Roundness measurement system, roundness measurement method, and program

Abstract

Even with measurement data affected by noise, it can measure the roundness of the inner surface of a cylindrical object with high accuracy. [Solution] The roundness measurement system comprises a measuring device for measuring the inner surface of a cylindrical object to be measured; a storage unit that stores a learning model generated by machine learning using known observed values ​​as learning data, which estimates the inclination direction and inclination angle of the object to be measured from the measurement data obtained by the measuring device; and a control unit that calculates the roundness of the inner surface of the cylindrical object to be measured from the measurement data obtained by measuring the inner surface of the object to be measured. The control unit estimates the inclination direction and inclination angle of the object to be measured by inputting the measurement data into the learning model, corrects the measurement data using the estimated inclination direction and inclination angle as correction values, and calculates the roundness of the inner surface of the object to be measured based on the corrected measurement data.

Description

Technical Field

[0001] The present invention relates to a roundness measurement system, a roundness measurement method, and a program.

Background Art

[0002] Among various products such as automobiles and electric products, there are some that use cylindrical members or parts provided with cylindrical holes. In order to confirm the manufacturing accuracy of these members and parts, it may be necessary to measure the roundness of the inner peripheral surface of a cylindrical measurement object. For example, parts such as a cylinder of an automobile engine and a brake master cylinder need to be manufactured with high precision, so the roundness of the inner peripheral surface of these parts is required to be above a certain value. And various measurement methods and measurement devices have been proposed for measuring the inner peripheral surface of such a cylindrical measurement object.

[0003] [[ID=1⑥For example, Patent Document 1 discloses a roundness measurement method in which light is irradiated onto the inner peripheral surface in the shape of a cylindrical surface of a measured work, the inner peripheral surface irradiated with the light is imaged to obtain image data, and the roundness of the inner peripheral surface is calculated and obtained based on the obtained image data.

[0004] Further, Patent Document 2 discloses a roundness measuring machine in which a work to be measured is placed on a table that is rotationally driven, a stylus is brought into contact with an arbitrary height position on the peripheral surface of the work, and the displacement of the stylus is detected by a detector to obtain the contour shape of the work surface as an output signal.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] Conventional roundness measurement systems and machines, as described above, require considerable effort to correct for misalignment or tilt of the center position of cylindrical objects being measured. In particular, when attempting to measure roundness precisely, even a slight tilt of the object being measured can affect the measurement accuracy. Therefore, in some cases, the misalignment of the center position, the direction of tilt, and the angle of tilt are calculated from the measurement data of the inner surface of the object being measured, and the measurement data is corrected using the obtained misalignment amount, tilt direction, and angle of tilt as correction values.

[0007] However, even when obtaining measurement data by optically measuring the inner surface of the object being measured, the resulting measurement data will be noisy due to the effects of vibration, temperature changes, ambient light, surface roughness, etc. Even if the measurement data is corrected by calculating the tilt direction and tilt angle using measurement data containing such noise, if the calculated tilt direction and tilt angle are incorrect in the first place, the roundness obtained based on the corrected measurement data will also contain errors. In particular, there are limits to how much correction can be achieved by calculating the minute tilt of the object being measured.

[0008] The object of the present invention is to provide a roundness measurement system, a roundness measurement method, and a program that can measure the roundness of the inner surface of a cylindrical object with high accuracy, even when the measurement data is affected by noise. [Means for solving the problem]

[0009] The roundness measurement system of the present invention comprises a measuring device for measuring the inner surface of a cylindrical object to be measured, and a storage unit that stores a learning model generated by machine learning using known observed values ​​as learning data, which estimates the inclination direction and inclination angle of the object to be measured from the measurement data obtained by the measuring device. The system comprises a control unit that calculates the roundness of the inner surface of a cylindrical object from measurement data obtained by measuring the inner surface of the object, The control unit, By inputting the measurement data into the learning model, the tilt direction and tilt angle of the object being measured are estimated. The estimated inclination direction and inclination angle are used as correction values ​​to correct the measurement data. The roundness of the inner surface of the object to be measured is calculated based on the corrected measurement data.

[0010] According to the roundness measurement system of the present invention, the inclination direction and inclination angle of an object to be measured can be estimated from measurement data using a pre-generated learning model, which is created by machine learning using observed values ​​with known inclination direction and inclination angle as training data. This allows for the estimation of the inclination direction and inclination angle of an object to be measured without being affected by noise contained in the measurement data. Therefore, according to the present invention, even with measurement data affected by noise, the roundness of the inner surface of a cylindrical object to be measured can be measured with high accuracy.

[0011] Furthermore, another roundness measurement system of the present invention includes a measuring device for measuring the inner surface of a cylindrical object to be measured, and a storage unit that stores multiple learning models which are generated for each combination of the range The system comprises a control unit that calculates the roundness of the inner surface of a cylindrical object from measurement data obtained by measuring the inner surface of the object, The control unit, Based on the aforementioned measurement data, the inclination direction and inclination angle of the object to be measured are calculated. From among the multiple learning models mentioned above, select one learning model that corresponds to the calculated slope direction and slope angle. By inputting the measurement data into a selected learning model, the inclination direction and inclination angle of the object being measured are estimated. The estimated inclination direction and inclination angle are used as correction values ​​to correct the measurement data. The roundness of the inner surface of the object to be measured is calculated based on the corrected measurement data.

[0012] According to the roundness measurement system of the present invention, a plurality of learning models for each combination of the range of the tilt direction and the range of the tilt angle are generated in advance by machine learning the observed values of the tilt direction and the tilt angle known as learning data. Therefore, compared with the case of generating one learning model corresponding to all ranges of the tilt direction and the range of the tilt angle, the calculation load required for generating the learning model and the memory amount and hardware resources required for the calculation can be reduced.

Advantages of the Invention

[0013] According to the present invention, it is possible to provide a roundness measurement system, a roundness measurement method, and a program capable of measuring the roundness of the inner peripheral surface of a cylindrical measurement object with high accuracy even for measurement data affected by noise.

Brief Description of the Drawings

[0014] [Figure 1] It is a diagram for explaining a measurement object 80 which is an object to be measured for roundness by the roundness measurement system according to an embodiment of the present invention. [Figure 2] It is a diagram showing the system configuration of the roundness measurement system according to an embodiment of the present invention. [Figure 3] It is a diagram showing the detailed configuration of the data measurement device 20 in an embodiment of the present invention. [Figure 4] It is a diagram showing the hardware configuration of the terminal device 10 for measuring roundness based on the measurement data obtained by the data measurement device 20. [Figure 5] It is a diagram showing an example of measurement data obtained by the data measurement device 20. [Figure 6] It is a diagram showing a state when measurement is performed with the central axis of the measurement object 80 inclined. [Figure 7] It is a diagram showing an example of measurement data before tilt correction and measurement data after tilt correction. [Figure 8] It is a diagram showing an example of measurement data including minute noise. [Figure 9]This figure shows the process of performing machine learning on the learning model 40 used in the first embodiment of the present invention. [Figure 10] This figure shows how the learning model 40 estimates the slope direction and slope angle. [Figure 11] This is a flowchart illustrating the operation of the roundness measurement system according to the first embodiment of the present invention. [Figure 12] This figure illustrates a plurality of learning models 51 to 56 used in a second embodiment of the present invention. [Figure 13] This figure shows the process of performing machine learning on the learning models 51 to 56 used in the second embodiment of the present invention. [Figure 14] This figure shows how the slope direction and slope angle are estimated by a single selected learning model 532. [Figure 15] This is a flowchart illustrating the operation of the roundness measurement system according to the second embodiment of the present invention. [Modes for carrying out the invention]

[0015] Next, embodiments of the present invention will be described in detail with reference to the drawings.

[0016] First, before describing the roundness measurement system of this embodiment, the object to be measured, 80, which is the object whose roundness is to be measured, will be described with reference to Figure 1.

[0017] As shown in Figure 1, the object to be measured 80 has a cylindrical shape with an inner circumferential surface. The roundness measurement system of this embodiment is a system for measuring the roundness of the inner circumferential surface of a cylindrical object to be measured 80 as shown in Figure 1.

[0018] Figure 1 shows the system configuration of a roundness measurement system according to one embodiment of the present invention.

[0019] As shown in Figure 1, the roundness measurement system of this embodiment consists of a terminal device 10 such as a personal computer and a data measurement device 20. The data measurement device 20 is a device for measuring the distance from the center of the object to be measured 80 to its inner surface. The data measurement device 20 comprises a main body 21, a rotating probe 22, and a support column 23.

[0020] The support column 23 is vertically supported on a base. A sliding stage is attached to the support column 23, and this sliding stage is configured to move vertically along the support column 23. An arm is provided horizontally from the sliding stage, and the main body 21 is attached to the tip of this arm.

[0021] The main body 21 is equipped with a rotating probe 22 for measuring the inner surface of the object to be measured 80. The rotating probe 22 is designed to measure the distance between itself and the inner surface of the object to be measured by rotating inside the object 80. The tip of the rotating probe 22 has an opening for emitting laser light, and this laser light is configured to scan the inner surface of the object to be measured 80 as the measurement light.

[0022] The measurement data of the inner surface of the object to be measured 80, acquired by the data measuring device 20, is then transferred to the terminal device 10. Based on the measurement data transferred from the data measuring device 20, the terminal device 10 measures the roundness of the inner surface of the object to be measured.

[0023] Next, Figure 3 shows the detailed configuration of the data measurement device 20. As shown in Figure 3, the main body 21 includes an operation control unit 31, a displacement measurement unit 32, a transfer unit 33, and a stepping motor 24. The rotating probe 22 is equipped with a dispersion lens unit 25 and a reflective mirror 26. The data measurement device 20 is a device that measures the distance from the center of the object to be measured 80 to its inner surface with high precision and without contact by using laser light for displacement measurement.

[0024] The motion control unit 31 controls the rotational movement of the stepping motor 24. The rotation of the stepping motor 24 causes the rotation of the rotating probe 22. Since a reflective mirror 26 is provided at the tip of the rotating probe 22, the optical path of the measurement light that has passed through the dispersion lens unit 25 changes laterally, and it is emitted from the opening and irradiates the inner surface of the object to be measured 80. The measurement light reflected from the inner surface of the cylindrical object to be measured 80 is then reflected by the reflective mirror 26 and incident on the dispersion lens unit 25, and is received by the displacement measurement unit 32 via an optical fiber or the like. The displacement measurement unit 32 performs displacement measurement based on the received measurement light and measures the reflection position of the measurement light.

[0025] Furthermore, the motion control unit 31 inputs rotational position information of the rotation probe 22 to the displacement measurement unit 32. This rotational position information is synchronized with the rotation angle of the stepping motor 24, allowing the displacement measurement unit 32 to identify the position on the inner surface of the object being measured 80 where the measurement result was obtained.

[0026] The data measuring device 20 is capable of measuring distances to the inner circumferential surface of the object 80 over a 360° radius with an arbitrary resolution, such as 360 to 36,000 steps. Note that the higher the resolution of the measurement, the longer the measurement time will be.

[0027] The slide stage is configured to move the support column 23 up and down based on a control signal from the motion control unit 31. As a result, the main body 21 and the rotating probe 22 also move up and down as the slide stage moves. Therefore, by controlling the slide stage, it is possible to measure the inner surface of the object to be measured at different positions.

[0028] The stepping motor 24 may be a hollow motor that can pass through the dispersion lens unit 25, or it may have a pulley structure.

[0029] The transfer unit 33 then transfers the measurement data obtained by the displacement measurement unit 32 to the terminal device 10.

[0030] Furthermore, the measurement method for measuring the distance to the inner surface of the object to be measured 80 is not limited to the method described above, and various measurement methods can be used. Moreover, the present invention is not limited to non-contact data measurement devices using laser light, but can be similarly applied to measurement data obtained by contact-type measurement devices, measurement data obtained by 3D scanners, and measurement data from image processing systems.

[0031] Next, Figure 4 shows the hardware configuration of the terminal device 10 that measures roundness based on the measurement data obtained by the data measurement device 20.

[0032] As shown in Figure 4, the terminal device 10 consists of a display unit 11, a memory 12, a storage device 13 such as a hard disk drive, a data transmission / reception unit 14, a control unit 15, an operation input unit 16, and a learning model storage unit 17. These components are connected to each other via a control bus 18.

[0033] The data transmission / reception unit 14 receives measurement data from an external device such as a data measurement device 20.

[0034] The control unit 15 is composed of a processor such as a CPU and calculates the roundness of the inner surface of the cylindrical object to be measured 80 from the measurement data obtained by measuring the inner surface of the object to be measured 80, which is received from the data measurement device 20 via the data transmission / reception unit 14.

[0035] The display unit 11 is controlled by the control unit 15 and displays various information to the user. The operation input unit 16 receives various operation information performed by the user.

[0036] The learning model storage unit 17 is generated by machine learning using known observed values ​​of the inclination direction and inclination angle of the object to be measured 80 as training data, and stores a learning model that estimates the inclination direction and inclination angle of the object to be measured 80 from measurement data obtained by measuring the inner surface of the cylindrical object to be measured 80. Various types of data can be used as known observed values ​​of inclination direction and inclination angle, including actual measurement data, data generated in a virtual space, and data generated by computational processing. However, the following explanation will focus on the case where known measurement data of the inclination direction and inclination angle of the object to be measured 80 is used as training data.

[0037] Next, before explaining in detail how the control unit 15 calculates the roundness of the inner surface of the object to be measured 80, an example of measurement data obtained by the data measuring device 20 is shown in Figure 5.

[0038] As shown in Figure 5, the measurement data is two-dimensional data, representing the distance to the inner surface in two dimensions, with the central axis of the object being measured 80 set to (0,0). Note that the measurement data shown in Figure 5 is actually composed of numerous measurements, but in Figure 5 and the figures below, consecutive points are sometimes simply displayed as a line. If the object being measured 80 is not tilted at all, the central axis of the object being measured 80 and the central axis of the rotating probe 22 are perfectly aligned, and there is no error in the laser displacement measurement, the measurement data shown in Figure 5 will be a perfect circle.

[0039] However, in actual measurements, if the central axis of the object being measured (80) is tilted, the resulting measurement data will be elliptical.

[0040] Figure 6 shows what happens when the measurement is performed with the central axis of the object being measured 80 tilted. Figure 6(A) shows what happens when the central axis of the object being measured 80 is tilted at an angle θ. Figure 6(B) shows an example of measurement data taken with the central axis tilted in this way. In Figure 6(B), it can be seen that the measurement data is elliptical. It can also be seen that the major axis of the ellipse is longest in the tilt direction φ, which is the direction in which the central axis of the object being measured 80 is tilted. The major and minor axes of this ellipse depend on the tilt angle θ and the tilt direction φ of the object being measured 80. The larger the tilt angle θ, the longer the major axis of the ellipse. However, the minor axis maintains the original diameter of a perfect circle. Furthermore, the tilt direction φ also changes according to the direction in which the object being measured 80 is tilted.

[0041] Even if the roundness is calculated based on measurement data obtained with the central axis of the object being measured 80 tilted in this manner, it is not possible to obtain an accurate value of roundness. Therefore, the control unit 15 performs tilt correction on the obtained measurement data to correct the measurement error caused by the tilt of the object being measured 80, and calculates a more accurate roundness.

[0042] Figure 7 shows examples of measurement data before and after tilt correction. Figure 7(A) shows the measurement data before tilt correction, which is elliptical because it is tilted at an angle θ in the tilt direction φ. By applying tilt correction to the measurement data in Figure 7(A), measurement data with the measurement error due to tilt removed, as shown in Figure 7(B), can be obtained. However, in order to perform such tilt correction, it is necessary to correctly identify both the tilt direction φ and the tilt angle θ of the object being measured 80.

[0043] By performing the following calculations based on the obtained measurement data, the direction and angle of inclination can be determined and the measurement data can be corrected. (1) After obtaining the measurement data, calculate the major and minor axes of the ellipse. (2) Calculate the tilt angle and tilt direction based on the ratio of the major axis to the minor axis. (3) The measurement data is corrected using the obtained tilt angle and tilt direction.

[0044] As shown in Figure 7(A), if the measurement data is an ideal ellipse, the tilt angle and tilt direction can be obtained with high accuracy using the calculation method described above. However, as shown in Figure 8, measurement data obtained by optically measuring the inner surface of the object to be measured generates noise due to the effects of vibration, temperature changes, ambient light, surface roughness, etc. Even if the tilt angle and tilt direction are calculated based on measurement data containing such noise, the resulting tilt direction and tilt angle may contain errors. In particular, if the tilt angle and tilt direction are minute, it is difficult to distinguish the difference from noise by calculation, and there is a possibility that the noise will be mistakenly calculated as the tilt of the object to be measured. As a result, if tilt correction is performed using the tilt direction and tilt angle containing errors as correction values, the roundness obtained based on the measurement data after correction processing will also contain errors, which is a problem.

[0045] Figure 8 shows an example of measurement data containing minute noise, and illustrates the case where the number of measurement values ​​is reduced for clarity. As shown in Figure 8, when the noise in the measurement data is minute and the tilt of the object being measured is also minute, it becomes quite difficult to calculate the amount and direction of the tilt of the object being measured.

[0046] To remove noise components from measurement data, it is common practice to smooth the data using filtering methods such as calculating the moving average or median of the measured values, or using low-pass filters. However, these processing methods remove not only noise components but also minute shape changes in the measurement data, and even minute tilt information of the object being measured is removed.

[0047] Therefore, in the roundness measurement system of this embodiment, the tilt direction and tilt angle of the object to be measured are estimated using a machine learning model that has been trained in advance, and tilt correction is performed, thereby enabling tilt correction with high accuracy even for measurement data that contains noise.

[0048] [First Embodiment] First, a roundness measurement system according to the first embodiment of the present invention will be described.

[0049] Figure 9 shows how the learning model 40 used in this embodiment performs machine learning. Figure 9 shows how the learning model 40 is input as training data, with each combination of tilt direction from 1° to 360° (in 1° increments) and tilt angle from 0.1° to 40° (in 0.1° increments) being input. For example, the training data for a tilt direction of 1° and tilt angle of 0.1° is ideal measurement data obtained when the object being measured is tilted at a tilt direction of 1° and tilt angle of 0.1°. Such training data is generated for each combination of tilt direction and tilt angle and input into the learning model 40 to perform machine learning. In other words, the number of training data is 144,000 (360 × 400) combinations, consisting of 360 steps for tilt direction and 400 steps for tilt angle.

[0050] The learning model 40 generated in this manner is then stored in the learning model storage unit 17.

[0051] Figure 10 shows how the learning model 40 generated in this way estimates the slope direction and slope angle. By inputting actual measurement data into the learning model 40 that has undergone machine learning as described above, the slope direction φ and slope angle θ of the object to be measured 80 can be obtained.

[0052] Next, the operation of the roundness measurement system of this embodiment is shown in the flowchart of Figure 11.

[0053] First, the data measuring device 20 measures the inner surface of the cylindrical object to be measured 80 and transfers the measurement result as measurement data to the terminal device 10 (step S101).

[0054] Then, the control unit 15 of the terminal device 10 estimates the tilt direction and tilt angle of the object to be measured 80 by inputting the measurement data into the learning model 40 stored in the learning model storage unit 17 (step S102).

[0055] Then, the control unit 15 corrects the tilt of the measured data using the estimated tilt direction and tilt angle as correction values ​​(step S103).

[0056] Finally, the control unit 15 calculates the roundness of the inner surface of the object to be measured 80 based on the measurement data after tilt correction (step S104).

[0057] According to the roundness measurement system of this embodiment, the inclination direction and inclination angle of the object to be measured 80 can be estimated from the measurement data using the learning model 40, without being affected by noise contained in the measurement data. Therefore, according to the roundness measurement system of this embodiment, even if the measurement data is affected by noise, the roundness of the inner surface of the cylindrical object to be measured 80 can be measured with high accuracy.

[0058] [Second Embodiment] First, a roundness measurement system according to a second embodiment of the present invention will be described.

[0059] In the first embodiment described above, the slope direction and slope angle are estimated from the measurement data using a single learning model 40. However, as shown in Figure 9, performing machine learning on the learning model 40 requires a massive amount of training data, 144,000 data points. However, the amount of memory required by the GPU (Graphics Processing Unit) used for machine learning increases as the amount of data handled increases. Therefore, a massive amount of memory storage capacity is required when the total amount of training data becomes enormous. Furthermore, the more measurement values ​​are added to a single measurement data point, the larger the total amount of training data and memory required for machine learning becomes. Moreover, the more convolution operations performed in the learning model are performed, the more memory is required. Therefore, attempting to perform machine learning on the learning model 40 with higher accuracy may become practically impossible due to hardware resource constraints.

[0060] Therefore, in the roundness measurement system of this embodiment, multiple learning models 511~561, 512~562, ..., 511 are generated for each combination of the range 400 ~56 400 Use this.

[0061] Learning models 511-561, 512-562, ..., 51 in this embodiment 400 ~56 400 The tilt direction is divided into 60° increments: 0°~60°, 61°~120°, ..., 301°~360°, and the tilt angle is divided into 0.1° increments: 0.1°, 0.2°, ..., 40°. In the following description, the learning models 511~561, 512~562, ..., 51 used in this embodiment will be used. 400 ~56 400 These will be referred to as learning models 51-56, omitting the accompanying characters.

[0062] In this embodiment, the total number of learning models 51 to 56 is 2400 (6 × 400). These learning models 51 to 56 are stored in the learning model storage unit 17, similar to the first embodiment.

[0063] Next, Figure 13 shows how machine learning is performed on the learning models 51 to 56 used in this embodiment. However, Figure 13 only shows how machine learning is performed on learning model 511. Figure 13 shows how machine learning is performed on learning model 511 by inputting training data generated for each combination of tilt direction from 1° to 60° (in 1° increments) and tilt angle of 0.1 as training data. In Figure 13, the number of training data is 60, consisting of 60 steps in tilt direction and 1 step in tilt angle.

[0064] As can be seen, in this embodiment, the amount of data required to perform machine learning on a single learning model is drastically reduced to 1 / 2400th compared to the case shown in Figure 9.

[0065] Therefore, according to this embodiment, compared to the case where a single learning model 40 corresponding to all ranges of inclination directions and inclination angles is generated, the computational load required when generating learning models 51 to 56, as well as the amount of memory and hardware resources required for the calculations, can be reduced.

[0066] By inputting actual measurement data into the machine learning models 51-56, the inclination direction and inclination angle of the object being measured can be obtained.

[0067] However, in this embodiment, it is necessary to select one learning model from among many learning models 51 to 56 to input the measurement data. Therefore, the control unit 15 first calculates the inclination direction and inclination angle of the object to be measured 80 based on the measurement data from the data measuring device 20. Then, the control unit 15 selects one learning model from among the multiple learning models 51 to 56 that corresponds to the calculated inclination direction and inclination angle.

[0068] Figure 14 shows a case where the calculated slope direction is 72° and the slope angle is 0.2°. Therefore, a learning model 532 corresponding to a slope direction range of 61-120° and a slope angle of 0.2° is selected, and the measurement data is input to the selected learning model 532. As a result, the learning model 532 outputs a more accurate slope direction and slope angle, for example, a slope direction of 72.5° and a slope angle of 0.23°.

[0069] Next, the operation of the roundness measurement system of this embodiment is shown in the flowchart of Figure 15.

[0070] First, the data measuring device 20 measures the inner surface of the cylindrical object to be measured 80 and transfers the measurement result as measurement data to the terminal device 10 (step S201).

[0071] Then, the control unit 15 calculates the inclination direction and inclination angle of the object to be measured 80 from the transmitted measurement data (step S202).

[0072] Then, the control unit 15 selects one learning model from among the multiple learning models 51 to 56 that corresponds to the calculated tilt direction and tilt angle (step S203).

[0073] Next, the control unit 15 estimates the inclination direction and inclination angle of the object to be measured 80 by inputting the measurement data into one of the selected learning models (step S204).

[0074] Then, the control unit 15 corrects the tilt of the measured data using the estimated tilt direction and tilt angle as correction values ​​(step S205).

[0075] Finally, the control unit 15 calculates the roundness of the inner surface of the object to be measured 80 based on the measurement data after tilt correction (step S206).

[0076] In the roundness measurement system of this embodiment, compared to the roundness measurement system of the first embodiment, the computational load required when generating a learning model by machine learning, as well as the amount of memory and hardware resources required for the calculation, can be reduced. [Explanation of Symbols]

[0077] 10 Terminal devices 11 Display section 12 memory 13 Storage device 14. Data transmission / reception unit 15 Control Unit 16 Operation Input Section 17 Learning Model Memory Unit 18 Control bus 20 Data measurement devices 21 Main body 22 Rotating probe 23 Pillar 24 Stepping motors 25-dispersion lens unit 26 Reflective mirrors 31 Operation Control Unit 32 Displacement Measurement Unit 33 Transfer section 40 Learning Models 51-56 Learning Models 80. Objects to be measured

Claims

1. A measuring device for measuring the inner surface of a cylindrical object, A storage unit stores a learning model that estimates the tilt direction and tilt angle of an object from measurement data obtained by the measuring device. This model is generated by machine learning using observed values ​​obtained from an object whose tilt direction (indicating the rotation angle around the central axis) and tilt angle (indicating the inclination of the central axis) are known, and the tilt direction and tilt angle are used as learning data. The system comprises a control unit that calculates the roundness of the inner surface of a cylindrical object from measurement data obtained by measuring the inner surface of the object, The control unit, By inputting the measurement data into the learning model, the tilt direction and tilt angle of the object being measured are estimated. The estimated inclination direction and inclination angle are used as correction values ​​to correct the measurement data. Based on the corrected measurement data, the roundness of the inner surface of the object to be measured is calculated. Roundness measurement system.

2. A measuring device for measuring the inner surface of a cylindrical object, A storage unit stores multiple learning models that estimate the tilt direction and tilt angle of an object from measurement data obtained by the measuring device. These models are generated for each combination of the range of the tilt direction and the range of the tilt angle by machine learning using observed values ​​obtained from a known object, where the tilt direction and tilt angle are known as learning data, and the measurement device stores multiple learning models that estimate the tilt direction and tilt angle of an object from measurement data obtained by the measuring device. The system comprises a control unit that calculates the roundness of the inner surface of a cylindrical object from measurement data obtained by measuring the inner surface of the object, The control unit, Based on the aforementioned measurement data, the inclination direction and inclination angle of the object to be measured are calculated. From among the multiple learning models mentioned above, select one learning model that corresponds to the calculated slope direction and slope angle. By inputting the measurement data into one selected learning model, the inclination direction and inclination angle of the object being measured are estimated. The estimated inclination direction and inclination angle are used as correction values ​​to correct the measurement data. Based on the corrected measurement data, the roundness of the inner surface of the object to be measured is calculated. Roundness measurement system.

3. A step of storing a learning model which is generated by machine learning using observed values ​​obtained from a measurement target whose tilt direction indicating the rotation angle around the central axis and tilt angle indicating the inclination of the central axis are known, and the tilt direction and tilt angle as learning data, and which estimates the tilt direction and tilt angle of a cylindrical measurement target from measurement data obtained by measuring the inner surface of the measurement target. When calculating the roundness of the inner circumferential surface of a cylindrical object from measurement data obtained by measuring the inner circumferential surface of the object, the steps include: inputting the measurement data into the learning model to estimate the inclination direction and inclination angle of the object; The steps include correcting the measurement data using the estimated tilt direction and tilt angle as correction values, A step of calculating the roundness of the inner surface of the object to be measured based on the corrected measurement data, A method for measuring roundness equipped with the following features.

4. A step of storing a plurality of learning models which are generated for each combination of the range of the inclination direction and the range of the inclination angle of a cylindrical object, by machine learning using observed values ​​obtained from a known object to be measured, which have an inclination direction indicating the rotation angle around the central axis and an inclination angle indicating the tilt of the central axis, and which estimate the inclination direction and the inclination angle of a cylindrical object to be measured from measurement data obtained by measuring the inner surface of the cylindrical object to be measured, When calculating the roundness of the inner circumferential surface of a cylindrical object from measurement data obtained by measuring the inner circumferential surface of the object, the steps include calculating the inclination direction and inclination angle of the object based on the measurement data, The steps include selecting one learning model from among the aforementioned multiple learning models that corresponds to the calculated slope direction and slope angle, The steps include: estimating the inclination direction and inclination angle of the object to be measured by inputting the measurement data into a selected learning model; The steps include correcting the measurement data using the estimated tilt direction and tilt angle as correction values, A step of calculating the roundness of the inner surface of the object to be measured based on the corrected measurement data, A method for measuring roundness equipped with the following features.

5. A step of storing a learning model which is generated by machine learning using observed values ​​obtained from a measurement target whose tilt direction indicating the rotation angle around the central axis and tilt angle indicating the inclination of the central axis are known, and the tilt direction and tilt angle as learning data, and which estimates the tilt direction and tilt angle of a measurement target from measurement data obtained by measuring the inner surface of a cylindrical measurement target, When calculating the roundness of the inner circumferential surface of a cylindrical object from measurement data obtained by measuring the inner circumferential surface of the object, the steps include: inputting the measurement data into the learning model to estimate the inclination direction and inclination angle of the object; The steps include correcting the measurement data using the estimated tilt direction and tilt angle as correction values, A step of calculating the roundness of the inner surface of the object to be measured based on the corrected measurement data, A program that causes a computer to execute something.

6. A step of storing a plurality of learning models which are generated for each combination of the range of the inclination direction and the range of the inclination angle of a cylindrical object, by machine learning using observed values ​​obtained from a measurement target object whose inclination direction indicating the rotation angle around the central axis and inclination angle indicating the inclination of the central axis are known, and the inclination direction and the inclination angle as learning data, and which estimate the inclination direction and the inclination angle of a measurement target object from measurement data obtained by measuring the inner surface of the cylindrical object, When calculating the roundness of the inner circumferential surface of a cylindrical object from measurement data obtained by measuring the inner circumferential surface of the object, the steps include calculating the inclination direction and inclination angle of the object based on the measurement data, The steps include selecting one learning model from among the aforementioned multiple learning models that corresponds to the calculated slope direction and slope angle, The steps include: estimating the inclination direction and inclination angle of the object to be measured by inputting the measurement data into a selected learning model; The steps include correcting the measurement data using the estimated tilt direction and tilt angle as correction values, A step of calculating the roundness of the inner surface of the object to be measured based on the corrected measurement data, A program that causes a computer to execute something.

Images