Data generation method, computer program, recording medium, and processing device

Abstract

The data generation method of the present invention includes the following steps: acquiring test processing condition information, the test processing condition information including the number of pulses to be irradiated at each of the multiple irradiation positions of the test workpiece in order to process the test workpiece into a target shape; measuring the shape of the test workpiece after processing based on the test processing condition information, i.e., the test workpiece shape; and calculating the target number of pulses to be irradiated at the multiple irradiation positions of the workpiece based on the test processing condition information, the test workpiece shape, and the predicted information of the shape of the part processed by irradiating a pulse beam of a unit number of pulses.

Description

Technical Field

[0001] This invention relates to the technical field of a data generation method, a computer program, a recording medium, and a processing apparatus. The data generation method generates control data to control a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. Background Technology

[0002] Patent document 1 describes a processing apparatus that processes an object by irradiating it with a laser. In this processing apparatus, it is required that the object be processed appropriately.

[0003] Background Technology Documents

[0004] Patent documents

[0005] Patent Document 1: U.S. Patent Application Publication No. 2002 / 0017509. Summary of the Invention

[0006] According to the first state, a data generation method is provided, which generates control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: acquiring test processing condition information, the test processing condition information including the number of pulses of the pulsed energy beam to be irradiated at each of multiple irradiation positions of the test workpiece in order to process the test workpiece into a target shape; measuring the shape of the test workpiece after processing it based on the test processing condition information, i.e., the test workpiece shape; and calculating the target number of pulses of the pulsed energy beam to be irradiated at the multiple irradiation positions of the workpiece based on the test processing condition information, the test workpiece shape, and prediction information of the shape of the portion processed by irradiating the pulsed energy beam with a unit number of pulses.

[0007] According to the second state, a data generation method is provided, which generates control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: inputting test processing condition information, which indicates the number of pulses of the pulsed energy beam to be irradiated at multiple irradiation positions of the test workpiece to process it into a target shape; measuring the shape of the test workpiece related to the shape of the test workpiece after processing with the pulsed energy beam based on the test processing condition information; performing machine learning for generating model information based on the test processing condition information and the shape of the test workpiece, the model information including prediction information of the shape processed by the pulsed energy beam with a unit number of pulses; and calculating the target number of pulses of the pulsed energy beam to be irradiated at multiple irradiation positions of the workpiece based on the model information generated by the machine learning and the target shape.

[0008] According to the third state, a data generation method is provided, which generates control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: acquiring model information representing the relationship between target pulse number information and target shape information, wherein the target pulse number information represents the target number of pulses that should be irradiated at each of multiple irradiation positions of the workpiece, and the target shape information is information about the target shape of the workpiece after removal; and generating the target pulse number information as the control data based on the model information and the target shape information.

[0009] According to the fourth state, a data generation method is provided, which generates control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: acquiring model information that can be used to make predictions, namely: a predicted value of the amount of material removed from the workpiece when the pulsed energy beam is irradiated at a first irradiation position among a plurality of irradiation positions of the workpiece with a unit number of pulses; a predicted value of the amount of material removed at the first irradiation position; and a predicted value of the amount of material removed at a position different from the first irradiation position; and generating target pulse number information as the control data based on the model information and target shape information related to the target shape of the workpiece after removal, wherein the target pulse number information represents the target number of pulses that should be irradiated at each of the plurality of irradiation positions of the workpiece.

[0010] According to the fifth state, a data generation method is provided, which generates control data for controlling a processing device capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: acquiring test processing condition information for processing a test workpiece into a target shape; measuring the shape of the test workpiece after processing it based on the test processing condition information, i.e., the test workpiece shape; and calculating the processing conditions of the workpiece based on the test processing condition information, the test workpiece shape, and prediction information of the shape processed by the pulsed energy beam.

[0011] According to the sixth state, a data generation method is provided, which generates control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam. The data generation method includes the following steps: generating first test processing condition information, the first test processing condition information indicating the number of pulses of the pulsed energy beam to be irradiated at each of multiple irradiation positions of a first test workpiece; acquiring first test shape information, the first test shape information being information about the shape of the first test workpiece after removal processing by irradiating the first test workpiece with the pulsed energy beam based on the first test processing condition information; generating... The process involves: obtaining second test processing condition information, which indicates the number of pulses of the pulsed energy beam to be irradiated at each of the multiple irradiation positions of the second test workpiece in order to process the second test workpiece into a target shape; obtaining second test shape information, which is information about the shape of the second test workpiece after removal processing by irradiating the second test workpiece with the pulsed energy beam based on the second test processing condition information; and calculating the target number of pulses of the pulsed energy beam to be irradiated at the multiple irradiation positions of the workpiece based on the first test shape information, the second test shape information, the second test processing conditions, and the target shape.

[0012] According to the seventh state, a computer program is provided that causes a computer to execute the data generation method provided in any one of the first to sixth states.

[0013] According to the eighth state, a recording medium is provided, which records the computer program provided by the seventh state.

[0014] According to the ninth state sample, a processing apparatus is provided that uses the control data generated by the data generation method provided by any one of the first to sixth state samples to perform removal processing on the workpiece to be processed. Attached Figure Description

[0015] Figure 1 This is a block diagram showing the overall configuration of the processing system in this embodiment.

[0016] Figure 2 This is a perspective view schematically showing the appearance of the processing apparatus of this embodiment.

[0017] Figure 3 This is a system configuration diagram showing the system configuration of the processing apparatus in this embodiment.

[0018] Figure 4 It is a three-dimensional diagram showing the structure of the illumination optical system.

[0019] Figure 5 (a) to Figure 5 (c) are cross-sectional views showing the state of the removal process performed on the workpiece.

[0020] Figure 6 This refers to the workpiece before and after the removal process.

[0021] Figure 7 This indicates the part of the workpiece that should be removed.

[0022] Figure 8 It is a cross-sectional view showing multiple removal layers that should be removed from the workpiece.

[0023] Figure 9 (a) to Figure 9 (d) are top views of the areas on the surface of workpiece W that are removed during the removal process of each removal layer.

[0024] Figure 10 (a) and Figure 10 (b) are top views showing the position of the irradiated target on the surface of the workpiece.

[0025] Figure 11 This is a block diagram representing the structure of the data generation server.

[0026] Figure 12 It is a flowchart representing the process of data generation.

[0027] Figure 13 (a) is a perspective view showing an example of a ribbed structure. Figure 13 (b) is a cross-sectional view showing an example of a ribbed structure.

[0028] Figure 14 Conceptually represents the state of machine learning used to generate model information.

[0029] Figure 15 An example of a conceptual representation of model information.

[0030] Figure 16 An example of a conceptual representation of predictive information.

[0031] Figure 17 Conceptually, this represents the state of calculating the target pulse number based on model information. Detailed Implementation

[0032] Hereinafter, embodiments of the data generation method, computer program, recording medium, and processing apparatus will be described with reference to the accompanying drawings. The following description uses a processing system SYS that processes a workpiece W using a processing light EL to illustrate embodiments of the data generation method, computer program, recording medium, and processing apparatus. However, the present invention is not limited to the embodiments described below.

[0033] Furthermore, in the following description, an XYZ orthogonal coordinate system defined by mutually orthogonal X, Y, and Z axes is used to explain the positional relationships of the various components constituting the machining system SYS. For ease of explanation, the X and Y axes are assumed to be horizontal (i.e., predetermined directions within the horizontal plane), and the Z axis is assumed to be vertical (i.e., a direction orthogonal to the horizontal plane, essentially a vertical direction). The directions of rotation (in other words, tilt directions) around the X, Y, and Z axes are referred to as θX, θY, and θZ, respectively. Alternatively, the Z-axis can be considered the direction of gravity. The XY plane can also be considered horizontal.

[0034] (1) Overall composition of the machining system SYS

[0035] First, refer to Figure 1 The overall structure of the machining system SYS is described. Figure 1 This is a block diagram representing the overall structure of the machining system SYS.

[0036] like Figure 1 As shown, the machining system SYS includes a machining device 1 and a data generation server 2. Furthermore, the machining system SYS includes a client device 3. However, the machining system SYS may also lack a client device 3. The machining device 1, the data generation server 2, and the client device 3 can communicate with each other via a communication network 4, which includes at least one of a wired communication network and a wireless communication network. In this case, the data generation server 2 can also function as a cloud server for the machining device 1 and the client device 3. When the data generation server 2 can function as a cloud server, it can also be referred to as a cloud server or a cloud system. However, at least one of the machining device 1, the data generation server 2, and the client device 3 may be unable to communicate with at least another of the machining device 1, the data generation server 2, and the client device 3.

[0037] The processing apparatus 1 is capable of processing the object (i.e., the workpiece) W (refer to) Figure 2 The workpiece W is processed. Furthermore, the workpiece W can also be called the workpiece to be processed. The workpiece W can be, for example, a metal, an alloy (e.g., duralumin), a semiconductor (e.g., silicon), a resin, a composite material such as CFRP (Carbon Fiber Reinforced Plastic), glass, ceramics, or any other material. As an example of any material, at least one of gypsum, polyurethane, rubber, and elastomers can be listed. Alternatively, a portion of the workpiece W can be made of one material, while other portions are made of a different material.

[0038] In this embodiment, an example of removal processing of workpiece W by processing apparatus 1 will be described. In particular, in this embodiment, an example of removal processing by processing apparatus 1 by irradiating workpiece W with processing light EL will be described. However, processing apparatus 1 may also perform processing on workpiece W different from removal processing. For example, processing apparatus 1 may also perform additional processing on workpiece W.

[0039] The processing apparatus 1 can form a rib-like structure on the workpiece W by processing it. The rib-like structure can be a structure capable of reducing the surface resistance of the workpiece W relative to the fluid (especially at least one of frictional resistance and turbulent frictional resistance). The rib-like structure can include a structure capable of reducing noise generated when the fluid moves relative to the surface of the workpiece W. For example, the rib-like structure can include a structure formed by arranging a plurality of grooves extending along a first direction (e.g., the Y-axis) along a second direction (e.g., the X-axis direction), the first direction being along the surface of the workpiece W, and the second direction being along the surface of the workpiece W and intersecting the first direction.

[0040] The processing apparatus 1 can form an arbitrary structure of arbitrary shape on the surface of the workpiece W by processing it. As an example of an arbitrary structure, a structure that causes vortices to form due to the flow of fluid on the surface of the workpiece W can be exemplified. As another example of an arbitrary structure, a structure used to impart hydrophobicity to the surface of the workpiece W can be exemplified. As yet another example of an arbitrary structure, a micro- or nanoscale fine texture structure (typically an uneven structure) formed regularly or irregularly can be exemplified. Such a fine texture structure can include at least one of a sharkskin structure and a pitted structure that reduces fluid (gas and / or liquid) resistance. The fine texture structure can also include a lotus leaf surface structure that has at least one of a liquid-repellent function and a self-cleaning function (e.g., exhibiting a lotus effect). The microtexture structure may also include at least one of the following structures: a microprotrusion structure with liquid transport function (refer to U.S. Patent Publication No. 2017 / 0044002), a convex-concave structure with hydrophilic function, a convex-concave structure with anti-fouling function, a moth-eye structure with at least one of reflectivity reduction function and liquid repellency function, a convex-concave structure that exhibits structural color by interfering to enhance light of only a specific wavelength, a column array structure with adhesive function utilizing van der Waals forces, a convex-concave structure with aerodynamic noise reduction function, a honeycomb structure with droplet capture function, and a convex-concave structure that improves adhesion to layers formed on the surface, etc.

[0041] Data generation server 2 is capable of generating control data for controlling processing device 1. The control data can be any data, as long as it can be used to control processing device 1. For example, the control data can include data that directly or indirectly specifies the processing conditions of processing device 1. For example, the control data can also include data that directly or indirectly specifies the actions of processing device 1. For example, the control data can also include data that can directly control processing device 1 (e.g., instruction data). For example, the control data can also include data that can be used to generate data actually used to control processing device 1 (e.g., slicing data, described later).

[0042] To generate control data, the data generation server 2 can obtain (i.e., receive) information (hereinafter referred to as "reference information") referenced by the data generation server 2 for generating control data from at least one of the processing device 1 and the client device 3 via the communication network 4. The reference information may, for example, include information related to the processing light EL used by the processing device 1 to process the workpiece W. The information related to the processing light EL may include at least one of the following: information related to the intensity of the processing light EL, information related to the shape of the processing light EL in a plane intersecting the direction of travel (in other words, the irradiation direction) of the processing light EL, information related to the light intensity distribution of the processing light EL in a plane intersecting the direction of travel of the processing light EL, information related to the luminous flux of the processing light EL, and information related to the luminous flux distribution of the processing light EL. The reference information may also, for example, include device identification information for identifying the processing device 1. The reference information may also, for example, include workpiece information for identifying the workpiece W processed by the processing device 1. The reference information may also, for example, include information related to the material of the workpiece W processed by the processing device 1. Information related to the material of workpiece W may include information related to the type of material constituting workpiece W (e.g., the type of metal material). Reference information may also include, for example, information related to the shape of workpiece W before removal processing by processing apparatus 1. Reference information may also include, for example, information related to the shape of workpiece W after removal processing by processing apparatus 1. Furthermore, in the following description, the shape of workpiece W after removal processing by processing apparatus 1 is referred to as the target shape. Reference information may also include information related to the processing quality of processing apparatus 1 (processing quality information). Processing quality information may include at least one of the following: information related to the processing resolution of processing apparatus 1, information related to the surface roughness of workpiece W after processing by processing apparatus 1, and information related to the processing accuracy of processing apparatus 1. Reference information may also include information related to the processing volume of processing apparatus 1 (processing volume information).

[0043] Data generation server 2 can obtain all the required reference information from client device 3. Data generation server 2 can also obtain all the required reference information from processing device 1. Data generation server 2 can also obtain a portion of the required reference information from client device 3 and another portion from processing device 1. Data generation server 2 can also obtain reference information from processing device 1 via client device 3. Data generation server 2 can also obtain reference information from client device 3 via processing device 1.

[0044] Data generation server 2 can store the acquired reference information. Furthermore, information related to the processing optical EL and identification information used to identify the processing device 1 are both related to the processing device 1. In this case, data generation server 2 can store the information related to the processing optical EL and the identification information in a state that establishes an association between the information related to the processing optical EL and the identification information. Alternatively, reference information can be obtained from at least one of the processing device 1 and the client device 3 via communication network 4 by a server different from data generation server 2 (e.g., a cloud server), and the acquired reference information can be stored. In this case, data generation server 2 can obtain the reference information from the cloud server storing the reference information.

[0045] Data generation server 2 can generate control data based on the obtained reference information. For example, data generation server 2 can generate control data by performing calculations based on the obtained reference information. As an example, data generation server 2 can calculate processing conditions based on the obtained reference information and generate control data to control processing device 1 in a manner that causes processing device 1 to operate according to the calculated processing conditions. As an example of the calculation, deconvolution operation can be exemplified. Alternatively, for example, if multiple candidate control data are prepared, data generation server 2 can select one control data based on the obtained reference information. As an example, data generation server 2 can select one processing condition from multiple pre-prepared candidate processing conditions based on the obtained reference information and select (generate) control data to control processing device 1 in a manner that causes processing device 1 to operate according to the selected processing condition. That is, in this embodiment, the action of generating control data can include at least one of the following actions: the action of regenerating control data through calculation, and the action of selecting pre-prepared control data. Furthermore, processing conditions can also be referred to as processing recipes.

[0046] Data generation server 2 can be located at a location where processing device 1 is located. Data generation server 2 can also be located at a different location than the location where processing device 1 is located. Data generation server 2 can be located at a location where client device 3 is located. Data generation server 2 can also be located at a different location than the location where client device 3 is located. As an example, data generation server 2 can be located in a different business than the business where at least one of processing device 1 and client device 3 is located. As another example, data generation server 2 can be located in a different country than the country where at least one of processing device 1 and client device 3 is located.

[0047] Client device 3 is a terminal device that can be used by the user of processing device 1. Client device 3 may include, for example, at least one of a personal computer, a smartphone, and a tablet terminal.

[0048] (2) Processing device 1

[0049] Next, the machining device 1 of the machining system SYS will be described.

[0050] (2-1) Composition of processing device 1

[0051] First, refer to Figure 2 and Figure 3 The configuration of processing device 1 will be described. Figure 2 This is a three-dimensional view showing the configuration of the processing device 1. Figure 2 This is a block diagram showing the system configuration of processing device 1.

[0052] like Figure 2 and Figure 3 As shown, the processing apparatus 1 includes a processing unit 11, a measuring unit 12, a platform unit 13, and a control unit 14. Furthermore, the processing unit 11, measuring unit 12, platform unit 13, and control unit 14 may also be referred to as a processing apparatus, a measuring apparatus, a platform apparatus, and a control apparatus, respectively. The processing unit 11, measuring unit 12, and platform unit 13 are housed in a housing 15. However, at least a portion of the processing unit 11, measuring unit 12, and platform unit 13 may not be housed in the housing 15. The processing apparatus 1 may also lack a housing 15 for housing the processing unit 11, measuring unit 12, and platform unit 13.

[0053] The processing unit 11, under the control of the control unit 14, removes workpiece W by irradiating it with processing light EL. Specifically, the processing unit 11, under the control of the control unit 14, removes workpiece W by irradiating its surface with processing light EL. For processing workpiece W, the processing unit 11 includes a processing light source 111, a processing head 112, and a head drive system 113.

[0054] Under the control of the control unit 14, the processing light source 111 emits at least one of infrared light, visible light, ultraviolet light, and extreme ultraviolet light as the processing light EL. However, other types of light may also be used as the processing light EL.

[0055] The processing light EL comprises pulsed light (i.e., multiple pulse beams). Furthermore, since light is an example of an energy beam, pulsed light can also be referred to as a pulsed energy beam. In this case, the processing light source 111 can emit pulsed light with pulse widths on the order of femtoseconds, picoseconds, or nanoseconds as the processing light EL. However, the processing light EL may also not contain pulsed light. For example, the processing light EL can also be continuous light.

[0056] The processing light EL can be a laser. In this case, the processing light source 111 can include a laser source (e.g., a semiconductor laser such as a laser diode (LD)). The laser source can include at least one of a fiber laser, a CO2 laser, a YAG laser, and an excimer laser. However, the processing light EL may not be a laser. The processing light source 111 can include any light source (e.g., at least one of an LED (Light Emitting Diode) and a discharge lamp).

[0057] Under the control of the control unit 14, the machining head 112 processes the workpiece W by irradiating the workpiece W with the machining light EL emitted from the machining light source 111. The machining head 112 performs removal processing on the workpiece W placed on the platform 132 (described later). That is, removal processing is performed on the platform 132. In this case, the machining head 112 can be positioned above the platform 132 on which the workpiece W is placed. For example, the machining head 112 can be mounted on a door-shaped support frame 16, which is positioned on the pressure plate 131 of the platform unit 13. The support frame 16 can include a pair of leg members 161 protruding from the pressure plate 131 along the Z-axis direction, and a beam member 162 connecting the pair of leg members 161 via their upper ends. The beam member 162 can be positioned above the platform 132. The machining head 112 can be mounted on this beam member 162. Furthermore, in Figure 2 In the example shown, the machining head 112 is mounted on the beam member 162 via the head drive system 113 described later. With the machining head 112 positioned above the platform 132, it can illuminate the workpiece W by emitting machining light EL downwards from the machining head 112. In other words, the machining head 112 can illuminate the workpiece W by emitting machining light EL that travels along the Z-axis.

[0058] To illuminate the workpiece W with machining light EL, the machining head 112 is equipped with an illumination optical system 1120. Here, see... Figure 4 The illumination optical system 1120 will be described. Figure 4 This is a three-dimensional diagram showing the configuration of the illumination optical system 1120.

[0059] like Figure 4 As shown, the illumination optical system 1120 may include, for example, a focusing optical system 1121, a galvanometer lens 1122, and an fθ lens 1123.

[0060] The focusing optical system 1121 is an optical component capable of changing the focal position (i.e., the convergence position) of the processing light EL along the forward direction of the processing light EL. The focusing optical system 1121 may, for example, include a plurality of lenses arranged along the forward direction of the processing light EL. In this case, the focal position of the processing light EL can be changed by moving at least one of the plurality of lenses along its optical axis.

[0061] The processing light EL passing through the focusing optical system 1121 is incident on the galvanometer mirror 1122. The galvanometer mirror 1122 changes the emission direction of the processing light EL from the galvanometer mirror 1122 by deflecting the processing light EL (i.e., changing the emission angle of the processing light EL). If the emission direction of the processing light EL from the galvanometer mirror 1122 changes, then the position of the processing light EL emitted from the processing head 112 changes. If the position of the processing light EL emitted from the processing head 112 changes, then the position of the target irradiated area EA on the surface of the workpiece W that is irradiated by the processing light EL changes.

[0062] The galvanometer mirror 1122 includes, for example, an X-scanning mirror 1122X and a Y-scanning mirror 1122Y. The X-scanning mirror 1122X and the Y-scanning mirror 1122Y are tilt-angle variable mirrors capable of changing the angle relative to the optical path of the processing light EL incident on each mirror. The X-scanning mirror 1122X reflects the processing light EL toward the Y-scanning mirror 1122Y. The X-scanning mirror 1122X can oscillate or rotate about a rotation axis along the Y-axis. By oscillating or rotating the X-scanning mirror 1122X, the processing light EL scans the surface of the workpiece W along the X-axis. By oscillating or rotating the X-scanning mirror 1122X, the target illumination area EA moves along the X-axis on the surface of the workpiece W. The Y-scanning mirror 1122Y reflects the processing light EL toward the fθ lens 1123. The Y-scanning mirror 1122Y can oscillate or rotate about a rotation axis along the X-axis. By oscillating or rotating the Y-scanning mirror 1122Y, the processing light EL scans the surface of the workpiece W along the Y-axis. By swinging or rotating the Y-scanning mirror 1122Y, the target irradiation area EA moves along the Y-axis on the surface of the workpiece W.

[0063] Through this galvanometer lens 1122, the processing light EL can scan or sweep across the processing area PSA determined based on the processing head 112. That is, through the galvanometer lens 1122, the target irradiation area EA can move within the processing area PSA determined based on the processing head 112. Furthermore, the processing area PSA represents the area (in other words, range) removed by the processing head 112 when the positional relationship between the processing head 112 and the workpiece W is fixed (i.e., without change). Typically, the processing area PSA is set to be the same as, or narrower than, the scanning range of the processing light EL biased by the galvanometer lens 1122 when the positional relationship between the processing head 112 and the workpiece W is fixed. Furthermore, by moving the processing head 112 using the head drive system 113 (described later) and / or by moving the platform 132 using the platform drive system 133 (described later), the processing area PSA (target irradiation area EA) can be moved relative to the surface of the workpiece W.

[0064] Furthermore, depending on the height of the workpiece W's surface, the machining head 112 can be moved along the Z-axis direction (the direction intersecting the surface of the workpiece W) via the head drive system 113, the platform 132 can be moved along the Z-axis direction via the platform drive system 133, or the focusing optical system 1121 can be used to change the focusing position. At least two of these three methods can also be used simultaneously.

[0065] The fθ lens 1123 is an optical system for directing the processing light EL from the galvanometer lens 1122 toward the workpiece W. Specifically, the fθ lens 1123 is an optical element capable of focusing the processing light EL from the galvanometer lens 1122 onto a focusing surface. Therefore, the fθ lens 1123 can also be referred to as a focusing optical system or an objective lens optical system. The focusing surface of the fθ lens 1123 can, for example, be set to the surface of the workpiece W. Alternatively, the focusing surface of the fθ lens 1123 can be set to a surface that moves away from the surface of the workpiece W along the optical axis AX of the fθ lens 1123. Furthermore, the focusing surface of the fθ lens 1123 can also be set to include the rear focal point position of the fθ lens 1123. In this case, the galvanometer lens 1122 can also be positioned at the front focal point position of the fθ lens 1123. When the galvanometer lens 1122 is equipped with multiple scanning mirrors (e.g., X scanning mirror 1122X and Y scanning mirror 1122Y), the front focal position of the fθ lens 1123 can also be set among the multiple scanning mirrors.

[0066] Back to Figure 2 and Figure 3In this process, the head drive system 113, under the control of the control unit 14, moves the processing head 112 (especially the irradiation optical system 1120) along at least one of the X-axis, Y-axis, Z-axis, θX, θY, and θZ directions. Therefore, the head drive system 113 can also be referred to as a moving device. Figure 2 An example is shown where the head drive system 113 moves the machining head 112 along the Z-axis. In this case, the head drive system 113 may, for example, include a Z-slider member 1131 extending along the Z-axis. The Z-slider member 1131 is disposed on a support frame 16 via a vibration damping device, and the support frame 16 is disposed on a pressure plate 131. The Z-slider member 1131 is disposed on a beam member 162, for example, via a support member 163 extending along the Z-axis. The Z-slider member 1131 connects to the machining head 112 and enables it to move along the Z-slider member 1131.

[0067] If the machining head 112 moves, the positional relationship between the machining head 112 and the platform 132 (described later) changes. Furthermore, if the machining head 112 moves, the positional relationship between the machining head 112 and the workpiece W placed on the platform 132 changes. Therefore, moving the machining head 112 can be considered equivalent to changing the positional relationship between the machining head 112 and the platform 132 and the workpiece W respectively. Additionally, if the machining head 112 moves, the target irradiation area EA and the machining area PSA on the surface of the workpiece W, which are irradiated by the machining light EL, move relative to the surface of the workpiece W.

[0068] The measuring unit 12 is capable of measuring the object under the control of the control unit 14. In order to measure the object, the measuring unit 12 includes a measuring head 121 and a head drive system 122.

[0069] The measuring head 121, under the control of the control unit 14, can measure (in other words, determine) the object to be measured. Specifically, the measuring head 121 can measure any characteristic of the object to be measured. As an example of a characteristic of the object to be measured, the position of the object to be measured can be cited. As another example of a characteristic of the object to be measured, the shape of the object to be measured (e.g., a two-dimensional shape or a three-dimensional shape) can be cited. As another example of a characteristic of the object to be measured, at least one of the reflectivity, transmittance, and surface roughness of the object to be measured can be cited.

[0070] The object to be measured may include a workpiece W. Specifically, the object to be measured may include at least one of the following workpieces: a workpiece W that has not undergone removal processing by the processing unit 11, a workpiece W that is being removed by the processing unit 11, and a workpiece W that has been removed by the processing unit 11. The object to be measured may also include a platform 132 capable of holding the workpiece W.

[0071] The measuring head 121 can measure the object to be measured using any measurement method. For example, the measuring head 121 can perform optical, electrical, magnetic, physical, chemical, or thermal measurements on the object to be measured. The measuring head 121 can measure the object to be measured without contacting it. The measuring head 121 can also measure the object to be measured by contacting it. In this embodiment, an example of optical measurement of the object to be measured by illuminating the object with measuring light ML without contacting it will be described. For example, the measuring head 121 can use optical sectioning to measure the object to be measured, which is a method of projecting measuring light ML, which is a slit light, onto the surface of the object to be measured, and simultaneously measuring the shape of the projected slit light. For example, the measuring head 121 can also use white light interferometry to measure the object to be measured, which is a method of measuring the interference pattern between white light after passing through the object to be measured, i.e., measuring light ML, and white light before the object to be measured. For example, the measuring head 121 can also use a pattern projection method to measure the object, which involves projecting a measuring light ML depicting a light pattern onto the surface of the object and measuring the shape of the projected pattern. Alternatively, the measuring head 121 can also use a time-of-flight method to measure the object, which involves performing the following actions at multiple locations on the object: emitting a measuring light ML onto the surface of the object and determining the distance to the object based on the time it takes for the emitted measuring light ML to return. For example, the measuring head 121 can also use at least one of the following methods to measure the object: Moiré topography (specifically, grating illumination or grating projection), holographic interferometry, autocollimation, stereochemistry, astigmatism, critical angle method, edge method, interferometry, and confocal method.

[0072] The measuring head 121 can be positioned above the platform 132 on which the workpiece W is placed. Specifically, the measuring head 121 can be mounted on the beam member 162 in the same manner as the machining head 112. Furthermore, in Figure 2 In the example shown, the measuring head 121 is mounted on the beam member 162 via the head drive system 122. When the measuring head 121 is positioned above the platform 132, it can measure the workpiece W from above. The measuring head 121 can also measure the platform 132 from above. When the measuring head 121 is positioned above the platform 132, it can illuminate at least one of the workpiece W and the platform 132 by emitting a measuring light ML downwards from the measuring head 121. That is, the measuring head 121 can illuminate at least one of the workpiece W and the platform 132 by emitting a measuring light ML that travels along the Z-axis.

[0073] Furthermore, the measuring head 121 may include multiple measuring instruments capable of measuring the object to be measured separately. These multiple measuring instruments may include at least two measuring instruments with different measurement resolutions (in other words, different measurement accuracies). Alternatively, the multiple measuring instruments may include at least two measuring instruments with different measurement area sizes.

[0074] Under the control of the control unit 14, the head drive system 122 moves the measuring head 121 along at least one of the X-axis, Y-axis, Z-axis, θX, θY, and θZ directions. Therefore, the head drive system 122 can also be referred to as a moving device. Figure 2 An example is shown where the head drive system 122 moves the measuring head 121 along the Z-axis. In this case, the head drive system 122 may, for example, include a Z-slider component 1221 extending along the Z-axis. The Z-slider component 1221 is disposed on a support frame 16 via a vibration damping device, and the support frame 16 is disposed on a pressure plate 131. The Z-slider component 1221 is disposed on a beam component 162, for example, via a support component 164 extending along the Z-axis. The Z-slider component 1221 connects to the measuring head 121 and enables it to move along the Z-slider component 1221.

[0075] If the measuring head 121 moves, the positional relationship between the measuring head 121 and the platform 132 (described later) changes. Furthermore, if the measuring head 121 moves, the positional relationship between the measuring head 121 and the workpiece W placed on the platform 132 changes. Therefore, moving the measuring head 121 can also be considered equivalent to changing the respective positional relationships between the measuring head 121, the platform 132, and the workpiece W.

[0076] Platform unit 13 includes pressure plate 131, platform 132 and platform drive system 133.

[0077] The pressure plate 131 is disposed on the bottom surface of the housing 15 (or on a supporting surface such as the base plate of the housing 15). A platform 132 is disposed on the pressure plate 131. A vibration damping device (not shown) may be provided between the bottom surface of the housing 15 or the supporting surface such as the base plate of the housing 15 and the pressure plate 131 to reduce the transmission of vibration from the pressure plate 131 to the platform 132. Furthermore, the support frame 16 described above may also be disposed on the pressure plate 131. Additionally, a foot member may be provided between the pressure plate 131 and the bottom surface of the housing 15 (or on a supporting surface such as the base plate of the housing 15). In this case, a vibration damping device may be provided between the foot member and the pressure plate 131 and / or between the foot member and the bottom surface (or supporting surface).

[0078] Platform 132 is a mounting device for placing workpiece W. Platform 132 may be able to hold the workpiece W placed on it. Alternatively, platform 132 may not be able to hold the workpiece W placed on it. In this case, workpiece W can be placed on platform 132 without the use of a fixture. When platform 132 is able to hold workpiece W, it may be equipped with at least one of a mechanical chuck, an electrostatic chuck, or a vacuum chuck to hold workpiece W.

[0079] Under the control of the control unit 14, the platform drive system 133 moves the platform 132. For example, under the control of the control unit 14, the platform drive system 133 can move the platform 132 in at least one of the following directions: X-axis, Y-axis, Z-axis, θX direction, θY direction, and θZ direction. Furthermore, the platform drive system 133 can also be referred to as a moving device.

[0080] exist Figure 2 In the example shown, the platform drive system 133 causes the platform 132 to move along the X-axis and Y-axis, respectively. That is, in... Figure 2 In the example shown, the platform drive system 133 moves the platform 132 along a direction intersecting the respective forward directions of the processing light EL and the measurement light ML in the XY plane. In this case, the platform drive system 133 may, for example, include an X-sliding member 1331 extending along the X-axis direction (in... Figure 2 In the example shown, two X-sliding components 1331 are arranged parallel to each other, and a Y-sliding component 1332 extends along the Y-axis direction (in... Figure 2 In the example shown, one Y-sliding member 1332 is used. Two X-sliding members 1331 are arranged on the pressure plate 131 in a manner aligned along the Y-axis. The Y-sliding member 1332 is connected to the two X-sliding members 1331 in a manner movable along them. The platform 132 is connected to the Y-sliding member 1332 in a manner movable along it. Furthermore, in Figure 2 In the example, multiple X-sliding components 1331 are provided, but there can also be only one X-sliding component 1331. Alternatively, the platform 132 can be suspended and supported on the pressure plate 131 by means of an air bearing.

[0081] If the platform drive system 133 moves the platform 132, the positional relationships between the machining head 112 and the measuring head 121 and the platform 132 and the workpiece W respectively change. Therefore, the platform drive system 133 can also be considered as a position changing device capable of changing the positional relationships between the machining head 112 and the measuring head 121 and the platform 132 and the workpiece W respectively. Furthermore, if the platform drive system 133 moves the platform 132, the platform 132 and the workpiece W respectively move relative to the machining area PSA where the machining head 112 performs the removal machining and the measurement area where the measuring head 121 performs the measurement.

[0082] Control unit 14 controls the operation of processing device 1. For example, control unit 14 can generate processing control information (e.g., processing path information) for processing workpiece W, and control processing unit 11 and platform unit 13 based on the processing control information to process workpiece W according to the generated processing control information. That is, control unit 14 can control the processing of workpiece W. For example, control unit 14 can generate measurement control information for measuring an object to be measured, and control processing unit 11 and platform unit 13 based on the measurement control information to measure the object to be measured according to the generated measurement control information. That is, control unit 14 can control the measurement of the object to be measured.

[0083] The control unit 14 can generate at least one of processing control information and measurement control information based on the control data generated by the data generation server 2. In this case, the control unit 14 can obtain control data from the data generation server 2 via the communication network 4. Alternatively, the control unit 14 can also obtain control data from the client device 3, which has already obtained control data from the data generation server 2, via the communication network 4.

[0084] The control unit 14 may include, for example, a computing device 141 and a storage device 142. The computing device 141 may include, for example, at least one of a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The control unit 14 functions as a device for controlling the operation of the processing device 1 by causing the computing device 141 to execute a computer program. This computer program is a computer program for causing the control unit 14 (e.g., the computing device 141) to perform (i.e., execute) the actions described later that the control unit 14 should perform. In other words, this computer program is a computer program for causing the control unit 14 to function in a manner that causes the processing device 1 to perform the actions described later. The computer program executed by the computing device 141 may be recorded in the storage device 142 (i.e., a recording medium) of the control unit 14, or it may be recorded on any storage medium (e.g., a hard disk or semiconductor memory) that can be built into or installed outside the control unit 14. Alternatively, the computing device 141 may also download the computer program to be executed from a device outside the control unit 14 via a network interface.

[0085] The control unit 14 may not be located inside the processing apparatus 1. For example, the control unit 14 may be located outside the processing apparatus 1 as a server or the like. For example, the control unit 14 may be a computer (e.g., a laptop computer) that can be connected to the processing apparatus 1. For example, the control unit 14 may be a computer (e.g., a laptop computer) located near the processing apparatus 1. In this case, the control unit 14 and the processing apparatus 1 can be connected via a wired and / or wireless network (or a data bus and / or communication loop). As a wired network, a network using a serial bus interface, such as at least one of IEEE 1394, RS-232x, RS-422, RS-423, RS-485, and USB, may be used. As a wired network, a network using a parallel bus interface may also be used. As a wired network, a network using an Ethernet interface (registered trademark), such as at least one of 10BASE-T, 100BASE-TX, and 1000BASE-T, may also be used. As a wireless network, a network using radio waves may be used. As an example of a network using radio waves, a network based on IEEE 802.1x (e.g., at least one of wireless LAN and Bluetooth (registered trademark)) can be cited. An infrared network can also be used as a wireless network. An optical communication network can also be used as a wireless network. In this case, the control unit 14 and the processing device 1 can be configured to send and receive various information via the network. Furthermore, the control unit 14 can send instructions, control parameters, and other information to the processing device 1 via the network. The processing device 1 can also have a receiving device that receives instructions, control parameters, and other information from the control unit 14 via the network. Alternatively, a first control device performing a portion of the processing performed by the control unit 14 can be located inside the processing device 1, while a second control device performing another portion of the processing performed by the control unit 14 can be located outside the processing device 1.

[0086] The control unit 14 can also function as a client device 3. For example, a computer can be used as both the control unit 14 and the client device 3. That is, the control unit 14 and the client device 3 can also be a single device (or a single system). However, typically, two different computers can also be used as the control unit 14 and the client device 3, respectively.

[0087] Within the control unit 14, a computer program can be executed via the computing device 141, and a computational model capable of being constructed through machine learning can be installed. As an example of a computational model capable of being constructed through machine learning, a computational model containing a neural network (so-called Artificial Intelligence (AI)) can be cited. In this case, the learning of the computational model can include the learning of the parameters of the neural network (e.g., at least one of weights and biases). The control unit 14 can use the computational model to control the operation of the processing device 1. That is, the action of controlling the operation of the processing device 1 can include the action of controlling the operation of the processing device 1 using the computational model. Furthermore, a computational model that has been constructed offline through machine learning using teaching data can also be installed in the control unit 14. Additionally, the computational model installed in the control unit 14 can be updated on the control unit 14 via online machine learning. Alternatively, the control unit 14 can also use a computational model from a device installed outside the control unit 14 (i.e., a device located outside the processing device 1) to control the operation of the processing device 1, in addition to using the computational model installed in the control unit 14 or as an alternative.

[0088] Furthermore, the recording medium for the computer program executed by the recording processing device 141 may be at least one of the following media: optical discs such as CD-ROM, CD-R, CD-RW, floppy disk, MO, DVD-ROM, DVD-RAM, DVD-R, DVD+R, DVD-RW, DVD+RW, and Blu-ray (registered trademark); magnetic media such as magnetic tape; semiconductor memory such as magneto-optical disc and USB memory; and any other media capable of storing programs. The recording medium may also include a device capable of recording computer programs (e.g., a general-purpose or special-purpose device in which the computer program is installed in an executable state in at least one form such as software and firmware). Furthermore, the various processes or functions included in the computer program may be implemented by the control unit 14 (i.e., the computer) executing the computer program by a logic processing block implemented within the control unit 14, or by hardware such as a predefined gate array (FPGA, ASIC) provided by the control unit 14, or by a hybrid form in which the logic processing block and a portion of the hardware elements of the implementation hardware coexist.

[0089] (2-2) Removal process performed by processing device 1

[0090] Next, refer to Figure 5 (a) to Figure 5Figure (c) illustrates an example of a removal process using processing light EL. Figure 5 (a) to Figure 5 Figure (c) is a cross-sectional view showing the state of the removal process performed on workpiece W.

[0091] like Figure 5 As shown in (a), the processing apparatus 1 irradiates a target irradiation area EA set (in other words, formed) on the surface of the workpiece W with a processing light EL. When the processing light EL is irradiated into the target irradiation area EA, the energy of the processing light EL is transferred to the target irradiation area EA and the portion of the workpiece W adjacent to the target irradiation area EA. If the heat generated by the energy of the processing light EL is transferred, the heat generated by the energy of the processing light EL causes the material constituting the target irradiation area EA and the portion of the workpiece W adjacent to the target irradiation area EA to melt. The molten material becomes droplets and disperses. Alternatively, the molten material evaporates due to the heat generated by the energy of the processing light EL. As a result, the target irradiation area EA and the portion of the workpiece W adjacent to the target irradiation area EA are removed. That is, as Figure 5 As shown in (b), a recess (in other words, a groove) with a depth equivalent to the reference machining amount ΔZ is formed on the surface of the workpiece W by the machining light EL. Here, when the machining light EL is a pulsed light, the reference machining amount Δz can be the depth to which the workpiece W is removed when the machining light EL is irradiated to the workpiece W with a unit number of pulses. Here, the unit number of pulses can be 1 or more. In addition, since the reference machining amount Δz is the amount processed by irradiating the machining light EL with a unit number of pulses, it can also be called the unit machining amount. In this case, the machining apparatus 1 can also be regarded as performing removal processing on the workpiece W using the so-called thermal processing principle. Furthermore, in the case of thermal processing, the machining light EL can include pulsed light or continuous light with a emission time of milliseconds or more. Furthermore, when the machining light EL is a continuous light, the reference machining amount Δz can be the amount (depth to be removed) to which the workpiece W is processed when the machining light EL with a unit energy is irradiated to the workpiece W with a unit time.

[0092] On the other hand, based on the characteristics of the processing light EL, the processing apparatus 1 can also process the workpiece W using a non-thermal processing principle (e.g., ablation processing). That is, the processing apparatus 1 can also perform non-thermal processing (e.g., ablation processing) on ​​the workpiece W. For example, if light with a high photon density (in other words, luminous flux) is used as the processing light EL, the material constituting the target irradiation region EA and the portion near the target irradiation region EA in the workpiece W will evaporate and disperse instantaneously. That is, the material constituting the target irradiation region EA and the portion near the target irradiation region EA in the workpiece W will evaporate and disperse in a sufficiently short time compared to the thermal diffusion time of the workpiece W. In this case, the material constituting the target irradiation region EA and the portion near the target irradiation region EA in the workpiece W can be emitted from the workpiece W in at least one form selected from ions, atoms, free radicals, molecules, clusters, and solid sheets. Furthermore, in the case of non-thermal processing, the processing light EL can contain pulsed light with an emission time of less than picoseconds (or, depending on the case, less than nanoseconds or femtoseconds). When using pulsed light with a emission time of less than picoseconds (or, depending on the case, less than nanoseconds or femtoseconds) as the processing light EL, the material constituting the target irradiation area EA and the portion near the target irradiation area EA in the workpiece W may sublimate directly without going through a molten state. Therefore, it is possible to process the workpiece W while minimizing the impact of the heat generated by the processing light EL energy on the workpiece W.

[0093] The processing apparatus 1 uses the galvanometer lens 1122 described above to move the target irradiation area EA on the surface of the workpiece W. That is, the processing apparatus 1 scans the surface of the workpiece W with processing light EL. The result is as follows: Figure 5 As shown in (c), at least partially, the surface of the workpiece W is removed along the scanning trajectory of the processing light EL (i.e., the movement trajectory of the target irradiation area EA). Therefore, by scanning the surface of the workpiece W with the processing light EL along a desired scanning trajectory corresponding to the area to be removed, the processing apparatus 1 can appropriately remove the portion of the workpiece W to be removed. By scanning the surface of the workpiece W with the processing light EL along a desired scanning trajectory corresponding to the area to be removed, the processing apparatus 1 can remove a removal layer SL from the workpiece W as the portion to be removed, which has a thickness equivalent to a reference processing amount Δz.

[0094] The processing apparatus 1 can process the workpiece W to achieve a desired shape by repeatedly removing a removal layer SL with a thickness equivalent to a reference processing amount Δz. In the following description, for ease of explanation, an example is given: processing the workpiece W to achieve a desired shape... Figure 6 The shape of the cuboid-shaped workpiece W shown on the left becomes Figure 6The right side shows a conical shape protruding from the plate-like component. In this case, as... Figure 7 As shown, the processing device 1 performs a removal process to remove the removal object portion W_rmv that should be removed from the workpiece W.

[0095] also, Figure 6 An example is shown where the surface of the workpiece W before removal processing (in particular, the surface of the workpiece W to be removed) is a plane along the XY plane. However, the surface of the workpiece W (in particular, the surface of the workpiece W to be removed) may also contain a plane inclined relative to the XY plane. The surface of the workpiece W (in particular, the surface of the workpiece W to be removed) may also contain a curved surface.

[0096] In order to perform the removal process for removing the W_rmv portion of the object to be removed, such as Figure 8 As shown, the processing apparatus 1 sequentially removes multiple removal layers SL obtained by slicing the removal target portion W_rmv along the Z-axis direction. Each removal layer SL can also be considered as a removal portion removed by scanning the processing light EL with the workpiece W and processing head 112 in a fixed positional relationship (in particular, the positional relationship between the workpiece W and the focused position of the processing light EL) in the Z-axis direction. In this case, firstly, the processing apparatus 1 performs a removal process to remove the uppermost removal layer SL#1 from the workpiece W. Then, the processing apparatus 1 moves the platform 132 and / or the processing head 112, causing the processing head 112 to approach the workpiece W at a distance equivalent to a reference processing amount Δz. Alternatively, the processing apparatus 1 controls the illumination optical system 1120 (in particular, the focusing optical system 1121), causing the focused position of the processing light EL to approach the workpiece W at a distance equivalent to a reference processing amount Δz. Then, the processing apparatus 1 performs a removal process to remove the second removal layer SL#2 from the workpiece W from which the first removal layer SL#1 has been removed. The processing apparatus 1 then repeats the same operation until all layer SL is removed (e.g., in...). Figure 8 In the example shown, Q (where Q is a constant representing an integer greater than 2) removal layers SL#1 to SL#Q are removed. The result is that, with Figure 6 The shape of the cuboid-shaped workpiece W shown on the left becomes Figure 6 The right side shows a conical shape protruding from the plate-like component.

[0097] The processing apparatus 1 can sequentially remove multiple removal layers SL based on slice data representing the area on the surface of the workpiece W to be removed during the removal process of each removal layer SL. Furthermore, the slice data can also be an example of processing control information used to process the workpiece W. For example, the processing apparatus 1 can remove multiple removal layers SL sequentially based on the first slice data representing the area on the surface of the workpiece W to be removed during the removal process of removal layer SL#1 (see reference). Figure 9 (a) ), the processing light EL is irradiated onto the area represented by the first slice data on the surface of the workpiece W, thereby removing the removal layer SL#1. Then, the processing apparatus 1 can use the second slice data (refer to) representing the area on the surface of the workpiece W to be removed during the removal of the removal layer SL#2. Figure 9 (b) ), the processing light EL is irradiated onto the area represented by the second slice data on the surface of the workpiece W, thereby removing the removal layer SL#2. Then, the processing apparatus 1 can use the third slice data (refer to) representing the area on the surface of the workpiece W to be removed during the removal of the removal layer SL#3. Figure 9 (c) The processing light EL is irradiated onto the area represented by the third slice data on the surface of the workpiece W, thereby removing the removal layer SL#3. Then, the processing apparatus 1 repeats the same operation. Finally, the processing apparatus 1 can determine the removal layer based on the Q slice data (refer to the Q slice data representing the area to be removed on the surface of the workpiece W during the removal of the removal layer SL#Q). Figure 9 (d) , irradiate the area represented by the Q-th slice data in the surface of the workpiece W with processing light EL, thereby removing the removal layer SL#Q. Furthermore, Figure 9 (a) to Figure 9 In each of the (d) figures, the area on the surface of the workpiece W to be removed (i.e., the area to be illuminated by the processing light EL) is shown in shaded areas.

[0098] The processing apparatus 1 uses a processing light EL to scan the processing area PSA defined on the surface of the workpiece W along the XY plane. In this case, as... Figure 10 As shown in (a), it can be considered that within the processing area PSA, the number of irradiation target positions (irradiation positions) C that can be irradiated by the processing light EL is set to correspond to the scanning speed and pulse frequency of the processing light EL. Furthermore, the processing apparatus 1 moves the processing area PSA along the XY plane on the surface of the workpiece W by moving at least one of the processing head 112 and the platform 132. Therefore, as Figure 10 As shown in (b), it can be considered that during the removal process of each removal layer SL, in the plane along the scanning direction of the processing light EL in the processing area PSA and the moving direction of the processing area PSA, and in the processing surface PL (typically, the plane along the XY plane) where the processing apparatus 1 is to perform the removal processing, the number of irradiation target positions C that can be irradiated by the processing light EL is set corresponding to the number of irradiation target positions C set in the processing area PSA and the number of times the processing area PSA moves on the surface of the workpiece W (i.e., the number of times the processing area PSA is set on the surface of the workpiece W). In this case, it can be considered that during the removal process of each removal layer SL, multiple irradiation target positions C are set on the surface of the workpiece W. In the following description, for ease of explanation, as Figure 10As shown in (b), an example is given of setting N (where N is a constant representing an integer greater than 2) irradiation target positions C (specifically, irradiation target positions C1 to CN) within the machining surface PL.

[0099] Figure 10 (a) and Figure 10 (b) Each target location C is schematically represented by a beam spot (i.e., a circular beam spot) of the processing light EL illuminating the target location C. The beam spot of the processing light EL may also refer to a region of the processing light EL whose intensity exceeds a predetermined intensity threshold. In this case, when each target location C is being irradiated by the processing light EL, the region outside the circle schematically representing each target location C can be irradiated by the processing light EL whose intensity does not exceed the predetermined intensity threshold. Alternatively, the beam spot of the processing light EL may also refer to a region corresponding to the full width at half maximum (FWHM) of the intensity distribution (e.g., a Gaussian distribution) of the processing light EL.

[0100] in addition, Figure 10 (a) and Figure 10 (b) preferably shows an example where an irradiated target position C and another adjacent irradiated target position C do not overlap, for the sake of readability of the accompanying drawings. However, in practice, an irradiated target position C and another adjacent irradiated target position C may at least partially overlap. That is, the beam spot of the processing light EL irradiating an irradiated target position C and the beam spot of the processing light EL irradiating another adjacent irradiated target position C may at least partially overlap.

[0101] Slice data can also be viewed as equivalent to information indicating whether processing light (EL) is irradiated onto each of the target locations C1 to CN. As an example, Figure 9 The Q-slice data shown in (d) can also be considered equivalent to the following information: representing the target positions C1 to CN to be irradiated. Figure 9 Processing light EL shall be irradiated at least one target location C in the shaded area shown in (d), but not at target locations C1 to CN that are not included in the shaded area. Figure 9 At least one other target location C in the shaded area shown in (d) is irradiated with processing light EL.

[0102] The data generation server 2 mentioned above can also generate data for generating... Figure 9 (a) to Figure 9The control data for the slice data shown in (d) is as follows. For example, Q slice data represent whether processing light EL should be irradiated onto each of the target irradiation positions C1 to CN. In other words, the Q slice data can be said to represent whether the number of times processing light EL should be irradiated onto each of the target irradiation positions C1 to CN during the removal process of each removal layer SL corresponding to the Q slice data is 1 or 0 (or whether it is a different number). Therefore, the Q slice data as a whole can be said to represent the number of times processing light EL should be irradiated onto each of the target irradiation positions C1 to CN. In this case, the data generation server 2 can also generate control data representing the number of times processing light EL should be irradiated onto each of the target irradiation positions C1 to CN. In this case, the control unit 14 can also generate Q slice data based on the control data, and remove Q removal layers SL sequentially based on the Q slice data.

[0103] Furthermore, as described above, since the processing light EL is a pulsed light, the number of times the processing light EL should be irradiated at the target location C can also refer to the number of times a unit number of pulses of processing light EL should be irradiated at the target location C. In other words, the number of times the processing light EL should be irradiated at the target location C can also refer to the number of pulses of the pulsed light contained in the processing light EL that should be irradiated at the target location C. That is to say, the number of times the processing light EL should be irradiated at the target location C can also refer to the number of pulses that should be irradiated at the target location C. In the following description, considering that the processing light EL is a pulsed light, for ease of explanation, the number of times the processing light EL should be irradiated at the target location C is referred to as the target pulse number. Furthermore, the target pulse number is a specific example of the processing conditions of the processing apparatus 1 described above.

[0104] (3) Data generation server 2

[0105] Next, the data generation server 2 of the machining system SYS will be described. The data generation server 2 that generates control data representing a target pulse number as an example of machining conditions will be described below. Specifically, the data generation server 2 that generates target pulse number information as at least a part of the control data will be described as follows: the target pulse number as the number of times machining light EL is irradiated to the target irradiation position C1, the target pulse number as the number of times machining light EL is irradiated to the target irradiation position C2, ..., the target pulse number as the number of times machining light EL is irradiated to the target irradiation position CN.

[0106] (3-1) Composition of data generation server 2

[0107] First, refer to Figure 11 The composition of data generation server 2 will be explained. Figure 11 This is a block diagram showing the structure of data generation server 2.

[0108] like Figure 11 As shown, the data generation server 2 includes a computing unit 21, a storage unit 22, and a communication unit 23. Furthermore, the data generation server 2 may also include an input unit 24 and an output unit 25. However, the data generation server 2 may also omit at least one of the input unit 24 and the output unit 25. The computing unit 21, storage unit 22, communication unit 23, input unit 24, and output unit 25 may also be connected via a data bus 26.

[0109] The computing device 21 includes, for example, at least one of a CPU and a GPU. The computing device 21 reads computer programs. For example, the computing device 21 can read computer programs stored in storage device 22. For example, the computing device 21 can also use a recording media reading device (not shown) to read computer programs stored on computer-readable and non-transitory recording media. The computing device 21 can also acquire (i.e., download or read) computer programs from a device (not shown) configured outside the data generation server 2 via a communication device 23. That is, the computing device 21 can also acquire (i.e., download or read) computer programs stored in storage devices of devices (not shown) configured outside the data generation server 2 via the communication device 23. The computing device 21 executes the read computer program. As a result, within the computing device 21, a logical function block for performing actions that the data generation server 2 should perform (e.g., data generation actions for generating control data) is implemented. In other words, the computing device 21 can function as a controller for implementing logical function blocks that perform actions that the data generation server 2 should perform. In this case, any device that executes a computer program (typically a computer) can function as a data generation server 2.

[0110] Figure 11 An example is shown of a logic function block implemented within the arithmetic unit 21 for performing the action of generating control data representing the target pulse number. Figure 11 As shown, a model generation unit 211 and a control data generation unit 212 are implemented within the computing device 21. Furthermore, the operation of each of the model generation unit 211 and the control data generation unit 212 will be explained below with reference to... Figure 12While details will be provided later, a brief overview is given below. The model generation unit 211 generates model information LM. Model information LM represents the relationship between the number of times processing light EL is irradiated at each of the irradiation target positions C1 to CN (as an example, the target pulse number) for processing the workpiece W and the processed shape of the workpiece W. The control data generation unit 212 calculates the target pulse number based on the model information LM. Specifically, the control data generation unit 212 calculates the target pulse number required to process the workpiece W to achieve the desired target shape based on the model information LM. In other words, the control data generation unit 212 generates control data representing the target pulse number.

[0111] Within the computing device 21, a computer program can be executed, and a computational model capable of being constructed through machine learning can be installed. As an example of a computational model capable of being constructed through machine learning, a computational model containing a neural network (so-called Artificial Intelligence (AI)) can be cited. In this case, the learning of the computational model may include the learning of the parameters of the neural network (e.g., at least one of weights and biases). The computing device 21 can use the computational model to perform data generation actions. That is, the data generation actions may include the action of using the computational model to generate control data. In other words, at least one of the model generation unit 211 and the control data generation unit 212 can be implemented using the computational model. In other words, the actions performed by at least one of the model generation unit 211 and the control data generation unit 212 can be performed using the computational model. Furthermore, a computational model that has been constructed through offline machine learning using teaching data can also be installed in the computing device 21. Additionally, the computational model installed in the computing device 21 can be updated on the computing device 21 through online machine learning. Alternatively, the computing device 21 may use the computing model of a device installed outside the computing device 21 (i.e., a device located outside the data generation server 2) in addition to or as a substitute for the computing model installed on the computing device 21 to generate control data.

[0112] Furthermore, at least a portion of the functional blocks (i.e., the model generation unit 211 and the control data generation unit 212) within the computing device 21 of the data generation server 2 may not be provided by the computing device 21 (i.e., the data generation server 2). For example, at least a portion of the functional blocks (i.e., the model generation unit 211 and the control data generation unit 212) within the computing device 21 may be provided by the client device 3. For example, at least a portion of the functional blocks (i.e., the model generation unit 211 and the control data generation unit 212) within the computing device 21 may be provided by the processing device 1 (e.g., the control unit 14).

[0113] Storage device 22 is capable of storing required data. For example, storage device 22 can temporarily store computer programs executed by computing device 21. Storage device 22 can also temporarily store data temporarily used by computing device 21 when executing computer programs. Storage device 22 can also store data that data generation server 2 will store long-term. In this embodiment, storage device 22 can also store model information LM generated by model generation unit 211. Furthermore, storage device 22 can include at least one of RAM (Random Access Memory), ROM (Read Only Memory), hard disk device, magneto-optical disk device, SSD (Solid State Drive), and disk array device. That is, storage device 22 can also include non-temporary recording media.

[0114] The communication device 23 can communicate with at least one of the processing device 1 and the client device 3 via the communication network 4. In this embodiment, the communication device 23 can receive reference information used by the data generation server 2 for generating control data from at least one of the processing device 1 and the client device 3 via the communication network 4. Furthermore, the communication device 23 can send the generated control data to at least one of the processing device 1 and the client device 3 via the communication network 4.

[0115] Input device 24 is a device for accepting information input from outside the data generation server 2 to the data generation server 2. For example, input device 24 may include an operating device (e.g., at least one of a keyboard, mouse, and touch panel) that can be operated by a server user. For example, input device 24 may also include a reading device capable of reading information recorded in data form on a recording medium that can be installed outside the data generation server 2.

[0116] Output device 25 is a device for outputting information to the external system of data generation server 2. For example, output device 25 can output information in the form of an image. That is, output device 25 may include a display device (so-called a monitor) capable of displaying an image representing the information to be output. For example, output device 25 can also output information in the form of audio. That is, output device 25 may also include an audio device (so-called a speaker) capable of outputting audio. For example, output device 25 can also output information onto paper. That is, output device 25 may also include a printing device (so-called a printer) capable of printing the desired information onto paper.

[0117] Data generation server 2 can also function as client device 3. For example, a computer can be used as both data generation server 2 and client device 3. That is, data generation server 2 and client device 3 can also be a single device (or a single system). However, typically, two different computers can also be used as data generation server 2 and client device 3, respectively.

[0118] (3-2) Data generation actions performed by data generation server 2

[0119] Next, refer to Figure 12 The data generation actions (i.e., actions to generate control data) performed by data generation server 2 are described. Figure 12 It is a flowchart representing the data generation actions (i.e., actions to generate control data) performed by data generation server 2.

[0120] like Figure 12 As shown, the data generation server 2 first generates model information LM for generating control data (steps S11 to S15). In this embodiment, the data generation server 2 can generate model information LM by performing machine learning using teaching data. In this case, the data generation server 2 first obtains information related to the actual test processing results performed by the processing device 1 on the workpiece W for test processing as teaching data.

[0121] Furthermore, in the following description, the workpiece W used for test processing will be referred to as test workpiece W_test, and the workpiece W actually processed using the control data generated by data generation server 2 will be referred to as the processing target workpiece W_target, thus distinguishing between the two. Test workpiece W_test can be a workpiece W of the same specifications as processing target workpiece W_target. Test workpiece W_test can also be a workpiece W of the same size as processing target workpiece W_target. Test workpiece W_test can also be a workpiece W of the same shape as processing target workpiece W_target. Test workpiece W_test can also be a workpiece W made of the same material as processing target workpiece W_target.

[0122] In order to obtain the test processing results of the workpiece W used for test processing by the processing device 1 as teaching data, the model generation unit 211 of the data generation server 2 acquires test processing condition information representing the processing conditions of the processing device 1 performing the test processing (step S11). That is, the model generation unit 211 acquires test processing condition information related to the processing conditions of the processing device 1 performing the test processing (step S11). The test processing condition information represents the processing conditions required to process the test workpiece W_test so that the shape of the test workpiece W_test becomes the desired test target shape. In other words, the test processing condition information represents the processing conditions for the test processing of the test workpiece W_test to make the shape of the test workpiece W_test become the desired test target shape.

[0123] The test target shape is the same as the shape that the machined workpiece W_target should have, i.e., the machining target shape. In other words, the test target shape is the same as the target shape of the machined workpiece W_target. However, the test target shape can also be a different shape from the machining target shape. As described below, when performing multiple test processes using multiple different test processing condition information, at least one of the following test processes can be performed: test processing to make the test workpiece W_test's shape the same as the machining target shape; and test processing to make the test workpiece W_test's shape a different shape from the machining target shape.

[0124] As described above, the processing apparatus 1 can form a rib-like structure on the workpiece W by processing the workpiece W_target. An example of the rib-like structure is shown below. Figure 13 (a) and Figure 13 (b). For example Figure 13 (a) and Figure 13As shown in (b), the rib-like structure may comprise a structure consisting of multiple grooves GV extending along a first direction (e.g., the X-axis direction) arranged along a second direction (e.g., the Y-axis direction), wherein the first direction is along the surface of the workpiece W_target, and the second direction is along the surface of the workpiece W_target and intersects the first direction. In other words, the rib-like structure may comprise a structure consisting of multiple convex structures LD extending along a first direction (e.g., the X-axis direction) arranged along a second direction (e.g., the Y-axis direction), wherein the first direction is along the surface of the workpiece W_target, and the second direction is along the surface of the workpiece W_target and intersects the first direction. In this case, at least one of the test target shape and the processing target shape may be a shape equivalent to the rib-like structure. For example, as at least one of the test target shape and the processing target shape, a three-dimensional shape of the rib-like structure may be used. For example, as at least one of the test target shape and the processing target shape, a three-dimensional distribution representing the three-dimensional shape of the rib-like structure (e.g., exhibiting...) may also be used. Figure 13 (a) shows the three-dimensional distribution of the three-dimensional shape.

[0125] At least one of the test target shape and the machining target shape can be a shape whose height varies in at least one dimension. Specifically, the test target shape can be a shape whose height (e.g., the height of the surface of the test workpiece W_test) varies in at least one dimension, and the machining target shape can be a shape whose height (e.g., the height of the surface of the machining target workpiece W_target) varies in at least one dimension. For example, as... Figure 13 (a) and Figure 13 As shown in (b), the shape of the rib-like structure described above becomes a shape in which the height (i.e., the dimension in the third direction intersecting the first and second directions, for example, the dimension in the Z-axis direction) varies along the second direction in which multiple grooves GV (multiple convex structures LD) are arranged. In this case, at least one of the test target shape and the machining target shape can be a shape in which the height varies along the second direction in which multiple grooves GV (multiple convex structures LD) are arranged. In other words, the test target shape can be as follows: Figure 13 (b) shows a cross-sectional distribution representing the height varying along the second direction of a plurality of grooves GV (plural convex structures LD) arranged in a plurality of grooves (in other words, a height distribution or a two-dimensional distribution, for example, representing...). Figure 13 (b) shows the two-dimensional distribution of the two-dimensional shape.

[0126] In this embodiment, the test processing condition information includes at least test pulse count information, which represents the number of times processing light EL should be irradiated onto the irradiation target position C for the purpose of testing the processing. Specifically, as described above, since N irradiation target positions C1 to CN are set on the processing surface PL of the workpiece W, N irradiation target positions C1 to CN are also set on the processing surface PL of the test workpiece W_test. In this case, the test processing condition information may include test pulse count information representing the following test pulse counts: the number of test pulses representing the number of times processing light EL is irradiated onto the irradiation target position C1, the number of test pulses representing the number of times processing light EL is irradiated onto the irradiation target position C2, ..., the number of test pulses representing the number of times processing light EL is irradiated onto the irradiation target position CN.

[0127] The model generation unit 211 can acquire test processing condition information that has been input to the data generation server 2 using the input device 24. For example, the model generation unit 211 can acquire test processing condition information that has been input to the data generation server 2 by the user of the processing system SYS using the input device 24.

[0128] The model generation unit 211 can use the communication device 23 to obtain test processing condition information sent to the data generation server 2. For example, the model generation unit 211 can obtain test processing condition information by using the communication device 23 to receive test processing condition information stored on a server outside the data generation server 2.

[0129] The model generation unit 211 can obtain test processing condition information by generating test processing condition information. For example, the model generation unit 211 can generate test processing condition information representing the set processing conditions by setting the processing conditions required to process the test workpiece W_test so that the shape of the test workpiece W_test becomes the desired test target shape.

[0130] Back to Figure 12In step S11, after obtaining the test processing condition information, the test workpiece W_test is processed based on the test processing condition information obtained in step S11 (step S12). Specifically, the model generation unit 211 controls the processing device 1 to process the test workpiece W_test based on the test processing condition information. For example, the model generation unit 211 can send the test processing condition information as control data to the processing device 1. The processing device 1 (especially the control unit 14) can generate processing control information for controlling the processing device 1 to perform test processing based on the test processing condition information sent from the data generation server 2. For example, the control unit 14 can generate processing control information for controlling the processing device 1 to process the test workpiece W_test so that the shape of the test workpiece W_test becomes the desired test target shape. Then, the processing device 1 can process the test workpiece W_test based on the generated processing control information. That is, the processing device 1 can perform test processing on the test workpiece W_test based on the generated processing control information.

[0131] After the test workpiece W_test is processed in step S12, the shape of the test workpiece W_test after the processing is measured, which is the test workpiece shape (step S13). In other words, the shape of the test workpiece W_test after processing based on the test processing condition information obtained in step S11 is the test workpiece shape (step S13). For example, the processing apparatus 1 can use the measurement unit 12 to measure the test workpiece shape. Then, the test workpiece shape information (test shape information), representing the measurement result of the test workpiece shape obtained by the measurement unit 12, can be sent from the processing apparatus 1 to the data generation server 2 via the communication network 4. The model generation unit 211 of the data generation server 2 can acquire the test workpiece shape information (test shape information), representing the measurement result of the test workpiece shape obtained by the measurement unit 12.

[0132] The obtained test workpiece shape information is related to the test processing result performed on the test workpiece W_test by the processing device 1. In other words, the model generation unit 211 acquires the test workpiece shape information as teaching data for generating model information LM. More specifically, the model generation unit 211 acquires data that establishes a correlation between the test processing condition information obtained in step S11 and the test workpiece shape information related to the test processing result based on the test processing condition information, and uses this data as teaching data for generating model information LM.

[0133] Then, the model generation unit 211 determines whether to end the test process (step S14). For example, the model generation unit 211 may determine to end the test process when the number of acquired teaching data exceeds a predetermined number of samples required to represent the number of teaching data needed to generate the model information LM. On the other hand, the model generation unit 211 may determine not to end the test process when the number of acquired teaching data is less than the predetermined number of samples. Typically, in order to generate the model information LM, it is preferable to use multiple teaching data. Therefore, the predetermined number of samples is preferably 2 or more.

[0134] If the determination in step S14 is that the test processing should not be terminated (step S14: No), then the actions in steps S11 to S13 are performed again. That is, the model generation unit 211 acquires new test processing condition information (step S11), performs test processing on the test workpiece W_test based on the newly acquired test processing condition information (step S12), and measures the shape of the test workpiece W_test after the test processing based on the newly acquired test processing condition information, which is the test workpiece shape (step S13).

[0135] Thus, in this embodiment, multiple test processing steps can be performed to obtain multiple teaching data. For example, first test processing condition information can be obtained (step S11), and a first test workpiece W_test can be tested and processed based on the first test processing condition information (step S12). The shape of the first test workpiece W_test after testing and processing based on the first test processing condition information is measured, which is the shape of the first test workpiece (step S13). As a result, first teaching data is obtained that establishes an association between the first test processing condition information and the test workpiece shape information representing the shape of the first test workpiece. Then, second test processing condition information different from the first test processing condition information can be obtained (step S11), and a second test workpiece W_test can be tested and processed based on the second test processing condition information (step S12). The shape of the second test workpiece W_test after testing and processing based on the second test processing condition information is measured, which is the shape of the second test workpiece (step S13). Furthermore, the second test workpiece W_test can be different from the first test workpiece W_test. As a result, second teaching data is obtained that establishes an association between the second test processing condition information and the test workpiece shape information representing the shape of the second test workpiece. Repeat this action until the required amount of teaching data has been obtained.

[0136] On the other hand, when the result of the determination in step S14 is the determination to end the test processing (step S14: yes), the model generation unit 211 performs machine learning by using a teaching dataset containing multiple teaching data obtained, and generates model information LM.

[0137] As described above, the model information LM represents the relationship between the number of times machining light EL is irradiated at each of the irradiation target positions C1 to CN for machining workpiece W and the shape of workpiece W after machining. In this case, the model generation unit 211 can use teaching data that establishes a correlation between test machining condition information (number of test pulses) and test workpiece shape information to perform machine learning, so that the model information LM appropriately represents the relationship between the number of times machining light EL is irradiated at each of the irradiation target positions C1 to CN for machining test workpiece W_test and the shape of test workpiece W_test after machining, i.e., the shape of the test workpiece. In other words, the model generation unit 211 can use teaching data that establishes a correlation between test machining condition information and test workpiece shape information to perform machine learning, so that the model information LM appropriately represents the relationship between test machining condition information and test workpiece shape information.

[0138] Here, when the model information LM is model information representing the relationship between the number of test pulses and the shape of the test workpiece, the model generation unit 211 can calculate (in other words, predict) the other of the number of test pulses and the shape of the test workpiece based on either the number of test pulses or the shape of the test workpiece and the model information LM. In this case, in order to perform machine learning for generating the model information LM, such as... Figure 14 As shown, the model generation unit 211 can predict the shape of the test workpiece based on test pulse count information representing the number of test pulses and model information LM. Furthermore, before model information LM is generated (i.e., ...), ... Figure 12 If step S15 is not performed even once, the model generation unit 211 can use the preset model information LM (in other words, the model information LM in its initial state) to predict the shape of the test workpiece. If the model information LM has already been generated (i.e.,... Figure 12 If step S15 has been performed at least once, the model generation unit 211 can use the generated model information LM to predict the shape of the test workpiece.

[0139] Here, the closer the model information LM is to the ideal model information LM, the smaller the difference between the predicted test workpiece shape obtained using the model information LM and the actual measurement result of the test workpiece shape (i.e., the test workpiece shape information), i.e., the test workpiece shape difference Δs_test. Therefore, the model generation unit 211 can generate the model information LM by performing machine learning in a way that minimizes (or reduces) the test workpiece shape difference Δs_test. As an example, when multiple teaching data are obtained, the model generation unit 211 can generate multiple prediction results for the test workpiece shape based on the multiple test pulse count information and the model information LM contained in each of the multiple teaching data. For example, the model generation unit 211 can generate a prediction result for the first test workpiece shape contained in the first teaching data based on the first test pulse count information and the model information LM contained in the first teaching data, and generate a prediction result for the second test workpiece shape contained in the second teaching data based on the second test pulse count information and the model information LM contained in the second teaching data. Then, the model generation unit 211 can calculate the differences between multiple predicted results of the test workpiece shape obtained using the model information LM and multiple measurement results of the test workpiece shape contained in each of the multiple teaching data (i.e., multiple test workpiece shape information), namely, multiple test workpiece shape differences Δs_test. For example, the model generation unit 211 can calculate the difference between the actual first test workpiece shape contained in the first teaching data and the predicted result of the first test workpiece shape, namely, the first test workpiece shape difference Δs_test, and calculate the difference between the actual second test workpiece shape contained in the second teaching data and the predicted result of the second test workpiece shape, namely, the second test workpiece shape difference Δs_test. Then, the model generation unit 211 can generate the model information LM by performing machine learning in a way that minimizes (or reduces) the loss determined based on the multiple test workpiece shape differences Δs_test. The loss determined based on the multiple test workpiece shape differences Δs_test can be a loss whose value is smaller as the multiple test workpiece shape differences Δs_test are smaller. As an example of loss, mean square error and mean absolute error can be cited. The mean square error is calculated by dividing the sum of the squares of multiple test workpiece shape differences Δs_test by the total number of test workpiece shape differences Δs_test. The mean absolute error is calculated by dividing the sum of the absolute values ​​of multiple test workpiece shape differences Δs_test by the total number of test workpiece shape differences Δs_test.

[0140] The model information LM can be any information that represents the relationship between the number of test pulses and the shape of the test workpiece. For example, the model information LM can contain model information that can predict the shape of the test workpiece based on the number of test pulses. Alternatively, the model information LM can contain information that can predict the number of test pulses based on the shape of the test workpiece. Or, the model information LM can contain model information that can output a predicted shape of the test workpiece when a number of test pulses is input. Finally, the model information LM can contain information that can output a predicted number of test pulses when a shape of the test workpiece is input.

[0141] An example of model information LM is shown below. Figure 15 In the middle. For example Figure 15 As shown, the model information LM can include prediction information PI representing the predicted amount of machining done on the test workpiece W_test when the target location C is irradiated by a unit number of machining lights EL. Specifically, as... Figure 15 As shown, the model information LM can include: prediction information PI1, representing the predicted amount of machining on the test workpiece W_test when the target position C1 is irradiated by a unit number of machining lights EL; prediction information PI2, representing the predicted amount of machining on the test workpiece W_test when the target position C2 is irradiated by a unit number of machining lights EL; ..., prediction information PIi, representing the predicted amount of machining on the test workpiece W_test when the target position Ci (where i is a variable representing an integer greater than 1 and less than n) is irradiated by a unit number of machining lights EL; ..., prediction information PIn, representing the predicted amount of machining on the test workpiece W_test when the target position Cn is irradiated by a unit number of machining lights EL. Furthermore, in this embodiment, an example with a unit number of 1 is described.

[0142] The "machining amount of test workpiece W_test" represented by the prediction information PI can refer to the distribution of the machining amount (specifically, the removal amount) of test workpiece W_test along the surface of test workpiece W_test in a direction (e.g., at least one of the X-axis and Y-axis directions). In other words, the "machining amount of test workpiece W_test" represented by the prediction information PI can refer to a one-dimensional distribution of the machining amount of test workpiece W_test. Alternatively, the "machining amount of test workpiece W_test" represented by the prediction information PI can refer to the distribution of the machining amount of test workpiece W_test within a surface of test workpiece W_test (e.g., a surface along the XY plane). In other words, the "machining amount of test workpiece W_test" represented by the prediction information PI can refer to a two-dimensional distribution of the machining amount of test workpiece W_test.

[0143] As an example, consider the conceptual representation of predictive information Pii. Figure 16 As shown, the prediction information Pii, representing the predicted amount of machining on the test workpiece W_test when the target position Ci is irradiated by a unit number of machining light ELs, can represent not only the predicted amount of machining at the target position Ci when the target position Ci is irradiated by a unit number of machining light ELs, but also the predicted amount of machining at the target position Cj (where j is a variable representing an integer greater than 1 and less than n, different from variable i) when the target position Ci is irradiated by a unit number of machining light ELs. In other words, the prediction information Pii can contain information that can be used to make predictions for: the predicted amount of machining at the target position Ci when the target position Ci is irradiated by a unit number of machining light ELs, and the predicted amount of machining at the target position Cj when the target position Ci is irradiated by a unit number of machining light ELs.

[0144] For example, prediction information PI1 can represent not only the predicted amount of processing at target position C1 when it is irradiated by a unit number of processing light ELs, but also the predicted amount of processing at target positions C2 to Cn when it is irradiated by a unit number of processing light ELs. Similarly, prediction information PI2 can represent not only the predicted amount of processing at target position C2 when it is irradiated by a unit number of processing light ELs, but also the predicted amount of processing at target positions C1 and C3 to Cn when it is irradiated by a unit number of processing light ELs. Likewise, prediction information Pii can represent not only the predicted amount of processing at target position Ci when it is irradiated by a unit number of processing light ELs, but also the predicted amount of processing at target positions C1 to Ci-1 and Ci+1 to Cn when it is irradiated by a unit number of processing light ELs. For example, the prediction information PIn can represent not only the predicted amount of processing at the target position Cn when the target position Cn is irradiated by a unit number of processing lights EL, but also the predicted amount of processing at each of the target positions C1 to Cn-1 when the target position Cn is irradiated by a unit number of processing lights EL.

[0145] Thus, as one reason for predicting the distribution of the machining amount of the test workpiece W_test by representing the information Pii, it can be exemplified that the machining light EL illuminating the target position Ci may not only machine a portion of the test workpiece W_test located at the target position Ci, but also machine another portion of the test workpiece W_test located at the target position Cj. Specifically, the reason is that when the target position Ci is irradiated by the machining light EL, if the luminous flux of the machining light EL at the target position Cj exceeds the lower threshold of the luminous flux that can machine the test workpiece W_test, then the machining light EL illuminating the target position Ci will machine the other portion of the test workpiece W_test located at the target position Cj.

[0146] Furthermore, the state of the processing light EL illuminating the target location Ci can refer to the state where the center of the beam point of the processing light EL is located at or near the target location Ci. The beam point of the processing light EL can also refer to the region of the processing light EL where the irradiation intensity exceeds a predetermined intensity threshold. The beam point of the processing light EL can also refer to the region corresponding to the full width at half maximum (FWHM) of the intensity distribution of the processing light EL (e.g., a Gaussian distribution).

[0147] When the prediction information Pii represents the distribution of the machining amount of the test workpiece W_test, the prediction information Pii can also be regarded as representing at least one of the width and depth of the removal target portion W_rmv removed from the test workpiece W_test when the target position Ci is irradiated by a unit number of machining light ELs. That is, the prediction information Pii can also be regarded as information including at least one of the width and depth of the removal target portion W_rmv removed from the test workpiece W_test when the target position Ci is irradiated by a unit number of machining light ELs. Furthermore, the width of the removal target portion W_rmv can refer to the size of the removal target portion W_rmv in at least one of the X-axis and Y-axis directions. The depth of the removal target portion W_rmv can refer to the size of the removal target portion W_rmv in the Z-axis direction. The depth of the removal target portion W_rmv can also be synonymous with the machining amount.

[0148] When the model information LM includes prediction information PI, the machine learning used to generate (update) the model information LM can include machine learning used to generate (update) the prediction information PI included in the model information LM. That is, the model generation unit 211 can generate the model information LM by performing machine learning to generate the prediction information PI included in the model information LM. For example, the model generation unit 211 can generate the model information LM by performing machine learning to generate the prediction information PI in a manner that minimizes (or minimizes) the loss determined based on the multiple test workpiece shape differences Δs_test described above.

[0149] The prediction information Pii can include model parameters A that specify the distribution of machining quantities for the test workpiece W_test. For example, as... Figure 16 As shown, the prediction information Pii can contain m model parameters A (specifically, model parameters Ai, 1 to Ai, m) multiplied by m (where m is a constant representing an integer greater than 1) basis functions PM (specifically, basis functions PM1 to PMm). The basis functions PM1 to PMm can each represent the unit processing quantity of the test workpiece W_test. In particular, the basis functions PM1 to PMm can each represent m distinct unit processing quantities. Figure 16 In the example shown, basis functions PM1 to PMm represent m units of machining quantity with different dimensions (i.e., machining quantity) in the width direction (i.e., at least one of the X-axis and Y-axis directions). However, basis functions PM1 to PMm can also represent m units of machining quantity with different dimensions (i.e., machining quantity) in the depth direction (i.e., Z-axis direction). In this case, the prediction information Pii can also represent the distribution of machining quantity obtained by adding the product values ​​of basis functions PM1 to PMm with model parameters Ai,1 to Ai,m, as the distribution of machining quantity for the test workpiece W_test. That is, the prediction information Pii can also represent the distribution of machining quantity obtained by adding the product values ​​of basis function PM1 with model parameter Ai,1, basis function PM2 with model parameter Ai,2, ..., basis function PMm with model parameter Ai,m, as the distribution of machining quantity for the test workpiece W_test.

[0150] When the prediction information PI includes model parameters A, the machine learning used to generate (update) the model information LM can include machine learning used to generate (update) the model parameters A included in the prediction information PI. In other words, the model generation unit 211 can generate the model information LM by performing machine learning to generate the model parameters A included in the prediction information PI. For example, the model generation unit 211 can generate the model information LM by performing machine learning to generate the model parameters A in a manner that minimizes (or reduces) the loss determined based on the multiple test workpiece shape differences Δs_test described above.

[0151] When the model information LM includes prediction information PI (i.e., prediction information PI1 to PIn), such as Figure 15 As shown, the following relationship essentially holds between the number of test pulses and the predicted shape of the test workpiece: Number of test pulses × Prediction information PI = Prediction result of the test workpiece shape. Specifically, as... Figure 15 As shown, the model generation unit 211 predicts the machining amount of the test workpiece W_test when the target irradiation position Ci is irradiated with the machining light EL the same number of times as the test pulses by multiplying the predicted information Pii by the number of times the machining light EL is irradiated towards the target irradiation position Ci. The model generation unit 211 predicts the machining amount of the test workpiece W_test when the target irradiation positions C1 to Cn are irradiated with the machining light EL the same number of times as the test pulses by repeatedly performing the same operation with the same number of target irradiation positions C. Then, the model generation unit 211 predicts the machining amount of the test workpiece W_test (i.e., the shape of the test workpiece) by adding the n predicted values ​​of the machining amount of the test workpiece W_test corresponding to the target irradiation positions C1 to Cn respectively. As a result, the model generation unit 211 can calculate the difference between the predicted result of the test workpiece shape and the actual measurement result of the test workpiece shape (i.e., the test workpiece shape information), namely the test workpiece shape difference Δs_test, and perform machine learning in a way that the loss determined based on the calculated test workpiece shape difference Δs_test is reduced (or minimized), thereby generating model information LM.

[0152] Through the machine learning described above, the model generation unit 211 can generate model information LM that appropriately represents the relationship between the number of test pulses and the shape of the test workpiece. In other words, the model generation unit 211 can acquire model information LM that appropriately represents the relationship between the number of test pulses and the shape of the test workpiece.

[0153] Back to Figure 12In step S15, after the model information LM is generated, the control data generation unit 212 of the data generation server 2 calculates (in other words, generates) the target pulse number based on the model information LM generated in step S15 (step S21). That is, the control data generation unit 212 generates target pulse number information representing the target pulse number based on the model information LM generated in step S15 (step S21).

[0154] Specifically, the test pulse number, like the target pulse number, represents the number of times the processing light EL is irradiated onto each of the irradiation target positions C1 to CN. Therefore, the model information LM, which appropriately represents the relationship between the test pulse number and the shape of the test workpiece, is assumed to represent the relationship between the target pulse number and the shape of the workpiece W_target after processing based on the target pulse number. Thus, the model generation unit 211 can also be viewed as generating, in effect, model information LM that appropriately represents the relationship between the test pulse number and the shape of the test workpiece, while essentially generating model information LM that appropriately represents the relationship between the target pulse number and the shape of the workpiece W_target after processing based on the target pulse number. In other words, the model information LM is assumed to represent the relationship between the target pulse number and the shape of the processing target. Therefore, the model generation unit 211 can also be viewed as generating, in effect, model information LM that appropriately represents the relationship between the test pulse number and the shape of the test workpiece, while essentially generating model information LM that appropriately represents the relationship between the target pulse number and the shape of the processing target.

[0155] Furthermore, the prediction information PI contained in the model information LM generated through machine learning is also assumed to be a predicted value representing the amount of workpiece W_target being processed when the target position C is irradiated by the processing light EL a unit number of times. In other words, the model generation unit 211 can generate model information LM containing prediction information PI representing the predicted amount of workpiece W_test being processed when the target position C is irradiated by the processing light EL a unit number of times, by generating model information LM containing prediction information PI representing the predicted amount of workpiece W_target being processed when the target position C is irradiated by the processing light EL a unit number of times.

[0156] Thus, since the model information LM, which represents the relationship between the number of test pulses and the shape of the test workpiece, can be regarded as representing the relationship between the number of target pulses and the shape of the machining target, the model information LM can be used to predict the number of target pulses recommended for machining the workpiece W_target so that its shape becomes the desired machining target shape, based on the target value of the shape of the machined workpiece W_target, i.e., the machining target shape. In other words, when the machining target shape information related to the machining target shape is known (i.e., the machining target shape has been set), the model information LM can be used to predict the number of target pulses recommended for machining the workpiece W_target so that its shape becomes the desired machining target shape, based on the machining target shape information. Therefore, the model information LM can also be regarded as prediction information that can be used to generate target pulse number information (i.e., predict the target pulse number) as unknown information based on the machining target shape information, which is known information. Therefore, in step S21, the control data generation unit 212 calculates the number of target pulses based on the machining target shape information representing the machining target shape and the model information LM. In other words, the control data generation unit 212 uses the machining target shape information and model information LM to predict the recommended number of target pulses to be used in order to process the workpiece W_target so that the shape of the workpiece W_target becomes the desired machining target shape. In other words, the control data generation unit 212 treats the model information LM, which represents the relationship between the test pulse number and the test workpiece shape, as representing the relationship between the target pulse number and the machining target shape, and calculates the target pulse number based on the machining target shape information and model information LM, which represent the machining target shape. For example, as... Figure 17 As shown, the control data generation unit 212 can calculate the target pulse number as the output (prediction result) of the model information LM by inputting the processing target shape information representing the processing target shape into the model information LM.

[0157] Here, it is unlikely that only one target pulse number exists when processing the workpiece W_target to achieve the desired shape. That is, it is possible that the workpiece W_target, processed with one target pulse number, achieves the desired shape, and simultaneously, the workpiece W_target, processed with a different target pulse number, also achieves the desired shape. Therefore, in step S21, the control data generation unit 212 generates multiple target pulse number information representing different target pulse numbers. However, in step S21, the control data generation unit 212 may not generate multiple target pulse number information. In step S21, the control data generation unit 212 may also generate a single target pulse number information.

[0158] Back to Figure 12 Then, based on the target pulse number information and model information LM generated in step S21, the control data generation unit 212 predicts the shape of the workpiece W_target after processing, assuming that the workpiece W_target is processed using the target pulse number represented by the target pulse number information (step S22). Furthermore, the shape of the workpiece W_target predicted in step S22 is referred to as the predicted workpiece shape. In particular, since multiple target pulse number information is generated in step S21, in step S22, the control data generation unit 212 predicts multiple predicted workpiece shapes corresponding to each of the multiple target pulse number information.

[0159] Specifically, as described above, since the model information LM can be considered as representing the relationship between the target pulse number and the shape of the processing target, the model information LM can be used to predict the shape of the processed workpiece W_target after processing, assuming that the workpiece W_target is processed using the target pulse number (in this case, the shape of the processing target corresponding to the target pulse number). That is, when the target pulse number information related to the target pulse number is known (i.e., the target pulse number has been set), the model information LM can be used to predict the shape of the processed workpiece W_target, assuming that the processing light EL is irradiated to the irradiation target positions C1 to CN with the same number of times as the target pulse number, based on the target pulse number information. Therefore, the model information LM can also be considered as prediction information that can be used to predict the shape of the processed workpiece W_target after processing, which is unknown information, based on the target pulse number information, which is known information. Therefore, in step S22, the control data generation unit 212 can predict the shape of the workpiece based on the target pulse number information and the model information LM generated in step S21. In other words, in step S22, the control data generation unit 212, by treating the model information LM representing the relationship between the test pulse number and the test workpiece shape as representing the relationship between the target pulse number and the processing target shape, can predict the predicted workpiece shape based on the target pulse number information and the model information LM. For example, such as Figure 17 As shown, the control data generation unit 212 can also calculate the predicted workpiece shape as the output (prediction result) of the model information LM by inputting the target pulse number information representing the target pulse number into the model information LM.

[0160] Back to Figure 12 Then, the control data generation unit 212 selects one target pulse number from the multiple target pulse number information generated in step S21 that satisfies the predetermined selection criterion based on the predicted workpiece shape predicted in step S22 (step S23). For example, as Figure 17As shown, the control data generation unit 212 can select a target pulse number information that satisfies a predetermined selection criterion based on the difference between the predicted workpiece shape and the processing target shape, i.e., the difference Δ_target in the shape of the workpiece. As an example, the control data generation unit 212 can select a target pulse number information corresponding to the predicted workpiece shape that satisfies the condition of minimizing the difference Δ_target in the shape of the workpiece, as a target pulse number information that satisfies the predetermined selection criterion. In other words, the control data generation unit 212 can select from multiple target pulse number information that predicts the shape of the workpiece W_target after processing is closest to the processing target shape when the target pulse number represented by the target pulse number information is used for processing, as a target pulse number information that satisfies the predetermined selection criterion. In this case, the action of selecting a target pulse number information from multiple target pulse number information can also be considered equivalent to generating a target pulse number information in a way that the predicted workpiece shape is close to the processing target shape (i.e., in a way that the difference between the predicted workpiece shape and the processing target shape, i.e., the difference Δ_target in the shape of the workpiece, becomes smaller).

[0161] Furthermore, the control data generation unit 212 can also select from multiple target pulse number information that satisfies a different selection criterion than the selection criterion based on the shape difference Δ_target of the workpiece being processed. For example, as described above, the processing apparatus 1 processes the workpiece W_target, forming a rib-like structure on the workpiece W_target. In this case, the control data generation unit 212 can select from multiple target pulse number information that satisfies a selection criterion based on the characteristics of the rib-like structure formed when processing the workpiece W_target using the target pulse number information. In this case, the control data generation unit 212 can calculate the characteristics of the rib-like structure based on the predicted workpiece shape predicted in step S22 (i.e., the predicted shape of the rib-like structure formed when processing the workpiece W_target using the target pulse number information), and select a target pulse number information that satisfies the selection criterion based on the characteristics of the rib-like structure based on the calculated characteristics of the rib-like structure. For example, as described above, the rib-like structure has the effect of reducing the surface resistance of the workpiece W_target relative to the fluid. In this case, the characteristics of the rib-like structure can include the drag reduction effect of the rib-like structure. In this case, the control data generation unit 212 can select one target pulse number information from multiple target pulse number information that satisfies the following condition: assuming that the drag reduction effect of the rib-like structure formed when machining the workpiece W_target using the target pulse number represented by the target pulse number information is the highest.

[0162] Furthermore, if the control data generation unit 212 generates no multiple target pulse count information (i.e., a single target pulse count information) in step S21, the control data generation unit 212 may not perform steps S22 to S23. In this case, the single target pulse count information generated in step S21 can be used as the target pulse count information selected in step S23.

[0163] Then, based on the target pulse number information selected in step S23, the workpiece W_target is processed (step S24). Specifically, the control data generation unit 212 controls the processing apparatus 1 to process the test workpiece W_test based on the target pulse number information. For example, the control data generation unit 212 can send the target pulse number information as control data to the processing apparatus 1. The processing apparatus 1 (especially the control unit 14) can generate processing control information for controlling the processing apparatus 1 based on the target pulse number information sent from the data generation server 2. For example, the control unit 14 can generate processing control information for controlling the processing apparatus 1 to process the workpiece W_target so that the shape of the workpiece W_target becomes the desired processing target shape. Then, the processing apparatus 1 can process the workpiece W_target based on the generated processing control information.

[0164] After machining the workpiece W_target, the shape of the machined workpiece W_target is measured, which is the workpiece shape (step S24). In other words, the shape of the machined workpiece W_target after machining based on the target pulse number information selected in step S23 is the workpiece shape (step S24). For example, the machining apparatus 1 can use the measurement unit 12 to measure the workpiece shape. Then, the workpiece shape information (machining shape information), representing the measurement result of the workpiece shape obtained by the measurement unit 12, can be sent from the machining apparatus 1 to the data generation server 2 via the communication network 4. The model generation unit 211 of the data generation server 2 can acquire the workpiece shape information representing the measurement result of the workpiece shape obtained by the measurement unit 12.

[0165] Then, the control data generation unit 212 determines whether the shape of the workpiece to be processed meets the predetermined specifications (step S25). For example, if the difference between the shape of the workpiece to be processed and the shape of the target to be processed is less than the allowable amount, it is highly likely that the workpiece W_target has been properly processed based on the target pulse number information. Therefore, in this case, the control data generation unit 212 can determine that the shape of the workpiece to be processed meets the predetermined specifications. On the other hand, for example, if the difference between the shape of the workpiece to be processed and the shape of the target to be processed is greater than the allowable amount, it is highly likely that the workpiece W_target has not been properly processed based on the target pulse number information. Therefore, in this case, the control data generation unit 212 can determine that the shape of the workpiece to be processed does not meet the predetermined specifications.

[0166] If the determination in step S25 is that the shape of the workpiece meets the predetermined specifications (step S25: Yes), it is assumed that the target pulse number information selected in step S25 is suitable as the target pulse number information used to process the workpiece W_target so that the shape of the workpiece W_target becomes the predetermined processing target shape. Therefore, in this case, the control data generation unit 212 can generate control data representing the target pulse number information selected in step S25 and output the generated control data to the processing device 1.

[0167] If the determination in step S25 is that the shape of the workpiece does not meet the predetermined specifications (step S25: No), it is assumed that the target pulse number information selected in step S25 is not suitable as the target pulse number information used to process the workpiece W_target so that the shape of the workpiece W_target becomes the predetermined processing target shape. In this case, the learning of the model information LM used to generate the target pulse number information may not be sufficient. Therefore, the data generation server 2 performs the action of step S15 again, and performs machine learning again to generate (in this case, update) the model information LM. In particular, when performing machine learning again, the model generation unit 211 can use, in addition to the teaching data obtained through the actions of steps S11 to S13 (i.e., the teaching data that establishes a correlation between the test processing condition information and the test workpiece shape information), the data that establishes a correlation between the target pulse number information selected in step S23 and the workpiece shape information representing the shape of the workpiece measured in step S24 as new teaching data.

[0168] After performing machine learning again to generate (in this case, update) the model information LM, the data generation server 2 performs steps S21 to S25 again. That is, the control data generation unit 212 regenerates the target pulse number information based on the updated model information LM. Furthermore, regenerating the target pulse number information can also be considered equivalent to updating the target pulse number information (i.e., updating the target pulse number represented by the target pulse number information).

[0169] Here, since the model information LM has been updated, the target pulse count information generated based on the updated model information LM is more likely to be suitable for use in machining the workpiece W_target to achieve the desired machining target shape compared to the target pulse count information generated based on the unupdated model information LM. In other words, the target pulse count information generated based on the updated model information LM is more likely to be suitable for use in machining the workpiece W_target to achieve the desired machining target shape compared to the target pulse count information generated based on the unupdated model information LM.

[0170] In this case, when generating the target pulse number information based on the updated model information LM, compared to generating the target pulse number information based on the unupdated model information LM, it is expected that the workpiece shape difference Δ_target corresponding to the target pulse number information selected in step S23 will decrease. Therefore, updating the model information LM and generating (in other words, updating) the target pulse number information based on the updated model information LM can also be considered equivalent to generating (in other words, updating) the target pulse number information in a way that the workpiece shape difference Δ_target decreases.

[0171] Furthermore, when generating target pulse count information based on the updated model information LM, compared to generating target pulse count information based on the unupdated model information LM, it is expected that the probability of determining that the shape of the workpiece meets the specifications in step S25 will increase. In other words, it is expected that the difference between the shape of the workpiece and the target shape will decrease. Therefore, updating the model information LM and generating (in other words, updating) the target pulse count information based on the updated model information LM can also be considered equivalent to generating (in other words, updating) the target pulse count information in a way that increases the probability of determining that the shape of the workpiece meets the specifications (e.g., the difference between the shape of the workpiece and the target shape decreases).

[0172] (4) Technical effects

[0173] As explained above, in this embodiment, the data generation server 2 can use model information LM, which represents the relationship between the number of times processing light EL is irradiated at each of the target irradiation positions C1 to CN for processing the workpiece W and the processed shape of the workpiece W, to calculate the target pulse number. Therefore, compared with the case where the model information LM is not used, the time required to calculate the target pulse number is shortened.

[0174] Specifically, without using model information LM, the operator of the machining system SYS needs to manually set the target pulse number. Then, the workpiece W needs to be machined based on the set target pulse number, the shape of the machined workpiece W is measured, and the suitability of the manually set target pulse number is evaluated based on the measurement results. Furthermore, if it is determined that the manually set target pulse number is unsuitable (i.e., the shape of the machined workpiece W differs from the target shape), the operator needs to manually set the target pulse number again, and the suitability of the re-set target pulse number is evaluated again. Therefore, setting a suitable target pulse number requires a significant amount of time.

[0175] On the other hand, in this embodiment, firstly, since the target pulse number is calculated using model information LM, it is not necessary for the operator to manually set the target pulse number. Furthermore, since the model information LM is generated by actually machining the test workpiece W_test and using the measurement results of the shape of the actually machined test workpiece W_test, the model information LM is highly likely to represent the relationship between the target pulse number and the target shape of the workpiece W. Therefore, the target pulse number calculated using the model information LM is highly likely to be a suitable target pulse number. Therefore, the possibility of having to set the target pulse number again is reduced compared to the case where the model information LM is not used. Furthermore, in order to obtain teaching data for generating the model information LM, the test workpiece W_test is machined, and the shape of the machined test workpiece W_test is measured. However, since the machining and measurement of the test workpiece W_test are for obtaining teaching data, the number of machining and measurement operations performed on the test workpiece W_test is less than the number of machining and measurement operations performed to evaluate whether the target pulse number is suitable. For these reasons, the time required to set the target pulse number is shortened.

[0176] (5) Other variations

[0177] In the above description, data generation server 2 generates model information LM by performing machine learning. However, data generation server 2 can also generate model information LM without performing machine learning. For example, data generation server 2 can generate model information LM representing the relationship between test processing condition information and test workpiece shape information based on test processing condition information and test workpiece shape information.

[0178] In the above description, data generation server 2 generates model information LM based on test processing condition information and test workpiece shape information, and then generates target pulse count information based on the generated model information LM and processing target shape information. However, data generation server 2 can also generate target pulse count information without generating model information LM. For example, data generation server 2 can generate target pulse count information based on test processing condition information, test workpiece shape information, and processing target shape information without generating model information LM.

[0179] In the above description, the prediction information PI included in the model information LM represents the predicted amount of workpiece W when the target position C is irradiated with the processing light EL once, and is used as the predicted amount of workpiece W when the target position C is irradiated with the processing light EL a unit number of times. That is, the above description illustrates an example where the unit number of times is once. However, the unit number of times can also be two or more. In this case, the target pulse number can represent the number of times a unit irradiation action of irradiating the target position C with the processing light EL a unit number of times should be performed. Specifically, the target pulse number information can represent: the target pulse number as the number of times the first unit irradiation action of irradiating the target position C1 with the processing light EL a unit number of times should be performed, the target pulse number as the number of times the second unit irradiation action of irradiating the target position C2 with the processing light EL a unit number of times should be performed, ..., the target pulse number as the number of times the nth unit irradiation action of irradiating the target position Cn with the processing light EL a unit number of times should be performed.

[0180] In the above description, data generation server 2 generates control data representing the target pulse number. However, data generation server 2 can also generate control data different from the control data representing the target pulse number. For example, data generation server 2 can generate control data specifying arbitrary processing conditions (processing recipes) for processing device 1. In this case, data generation server 2 can generate model information LM representing the relationship between arbitrary processing conditions for processing workpiece W and the processed shape of workpiece W, based on test processing condition information representing arbitrary processing conditions and test workpiece shape, and calculate the target value of arbitrary processing conditions (target processing conditions, workpiece processing conditions) based on the generated model information and processing target shape information. That is, data generation server 2 can generate control data representing the target value (target processing conditions) of arbitrary processing conditions. Processing conditions may include conditions of processing light EL irradiating workpiece W (irradiation conditions). Irradiation conditions may include at least one of the following conditions: conditions related to the intensity of processing light EL, conditions related to the energy (e.g., pulse energy) of processing light EL, conditions related to the number of times processing light EL is irradiated, conditions related to the irradiation position of processing light EL, and conditions related to the burst mode of processing light EL. Furthermore, for example, as described in Japanese Patent Application Publication No. 2016-524864, a burst mode can refer to an operating mode capable of dividing each pulse constituting the processing light EL into a desired number of pulses. In this case, the conditions related to the burst mode can include conditions related to the number of pulse divisions. Additionally, the target number of pulses for irradiating the processing light EL to each of the target irradiation positions C1 to CN can also be considered equivalent to conditions related to the number of irradiations and the irradiation position of the processing light EL. In this case, the control data indicating the target number of pulses can also be considered as control data specifying the processing conditions. The processing conditions can include the movement conditions of at least one of the processing head 112 and the platform 132. The movement conditions can include, for example, at least one of the following conditions: conditions related to movement speed, conditions related to movement amount, conditions related to movement direction, and conditions related to movement timing.

[0181] In the above description, the processing apparatus 1 processes the workpiece W by irradiating it with a processing beam containing a single processing light EL. However, the processing apparatus 1 can also process the workpiece W by irradiating it with a processing beam containing multiple processing light ELs. In this case, the data generation server 2 can also process the workpiece W by performing... Figure 12The data generation action shown generates control data. However, in this case, the model information LM can represent the relationship between the number of times a processing beam containing multiple processing light ELs is irradiated onto the target positions C1 to CN for processing the workpiece W, and the processed shape of the workpiece W. The prediction information PI contained in the model information LM can represent the predicted value of the processing amount of the test workpiece W_test when the target position C is irradiated by a processing beam containing multiple processing light ELs a unit number of times.

[0182] In the case where the processing apparatus 1 processes the workpiece W by irradiating it with a processing beam containing multiple processing light ELs, the processing apparatus 1 can process the workpiece W by forming interference fringes on the surface of the workpiece W obtained by interfering with the multiple processing light ELs. Furthermore, an example of a processing apparatus that processes the workpiece W by forming interference fringes on the surface of the workpiece W is described in U.S. Patent Application Publication No. 2022 / 0258289. In this case, the data generation server 2 can also... Figure 12 The data generation action shown generates control data. However, in this case, the model information LM can also represent the relationship between the number of interference fringes formed at the irradiation target positions C1 to CN respectively for machining the workpiece W, and the machined shape of the workpiece W. The prediction information PI contained in the model information LM can also represent the predicted value of the machining amount of the test workpiece W_test when interference fringes are formed at the irradiation target position C a unit number of times.

[0183] In the above description, the processing apparatus 1 processes the workpiece W by irradiating it with processing light EL. However, the processing apparatus 1 can also process the workpiece W by irradiating it with an arbitrary energy beam. In this case, the processing apparatus 1 may have a beam source capable of irradiating an arbitrary energy beam, either in addition to or in place of the processing light source 111. As an example of an arbitrary energy beam, at least one of charged particle beams and electromagnetic waves can be cited. As an example of a charged particle beam, at least one of electron beams and ion beams can be cited.

[0184] In the description above, the measuring unit 12 is provided separately from the processing unit 11. However, the measuring unit 12 may also be integrated with the processing unit 11, and the workpiece W may be measured via the focusing optical system (fθ lens 1123) of the processing unit 11 that focuses the processing light EL. Such a processing and measuring apparatus is disclosed, for example, in International Publication No. 2021 / 024480.

[0185] In the description above, the information related to the processing light EL included in the reference information may also include information related to the light intensity distribution or luminous flux distribution of the processing light EL in the forward direction (in other words, the illumination direction) of the processing light EL.

[0186] In the embodiments described above, the data generation server 2 may also calculate the light penetration depth based on the processing results of the test workpiece W_test, or alternatively, based on the processing results of the workpiece W. In this case, the data generation server 2 can process and measure the predetermined portion to be removed from the workpiece W to determine the light penetration depth. To determine the light penetration depth, the workpiece W after processing can be measured to obtain information related to the initial shape of the workpiece W.

[0187] In the above description, the data generation server 2 uses the response surface methodology to determine the optimal processing conditions, but it is not limited to the response surface methodology. For example, other methods such as design of experiments, machine learning, and Bayesian inference can be used.

[0188] As described above, platform unit 13 may also have multiple platforms 132.

[0189] (5) Notes

[0190] The following notes further disclose the implementation methods described above.

[0191] [Postscript 1]

[0192] A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Input first information related to a target shape with height variation in at least one dimension, one or more second pieces of information related to processing conditions for processing the target shape, and one or more third pieces of information related to the height distribution in the at least one dimension of the object processed using the processing conditions; and The output consists of one or more processing conditions and processing shape prediction results corresponding to the one or more processing conditions, obtained by inputting the first information into a machine learning model that has been learned using the one or more second information and the one or more third information.

[0193] The requirements of the above-described embodiments can be appropriately combined. It is also possible to omit some of the requirements of the above-described embodiments. The requirements of the above-described embodiments can be appropriately substituted with the requirements of other embodiments. Furthermore, all publicly available publications and U.S. patent publications relating to the devices referenced in the above-described embodiments may be incorporated into this specification, provided that permitted by law.

[0194] Furthermore, the present invention may be appropriately modified without departing from the spirit or concept of the invention as can be read from the claims and the specification in their entirety, and the data generation method, cloud system, data generation device, computer program, recording medium and processing device that are modified accordingly are also included within the technical concept of the present invention.

[0195] [Explanation of Symbols]

[0196] SYS: Machining System

[0197] 1: Processing equipment

[0198] 11: Processing Unit

[0199] 12: Measurement Unit

[0200] 13: Platform Unit

[0201] 14: Control Unit

[0202] 2: Data generation server

[0203] 21: Computing device

[0204] 211: Model Generation Department

[0205] 212: Control Data Generation Unit

[0206] 3: Client device

[0207] EL: Processing light

[0208] W: Workpiece

[0209] W_test: Test artifact

[0210] W_target: The workpiece to be processed.

[0211] C: Location of the irradiated target

[0212] LM: Model Information

Claims

1. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Obtain test processing condition information, which includes the number of pulses of the pulsed energy beam that should be irradiated at each of the multiple irradiation positions of the test workpiece in order to process the test workpiece into the target shape; The shape of the test workpiece after processing based on the aforementioned test processing conditions is measured, i.e., the shape of the test workpiece; and Based on the test processing condition information, the shape of the test workpiece, and the predicted shape of the portion processed by irradiating the pulse energy beam with a unit number of pulses, the target number of pulses to be irradiated by the pulse energy beam at multiple irradiation positions of the workpiece to be processed is calculated.

2. The data generation method according to claim 1, wherein The prediction information includes information on at least one of the width and depth of the portion processed by the pulse energy beam of the unit pulse number.

3. The data generation method according to claim 1 or 2, wherein The prediction information represents a predicted value of the amount of workpiece processed when the pulsed energy beam is irradiated with a unit number of pulses.

4. The data generation method according to claim 3, wherein The prediction information includes a predicted value of the amount of machining at the first test position and a predicted value of the amount of machining at the second test position of the test workpiece when the pulsed energy beam is irradiated with a unit number of pulses at a first test position and a second test position different from the first test position.

5. The data generation method according to claim 4, wherein... The target pulse number is calculated using the difference between the target shape and the predicted shape of the workpiece being processed based on the predicted information.

6. The data generation method according to any one of claims 1 to 5, comprising the following steps: The prediction information is generated in such a way that the difference between the shape of the test workpiece and the predicted shape of the test workpiece based on the prediction information becomes smaller.

7. The data generation method according to any one of claims 1 to 6, wherein Calculating the target pulse number includes the following steps: Based on the test processing condition information and the shape of the test workpiece, the prediction information is generated; and Based on the predicted information, the target pulse number is calculated.

8. The data generation method according to any one of claims 1 to 7, wherein Generating the target pulse count includes the following steps: further generating target pulse count information based on the target shape, wherein the target pulse count information represents the target number of pulse energy beams that should be irradiated onto each of the plurality of irradiation positions in order to set the shape of the workpiece to the target shape by removing the machining process.

9. The data generation method according to any one of claims 1 to 8, further comprising the following steps: Based on the test processing condition information, the predicted shape of the test workpiece is predicted; and The prediction information includes model parameters. The model parameters are generated by reducing the difference between the test workpiece shape and the predicted shape.

10. The data generation method according to claim 9, comprising the following steps: The model parameters are updated in such a way that the difference between the predicted shape and the target shape decreases; and The predicted shape is based on the predicted information, predicting the shape of the test workpiece when it is subjected to removal processing by irradiating the test workpiece with the pulsed energy beam.

11. The data generation method according to claim 9 or 10, wherein Generating the target pulse count includes the following steps: by considering the relationship between the shape of the test workpiece and the test pulse count as representing the relationship between the target shape and the target pulse count, the target pulse count is generated based on the target shape and the prediction information.

12. The data generation method according to any one of claims 1 to 8, wherein Generating the target pulse number includes the following steps: generating the target pulse number information in such a way that the predicted shape of the workpiece being processed, as predicted by the prediction information based on the prediction information, is close to the target shape when the removal process is performed by irradiating the plurality of irradiation positions with the pulse energy beam represented by the target pulse number.

13. The data generation method according to any one of claims 9 to 12, wherein The prediction information includes information that can be used to make the following predictions: as a predicted value of the amount of workpiece processed when the pulse energy beam is irradiated at the first irradiation position of the plurality of irradiation positions with a unit number of pulses, a predicted value of the amount of workpiece processed at the first irradiation position, and a predicted value of the amount of workpiece processed at a position different from the first irradiation position.

14. The data generation method according to claim 13, wherein The prediction information includes information that can be used to make the following predictions: as a prediction of the amount of processing of the test workpiece when the pulse energy beam is irradiated at a second irradiation position, which is different from the first irradiation position, with a unit number of pulses, a prediction of the amount of processing at the second irradiation position, and a prediction of the amount of processing at a position different from the second irradiation position.

15. The data generation method according to any one of claims 9 to 14, wherein Generating the target pulse number includes the following steps: generating at least one target pulse number information.

16. The data generation method according to claim 15, wherein Generating the at least one target pulse number information includes the following steps: Based on the predicted information and the target shape, multiple target pulse count information are generated; and From the plurality of target pulse count information, at least one target pulse count information that satisfies a predetermined selection criterion is selected as the at least one target pulse count information.

17. The data generation method according to claim 16, wherein The selection criterion includes a first criterion, which is a criterion relating to the difference between the predicted shape of the workpiece to be processed, as predicted based on the prediction information, and the target shape, when removal processing is performed by irradiating the plurality of irradiation positions with the pulse energy beam represented by the target pulse number information.

18. The data generation method according to claim 16 or 17, wherein The processing apparatus removes material from the workpiece, thereby forming a rib-like structure on the surface of the workpiece. This rib-like structure reduces the friction between the workpiece surface and the fluid. The selection criterion includes a second criterion, which is a criterion relating to the characteristics of the rib-like structure formed on the surface of the workpiece being processed during removal processing, assuming that the pulse energy beam is irradiated to each of the plurality of irradiation positions by the number of pulses represented by the target pulse number information.

19. The data generation method according to any one of claims 1 to 18, wherein The test workpiece is the first test workpiece, the test processing condition information is the first test processing condition information, and the shape of the test workpiece is the shape of the first test workpiece. The data generation method includes the following steps: Obtain second test processing condition information, wherein the second test processing condition information indicates the number of pulses of the pulse energy beam to be irradiated at each of the multiple irradiation positions of the second test workpiece, which is different from the first test workpiece; The pulsed energy beam is irradiated onto the second test workpiece based on the second test processing condition information; and The shape of the second test workpiece is measured in relation to the shape of the second test workpiece after it has been processed by the pulsed energy beam; and The prediction information is generated based on the second test processing conditions information and the shape of the second test workpiece.

20. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Input test processing condition information, which indicates the number of pulses of the pulse energy beam that should be irradiated at each of the multiple irradiation positions of the test workpiece in order to process the test workpiece into the target shape; The shape of the test workpiece is measured in relation to the shape of the test workpiece after it has been processed by the pulsed energy beam based on the test processing condition information. Based on the test processing condition information and the shape of the test workpiece, machine learning is performed to generate model information, which includes prediction information of the shape processed by the pulse energy beam with a unit number of pulses; as well as Based on the model information generated through the machine learning and the target shape, the target number of pulses of the pulsed energy beam to be irradiated at multiple irradiation positions of the workpiece to be processed is calculated.

21. The data generation method according to claim 20, wherein The model information includes information on at least one of the width and depth of the portion processed by the pulse energy beam of the unit pulse number.

22. The data generation method according to claim 20 or 21, wherein The model information represents a predicted value of the amount of workpiece processed when the pulsed energy beam is irradiated with a unit number of pulses.

23. The data generation method according to claim 22, wherein The model information includes a predicted value of the amount of machining at the first test position and a predicted value of the amount of machining at the second test position of the test workpiece when the pulsed energy beam is irradiated with a unit number of pulses at a first test position and a second test position different from the first test position.

24. The data generation method according to claim 23, wherein The target pulse number is calculated using the difference between the target shape and the predicted shape of the workpiece being processed based on the model information.

25. The data generation method according to any one of claims 20 to 24, comprising the following steps: The model information is generated in such a way that the difference between the shape of the test workpiece and the predicted shape of the test workpiece based on the model information becomes smaller.

26. The data generation method according to any one of claims 20 to 25, wherein Calculating the target pulse number includes the following steps: Based on the test processing condition information and the shape of the test workpiece, the model information is generated; and Based on the model information, the target pulse number is calculated.

27. The data generation method according to any one of claims 20 to 26, wherein Generating the target pulse count includes the following steps: further generating target pulse count information based on the target shape, wherein the target pulse count information represents the target number of pulse energy beams that should be irradiated onto each of the plurality of irradiation positions in order to set the shape of the workpiece to the target shape by removing the machining process.

28. The data generation method according to any one of claims 20 to 27, further comprising the following steps: Based on the test processing condition information, the predicted shape of the test workpiece is predicted; and The model information includes model parameters. The model parameters are generated by reducing the difference between the test workpiece shape and the predicted shape.

29. The data generation method according to claim 28, comprising the following steps: The model parameters are updated in such a way that the difference between the predicted shape and the target shape decreases; and The predicted shape is based on the model information and predicts the shape of the test workpiece when it is subjected to removal processing by irradiating the test workpiece with the pulsed energy beam.

30. The data generation method according to claim 28 or 29, wherein Generating the target pulse count includes the following steps: by considering the relationship between the shape of the test workpiece and the test pulse count as representing the relationship between the target shape and the target pulse count, the target pulse count is generated based on the target shape and the prediction information.

31. The data generation method according to any one of claims 20 to 27, wherein Generating the target pulse number includes the following steps: generating the target pulse number information in such a way that the predicted shape of the workpiece being processed, as predicted by the model information during the removal process by irradiating the plurality of irradiation positions with the pulse energy beam represented by the target pulse number, is close to the target shape.

32. The data generation method according to any one of claims 28 to 31, wherein The model information includes information that can be used to make the following predictions: a predicted value of the amount of workpiece processed when the pulsed energy beam is irradiated at the first irradiation position of the plurality of irradiation positions with a unit number of pulses; a predicted value of the amount of workpiece processed at the first irradiation position; and a predicted value of the amount of workpiece processed at a position different from the first irradiation position.

33. The data generation method according to claim 32, wherein... The model information includes information that can be used to make the following predictions: a predicted value of the amount of processing of the test workpiece when the pulse energy beam is irradiated at a second irradiation position, which is different from the first irradiation position, with a unit number of pulses; a predicted value of the amount of processing at the second irradiation position; and a predicted value of the amount of processing at a position different from the second irradiation position.

34. The data generation method according to any one of claims 28 to 33, wherein Generating the target pulse number includes generating at least one target pulse number information.

35. The data generation method according to claim 34, wherein Generating the at least one target pulse number information includes the following steps: Based on the model information and the target shape, multiple target pulse count information are generated; and From the plurality of target pulse count information, at least one target pulse count information that satisfies a predetermined selection criterion is selected as the at least one target pulse count information.

36. The data generation method according to claim 35, wherein The selection criterion includes a first criterion, which is a criterion relating to the difference between the predicted shape of the workpiece to be processed, as predicted based on the prediction information, and the target shape, when removal processing is performed by irradiating the plurality of irradiation positions with the pulse energy beam represented by the target pulse number information.

37. The data generation method according to claim 35 or 36, wherein The processing apparatus removes material from the workpiece, thereby forming a rib-like structure on the surface of the workpiece. This rib-like structure reduces the friction between the workpiece surface and the fluid. The selection criterion includes a second criterion, which is a criterion relating to the characteristics of the rib-like structure formed on the surface of the workpiece being processed during removal processing, assuming that the pulse energy beam is irradiated to each of the plurality of irradiation positions by the number of pulses represented by the target pulse number information.

38. The data generation method according to any one of claims 20 to 37, wherein The test workpiece is the first test workpiece, the test processing condition information is the first test processing condition information, and the shape of the test workpiece is the shape of the first test workpiece. The data generation method includes the following steps: Obtain second test processing condition information, wherein the second test processing condition information indicates the number of pulses of the pulse energy beam to be irradiated at each of the multiple irradiation positions of the second test workpiece, which is different from the first test workpiece; Based on the second test processing condition information, the pulsed energy beam is irradiated onto the second test workpiece; and The shape of the second test workpiece is measured in relation to the shape of the second test workpiece after it has been processed by the pulsed energy beam; and The model information is generated based on the second test processing conditions information and the shape of the second test workpiece.

39. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Obtain model information representing the relationship between target pulse count information and target shape information, wherein the target pulse count information represents the target number of pulses that should be irradiated onto each of the multiple irradiation positions of the workpiece, and the target shape information is information about the target shape of the workpiece after removal processing; and Based on the model information and the target shape information, the target pulse count information is generated as the control data.

40. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Obtain model information that can be used to make the following predictions: a predicted value of the amount of workpiece processed when the pulsed energy beam is irradiated at a first irradiation position out of a plurality of irradiation positions of the workpiece at a unit number of pulses; a predicted value of the amount of workpiece processed at the first irradiation position; and a predicted value of the amount of workpiece processed at a position different from the first irradiation position; and Based on the model information and the target shape information related to the target shape of the workpiece after removal processing, target pulse number information is generated as the control data. The target pulse number information indicates the target number of pulses that should be irradiated to each of the multiple irradiation positions of the workpiece.

41. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Obtain test processing condition information for machining the test workpiece into the target shape; The shape of the test workpiece after processing based on the aforementioned test processing conditions is measured, i.e., the shape of the test workpiece; and Based on the test processing condition information, the shape of the test workpiece, and the predicted shape of the workpiece processed by the pulse energy beam, the processing conditions of the workpiece are calculated.

42. A data generation method for generating control data for controlling a processing apparatus capable of removing material from a workpiece by irradiating the surface of the workpiece with a pulsed energy beam, the data generation method comprising the following steps: Generate first test processing condition information, wherein the first test processing condition information indicates the number of pulses of the pulse energy beam to be irradiated at each of the multiple irradiation positions of the first test workpiece; Obtain first test shape information, which is information about the shape of the first test workpiece after removal processing by irradiating the first test workpiece with the pulsed energy beam based on the first test processing conditions information; Generate second test processing condition information, which indicates the number of pulses of the pulse energy beam that should be irradiated at each of the multiple irradiation positions of the second test workpiece in order to process the second test workpiece into the target shape; Acquire second test shape information, which is information about the shape of the second test workpiece after removal processing by irradiating the second test workpiece with the pulsed energy beam based on the second test processing conditions; and Based on the first test shape information, the second test shape information, the second test processing conditions, and the target shape, calculate the target number of pulses that should be irradiated by the pulse energy beam at multiple irradiation positions of the workpiece to be processed.

43. A computer program that causes a computer to perform the data generation method according to any one of claims 1 to 42.

44. A recording medium having a computer program as described in claim 43.

45. A processing apparatus for removing material from a workpiece using control data generated by the data generation method according to any one of claims 1 to 42.

46. ​​The processing apparatus according to claim 45, further comprising a control device for generating the control data by performing the data generation method.

Images