Torque fastening place detection system

Abstract

To provide a torque fastening place detection system capable of detecting the fastening position of a component to be fastened without being affected by a surrounding environment.SOLUTION: A fastening tool (10) includes a 3D scanner (13) that generates 3D scan data (D1) representing a shape of a workpiece (51) by means of data of a distance to each point on the surface of the workpiece (51) which is a product to be worked. A torque detection unit (11) outputs a torque fastening signal (S1) when the torque of torque fastening of the fastening tool (10) reaches a prescribed value. A determination unit (32) identifies a torque fastening place on the workpiece (51) by comparing the 3D scan data (D1) generated when the torque fastening signal (S1) is output with correct answer 3D shape data (C1) presenting a correct shape of the workpiece (51).SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a torque fastening point detection system using a fastening tool such as a torque driver or a torque wrench. [Background technology]

[0002] In the assembly of many products, screw tightening is used as a fastening method. Some products require torque management for screw tightening, and torque management for screw tightening is one of the important control items for product quality. Screw tightening torque is inappropriate if it is too strong or too weak, and torque is generally managed by using dedicated fastening tools such as torque drivers and torque wrenches.

[0003] In actual screw tightening, for example, when attaching a part to a product, it is common to first provisionally tighten all the screws with a relatively weak force, and then use a fastening tool to tighten them to the torque. However, with this method, it is not possible to visually distinguish between the state of provisionally tightened screws and the state of torque-tightened screws, so even if some screws are forgotten to be torque-tightened, all the screws must be torque-tightened again.

[0004] In order to ensure that torque tightening is performed correctly in product quality control, a mechanism to prevent or detect forgetting to torque tighten (sometimes called "poka-yoke") is required. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2020-175488 A Summary of the Invention [Problem to be solved by the invention]

[0006] A widely used mechanism is to generate a signal in the fastening tool when the screw tightening torque reaches a specified value, and to prevent forgetting to tighten the screw by checking the signal. However, it is technically difficult to distinguish between screws that have been torqued and those that have not.

[0007] For example, if the position of a screw that has been torqued is detected and recorded, it is possible to distinguish between screws that have been torqued and screws that have not been torqued. Patent Document 1 proposes a method of registering an image of the vicinity of a screw tightening point taken in advance as basic information, and comparing an image of the vicinity of the screw tightening point newly taken with a camera attached to a screwdriver with the image of the basic information to determine whether the newly taken image of the screw tightening point is a specified screw tightening point. However, the technology of Patent Document 1 requires that an image of the reference information be taken in advance, and, as with general image recognition, a constant work environment is required to prevent differences in images due to differences in brightness around the screw tightening point or differences in how the light hits it.

[0008] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a torque fastening point detection system that is capable of detecting the fastening position of a fastening part without being affected by the surrounding environment. [Means for solving the problem]

[0009] The torque tightening point detection system of the present disclosure comprises a fastening tool that performs torque tightening of fastening parts, a torque detection unit that detects the torque of the torque tightening and outputs a torque tightening signal when the torque reaches a specified value, a 3D scanner mounted on the fastening tool and generating 3D scan data representing the shape of the workpiece based on data of the distance to each point on the surface of the workpiece, which is the product to be worked on, a correct data memory unit that stores correct 3D shape data representing the correct shape of the workpiece, and a determiner that identifies torque tightening points on the workpiece by comparing the 3D scan data generated when the torque tightening signal is output with the correct 3D shape data. Effect of the Invention

[0010] According to the present disclosure, it is possible to detect the tightening position of a fastening part while suppressing the influence of the surrounding environment. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram showing a configuration of a torque fastening point detection system according to a first embodiment. [Diagram 2] FIG. 1 is a diagram showing a configuration of a torque fastening point detection system according to a first embodiment. [Diagram 3] FIG. 2 is a diagram showing an example of a scanning range of the 3D scanner according to the first embodiment. [Figure 4] FIG. 2 is a diagram showing an example of 3D scan data according to the first embodiment. [Diagram 5] FIG. 4 is a diagram showing an operation flow of a determiner according to the first embodiment. [Figure 6] FIG. 11 is a diagram showing a configuration of a torque fastening location detection system according to a second embodiment. [Figure 7] FIG. 11 is a diagram showing a configuration of a torque fastening location detection system according to a second embodiment. [Figure 8] FIG. 11 is a diagram showing an example of a scanning range of a 3D scanner according to the second embodiment. [Figure 9] FIG. 11 is a diagram showing an example of 3D scan data according to the second embodiment. [Figure 10] FIG. 11 is a diagram showing an operation flow of a determiner according to the second embodiment. [Figure 11] FIG. 11 is a diagram showing a configuration of a torque fastening location detection system according to a third embodiment. [Figure 12] FIG. 11 is a diagram showing a configuration of a torque fastening location detection system according to a third embodiment. [Figure 13] FIG. 11 is a diagram showing an operation flow of a determiner according to the third embodiment. [Figure 14] FIG. 13 is a diagram illustrating a configuration of a learning device according to a fourth embodiment. [Figure 15]FIG. 13 is a diagram showing a configuration of a neural network used in a learning device according to a fourth embodiment. [Figure 16] FIG. 13 is a diagram showing an operation flow of a learning device according to embodiment 4. [Figure 17] FIG. 13 is a diagram illustrating a configuration of an inference device according to a fourth embodiment. [Figure 18] FIG. 13 is a diagram showing an operational flow of the inference device according to the fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] <Embodiment 1> Figures 1 and 2 are diagrams showing the configuration of a torque tightening point detection system according to embodiment 1. As shown in Figure 1, the torque tightening point detection system includes a fastening tool 10, a signal receiver 20, and a PC (Personal Computer) 30. Figure 2 shows a functional block diagram of the fastening tool 10 and the PC 30.

[0013] The fastening tool 10 is a tool such as a torque driver or a torque wrench that tightens screws in a process of assembling a product. The fastening parts tightened by the fastening tool 10 may be bolts, nuts, etc. in addition to screws. In this specification, they are collectively referred to as "screws." In FIG. 1, the fastening tool 10 has a shape of a general electric screwdriver, but the power for the fastening tool 10 to tighten the screws may be any type, such as manual or electric, and the shape of the fastening tool 10 may also be any type.

[0014] The fastening tool 10 includes a torque detection unit 11, a controller 12, a 3D scanner 13 and a signal transmitter 14.

[0015] The torque detection unit 11 detects the value of the screw tightening torque, and when it detects that the screw tightening torque has reached a specified value, outputs a torque tightening signal S1 to the controller 12. The fastening tool 10 may have a function of stopping the application of torque when the torque detection unit 11 outputs the torque tightening signal S1 (such as by idling in the case of a manual type or by stopping operation in the case of an electric type).

[0016] The 3D scanner 13 is a shape detection means for detecting the shape of the workpiece, which is the product to be worked on, in the vicinity of the torque tightening point. In the example of Fig. 1, the 3D scanner 13 detects the shape of the workpiece 51 in the vicinity of the torque tightening point, i.e., the shape of the workpiece 51 including the screw 53 to be torque tightened and the part 52 fixed to the workpiece 51 by the screw 53, and generates 3D scan data D1 representing that shape.

[0017] Here, the 3D scanner 13 is a device that has a function of measuring the distance to the measurement object by detecting the reflected light when light or laser is applied to the measurement object, and obtains shape information of the measurement object by scanning and measuring the distance to each measurement point on the surface of the measurement object one by one. Therefore, the 3D scan data D1 obtained by the 3D scanner 13 is a collection of data on the distance to each point on the surface of the measurement object, and is two-dimensional information in a plane in which distance information for each measurement point is arranged. If the workpiece 51, part 52, and screw 53 shown in FIG. 1 have a shape as shown in FIG. 3, and the scan range of the 3D scanner 13 is within the dotted frame in FIG. 3, it is assumed that the obtained 3D scan data D1 will be represented three-dimensionally as data as shown in FIG. 4 (the intersections of the straight lines in FIG. 4 are the measurement points).

[0018] When the controller 12 receives the torque tightening signal S1 from the torque detection unit 11, the controller 12 causes the 3D scanner 13 to detect the shape of the workpiece 51 near the torque tightening point, and acquires 3D scan data D1, which is the detection result, from the 3D scanner 13. Furthermore, the controller 12 uses the signal transmitter 14 to transmit the 3D scan data D1 to the signal receiver 20.

[0019] The signal receiver 20 receives the 3D scan data D1 transmitted from the fastening tool 10 and inputs the 3D scan data D1 to the PC 30.

[0020] The PC 30 includes a correct answer data storage unit 31 and a determiner 32 .

[0021] The correct data storage unit 31 is a storage medium in which correct 3D shape data C1, which is 3D data representing the correct shape of the product (workpiece) to be worked on, is stored in advance. The correct 3D shape data C1 includes at least data on the torque tightening points of the product. In this embodiment, the correct 3D shape data C1 is created for each torque tightening point, and each correct 3D shape data C1 is assigned an individual number (identification number) indicating the corresponding torque tightening point.

[0022] The correct 3D shape data C1 can be obtained, for example, by measuring the shape of the actual product with a 3D scanner (which may be the 3D scanner 13 of the fastening tool 10).

[0023] Alternatively, the correct 3D shape data C1 may be created from 3D CAD data, which is design data created when the product is designed. The 3D CAD data is basically composed of a combination of surface data. In the coordinate system of the 3D CAD data, the distance from the position coordinates of the 3D scanner 13 assumed when the fastening tool 10 is torqued to the coordinates of each point on the surface of the product is calculated, and the calculated distance data of each point is converted into two-dimensional information in which the distance information of each point is arranged in a plane, and the correct 3D shape data is obtained. By constructing a calculation mechanism that automatically generates the correct 3D shape data C1 from the 3D CAD data, it is possible to obtain the correct 3D shape data C1 without performing a 3D scan of the actual product. This simplifies the work of preparing the correct 3D shape data C1.

[0024] The determiner 32 includes a determination processing unit 32a that collates the 3D scan data D1 acquired from the fastening tool 10 with the correct 3D shape data C1 stored in the correct data storage unit 31 and determines whether the two match or not, thereby identifying the torque tightening location by the fastening tool 10. The determiner 32 may be realized by the processor of the PC 30 executing software (program) stored in a memory, or may be realized by a processing circuit as hardware.

[0025] 5 is a flowchart showing the operation flow of the determiner 32. The operation of the determiner 32 will be described below with reference to the flowchart of FIG.

[0026] When the determiner 32 acquires the 3D scan data D1 transmitted from the fastening tool 10 (step ST1), the determination process starts. First, the determiner 32 sets the number of the first torque tightening point as a candidate for the torque tightening point (step ST2).

[0027] The judgment processing unit 32a reads the correct 3D shape data C1 of the torque tightening point corresponding to the set number from the correct data storage unit 31 (step ST3), and judges whether the correct 3D shape data C1 matches the 3D scan data D1 acquired in step ST1 (step ST4). If they do not match, the number of the next torque tightening point is set (step ST5), and the process returns to step ST3.

[0028] If the correct 3D shape data C1 read from the correct data storage unit 31 matches the 3D scan data D1 acquired in step ST1, the judgment processing unit 32a outputs the number of that location as torque tightening location information T1 (step ST6). This mechanism makes it possible to identify the location where torque tightening has been completed almost simultaneously with the torque tightening work. In addition, the torque tightening location information T1 is recorded and managed in the PC 30. This makes it possible to detect torque tightening work errors.

[0029] Generally, image data obtained from a camera is based on color and brightness information, and is therefore significantly affected by the brightness of the surroundings of the object to be measured and the amount of light hitting it. In contrast, the 3D scan data D1 obtained from the 3D scanner 13 is distance information, and is less affected by the brightness of the surroundings of the object to be measured and the amount of light hitting it. Therefore, the torque tightening point detection system according to this embodiment makes it possible to detect the tightening position of a fastening part while suppressing the influence of the surrounding environment.

[0030] In this embodiment, the determiner 32 is a device separate from the fastening tool 10, but the determiner 32 may be built into the fastening tool 10. In that case, the fastening tool 10 outputs the torque tightening location information T1 and transmits it to the PC 30.

[0031] <Embodiment 2> Figures 6 and 7 are diagrams showing the configuration of a torque fastening point detection system according to embodiment 2. In these figures, components corresponding to those shown in Figures 1 and 2 are given the same reference numerals. Figure 7 shows a functional block diagram of the fastening tool 10 and the PC 30.

[0032] In the second embodiment, a plurality of 3D scanners 13 are provided in the fastening tool 10. Here, an example is shown in which two scanners, a first 3D scanner 13a and a second 3D scanner 13b, are provided in the fastening tool 10 as the plurality of 3D scanners 13.

[0033] When the controller 12 receives the torque tightening signal S1 from the torque detection unit 11, the controller 12 causes the first 3D scanner 13a and the second 3D scanner 13b to detect the shape of the workpiece 51 near the torque tightening point, and acquires the first 3D scan data D1a output by the first 3D scanner 13a and the second 3D scan data D1b output by the second 3D scanner 13b. Furthermore, the controller 12 transmits the first 3D scan data D1a and the second 3D scan data D1b to the signal receiver 20 using the signal transmitter 14.

[0034] The first 3D scan data D1a and the second 3D scan data D1b are input to the PC 30 via the signal receiver 20. The determiner 32 of the PC 30 includes a synthesis processor 32b that synthesizes the first 3D scan data D1a and the second 3D scan data D1b to convert them into one 3D scan data D1.

[0035] For example, as shown in FIG. 8, if the scanning range of the first 3D scanner 13a for the workpiece 51, part 52, and screw 53 partially overlaps with the scanning range of the second 3D scanner 13b, the second 3D scanner 13b deletes the overlapping portions from the first 3D scan data D1a and the second 3D scan data D1b, as shown in FIG. 9, and then combines the first 3D scan data D1a and the second 3D scan data D1b to synthesize them into a single 3D scan data D1.

[0036] Generally, since the main body of the fastening tool 10, such as a torque driver or a torque wrench, is often cylindrical, it is assumed that the mounting position of the 3D scanner 13 is the side of the cylinder. In this case, a part of the scanning range of the 3D scanner 13 is blocked by the main body of the fastening tool 10, and it is difficult to obtain 3D scan data D1 of the entire workpiece with one 3D scanner 13. For example, as in this embodiment, the first 3D scanner 13a and the second 3D scanner 13b are installed so as to sandwich the cylindrical main body of the fastening tool 10, and the synthesis processing unit 32b synthesizes the first 3D scan data D1a and the second 3D scan data D1b to generate 3D scan data D1, thereby making it possible to obtain 3D scan data D1 of the entire workpiece.

[0037] Fig. 10 shows a flowchart of the operation flow of the decision unit 32 in the embodiment 2. In the flowchart of Fig. 10, step ST1 of the flowchart of Fig. 5 is replaced with steps ST11 to ST13.

[0038] In step ST11, the determiner 32 acquires a plurality of 3D scan data (e.g., the first 3D scan data D1a and the second 3D scan data D1b) acquired by a plurality of 3D scanners 13 (e.g., the first 3D scanner 13a and the second 3D scanner 13b). In step ST12, the synthesis processor 32b deletes unnecessary parts (e.g., overlapping parts) from the plurality of 3D scan data. In step ST13, the synthesis processor 32b combines the plurality of 3D scan data from which the unnecessary parts have been deleted to generate one 3D scan data. The subsequent steps are the same as those in FIG. 5.

[0039] <Embodiment 3> Figures 11 and 12 are diagrams showing the configuration of a torque fastening point detection system according to embodiment 3. In these figures, the same reference numerals are used for components corresponding to those shown in Figures 1 and 2. Figure 12 shows a functional block diagram of the fastening tool 10 and the PC 30.

[0040] In the first and second embodiments, it is assumed that the fastening tool 10 is used in a fixed orientation, but the fastening tool 10 may be used in various orientations depending on the situation. Therefore, as the 3D scan data D1 of the vicinity of the torque tightening point, there is a possibility that the 3D scan data D1 rotated at any angle in the range of 0 to 360° around the head of the screw to be torque tightened may be obtained.

[0041] In this case, the determiner 32 rotates the 3D scan data D1 stepwise in the range of 0 to 360 degrees, and compares the 3D scan data D1 rotated at each angle with the correct 3D shape data C1, making it possible to determine the torque tightening location. However, this method requires multiple rotation and determination processes, which increases the software processing load required to realize the determiner 32.

[0042] In the third embodiment, a direction sensor 15 such as a gyro sensor is provided in the fastening tool 10 to detect the direction of the fastening tool 10 (for example, the direction directly above the fastening tool 10 in a plane perpendicular to the direction of the screw to be fastened). Information on the direction of the fastening tool 10 detected by the direction sensor 15 is input to the controller 12 as direction data D2, and the controller 12 transmits the direction data D2 to the PC 30 together with the 3D scan data D1.

[0043] Meanwhile, the correct data memory unit 31 of the PC 30 stores, in addition to the correct 3D shape data C1, correct direction data C2, which is information on the orientation of the correct 3D shape data C1 (for example, the upward direction in a plane horizontal to the workpiece surface (perpendicular to the screw direction)).

[0044] The determiner 32 also includes a rotation processor 32c that performs rotation processing of the 3D scan data D1. The determiner 32 calculates the angle difference between the direction data D2 and the correct direction data C2, and converts (rotates) the 3D scan data D1 according to the angle difference. The determination processor 32a compares the rotated 3D scan data D1 with the correct 3D shape data C1.

[0045] Fig. 13 shows a flowchart of the operation flow of the decision unit 32 in the embodiment 3. In the flowchart of Fig. 13, step ST1 of the flowchart of Fig. 5 is replaced with step ST21, and step ST3 is replaced with steps ST22 to ST24.

[0046] When the determiner 32 acquires the 3D scan data D1 and the direction data D2 transmitted from the fastening tool 10 (step ST21), the determination process starts. First, the determiner 32 sets the number of the first torque tightening point as a candidate for the torque tightening point (step ST2).

[0047] The determiner 32 reads the correct 3D shape data C1 and the correct direction data C2 corresponding to the set number from the correct data storage unit 31 (step ST22). At this time, the correct 3D shape data C1 is sent to the determination processing unit 32a, and the correct direction data C2 is sent to the rotation processing unit 32c.

[0048] The rotation processing unit 32c calculates a rotation angle from the difference between the direction data D2 and the correct direction data C2 (step ST23), and performs a rotation process to rotate the 3D scan data D1 by the calculated rotation angle (step ST24).

[0049] In step ST4, the judgment processing unit 32a judges whether or not the correct 3D shape data C1 read in step ST22 matches the 3D scan data D1 rotated in step ST24 (step ST4). If they do not match, the number of the next torque tightening point is set (step ST5), and the process returns to step ST22.

[0050] If the correct 3D shape data C1 read in step ST22 matches the 3D scan data D1 rotated in step ST24, the determination processing unit 32a outputs the number of that location as torque tightening location information T1 (step ST6).

[0051] According to the torque fastening point detection system of the third embodiment, the rotation angle in the rotation process of the 3D scan data D1 is calculated based on the correct direction data C2, thereby reducing the number of times the rotation process and the judgment process are executed. This reduces the load on the software processing, and is expected to prevent erroneous judgment by reducing unnecessary judgment processes.

[0052] <Fourth embodiment> In the first to third embodiments, it is assumed that the determination processing unit 32a uses general-purpose pattern recognition software to identify the torque-tightened portion. However, an inference device using artificial intelligence (AI) may be used as a mechanism for identifying the torque-tightened portion. In the fourth embodiment, the determination processing unit 32a is realized by an inference device.

[0053] However, in order to use the inference device, it is necessary to use a learning device to create a trained model that learns the previously generated 3D scan data D1 and the corresponding correct 3D shape data C1. When acquiring the correct 3D shape data C1 from an actual product, by using as much correct 3D shape data C1 as possible for training, it becomes possible to identify the screw fastening locations with high accuracy and with little influence of variation.

[0054] [Learning Phase] 14 is a configuration diagram of a learning device 100 that creates a trained model to be used in an inference device serving as a judgment processing unit 32a. The learning device 100 includes a data acquisition unit 101, a model generation unit 102, and a trained model storage unit 300. The learning device 100 is realized by the processor of the PC 30 executing software (programs) stored in memory.

[0055] The data acquisition unit 101 acquires, as learning data, 3D scan data D1 (hereinafter referred to as "input data") of the torque fastening points obtained from the 3D scanner 13 attached to the fastening tool 10 and correct 3D shape data C1 (hereinafter referred to as "correct data") of the correct torque fastening points acquired in advance.

[0056] The model generation unit 102 learns the judgment result of the torque fastening point information T1 (hereinafter referred to as the "judgment result") based on the learning data created based on a combination of the input data and the supervised data acquired in advance, which is output from the data acquisition unit 101. That is, the model generation unit 102 generates a trained model M1 that infers the optimal judgment result from the input data and the supervised data of the torque fastening point detection system. Here, the learning data is data in which the input data and the supervised data are associated with each other. Also, the judgment result is data having two values, match / mismatch.

[0057] The learning algorithm used by the model generation unit 102 may be a known algorithm such as supervised learning, unsupervised learning, reinforcement learning, etc. Here, as an example, a case where a neural network is applied will be described.

[0058] The model generation unit 102 learns the judgment results by so-called supervised learning, for example, according to a neural network model. Here, supervised learning refers to a method of providing a learning device with a set of input and result (label) data, learning the features of the learning data, and inferring the result from the input.

[0059] A neural network is composed of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer, or two or more layers.

[0060] For example, in a three-layered neural network as shown in Figure 15, when multiple inputs are input to the input layer (X1-X3), the values ​​are multiplied by weight W1 (w11-w16) and input to the intermediate layer (Y1-Y2), and the result is further multiplied by weight W2 (w21-w26) and output from the output layer (Z1-Z3). This output result changes depending on the values ​​of weights W1 and W2.

[0061] In this embodiment, the neural network learns the judgment result by so-called supervised learning according to learning data created based on a combination of input data and supervised data acquired by the data acquisition unit 101. That is, the neural network learns by inputting input data to the input layer and adjusting the weights W1 and W2 so that the result output from the output layer approaches the supervised data.

[0062] The model generation unit 102 executes the above-described learning to generate and output a trained model M1.

[0063] The trained model M1 output by the model generation unit 102 is stored in the trained model storage unit 300. The trained model storage unit 300 is a storage medium for storing the trained model M1.

[0064] Next, the learning process performed by the learning device 100 will be described with reference to Fig. 16. Fig. 16 is a flowchart showing the learning process performed by the learning device 100.

[0065] In step ST201, the data acquisition unit 101 acquires input data and correct answer data. Here, the input data and the correct answer data are acquired simultaneously, but as long as the input data and the correct answer data can be input in association with each other, the input data and the correct answer data may be acquired at different times.

[0066] Next, in step ST202, the model generation unit 102 learns the judgment results by so-called supervised learning according to the learning data created based on a combination of the input data and the correct answer data acquired by the data acquisition unit 101, and generates a learned model M1.

[0067] Then, in step ST203, the trained model storage unit 300 stores the trained model M1 generated by the model generation unit 102.

[0068] In this embodiment, the learning device 100 is constructed on the PC 30 and learns the judgment result output from the judgment processing unit 32a on the PC 30. However, the learning device 100 may be a device separate from the PC 30, for example, connected to the PC 30 via a network. Alternatively, the learning device 100 may exist on a cloud server.

[0069] [Utilization phase] 17 is a configuration diagram of an inference device 200 which serves as the judgment processing unit 32a of the torque fastening point detection system. It includes a data acquisition unit 201 and an inference unit 202. The inference device 200 is realized by the processor of the PC 30 executing software (program) stored in the memory.

[0070] The data acquisition unit 201 acquires 3D scan data D1 (hereinafter referred to as “input data”) of the torque fastening point obtained from the 3D scanner 13 attached to the fastening tool 10.

[0071] The inference unit 202 infers the judgment result of the torque tightening point corresponding to the input data (i.e., the torque tightening point information T1) by using the trained model M1 that the learning device 100 has created in advance and stored in the trained model storage unit 300. That is, the inference unit 202 can output the judgment result inferred from the input data by inputting the input data acquired by the data acquisition unit 201 to the trained model M1.

[0072] Next, the process for obtaining a judgment result using inference device 200 will be described with reference to FIG.

[0073] In step ST201, the data acquisition section 201 acquires input data.

[0074] Next, in step ST202, the inference unit 202 inputs the input data to the trained model M1 stored in the trained model storage unit 300, and obtains a judgment result.

[0075] Then, in step ST203, the inference unit 202 outputs the determination result (torque tightening location information T1) obtained by the trained model M1.

[0076] In this embodiment, inference device 200 infers a judgment result using trained model M1 created by learning device 100 constructed on PC 30, but trained model M1 acquired from an external source, such as an inference device 200 separate from PC 30, may also be used. In addition, inference device 200 may be a device separate from PC 30, for example, connected to PC 30 via a network. Alternatively, inference device 200 may exist on a cloud server.

[0077] In this embodiment, a case where supervised learning is applied to the learning algorithm used by the model generation unit 102 has been described, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to supervised learning.

[0078] The model generation unit 102 may learn the judgment result according to the learning data acquired from the multiple judgment processing units 32a. In this case, the model generation unit 102 may acquire the learning data from the multiple judgment processing units 32a used in the same area, or may collect the learning data from the multiple judgment processing units 32a operating independently in different areas. The judgment processing units 32a from which the learning data is to be collected may be added or removed midway. Furthermore, the learning device 100 that has learned the judgment result for a certain judgment processing unit 32a may be applied to another judgment processing unit 32a, and the judgment result for the other judgment processing unit 32a may be re-learned to update the learned model M1.

[0079] In addition, the learning algorithm used in the model generation unit 102 may be deep learning, which learns to extract the features themselves, or machine learning may be performed according to any other method, such as genetic programming, functional logic programming, or support vector machines.

[0080] It should be noted that the embodiments can be freely combined, and each embodiment can be modified or omitted as appropriate.

[0081] <Additional Notes> Various aspects of the present disclosure are summarized below as appendices.

[0082] (Appendix 1) A fastening tool for torque tightening of fastening parts; A torque detection unit that detects the torque of the torque tightening and outputs a torque tightening signal when the torque reaches a specified value; A 3D scanner is mounted on the fastening tool and generates 3D scan data representing the shape of the workpiece based on data of distances to each point on the surface of the workpiece, which is the product to be worked on; A correct data storage unit that stores correct 3D shape data representing a correct shape of the workpiece; A determiner that identifies a torque tightening location on the workpiece by comparing the 3D scan data generated when the torque tightening signal is output with the correct 3D shape data; A torque fastening point detection system.

[0083] (Appendix 2) The correct 3D shape data is created from 3D CAD data which is design data of the workpiece. 2. A torque fastening point detection system as described in appendix 1.

[0084] (Appendix 3) The fastening tool is equipped with a plurality of the 3D scanners, The determiner compares data obtained by combining the plurality of 3D scan data generated by the plurality of 3D scanners with the correct 3D shape data. 2. A torque fastening point detection system according to claim 1.

[0085] (Appendix 4) The fastening tool is equipped with a direction sensor that detects the orientation of the fastening tool, The determiner compares the rotated 3D scan data based on the orientation of the fastening tool detected by the direction sensor with the correct 3D shape data. 4. A torque fastening point detection system according to any one of claims 1 to 3.

[0086] (Appendix 5) The determiner includes an inference device that infers the torque tightening location on the workpiece from the 3D scan data generated when the torque tightening signal is output, using a learned model that has learned the 3D scan data generated in advance and the corresponding correct 3D shape data. 5. A torque fastening point detection system according to any one of claims 1 to 4.

[0087] (Appendix 6) The method further includes a learning device that creates the trained model by learning the 3D scan data generated in advance and the corresponding correct 3D shape data. 6. A torque fastening point detection system as described in appendix 5. [Explanation of symbols]

[0088] 10 Fastening tool, 11 Torque detection unit, 12 Controller, 13 3D scanner, 13a First 3D scanner, 13b Second 3D scanner, 14 Signal transmitter, 15 Direction sensor, 20 Signal receiver, 30 PC, 31 Correct data storage unit, 32 Judge, 32a Judgment processing unit, 32b Synthesis processing unit, 32c Rotation processing unit, S1 Torque fastening signal, D1 3D scan data, D1a First 3D scan data, D1b Second 3D scan data, D2 Direction data, C1 Correct 3D shape data, C2 Correct direction data, T1 Torque fastening location information, 51 Workpiece, 52 Part, 53 Screw, 100 Learning device, 101 Data acquisition unit, 102 Model generation unit, 200 Inference device, 201 Data acquisition unit, 202 Inference unit, 300 Learned model storage unit, M1 Trained model.

Claims

1. A fastening tool for torque tightening of fastening parts; A torque detection unit that detects the torque of the torque tightening and outputs a torque tightening signal when the torque reaches a specified value; A 3D scanner is mounted on the fastening tool and generates 3D scan data representing the shape of the workpiece based on data of distances to each point on the surface of the workpiece, which is a product to be worked on; A correct data storage unit that stores correct 3D shape data representing a correct shape of the workpiece; A determiner that identifies a torque tightening location on the workpiece by comparing the 3D scan data generated when the torque tightening signal is output with the correct 3D shape data; A torque fastening point detection system.

2. The correct 3D shape data is created from 3D CAD data which is design data of the workpiece.

2. The torque fastening point detection system according to claim 1.

3. The fastening tool is equipped with a plurality of the 3D scanners, The determiner compares data obtained by combining the plurality of 3D scan data generated by the plurality of 3D scanners with the correct 3D shape data. The torque fastening point detection system according to claim 1 or 2.

4. The fastening tool is equipped with a direction sensor that detects the orientation of the fastening tool, The determiner compares the rotationally processed 3D scan data with the correct 3D shape data based on the orientation of the fastening tool detected by the direction sensor. The torque fastening point detection system according to claim 1 or 2.

5. The determiner includes an inference device that infers the torque tightening location on the workpiece from the 3D scan data generated when the torque tightening signal is output, using a learned model that has learned the 3D scan data generated in advance and the corresponding correct 3D shape data. The torque fastening point detection system according to claim 1 or 2.

6. The method further includes a learning device that creates the learned model by learning the 3D scan data generated in advance and the corresponding correct 3D shape data. The torque fastening point detection system according to claim 5.

Images