead and the surrounding weld bead that has already been formed. For example, creating a database that specifies the combinations of welding conditions and weld bead configuration patterns is conceivable. However, there are a large number of weld bead configuration patterns, making it complex and cumbersome to create a database that corresponds to the welding conditions. Patent Document 1 mentioned above does not consider the changing trends in weld bead shape that occur with the configuration pattern, and this also leaves room for improvement.

[0009] In view of the above-mentioned problems, an object of the present invention is to determine more appropriate welding conditions when forming a laminated object.

[0010] Solutions to Problems

[0011] In order to solve the above-mentioned problems, the present invention has the following configuration.

[0012] (1) A machine learning device that performs machine learning on welding conditions when a laminated object is formed by depositing a filler material and laminating weld beads,

[0013] The machine learning device is characterized in that

[0014] A learning processing unit is provided for performing a learning process for generating a learned model using welding conditions of a weld bead and a block pattern composed of the weld bead as input data and shape data of the weld bead as output data.

[0015] Furthermore, another embodiment of the present invention has the following configuration.

[0016] (2) A laminate molding system that forms a laminated object by depositing a filler material and laminating weld beads,

[0017] The stacking molding system is characterized by having:

[0018] a production mechanism for producing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object;

[0019] a determination unit that determines welding conditions when forming weld passes corresponding to the plurality of pass data generated by the generation unit;

[0020] a determining mechanism for determining a block pattern formed by the weld bead based on the configuration when the weld bead is formed;

[0021] a deriving unit that inputs the welding conditions corresponding to the pass data determined by the determining unit and the block pattern determined by the determining unit into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0022] An adjusting means adjusts the welding conditions so that a difference between the shape data derived by the deriving means and the shape data represented by the pass data produced by the producing means does not exceed a predetermined threshold value.

[0023] Furthermore, another embodiment of the present invention has the following configuration.

[0024] (3) A machine learning method for determining welding conditions when a laminated object is formed by depositing a filler material and laminating weld beads,

[0025] The machine learning method is characterized in that

[0026] The method includes a learning processing step of performing a learning process for generating a learned model using welding conditions of a weld bead and a block pattern composed of the weld bead as input data and shape data of the weld bead as output data.

[0027] Furthermore, another embodiment of the present invention has the following configuration.

[0028] (4) A method for determining welding conditions in a stacking molding system for molding a stacked object by depositing a filler material and stacking weld beads,

[0029] The method for determining welding conditions is characterized by comprising:

[0030] a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object;

[0031] a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step;

[0032] a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed;

[0033] a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0034] The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

[0035] Furthermore, another embodiment of the present invention has the following configuration.

[0036] (5) A storage medium recording a program, wherein:

[0037] The program causes a computer to execute a learning processing step for generating a learned model having welding conditions of weld beads constituting a stacked shaped object and a block pattern constituted by the weld beads as input data and shape data of the weld beads as output data.

[0038] Furthermore, another embodiment of the present invention has the following configuration.

[0039] (6) A storage medium recording a program, wherein:

[0040] The program is used to make the computer execute the following steps:

[0041] a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld beads constituting the stacked shaped object based on design data of the stacked shaped object formed by depositing a filler material and stacking weld beads;

[0042] a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step;

[0043] a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed;

[0044] a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0045] The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

[0046] Effects of the Invention

[0047] According to the present invention, it is possible to determine more appropriate welding conditions when forming a laminated object. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a schematic structural diagram showing a structural example of a stacking molding system according to one embodiment of the present invention.

[0049] Figure 2 This is a conceptual diagram for explaining the shape data of a weld bead.

[0050] Figure 3 This is a conceptual diagram showing an example of a block pattern according to one embodiment of the present invention.

[0051] Figure 4 This is a schematic diagram for explaining the concept of learning in one embodiment of the present invention.

[0052] Figure 5 This is a schematic diagram for explaining the concept of obtaining weld bead elements from design data of a stacked structure according to one embodiment of the present invention.

[0053] Figure 6This is a schematic diagram for explaining the association relationship of block patterns according to one embodiment of the present invention.

[0054] Figure 7 This is a flowchart of a stacking condition determination process according to one embodiment of the present invention.

[0055] Figure 8 This is a flowchart of a process for generating learning data according to one embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following describes specific embodiments of the present invention with reference to the accompanying drawings. It should be noted that the embodiments described below are intended to illustrate one embodiment of the present invention and are not intended to limit the present invention. Furthermore, not all structures described in each embodiment are necessarily required to solve the problems of the present invention. In the accompanying drawings, identical components are assigned the same reference numerals to indicate correspondence.

[0057] <First embodiment>

[0058] Hereinafter, a first embodiment of the present invention will be described.

[0059] [System Structure]

[0060] Hereinafter, one embodiment of the present invention will be described in detail with reference to the accompanying drawings. Figure 1 This is a schematic structural diagram of a stacking molding system to which the present invention can be applied.

[0061] The laminated molding system 1 of the present embodiment is configured to include a laminated molding apparatus 100 and an information processing apparatus 200 for comprehensively controlling the laminated molding apparatus 100 .

[0062] The stacking molding apparatus 100 includes a welding robot 104 , a filler material supply unit 105 that supplies a filler material (welding wire) M to the welding torch 102 , a robot controller 106 that controls the welding robot 104 , and a power source 107 .

[0063] Welding robot 104 is a multi-jointed robot. A welding torch 102, mounted on a distal shaft, supports filler material M in a continuous supply manner. Welding torch 102 holds filler material M protruding from the distal end. The position and posture of welding torch 102 can be freely set three-dimensionally within the range of freedom of the robot arm of welding robot 104.

[0064] The welding torch 102 includes a shield nozzle (not shown) from which shielding gas is supplied. The shielding gas blocks the atmosphere, preventing oxidation and nitridation of the molten metal during welding, thereby suppressing welding defects. The arc welding method used in this embodiment can be any of consumable electrode methods such as covered arc welding and carbon dioxide gas arc welding, or non-consumable electrode methods such as TIG welding and plasma arc welding. The method is appropriately selected depending on the desired laminated object W.

[0065] A shape sensor 101 that can move in accordance with the movement of the welding torch 102 is provided near the welding torch 102. The shape sensor 101 detects the shape of the stacked object W formed on the base 103. In this embodiment, the height, position, width, etc. of the weld bead 108 (also referred to as "weld bead") constituting the stacked object W can be detected by the shape sensor 101. The information detected by the shape sensor 101 is sent to the information processing device 200. It should be noted that the structure of the shape sensor 101 is not particularly limited, and it can be a structure that detects shape by contact (contact sensor) or a structure that detects shape by laser or the like (non-contact sensor).

[0066] In welding robot 104, when the arc welding method is consumable electrode, a contact tip is placed inside the shielded nozzle, and filler material M, which supplies the melting current, is held in the contact tip. While holding the filler material M, welding torch 102 generates an arc from the tip of the filler material M under a shielding gas atmosphere. The filler material M is fed from a filler material supply unit 105 to welding torch 102 via a delivery mechanism (not shown) attached to a robot arm, etc. As the continuously fed filler material M melts and solidifies while welding torch 102 is moved, a linear weld bead 108, representing the molten solidified filler material M, is formed on base 103. By stacking weld beads 108, a stacked object W is formed.

[0067] It should be noted that the heat source for melting the filler material M is not limited to the aforementioned arc. For example, other heat sources such as a combined arc and laser heating method, a plasma heating method, an electron beam heating method, or a laser heating method may also be employed. When using an electron beam or laser for heating, the amount of heat can be more precisely controlled, the state of the weld bead 108 can be more appropriately maintained, and the quality of the laminated object W can be further improved.

[0068] Based on instructions from the information processing device 200, the robot controller 106 drives the welding robot 104 using a specified driver program to form a stacked object W on the base 103. Specifically, the welding robot 104 moves the welding torch 102 while melting the filler material M using an arc, according to instructions from the robot controller 106. The power supply 107 is a welding power source that supplies the power required for welding to the robot controller 106. The power supply 107 can operate in multiple control modes and can switch the power (current, voltage, etc.) supplied to the robot controller 106 according to the control mode. The filler material supply unit 105 controls the supply of filler material M to the welding torch 102 of the welding robot 104 and the feed speed based on instructions from the information processing device 200.

[0069] The information processing device 200 may be, for example, an information processing device such as a PC (Personal Computer). Figure 1 Each of the functions shown can also be implemented by a control unit (not shown) reading and executing a program for the functions of this embodiment stored in a storage unit (not shown). The storage unit may include a RAM (Random Access Memory) as a volatile storage area, a ROM (Read Only Memory) as a non-volatile storage area, an HDD (Hard Disk Drive), etc. In addition, a CPU (Central Processing Unit), a GPU (Graphical Processing Unit), or a GPGPU (General-Purpose Computing on Graphics Processing Units) may also be used as the control unit.

[0070] The information processing device 200 includes a shaping control unit 201, a power supply control unit 202, a feed control unit 203, a database management unit 204, a shape data acquisition unit 205, a learning data management unit 206, a learning processing unit 207, and a welding condition derivation unit 208. The shaping control unit 201 generates control signals for the robot controller 106 during shaping based on the design data (e.g., CAD / CAAM data) of the stacked object W to be shaped. These control signals include the movement trajectory of the welding torch 102 of the welding robot 104, welding conditions during the formation of the weld bead 108, and the feed rate of the filler material M by the filler material supply unit 105. The movement trajectory of the welding torch 102 is not limited to the trajectory of the welding torch 102 during the formation of the weld bead 108 on the base 103; for example, it also includes the trajectory of the welding torch 102 toward the starting position for forming the weld bead 108.

[0071] The power supply control unit 202 controls the power supply (control mode) from the power supply 107 to the robot controller 106. Depending on the control mode, the current, voltage, current waveform (pulse), and other characteristics may vary when forming weld beads of the same shape. Furthermore, the power supply control unit 202 obtains information on the current and voltage supplied to the robot controller 106 from the power supply 107 as appropriate.

[0072] The feed control unit 203 controls the feed speed and timing of the filler material M by the filler material supply unit 105. Feed control of the filler material M here includes not only forward feeding (forward feeding) but also return feeding (reverse feeding). The DB management unit 204 manages the database (DB) of this embodiment. Details of the DB of this embodiment will be described later. The shape data acquisition unit 205 acquires the shape data of the weld bead 108 formed on the base 103, detected by the shape sensor 101.

[0073] The learning data management unit 206 generates and manages learning data used in the learning process performed by the learning processing unit 207. The learning processing unit 207 performs learning using the learning data managed by the learning data management unit 206. Details of the learning data and learning process in this embodiment will be described later. Furthermore, the learning processing unit 207 manages the learned model obtained as a result of the learning process.

[0074] The welding condition derivation unit 208 uses the learned model generated by the learning processing unit 207 to derive the welding conditions used by the formation control unit 201 and notifies the formation control unit 201. The processing performed by the welding condition derivation unit 208 will be described later.

[0075] In this embodiment, if Figure 1 As shown in FIG, the structure in which the welding torch 102 is moved on the cylindrical base 103 to form the weld bead 108 and shape the stacked object W is described as an example. Figure 1 In this embodiment, the structure in which the base 103 forms the stacked object W on a cylindrical plane is shown, but the present invention is not limited to this. For example, the base 103 may be cylindrical with a weld bead 108 formed on the outer periphery of its side surface. Furthermore, the coordinate system in the design data of this embodiment is associated with the coordinate system on the base 103 used to form the stacked object W. The three axes of the coordinate system (X-axis, Y-axis, and Z-axis) are set so that a three-dimensional position is defined with an arbitrary position as the origin.

[0076] The stacking molding system 1 of the above configuration melts the filler material M while moving the welding torch 102 according to the movement trajectory of the welding torch 102 specified by the set molding data, driven by the welding robot 104, and supplies the melted filler material M onto the base 103. This forms a stacked molding W by arranging and stacking a plurality of linear weld beads 108 on the upper surface of the base 103.

[0077] [Control parameters during modeling]

[0078] When forming a laminated object W, various control parameters must be determined and adjusted depending on the operating state of the power supply 107, the inherent characteristics of the device, the structure of the laminated object W, and other factors. The following describes examples of control parameters considered during laminated forming and data items representing the shape of the resulting weld bead.

[0079] (Control Parameters)

[0080] Welding conditions (wire feed speed, welding speed, torch travel speed, etc.)

[0081] Type of welding power source, control curve for current and voltage

[0082] Start / end processing conditions (current and voltage application conditions when forming the start and end of the weld bead)

[0083] Welding wire used (wire type, wire quality, etc.), wire composition (physical properties such as viscosity and surface tension)

[0084] Temperature and time between passes

[0085] Temperature of adjacent weld beads in the vertical, horizontal, or oblique directions (hereinafter referred to as "adjacent weld beads")

[0086] Angle of the welding torch

[0087] ·Amount of filler material deposited

[0088] Target position (distance to adjacent welds, etc.)

[0089] Oscillation conditions (oscillation frequency, amplitude)

[0090] (Shape data)

[0091] Weld bead height, width, root angle, and volume

[0092] Lower layer shape

[0093] ·Surface shape of the laminated member and the size of the surface irregularities

[0094] The overall shape of the laminated component (overall height, width, volume, etc.)

[0095] The shape of the weld bead start / end

[0096] ·Presence and size of internal defects

[0097] The shape (depth and width) of the weld zone and heat-affected zone (HAZ)

[0098] In addition, each item shown above is an example, and a part of them or other items may be used.

[0099] Figure 2 This is a conceptual diagram for explaining the shape data of a weld bead. Figure 2 FIG. 1 shows a cross section of a state where a weld bead 108 is formed on the base 103, viewed from the direction of travel of the welding torch 102 during the formation. Figure 2 As shown, as the shape data of the weld bead 108 , information such as the height h, the width w, the angle α at the root, and surface irregularities can be used.

[0100] [database]

[0101] In this embodiment, a database showing the relationship between welding conditions and the shape information of a weld bead formed using the welding conditions is used. This database is managed by the DB management unit 204 and is defined in advance.

[0102] The database stores information about the shape of the weld bead formed when welding using predefined control parameters as welding conditions in a corresponding relationship. The welding condition items include the aforementioned control parameters. Furthermore, the weld bead shape information items also include the aforementioned shape data items. In this embodiment, the database may correspond to the welding conditions and shape data for when weld bead 108 is formed on base 103. Alternatively, the database may correspond to the welding conditions and shape data for when weld bead 108 is formed by stacking a predetermined shape on weld bead 108. Alternatively, both types of information may be included.

[0103] [Block pattern]

[0104] As described above, in this embodiment, a stacked object W is formed by stacking multiple weld beads. When focusing on a particular weld bead, multiple patterns can be defined based on the positional relationship with surrounding weld beads, the position on the base 103, and the like. In this embodiment, this pattern is referred to as a block pattern and will be described.

[0105] Figure 3This figure shows an example of the basic classification of block patterns in this embodiment, and is a schematic diagram showing the cross-sectional shape viewed from the direction of weld bead formation. For simplicity of explanation, the cross-section of the weld bead is simplified to a trapezoidal shape, etc. Block patterns are defined based on the number of weld beads (passes). In this embodiment, 10 types of block patterns consisting of one to five passes are used as examples for explanation.

[0106] A block pattern consisting of one pass is one type (block pattern a), and has no adjacent weld beads. Block patterns consisting of two passes are of three types (block patterns b to d). More specifically, block pattern b has a structure with two weld beads arranged in the width direction of the weld bead (1 layer × 2 rows). Block pattern c has a structure with two weld beads overlapping in the height direction, and the center positions of the weld beads in the width direction are aligned (2 layers × 1 row). Block pattern c has a structure with two weld beads overlapping in the height direction, and the center positions of the weld beads in the width direction are different (2 layers × 1 row).

[0107] There are three types of block patterns consisting of three passes (block patterns e to g). More specifically, block pattern e has a structure in which two weld beads are arranged in the width direction of the weld bead, and two weld beads overlap in the height direction on one side (2 layers × 1 row + 1 layer × 1 row). Block pattern f has a structure in which three weld beads overlap in the height direction, and the centers of the weld beads in the width direction are aligned (3 layers × 1 row). Block pattern g has a structure in which three weld beads overlap in the height direction, and the centers of the weld beads in the width direction are at different positions (3 layers × 1 row).

[0108] There are two types of block patterns consisting of four passes (block patterns h to i). More specifically, block pattern h is a structure in which two welds are arranged in the width direction of the weld and two welds are overlapped in the height direction on both sides (2 layers × 2 columns). Block pattern i is a structure in which three welds are overlapped in the height direction and one weld is adjacent in the width direction relative to the weld located in its uppermost layer (3 layers × 1 column + 1 column of adjacent welds adjacent to the uppermost layer). There is one type of block pattern consisting of five passes (block pattern j). More specifically, it is a structure in which three welds are overlapped in the height direction and two welds are arranged and adjacent in the width direction relative to the weld located in its uppermost layer (3 layers × 1 column + 2 columns of adjacent welds adjacent to the uppermost layer). It should be noted that the weld of the lowest layer of the stacked welds is not limited to the state formed on the base 103.

[0109] In this embodiment, a weld bead is associated with at least one of the above-mentioned block patterns. Figure 3 The block pattern classification shown is an example and is not limited thereto. For example, block patterns other than the above may be used depending on the shape and size of the laminated object W to be formed, the material of the filler M, and the like.

[0110] [Learning Process]

[0111] In this embodiment, a deep learning method based on a neural network in machine learning is used as a learning method, and training learning is used as an example for explanation. It should be noted that the more specific method (algorithm) of deep learning is not particularly limited, and for example, well-known methods such as convolutional neural networks (CNNs) can also be used. In addition, there are no particular restrictions on the type and number of layers that constitute the neural network.

[0112] Figure 4 This is a schematic diagram for explaining the concept of the learning process of this embodiment. First, learning data for learning is prepared. The learning data is prepared by converting the original data into a form that matches the learning. As the original data, welding conditions, the shape of the weld bead formed based on the welding conditions, and the block pattern of the weld bead are prepared. Here, as the weld bead shape, the height of the weld bead, which is one of its elements, is taken as an example for explanation. In this embodiment, as the learning data for learning, the welding conditions and block pattern in the original data are used as input data, and the shape of the weld bead is used as training data. A plurality of such learning data are prepared. The block pattern is set to use Figure 3 The structure is described.

[0113] In this embodiment, the learning process is performed using the above-mentioned learning data. When the input data prepared as learning data (here, welding conditions and block patterns) is input to the learning model, the shape data of the weld bead is output as output data relative to the input data. This output data corresponds to the weld bead shape (height). Next, using this output data and the training data prepared as learning data (here, weld bead shape (height)), an error is derived using a loss function. In addition, each parameter in the learning model is adjusted in such a way as to reduce the error. For example, the error back propagation method can also be used to adjust the parameters. In this way, a learned model is generated by repeatedly learning using multiple learning data. The learned model is updated each time the learning process is executed, so the parameters constituting the learned model change according to the timing of use, and the output result relative to the input data is also different.

[0114] It should be noted that the learning process does not necessarily need to be performed by the information processing device 200. For example, the information processing device 200 may provide learning data to a learning server (not shown) provided outside the information processing device 200, and the learning process may be performed on the server side. In addition, the server may provide the learned model to the information processing device 200 as needed. Such a learning server may be located on a network such as the Internet (not shown), and the server and the information processing device 200 may be communicatively connected. That is, the information processing device 200 may act as a machine learning device, or an external device may act as a machine learning device. In either case, the information processing device 200 can obtain the learned model obtained by the learning process and use it when shaping the stacked shape object W.

[0115] Figure 5 This is a diagram for explaining the concept of extracting weld bead elements from design data in the process of this embodiment. Design data is data representing the design shape of the stacked object W. The stacked object W is formed by stacking in a predetermined stacking direction based on the design data and forming weld beads. Figure 5 In the example of FIG, the case where weld beads are stacked in the direction indicated by the arrows will be described. It should be noted that the stacking direction can be arbitrarily set according to the shape of the stacked shaped object W and the like.

[0116] First, the stacking direction is specified in the design data, and the material is divided (sliced) into one or more layers along a direction perpendicular to this stacking direction. This determines one or more slice data items. The number of slice data items (layers) varies depending on the size, shape, and layer thickness of the stacked object W represented by the design data. Each slice data item is further divided into weld bead elements corresponding to a single pass during weld bead formation.

[0117] Figure 6 This is a diagram for explaining a block pattern that corresponds to the weld bead 600 of interest. Here, the block pattern that corresponds to the weld bead 600 at the time when the weld bead 600 is formed is used as an example for explanation. Each weld bead is formed sequentially from the bottom layer, and other weld beads are already formed around the weld bead 600. In this case, the weld bead 600 can be Figure 3 Among the block patterns shown, a block pattern h of four passes (dashed line 601 ), a block pattern of three passes (dashed line 602 ), and a block pattern c of two passes (dashed line 603 ) establish a corresponding relationship.

[0118] [Processing Flow]

[0119] (Determination of welding conditions)

[0120] Figure 7This is a flowchart of the welding condition determination process of this embodiment. This process is executed and controlled by the information processing device 200. For example, the processing unit such as the CPU and GPU of the information processing device 200 can read and execute the data from the storage unit (not shown) to achieve the desired result. Figure 1 The program of each part shown is thus realized. In addition, before starting this processing flow, the above-mentioned learning process is performed to generate a learned model. In addition, this process is executed as the formation of the stacked object W begins.

[0121] In S701, the information processing device 200 obtains design data for a stacked object W. This design data specifies the shape of the stacked object W and is created based on user instructions. The design data can be input from an external device (not shown) that is communicatively connected, or created on the information processing device 200 using a predetermined application (not shown).

[0122] In S702, the information processing apparatus 200 generates one or more slice data based on the design data acquired in S701. Figure 5 As shown, by dividing the set data in a direction perpendicular to the predetermined stacking direction, one or more slice data are generated. The number of layers, layer thickness, and other division conditions used during division are not particularly limited, but multiple settings can be selected based on the functionality of the stacking molding system 1. The division conditions, such as layer thickness, used when generating slice data are stored together with the generated slice data in a storage unit (not shown).

[0123] In S703, the information processing device 200 generates a plurality of pass data according to the one or more slice data generated in S702. Figure 5 As shown, by dividing a single slice data, multiple pass data corresponding to one pass of the weld bead are generated. The division conditions, such as the number of passes and the width corresponding to one pass, are not particularly limited, but multiple settings can be selected based on the functions of the stacking molding system 1. The division conditions, such as the number of passes and width, when generating the pass data are stored together with the generated pass data in a storage unit (not shown). In addition to shape data representing the shape of the weld bead, the pass data may also include information such as the movement trajectory of the welding torch 102. The shape data generated here corresponds to the design value.

[0124] In S704, the information processing device 200 focuses on unprocessed pass data (hereinafter referred to as "focused pass data") from the plurality of pass data generated in S703. The focus order here may be, for example, the order in which weld beads corresponding to the pass data are formed.

[0125] In S705, the information processing device 200 refers to the DB and determines the welding conditions for forming the weld bead having the shape indicated by the target pass data. As described above, the DB associates the welding conditions with the weld bead shape data, and the welding conditions can be determined by specifying the shape data.

[0126] At S706, information processing device 200 determines one or more block patterns corresponding to the configuration of the target pass data. Depending on the order in which the weld beads corresponding to the pass data are actually formed, adjacent weld beads formed may differ. Therefore, one or more block patterns corresponding to the target pass data are determined based on the order in which the weld beads are formed.

[0127] In S707, the information processing device 200 selects a block pattern from the block patterns determined in S706 that has not yet been processed. The selection method is not particularly limited, but for example, a configuration may be employed in which a priority is pre-set for each block pattern and selection is performed based on that priority. More specifically, a configuration may be employed in which block patterns are selected in order of the number of passes that constitute the block pattern.

[0128] In S708, the information processing device 200 inputs the welding conditions determined in S705 and the block pattern selected in S707 into the already generated learned model, thereby deriving shape data corresponding to the target pass data as output data. As described above, the shape data output here corresponds to the height of the weld bead corresponding to the target pass data and serves as a predicted value when the weld bead is formed using the welding conditions determined in S705.

[0129] In S709, the information processing device 200 compares the weld bead shape (design value) represented by the target pass data with the weld bead shape (predicted value) derived in S708 to determine whether the difference (|design value - predicted value|; it should be noted that |X| represents the absolute value of X) is greater than a predetermined threshold. This threshold is predefined and stored and managed in a storage unit (not shown). If the difference is greater than the threshold (YES in S709), the information processing device 200 proceeds to S710. On the other hand, if the difference is less than the threshold (NO in S709), the information processing device 200 proceeds to S711. If the shape data includes multiple items such as height and width, the differences for each item are derived. Furthermore, if multiple items are used for determination in the shape data, a YES determination can be made if all items are greater than the threshold when compared with the threshold. In this case, a threshold is set for each item.

[0130] In S710, the information processing device 200 determines whether the processes S707 to S709 have been performed on all the block patterns identified in S706. If there are any unprocessed block patterns (YES in S710), the information processing device 200 returns to S707 and repeats the subsequent processes. On the other hand, if there are no unprocessed block patterns (NO in S710), the information processing device 200 returns to S702 and repeats the subsequent processes. In this case, the segmentation conditions used in the segmentation process for the slice data (S702) and the segmentation process for the pass data (S703) are set to be different from the segmentation conditions used in the previous process. In other words, the segmentation process is performed again so that the design value of the weld bead shape is changed. As described above, the information processing device 200 of this embodiment can set multiple segmentation conditions, and thus selects an unused segmentation condition from among them. The method for changing the segmentation conditions is not particularly limited. For example, a configuration may be employed in which the segmentation conditions for the slice data segmentation process are changed first, and the segmentation conditions for the pass data segmentation process are changed based on the result, or vice versa.

[0131] In S711 , the information processing apparatus 200 determines the currently determined welding conditions as the welding conditions based on the pass data.

[0132] In S712, the information processing device 200 determines whether processing of all pass data is complete. If processing of all pass data is complete (YES in S712), this process flow ends. On the other hand, if unprocessed pass data exists (NO in S712), the information processing device 200 returns to S704 and repeats the subsequent processing.

[0133] In the above flowchart, if the result of the determination process in S710 is negative, the process returns to the division process in S702. However, the process is not limited to this structure, and the process may also return to the division process in S703.

[0134] In addition, in the above flowchart, the welding conditions are determined based on the design data of the stacked object W being aggregated and divided into pass data, but the present invention is not limited to this. For example, the design data of the stacked object W may be divided into a plurality of parts, and the welding conditions may be determined based on the plurality of parts. Figure 7 Processing as shown.

[0135] Furthermore, the above flowchart illustrates an example of adjusting welding conditions by re-dividing the pass data corresponding to the weld bead, but the present invention is not limited to this. For example, adjustment can also be performed by changing the order in which the pass data are generated and thereby changing the block pattern corresponding to the target pass data.

[0136] (Data generation process for learning)

[0137] Figure 8 This is a flowchart of the process of generating learning data for learning processing in this embodiment. This process is executed and controlled by the information processing device 200. For example, the processing unit such as the CPU of the information processing device 200 can read and execute the data from the storage unit (not shown) to achieve the purpose. Figure 1 This is achieved by the program for each part shown. In this embodiment, the description assumes that, in parallel with the shaping operation, the shape sensor 101 is used to obtain the shape data of the weld bead formed when shaping the stacked object W. This processing flow is initiated upon receiving an instruction to start shaping the stacked object W. It should be noted that whether or not to generate learning data can be determined based on an instruction from the user of the stacked object shaping system 1, or can be configured so that the user can select it.

[0138] In S801 , the information processing device 200 selects unprocessed pass data from a plurality of pass data for forming the laminated object W in accordance with a predetermined forming order, and obtains welding conditions associated with the pass data.

[0139] In S802 , the information processing device 200 causes the welding robot 104 to form a weld bead based on the welding conditions acquired in S801 .

[0140] In S803, information processing device 200 uses shape sensor 101 to measure the shape of the weld bead formed in S802 and obtains the measurement results as shape data. The measurement results may include at least one of the following: height, width, volume, root angle, surface irregularities, and other information.

[0141] In S804, the information processing device 200 determines one or more block patterns corresponding to the weld bead formed in S803 based on the positional relationship between the weld bead formed in S803 and the surrounding weld bead that has been formed. The block pattern is determined by Figure 6 In addition, the types of block patterns are determined by the method shown in FIG. Figure 3 Predetermined as shown.

[0142] In S805 , the information processing device 200 associates the pass data selected in S801 , the shape data acquired in S803 , and the block pattern determined in S804 , and stores them in a storage unit (not shown).

[0143] In S806, the information processing device 200 determines whether the formation of the weld beads corresponding to all the pass data is complete. If there is any unprocessed pass data (No in S806), the information processing device 200 returns to S801 and repeats the processing of the unprocessed pass data. If the processing of all the pass data is complete, that is, if the formation of the stacked object W is complete (Yes in S806), the information processing device 200 proceeds to S807.

[0144] In S807, the information processing device 200 generates learning data using the stored data. Figure 4 As described, the learning data in this embodiment assumes training and learning, and is composed of pairs of input data (welding conditions, block patterns) and training data (shape data). Learning data is generated to match this structure. Alternatively, the stored data can be extracted to only those items required for learning, from among those included in the welding conditions and shape data. The generated learning data is used in subsequent learning processes. This process flow ends.

[0145] It should be noted that in Figure 8 , an example of forming a weld bead and generating learning data through a series of processes is shown. However, this configuration is not limiting; learning data may be generated when predetermined data is accumulated. Alternatively, information processing device 200 may provide stored data to an external device (not shown), which may then generate learning data.

[0146] As described above, according to this embodiment, more appropriate welding conditions can be determined when forming a laminated object.

[0147] <Other Implementation Methods>

[0148] In addition, in the present invention, it can also be achieved by supplying a program or application for implementing the functions of one or more of the above-mentioned embodiments to a system or device using a network or storage medium, and one or more processors in a computer of the system or device read and execute the program.

[0149] Alternatively, the present invention may be realized by a circuit that realizes one or more functions (for example, an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array)).

[0150] As described above, the following matters are disclosed in this specification.

[0151] (1) A machine learning device that performs machine learning on welding conditions when a laminated object is formed by depositing a filler material and laminating weld beads,

[0152] The machine learning device is characterized in that

[0153] A learning processing unit is provided for performing a learning process for generating a learned model using welding conditions of a weld bead and a block pattern composed of the weld bead as input data and shape data of the weld bead as output data.

[0154] This configuration enables more appropriate welding conditions to be determined when shaping a laminated object. In particular, a learned model can be generated for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0155] (2) The machine learning device according to (1), characterized in that

[0156] The block pattern includes any one of a pattern consisting of two passes of 1 layer×2 columns, a pattern consisting of two passes of 2 layers×1 column, a pattern consisting of three passes of 3 layers×1 column, and a pattern consisting of four passes of 2 layers×2 columns.

[0157] According to this configuration, by generating a learned model based on a predetermined relatively simple block pattern, the processing load in the learning process can be reduced, and efficient learning can be performed.

[0158] (3) The machine learning device according to (1) or (2), characterized in that

[0159] The shape data includes at least any one of a height, a width, and a volume of the weld bead.

[0160] According to this configuration, a learned model focusing on an arbitrary shape of a weld bead can be generated.

[0161] (4) The machine learning device according to any one of (1) to (3), characterized in that

[0162] The welding conditions include at least one of a feed rate of the filler material, a welding rate, a target position on a base for molding the stacked object, a heat input during molding, a moving speed of the welding torch, an inter-pass temperature, and an inter-pass time.

[0163] According to this configuration, a learned model can be generated by focusing on any item among various welding condition items.

[0164] (5) The machine learning device according to any one of (1) to (4), characterized in that

[0165] The learning processing unit performs the learning process using a training learning method using a neural network.

[0166] According to this configuration, machine learning corresponding to training learning using a neural network can be performed.

[0167] (6) A laminate molding system that forms a laminated object by depositing a filler material and laminating weld beads,

[0168] The stacking molding system is characterized by having:

[0169] a production mechanism for producing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object;

[0170] a determination unit that determines welding conditions when forming weld passes corresponding to the plurality of pass data generated by the generation unit;

[0171] a determining mechanism for determining a block pattern formed by the weld bead based on the configuration when the weld bead is formed;

[0172] a deriving unit that inputs the welding conditions corresponding to the pass data determined by the determining unit and the block pattern determined by the determining unit into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0173] An adjusting means adjusts the welding conditions so that a difference between the shape data derived by the deriving means and the shape data represented by the pass data produced by the producing means does not exceed a predetermined threshold value.

[0174] This configuration allows for determining more appropriate welding conditions when shaping a laminated object. In particular, it allows for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0175] (7) The stacking molding system according to (6), characterized in that:

[0176] The adjustment mechanism adjusts the welding conditions by repeatedly performing a process of changing the conditions when the pass data is generated by the creation mechanism.

[0177] According to this configuration, by adjusting the conditions when generating pass data based on design data, it is unnecessary to adjust the values ​​of various items included in the numerous welding conditions for forming a weld bead, thereby reducing the processing load and facilitating the determination of welding conditions.

[0178] (8) The stacking molding system according to (6) or (7), characterized in that:

[0179] The production organization has:

[0180] a first dividing unit for generating one or more slice data by dividing the design data of the stacked object into one or more layers in a direction perpendicular to a predetermined stacking direction; and

[0181] The second segmentation mechanism segments the one or more slice data generated by the first segmentation mechanism into a plurality of pass data.

[0182] According to this configuration, arbitrary pass data can be generated according to the design data of the stacked molded object.

[0183] (9) The stacking molding system according to any one of (6) to (8), characterized in that:

[0184] The determining mechanism determines one or more block patterns formed by the weld bead corresponding to the pass data.

[0185] According to this configuration, one or more block patterns can be identified for one piece of pass data to perform welding condition determination processing.

[0186] (10) The stacking molding system according to (9), characterized in that:

[0187] The deriving means sequentially derives shape data corresponding to the pass data for one or more block patterns determined by the determining means based on a priority predetermined for the block patterns.

[0188] According to this configuration, one or more block patterns determined for one pass data can be sequentially subjected to welding condition determination processing based on arbitrary priorities predetermined for the block patterns.

[0189] (11) The stacking molding system according to any one of (6) to (10), characterized in that:

[0190] The determination unit determines welding conditions for forming a weld bead corresponding to the pass data created by the creation unit using a database in which the shape of the weld bead and the welding conditions are previously associated.

[0191] According to this configuration, by determining welding conditions using a database in which weld bead shape data and welding conditions are previously associated, it is unnecessary to adjust each condition item individually when determining welding conditions, thereby reducing the processing load.

[0192] (12) The stacking molding system according to any one of (6) to (11), characterized in that:

[0193] The stacking molding system also has:

[0194] an acquisition mechanism for acquiring shape data of a weld bead when the weld bead is formed;

[0195] a determining mechanism that determines a block pattern formed by the weld bead based on a positional relationship with other weld beads that have already been formed when the weld bead is formed; and

[0196] A generating unit generates learning data used when performing the learning process based on the pass data of the weld bead, the shape data acquired by the acquiring unit, and the patch pattern determined by the determining unit.

[0197] According to this configuration, learning data for subsequent learning processing can be generated simultaneously with the shaping of the stacked shaped object. By repeating the learning processing using this learning data, more appropriate welding conditions can be determined.

[0198] (13) A machine learning method for determining welding conditions when a laminated object is formed by depositing a filler material and laminating weld beads,

[0199] The machine learning method is characterized in that

[0200] The method includes a learning processing step of performing a learning process for generating a learned model using welding conditions of a weld bead and a block pattern composed of the weld bead as input data and shape data of the weld bead as output data.

[0201] This configuration enables more appropriate welding conditions to be determined when shaping a laminated object. In particular, a learned model can be generated for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0202] (14) A method for determining welding conditions in a stacking molding system for molding a stacked object by depositing a filler material and stacking weld beads.

[0203] The method for determining welding conditions is characterized by comprising:

[0204] a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object;

[0205] a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step;

[0206] a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed;

[0207] a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0208] The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

[0209] This configuration allows for determining more appropriate welding conditions when shaping a laminated object. In particular, it allows for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0210] (15) A program, wherein:

[0211] The program causes a computer to execute a learning processing step for generating a learned model having welding conditions of weld beads constituting a stacked shaped object and a block pattern constituted by the weld beads as input data and shape data of the weld beads as output data.

[0212] This configuration enables more appropriate welding conditions to be determined when shaping a laminated object. In particular, a learned model can be generated for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0213] (16) A program, wherein:

[0214] The program is used to make the computer execute the following steps:

[0215] a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld beads constituting the stacked shaped object based on design data of the stacked shaped object formed by depositing a filler material and stacking weld beads;

[0216] a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step;

[0217] a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed;

[0218] a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; and

[0219] The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

[0220] This configuration allows for determining more appropriate welding conditions when shaping a laminated object. In particular, it allows for determining more appropriate welding conditions that take into account the patch pattern formed by the weld beads corresponding to the pass data.

[0221] While various embodiments have been described above with reference to the accompanying drawings, the present invention is not limited to these examples. Those skilled in the art will readily be able to devise various variations or modifications within the scope of the technical solutions, which are also understood to fall within the technical scope of the present invention. Furthermore, the various components of the above embodiments may be arbitrarily combined without departing from the spirit of the invention.

[0222] It should be noted that the present application is based on Japanese patent application (Japanese Patent Application No. 2020-123775) filed on July 20, 2020, the contents of which are incorporated herein by reference.

[0223] Description of Reference Numerals

[0224] 1Layered styling system

[0225] 100-layer modeling device

[0226] 101 Shape Sensor

[0227] 102 welding torch

[0228] 103 base

[0229] 104 welding robot

[0230] 106 robot controller

[0231] 107 Power Supply

[0232] 108 welds

[0233] 200 Information Processing Device

[0234] 201 Modeling Control Department

[0235] 202 Power Control Unit

[0236] 203 feed control unit

[0237] 204DB (Database) Management Department

[0238] 205 shape data acquisition unit

[0239] 206 Learning Data Management Department

[0240] 207 Learning Processing Department

[0241] 208 welding condition export section

[0242] W stacked shape

[0243] M filling material.

Claims

1. A machine learning device that performs machine learning on welding conditions when forming a laminated object by depositing a filler material and laminating weld beads. The machine learning device is characterized in that The machine learning device includes a learning processing mechanism that performs learning processing for generating a learned model that uses welding conditions of a weld bead and a block pattern formed by the weld bead as input data and uses shape data of the weld bead as output data, and adjusts the welding conditions in such a manner that a difference between the shape data output by the learned model and shape data represented by pass data created based on design data of the stacked molded object does not exceed a predetermined threshold value.

2. The machine learning device according to claim 1, wherein The block pattern includes any one of a pattern consisting of two passes of 1 layer×2 columns, a pattern consisting of two passes of 2 layers×1 column, a pattern consisting of three passes of 3 layers×1 column, and a pattern consisting of four passes of 2 layers×2 columns.

3. The machine learning device according to claim 1, wherein The shape data includes at least any one of a height, a width, and a volume of the weld bead.

4. The machine learning device according to claim 1, wherein The welding conditions include at least one of a feed rate of the filler material, a welding rate, a target position on a base for molding the stacked object, a heat input during molding, a moving speed of the welding torch, an inter-pass temperature, and an inter-pass time.

5. The machine learning device according to claim 1, wherein The learning processing unit performs the learning process using a training learning method using a neural network.

6. A stacking molding system for molding a stacked object by depositing a filler material and stacking weld beads. The stacking molding system is characterized by having: a production mechanism for producing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object; a determination unit that determines welding conditions when forming weld passes corresponding to the plurality of pass data generated by the generation unit; a determining mechanism for determining a block pattern formed by the weld bead based on the configuration when the weld bead is formed; a deriving unit that inputs the welding conditions corresponding to the pass data determined by the determining unit and the block pattern determined by the determining unit into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; as well as An adjusting means adjusts the welding conditions so that a difference between the shape data derived by the deriving means and the shape data represented by the pass data produced by the producing means does not exceed a predetermined threshold value.

7. The stacking molding system according to claim 6, characterized in that: The adjustment mechanism adjusts the welding conditions by repeatedly performing a process of changing the conditions when the pass data is generated by the creation mechanism.

8. The stacked molding system according to claim 6, characterized in that: The production mechanism has: a first dividing unit for generating one or more slice data by dividing the design data of the stacked object into one or more layers in a direction perpendicular to a predetermined stacking direction; as well as The second segmentation mechanism segments the one or more slice data generated by the first segmentation mechanism into a plurality of pass data.

9. The stacked molding system according to claim 6, characterized in that: The determining mechanism determines one or more block patterns formed by the weld bead corresponding to the pass data.

10. The stacking molding system according to claim 9, characterized in that: The deriving means sequentially derives shape data corresponding to the pass data for one or more block patterns determined by the determining means based on a priority predetermined for the block patterns.

11. The stacking molding system according to claim 6, characterized in that: The determination unit determines welding conditions for forming a weld bead corresponding to the pass data created by the creation unit using a database in which the shape of the weld bead and the welding conditions are previously associated.

12. The stacking molding system according to claim 6, characterized in that: The stacking molding system also has: an acquisition mechanism for acquiring shape data of a weld bead when the weld bead is formed; a setting mechanism for setting a block pattern formed by the weld bead based on a positional relationship with other weld beads already formed when the weld bead is formed; as well as Generating means generates learning data used in the learning process based on the pass data of the weld bead, the shape data acquired by the acquiring means, and the block pattern determined by the setting means.

13. A machine learning method for determining welding conditions when laminating a model by depositing a filler material and laminating weld beads. The machine learning method is characterized in that The machine learning method includes a learning processing step, wherein the learning processing step performs a learning process for generating a learned model that uses welding conditions of a weld bead and a block pattern formed by the weld bead as input data and uses shape data of the weld bead as output data, and adjusts the welding conditions in such a manner that a difference between the shape data output by the learned model and shape data represented by pass data created based on design data of the stacked shaped object does not exceed a predetermined threshold value.

14. A method for determining welding conditions in a stacking molding system for molding a stacked object by depositing a filler material and stacking weld beads. The method for determining welding conditions is characterized by comprising: a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld passes constituting the stacked shaped object based on the design data of the stacked shaped object; a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step; a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed; a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; as well as The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

15. A storage medium recording a program, wherein: The program is used to cause a computer to execute a learning processing step, wherein the learning processing step performs a learning process for generating a learned model that uses welding conditions of a weld bead constituting a stacked shaped object and a block pattern constituted by the weld bead as input data and uses shape data of the weld bead as output data, and adjusts the welding conditions in such a manner that a difference between the shape data output by the learned model and shape data represented by pass data created based on design data of the stacked shaped object does not exceed a predetermined threshold value.

16. A storage medium recording a program, wherein: The program is used to make the computer execute the following steps: a manufacturing step of manufacturing a plurality of pass data corresponding to a plurality of weld beads constituting the stacked shaped object based on design data of the stacked shaped object formed by depositing a filler material and stacking weld beads; a determination step of determining welding conditions for forming weld passes corresponding to the plurality of pass data produced in the production step; a determining step of determining a block pattern formed by the weld bead based on a configuration when the weld bead is formed; a deriving step of inputting the welding conditions corresponding to the pass data determined in the determining step and the block pattern determined in the determining step into a learned model generated by performing a learning process using the welding conditions of the weld bead and the block pattern formed by the weld bead as input data and the shape data of the weld bead as output data, thereby deriving shape data corresponding to the pass data; as well as The adjusting step adjusts the welding conditions so that a difference between the shape data derived in the deriving step and the shape data represented by the pass data produced in the producing step does not exceed a predetermined threshold value.

Images