Parameter adjustment device, learning device, measurement system, parameter adjustment method, and program product

Abstract

The parameter adjustment device (500) adjusts a parameter related to control of laser light projected toward an object by the measurement sensor (210). The parameter calculation section (520) applies waveform data newly acquired in a new state to a trained model generated by machine learning using learning data including the waveform data including the light-receiving amount of the measurement sensor (210) and data indicating the parameter at the time of acquisition of the waveform data, and calculates a parameter with which the object can be measured using the measurement sensor (210) in the new state. The parameter output section (530) outputs data indicating the parameter calculated by the parameter calculation section (520).

Description

TECHNICAL FIELD

[0001] The present application relates to a parameter adjustment device, a learning device, a measurement system, a parameter adjustment method, and a program product. BACKGROUND

[0002] A laser displacement sensor that measures an object using laser light, for example, if measurement is performed using a plurality of sensor heads, or measurement is performed on a transparent object, laser light interferes and there are two or more peak waveforms of light receiving amount at a place where there is originally one, and it is possible that a stable measurement result cannot be obtained. Therefore, a user uses an information processing system including a computer device to analyze the waveforms of the light receiving amount, and then adjusts parameters related to the interference of the laser light.

[0003] As one example of the information processing system as described above, an optical displacement measurement system capable of setting a photographing condition is disclosed in Patent Literature 1. The optical displacement measurement system described in Patent Literature 1 calculates the reliability of a photographing parameter on which appropriate measurement can be performed in accordance with a photographing result of the photographing parameter set in advance, arranges and displays a plurality of photographing parameters in order from high to low in reliability, and allows a user to select an arbitrary photographing parameter. In addition, as one example of the computer device as described above, an inspection device capable of adjusting a control parameter of a displacement meter that acquires a two-dimensional profile indicating the cross-sectional shape of an object is disclosed in Patent Literature 2. The inspection device described in Patent Literature 2 adjusts the shape and area of an image mask, which is one example of an adjustment parameter, and thereby can reduce the influence of stray light or insufficient light receiving amount.

[0004] Patent Literature 1: Japanese Patent Application Publication No. 2016-161473

[0005] Patent Literature 2: Japanese Patent Application Publication No. 2015-163841 SUMMARY

[0006] The optical displacement measurement system described in Patent Literature 1 and the inspection device described in Patent Literature 2 have parameters set manually by a user, and it is possible that the parameters cannot be appropriately set depending on the proficiency of the user, and the object cannot be appropriately measured. In particular, the optical displacement measurement system described in Patent Literature 1 has a problem that the user can only select a photographing parameter from a plurality of kinds of photographing parameters set in advance, and even if the user selects the best photographing parameter, appropriate measurement cannot be performed. In addition, the inspection device described in Patent Literature 2 has a problem that even if the shape and area of the image mask, which is an adjustment parameter, are adjusted, the light receiving amount itself cannot be increased, and it is possible that the light receiving amount required for acquisition of the two-dimensional profile is insufficient.

[0007] The present application has been made in view of the above-described circumstances, and has an object to appropriately perform measurement on an object.

[0008] To achieve the above object, a parameter adjustment device according to the present application adjusts a parameter related to control of laser light projected by a measurement sensor toward an object. The parameter adjustment device has a parameter calculation section and a parameter output section. The parameter calculation section applies, to a waveform data newly acquired in a new state in which laser light is projected by the measurement sensor toward the object, a trained model generated by machine learning using learning data including the waveform data including an amount of light received by the measurement sensor from the object and data indicating a parameter at the time of acquisition of the waveform data, and calculates a parameter with which the measurement sensor can perform measurement on the object in the new state. The parameter output section outputs data indicating the parameter calculated by the parameter calculation section.

[0009] Effects of the Invention

[0010] According to the present application, to a waveform data newly acquired in a new state, a trained model generated by machine learning using learning data including the waveform data and data indicating a parameter is applied, and a parameter with which the measurement sensor can perform measurement on the object in the new state is calculated and output. Therefore, by the parameter determined from the output data, the measurement sensor performs measurement on the object, and thus the object is appropriately measured. BRIEF DESCRIPTION OF DRAWINGS

[0011] Figure 1 is a diagram showing a functional configuration of a measurement system according to an embodiment of the present application.

[0012] Figure 2 is a block diagram showing a hardware configuration of a measurement device, a learning device, and a parameter adjustment device according to the present embodiment.

[0013] Figure 3 is a diagram showing an example of a parameter according to the present embodiment.

[0014] Figure 4A is a diagram showing a display example 1 of waveform data according to the present embodiment.

[0015] Figure 4B is a diagram showing a display example 2 of waveform data according to the present embodiment.

[0016] Figure 5A is an explanation of adjustment of a projection amount and a measurement center without specifying a projection amount adjustment region Figure 1 .

[0017] Figure 5Bis an explanatory view of adjustment of the projection amount and the measurement center when the projection amount adjustment region is not specified Figure 2 .

[0018] Figure 5C is an explanatory view of adjustment of the projection amount when the projection amount adjustment region is specified

[0019] Figure 5D is an explanatory view of adjustment of the projection amount when the interference avoidance function is set to be active

[0020] Figure 6A is a view showing a success example 1 of the measurement according to the present embodiment.

[0021] Figure 6B is a view showing a failure example 1 of the measurement according to the present embodiment.

[0022] Figure 7A is a view showing a success example 2 of the measurement according to the present embodiment.

[0023] Figure 7B is a view showing a failure example 2 of the measurement according to the present embodiment.

[0024] Figure 8A is a view showing a success example 3 of the measurement according to the present embodiment.

[0025] Figure 8B is a view showing a failure example 3 of the measurement according to the present embodiment.

[0026] Figure 9 is a view showing a display example of a measurement result display screen according to the present embodiment.

[0027] Figure 10 is a flowchart showing a trained model generation process according to the present embodiment.

[0028] Figure 11 is a flowchart showing a parameter adjustment process according to the present embodiment.

[0029] Figure 12 is a flowchart showing a measurement process according to the present embodiment. DETAILED DESCRIPTION

[0030] Hereinafter, a parameter adjustment device, a learning device, a measurement system, a parameter adjustment method, and a program according to the present embodiment will be described in detail with reference to the drawings. In the drawings, the same or equivalent parts are denoted by the same reference numerals.

[0031] The measurement system according to an embodiment of the present application uses a measurement sensor to measure an object. In addition, the parameter adjustment device, the parameter adjustment method, and the program according to an embodiment of the present application adjust a parameter related to control of laser light projected by the measurement sensor toward the object at the time of measurement. In addition, the learning device according to an embodiment of the present application generates a trained model used by the parameter adjustment device, the parameter adjustment method, and the program to adjust the parameter through machine learning.

[0032] Figure 1 is a diagram showing a functional configuration of the measurement system according to an embodiment of the present application.

[0033] The measurement system 100 according to the present embodiment has a measurement device 200 including a laser displacement sensor that measures an object, as shown in Figure 1 Here, the laser displacement sensor is a sensor that emits laser light from a laser diode of a sensor head, receives light from the object by a CMOS sensor as a light-receiving element, and measures the object. In addition, the measurement system 100 has a learning device 300 that generates a trained model through machine learning using learning data, a trained model storage DB 400 that stores the trained model, and a parameter adjustment device 500 that adjusts a parameter at the time of measurement using the trained model. The measurement device 200, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 can transmit and receive data via a LAN (Local Area Network) not shown.

[0034] The measurement device 200 has a measurement sensor 210 that measures an object, a parameter setting section 220 that sets a parameter of the measurement sensor 210, and a waveform data generation section 230 that generates waveform data of an amount of light received by the measurement sensor 210 from the object. In addition, the measurement device 200 has a measurement section 240 that measures the object based on the waveform data and a measurement result display section 250 that displays a measurement result.

[0035] The learning device 300 has a learning data acquisition section 310 that acquires learning data and a trained model generation section 320 that generates a trained model through machine learning using the acquired learning data.

[0036] The trained model storage DB 400 has a trained model storage section 410 that stores the trained model generated by the trained model generation section 320.

[0037] The parameter adjustment device 500 has a waveform data acquisition section 510 that acquires waveform data, a parameter calculation section 520 that calculates parameters by applying the acquired waveform data to the trained model stored in the trained model storage section 410, and a parameter output section 530 that outputs data indicating the calculated parameters.

[0038] Figure 2 is a block diagram indicating the hardware structure of the measurement device, the learning device, and the parameter adjustment device according to the present embodiment.

[0039] Figure 1 The measurement device 200, the learning device 300, and the parameter adjustment device 500 illustrated in Figure 2 are configured by the hardware illustrated in FIG. 5. The measurement device 200 has a control section 51, a main storage section 52, an external storage section 53, an operation section 54, a display section 55, and a transceiver section 56. Also, the learning device 300 has the control section 51, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiver section 56, similarly to the measurement device 200. Also, the parameter adjustment device 500 has the control section 51, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiver section 56, similarly to the measurement device 200 and the learning device 300. Furthermore, although not illustrated, the trained model storage DB 400 has at least the external storage section 53.

[0040] The control section 51 performs processing in accordance with a control program 59. The control section 51 has a CPU (Central Processing Unit). The control section 51 functions as the parameter setting section 220, the waveform data generation section 230, and the measurement section 240 of the measurement device 200 in accordance with the control program 59. For example, the control section 51 performs the parameter setting step performed by the parameter setting section 220, the waveform data generation step performed by the waveform data generation section 230, and the measurement step performed by the measurement section 240. Also, the control section 51 functions as the learning data acquisition section 310 and the trained model generation section 320 of the learning device 300 in accordance with the control program 59. For example, the control section 51 performs the learning data acquisition step performed by the learning data acquisition section 310 and the trained model generation step performed by the trained model generation section 320. Also, the control section 51 functions as the waveform data acquisition section 510, the parameter calculation section 520, and the parameter output section 530 of the parameter adjustment device 500 in accordance with the control program 59. For example, the control section 51 performs the waveform data acquisition step performed by the waveform data acquisition section 510, the parameter calculation step performed by the parameter calculation section 520, and the parameter output step performed by the parameter output section 530.

[0041] The main storage section 52 loads the control program 59, which is used as a work area of the control section 51. The main storage section 52 has a RAM (Random-Access Memory).

[0042] The external storage section 53 stores the control program 59 in advance. The external storage section 53 supplies data stored by the program to the control section 51 and stores data supplied from the control section 51, according to an instruction from the control section 51. The external storage section 53 has a nonvolatile memory such as a flash memory, an HDD (Hard Disk Drive), an SSD (Solid State Drive), and the like. The external storage section 53 functions as a trained model storage section 410 in the trained model storage DB 400.

[0043] The operation section 54 supplies input information to the control section 51. The operation section 54 has an information inputting section such as a keyboard, a mouse, a touch panel, an operation button, and the like. The operation section 54 functions as a measurement sensor 210 in the measurement device 200.

[0044] The display section 55 displays information input via the operation section 54, information output by the control section 51, and the like. The display section 55 has a display device such as an LCD (Liquid Crystal Display), an organic EL (Electro-Luminescence) display, and the like. The display section 55 functions as a measurement result display section 250 in the measurement device 200.

[0045] The transceiving section 56 transmits and receives information. The transceiving section 56 has an information communication section such as a network terminal device connected to a network, a wireless communication device, and the like.

[0046] In the measurement device 200, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiving section 56 are connected to the control section 51 via the internal bus 50. Also in the learning device 300, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiving section 56 are connected to the control section 51 via the internal bus 50, as in the measurement device 200. Also in the parameter adjustment device 500, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiving section 56 are connected to the control section 51 via the internal bus 50, as in the measurement device 200 and the learning device 300.

[0047] Figure 1The control section 51 of the measurement apparatus 200 uses the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiver section 56 as resources, thereby realizing the functions of the measurement sensor 210, the parameter setting section 220, the waveform data generation section 230, the measurement section 240, and the measurement result display section 250. Further, the control section 51 of the learning apparatus 300 uses the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiver section 56 as resources, thereby realizing the functions of the learning data acquisition section 310 and the trained model generation section 320. Further, the control section 51 of the parameter adjustment apparatus 500 uses the main storage section 52, the external storage section 53, the operation section 54, the display section 55, and the transceiver section 56 as resources, thereby realizing the functions of the waveform data acquisition section 510, the parameter calculation section 520, and the parameter output section 530.

[0048] Returning to Figure 1 , the measurement sensor 210 is, for example, a laser displacement sensor that measures a displacement, i.e., a distance up to an object, by a triangulation method. The measurement sensor 210 has a first measurement sensor 211 and a second measurement sensor 212. The first measurement sensor 211 has Figure 6A , Figure 6B , Figure 7A , Figure 7B , Figure 8A , Figure 8B a first sensor head 211a and a not-illustrated first sensor head controller that controls the first sensor head 211a. Further, the second measurement sensor 212 has Figure 6A , Figure 6B , Figure 7A , Figure 7B a second sensor head 212a and a not-illustrated second sensor head controller that controls the second sensor head 212a. Further, the first sensor head controller and the second sensor head controller are, for example, a PLC (Programmable Logic Controller). Further, in the present embodiment, the first sensor head controller and the second sensor head controller are provided for each kind of the measurement sensor 210, but are not limited thereto, and for example, the measurement sensor 210 can have a sensor head controller that controls both the first sensor head 211a and the second sensor head 212a.

[0049] Figure 3 is a diagram that shows an example of a parameter related to the present embodiment.

[0050] Returning to Figure 1 , the parameter setting section 220 sets, for example, a parameter for each kind of the measurement sensor 210 Figure 3The illustrated projection amount adjustment region designation parameter and the interference avoidance parameter are parameters of the measurement sensor 210. That is, the parameter setting section 220 sets the projection amount adjustment region designation parameter and the interference avoidance parameter of the first measurement sensor 211 and the projection amount adjustment region designation parameter and the interference avoidance parameter of the second measurement sensor 212. Here, the projection amount adjustment region designation parameter is a parameter that designates a region in which the projection amount of the laser light projected by the measurement sensor 210 is adjusted. For example, in the projection amount adjustment region designation parameter, at least for each kind of the measurement sensor 210, parameters corresponding to items indicating the "projection amount" of the projection amount, the "measurement center position" indicating the center position of the measurement corresponding to the measurement center distance, the "measurement start position" indicating the start position of the measurement, and the "measurement end position" indicating the end position of the measurement are included. In addition, the interference avoidance parameter is a parameter that avoids interference of the reception of the laser light by one measurement sensor with the reception of the laser light by another measurement sensor. For example, in the interference avoidance parameter, at least for each kind of the measurement sensor 210, a parameter corresponding to the item of the "sampling interval" indicating the sampling interval at the time of the measurement is included.

[0051] For example, the value of the projection amount of the first measurement sensor 211 is LI01, the value of the measurement center position is CT01, the value of the measurement start position is ST01, the value of the measurement end position is ED01,..., and the value of the sampling interval is SI01,.... In addition, for example, the value of the projection amount of the second measurement sensor 212 is LI02, the value of the measurement center position is CT02, the value of the measurement start position is ST02, the value of the measurement end position is ED02,..., and the value of the sampling interval is SI02,.... That is, the parameter setting section 220 sets the values of the projection amount, the measurement center position, the measurement start position, the measurement end position, the sampling interval, and the like as parameters for each kind of the measurement sensor 210.

[0052] Returning to Figure 1 The waveform data generation section 230 generates waveform data of the amount of reception of the laser light by the measurement sensor 210 from the object when the laser light is projected by the measurement sensor 210 toward the object by the parameters set by the parameter setting section 220. The waveform data generation section 230 generates waveform data of the amount of reception of the first measurement sensor 211 and waveform data of the amount of reception of the second measurement sensor 212.

[0053] Figure 4A is a diagram indicating a display example 1 of the waveform data according to the present embodiment. Figure 4B is a diagram indicating a display example 2 of the waveform data according to the present embodiment.

[0054] The waveform data generating section 230 generates waveform data based on, for example, the amount of light received by the laser light received by the first measurement sensor 211, as shown in Figure 4A , Figure 4B . Figure 4A , Figure 4B The waveform data shown in FIG. 6 is displayed by a two-dimensional graph in which the vertical axis represents the amount of light received and the horizontal axis represents the imaging position of the light receiving element. In addition, the waveform data shown in FIG. 6 has a peak waveform of an amount of light received of about 600 at a position of about 260, on the other hand, the waveform data shown in FIG. 7 has a peak waveform of an amount of light received of about 800 at a position of about 300, and a peak waveform of an amount of light received of about 350 at a position of about 420. In addition, in this case, the actual position of the object can be determined based on the distance from the first measurement sensor 211 to the object determined from the peak waveform of the waveform data shown in FIG. 6 at a position of about 260. Figure 4A Figure 4B Figure 4A

[0055] Figure 5A is a description of adjustment of the projection amount and the measurement center when the projection amount adjustment region is not specified Figure 1 .

[0056] In addition, the waveform data generated by the waveform data generating section 230 can be adjusted by the parameters set by the parameter setting section 220 described above. For example, as shown in Figure 5A , in the case where a peak waveform of an amount of light received of about 600 is generated at a position of about 160, if the projection amount and the measurement center position are adjusted, a peak waveform of an amount of light received of about 600 can be newly generated at a position of 300, and the peak waveform can be adjusted to be suitable for position detection.

[0057] Figure 5B is a description of adjustment of the projection amount and the measurement center when the projection amount adjustment region is not specified Figure 2 .

[0058] However, for example, as shown in Figure 5B , in the case where a plurality of peak waveforms are generated, adjustment of the projection amount and the measurement center position in matching with which peak waveform becomes a problem. For example, as shown in Figure 5B ​​​As shown, consider the case where, although the peak waveform of the measurement object is generated at a position of 300, a peak waveform of the outside of the measurement object is also generated at a position of about 160. In this case, for example, if the measurement object is at the surface of the transparent substrate, on the other hand, the circuit pattern of the metal is at the outside of the measurement object, the light receiving amount of the peak waveform of the outside of the measurement object is larger, 400, than the light receiving amount of the peak waveform of the measurement object, about 200. Therefore, there is a problem that adjustment is made in accordance with the peak waveform of the outside of the measurement object, which has a large light receiving amount, as Figure 5B As shown, the projection amount and the measurement center position are adjusted, and a peak waveform of the outside of the measurement object, which has a light receiving amount of about 800, is newly generated at a position of 300.

[0059] Figure 5C is an explanatory diagram when the projection amount is adjusted by specifying the projection amount adjustment region.

[0060] On the other hand, if the projection amount adjustment region specification parameter is valid, as shown in Figure 5C by specifying the projection amount adjustment region from the measurement start position to the measurement end position, for example, the peak waveform of the outside of the measurement object is masked, and it is possible to leave the peak waveform of the measurement object within the projection amount adjustment region and set the peak waveform of the outside of the measurement object outside the projection amount adjustment region. In this case, if the measurement center position is not adjusted but only the projection amount is adjusted, it is possible to newly generate a peak waveform of the measurement object, which has a light receiving amount of about 600, at a position of 300, and it is possible to adjust the peak waveform to be suitable for position detection.

[0061] Figure 5D is an explanatory diagram when the projection amount is adjusted with the interference avoidance function set to be valid.

[0062] In addition, for example, measurement is performed not only using the first measurement sensor 211 but also using the second measurement sensor 212, and as a result, sometimes the laser light projected by one measurement sensor interferes with the light receiving of the other measurement sensor, and thus multiple peak waveforms are generated. For example, as shown in Figure 5D although the peak waveform of the measurement object is generated at a position of 300, sometimes a peak waveform of the outside of the measurement object is also generated at a position of about 160. On the other hand, if the interference avoidance parameter is valid, as shown in Figure 5D by staggering the sampling interval of one measurement sensor and the sampling interval of the other measurement sensor, a peak waveform of the outside of the measurement object is not generated. In this state, if the measurement center position is not adjusted but only the projection amount is adjusted, it is possible to newly generate a peak waveform of the measurement object, which has a light receiving amount of about 600, at a position of 300, and it is possible to adjust the peak waveform to be suitable for position detection.

[0063] Figure 6A This is a graph showing a successful example 1 of the measurement involved in this embodiment. Figure 6B This is a diagram showing example 1 of the measurement failure involved in this embodiment.

[0064] The measuring unit 240 measures the object based on waveform data. For example, the measuring unit 240... Figure 6A , Figure 6B As shown, when the first sensor head 211a and the second sensor head 212a are arranged at the position where the object is clamped in the vertical direction, the thickness of the object is measured based on the waveform data of the light received by the first measuring sensor 211 and the waveform data of the light received by the second measuring sensor 212.

[0065] Specifically, firstly, the measuring unit 240 determines the position of the upper surface of the object based on the position of the peak waveform of the waveform data of the light received by the first measuring sensor 211 and the actual measuring position of the first sensor head 211a. Next, the measuring unit 240 determines the position of the lower surface of the object based on the position of the peak waveform of the waveform data of the light received by the second measuring sensor 212 and the actual measuring position of the second sensor head 212a. Furthermore, the measuring unit 240 calculates the thickness of the object by measuring the distance between the positions of the upper and lower surfaces.

[0066] Furthermore, in this embodiment, the measuring unit 240 measures the thickness of the object, but it is not limited to this. For example, it may also measure the displacement of the object, i.e., the distance to the object. In addition, it may also measure physical quantities such as the height and width of the object.

[0067] Here, we will describe a measurement example of the object involved in the measurement unit 240. First, we will consider the measurement example of the object as... Figure 6A The thickness of a workpiece 600, an example of an object shown, is measured. The workpiece 600 is a planar plate-shaped object having an upper surface and a lower surface. The workpiece 600 is an object into which laser light is projected from a first sensor head 211a from above, and reflected from the upper surface is directly received by the first sensor head 211a; and into which laser light is projected from a second sensor head 212a from below, and reflected from the lower surface is directly received by the second sensor head 212a. In this case, the imaging position and light distribution of the laser light in the first sensor head 211a and the second sensor head 212a are stable, and the waveform data generation unit 230 can generate… Figure 4A The waveform data shown generates a single peak waveform, and the measuring unit 240 is able to stably measure the thickness of the workpiece 600.

[0068] On the other hand, consider the role of Figure 6BThe thickness of the infiltrated workpiece 601, as shown in the example, was measured. Furthermore, the infiltrated workpiece 601 is a planar plate-shaped object, similar to workpiece 600, having an upper and lower surface; however, unlike workpiece 600, it is an object formed by material infiltrating inward from the projection surface using a projected laser. In this case, for example... Figure 6A As shown, if laser light is projected from the upper first sensor head 211a, a portion of the laser light is reflected at the upper surface, while the remaining laser light penetrates into the interior. This penetrated laser light may be received by the lower second sensor head 212a. Therefore, the penetrated laser light may interfere with the laser light projected from the lower second sensor head 212a and reflected at the lower surface. As a result, the imaging position and light distribution of the laser light in the first sensor head 211a and the second sensor head 212a become unstable, and the waveform data generation unit 230 generates… Figure 4B , Figure 5D The waveform data shown generates multiple peak waveforms, which may prevent the measuring unit 240 from measuring the thickness of the penetrated workpiece 601. Therefore, in order to stably measure the thickness of the penetrated workpiece 601, it is necessary to avoid interference between the laser emitted from one measuring sensor and the light received by the other measuring sensor, at least regarding the laser emitted from the first sensor head 211a and the second sensor head 212a. The interference avoidance parameter needs to be adjusted to an appropriate value. In addition, in order to stabilize the amount of light received by the laser in the first sensor head 211a and the second sensor head 212a, the value of the "projection amount" parameter specified in the projection amount adjustment area can be adjusted to an appropriate value.

[0069] Figure 7A This is a graph showing a successful example 2 of the measurement involved in this embodiment. Figure 7B This is a diagram showing example 2 of the measurement failure involved in this embodiment.

[0070] Additionally, consider arranging the first sensor head 211a and the second sensor head 212a above to... Figure 7A The displacement of the upper surface of the workpiece 600 shown is measured. In this case, the laser light projected from the first sensor head 211a and reflected at the upper surface is directly received by the first sensor head 211a, and the laser light projected from the second sensor head 212a and reflected at a different position from the upper surface is directly received by the second sensor head 212a. Therefore, the imaging position and light distribution of the laser light in the first sensor head 211a and the second sensor head 212a are stable, and the waveform data generation unit 230 can generate... Figure 4A The waveform data shown generates a single peak waveform, and the measuring unit 240 can stably measure the displacement of the upper surface of the workpiece 600.

[0071] On the other hand, consider a case where the displacement amount of the upper surface of the scattering workpiece 602, which is one example of the object shown in FIG. 6, is measured. Further, the scattering workpiece 602 is a planar plate-shaped object having an upper surface and a lower surface like the workpiece 600, and on the other hand, is an object having fine irregularities formed on the upper surface unlike the workpiece 600. In this case, for example, as shown in FIG. 7, if laser light is projected from the upper first sensor head 211a, the laser light is diffusely reflected and scattered at the upper surface, and a part of the scattered laser light can be received by the second sensor head 212a. Therefore, the scattered laser light can interfere with laser light that is projected from the second sensor head 212a and reflected at the lower surface in the vicinity of the first sensor head 211a. As a result, like the case shown in FIG. 6, the measurement unit 240 can not be able to measure the displacement amount of the upper surface of the scattering workpiece 602. Figure 7B Figure 7A Figure 6B Figure 6B

[0072] Figure 8A is a graph showing a success example 3 of the measurement according to the present embodiment, Figure 8B is a graph showing a failure example 3 of the measurement according to the present embodiment.

[0073] Further, consider a case where the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603 shown in FIG. 8 are measured by only the first measurement sensor. Further, the transparent workpiece 603 is a planar plate-shaped transparent object having an upper surface and a lower surface. In this case, laser light that is projected from the first sensor head 211a and reflected at the upper surface is received by the first sensor head 211a, and laser light that is transmitted and reflected at the lower surface is also received by the first sensor head 211a. At this time, as shown in FIG. 9, the first sensor head 211a can receive laser light that is projected from the second sensor head 212a and reflected at the lower surface. Therefore, the laser light that is projected from the second sensor head 212a can interfere with the laser light that is projected from the first sensor head 211a and reflected at the upper surface. As a result, like the case shown in FIG. 8, the measurement unit 240 can not be able to measure the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603. Figure 8A Figure 8A ​​​​​As illustrated, the directions of the laser light reflected from the upper surface and the directions of the laser light reflected from the lower surface are different, and the imaging positions of the laser light in the first sensor head 211a become different. Therefore, the distribution of the imaging positions and the light receiving amounts of the laser light in the first sensor head 211a becomes stable, and the waveform data generation section 230, although omitted from illustration, is able to generate waveform data in which two largest peak waveforms are generated, and the measurement section 240 is able to stably measure the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603 on the basis of the waveform data.

[0074] On the other hand, a case in which the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603 whose lower surface is fixed with the reflection plate 604 illustrated are measured is considered. In this case, a part of the laser light that is transmitted and reflected by the reflection plate 604 is received by the first sensor head 211a, and the remaining laser light is likely to be received by the first sensor head 211a after being reflected by the upper surface and the reflection plate 604 in this order. At this time, as illustrated, the directions of the laser light that is reflected by the upper surface and the reflection plate 604 are different from the direction of the laser light that is reflected only at the upper surface and the direction of the laser light that is reflected only by the reflection plate 604, and the imaging positions of the laser light in the first sensor head 211a all become different. Therefore, the distribution of the imaging positions and the light receiving amounts of the laser light in the first sensor head 211a becomes unstable, and the waveform data generation section 230, although omitted from illustration, generates waveform data in which three or more peak waveforms are generated, and the measurement section 240 is likely to be unable to stably measure the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603 on the basis of the waveform data. Thus, in order to stably measure the displacement amounts of the upper surface and the lower surface of the transparent workpiece 603, it is necessary to adjust the values of "projection amount", "measurement center position", "measurement start position", and "measurement end position" of the projection amount adjustment region designation parameter to appropriate values. Figure 8B Figure 8B

[0075] As described in the above description, in order to stably measure the object by the measurement section 240, it is necessary to adjust the parameters in correspondence with the characteristics of the object of measurement. Therefore, the measurement system 100 in the present embodiment cooperates with the learning device 300 that is able to perform transmission and reception of data with the measurement device 200, the trained model storage DB 400, and the parameter adjustment device 500 to adjust the parameters set in the measurement device 200.

[0076] Figure 9 is a diagram that shows a display example of a measurement result display screen related to the present embodiment.

[0077] The measurement result display section 250, for example, generates a display screen by Figure 9 ​​The measurement result display screen shown displays the measurement result measured by the measurement unit 240, and thus displays the measurement result. In the measurement result display screen shown, for example, a character image representing the name of an item such as "object" or "thickness" and a character image representing the value of an item such as "P01" or "T01" are displayed. Figure 9 In the measurement result display screen shown, for example, a character image representing the name of an item such as "object" or "thickness" and a character image representing the value of an item such as "P01" or "T01" are displayed.

[0078] Further, in the present embodiment, the measurement result display unit 250 displays the thickness of the object in the measurement result display screen, but is not limited thereto. For example, when the measurement unit 240 measures a physical quantity such as the displacement amount, height, or width of the object, these physical quantities can be displayed as the measurement result.

[0079] The learning data acquisition unit 310 acquires, as learning data, the waveform data generated by the waveform data generation unit 230 and data representing the parameters set by the parameter setting unit 220 when the waveform data generation unit 230 generates the waveform data. Specifically, the learning data acquisition unit 310 acquires learning data including waveform data of the light-receiving amount of the first measurement sensor 211 and data representing the projection amount adjustment region designation parameter and the interference avoidance parameter of the first measurement sensor 211 at the time of acquisition of the waveform data. In addition, the learning data acquisition unit 310 acquires learning data including waveform data of the light-receiving amount of the second measurement sensor 212 and data representing the projection amount adjustment region designation parameter and the interference avoidance parameter of the second measurement sensor 212 at the time of acquisition of the waveform data. That is, the learning data acquisition unit 310 acquires, for each kind of measurement sensor 210, learning data including waveform data and data representing the parameters at the time of acquisition of the waveform data.

[0080] The trained model generation unit 320 generates a trained model by machine learning using the learning data acquired by the learning data acquisition unit 310. The trained model generation unit 320 generates a trained model capable of estimating the parameters that enable measurement of the object using the measurement sensor 210 in a new state in which new waveform data is acquired, based on the learning data including the waveform data and the data representing the parameters at the time of acquisition of the waveform data.

[0081] The trained model generation unit 320 generates the trained model using, as an example of an algorithm of a known reinforcement learning (Reinforcement Learning) used in machine learning, Q-learning. Here, reinforcement learning is machine learning in which an agent observes a parameter of an environment, i.e., a current state, and determines an action to be taken. In reinforcement learning, the environment dynamically changes through the action of the agent, and a reward is given to the agent in correspondence with the change in the environment. Moreover, in reinforcement learning, the agent repeatedly performs the above action, and learns an action policy that gives the most reward through a series of actions.

[0082] In addition, in Q-learning, an action policy that gives the most reward is calculated as an action value based on an action value function. Here, let the state of the environment at time t be s t , let the action at time t be a t , let the state changed through the action a t be s t+1 , let the state change from s t to s t+1 , and let the reward obtained thereby be r t+1 . Let the discount rate be γ, and let the learning coefficient be α, 0 < γ ≤ 1 and 0 < α ≤ 1 hold. Moreover, in the case where the action value function is Q(s t , a t ), a general update formula of the action value function Q(s t , a t ) is represented by the following Equation 1.

[0083] [Equation 1]

[0084]

[0085] In addition, in Q-learning, in the case where the action value is Q, if the action value Q of the action a t+1 that gives the highest value at time t + 1 is greater than the action value Q of the action a t performed at time t, the action value Q is increased, and if the action value Q of the action a t+1 is smaller than the action value Q of the action a t , the action value Q is decreased. In other words, in Q-learning, in order to make the action value Q of the action a t at time t close to the best action value at time t + 1, the action value function Q(s t , a t ) is updated. As a result, the best action value Q in a certain environment is continuously propagated as the action value Q in the environment before it.

[0086] The trained model generation unit 320 causes the value based on the waveform data included in the learning data to be fitted to the state s t , and causes the value based on the data indicating the parameter included in the learning data to be fitted to the action a t , thereby generating the trained model using Q-learning. Furthermore, regarding the fitting of the value based on the waveform data to the state s t , arbitrary fitting can be performed. For example, the values of the light receiving amount for each position included in the waveform data of the light receiving amount of the first measurement sensor 211 are set to wa1, wa2,..., wan, the values of the light receiving amount for each position included in the waveform data of the light receiving amount of the second measurement sensor 212 are set to wb1, wb2,..., wbn, and predetermined constants are set to ua1, ua2,..., uan, ub1, ub2,..., ubn. At this time, regarding the state s t , st = ua1 x wa1 + ua2 x wa2 +... + uan x wan + ub1 x wb1 + ub2 x wb2 +... + ubn x wbn can be established.

[0087] In addition, regarding the fitting of the value based on the data indicating the parameter to the action a t , as long as the action a t can be calculated based on the action value function Q(s t , a t ) and the state s t and the values of the projection amount adjustment region specification parameter and the interference avoidance parameter can be calculated based on the action a t , arbitrary fitting can be performed. For example, predetermined constants are set to va1, va2, va3, vb4,..., van,..., vb1, vb2, vb3, vb4,..., vbn,.... At this time, regarding the action a t , a t = va1 x LI01 + va2 x CT01 + va3 x ST01 + va4 x ED01 +... + van x SI01 +... + vb1 x LI02 + vb2 x CT01 + vb3 x ST02 + vb4 x ED02 +... + vbn x SI01 +... can be established. Furthermore, in this case, in order to calculate the value of the projection amount adjustment region specification parameter and the interference avoidance parameter for each kind of measurement sensor 210 according to the action a t , arbitrary fitting can be performed. Figure 3The values ​​of the projection amount LI01, LI02, the values ​​of the measurement center position CT01, CT02, the values ​​of the measurement start position ST01, ST02, the values ​​of the measurement end position ED01, ED02, ..., and the values ​​of the sampling interval SI01, SI02, ... need to be calculated, and the constants va1, va2, va3, va4, ..., van, ..., vb1, vb2, vb3, vb4, ..., vbn, ... need to be set.

[0088] Return to Figure 1 The trained model generation unit 320 has: a reward calculation unit 321, which calculates the reward r t+1 Perform calculations; and function update unit 322, which updates the action value function Q(s) t a t Update it.

[0089] The return calculation unit 321 calculates the return r based on the waveform data for each type of measuring sensor 210 contained in the learning data, the values ​​of the projection amount adjustment area specified parameters for each type of measuring sensor 210 determined according to the data representing the parameters, and the values ​​of the interference avoidance parameters. t+1 Calculations are performed. The return calculation unit 321 calculates the return r based on the number of peak waveforms generated in the waveform data. t+1 Calculations are performed. Specifically, the return calculation unit 321 increases the return r when the number of peak waveforms is one. t+1 For example, a reward of +1 can be given; on the other hand, the reward r can be reduced when the number of peak waveforms is 0 or greater than or equal to 2. t+1 For example, giving a reward of -1.

[0090] The function update unit 322 updates the return r calculated by the return calculation unit 321. t+1 For the action value function Q(s) t a t The action value function Q(s) is updated accordingly. t a t ) is a function used to determine the values ​​of the projection amount adjustment region specification parameter and the interference prevention parameter in the new state of the input waveform data. Additionally, the function update unit 322 will represent the action value function Q(s) t a t The data is output as a trained model to the trained model storage DB 400 and stored in the trained model storage section 410.

[0091] The trained model generation unit 320 repeatedly performs a feedback process each time the learning data acquisition unit 310 acquires learning data from the measurement device 200. t+1 The calculation and action value function Q(s)t , a t ). In addition, the trained model generation unit 320 outputs the updated action value function Q(s t , a t ) to the trained model storage DB 400 as a generated trained model each time the action value function Q(s t , a t ) is updated by the update formula shown in the above-described equation 1, and stores it in the trained model storage unit 410.

[0092] The waveform data acquisition unit 510 acquires the waveform data generated by the waveform data generation unit 230. Specifically, the waveform data acquisition unit 510 acquires the waveform data of the light receiving amount for each kind of the measurement sensor 210. For example, the waveform data acquisition unit 510 newly acquires the waveform data containing the values of the light receiving amount for each position of each kind of the measurement sensor 210.

[0093] The parameter calculation unit 520 calculates the parameters with respect to the trained model stored in the trained model storage unit 410, using the waveform data acquired by the waveform data acquisition unit 510. The parameter calculation unit 520 calculates the parameter estimated to be determined so that the peak waveform becomes 1 according to the waveform data acquired in the new state. Specifically, first, the parameter calculation unit 520 acquires the trained model, that is, the action value function Q(s t , a t ) stored in the trained model storage unit 410. Next, the parameter calculation unit 520 calculates the new state s t from the values of the light receiving amount for each position of each kind of the measurement sensor 210 contained in the newly acquired waveform data. Furthermore, the parameter calculation unit 520 calculates the best action a t based on the action value function Q(s t , a t ) and the new state s t , and calculates the values of the projection amount, the measurement center position, the measurement start position, the measurement end position,..., the sampling interval,... for each kind of the measurement sensor 210 based on the action a t .

[0094] The parameter output unit 530 outputs data indicating the parameters calculated by the parameter calculation unit 520 to the measurement apparatus 200. At this time, the parameter setting unit 220 of the measurement apparatus 200 updates the settings in the parameters determined from the data output from the parameter output unit 530.

[0095] Figure 10is a flowchart showing a process of generating a trained model according to the present embodiment.

[0096] Next, the action of generating a trained model by the learning device 300 will be described using the flowchart shown in Figure 10 Fig. 7. First, as shown in Fig. 7, the learning data acquisition section 310 acquires learning data from the measurement device 200 (step S101). For example, the learning data acquisition section 310 acquires the waveform data shown in Fig. 2 for each kind of the measurement sensor 210. Figure 10 Figure 3 The values of the light-receiving amount for each position, wa1, wa2,..., wan, wb1, wb2,..., wbn, the values of the projection amount, LI01, LI02, the values of the measurement center position, CT01, CT02, the values of the measurement start position, ST01, ST02, the values of the measurement end position, ED01, ED02,..., and the values of the sampling interval, SI01, SI02,... included in the waveform data shown in Fig. 2 are calculated.

[0097] After the learning data is acquired, the reward calculation section 321 calculates the reward r t+1 based on the waveform data included in the acquired learning data (step S102). For example, the reward calculation section 321 calculates the number of peak waveforms from the values of the light-receiving amount for each position, wa1, wa2,..., wan, wb1, wb2,..., wbn, for each kind of the measurement sensor 210. Also, the reward calculation section 321 increases the reward r t+1 by assigning a reward of +1 in the case where the number of peak waveforms is 1, and decreases the reward r t+1 by assigning a reward of -1 in the case where the number of peak waveforms is 0 or 2 or more.

[0098] After the reward is calculated, the function update section 322 updates the action value function Q(s t+1 , a t based on the calculated reward r t (step S103). For example, the function update section 322 calculates the state s t from the values of the light-receiving amount, wa1, wa2,..., wan, wb1, wb2,..., wbn, and calculates the action a t from the values of the projection amount, LI01, LI02, the values of the measurement center position, CT01, CT02, the values of the measurement start position, ST01, ST02, the values of the measurement end position, ED01, ED02,..., and the values of the sampling interval, SI01, SI02,.... Also, the function update section 322 updates the action value function Q(s t , a t ​) is updated. After the function is updated, the trained model generation unit 320 outputs a trained model based on the updated action value function Q(s t , a t ) to the trained model storage DB 400 and stores it in the trained model storage unit 410 (step S104), ending the trained model generation processing.

[0099] Figure 11 is a flowchart showing the parameter adjustment processing according to the present embodiment.

[0100] Next, the operation of adjusting the parameters by the parameter adjustment device 500 will be described using the flowchart shown in Figure 11 . First, as shown in Figure 11 , the waveform data acquisition unit 510 acquires waveform data from the measurement device 200 (step S201). For example, the waveform data acquisition unit 510 newly acquires waveform data containing values of the light receiving amount for each position of each kind of the measurement sensor 210.

[0101] After the waveform data is acquired, the parameter calculation unit 520 calculates the parameters by applying the waveform data acquired by the waveform data acquisition unit 510 to the trained model stored in the trained model storage unit 410 (step S202). For example, the parameter calculation unit 520 acquires the action value function Q(s t , a t ) stored in the trained model storage unit 410, and calculates a new state s t from the values of the light receiving amount for each position of each kind of the measurement sensor 210 contained in the newly acquired waveform data. Further, the parameter calculation unit 520 calculates the best action a t based on the action value function Q(s t , a t ) and the new state s t , and calculates the values of the projection amount, the measurement center position, the measurement start position, the measurement end position, …, the sampling interval, … for each kind of the measurement sensor 210 based on the action a t . After the parameters are calculated, the parameter output unit 530 outputs data showing the calculated parameters to the measurement device 200 (step S203), and ends the parameter adjustment processing with the parameters set updated based on the data output to the parameter setting unit 220.

[0102] Figure 12 is a flowchart showing the measurement processing according to the present embodiment.

[0103] Next, the operation of adjusting the parameters by the parameter adjustment device 500 will be described using the flowchart shown in Figure 12The flowchart shown illustrates the operation of measuring the object by the measuring device 200. First, as... Figure 12 As shown, the parameter setting unit 220 sets arbitrary values ​​as parameters of the measuring sensor 210 (step S301). After setting the parameters, the waveform data generation unit 230 generates waveform data of the amount of laser light received by the measuring sensor 210 from the object when the measuring sensor 210 projects laser light onto the object (step S302). At this time, the diagram is omitted, but the waveform data generation unit 230 outputs the generated waveform data to the parameter adjustment device 500.

[0104] After the waveform data is output, the parameter setting unit 220 determines whether data has been acquired. This data represents the parameters that have been adjusted based on the waveform data generated from the parameter adjustment device 500 (step S303). If no data representing the parameters is acquired (step S303; N), the parameter setting unit 220 repeats step S303 until data representing the parameters is acquired. If data representing the parameters is acquired (step S303; Y), the settings are updated based on the parameters determined by the acquired data (step S304). After the parameter settings are updated, the waveform data generation unit 230 generates waveform data of the light received by the measuring sensor 210 by setting the updated parameters (step S305). After the waveform data is generated, the measuring unit 240 measures the object based on the generated waveform data (step S306), and the measuring result display unit 250 displays the measuring result on the measuring result display screen (step S307), ending the measuring process.

[0105] As described above, in the measurement system 100 according to this embodiment, the trained model storage unit 410 stores the trained model generated by the trained model generation unit 320. Furthermore, the parameter calculation unit 520 calculates parameters that enable the measurement sensor 210 to measure the object in the new state, based on the new waveform data acquired by the waveform data acquisition unit 510 and the trained model stored in the trained model storage unit 410. Moreover, the parameter output unit 530 outputs data representing the parameters calculated by the parameter calculation unit 520 to the measurement device 200.

[0106] By setting it in the manner described above, the measuring system 100 appropriately measures the object using the measuring sensor 210 based on the parameters determined by the data output from the parameter adjustment device 500.

[0107] In particular, according to the measurement system 100 according to the present embodiment, the trained model generation unit 320 generates a trained model by machine learning using learning data including waveform data of the light receiving amount of the measurement sensor 210 and data indicating parameters at the time of acquiring the waveform data.

[0108] By being configured as described above, the measurement system 100 can appropriately measure the object using the measurement sensor 210 using the parameters calculated by the trained model generated by the learning device 300.

[0109] Here, in the conventionally known measurement system, the user manually selects and sets the parameters until the object can be appropriately measured based on the waveform data, and it is necessary to analyze the relationship between the waveform data and the parameters by human power.

[0110] In contrast, in the measurement system 100 according to the present embodiment, the learning device 300 automatically acquires learning data including waveform data and data indicating parameters at the time of acquiring the waveform data from the measurement device 200, and automatically generates a trained model by machine learning using the acquired learning data. In addition, in the measurement system 100, the parameter adjustment device 500 automatically acquires new waveform data from the measurement device 200, automatically calculates parameters adjusted by applying the acquired waveform data to the trained model, and outputs the parameters to the measurement device 200, and the parameters automatically adjusted by the measurement device 200 are updated in the setting.

[0111] Therefore, the measurement system 100 can automatically adjust the parameters manually adjusted in the conventionally known measurement system, and can appropriately measure the object by the measurement device 200 using the adjusted parameters, so that the user's work load and work time can be reduced compared to the conventionally known measurement system.

[0112] In addition, according to the measurement system 100 according to the present embodiment, the projection amount adjustment region designation parameter is included in the parameters. Furthermore, the parameter adjustment device 500 of the measurement system 100 adjusts the projection amount adjustment region designation parameter, whereby as shown in Figure 4B , Figure 5C

[0113] ​By being configured in the above-described manner, the measurement system 100 is able to easily measure the object by setting the adjusted projection amount adjustment region designation parameter compared to a measurement system that does not measure the object. In addition, by being configured in the above-described manner, the measurement system 100 is able to automatically designate the projection amount adjustment region, and is able to appropriately measure the object based on the waveform data that determines the peak waveform of the measurement object even if the user of the measurement system does not manually designate the projection amount adjustment region.

[0114] In particular, according to the measurement system 100 related to the present embodiment, the projection amount adjustment region designation parameter includes parameters corresponding to the items of "projection amount", "measurement center position", "measurement start position", and "measurement end position".

[0115] By being configured in the above-described manner, the measurement system 100 is able to reduce the cases in which the light reception amount is insufficient for the measurement of the object compared to a measurement system that is unable to designate the projection amount itself and a measurement system that is unable to designate the region in which the projection amount is adjusted, and is able to easily measure the object.

[0116] Furthermore, in the present embodiment, the projection amount adjustment region designation parameter includes parameters corresponding to the items of "projection amount", "measurement center position", "measurement start position", and "measurement end position", but is not limited thereto as long as the region in which the projection amount is adjusted is designated. For example, a "projection angle" that indicates the angle at which the laser is projected can be included in the projection amount adjustment region designation parameter.

[0117] Furthermore, as in the present embodiment, in order to easily measure the object, it is preferable that the projection amount adjustment region designation parameter be included in the parameters, but as long as the parameter is one that enables the measurement of the object by adjustment, it is not limited thereto. For example, the interference avoidance parameter of the first measurement sensor 211 and the second measurement sensor 212 can be included in the parameters, and on the other hand, the projection amount adjustment region designation parameter can not be included.

[0118] In addition, according to the measurement system 100 related to the present embodiment, the interference avoidance parameter of the first measurement sensor 211 and the second measurement sensor 212 is included in the parameters. Furthermore, the parameter adjustment device 500 of the measurement system 100 adjusts the interference avoidance parameter, and as a result, as shown in Figs. 21 and 22, the light projected by the first measurement sensor 211 is prevented from interfering with the light reception of the second measurement sensor 212, and the light projected by the second measurement sensor 212 is prevented from interfering with the light reception of the first measurement sensor 211. Figure 6B 、 Figure 7B

[0119] ​By being configured in the above-described manner, the measurement system 100 is able to easily perform measurement on the object by setting the adjusted interference avoidance parameters compared to a measurement system that does not measure the object. In addition, by being configured in the above-described manner, the measurement system 100 is able to automatically set the interference avoidance function to be active, and even if the user of the measurement system does not manually set the interference avoidance function to be active, the object is appropriately measured based on the waveform data that determines the peak waveform of the measurement object.

[0120] In particular, according to the measurement system 100 according to the present embodiment, the interference avoidance parameters include parameters corresponding to the item of "sampling interval" for each type of measurement sensor 210.

[0121] By being configured in the above-described manner, the measurement system 100 is able to easily avoid interference of the light projected from one measurement sensor with the light reception of the other measurement sensor, and easily perform measurement on the object, compared to a measurement system that is not able to specify the sampling interval.

[0122] Further, in the present embodiment, the interference avoidance parameters include parameters corresponding to the item of "sampling interval" for each type of measurement sensor 210, but as long as interference of the light projected from one measurement sensor with the light reception of the other measurement sensor is able to be avoided, it is not limited thereto. For example, the interference avoidance parameters can include a "projection angle" indicating the angle of the light projected for each type of measurement sensor 210, a "moving speed" indicating the moving speed at the time of measurement for each type of measurement sensor 210, and the like.

[0123] Further, as in the present embodiment, in order to easily perform measurement on the object, it is preferable that the interference avoidance parameters of the first measurement sensor 211 and the second measurement sensor 212 be included in the parameters, but as long as the parameters are able to be adjusted to perform measurement on the object, it is not limited thereto. For example, only the projection amount adjustment region specification parameters can be included in the parameters, and on the other hand, the interference avoidance parameters of the first measurement sensor 211 and the second measurement sensor 212 can not be included.

[0124] In addition, according to the measurement system 100 according to the present embodiment, the trained model generation unit 320 is able to determine the peak waveform based on the waveform data included in the learning data. In addition, the reward calculation unit 321 increases the evaluation of the learning data for which one peak waveform is determined, and on the other hand, decreases the evaluation of the learning data for which no peak waveform is determined and the learning data for which multiple peak waveforms are determined, and performs calculation of the reward r. t+1

[0125] ​By being configured in the above-described manner, the parameter calculation section 520 of the measurement system 100 is able to easily calculate the parameters at the time of acquiring the waveform data in which one peak waveform is determined when the new waveform data acquired is applied to the trained model generated by the trained model generation section 320 in the new state. As a result, the measurement system 100 is able to easily perform measurement of the object using the measurement sensor 210 compared to a measurement system that does not improve the evaluation of the learning data in which one peak waveform is determined, and a measurement system that does not reduce the evaluation of the learning data in which no peak waveform is determined and the learning data in which multiple peak waveforms are determined.

[0126] Further, as in the present embodiment, in order to easily measure the object, the reward calculation section 321 preferably improves the evaluation of the learning data in which one peak waveform is determined, and reduces the evaluation of the learning data in which no peak waveform is determined and the learning data in which multiple peak waveforms are determined, but is not limited thereto. For example, even if the learning data in which multiple peak waveforms are determined, if the difference between the maximum value of the light receiving amount of the largest peak waveform and the maximum value of the light receiving amount of the other peak waveforms exceeds a predetermined threshold value, the reward calculation section 321 can improve the evaluation of the learning data. In addition, for example, even if the learning data in which multiple peak waveforms are determined, if the maximum value of the light receiving amount of the largest peak waveform becomes a value within a predetermined range, and on the other hand, the maximum value of the light receiving amount of the other peak waveforms becomes a value outside the range, the reward calculation section 321 can improve the evaluation of the learning data.

[0127] In addition, according to the measurement system 100 according to the present embodiment, the parameter calculation section 520 calculates the parameters in which the peak waveform determined from the waveform data acquired in the new state is estimated to be one using the trained model generated by the trained model generation section 320.

[0128] By being configured in the above-described manner, the measurement system 100 is able to easily perform measurement of the object using the measurement sensor 210 compared to a measurement system that does not calculate the parameters in which the peak waveform determined from the waveform data acquired in the new state is estimated to be one using the trained model.

[0129] Further, as in the present embodiment, in order to easily determine the object, the parameter calculation section 520 preferably calculates a parameter in which it is presumed that the peak waveform determined based on the newly acquired waveform data is one, but is not limited thereto as long as a parameter in which the object can be determined is calculated. For example, even if a plurality of peak waveforms determined based on the newly acquired waveform data are determined, the parameter calculation section 520 can calculate a parameter in which it is presumed that a difference between the maximum value of the light receiving amount of the largest peak waveform and the maximum value of the light receiving amount of the other peak waveforms exceeds a predetermined threshold value. Further, for example, even if a plurality of peak waveforms determined based on the newly acquired waveform data are determined, the parameter calculation section 520 can calculate a parameter in which it is presumed that the maximum value of the light receiving amount of the largest peak waveform is a value within a predetermined range, and on the other hand, the maximum value of the light receiving amount of the other peak waveforms is a value outside the range.

[0130] Further, in the present embodiment, the trained model generation section 320 generates the trained model using Q-learning which is one example of an algorithm of reinforcement learning, but is not limited thereto, and can generate the trained model using other algorithms of reinforcement learning. For example, the trained model generation section 320 can generate the trained model using TD-learning. Further, the trained model generation section 320 generates the trained model using an algorithm of reinforcement learning, but is not limited thereto, and for example, can generate the trained model using a known algorithm of teacher learning, teacherless learning. Further, for example, can generate the trained model using a known learning algorithm of deep learning, neural network, genetic programming, functional logic programming, support vector machine, and the like.

[0131] Further, in the present embodiment, the measurement device 200, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 can perform transmission and reception of data via a LAN, but the structure of the data transmission and reception is not limited thereto. For example, the measurement device 200, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 can perform transmission and reception of data via a communication cable connecting each other. Further, for example, the measurement device 200, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 can perform transmission and reception of data via the Internet. In this case, for example, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 can function as a so-called cloud server, and can generate and store the trained model by machine learning based on the learning data acquired from the measurement device 200, and output the parameter adjusted based on the newly acquired waveform data from the measurement device 200.

[0132] Further, in the present embodiment, the learning device 300 has the trained model generating section 320, but is not limited thereto, and the parameter adjustment device 500 can also have the trained model generating section 320. That is, the parameter adjustment device 500 can have the learning data obtaining section 310 and the trained model generating section 320, and the trained model generating process illustrated in FIG. 6 can be executed in the parameter adjustment device 500. Figure 10

[0133] Further, in the present embodiment, the measurement device 200, the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 are provided as separate devices, but are not limited thereto, and can be provided as an integrated device. That is, the measurement device 200 of the measurement system 100 can have not only the measurement sensor 210, the parameter setting section 220, the waveform data generating section 230, the measurement section 240, and the measurement result display section 250, but also the learning data obtaining section 310, the trained model generating section 320, the trained model storage section 410, the waveform data obtaining section 510, the parameter calculation section 520, and the parameter output section 530. In addition, for example, in a case where the learning device 300, the trained model storage DB 400, and the parameter adjustment device 500 are provided as a cloud server, only these devices can be provided as an integrated device. That is, the cloud server can have the learning data obtaining section 310, the trained model generating section 320, the trained model storage section 410, the waveform data obtaining section 510, the parameter calculation section 520, and the parameter output section 530.

[0134] Further, as in the present embodiment, the learning data obtaining section 310 preferably obtains learning data from the measurement device 200, but is not limited thereto as long as the learning data includes waveform data and data indicating parameters used when the waveform data is generated. For example, the measurement device 200 can obtain learning data including waveform data and data indicating parameters of a measurement sensor used when the waveform data is generated from a different measurement device. That is, the learning device 300 can obtain learning data from a measurement device or a measurement system different from the measurement system 100, and generate a trained model by machine learning using the obtained learning data. Thereby, the measurement system 100 can use a trained model generated in advance by another measurement system, and adjust parameters used in the measurement device 200 by the parameter adjustment device 500. In this case, the measurement device 200, in order to appropriately measure an object, preferably obtains new learning data from the measurement device 200, and updates the trained model generated in advance by the learning device 300.

[0135] ​Furthermore, in this embodiment, the measuring device 200 has two types of measuring sensors 210: a first measuring sensor 211 and a second measuring sensor 212. However, it is not limited to this; there may be one type of measuring sensor 210, or there may be three or more types. Furthermore, when there are three or more types of measuring sensors 210, such as... Figure 6A , Figure 6B As shown, multiple measuring sensors can be set up in the same area, such as... Figure 7A , Figure 7B As shown, measurement sensors can also be set in different areas. Furthermore, when there is only one type of measurement sensor 210, the learning device 300 generates a trained model using machine learning with learning data related to the one type of measurement sensor 210, and the parameter adjustment device 500 calculates the parameters of the one type of measurement sensor 210 by applying newly acquired waveform data to the generated trained model. Additionally, when n is set to any integer greater than or equal to 3, when there are n types of measurement sensors 210, the learning device 300 generates a trained model using machine learning with learning data related to n types of measurement sensors 210, and the parameter adjustment device 500 calculates the parameters of the n types of measurement sensors 210 by applying newly acquired waveform data to the generated trained model.

[0136] Furthermore, as in this embodiment, it is preferable that the number of measuring sensors 210 used when measuring an object by the measuring device 200 remains unchanged, but this is not a limitation; the number of measuring sensors 210 used when measuring an object by the measuring device 200 can also be increased or decreased. When the number of measuring sensors 210 is reduced, the measuring device 200 only needs to avoid using the parameters of the unused measuring sensors from the parameters output from the parameter adjustment device 500. On the other hand, when the number of measuring sensors 210 is increased, the measuring device 200 needs to have the learning device 300 acquire learning data related to the newly added measuring sensors to generate a new trained model, and have the parameter adjustment device 500 acquire waveform data newly obtained using the newly added measuring sensors, and output the parameters adjusted using the new trained model.

[0137] Further, in the present embodiment, the learned model is updated each time the learning device 300 newly acquires learning data, and the learning data acquired at the time of updating the learned model is not stored, but the acquired learning data can be stored. For example, the learning device 300 can have a learning data storage section that stores all the learning data acquired by the learning data acquisition section 310 up to this point. In this case, the learning data storage section needs to store, for example, the learning data for each kind of the measurement sensor 210 together with the acquired date and time data.

[0138] Further, the central part that performs the processes of the measurement device 200, the learning device 300, and the parameter adjustment device 500 having the control section 51, the main storage section 52, the external storage section 53, the operation section 54, the display section 55, the transceiver section 56, the internal bus 50, and the like is not limited to a dedicated system, and can be implemented using a general computer system. For example, a computer program for executing the aforementioned actions is distributed by being stored in a recording medium readable by a computer, such as a floppy disk, a DVD-ROM (Read-Only Memory), and the like, and by being installed in a computer, the measurement device 200, the learning device 300, and the parameter adjustment device 500 that perform the aforementioned processes can be constituted. Alternatively, the computer program can be stored in a storage device of a server device on a communication network, and downloaded by a general computer system, whereby the measurement device 200, the learning device 300, and the parameter adjustment device 500 can be constituted.

[0139] Further, in a case where the functions of the measurement system 100 are implemented by sharing of an OS (operating system) and an application program, or in a case where the functions are implemented by cooperative action of the OS and the application program, only the application program part can be stored in a recording medium, a storage device.

[0140] Further, the computer program can be provided by being superimposed on a carrier wave and provided via a communication network. For example, the aforementioned computer program is announced on a bulletin board (BBS, Bulletin Board System) on a communication network, and the aforementioned computer program is provided via a network. Then, the computer program is activated under the control of an OS, and can be executed like other application programs, whereby the aforementioned processes are executed.

[0141] The present application can be implemented in various embodiments and modifications without departing from the broad spirit and scope of the present application. Further, the above-described embodiments are for describing the present application, and do not limit the scope of the present application. That is, the scope of the present application is not the embodiments, but is shown by the claims. Moreover, various modifications implemented within the scope of the claims and the meaning equivalent thereto are to be construed as being within the scope of the present application.

[0142] Explanation of reference signs

[0143] 50 internal bus, 51 control section, 52 main storage section, 53 external storage section, 54 operation section, 55 display section, 56 transceiving section, 59 control program, 100 measurement system, 200 measurement device, 210 measurement sensor, 211 first measurement sensor, 211a first sensor head, 212 second measurement sensor, 212a second sensor head, 220 parameter setting section, 230 waveform data generation section, 240 measurement section, 250 measurement result display section, 300 learning device, 310 learning data acquisition section, 320 trained model generation section, 321 reward calculation section, 322 function update section, 400 trained model storage DB, 410 trained model storage section, 500 parameter adjustment device, 510 waveform data acquisition section, 520 parameter calculation section, 530 parameter output section, 600 workpiece, 601 infiltrated workpiece, 602 scattering workpiece, 603 transparent workpiece, 604 reflection plate.

Claims

1. A parameter adjustment device that adjusts a parameter related to control of laser light projected by a measurement sensor toward an object, the parameter adjustment device having: a parameter calculation section that applies, to waveform data newly acquired in a new state in which laser light is projected by the measurement sensor toward a measurement object included in the object, a trained model generated by machine learning using learning data including waveform data of an amount of light of laser light received by a learning object included in the object from the measurement sensor and data indicating the parameter at the time of acquisition of the waveform data, and calculates the parameter with which the measurement object can be measured using the measurement sensor in the new state; and a parameter output section that outputs data indicating the parameter calculated by the parameter calculation section, the parameter including a projection amount adjustment region specifying parameter that specifies a region in which a projection amount of laser light projected by the measurement sensor is adjusted.

2. The parameter adjustment device according to claim 1, wherein the parameter includes an interference avoidance parameter that avoids interference of reception of laser light projected by one measurement sensor with reception of laser light by the other measurement sensor when a first measurement sensor and a second measurement sensor are used as the measurement sensor.

3. A parameter adjustment device that adjusts a parameter related to control of laser light projected by a measurement sensor toward an object, the parameter adjustment device having: a parameter calculation section that applies, to waveform data newly acquired in a new state in which laser light is projected by the measurement sensor toward a measurement object included in the object, a trained model generated by machine learning using learning data including waveform data of an amount of light of laser light received by a learning object included in the object from the measurement sensor and data indicating the parameter at the time of acquisition of the waveform data, and calculates the parameter with which the measurement object can be measured using the measurement sensor in the new state; and a parameter output section that outputs data indicating the parameter calculated by the parameter calculation section, the parameter including an interference avoidance parameter that avoids interference of reception of laser light projected by one measurement sensor with reception of laser light by the other measurement sensor when a first measurement sensor and a second measurement sensor are used as the measurement sensor.

4. The parameter adjustment device according to any one of claims 1 to 3, further having a trained model generation section that generates the trained model by machine learning using the learning data.

5. The parameter adjustment device according to claim 4, wherein ​ The trained model generation section can perform machine learning so as to increase the evaluation of the learning data in which one peak waveform is determined from the waveform data, and decrease the evaluation of the learning data in which no peak waveform is determined and the learning data in which a plurality of peak waveforms are determined.

6. The parameter adjustment device according to claim 5, wherein The parameter calculation section calculates the parameter in which the peak waveform determined from the waveform data acquired in the new state is one.

7. A learning device having a trained model generation section that generates a trained model by machine learning using learning data including waveform data of a light-receiving amount of a laser light received by a learning object included in an object from which the laser light is received by a measurement sensor and data indicating a parameter at the time of acquisition of the waveform data, before measurement of the learning object is performed with the parameter set in relation to control of the laser light projected by the measurement sensor toward the learning object in a state in which the learning object is measured based on waveform data of a light-receiving amount of the laser light received by a measurement object included in an object to be measured, The parameter includes a projection amount adjustment region designation parameter that designates a region in which the projection amount of the laser light projected by the measurement sensor is adjusted.

8. A learning device having a trained model generation section that generates a trained model by machine learning using learning data including waveform data of a light-receiving amount of a laser light received by a learning object included in an object from which the laser light is received by a measurement sensor and data indicating a parameter at the time of acquisition of the waveform data, before measurement of the learning object is performed with the parameter set in relation to control of the laser light projected by the measurement sensor toward the learning object in a state in which the learning object is measured based on waveform data of a light-receiving amount of the laser light received by a measurement object included in an object to be measured, The parameter includes an interference avoidance parameter that avoids interference of light received by one measurement sensor by laser light projected by the other measurement sensor when a first measurement sensor and a second measurement sensor are used as the measurement sensor.

9. A measurement system having: a trained model generation section that generates a trained model by machine learning using learning data including waveform data of a light-receiving amount of a laser light received by a learning object included in an object from which the laser light is received by a measurement sensor and data indicating a parameter in relation to control of the laser light projected by the measurement sensor toward the learning object at the time of acquisition of the waveform data; a parameter calculation section that calculates the parameter for a new state, with respect to the trained model generated by the trained model generation section, by applying waveform data newly acquired in the new state in which laser light is projected by the measurement sensor toward a measurement target object included in the object; a parameter setting section that sets the parameter calculated by the parameter calculation section to the measurement sensor; and a measurement section that measures the measurement target object based on waveform data acquired using the measurement sensor, by the parameter set by the parameter setting section, the parameter includes a projection amount adjustment region designation parameter that designates a region in which the projection amount of laser light projected by the measurement sensor is adjusted.

10. A measurement system that has: a trained model generation section that generates a trained model by machine learning using learning data including waveform data including the amount of light received by a measurement sensor from a learning target object included in an object, and data representing a parameter related to control of laser light projected by the measurement sensor toward the learning target object at the time of acquisition of the waveform data; a parameter calculation section that calculates the parameter for a new state, with respect to the trained model generated by the trained model generation section, by applying waveform data newly acquired in the new state in which laser light is projected by the measurement sensor toward a measurement target object included in the object; a parameter setting section that sets the parameter calculated by the parameter calculation section to the measurement sensor; and a measurement section that measures the measurement target object based on waveform data acquired using the measurement sensor, by the parameter set by the parameter setting section, the parameter includes an interference avoidance parameter that avoids interference of light projected by one measurement sensor with light reception by the other measurement sensor, when a first measurement sensor and a second measurement sensor are used as the measurement sensor.

11. A parameter adjustment method that adjusts a parameter related to control of laser light projected by a measurement sensor toward an object, the parameter adjustment method performs the following steps: a parameter calculation step that calculates the parameter with which the measurement target object included in the object can be measured using the measurement sensor in a new state, with respect to a trained model generated by machine learning using learning data including waveform data including the amount of light received by the measurement sensor from a learning target object included in the object, and data representing the parameter at the time of acquisition of the waveform data, by applying waveform data newly acquired in the new state; and a parameter output step that outputs data representing the parameter calculated by the parameter calculation step, the parameter includes a projection amount adjustment region designation parameter that designates a region in which the projection amount of laser light projected by the measurement sensor is adjusted. ​ The parameters include a projection amount adjustment region specifying parameter that specifies a region in which the projection amount of laser light projected by the measurement sensor is adjusted.

12. A parameter adjustment method that adjusts parameters related to control of laser light projected by a measurement sensor toward an object, The parameter adjustment method performs the following steps: a parameter calculation step that applies waveform data newly acquired in a new state to a trained model generated through machine learning using learning data including waveform data of the amount of light of laser light received by a learning-use object included in the object and data indicating the parameters at the time of acquisition of the waveform data, and calculates the parameters with which the measurement-use object included in the object can be measured using the measurement sensor in the new state; and a parameter output step that outputs data indicating the parameters calculated by the parameter calculation step, The parameters include an interference avoidance parameter that avoids interference of light received by one measurement sensor with light projected by the other measurement sensor when a first measurement sensor and a second measurement sensor are used as the measurement sensor.

13. A program product that causes a computer that adjusts parameters related to control of laser light projected by a measurement sensor toward an object to function as the following: a parameter calculation section that applies waveform data newly acquired in a new state to a trained model generated through machine learning using learning data including waveform data of the amount of light of laser light received by a learning-use object included in the object and data indicating the parameters at the time of acquisition of the waveform data, and calculates the parameters with which the measurement-use object included in the object can be measured using the measurement sensor in the new state; and a parameter output section that outputs data indicating the parameters calculated by the parameter calculation section, The parameters include a projection amount adjustment region specifying parameter that specifies a region in which the projection amount of laser light projected by the measurement sensor is adjusted.

14. A program product that causes a computer that adjusts parameters related to control of laser light projected by a measurement sensor toward an object to function as the following: a parameter calculation section that applies waveform data newly acquired in a new state to a trained model generated through machine learning using learning data including waveform data of the amount of light of laser light received by a learning-use object included in the object and data indicating the parameters at the time of acquisition of the waveform data, and calculates the parameters with which the measurement-use object included in the object can be measured using the measurement sensor in the new state; and a parameter output section that outputs data indicating the parameters calculated by the parameter calculation section, The parameters include a projection amount adjustment region specifying parameter that specifies a region in which the projection amount of laser light projected by the measurement sensor is adjusted. The parameters include an interference avoidance parameter that, when the first assay sensor and the second assay sensor are used as the assay sensor, avoids interference of light projected from one assay sensor with light reception of the other assay sensor. The parameters include an interference avoidance parameter that, when the first assay sensor and the second assay sensor are used as the assay sensor, avoids interference of light projected from one assay sensor with light reception of the other assay sensor.

Images