Processability determination device, processability training device, processability determination method, processability training method, processability determination program, and processability training program

Abstract

A processability determination device (100) according to the present invention comprises: a three-dimensional shape extraction unit (10) that extracts a three-dimensional shape from three-dimensional shape data indicating a planned processing shape; a three-dimensional shape rotation unit (12) that rotates the three-dimensional shape so that a processing surface of the three-dimensional shape faces a desired processing axis direction; a depth map conversion unit (14) that generates a depth map obtained by adding information about depth in the desired processing axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthographic projection of the processing surface onto a plane orthogonal to the desired processing axis direction; and an inference model (16) that is built by machine learning using training processing axis direction indications, a training depth map extracted from training three-dimensional shape data, and processability information indicating the possibility of actual processing performed according to the training processing axis direction indications and the training three-dimensional shape data, and determines whether processing into an intended shape is possible by inference using the desired processing axis and the depth map generated by the depth map conversion unit (14).

Description

Machining possibility determination device, machining possibility learning device, machining possibility determination method, machining possibility learning method, machining possibility determination program, and machining possibility learning program

[0001] The present disclosure relates to a processing possibility determination device, a processing possibility learning device, a processing possibility determination method, a processing possibility learning method, a processing possibility determination program, and a processing possibility learning program.

[0002] In cutting processes, which involve cutting a material (workpiece) with a cutting tool such as a drill, milling cutter, or cutting tool (hereinafter referred to as "tool"), cutting with the tool can be difficult depending on the design shape. Furthermore, heat generated during cutting can cause expansion of the material or deformation of the tool, and further processing defects can occur due to vibration or deflection of the tool.

[0003] Patent Document 1 discloses an electrode manufacturing method for determining whether an electrode for electric discharge machining can be manufactured. Non-Patent Document 1 also discloses a technology for determining whether cutting is possible using a device that inputs voxel data representing a three-dimensional shape into a 3D-CNN (three-dimensional convolutional neural network) model and allows it to learn.

[0004] JP 2010-105080 A

[0005] Computer Aided Geometric Design 62 (2018) 263-275

[0006] However, the technology described in Patent Document 1 extracts all of the ridge lines that make up the electrode shape from the input three-dimensional electrode data and shape identification information, and determines whether or not the electrode contour identified from the extracted ridge lines can be machined, which has the problem of increasing the computational load of extracting the electrode contour from the three-dimensional electrode data, etc. Furthermore, the technology described in Non-Patent Document 1 requires a large-scale 3D-CNN model to handle voxel data, which is three-dimensional data, which has the problem of increasing the computational load and the number of data required for learning.

[0007] The present disclosure aims to provide a machining feasibility determination device, a machining feasibility determination method, and a machining feasibility determination program that determine whether cutting processing is possible (whether cutting processing is possible or not), and a machining feasibility learning device, a machining feasibility learning method, and a machining feasibility learning program that construct the machining feasibility determination device through learning that reduces the number of learning data and the computational load.

[0008] a depth map conversion unit that generates a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane orthogonal to the desired machining axis direction; and an inference unit that is constructed by machine learning using a learning machining axis direction instruction including instruction information for the learning machining axis direction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information describing the feasibility of actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data, and that determines whether the to-be-machined shape can be machined by inference using the desired machining axis direction and the depth map generated by the depth map conversion unit.

[0009] The machining feasibility determination method disclosed herein is a machining feasibility determination method executed by a computer, and includes the steps of: extracting a three-dimensional shape from three-dimensional shape data indicating a shape to be machined; rotating the three-dimensional shape so that a machining surface in the three-dimensional shape faces a desired machining axis direction; generating a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the desired machining axis direction; and determining whether the shape to be machined can be machined by inference using the desired machining axis direction and the depth map, the depth map being constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information describing whether actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data is possible.

[0010] The machining feasibility determination program of the present disclosure causes a computer to execute the following steps: extracting a three-dimensional shape from three-dimensional shape data indicating a shape to be machined; rotating the three-dimensional shape so that a machining surface in the three-dimensional shape faces a desired machining axis direction; generating a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the desired machining axis direction; and determining whether the shape to be machined can be machined by inference using the desired machining axis direction and the depth map, the depth map being constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information describing whether actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data is possible.

[0011] According to the present disclosure, it is possible to provide a machining feasibility determination device, a machining feasibility determination method, and a machining feasibility determination program that determine whether cutting processing is possible through learning that reduces the number of learning data and the calculation load by using a depth map from the machining axis direction, as well as a machining feasibility learning device, a machining feasibility learning method, and a machining feasibility learning program that perform the learning.

[0012] 6A is a functional configuration diagram showing a machining feasibility determination device according to a first embodiment, and FIG. 6B is a functional configuration diagram showing a machining feasibility learning device according to the first embodiment. It is a hardware configuration diagram showing a machining feasibility determination device and a machining feasibility learning device according to the first embodiment. (a), (b), and (c) are explanatory diagrams illustrating cutting processing using a cutting machine such as a milling machine or a machining center. (a), (b), (c), (d), and (e) are explanatory diagrams illustrating cases in which machining is not possible. It is a flowchart showing an example of a determination process of a learned machining feasibility determination device according to the first embodiment. (a) is a schematic diagram showing an example of a three-dimensional shape extracted from three-dimensional CAD data, and (b) is an example of a depth map generated from the three-dimensional shape shown in FIG. 6A. (a) is a schematic diagram showing an example of a case in which a tool accesses a material from the machining axis direction, and (b) is an explanatory diagram showing that shape information of a machining surface can be represented in a depth map by orthographic projection. It is a functional configuration diagram showing a machining feasibility determination device according to a second embodiment. 13A is an explanatory diagram illustrating an example of a location where cutting is difficult in the current machining axis direction, and FIG. 13B is an explanatory diagram illustrating a case where machining becomes possible by changing the machining axis direction. FIG. 13B is a flowchart illustrating an example of a judgment process of a learned machining feasibility determination device according to embodiment 2. FIG. 13A is a functional configuration diagram illustrating a machining feasibility determination device according to embodiment 3, and FIG. 13B is a functional configuration diagram illustrating a machining feasibility learning device according to embodiment 3. FIG. 13B is a functional configuration diagram illustrating a machining feasibility learning device according to embodiment 4. FIG. 13A is a schematic diagram illustrating an example of a depth map, and FIG. 13B is a schematic diagram illustrating an example of a heat map of inappropriate machining locations, which is machining feasibility information corresponding to the depth map of FIG. 13A. FIG. 13C is a functional configuration diagram illustrating a machining feasibility determination device according to embodiment 5. 15A is an explanatory diagram showing the planned machining shape for the material shape, (b) is an explanatory diagram when the machining surface / machining axis direction recording unit records the machining axis direction in the machining surface list created by the machining surface division unit, (c) is an explanatory diagram when the material shape shown in FIG. 15A is rotated 90° to the left on the figure by the three-dimensional shape rotation unit, and (d) is an explanatory diagram when the machining surface / machining axis direction recording unit records the machining axis direction in the machining surface list with the material shape rotated 90°.13 is a flowchart showing an example of a determination process of a trained machining possibility determination device according to embodiment 5.

[0013] The following describes a processing feasibility determination device and a processing feasibility learning device according to embodiments, with reference to the drawings. The following embodiments are merely examples, and the embodiments can be appropriately combined and modified.

[0014] 1A is a functional configuration diagram showing a machining feasibility determination device 100 according to embodiment 1. The machining feasibility determination device 100 includes a three-dimensional shape extraction unit 10 that extracts post-machining shape information (planned machining shape) of a material to be machined from input 3D CAD data, a three-dimensional shape rotation unit 12 that orients a surface to be machined (hereinafter referred to as a "machining surface") in the extracted shape information in a desired machining axis direction (e.g., Z-axis direction) based on an input machining axis instruction, a depth map conversion unit 14 that converts shape information in which the machining surface is oriented in the machining axis direction to generate a depth map in the machining axis direction as shown in FIG. 6B, and an inference model 16 that determines whether the generated depth map can be machined and outputs the determination result. As will be described later, the inference model 16 is constructed by training a mathematical model such as a CNN through machine learning using learning machining axis direction instructions, each of which is prepared in advance for learning, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instructions, and machining feasibility information that describes whether or not actual machining performed in accordance with the learning machining axis direction instructions and the learning three-dimensional shape data is possible.

[0015] 1(b) is a functional configuration diagram showing a machining feasibility learning device 110 according to embodiment 1. In addition to the above-mentioned three-dimensional shape extraction unit 10, three-dimensional shape rotation unit 12, depth map conversion unit 14, and inference model 16, the machining feasibility learning device 110 includes a shape / machining axis direction database (DB) 20, a machining feasibility information DB 24 that stores machining feasibility information, which is training data, and a learning device 22 that updates the inference model 16 with the machining feasibility information.

[0016] The shape / machining axis direction DB20 includes a shape information DB20A that stores three-dimensional CAD data, which is three-dimensional shape data for learning, and a machining axis direction instruction DB20B that stores machining axis direction instructions, which are machining axis direction instructions for learning.

[0017] The data stored in each of the shape information DB 20A, the machining axis direction instruction DB 20B, and the machining feasibility information DB 24 is past machining case data. For example, the 3D CAD data stored in the shape information DB 20A is data of a learning 3D shape actually used in machining, and the machining axis direction instructions stored in the machining axis direction instruction DB 20B are also data of learning machining axis direction instructions actually used in machining. The machining feasibility information DB 24 stores the feasibility of machining results using the 3D CAD data stored in the shape information DB 20A and the machining axis direction instructions stored in the machining axis direction instruction DB 20B. The learning machining feasibility information stored in the machining feasibility information DB 24 corresponds to each of the 3D CAD data stored in the shape information DB 20A and the machining axis direction instructions stored in the machining axis direction instruction DB 20B. In embodiment 1, the learning device 22 evaluates the inference model 16 by comparing the results of the inference model 16's judgment of the learning depth map obtained using past machining case data, namely, 3D CAD data and machining axis direction instructions, with the machining feasibility information corresponding to the 3D CAD data and machining axis direction instructions used by the inference model 16 for judgment, and updates the inference model 16.

[0018] 2 is a hardware configuration diagram showing the manipulability determination device 100 and the manipulability learning device 110 according to embodiment 1. The manipulability determination device 100 and the manipulability learning device 110 are composed of a computer including a processor 210, a storage device 220, and an input / output interface 230. The manipulability determination device 100 may be composed of multiple computers.

[0019] The processor 210 is an integrated circuit (IC) that performs arithmetic processing. Specific examples of the processor 210 include a central processing unit (CPU), a digital signal processor (DSP), and a graphics processing unit (GPU). The processor 210 can render 3D CAD data and reorient the imaged 3D CAD data in any machining axis direction. By running a program that generates a depth map from the 3D CAD data, the processor 210 functions as the 3D shape extraction unit 10, the 3D shape rotation unit 12, and the depth map conversion unit 14. The processor 210 also functions as the learner 22 by running a program that performs machine learning using training data. Based on the results of the machine learning, the processor 210 constructs an inference model 16 that determines whether machining is possible. In the first embodiment, the operation of the machine learning-related manipulability learning program causes the processor 210 to function as the three-dimensional shape extraction unit 10, the three-dimensional shape rotation unit 12, the depth map conversion unit 14, and the learner 22 that updates the inference model 16. Also, in the first embodiment, after the machine learning, the operation of the manipulability determination program causes the processor 210 to function as the three-dimensional shape extraction unit 10, the three-dimensional shape rotation unit 12, the depth map conversion unit 14, and the inference model 16. Also, the manipulability determination program and the manipulability learning program are provided, for example, by a recording medium on which they are recorded.

[0020] The storage device 220 is configured by a volatile storage device such as a RAM (Random Access Memory), a ROM (Read Only Memory), a HDD (Hard Disk Drive), or a non-volatile storage device such as a flash memory.

[0021] The input / output interface 230 is a port to which the input device 300 and the output device 310 are connected. Specific examples of the input / output interface 230 include a USB (Universal Serial Bus) terminal, an IEEE 1394 terminal, a Thunderbolt terminal, or the like, and further includes a communication interface such as Ethernet. The input device 300 is a touch panel, a keyboard, a mouse, or the like. The output device 310 is a display, a printer, or the like. During learning, the shape / machining axis direction DB 20 and the machining feasibility information DB 24 are each connected to the input / output interface 230. The shape / machining axis direction DB 20 and the machining feasibility information DB 24 may each be constructed in the storage device 220.

[0022] 3(a), (b), and (c) are explanatory diagrams illustrating cutting processing using a cutting machine such as a milling machine or a machining center. Fig. 3(a) is an explanatory diagram of plane machining in which a tool 30 such as a flat milling cutter is used to cut a planar cutting surface 42 of a material 40. In plane machining, the tool 30 is moved in a direction parallel to a plane of the material 40 that is perpendicular to the machining axis direction, as indicated by the arrow, to cut the cutting surface 42.

[0023] 3B is an explanatory diagram of side surface machining in which a tool 32 such as a corner milling cutter is used to cut the side surface of a material 40. In side surface cutting, the tool 32 is moved in the direction of the arrow to cut a cutting surface 44.

[0024] 3C is an explanatory diagram of drilling / pocketing in which a tool 34 such as an end mill cuts a blank 40 in the machining axis direction. In drilling / pocketing, the tool 34 is fed in the machining axis direction shown by the arrow to cut a cutting surface 46 of the blank 40.

[0025] In the first embodiment, the possibility of machining other than the above is also determined. For example, it may be determined whether or not the machining is possible when forming a three-dimensional shape by electrical discharge machining, which processes a material by arc discharge between an electrode and the material.

[0026] 4(a), (b), (c), (d), and (e) are explanatory diagrams illustrating examples of cases in which machining is impossible. Fig. 4(a) shows a case of tool interference, in which a portion of the tool 34 other than the cutting edge interferes with the blank 40, and Fig. 4(b) shows a case in which the tool 34 is not suitable for the machining shape of the blank 40. In the cases shown in Figs. 4(a) and 4(b), the problem can be solved by applying machining rules, such as changing the tool 34 to one that is more suitable for the situation, so it is possible to determine whether machining is possible without using the machining possibility determination device 100 according to the first embodiment.

[0027] FIG. 4( c ) shows a case where the cutting surface 44 cannot be accurately cut due to vibration of the tool 32, and FIG. 4( d ) shows a case where the tool 34 bends and the material 40 cannot be cut properly. FIG. 4( e ) shows an example where the processed shape of the material 40 is complex, making it difficult to make a judgment based on the processing rules. In the cases shown in FIGS. 4( c ), 4 ( d ), and 4 ( e ), a solution based on the processing rules, such as changing the tool, is not possible, so in the past, skilled technicians have judged whether or not the processing can be performed based on their own experience. In the first embodiment, the learned processing feasibility determination device 100 determines whether or not the processing can be performed in the cases shown in FIGS. 4( c ), 4 ( d ), and 4 ( e ), without relying on skilled technicians.

[0028] 5 is a flowchart showing an example of the determination process of the trained machining feasibility determination device 100 according to Embodiment 1. In step S101, three-dimensional CAD data, which is shape information, is input to the three-dimensional shape extraction unit 10.

[0029] In step S102, the three-dimensional shape extraction unit 10 renders the input three-dimensional CAD data to extract a three-dimensional shape. Fig. 6A is a schematic diagram showing an example of a three-dimensional shape extracted from the three-dimensional CAD data.

[0030] In step S103, the three-dimensional shape rotation unit 12 rotates the three-dimensional shape extracted in step S102 in accordance with the machining axis direction instruction. In the first embodiment, as an example, the machining surface of the material 40 is oriented in the Z axis direction. More specifically, the machining surface of the material 40 is made to face the Z axis direction. The machining axis direction instruction may be set in advance to the Z axis, or may be input at the stage of step S101.

[0031] In step S104, the depth map conversion unit 14 generates a depth map of a three-dimensional shape with the processing surface oriented in the Z-axis direction. FIG. 6B is an example of a depth map generated from the three-dimensional shape shown in FIG. 6A. The depth map is two-dimensional image data generated by orthogonally projecting a three-dimensional shape with the processing surface oriented in the Z-axis direction onto a plane perpendicular to the Z-axis. The depth map includes information on depth in the Z-axis direction attached to two-dimensional image data showing the processing surface. The information on depth in the Z-axis direction is extracted from the input three-dimensional CAD data, which is the shape information. In FIG. 6B, the depth information is hue or brightness. In the depth map, shallow areas are represented by a light hue such as yellow, and deep areas are represented by a dark hue such as dark blue. When depth is represented by a single color such as gray, shallow areas are represented by a high brightness, and deep areas are represented by a low brightness.

[0032] In a cutting machine such as a milling machine, as shown in Figure 7(a), a tool 32 approaches and processes a material 40 from the processing axis direction (Z-axis direction), so the information necessary to determine whether or not processing is possible is shape information of the processing surface as viewed from the Z-axis direction, and shape information of the side or back surface of the material 40 other than the processing surface is not necessary for the time being. As shown by the arrow in Figure 7(b), shape information of the processing surface can be expressed in a depth map by orthogonal projection, so whether or not processing is possible can be determined using a depth map, which is two-dimensional data, instead of data indicating a three-dimensional shape.

[0033] In step S105, the inference model 16 infers whether machining is possible. As described above, in the first embodiment, the inference model 16 is trained using the 3D CAD data, the machining axis direction instruction, and the machining possibility information corresponding to each of the 3D CAD data and the machining axis direction instruction. The trained inference model 16 performs inference based on the machining possibility information used for training and outputs the inference result.

[0034] In step S106, the inference model 16 determines whether the inference result output in step S105 can be processed. If the determination result is processable in step S106, the procedure proceeds to step S107, and if the determination result is not processable, the procedure proceeds to step S108.

[0035] In step S107, the inference model 16 outputs the processing possibility determination to the output device 310 and ends the process. In step S108, the inference model 16 outputs the processing impossibility determination to the output device 310 and ends the process.

[0036] As described above, in the first embodiment, the possibility of processing is determined using a depth map, which is two-dimensional data generated from a three-dimensional shape. Since the data is two-dimensional, the data size is smaller than that of a three-dimensional shape, so the number of training data and the computational load during training of the inference model 16 can be reduced. In addition, the computational load for determining whether processing is possible in the processing possibility determination device 100 after training can be reduced.

[0037] When realizing an inference model 16 for determining whether cutting is possible, conventional models that input voxel data require a large amount of calculation and training data because they handle data representing three-dimensional shapes. However, in the first embodiment, the amount of calculation and training data can be reduced by inputting a depth map, which is two-dimensional data. Furthermore, for the same amount of training data, the number of neural network parameters is reduced, making it possible to achieve higher determination accuracy than conventional models.

[0038] Second Embodiment Next, a machining feasibility determination device 120 according to the second embodiment will be described. The machining feasibility determination device 120 according to the second embodiment shown in FIG. 8 includes a machining axis direction adjustment unit 26 that adjusts the machining axis direction in accordance with the machining feasibility determination output by the inference model 16, and differs from the first embodiment in that the three-dimensional shape rotation unit 28 rotates the three-dimensional shape in the machining axis direction adjusted by the machining axis direction adjustment unit 26. However, since the other configurations are the same as those of the first embodiment, the same components as those of the first embodiment are denoted by the same reference numerals as those of the first embodiment, and detailed description thereof will be omitted. Furthermore, since the hardware configuration of the second embodiment is the same as that of the first embodiment, detailed description thereof will be omitted. However, in the second embodiment, the processor 210 functions as the three-dimensional shape extraction unit 10, the three-dimensional shape rotation unit 28, the depth map conversion unit 14, and the inference model 16, and also functions as the machining axis direction adjustment unit 26.

[0039] In the first embodiment, the machining possibility was determined using the Z-axis as the machining axis. However, when determining whether machining is possible, there may be a portion that is difficult to cut with the current machining axis, such as the undercut shape 38 shown in FIG. 9( a). Even in the case shown in FIG. 9( a), machining may be possible depending on the machining axis direction, as shown in FIG. 9( b). In the second embodiment, when design data is provided, the machining possibility determination is repeated while automatically changing the machining axis direction (the orientation of the material 40), thereby verifying all machining axis directions. For example, as shown in FIG. 8, the machining axis direction is changed by rotating the material 40 in a vertical direction 50 or a horizontal direction 52. Some cutting machines allow the machining axis to be changed while the material 40 is fixed. However, even with such machines, rotating the material 40 to change the machining surface allows for easier and faster machining control than changing the machining axis. In the second embodiment and the fifth embodiment described below, the actual machining axis on which the tool is set is fixed, for example, in the Z-axis direction, and the three-dimensional shape is rotated so that the machining surface is the same as when the machining axis is changed from the Z-axis.

[0040] Fig. 10 is a flowchart showing an example of the determination process of the trained machining feasibility determination device 120 according to embodiment 2. The flowchart shown in Fig. 10 differs from the flowchart of embodiment 1 shown in Fig. 5 in that it includes step S203 instead of step S103 in embodiment 1, step S204 for determining whether the machining axis direction has been changed in all directions, and step S205 for changing the machining axis direction instruction. However, since the other steps are the same as those in embodiment 1, the same steps as those in embodiment 1 are denoted by the same reference numerals as in embodiment 1, and detailed description thereof will be omitted.

[0041] In step S203, the three-dimensional shape rotation unit 28 rotates the three-dimensional shape extracted in step S102 in accordance with the machining axis direction instruction. In the second embodiment, in the first determination process, the machining axis direction instruction is set to the Z axis in advance. The machining axis direction instruction may be input at the stage of step S101. As will be described later, in the second embodiment, if the determination result indicates that machining is not possible with the current machining axis direction instruction, the machining axis direction adjustment unit 26 changes the machining axis direction instruction, and in step S203, the three-dimensional shape is rotated in accordance with the changed machining axis direction instruction.

[0042] If the determination result in step S106 is that machining is not possible (if machining is not possible), the machining axis direction adjustment unit 26 determines in step S204 whether or not machining is possible in all machining axis directions. If the material 40 is considered to be a rectangular parallelepiped, there are a total of six machining axis direction instructions. Of these six machining axis direction instructions, the machining axis direction adjustment unit 26 registers in the storage device 220 the Z axis set in the initial determination and the machining axis direction instruction changed in the subsequent step S205. In step S204, the machining axis direction adjustment unit 26 determines whether or not machining is possible in all machining axis directions by referring to the storage device 220.

[0043] If it is determined in step S204 that the feasibility of machining has been examined in all machining axis directions, the procedure proceeds to step S108. In step S108, as in the first embodiment, a machining impossibility determination is output to the output device 310, and the process ends.

[0044] If it is determined in step S204 that the feasibility of machining has not been considered for all machining axis directions, the procedure proceeds to step S205. In step S205, the machining axis direction adjustment unit 26 changes the machining axis direction instruction and inputs the changed machining axis direction instruction to the three-dimensional shape rotation unit 28.

[0045] In step S203, the three-dimensional shape is rotated according to the changed machining axis direction instruction. Specifically, the three-dimensional shape is rotated so that a new machining surface when machining the three-dimensional shape in the machining axis direction indicated by the machining axis direction instruction input from the machining axis direction adjustment unit 26 faces the desired machining axis direction (the Z-axis direction in the second embodiment). In the subsequent steps, the depth map conversion unit 14 orthogonally projects the new machining surface onto a plane perpendicular to the Z-axis direction to generate a new depth map, and the inference model 16 uses the new depth map to determine whether the to-be-machined shape can be machined. If the determination result in step S106 is that machining is possible, a machining possibility determination is output to the output device 310 in step S107, and the process ends.

[0046] As described above, in the second embodiment, when design data is provided, the machining axis direction is automatically changed while repeatedly determining whether machining is possible, thereby verifying all machining axis directions. As a result, even in the case of a shape whose machining ability changes depending on the machining axis direction, by making a determination for all machining axis directions, it is possible to determine at what angle the part shape indicated by the design data cannot be machined. Furthermore, in the second embodiment, it is possible to determine whether machining is possible for the side and back shapes, which are lost as information when a three-dimensional shape is converted into a depth map.

[0047] Next, a description will be given of a machining feasibility determination device 130 and a machining feasibility learning device 140 according to Embodiment 3. The machining feasibility learning device 140 according to Embodiment 3 shown in Fig. 11(b) differs from Embodiment 1 in that the learner 22 evaluates the inference model 58 by comparing a result of judgment made by the inference model 58 using the depth map obtained using the 3D CAD data and machining axis direction instruction, which are past machining case data, respectively, the machining axis direction indicated by the machining axis direction instruction, and information on the material of the blank 40, tool information (tool type, tool material, tool diameter, tool length, etc.) and cutting parameters (tool feed rate, tool rotation speed, etc.) stored in the machining information DB 20C included in the shape / machining axis direction / machining information DB 48, with machining feasibility information stored in the machining feasibility information DB 24, and thereby updates the inference model 58. 11(a) differs from the first embodiment in that the learned inference model 58 determines a depth map obtained using the 3D CAD data and the machining axis direction instruction, the machining axis direction indicated by the machining axis direction instruction, and machining information including material information of the blank 40, tool information, and cutting parameters. However, since the other configurations are the same as those of the first embodiment, the same components as those of the first embodiment are denoted by the same reference numerals as those of the first embodiment, and detailed description thereof will be omitted. Furthermore, since the hardware configuration of the third embodiment is the same as that of the first embodiment, detailed description thereof will be omitted.

[0048] The quality of the blank 40, the tool information, and the cutting parameters all have a significant effect on the determination of whether or not machining is possible. For example, if an appropriate tool is not used for a difficult-to-cut material such as stainless steel, or if the feed rate or rotation speed of the tool is not appropriate for the difficult-to-cut material, there is a risk that the cutting surface 44 will not be cut accurately due to vibrations of the tool 32, as shown in Fig. 4(c), or that the tool 34 will bend, as shown in Fig. 4(d), and the blank 40 will not be cut correctly.

[0049] In this way, even if the design data is the same, whether cutting is possible or not varies depending on the information on the material of the raw material 40, the tool information, and the cutting parameters. In the machining possibility determination device 130 and the machining possibility learning device 140 according to the third embodiment, at least one of the information on the material of the raw material 40, the tool information, and the cutting parameters is added to the input to the inference model 58 to determine whether machining is possible or not.

[0050] As described above, according to the third embodiment, even when the possibility of machining varies depending not only on the design data but also on the material of the blank 40, the tool, and the cutting parameters, an accurate possibility determination can be made.

[0051] Fourth Embodiment Next, a processing feasibility learning device 150 according to the fourth embodiment will be described. The processing feasibility learning device 150 according to the fourth embodiment shown in FIG. 12 differs from the first embodiment in that a heat map representing unsuitable processing locations is stored in the processing feasibility information DB 56, and the learning device 54 updates the inference model 16 using the heat map as training data. However, since the other configurations are the same as those of the first embodiment, the same components as those of the first embodiment are assigned the same reference numerals, and detailed descriptions thereof will be omitted. Furthermore, the processing feasibility determination device constructed by the processing feasibility learning device 150 has the same configuration as the processing feasibility determination device 100 according to the first embodiment, except for the fact that the inference model 16 outputs a determination result identifying unsuitable processing locations through learning using a heat map representing unsuitable processing locations as processing feasibility information to update the inference model 16, and therefore detailed descriptions thereof will be omitted. Furthermore, the hardware configuration of the fourth embodiment is the same as that of the first embodiment, and therefore detailed descriptions thereof will be omitted.

[0052] Figure 13(a) is a schematic diagram showing an example of a depth map, and Figure 13(b) is a schematic diagram showing an example of a heat map of areas that are unsuitable for processing, which is processing feasibility information corresponding to the depth map of Figure 13(a).

[0053] 13A, sharp edges 60, known as pin angles, are marked at locations corresponding to recesses. In cutting, recesses are cut with a tool such as an end mill, but because the tool such as an end mill cuts while rotating, it is inevitable that the corners cut out of the material 40 will be rounded, and it is not possible to reproduce pin angles such as the edges 60.

[0054] In the fourth embodiment, a heat map corresponding to the depth map is used as training data for training the inference model 16. The heat map is obtained by dividing the range corresponding to the depth map, such as that shown in FIG. 13( a), into a grid of predetermined cell sizes, and associating each cell with a numerical value indicating the possibility of machining. For example, in FIG. 13( b), the white cells are machinable areas 62 associated with numerical values ​​close to 0, and the black cells are unmachinable areas 66 associated with numerical values ​​close to 1. In addition, the gray cells in FIG. 13( b) are intermediate areas 64 associated with numerical values ​​between 0 and 1 that are somewhat difficult to machine. The numerical values ​​associated with each cell of the heat map, which represents machining feasibility information, are determined based on the results of actual manual verification of the feasibility of machining.

[0055] In embodiment 4, the learning device 54 evaluates the inference model 16 by comparing the results of the inference model 16's judgment of a depth map obtained using three-dimensional CAD data and machining axis direction instructions, which are past machining case data, with a heat map, which is machining feasibility information corresponding to the three-dimensional CAD data and machining axis direction instructions used for the judgment by the inference model 16, and updates the inference model 16.

[0056] The processing possibility determination device constructed by the above learning outputs a heat map in which the degree of processing improper parts is quantified on a cell-by-cell basis as a processing possibility determination result. As a result, according to the fourth embodiment, it is possible to precisely indicate which parts of the three-dimensional shape or depth map are processing improper parts.

[0057] Fifth Embodiment Next, a machining possibility determination device 160 according to a fifth embodiment will be described. The machining feasibility determination device 160 according to the fifth embodiment shown in FIG. 14 includes: a machining surface division unit 70 that creates a machining surface list by listing all machining surfaces of the material 40 from the three-dimensional shape extracted by the three-dimensional shape extraction unit 10; a machining surface / machining axis direction recording unit 74 that receives the machining surface list created by the machining surface division unit 70 and records the machining axis directions of the machining surfaces that are determined to be machineable in the machining feasibility determination output by the inference model 16 in the machining surface list; a termination determination unit 76 that makes an end determination when it is determined that all machining surfaces listed in the machining surface list are machineable or when the machineability in all machining axis directions has been verified, and outputs the machineability determination for the three-dimensional shape and the machining axis directions of each machining surface; and a machining axis direction adjustment unit 78 that outputs an instruction for a machining axis direction other than the machining axis direction whose machineability has been verified when the termination determination unit 76 determines that all machining axis directions have not been verified. The three-dimensional shape rotation unit 72 differs from the first embodiment in that it rotates the three-dimensional shape in the machining axis direction adjusted by the machining axis direction adjustment unit 78. However, since the other configurations are the same as those of embodiment 1, the same configurations as those of embodiment 1 are assigned the same reference numerals as those of embodiment 1, and detailed description thereof will be omitted. Furthermore, since the hardware configuration of embodiment 5 is the same as that of embodiment 1, detailed description thereof will be omitted, but in embodiment 5, the processor 210 functions as a three-dimensional shape extraction unit 10, a three-dimensional shape rotation unit 72, a depth map conversion unit 14, and an inference model 16, and also functions as a machining surface division unit 70, a machining surface / machining axis direction recording unit 74, an end determination unit 76, and a machining axis direction adjustment unit 78.

[0058] Fig. 15(a) is an explanatory diagram showing a to-be-machined shape 82 for a material shape 80. Machining surfaces (1), (2), (3), (4), (5), (6), and (7) are set in the to-be-machined shape 82, but in the state shown in Fig. 15(a), the surface that can be machined by the tool 34 is the machinable surface 84, and machining surfaces (5), (6), and (7) cannot be machined.

[0059] 15(b) is an explanatory diagram of a case where the machining surface / machining axis direction recording unit 74 records the machining axis direction in the machining surface list created by the machining surface dividing unit 70. In the case shown in FIG. 15(a), machining of machining surfaces (1) to (4) is possible, so the machining axis direction in the machining surface list is set to 0°, which indicates the Z axis direction, for example. However, in the case shown in FIG. 15(a), machining of machining surfaces (5) to (7) is not possible, so the machining axis direction in the machining surface list is left blank.

[0060] Figure 15(c) is an explanatory diagram of the case where the material shape 80 shown in Figure 15(a) is rotated 90 degrees to the left in the figure by the three-dimensional shape rotation unit 72. By rotating the material shape 80 by 90 degrees, it becomes possible to machine the machining surfaces (5) to (7) that were impossible to machine before the rotation.

[0061] 15(d) is an explanatory diagram of a case where the machining plane / machining axis direction recording unit 74 records the machining axis direction in the machining plane list when the material shape 80 is rotated 90°. As shown in FIG. 15(c), when the material shape 80 is rotated 90°, machining of machining planes (5) to (7) is possible, so the machining axis direction in the machining plane list is recorded as +90°, which is the rotation angle with respect to the Z axis.

[0062] 16 is a flowchart showing an example of the judgment processing of the learned machining feasibility judgment device 160 according to the fifth embodiment. The flowchart shown in FIG. 16 differs from the flowchart of the first embodiment shown in FIG. 5 in that it includes step S302 instead of step S103 in the first embodiment, step S301 in which the three-dimensional shape extraction unit 10 extracts machining surfaces from the extracted three-dimensional shape and creates a machining surface list, step S304 in which the machining surface / machining axis direction recording unit 74 records the machining surfaces and machining axis directions determined to be machineable in the machining surface list, step S305 in which the termination judgment unit 76 determines whether all machining surfaces have been determined to be machineable, step S306 in which the machining feasibility judgment and the machining axis directions of each machining surface are output, step S307 in which it determines whether the machining axis directions have been changed in all directions, and step S308 in which the machining axis direction instruction is changed. However, since the other steps are the same as those in the first embodiment, the same steps as those in the first embodiment are denoted by the same reference numerals as those in the first embodiment, and detailed description thereof will be omitted.

[0063] In step S301, the machining surface division unit 70 extracts machining surfaces from the three-dimensional shape extracted by the three-dimensional shape extraction unit 10 in step S102, and creates a machining surface list as shown in Figures 15(b) and 15(d).

[0064] In step S302, the three-dimensional shape rotation unit 72 rotates the three-dimensional shape extracted in step S102 in accordance with the machining axis direction instruction. In the fifth embodiment, in the initial determination process, the machining axis direction instruction is set to the Z axis in advance. The machining axis direction instruction may be input at the stage of step S101. As will be described later, in the fifth embodiment, if the determination result indicates that machining is not possible with the current machining axis direction instruction, the machining axis direction adjustment unit 78 changes the machining axis direction instruction, and in step S302, the three-dimensional shape is rotated in accordance with the changed machining axis direction instruction.

[0065] In step S304, based on the machining feasibility determination output by the inference model 16 in step S105, the machining surface / machining axis direction recording unit 74 records the machining surface and machining axis direction that are determined to be machineable in the machining surface list.

[0066] In step S305, the termination determination unit 76 determines whether all machining surfaces have been determined to be machineable. For example, as shown in Fig. 15(d), the termination determination unit 76 determines that all machining surfaces have been determined to be machineable when all columns in the machining axis direction of the machining surface list are filled with information on significant angles such as 0° or +90°. In step S305, if it is determined that all machining surfaces have been determined to be machineable, the procedure proceeds to step S306, and if it is determined that all machining surfaces have not been determined to be machineable, the procedure proceeds to step S307.

[0067] In step S306, the process ends by outputting the final machining feasibility decision of the three-dimensional shape and the machining axis direction of each machining surface to the output device 310. The information on the machining feasibility decision and the machining axis direction of each machining surface output in step S306 is, for example, a machining surface list showing the machining axis direction corresponding to each machining surface, as shown in FIG.

[0068] If it is determined in step S305 that not all machining surfaces are determined to be machineable, the termination determination unit 76 determines in step S307 whether the machining axis direction has been changed in all directions. If the material 40 is considered to be a rectangular parallelepiped, there are a total of six possible machining axis direction instructions. As in the second embodiment, the machining axis direction adjustment unit 78 registers, in the storage device 220, the Z axis set in the initial determination and the machining axis direction instruction changed in the subsequent step S308, among these six possible machining axis direction instructions. In step S307, the termination determination unit 76 determines whether machining feasibility has been considered in all machining axis directions by referring to the storage device 220. Alternatively, a separate column for machining axis change history may be provided in the machining surface list, and the termination determination unit 76 may determine whether machining feasibility has been considered in all machining axis directions by referring to this column.

[0069] If it is determined in step S307 that the machining axis direction has been changed in all directions, the procedure proceeds to step S108. In step S108, as in the first embodiment, a machining impossibility determination is output to the output device 310, and the process ends.

[0070] If it is determined in step S307 that the machining axis direction has not been changed in all directions, the procedure proceeds to step S308. In step S308, the machining axis direction adjustment unit 78 changes the machining axis direction instruction and inputs the changed machining axis direction instruction to the three-dimensional shape rotation unit 72.

[0071] In step S302, the three-dimensional shape rotation unit 72 rotates the three-dimensional shape in accordance with the changed machining axis direction instruction, and in the subsequent steps, the depth map is converted and a determination is made as to whether machining is possible.If it is determined in step S305 that all machining surfaces are machineable, the machining feasibility determination and the machining axis direction of each machining surface are output to the output device 310 in step S306, and the processing ends.

[0072] As described above, in the fifth embodiment, as in the second embodiment, even when the shape of the part changes depending on the machining axis direction, by making a judgment for all machining axis directions, it is possible to determine at what angle the part shape indicated by the design data cannot be machined. Furthermore, by clarifying not only whether the part can be machined but also from what direction each machining surface can be machined, it is possible to facilitate the setup for actual machining.

[0073] The inference section in the claims corresponds to the inference models 16 and 58 described in the detailed description of the invention.

[0074] 10 Three-dimensional shape extraction unit, 12 Three-dimensional shape rotation unit, 14 Depth map conversion unit, 16 Inference model, 20A Shape information DB, 20B Machining axis direction instruction DB, 20C Machining information DB, 22 Learning device, 24 Machining feasibility information DB, 26 Machining axis direction adjustment unit, 30, 32, 34 Tool, 40 Material, 54 Learning device, 56 Machining feasibility information DB, 58 Inference model, 70 Machining surface division unit, 72 Three-dimensional shape rotation unit, 74 Machining surface / machining axis direction recording unit, 76 End determination unit, 78 Machining axis direction adjustment unit, 100 Machining feasibility determination device, 110 Machining feasibility learning device, 120 Machining feasibility determination device, 130 Machining feasibility determination device, 140 Machining feasibility learning device, 150 Machining feasibility learning device, 160 Machining feasibility determination device

Claims

a depth map conversion unit that generates a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the desired machining axis direction, the depth map being an inference unit that is constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from the learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information indicating whether or not actual machining was possible in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data, and that determines whether or not the to-be-machined shape can be machined by inference using the desired machining axis direction instruction and the depth map generated by the depth map conversion unit.

2. The machining feasibility determination device of claim 1, further comprising a machining axis direction adjustment unit which outputs an instruction for a machining axis direction different from the desired machining axis direction when the inference unit outputs a determination that machining is impossible in the inference about the machining surface directly facing the desired machining axis direction, wherein the three-dimensional shape rotation unit rotates the three-dimensional shape so that a new machining surface when machining the three-dimensional shape in the machining axis direction indicated by the machining axis direction instruction input from the machining axis direction adjustment unit faces directly toward the desired machining axis direction, the depth map conversion unit orthogonally projects the new machining surface onto a plane perpendicular to the desired machining axis direction to generate a new depth map, and the inference unit uses the new depth map to determine whether the to-be-machined shape can be machined.

3. The machining feasibility determination device of claim 1, wherein the inference unit is constructed by machine learning using the learning machining axis direction instruction, a learning depth map extracted from the learning three-dimensional shape data in accordance with the learning machining axis direction instruction, learning machining information consisting of tool information including tool type, tool material, tool diameter, and tool length, information on the material of the object to be machined, and cutting parameters including the tool feed speed and tool rotation speed, and machining feasibility information describing whether actual machining performed in accordance with the learning machining axis direction instruction, the learning three-dimensional shape data, and the learning machining information, and determines whether the to-be-machined shape can be machined by inference using the desired machining axis direction, the depth map generated by the depth map conversion unit, and newly input machining information consisting of tool information including tool type, tool material, tool diameter, and tool length, information on the material of the object to be machined, and cutting parameters including the tool feed speed and tool rotation speed.

4. The machining feasibility determination device of claim 1, wherein the machining feasibility information is a heat map showing unsuitable machining locations on the machining surface, and the inference unit is constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and the heat map showing the feasibility of actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data, and outputs a determination result identifying unsuitable machining locations on the planned machining shape by inference using the desired machining axis direction and the depth map generated by the depth map conversion unit.

5. The present invention further comprises a machining surface division unit which creates a machining surface list by listing all machining surfaces of a machining object from the three-dimensional shape extracted by the three-dimensional shape extraction unit; a machining surface / machining axis direction recording unit which records in the machining surface list the machining axis directions of machining surfaces which are judged to be machineable in the machining feasibility judgment output by the inference unit; a termination judgment unit which makes an end judgment when a machining feasibility judgment has been made for all machining surfaces listed in the machining surface list, or when the machining feasibility in all machining axis directions has been verified, and outputs a machining feasibility judgment for the three-dimensional shape and the machining axis directions of each machining surface; and a machining axis direction adjustment unit which outputs an instruction for a machining axis direction other than the machining axis direction whose machining feasibility has been verified, when the termination judgment unit judges that the machining feasibility in all machining axis directions has not been verified; and the three-dimensional shape rotation unit rotates the three-dimensional shape so that a new machining surface when machining the three-dimensional shape in the machining axis direction indicated by the machining axis direction instruction input from the machining axis direction adjustment unit faces the desired machining axis direction, The machining feasibility determination device according to claim 1, wherein the depth map conversion unit generates a new depth map by orthogonally projecting the new machining surface onto a plane perpendicular to the desired machining axis direction, and the inference unit uses the new depth map to determine whether the planned machining shape can be machined.

6. A machining feasibility learning device comprising: a three-dimensional shape extraction unit which extracts a learning three-dimensional shape from input learning three-dimensional shape data; a three-dimensional shape rotation unit which rotates the three-dimensional shape so that the machining surface of the three-dimensional shape is directly facing the learning machining axis direction indicated by the input learning machining axis direction instruction; a depth map conversion unit which generates a learning depth map by adding depth information in the learning machining axis direction extracted from the learning three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the learning machining axis direction; and an inference unit which is updated by a learner that performs machine learning using the learning machining axis direction instruction, the learning depth map, and machining feasibility information describing whether actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data is possible or not, wherein the machining feasibility learning device constructs the machining feasibility determination device described in claim 1 by the machine learning.

7. A three-dimensional shape extraction unit that extracts a learning three-dimensional shape from input learning three-dimensional shape data; a three-dimensional shape rotation unit that rotates the three-dimensional shape so that a machining surface of the three-dimensional shape faces the learning machining axis direction indicated by the input learning machining axis direction instruction; a depth map conversion unit that generates a learning depth map by adding depth information in the learning machining axis direction extracted from the learning three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the learning machining axis direction; and an inference unit that is updated by a learner that performs machine learning using the learning machining axis direction instruction, the learning depth map, learning machining information consisting of tool information including tool type, tool material, tool diameter, and tool length, information on the material of the object to be machined, and cutting parameters including the tool feed rate and tool rotation speed, and machining feasibility information indicating whether or not actual machining performed in accordance with the learning machining axis direction instruction, the learning three-dimensional shape data, and the learning machining information, A machining acceptability learning device that constructs the machining acceptability determination device according to claim 3 by machine learning.

8. A machining feasibility learning device comprising: a three-dimensional shape extraction unit which extracts a learning three-dimensional shape from input learning three-dimensional shape data; a three-dimensional shape rotation unit which rotates the three-dimensional shape so that the machining surface of the three-dimensional shape is directly facing the learning machining axis direction indicated by the input learning machining axis direction instruction; a depth map conversion unit which generates a learning depth map by adding depth information in the learning machining axis direction extracted from the learning three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the learning machining axis direction; and an inference unit which is updated by a learner that performs machine learning using the learning machining axis direction instruction, the learning depth map, and a heat map representing improper machining locations in actual machining performed in accordance with the learning machining axis direction instruction, and the learning three-dimensional shape data, wherein the machining feasibility learning device constructs the machining feasibility determination device described in claim 4 by the machine learning.

9. A machining feasibility determination method executed by a computer, comprising: a step of extracting a three-dimensional shape from three-dimensional shape data indicating a shape to be machined; a step of rotating the three-dimensional shape so that a machining surface of the three-dimensional shape faces a desired machining axis direction; a step of generating a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the desired machining axis direction, the depth map being constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information describing whether actual machining can be performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data, and determining whether the shape to be machined can be machined by inference using the desired machining axis direction and the depth map.

10. A machining feasibility determination program that causes a computer to execute the steps of: extracting a three-dimensional shape from three-dimensional shape data indicating a shape to be machined; rotating the three-dimensional shape so that the machining surface of the three-dimensional shape faces a desired machining axis direction; generating a depth map by adding depth information in the desired machining axis direction extracted from the three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the desired machining axis direction, the depth map being constructed by machine learning using a learning machining axis direction instruction, a learning depth map extracted from learning three-dimensional shape data in accordance with the learning machining axis direction instruction, and machining feasibility information describing whether actual machining was possible in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data, and determining whether the shape to be machined can be machined by inference using the desired machining axis direction and the depth map.

11. A machining feasibility learning method executed by a computer, comprising: a step of extracting a learning three-dimensional shape from input learning three-dimensional shape data; a step of rotating the three-dimensional shape so that the machining surface of the three-dimensional shape is directly facing the learning machining axis direction indicated by the input learning machining axis direction instruction; a step of generating a learning depth map by adding depth information in the learning machining axis direction extracted from the learning three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the learning machining axis direction; and a step of updating an inference unit that judges whether or not a shape to be machined can be machined by a learning device that performs machine learning using the learning machining axis direction instruction, the learning depth map, and machining feasibility information describing whether or not actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data is possible; and a machining feasibility learning method that constructs a machining feasibility judgment device described in claim 1 by the machine learning.

12. A machining feasibility learning program for constructing a machining feasibility determination device as described in claim 1 by having a computer execute the following steps: extracting a learning three-dimensional shape from input learning three-dimensional shape data; rotating the three-dimensional shape so that the machining surface of the three-dimensional shape is directly facing the learning machining axis direction indicated by the input learning machining axis direction instruction; generating a learning depth map by adding depth information in the learning machining axis direction extracted from the learning three-dimensional shape data to a two-dimensional image obtained by orthogonally projecting the machining surface onto a plane perpendicular to the learning machining axis direction; and updating an inference unit that determines whether or not a shape to be machined can be machined by a learning device that performs machine learning using the learning machining axis direction instruction, the learning depth map, and machining feasibility information describing whether or not actual machining performed in accordance with the learning machining axis direction instruction and the learning three-dimensional shape data is possible.

Images