Method for managing calibration data and shape measuring instrument

Abstract

To provide a method for managing calibration data capable of easily determining quality of calibration data of a probe, and a shape measuring instrument.SOLUTION: A method for managing calibration data comprises: a calibration data acquisition step (calibration data acquisition unit 54) for acquiring new calibration data 42 for a probe 26; a prediction step (automatic trend management unit 62) for predicting measurement data 39 based on the calibration data 42 acquired in the calibration data acquisition step, using a machine learning model 44 that inputs the calibration data 42 and outputs the measurement data 39; and a determination step (automatic trend management unit 62) for determining quality of the calibration data 42 acquired in the calibration data acquisition step based on the measurement data 39 predicted in the prediction step and design value information 46 of a workpiece W.SELECTED DRAWING: Figure 16

Description

Technical Field

[0001] The present invention relates to a method for managing calibration data of a probe of a shape measuring machine and a shape measuring machine.

Background Art

[0002] Conventionally, there is a three-dimensional coordinate measuring machine (shape measuring machine) having a drive unit for displacing the position and orientation of a probe, and bringing this probe into contact (probing) with a plurality of measurement elements (for example, a straight line, a circular hole, a plane, a sphere, etc.) formed on a workpiece (object to be measured) to perform various measurements such as dimensions and shapes of the measurement elements. This three-dimensional coordinate measuring machine acquires measurement data of the workpiece corresponding to the measurement items based on the coordinate values of the probing points obtained by measuring one or more measurement items (radius, diameter, roundness, etc.) of the measurement elements of the workpiece with the probe according to a predetermined measurement plan and the calibration data of the probe (see Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the three-dimensional coordinate measuring machine described in Patent Document 1, calibration work of the probe and acquisition of calibration data are periodically executed. At this time, conventionally, the user sets thresholds such as calibration data (offset amounts in the XYZ directions, probe diameter, etc.) and compares the newly acquired calibration data with the thresholds to manage the quality and trends of this calibration data. However, in this case, since the method for setting the thresholds is not defined, the user does not know what thresholds should be set. Also, for the user, setting the thresholds of the calibration data itself is complicated.

[0005] The present invention has been made in view of such circumstances, and an object thereof is to provide a calibration data management method and a shape measuring machine capable of easily determining the quality of calibration data of a probe.

Means for Solving the Problems

[0006] A calibration data management method for achieving the object of the present invention is a calibration data acquisition step of acquiring new calibration data of a probe, and based on the calibration data acquired in the calibration data acquisition step, using a machine learning model that takes the calibration data as an input and outputs measurement data, a prediction step of predicting measurement data, and a determination step of determining the quality of the calibration data acquired in the calibration data acquisition step based on the measurement data predicted in the prediction step and the design value information of the work, in a shape measuring machine that acquires measurement data of a work corresponding to the measurement items of a measurement plan based on the coordinate values of probing points obtained by measuring one or more measurement items of the work with a probe provided for coordinate measurement and the calibration data of the probe.

[0007] According to this calibration data management method, it is possible to automatically determine the quality of calibration data without the user setting a threshold value for the calibration data.

[0008] In a calibration data management method according to another aspect of the present invention, a first data collection step of collecting a plurality of first correspondence data in which a plurality of measurement data are acquired by a shape measuring machine according to a measurement plan and the measurement data and the calibration data used by the shape measuring machine to acquire the measurement data are associated with each other, and a machine learning model generation step of generating a machine learning model based on the plurality of first correspondence data collected in the first data collection step, and in the prediction step, the measurement data is predicted using the machine learning model generated in the machine learning model generation step. Thereby, a machine learning model can be generated.

[0009] In the calibration data management method according to another aspect of the present invention, there are a plurality of measurement plans. In the first data collection step, a plurality of first corresponding data are collected for each measurement plan. In the machine learning model generation step, a machine learning model is generated for each measurement plan. In the prediction step, calibration data for each measurement plan is input into the machine learning model corresponding to the measurement plan to predict measurement data for each measurement plan. In the determination step, based on the measurement data for each measurement plan predicted in the prediction step and the design value information, the quality of the calibration data is determined. Thereby, the quality of the calibration data can be easily determined.

[0010] In the calibration data management method according to another aspect of the present invention, when the calibration data is determined to be good in the determination step, a measurement step of acquiring measurement data with a shape measuring machine according to the measurement plan, and a second data collection step of collecting second corresponding data associating the measurement data measured in the measurement step and the calibration data determined to be good in the determination step. And a re-learning step of performing re-learning of the machine learning model corresponding to the measurement plan based on the second corresponding data collected in the second data collection step. By performing re-learning of the machine learning model, the prediction accuracy of the measurement data, that is, the determination accuracy of the quality of the calibration data can be improved.

[0011] In the calibration data management method according to another aspect of the present invention, when the calibration data is determined to be good in the determination step, a measurement step of acquiring measurement data with a shape measuring machine according to the measurement plan, and a second data collection step of collecting second corresponding data associating the measurement data measured in the measurement step and the calibration data determined to be good in the determination step. And a re-learning step of performing re-learning of the machine learning model based on the second corresponding data collected in the second data collection step. By performing re-learning of the machine learning model, the prediction accuracy of the measurement data, that is, the determination accuracy of the quality of the calibration data can be improved.

[0012] In the calibration data management method according to another aspect of the present invention, when it is determined in the determination step that the calibration data is negative, the calibration data acquisition step, the prediction step, and the determination step are repeatedly executed. Thereby, good calibration data can be obtained.

[0013] In the calibration data management method according to another aspect of the present invention, when it is determined in the determination step that the calibration data is negative, it has a notification step of notifying warning information. Thereby, the user can recognize that the calibration data is defective.

[0014] A shape measuring machine for achieving the object of the present invention includes a probe for coordinate measurement, and based on the coordinate values of probing points obtained by measuring one or more measurement items of a workpiece with the probe according to a predetermined measurement plan and the calibration data of the probe, in a shape measuring machine that acquires measurement data of the workpiece corresponding to the measurement items of the measurement plan, a calibration data acquisition unit that acquires new calibration data of the probe, and based on the calibration data acquired by the calibration data acquisition unit, a prediction unit that predicts measurement data using a machine learning model that takes the calibration data as an input and outputs the measurement data, and a determination unit that determines the quality of the calibration data acquired by the calibration data acquisition unit based on the measurement data predicted by the prediction unit and the design value information of the workpiece.

[0015] In a shape measuring machine according to another aspect of the present invention, a first data collection unit that collects a plurality of first correspondence data in which a plurality of measurement data are acquired according to a measurement plan and the measurement data and the calibration data used by the shape measuring machine to acquire the measurement data are associated, and a machine learning model generation unit that generates a machine learning model based on the plurality of first correspondence data collected by the first data collection unit, and the prediction unit predicts the measurement data using the machine learning model generated by the machine learning model generation unit.

Effect of the Invention

[0016] The present invention can easily determine the quality of the calibration data of the probe.

Brief Description of the Drawings

[0017] [Figure 1] It is a schematic diagram of a three-dimensional coordinate measuring machine. [Figure 2] It is an enlarged perspective view of a probe head and a probe. [Figure 3] It is a functional block diagram of a control device. [Figure 4] It is an explanatory diagram for explaining the acquisition of probing data of a measurement point (probing point) by a probing data acquisition unit. [Figure 5] It is a graph showing the relationship between calibration data (probe diameter) and measurement data (radius). [Figure 6] It is an explanatory diagram for explaining the generation of a machine learning model. [Figure 7] It is a flowchart showing the flow of the collection process of corresponding data for each measurement item of a measurement plan by a first data collection unit. [Figure 8] It is a flowchart showing the flow of probe calibration (step S1) in FIG. 7. [Figure 9] It is an explanatory diagram for explaining the association of probing data for each measurement item of measurement plan A with calibration data by a calibration ID. [Figure 10] It is a diagram showing an example of corresponding data for each measurement plan used for machine learning of a machine learning model. [Figure 11] It is an explanatory diagram for explaining the generation of a machine learning model for each measurement plan by a machine learning model generation unit. [ [Figure 12] It is an explanatory diagram for explaining the prediction of measurement data for each measurement plan by a machine learning model. [Figure 13] It is a diagram showing a part of the actual corresponding data used for the generation of a machine learning model. [Figure 14] It is a table showing the generalization performance for each measurement item of a machine learning model generated using the corresponding data shown in FIG. 13. [Figure 15] It is an explanatory diagram for explaining the determination of the quality of calibration data and the re-learning of a machine learning model. [Figure 16]It is an explanatory diagram for explaining the function of the automatic trend management unit in FIG. 15. [Figure 17] It is an explanatory diagram for explaining an example of individual determination by the automatic trend management unit. [Figure 18] It is an explanatory diagram for explaining the relearning of the machine learning model when the automatic trend management unit makes a good determination on the calibration data. [Figure 19] It is an explanatory diagram for explaining the functions of the notification control unit and the repetition control unit when the automatic trend management unit makes a bad determination on the calibration data. [Figure 20] It is an explanatory diagram showing an example of warning information displayed on the display unit. [Figure 21] It is a flowchart showing the flow of generation of the machine learning model and management of calibration data in the three-dimensional coordinate measuring machine.

Embodiments for Carrying Out the Invention

[0018] [Overall Configuration of Three-Dimensional Coordinate Measuring Machine] FIG. 1 is a schematic diagram of a three-dimensional coordinate measuring machine 10 corresponding to the shape measuring machine of the present invention. The XYZ axes orthogonal to each other in FIG. 1 indicate the XYZ directions of the machine coordinate system defined based on the machine coordinate origin unique to the three-dimensional coordinate measuring machine 10.

[0019] As shown in FIG. 1, the three-dimensional coordinate measuring machine 10 performs shape measurement of one or more measurement items (measurement elements) of the work W according to one or more predetermined measurement plans using a probe 26 for coordinate measurement corresponding to contact measurement. Here, in the present embodiment, a plurality of measurement plans for the work W are prepared, and it is assumed that a plurality of measurement items are defined for each individual measurement plan. The measurement items of the work W include, for example, the radius, diameter, and roundness of the work W. In addition, the shape of the work W here includes, for example, various dimensional shapes such as length and diameter in addition to the three-dimensional shape, two-dimensional shape, surface shape, and contour shape of the work W. Note that the shape and type of the work W to be measured are not particularly limited.

[0020] The three-dimensional coordinate measuring machine 10 includes a gantry 12, a table 14 (surface plate) provided on the gantry 12, a right Y carriage 16R and a left Y carriage 16L erected at both ends of the table 14, and an X guide 18 connecting the upper parts of the right Y carriage 16R and the left Y carriage 16L. The right Y carriage 16R, the left Y carriage 16L, and the X guide 18 constitute a gantry frame 19.

[0021] On the upper and side surfaces of both ends of the table 14 in the X direction, sliding surfaces are formed along the Y direction on which the right Y carriage 16R and the left Y carriage 16L slide. Note that air bearings (not shown) are provided on the right Y carriage 16R and the left Y carriage 16L at positions facing the sliding surfaces of the table 14. Thereby, the right Y carriage 16R and the left Y carriage 16L are movable in the Y direction together with the X guide 18.

[0022] An X carriage 20 is attached to the X guide 18. On the X guide 18, a sliding surface along which the X carriage 20 slides is formed in the X direction. Also, air bearings (not shown) are provided on the X carriage 20 at positions facing the sliding surface of the X guide 18. Thereby, the X carriage 20 is movable in the X direction.

[0023] A Z carriage 22 (also referred to as a Z spindle) is attached to the X carriage 20. Also, air bearings (not shown) for guiding the Z carriage 22 in the Z direction are provided on the X carriage 20. Thereby, the Z carriage 22 is held movably in the Z direction by the X carriage 20. At the lower end of this Z carriage 22, a probe head 24 for selectively holding various probes including the probe 26 of the present invention is provided.

[0024] In addition, the three-dimensional coordinate measuring machine 10 is provided with an XYZ drive unit 27 (see FIG. 3) that executes the movement of the gantry frame 19 in the Y direction, the movement of the X carriage 20 in the X direction, and the movement of the Z carriage 22 in the Z direction. The XYZ drive unit 27 is a known actuator composed of, for example, a motor or the like. By driving the XYZ drive unit 27, the probe head 24 (probe 26) can be moved in the XYZ directions.

[0025] Furthermore, although not shown in the figure, a Y linear scale is provided at the end of the right Y carriage 16R side of the table 14, an X linear scale is provided on the X guide 18, and a Z linear scale is provided on the Z carriage 22. Additionally, the three-dimensional coordinate measuring machine 10 is provided with a reading head 27A (see FIG. 3) that reads the XYZ linear scales respectively. The detection result of this reading head 27A is output to the control device 30 via the controller 29.

[0026] FIG. 2 is an enlarged perspective view of the probe head 24 and the probe 26. As shown in FIG. 2, the probe head 24 is, for example, a 5-axis simultaneous control head equipped with a stepless positioning mechanism. The probe head 24 is provided with a rotation drive unit 28 (see FIG. 3) such as a motor that rotates the probe 26 in the circumferential direction θ1 around the rotation axis parallel to the Z direction and in the circumferential direction θ2 around the rotation axis perpendicular to the Z direction respectively. By driving the rotation drive unit 28, the postures of the probe 26 in the circumferential directions θ1 and θ2 around the axis can be displaced.

[0027] In addition, the probe head 24 is provided with a rotation angle detection unit 24A (see FIG. 3) such as a rotary encoder that detects the rotation angles of the probe 26 in the circumferential directions θ1 and θ2 around the axis respectively. The detection result of this rotation angle detection unit 24A is output to the control device 30 via the controller 29.

[0028] The probe 26 (stylus) is detachably attached to the probe head 24. This probe 26 is a contact type touch trigger probe and has a known contact 26a (probing sphere). Note that the type of the probe 26 is not particularly limited.

[0029] Returning to FIG. 1, the three-dimensional coordinate measuring machine 10 includes a controller 29 that controls the XYZ drive unit 27 and the rotation drive unit 28 shown in FIG. 3 described later to control the movement of the probe head 24, that is, the displacement of the position and orientation of the probe 26.

[0030] When the three-dimensional coordinate measuring machine 10 is in the manual measurement mode, the controller 29 drives the XYZ drive unit 27 and the rotation drive unit 28 according to an operation input from a user (operator), and for each measurement item of the workpiece W defined in the measurement plan, the probe 26 is brought into contact with a plurality of measurement points (probing points) corresponding to the measurement item. Further, when the three-dimensional coordinate measuring machine 10 is in the automatic measurement mode, the controller 29 drives the XYZ drive unit 27 and the rotation drive unit 28 under the control of the control device 30, and for each measurement item of the workpiece W according to the measurement plan, the probe 26 is brought into contact with a plurality of measurement points corresponding to the measurement item respectively.

[0031] Connected to the controller 29 are a contact detection sensor (not shown) of the contact touch trigger type probe 26, the aforementioned reading head 27A, and the rotation angle detection unit 24A. Then, at the moment when the controller 29 detects that the probe 26 has come into contact with the measurement points of each measurement item of the workpiece W by the contact detection sensor, the controller 29 acquires the detection results of the reading head 27A and the rotation angle detection unit 24A respectively and outputs them to the control device 30.

[0032] The control device 30 (computer) is communicably connected to the controller 29 via a known communication interface such as a LAN (Local Area Network). This control device 30 comprehensively controls the operations of each part of the three-dimensional coordinate measuring machine 10. The control device 30 includes an arithmetic circuit composed of various processors and memories. The various processors include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), and programmable logic devices [for example, SPLD (Simple Programmable Logic Devices), CPLD (Complex Programmable Logic Device), and FPGA (Field Programmable Gate Arrays)]. Note that the various functions of the control device 30 may be realized by one processor or by a plurality of processors of the same type or different types.

[0033] [Configuration of Control Device] Figure 3 is a functional block diagram of the control device 30. As shown in Figure 3, in addition to the aforementioned controller 29, an operation unit 32, a display unit 34, and a storage unit 36 are connected to the control device 30.

[0034] The operation unit 32 receives input operations by the user. Note that the operation unit 32 may be integrated with the controller 29. The display unit 34 displays various operation screens, measurement screens, menu screens, and warning information 70 (see Figure 20) described later.

[0035] In addition to a control program (not shown) for operating the control device 30, a measurement program 38, measurement data 39, a calibration program 40, calibration data 42, a machine learning model 44, and design value information 46 are stored in the storage unit 36.

[0036] The measurement program 38 has a plurality of measurement plans that define a plurality of measurement items for the workpiece W. The measurement plan defines the movement path of the probe 26 during the shape measurement of the workpiece W [for example, the measurement order of each measurement item, the coordinate values of all measurement points for each measurement item, and the coordinate values of the intermediate points that are the movement path points of the probe 26, etc.]. The measurement data 39 is the shape measurement result for each measurement item of each individual measurement plan.

[0037] The calibration program 40 is a program that defines the processing necessary for obtaining (generating) the calibration data 42 of the probe 26, including the contact method (position and orientation control) of the probe 26 with a known calibration jig such as a calibration sphere and the calculation method of the calibration data 42. In the three-dimensional coordinate measuring machine 10 of the present embodiment, the acquisition of the calibration data 42 is executed periodically. The calibration data 42 is obtained by the calibration data acquisition unit 54 executing the calibration program 40, which will be described in detail later.

[0038] The machine learning model 44 is used, as will be described in detail later, for predicting the measurement data 39 for each measurement item of each measurement plan of the workpiece W from the calibration data 42 by the automatic trend management unit 62 described later. The design value information 46 is data that registers the design values and tolerances (tolerance ranges) for each measurement item of each measurement plan of the workpiece W (see FIG. 17), and is used, as will be described in detail later, for determining the quality of the calibration data 42 by the automatic trend management unit 62.

[0039] The control device 30 functions as a drive control unit 50, a probing data acquisition unit 52, a calibration data acquisition unit 54, a measurement data calculation unit 56, a first data collection unit 58, a machine learning model generation unit 60, an automatic trend management unit 62, a second data collection unit 64, a notification control unit 66, and a repetition control unit 68 by executing a control program (not shown) stored in the storage unit 36.

[0040] The drive control unit 50 operates in the automatic measurement mode of the three-dimensional coordinate measuring machine 10. Based on the measurement program 38 in the storage unit 36, the drive control unit 50 drives the rotary drive unit 28 and the XYZ drive unit 27 via the controller 29, thereby executing a probing process for each measurement plan of the workpiece W. The probing process is a process of sequentially bringing the probe 26 into contact with all the measurement points of the workpiece W corresponding to the measurement item for each measurement item defined in the measurement plan.

[0041] FIG. 4 is an explanatory diagram for explaining the acquisition of probing data 47 of measurement points (probing points) by the probing data acquisition unit 52. During the probing process for each measurement plan of the workpiece W (in the automatic measurement mode and the manual measurement mode), each time the probe 26 contacts a measurement point of the workpiece W, the probing data acquisition unit 52 acquires the probing data 47 including the XYZ coordinate values of the measurement point (probing point) where the probe 26 has contacted and outputs it to the measurement data calculation unit 56.

[0042] For example, as shown in FIG. 4, each time the probe 26 contacts a measurement point of the workpiece W, the probing data acquisition unit 52 acquires the detection results of the above-mentioned reading head 27A and the rotation angle detection unit 24A from the controller 29. Then, the probing data acquisition unit 52 detects the XYZ coordinate values (refer to symbol P1) of the probe head 24 based on the detection result of the reading head 27A. Next, the probing data acquisition unit 52 detects the XYZ coordinate values of the center of the contact 26a (refer to symbol P2) based on the XYZ coordinate values of the probe head 24, the detection result of the rotation angle detection unit 24A, and the known length and radius of the probe 26. After that, the probing data acquisition unit 52 detects the XYZ coordinate values (refer to symbol P3) of the measurement point (probing point) of the workpiece W.

[0043] Therefore, the probing data acquisition unit 52 offsets the XYZ coordinate values of the probe head 24 detected by the reading head 27A by offset values F1 and F2 (including the offset direction) determined based on the detection result of the rotation angle detection unit 24A and the length and radius of the probe 26, thereby detecting the XYZ coordinate values of the measurement point (probing point) of the workpiece W. Then, in addition to the XYZ coordinate values of the measurement point, the probing data acquisition unit 52 outputs probing data 47 (see FIG. 9) including, for example, the direction of the probe 26, the temperature of the XYZ scale described above, and the temperature of the workpiece W to the measurement data calculation unit 56.

[0044] Returning to FIG. 3, the calibration data acquisition unit 54 drives the rotation drive unit 28 and the XYZ drive unit 27 via the controller 29 in accordance with the automatic drive or manual operation based on the calibration program 40 in the storage unit 36, and performs probing (calibration work) on a calibration jig such as a calibration sphere by the probe 26. Then, the calibration data acquisition unit 54 acquires (generates) calibration data 42 of the probe 26 by a known method based on the probing data 47 acquired by the probing data acquisition unit 52 in the calibration work. Since the method for acquiring the calibration data 42 is a known technique, a specific description thereof is omitted here.

[0045] The calibration data 42 is a calibration value for calibrating the XYZ coordinate values (or offset values F1 and F2) of each measurement point (probing point) included in the probing data 47. This calibration data 42 includes, for example, the probe diameter which is the diameter of the contact 26a, the offset amounts in the XYZ directions of the center position of the probe (stylus) viewed from the reference probe, and the variation (standard deviation) in the measurement of the calibration sphere during the calibration of the probe 26 (see FIG. 9). In addition, calibration ID (identification), which is its unique identification information, is assigned to the calibration data 42 stored in the storage unit 36 (see FIG. 9).

[0046] The acquisition of the calibration data 42 by the calibration data acquisition unit 54 is executed periodically (it can also be non-periodically). When the machine learning model 44 described later is not stored in the storage unit 36, the calibration data 42 acquired by the calibration data acquisition unit 54 is output to the storage unit 36 (it can also be the first data collection unit 58 described later). When the machine learning model 44 is already stored in the storage unit 36, it is output to the automatic trend management unit 62 described later.

[0047] The measurement data calculation unit 56 calculates the measurement data 39 of the work W corresponding to all the measurement items defined in the measurement plan for each measurement plan of the work W. For example, the measurement data calculation unit 56 calculates the measurement data 39 by a known method based on the probing data 47 of all the measurement points (probing points) acquired by the probing data acquisition unit 52 and the calibration data 42 stored in the storage unit 36 for each measurement item of each individual measurement plan.

[0048] The first data collection unit 58 and the machine learning model generation unit 60 operate when generating the machine learning model 44. The machine learning model 44 is used in a prediction process for predicting the measurement data 39 for each measurement item of each individual measurement plan from the calibration data 42 newly acquired by the calibration data acquisition unit 54. Since the measurement data 39 is calculated based on the probing data 47 (XYZ coordinate values) of the measurement points (probing points) calibrated by the calibration data 42, there may be a correlation between the calibration data 42 and the measurement data 39.

[0049] FIG. 5 is a graph showing the relationship between the calibration data 42 (probe diameter) and the measurement data 39 (radius). As shown in FIG. 5, a positive correlation is confirmed between the calibration data 42 (probe diameter) and the measurement data 39 (radius). Therefore, the measurement data 39 (radius) increases or decreases in proportion to the increase or decrease of the calibration data 42 (probe diameter). The fact that there is a correlation between the calibration data 42 and the measurement data 39 in this way suggests the possibility of predicting the measurement data 39 from the calibration data 42.

[0050] Also, if the calibration data 42 is defective for some reason, the measurement data 39 for each measurement item calculated by the measurement data calculation unit 56 based on this calibration data 42 is likely not to be the correct value. And in that case, the difference between the measurement data 39 for each measurement item and the design value is outside the tolerance range, that is, the measurement data 39 is likely to be outside the tolerance range. Therefore, when the measurement data 39 for each measurement item of the workpiece W predicted from the calibration data 42 is outside the tolerance range, it suggests that the calibration data 42 may be defective.

[0051] Therefore, in this embodiment, the measurement data 39 for each measurement item of each measurement plan of the workpiece W is predicted from the calibration data 42 newly acquired by the calibration data acquisition unit 54 using the machine learning model 44, and the quality of this calibration data 42 is determined by comparing the predicted measurement data 39 with the design value information 46.

[0052] <Generation of Machine Learning Model> FIG. 6 is an explanatory diagram for explaining the generation of the machine learning model 44. In FIG. 6, in order to prevent complication of the drawing, illustration of some components of the three-dimensional coordinate measuring machine 10 (such as the XYZ drive unit 27, the rotation drive unit 28, the controller 29, the operation unit 32, the display unit 34, and the storage unit 36, etc.) is omitted.

[0053] As shown in FIG. 6, the first data collection unit 58 controls the "calibration data acquisition unit 54" and the "drive control unit 50, the probing data acquisition unit 52, and the measurement data calculation unit 56" to collect a plurality of corresponding data 48 for each measurement item of the workpiece W for each measurement plan. The corresponding data 48 corresponds to the first corresponding data of the present invention and is teacher data (training data) used for the generation of the machine learning model 44. Specifically, the corresponding data 48 is data in which the measurement data 39 of the workpiece W for each measurement item defined in the measurement plan is associated with the calibration data 42 used by the measurement data calculation unit 56 for the calculation of each measurement data 39 (see FIG. 10 described later).

[0054] FIG. 7 is a flowchart showing the flow of the collection process of the corresponding data 48 for each measurement item of the measurement plan A by the first data collection unit 58. FIG. 8 is a flowchart showing the flow of the probe calibration (step S1) in FIG. 7.

[0055] As shown in FIGS. 7 and 8 and the aforementioned FIG. 6, the first data collection unit 58 controls the calibration data acquisition unit 54 to execute the acquisition of the calibration data 42 (step S1). First, as shown in FIG. 8, the calibration data acquisition unit 54 drives the rotation drive unit 28 and the XYZ drive unit 27 via the controller 29 based on the calibration program 40 in the storage unit 36, either automatically or manually, to perform a calibration operation of bringing the probe 26 into contact with a calibration jig such as a calibration sphere a plurality of times from a predetermined direction in a predetermined posture (step S1A). As a result, each time the probe 26 is brought into contact with the calibration jig, the probing data acquisition unit 52 acquires the probing data 47 of the probing points of the probe 26 with respect to the calibration jig and outputs it to the calibration data acquisition unit 54.

[0056] Then, the calibration data acquisition unit 54 calculates the calibration data 42 by a known method based on the plurality of probing data 47 input from the probing data acquisition unit 52 (step S1B). Next, after attaching a calibration ID to the calibration data 42 (step S1C), the calibration data acquisition unit 54 stores this calibration data 42 in the storage unit 36 (step S1D). Thus, the acquisition of the calibration data 42 by the calibration data acquisition unit 54 is completed.

[0057] Returning to FIGS. 6 and 7, the first data collection unit 58 controls the drive control unit 50, the probing data acquisition unit 52, and the measurement data calculation unit 56 to start the measurement of the measurement data 39 of the work W for each measurement plan (measurement plan A, measurement plan B, measurement plan C,...). First, the measurement of the measurement data 39 of the work W for each measurement item defined in the measurement plan A is started (step S2).

[0058] First, the drive control unit 50 drives the rotation drive unit 28 and the XYZ drive unit 27 via the controller 29 based on the measurement program 38 in the storage unit 36, thereby executing a probing process of bringing the probe 26 into contact with all the measurement points of the workpiece W corresponding to the measurement item A of the measurement plan A (step S3). As a result, the probing data acquisition unit 52 acquires the probing data 47 for each measurement point (probing point) of the measurement item A (step S4).

[0059] FIG. 9 is an explanatory diagram for explaining the association of the probing data 47 for each measurement item of the measurement plan A with the calibration data 42 by the calibration ID. As shown in FIG. 9 and the aforementioned FIGS. 6 and 7, the probing data acquisition unit 52 assigns the calibration ID of the calibration data 42 stored in the storage unit 36 to the probing data 47 corresponding to the measurement item A (step S5). Thereby, the probing data 47 corresponding to the measurement item A and the calibration data 42 used for calculating the measurement data 39 of the measurement item A by the measurement data calculation unit 56 are associated with each other.

[0060] Similarly hereinafter, for each of the remaining measurement items (measurement item B, measurement item C,...) of the measurement plan A, the probing process, the acquisition process of the probing data 47, and the assignment process of the calibration ID are repeatedly executed (step S6).

[0061] When the probing data 47 for all the measurement items of the measurement plan A are acquired, the measurement data calculation unit 56 calculates the measurement data 39 by a known method based on all the probing data 47 of all the measurement points acquired by the probing data acquisition unit 52 and the calibration data 42 stored in the storage unit 36 for each measurement item of the measurement plan A (step S7). At this time, based on the calibration ID assigned to the probing data 47, for each measurement item, the measurement data 39 and the calibration data 42 used for calculating the measurement data 39 are associated with each other. Note that a plurality of pieces of measurement data 39 are acquired for each measurement item of the measurement plan A (see FIG. 10).

[0062] FIG. 10 is a diagram showing an example of the corresponding data 48 for each measurement plan used for the machine learning of the machine learning model 44. As shown in FIG. 10 and the aforementioned FIGS. 6 and 7, the first data collection unit 58 generates corresponding data 48 in which the measurement data 39 for each measurement item of the measurement plan A is associated with the calibration data 42 used for the calculation of the measurement data 39 (step S8). Thereby, the collection of the corresponding data 48 of the measurement plan A by the first data collection unit 58 is completed.

[0063] Similarly, by collecting the corresponding data 48 of other measurement plans (measurement plan B, measurement plan C, etc.) by the first data collection unit 58, a plurality of corresponding data 48 for each measurement plan are collected.

[0064] FIG. 11 is an explanatory diagram for explaining the generation of the machine learning model 44 for each measurement plan by the machine learning model generation unit 60. As shown in FIG. 11 and the aforementioned FIG. 6, the machine learning model generation unit 60 uses the corresponding data 48 for each measurement plan collected by the first data collection unit 58 (explanatory variable: calibration data 42, target variable: measurement data 39) as teacher data, and for each measurement plan, generates a machine learning model 44 that predicts the measurement data 39 for each measurement item from the calibration data 42. As a method for generating each machine learning model 44, machine learning (supervised learning) using a known multiple regression model (multiple regression equation, multiple regression analysis), or a machine learning algorithm such as a known convolutional neural network (CNN) is adopted. Also, unsupervised learning, reinforcement learning, transfer learning, etc. may be performed as machine learning. The machine learning model 44 generated by the machine learning model generation unit 60 is stored in the storage unit 36.

[0065] FIG. 12 is an explanatory diagram for explaining the prediction of measurement data 39 for each measurement plan by the machine learning model 44. As shown in FIG. 12, for each measurement plan, the machine learning model 44 takes calibration data 42 as an input (INPUT) and outputs measurement data 39 for each measurement item defined in the measurement plan as an output (OUTPUT). Thereby, based on the calibration data 42, the measurement data 39 for each measurement item of each individual measurement plan can be predicted.

[0066] FIG. 13 is a diagram showing a part of the actual correspondence data 48 used for generating the machine learning model 44. FIG. 14 is a table showing the generalization performance (coefficient of determination R 2 : accuracy evaluation index) for each measurement item of the machine learning model 44 generated using the correspondence data 48 shown in FIG. 13. Here, in FIG. 14, the coefficient of determination for each measurement item when using the teacher data and the coefficient of determination for each measurement item when using unknown test data are shown. Note that since the specific calculation method of the coefficient of determination is a known technique, a specific explanation is omitted here.

[0067] As shown in FIGS. 13 and 14, the numerical values of the coefficient of determination for each measurement item when using unknown test data are approximately the same as the numerical values of the coefficient of determination for each measurement item when using teacher data. Therefore, it can be seen that even when using calibration data 42 that has not been used for learning the machine learning model 44, the measurement data 39 can be predicted with excellent accuracy.

[0068] <Good or Bad Judgment of Calibration Data and Re - learning of Machine Learning Model> FIG. 15 is an explanatory diagram for explaining the good or bad judgment of calibration data 42 and the re - learning of the machine learning model 44. As shown in FIG. 15, when the calibration data acquisition unit 54 newly acquires calibration data 42 and outputs this calibration data 42 to the automatic trend management unit 62 in a state where the machine learning model 44 is stored in the storage unit 36, the automatic trend management unit 62 operates, and at the same time, the second data collection unit 64, the notification control unit 66, and the repetition control unit 68 become operable states. Note that in this case, no calibration ID is assigned to the calibration data 42 acquired by the calibration data acquisition unit 54.

[0069] FIG. 16 is an explanatory diagram for explaining the function of the automatic trend management unit 62 in FIG. 15. As shown in FIG. 16 and the aforementioned FIG. 15, the automatic trend management unit 62 functions as the prediction unit and determination unit of the present invention. When new calibration data 42 is input from the calibration data acquisition unit 54, the automatic trend management unit 62 inputs this calibration data 42 into the machine learning models 44 for each measurement plan as shown in FIG. 12 described above, respectively, and predicts the measurement data 39 for each measurement item of each measurement plan. In this case, the automatic trend management unit 62 functions as the prediction unit of the present invention.

[0070] Next, the automatic trend management unit 62 compares the measurement data 39 for each measurement item of each measurement plan with the design value information 46, and makes an individual determination to determine the quality of the measurement data 39 for each measurement item.

[0071] FIG. 17 is an explanatory diagram for explaining an example of the individual determination by the automatic trend management unit 62. In FIG. 17, only the "probe diameter" is exemplified as the calibration data 42 and the "radius of the circle" is exemplified as the measurement item of the measurement data 39 in order to prevent complication of the explanation.

[0072] As shown in FIG. 17, when the difference between the measurement data 39 and the design value is outside the tolerance range, that is, when the measurement data 39 is outside the tolerance range, the automatic trend management unit 62 determines the measurement data 39 as defective. Conversely, when the difference between the measurement data 39 and the design value is within the tolerance range, that is, when the measurement data 39 is within the tolerance range, the automatic trend management unit 62 determines the measurement data 39 as good. Similarly, the automatic trend management unit 62 makes an individual determination to determine the quality of the measurement data 39 for each measurement item of each measurement plan.

[0073] Returning to FIG. 16, when the automatic trend management unit 62 determines that the measurement data 39 of all measurement items of all measurement plans is good (when there is no measurement data 39 that is determined to be bad), the calibration data 42 is determined to be good. Conversely, when there is one or more pieces of measurement data 39 that are determined to be bad, the automatic trend management unit 62 determines that the calibration data 42 is bad. In this case, the automatic trend management unit 62 functions as the determination unit of the present invention.

[0074] (Calibration data: Good determination) FIG. 18 is an explanatory diagram for explaining the relearning of the machine learning model 44 when the automatic trend management unit 62 determines that the calibration data 42 is good. As shown in FIG. 18 and the aforementioned FIG. 15, when the automatic trend management unit 62 determines that the calibration data 42 is good, it assigns a calibration ID to this calibration data 42 and then stores it in the storage unit 36. Thereby, the measurement data calculation unit 56 calculates the measurement data 39 for each measurement item of each measurement plan based on the calibration data 42 newly stored in the storage unit 36.

[0075] After the new calibration data 42 is stored in the storage unit 36 by the automatic trend management unit 62, when the second data collection unit 64 acquires the measurement data 39 for each measurement item according to an arbitrary measurement plan with the three-dimensional coordinate measuring machine 10, the second data collection unit 64 generates correspondence data 48 (corresponding to the second correspondence data of the present invention) in which the new calibration data 42 and the measurement data 39 for each measurement item are associated. Next, the second data collection unit 64 outputs this correspondence data 48 to the machine learning model generation unit 60.

[0076] Based on the correspondence data 48 input from the second data collection unit 64, the machine learning model generation unit 60 performs relearning of the machine learning model 44 corresponding to the measurement plan of this correspondence data 48. Since the method of relearning the machine learning model 44 is a known technique, a specific description is omitted here. Hereinafter, each time the measurement data 39 for each measurement item is acquired according to an arbitrary measurement plan with the three-dimensional coordinate measuring machine 10, relearning of the machine learning model 44 corresponding to this measurement plan is executed.

[0077] (Calibration data: Bad determination) FIG. 19 is an explanatory diagram for explaining the functions of the notification control unit 66 and the repetition control unit 68 when the automatic trend management unit 62 determines that the calibration data 42 is defective. As shown in FIG. 19 and the aforementioned FIG. 15, when the automatic trend management unit 62 determines that the calibration data 42 is defective, the notification control unit 66 causes the display unit 34 to display warning information 70 (see FIG. 20) indicating that fact. Instead of causing the warning information 70 to be displayed on the display unit 34, the warning information 70 may be notified by a known method such as voice output from a speaker (not shown).

[0078] FIG. 20 is an explanatory diagram showing an example of the warning information 70 displayed on the display unit 34. As shown in FIG. 20, the warning information 70 includes, for example, a warning message 70a, measurement item information 70b, and a selection icon 70c.

[0079] The warning message 70a includes, for example, a message indicating that the calibration data 42 is defective and a message asking the user whether to adopt this calibration data 42. The measurement item information 70b displays the measurement item corresponding to the measurement data 39 for which the automatic trend management unit 62 has determined a defect.

[0080] The selection icon 70c is a selection operation unit for the user to select whether to adopt the calibration data 42 determined to be defective by the automatic trend management unit 62. When the user selects to adopt the calibration data 42 with the selection icon 70c, the automatic trend management unit 62 assigns a calibration ID to this calibration data 42 and then stores it in the storage unit 36. This calibration data 42 is used for re-learning of the machine learning model 44.

[0081] Conversely, when the user selects not to adopt the calibration data 42 with the selection icon 70c, the repetition control unit 68 operates.

[0082] Returning to FIGS. 19 and 15, the repetition control unit 68 activates the calibration data acquisition unit 54 and the automatic trend management unit 62 again to execute re-acquisition of the calibration data 42 by the calibration data acquisition unit 54 and determination of the quality of the new calibration data 42 by the automatic trend management unit 62. In this embodiment, the repetition control unit 68 is activated when the user selects not to adopt the calibration data 42 using the selection icon 70c. However, the repetition control unit 68 may be immediately activated when the automatic trend management unit 62 determines that the calibration data 42 is defective.

[0083] [Operation of the three-dimensional coordinate measuring machine] FIG. 21 is a flowchart showing the flow of generation of the machine learning model 44 and management of the calibration data 42 in the three-dimensional coordinate measuring machine 10 having the above configuration according to the calibration data 42 management method of the present invention. When the machine learning model 44 is already stored in the storage unit 36 (including the case where it is stored by the manufacturer), the process starts from step S12. Also, steps S10 and S11 may be performed externally (for example, by the manufacturer).

[0084] As shown in FIG. 21, the first data collection unit 58 executes the processes from step S1 to step S8 described in FIGS. 7 and 8 above, and collects a plurality of corresponding data 48 for each measurement item of the workpiece W for each measurement plan as shown in FIG. 10 above (step S10, corresponding to the first data collection step of the present invention).

[0085] Next, the machine learning model generation unit 60 generates a machine learning model 44 for each measurement plan as shown in FIG. 11 above, using the corresponding data 48 for each measurement plan collected by the first data collection unit 58 as teacher data (step S11, corresponding to the machine learning model generation step of the present invention). This machine learning model 44 is stored in the storage unit 36.

[0086] In the three-dimensional coordinate measuring machine 10, the probe 26 is calibrated regularly. In this case, the calibration data acquisition unit 54 acquires new calibration data 42 of the probe 26 as described in steps S1A to S1B of FIG. 8 described above, and outputs this calibration data 42 to the automatic trend management unit 62 (step S12, corresponding to the calibration data acquisition step of the present invention).

[0087] Next, based on the calibration data 42 newly input from the calibration data acquisition unit 54, the automatic trend management unit 62 predicts the measurement data 39 for each measurement item of each measurement plan using the machine learning model 44 for each measurement plan as shown in FIG. 12 described above (step S13, corresponding to the prediction step of the present invention). Then, the automatic trend management unit 62 compares the measurement data 39 for each measurement item of each measurement plan with the design value information 46, and performs an individual determination to determine the quality of the measurement data 39 for each measurement item. Based on the results of these individual determinations, when the automatic trend management unit 62 determines that the measurement data 39 for all measurement items of all measurement plans is good, it determines that the calibration data 42 is good, and conversely, when there is one or more pieces of measurement data 39 that are determined to be bad, it determines that the calibration data 42 is bad (step S15, corresponding to the determination step of the present invention).

[0088] When the automatic trend management unit 62 determines that the calibration data 42 is good, it assigns a calibration ID to this calibration data 42 and then stores it in the storage unit 36 (YES in step S15, step S16).

[0089] Next, the shape measurement of the workpiece W by the three-dimensional coordinate measuring machine 10 is started according to an arbitrary measurement plan. In this case, the second data collection unit 64 generates corresponding data 48 in which new calibration data 42 and measurement data 39 for each measurement item corresponding to an arbitrary measurement plan are associated by executing the processes after step S2 described in FIG. 7 described above (step S17, corresponding to the measurement step and the second data collection step of the present invention).

[0090] Then, the second data collection unit 64 outputs the corresponding data 48 to the machine learning model generation unit 60. Thereby, the machine learning model generation unit 60 executes re-learning of the machine learning model 44 based on the corresponding data 48 input from the second data collection unit 64 (step S18, corresponding to the re-learning step of the present invention). By performing re-learning of the machine learning model 44, the prediction accuracy of the measurement data 39 by the machine learning model 44 (the determination accuracy of the quality of the calibration data 42) can be improved.

[0091] On the other hand, when the automatic trend management unit 62 determines that the calibration data 42 is defective (NO in step S15), the notification control unit 66 operates to display warning information 70 on the display unit 34 as shown in FIG. 20 described above (step S19, corresponding to the notification step of the present invention). Thereby, the user can recognize that the calibration data 42 is defective. When the user operates the operation unit 32 and inputs a selection operation for selecting the adoption of the calibration data 42 to the selection icon 70c (YES in step S20), the processes from step S16 to step S18 described above are repeatedly executed.

[0092] Conversely, when the user operates the operation unit 32 and inputs a selection operation for selecting non-adoption of the calibration data 42 to the selection icon 70c (NO in step S20), the loop control unit 68 operates. Thereby, the processes from step S12 to step S14 described above, that is, the re-acquisition of the calibration data 42 by the calibration data acquisition unit 54 and the determination of the quality of the new calibration data 42 by the automatic trend management unit 62, are executed. Hereinafter, the re-acquisition and quality determination of the calibration data 42 are repeatedly executed until "YES" is obtained in step S15 or step S20. As a result, good calibration data 42 can be obtained.

[0093] Then, every time the calibration of the probe 26 is periodically executed by the three-dimensional coordinate measuring machine 10 (step S2), the processes after step S13 described above are repeatedly executed.

[0094] As described above, in the three-dimensional coordinate measuring machine 10 of the present embodiment, by generating and using a machine learning model 44 that takes calibration data 42 as input and outputs measurement data 39 for each measurement item of each measurement plan, it is possible to easily determine the quality of the calibration data 42 without setting a threshold value or the like of the calibration data 42 as in the prior art. In addition, the trend management of the calibration data 42 can be automated. Furthermore, since it is possible to prevent the shape measurement of the workpiece W from being performed using defective calibration data 42, the execution of unnecessary measurements is prevented.

[0095] [Others] In the above embodiment, a contact probe has been described as an example of the probe 26 of the three-dimensional coordinate measuring machine 10, but a known non-contact (optical) probe may be used as the probe 26 (see Japanese Unexamined Patent Application Publication No. 2020-098180). In this case, the probing point (measurement point) is a point on the surface of the workpiece W irradiated with measurement light from the non-contact probe.

[0096] In the above embodiment, the machine learning model 44 is generated by the three-dimensional coordinate measuring machine 10. However, the generation of the machine learning model 44 may be performed by a manufacturer or the like, and the three-dimensional coordinate measuring machine 10 may only perform the determination of the quality of the calibration data 42 using the machine learning model 44.

[0097] In the three-dimensional coordinate measuring machine 10 of the above embodiment, a plurality of measurement plans for the workpiece W are prepared, and a plurality of measurement items are set for each measurement plan. However, the number of measurement plans and measurement items may be one. Even in this case, similar to the above embodiment, it is possible to generate the machine learning model 44, predict the calibration data 42, and determine its quality.

[0098] In the above embodiment, the three-dimensional coordinate measuring machine 10 is cited as an example of the shape measuring machine of the present invention. However, the present invention can also be applied to a shape measuring machine that measures the shape (including surface roughness and roundness) of various workpieces (measurement objects) using the probe 26.

Explanation of Reference Numerals

[0099] 10…Coordinate measuring machine, 12…Base, 14…Table, 16L…Left Y carriage, 16R…Right Y carriage, 18…X guide, 19…Gantry frame, 20…X carriage, 22…Z carriage, 24…Probe head, 24A…Rotation angle detector, 26…Probe, 26a…Contact, 27…XYZ drive unit, 27A…Reading head, 28…Rotation drive unit, 29…Controller, 30…Control device, 32…Operation unit, 34…Display unit, 36…Memory unit, 38…Measurement program, 39…Measurement data, 40…Calibration program, 42…Calibration data, 44…Machine learning model, 46…Design value information, 47…Probing data, 48…Corresponding data, 50…Drive control unit, 52…Probing data acquisition unit, 54…Calibration data acquisition unit, 56…Measurement data calculation unit, 58…First data collection unit, 60…Machine learning model generation unit, 62…Automatic trend management unit, 64…Second data collection unit, 66…Notification control unit, 68…Repetition control unit, 70…Warning information, 70a…Warning message, 70b…Measurement item information, 70c…Selection icon, F1…Offset value, F2…Offset value

Claims

1. A method for managing calibration data in a shape measuring machine that includes a probe for coordinate measurement, and based on coordinate values of probing points obtained by measuring one or more measurement items of a workpiece with the probe according to a predetermined measurement plan and calibration data of the probe, acquires measurement data of the workpiece corresponding to the measurement items of the measurement plan, wherein: a calibration data acquisition step of acquiring new calibration data of the probe; a prediction step of predicting the measurement data using a machine learning model that takes the calibration data as an input and outputs the measurement data based on the calibration data acquired in the calibration data acquisition step; a determination step of determining the quality of the calibration data acquired in the calibration data acquisition step based on the measurement data predicted in the prediction step and design value information of the workpiece; A method for managing calibration data having the above.

2. a first data collection step of collecting a plurality of first correspondence data in which a plurality of measurement data are acquired by the shape measuring machine according to the measurement plan and the measurement data and the calibration data used by the shape measuring machine for acquiring the measurement data are associated with each other; a machine learning model generation step of generating the machine learning model based on the plurality of first correspondence data collected in the first data collection step; having The method for managing calibration data according to claim 1, wherein in the prediction step, the measurement data is predicted using the machine learning model generated in the machine learning model generation step.

3. There are a plurality of the measurement plans, in the first data collection step, a plurality of the first correspondence data are collected for each measurement plan, in the machine learning model generation step, the machine learning model is generated for each measurement plan, in the prediction step, the calibration data is input to the machine learning model corresponding to the measurement plan for each measurement plan to predict the measurement data for each measurement plan, in the determination step, the quality of the calibration data is determined based on the measurement data for each measurement plan predicted in the prediction step and the design value information. The method for managing calibration data according to claim 2.

4. a measurement step of acquiring the measurement data with the shape measuring machine according to the measurement plan when it is determined in the determination step that the calibration data is good; A second data collection step of collecting second correspondence data associating the measurement data measured in the measurement step and the calibration data determined to be good in the determination step; A re-learning step of performing re-learning of the machine learning model corresponding to the measurement plan based on the second correspondence data collected in the second data collection step; The calibration data management method according to claim 3, comprising:

5. A measurement step of obtaining the measurement data with the shape measuring machine according to the measurement plan when the calibration data is determined to be good in the determination step; A second data collection step of collecting second correspondence data associating the measurement data measured in the measurement step and the calibration data determined to be good in the determination step; A re-learning step of performing re-learning of the machine learning model based on the second correspondence data collected in the second data collection step; The calibration data management method according to claim 1, comprising:

6. The calibration data management method according to any one of claims 1 to 5, wherein when the calibration data is determined to be no in the determination step, the calibration data acquisition step, the prediction step, and the determination step are repeatedly executed.

7. The calibration data management method according to any one of claims 1 to 5, comprising a notification step of notifying warning information when the calibration data is determined to be no in the determination step.

8. In a shape measuring machine that includes a probe for coordinate measurement and acquires measurement data of the workpiece corresponding to the measurement items of the measurement plan based on coordinate values of probing points obtained by measuring one or more measurement items of the workpiece with the probe according to a predetermined measurement plan and calibration data of the probe, A calibration data acquisition unit that acquires new calibration data of the probe; A prediction unit that predicts the measurement data using a machine learning model that takes the calibration data as an input and outputs the measurement data based on the calibration data acquired by the calibration data acquisition unit; A determination unit that determines the quality of the calibration data acquired by the calibration data acquisition unit based on the measurement data predicted by the prediction unit and design value information of the workpiece; A shape measuring machine comprising:

9. A first data collection unit that acquires a plurality of the measurement data according to the measurement plan and collects a plurality of first correspondence data in which the measurement data is associated with the calibration data used by the shape measuring machine to acquire the measurement data; A machine learning model generation unit that generates the machine learning model based on the plurality of the first correspondence data collected by the first data collection unit; Comprising; The shape measuring machine according to claim 8, wherein the prediction unit predicts the measurement data using the machine learning model generated by the machine learning model generation unit.

Images