Axial force estimation device

Abstract

To provide an axial force estimation device which is configured to estimate an axial force of a screw when tightened according to pre-learned determination criteria, and is applicable to small screws.SOLUTION: An axial force estimation device 1 is provided, comprising data acquisition means 11 for acquiring measurement data on an electric tool, and axial force estimation means 12 configured to estimate an axial force when a screw is tightened using the electric tool according to pre-learned determination criteria on the basis of the measurement data acquired by the data acquisition means 11.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an axial force estimation device, a screw fastening system, an axial force estimation method, a program, and a recording medium. [Background technology]

[0002] As a technique for measuring the axial force during screw tightening, a method has been proposed that uses a strain gauge placed inside a hole provided in the axial direction from the head of the screw (Patent Documents 1 and 2). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2012-202688 A [Patent Document 2] Japanese Patent Application Publication No. 6-347349 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the techniques described in Patent Documents 1 and 2 require holes to be provided in the screw for arranging the strain gauge, making it difficult to apply them to small screws. Furthermore, there has been no technique to estimate the axial force during screw tightening using criteria obtained by machine learning or the like.

[0005] Therefore, an object of the present invention is to provide an axial force estimation device, a screw fastening system, an axial force estimation method, a program, and a recording medium that are applicable to small screws and that estimate the axial force during screw fastening based on pre-learned criteria. [Means for solving the problem]

[0006] In order to achieve the above object, the axial force estimation device of the present invention comprises: A data acquisition means for acquiring actual measurement data relating to the power tool; an axial force estimation means for estimating an axial force at the time of fastening the screw by the electric tool according to a previously learned judgment criterion based on the actual measurement data acquired by the data acquisition means; The present invention is characterized by comprising:

[0007] The screw fastening system of the present invention comprises: The present invention includes an electric power tool and an axial force estimation device, The axial force estimation device is characterized in that it is the axial force estimation device of the present invention.

[0008] The axial force estimation method of the present invention includes the steps of: A data acquisition step of acquiring actual measurement data relating to the power tool; an axial force estimation step of estimating an axial force at the time of fastening the screw by the electric tool based on the actual measurement data acquired in the data acquisition step and a previously learned judgment criterion; The present invention is characterized by comprising:

[0009] The program of the present invention is characterized by causing a computer to execute the axial force estimation method of the present invention.

[0010] The recording medium of the present invention is a computer-readable recording medium having the program of the present invention recorded thereon. Effect of the Invention

[0011] According to the present invention, it is possible to provide an axial force estimation device, a screw fastening system, an axial force estimation method, a program, and a recording medium that can be applied to small screws and estimate the axial force during screw fastening based on pre-learned judgment criteria. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1(A) is a block diagram showing an example of the configuration of an axial force estimation device of the present invention, and FIG. 1(B) is a flowchart showing an example of an axial force estimation method of the present invention. [Diagram 2] FIG. 2 is a diagram for explaining an example of measurement of actual measurement data in the present invention. [Diagram 3] FIG. 3 is another diagram for explaining an example of measurement of actual measurement data in the present invention. [Figure 4] FIG. 4 is a graph illustrating an example of acquiring teacher data in the present invention. [Diagram 5] FIG. 5 is a diagram illustrating an example of creating an axial force estimation model based on deep learning in the present invention. [Figure 6] FIG. 6 is a diagram illustrating an example of LSTM (long short-term memory) in the present invention. [Figure 7] FIG. 7 is a diagram illustrating an example of Dense in the present invention. [Figure 8] FIG. 8 is a diagram for explaining an example of an MLP (multilayer perceptron) in the present invention. [Figure 9] FIG. 9 is a diagram illustrating another example of creating an axial force estimation model based on deep learning in the present invention. [Figure 10] FIG. 10 is a diagram illustrating an example of a CNN (convolutional neural network) in the present invention. [Figure 11] FIG. 11 is another diagram illustrating an example of a CNN in the present invention. [Figure 12] FIG. 12 is a diagram illustrating an example of an MLP in a CNN according to the present invention. [Figure 13] FIG. 13 is a diagram illustrating yet another example of creating an axial force estimation model based on deep learning in the present invention. [Figure 14] FIG. 14 is a graph showing an example of an image of a spectrogram in the present invention. [Figure 15] FIG. 15 is a diagram illustrating an example of an MLP in the LSTM of the present invention. [Figure 16] FIG. 16 is a diagram illustrating yet another example of creating an axial force estimation model based on deep learning in the present invention. [Figure 17] FIG. 17 is a diagram for explaining another example of Dense in the present invention. [Figure 18]FIG. 18 is a diagram illustrating another example of the MLP in the present invention. [Figure 19] FIG. 19 is a graph showing the correlation between the actual measured value and the estimated value of the axial force in the first embodiment. [Figure 20] FIG. 20 is a graph showing the correlation between the actual measured value and the estimated value of the axial force in the second embodiment. [Figure 21] FIG. 21 is a graph showing the correlation between the measured axial force value and the estimated axial force value in the third embodiment. [Figure 22] FIG. 22 is a graph showing the correlation between the measured axial force values ​​and the estimated axial force values ​​in the fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] The axial force estimation device, screw fastening system, axial force estimation method, program, and recording medium of the present invention will be described with examples. However, the present invention is not limited to the following examples. In the following Figs. 1 to 22, the same parts are given the same reference numerals, and their description may be omitted. In addition, in the drawings, for convenience of description, the structure of each part may be appropriately simplified, and the dimensional ratio of each part may be shown diagrammatically, different from the actual one. In addition, the description of each embodiment can be used for each other, unless otherwise specified.

[0014] (Embodiment 1: Axial force estimation device and axial force estimation method) Fig. 1(A) is a block diagram showing an example of the configuration of an axial force estimation device of the present invention. As shown in Fig. 1(A), the axial force estimation device 1 of this embodiment includes a data acquisition means 11 and an axial force estimation means 12.

[0015] The data acquisition means 11 acquires actual measurement data related to the power tool. The power tool may be any tool capable of electrically tightening a screw, such as an electric driver (e.g., a screwdriver, an impact driver, an impact wrench, etc.), an electric torque wrench (nut runner), etc.

[0016] Examples of the measured data include time series data of screw tightening torque, time series data of screw tightening pressing force, time series data of power tool body vibration, time series data of fastened body vibration, etc. The measured data may be of only one type, or of two or more types. The data acquisition means 11 may be, for example, a measuring device for the measured data, and examples of such devices include a piezoelectric measuring element (load washer) that measures the time series data of the screw tightening torque and the screw tightening pressing force, and an acceleration pickup that measures the time series data of the power tool body vibration and the fastened body vibration.

[0017] An example of the measurement of the actual measurement data will be described with reference to Figures 2 and 3. Figure 3 is a schematic diagram showing an enlarged cross section of the lower part of Figure 2.

[0018] As shown in Fig. 3, the time series data of the screw tightening torque and the screw tightening pressing force may be measured, for example, by using a piezoelectric measuring element (load washer) 12c arranged between the fastened body 41 and the fastening body 42. In Fig. 3, 3 indicates a screw. As shown in Fig. 2, the time series data of the vibration of the power tool body and the vibration of the fastened body may be measured by using an acceleration pickup 12a installed on the power tool body 2 and an acceleration pickup 12b installed on the fastened body 41, respectively. The time series data of the vibration of the power tool body and the vibration of the fastened body may be acquired by using the acceleration pickup 12a and the acceleration pickup 12b, respectively, as acceleration data in three directions, namely, the axial direction (z direction) of the power tool body 2, the left-right direction (x direction) perpendicular to the axial direction, and the depth direction (y direction) (hereinafter, sometimes referred to as "three-axis acceleration data").

[0019] The axial force estimation means 12 estimates the axial force at the time of screw fastening by the electric tool according to a previously learned judgment criterion based on the actual measurement data acquired by the data acquisition means 11. The axial force estimation means 12 can be configured, for example, using dedicated hardware (for example, a central processing unit (CPU) or the like), or can be realized on a computer by software processing. The process in which the axial force estimation means 2 estimates the axial force according to the judgment criterion can be performed using a conventionally known method, for example, a method such as a CNN (convolutional neural network) or an RNN (recurrent neural network) can be used.

[0020] 1B is a flowchart of the axial force estimation method of this embodiment. In this embodiment, for example, the axial force is estimated from the measured data in the following manner.

[0021] First, prior to carrying out the axial force estimation method, teacher data is acquired. An example of a method for acquiring the teacher data will be described with reference to FIG. 4. In this example, as shown in FIG. 4, the time series data of the axial force is measured using a piezoelectric measuring element (load washer) 12c of the axial force estimation device shown in FIG. 3, and the average value is used as the teacher data. The average value may be calculated by sampling any point in the part showing the axial force in FIG. 4, and may be calculated from the last 10 points, for example. The screw tightening torque, rotation speed, and screw tightening pressing force in the measurement of the time series data of the axial force may be set according to their values ​​in the data acquisition process (step S1) described later. In addition, the measurement interval (sampling rate) in the measurement of the time series data of the axial force may also be set according to their values ​​in the data acquisition process (step S1) described later.

[0022] Next, the measured data is acquired (step S1). Here, an example will be described in which the measured data includes two types of data: time series data of the screw tightening torque and time series data of the screw tightening pressing force.

[0023] The actual measurement data acquired in this example is, for example, as follows: Time series data of screw tightening torque: Approximately 180 points (measurement interval (sampling rate) depends on the rotation speed set on the power tool as shown below) Time series data of screw tightening pressure: 30 to 90 points (measurement interval (sampling rate) is 20 to 40 Hz) Sampling rate (Hz) = {rotation speed (rpm (revolutions per minute)) / 60} x number of samples per rotation of the power tool

[0024] Next, as preprocessing for creating an axial force estimation model based on deep learning, the actual measurement data is standardized and the time series data is padded.

[0025] First, the measured data is standardized using the following formula.

number

[0026] Next, in order to match the input size to the model, the time series data is padded by adding 0 (zero) to the end until it reaches a predetermined length. The length of each time series data after padding is, for example, as follows: Time series data of screw tightening torque: 1024 (=2 10 ) Time series data of screw tightening pressure: 256 (=2 8 )

[0027] Next, an axial force estimation model is created based on deep learning (step S2). An example is shown in Fig. 5. Fig. 5 shows an example in which LSTM (long short-term memory), a type of RNN (recurrent neural network), is used for the model structure. In Fig. 5, torque indicates time series data of screw tightening torque, and pressing force indicates time series data of screw tightening pressing force.

[0028] In addition, in the LSTM in Fig. 5, for example, two LSTM layers are used as shown in Fig. 6, and the final output H of the final layer becomes the feature. Details of Dense enclosed in the upper dashed rectangle in Fig. 5 are as shown in Fig. 7. In Fig. 7, ReLU indicates an activation function, and is the same in Figs. 10, 12, and 15.

[0029] In FIG. 5, Concatenate indicates concatenation, combining the outputs from LSTM and Dense.

[0030] Furthermore, in the MLP (Multilayer Perceptron) enclosed in the lower dashed rectangle in Fig. 5, for example, the fully connected layers are three layers as shown in Fig. 8. Note that in Fig. 8, tanh indicates an activation function, and the same is true in Fig. 18.

[0031] In this way, an estimate of the axial force Y is obtained.

[0032] In the axial force estimation device 1 of this embodiment, the data acquisition means 11 may further acquire attribute data relating to the power tool.

[0033] Examples of the attribute data include a screw tightening torque, a rotation speed, a screw tightening pressing force, etc., that are set in the power tool. The attribute data may be of one type or of two or more types. The data acquisition means 11 may be, for example, an input means that inputs the attribute data set in the power tool to the axial force estimation means 12.

[0034] When acquiring the attribute data, the axial force estimation means 12 estimates the axial force when the electric tool tightens a screw based on the attribute data acquired by the data acquisition means 11 and the actual measurement data, using pre-learned judgment criteria.

[0035] Next, an example of the axial force estimation method of the present embodiment when acquiring the attribute data will be described. In this example, the attribute data is three types of data, namely, the screw tightening torque, the rotation speed, and the screw tightening pressure force set in the power tool, and the actual measurement data is four types of data, namely, time series data of the screw tightening torque, time series data of the screw tightening pressure force, time series data of the power tool main body vibration, and time series data of the fastened body vibration.

[0036] First, in the same manner as in the case where the attribute data is not acquired, prior to carrying out the axial force estimation method, teacher data is acquired.

[0037] Next, the attribute data and the actual measurement data are acquired (step S1).

[0038] The attribute data acquired in this step is, for example, as follows: Screw tightening torque: 50, 75, 100, 125, 150 [mN m] Rotation speed: 100, 300, 500, 700, 900 [rpm (revolutions per minute)] Screw tightening pressure: 20, 60, 100% (200N at 100%)

[0039] The actual measurement data obtained in this step is, for example, as follows: Time series data of screw tightening torque: Approximately 550 to 650 points (measurement interval (sampling rate) depends on the rotation speed as shown below) Time series data of screw tightening pressure: Approximately 30 to 150 points (measurement interval (sampling rate) is 20 to 40 Hz) Time series data of vibration of fastened body and power tool body (3-axis acceleration data): 512,000 points (measurement interval (sampling rate) is 51,200 Hz) Sampling rate (Hz) = {rotation speed (rpm (revolutions per minute)) / 60} x number of samples per rotation of the power tool (For example, if the rotation speed is 500 rpm (revolutions per minute) and the number of samples per rotation of the power tool is 90, the sampling rate = (500 / 60) x 90 = 750 Hz)

[0040] Next, as preprocessing for creating an axial force estimation model based on deep learning, standardization of each data and padding of the time series data are performed.

[0041] First, the attribute data is standardized using the following formula:

number

[0042] Next, the actual measurement data is standardized in the same manner as in the case where the attribute data is not acquired.

[0043] Next, padding is performed on the time series data. The length of the time series data of the screw tightening torque and the screw tightening pressure force after padding is the same as in the case where the attribute data is not acquired as described above. In addition, the length of the time series data of the fastened body vibration and the power tool body vibration after padding is, for example, as follows: Time series data of vibration of fastened body and power tool body (3-axis acceleration data): 524288 (=2 19 )

[0044] Next, an axial force estimation model is created based on deep learning (step S2). An example is shown in Fig. 9. Fig. 9 is an example in which CNN (convolutional neural network) is used for the model structure. In Fig. 9, torque indicates time series data of screw tightening torque, vibration indicates time series data (three-axis acceleration data) of the vibration of the fastened body and the vibration of the power tool body, and pressing force indicates time series data of the screw tightening pressing force.

[0045] In addition, the CNN enclosed in the upper dashed rectangle in Figure 9 is constructed using a one-dimensional CNN based on VGG, which is used for image classification tasks. A detailed example is shown in Figures 10 and 11.

[0046] Also, in FIG. 9, “Flatten” indicates smoothing and converts the output of the CNN into a one-dimensional vector, and “Concatenate” indicates concatenation and combines outputs from multiple different convolutional layers.

[0047] In the MLP (Multilayer Perceptron) enclosed in the lower dashed rectangle in Fig. 9, the error between the output result obtained by inputting the training data (the attribute data and the measured data) and the teacher data is calculated, and the weights of the neural network are updated to minimize this error. An example of the details is shown in Fig. 12.

[0048] In this way, an estimate of the axial force Y is obtained.

[0049] In creating an axial force estimation model based on deep learning when acquiring the attribute data (step S2), the model structure may be LSTM (long short-term memory), which is a type of RNN (recurrent neural network), in addition to CNN. Fig. 13 shows an example using LSTM. In Fig. 13, torque indicates time series data of screw tightening torque, triaxial acceleration data indicates time series data (triaxial acceleration data) of the vibration of the fastened body and the vibration of the power tool body, and pressing force indicates time series data of the screw tightening pressing force.

[0050] The spectrogram in FIG. 13 is a three-dimensional graph composed of elements of time, frequency, and amplitude level. As the data input to the LSTM becomes longer, the computational resources of the computer increase proportionally, causing problems such as unstable calculations. Therefore, in this example, a spectrogram is used to learn the time-series data (three-axis acceleration data) of the vibration of the fastened body and the vibration of the power tool body, which can reduce the number of data samples without reducing the amount of information. Except for this point, the process before the creation of the axial force estimation model based on deep learning (step S2) in this example is the same as in the above-mentioned CNN example. FIG. 14 shows an example of an image of a spectrogram.

[0051] In addition, in the LSTM enclosed in the upper dashed rectangle in Figure 13, for example, two LSTM layers are stacked as described above with reference to Figure 6, and the final output H of the final layer becomes the feature.

[0052] In FIG. 13, Concatenate indicates concatenation, combining the outputs from each LSTM.

[0053] Furthermore, in the MLP (multilayer perceptron) enclosed in the lower dashed rectangle in FIG. 13, for example, the number of fully connected layers is set to three as shown in FIG.

[0054] In this way, an estimate of the axial force Y is obtained.

[0055] From here on, an example will be described in which, when acquiring the attribute data, the actual measurement data is two types of time series data of the screw tightening torque and time series data of the screw tightening pressing force.

[0056] The acquisition of the teacher data in this example can be carried out in the same manner as described above with reference to Figure 4 in the example in which the attribute data is not acquired, for example, as in the example in which there are four types of actual measurement data.

[0057] Next, the attribute data acquired in acquiring the attribute data and the actual measurement data in this example (step S1) is as shown in Table 1, for example. [Table 1]

[0058] The actual measurement data acquired in this example is, for example, as follows: Time series data of screw tightening torque: Approximately 180 points (measurement interval (sampling rate) depends on the number of rotations as shown below) Time series data of screw tightening pressure: 30 to 90 points (measurement interval (sampling rate) is 20 to 40 Hz) Sampling rate (Hz) = {rotation speed (rpm (revolutions per minute)) / 60} x number of samples per rotation of the power tool

[0059] In this example, the standardization of each data and the padding of the time series data as preprocessing for creating an axial force estimation model based on deep learning can be performed in the same manner as in the example in which there are four types of actual measurement data described above.

[0060] Next, an axial force estimation model is created based on deep learning (step S2). An example is shown in Fig. 16. In Fig. 16, the torque indicates time series data of the screw tightening torque, and the pressing force indicates time series data of the screw tightening pressing force.

[0061] In addition, in the LSTM in Fig. 16, for example, two LSTM layers are stacked as described above with reference to Fig. 6, and the final output H of the final layer becomes the feature. Details of Dense enclosed in the upper dashed rectangle in Fig. 16 are as shown in Fig. 17, for example.

[0062] In FIG. 16, Concatenate indicates concatenation, and combines the outputs from the LSTM and each Dense.

[0063] Furthermore, in the MLP (Multilayer Perceptron) enclosed in the lower dashed rectangle in FIG. 16, for example, the number of fully connected layers is set to three as shown in FIG.

[0064] In this way, an estimate of the axial force Y is obtained.

[0065] Unlike the technologies described in Patent Documents 1 and 2, the axial force estimation device and axial force estimation method of the present invention do not require holes for arranging strain gauges in the screws to which they are applied, and therefore can be applied to small screws such as M2 screws and even smaller screws with a nominal length of about 1 mm.

[0066] (Embodiment 2: Screw fastening system) The screw fastening system of the present embodiment includes an electric power tool and an axial force estimation device, and the axial force estimation device is the axial force estimation device of the present invention. As described above, the electric power tool may be any tool that can fasten a screw electrically, and examples of the electric power tool include an electric driver (e.g., a screwdriver, an impact driver, an impact wrench, etc.) and an electric torque wrench (nut runner).

[0067] (Embodiment 3: Program and recording medium) The program of the present embodiment is characterized by causing a computer to execute the axial force estimation method of the present invention. The program of the present embodiment may be recorded in, for example, a computer-readable recording medium. EXAMPLES

[0068] [Example 1] In the same manner as in the example of the first embodiment in which the attribute data is not acquired, the axial force during screw tightening using an electric screwdriver (ADVANCE-Z manufactured by Vanguard Systems Co., Ltd.) was estimated. The batch size was 1, the number of epochs was 1000, the loss function was MSE (mean square error), the optimization algorithm was Adam, and the learning rate was 0.001 (1 -3 ) was decided.

[0069] The learning results are as follows, and the correlation between the measured and estimated axial force values ​​is shown in Figure 19. As shown below, a high estimation accuracy of 4.43% (mean absolute percentage error) was obtained. MAE (Mean Absolute Error): 10.9N MAPE (Mean Absolute Percentage Error): 4.43% RMSE(Root Mean Square Error): 14.6N R 2 (Determination Index): 0.965

[0070] [Example 2] The attribute data in the above-mentioned embodiment 1 was obtained, and the axial force during screw tightening using an electric screwdriver (ADVANCE-Z manufactured by Vanguard Systems Co., Ltd.) was estimated in the same manner as in the example in which CNN was used for the model structure. In the learning conditions at that time, the attribute data was randomly divided into 5×5×3×3=225 pieces of data for three times each for all combinations of the following five screw tightening torques, five rotation speeds, and three screw tightening pressure forces, and 179 pieces were used for training, 23 pieces for verification, and 23 pieces for testing. In addition, the batch size was 1, the number of epochs was 500, the loss function was MSE (mean square error), the optimization algorithm was Adam, and the learning rate was 10 -3 It was decided. Screw tightening torque: 50, 75, 100, 125, 150 [mN m] Rotation speed: 100, 300, 500, 700, 900 [rpm (revolutions per minute)] Screw tightening pressure: 20, 60, 100% (200N at 100%)

[0071] The learning results are as follows (the evaluation index is the value for the test data), and the correlation between the measured value and the estimated value of the axial force is shown in Figure 20. As shown below, a high estimation accuracy of 4.63% (MAPE (Mean Absolute Percentage Error)) was obtained. MAE (Mean Absolute Error): 6.058N MAPE (Mean Absolute Percentage Error): 4.63% RMSE(Root Mean Square Error): 7.48N R 2 (Determination coefficient): 0.982

[0072] [Example 3] The attribute data in the above-mentioned embodiment 1 was obtained, and the axial force during screw tightening by an electric screwdriver (ADVANCE-Z manufactured by Vanguard Systems Co., Ltd.) was estimated in the same manner as in the example using LSTM for the model structure. In the learning conditions at that time, the attribute data was randomly divided into 588 data pieces (5×8×3×3=360 pieces of data for all combinations of the following 5 screw tightening torques, 8 rotation speeds, and 3 screw tightening pressure forces, each of which was performed three times) in addition to the 225 pieces of data in the same manner as in the embodiment 2, 465 pieces of data were used for training, 60 pieces of data were used for verification, and 63 pieces of data were used for testing. In addition, the measurement interval (sampling rate) of the time series data (three-axis acceleration data) of the vibration of the fastened body and the vibration of the electric tool body was changed to 12,800 Hz instead of 51,200 Hz described in the above-mentioned embodiment 1. The batch size is 1, the number of epochs is 200, the loss function is MSE (mean squared error), the optimization algorithm is Adam, and the learning rate is 10 -3 It was decided. Screw tightening torque: 50, 75, 100, 125, 150 [mN m] Rotation speed: 100, 300, 500, 700, 900, 1100, 1300, 1500 [rpm (revolutions per minute)] Screw tightening pressure: 20, 60, 100% (200N at 100%)

[0073] The learning results are as follows (the evaluation index is the value for the test data), and the correlation between the measured value and the estimated value of the axial force is shown in Figure 21. As shown below, a high estimation accuracy of 7.23% (MAPE (Mean Absolute Percentage Error)) was obtained. MAE(Mean Absolute Error): 18.040N MAPE (Mean Absolute Percentage Error): 7.23% RMSE(Root Mean Square Error): 26.81N R 2 (Determination coefficient): 0.9881

[0074] [Example 4] The attribute data in the above-mentioned embodiment 1 was acquired, and the axial force during screw tightening using an electric screwdriver (ADVANCE-Z manufactured by Vanguard Systems Co., Ltd.) was estimated in the same manner as in the example in which two types of time-series data of the screw tightening torque and the screw tightening pressing force were used as the measured data. The attribute data included 11 screw tightening torques, 7 rotation speeds, and the screw tightening pressing force, which were determined according to the screw tightening torque. 11 x 7 x 3 = 231 pieces of data, each of which was three times, were randomly divided into 186 pieces for training, 21 pieces for verification, and 23 pieces for testing. The batch size was 1, the number of epochs was 1000, the loss function was MSE (mean square error), the optimization algorithm was Adam, and the learning rate was 0.001 (1 -3 ) was decided.

[0075] The learning results are as follows (the evaluation index is the value for the test data), and the correlation between the measured value and the estimated value of the axial force is as shown in Figure 22. As shown below, a high estimation accuracy of 3.63% (MAPE (Mean Absolute Percentage Error)) was obtained. MAE (Mean Absolute Error): 8.95N MAPE (Mean Absolute Percentage Error): 3.63% RMSE(Root Mean Square Error): 13.9N R 2 (Determination coefficient): 0.969

[0076] Although the present invention has been described above with reference to the embodiments and examples, the present invention is not limited to the above-mentioned embodiments and examples. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

[0077] All patents, patent applications, and publications cited in this specification are hereby incorporated by reference into this specification as if the contents of the patents, patent applications, and publications were specifically set forth herein.

[0078] <Additional Notes> A part or all of the above-described embodiments and examples can be described as follows, but is not limited to the following. <Axial force estimation device> (Appendix 1) A data acquisition means for acquiring actual measurement data relating to the power tool; an axial force estimation means for estimating an axial force at the time of fastening the screw by the electric tool based on the actual measurement data acquired by the data acquisition means and a previously learned judgment criterion; An axial force estimation device comprising: (Appendix 2) 2. The axial force estimation device according to claim 1, wherein the actual measurement data includes time-series data of a screw tightening torque and a screw tightening pressing force. (Appendix 3) The axial force estimation device according to claim 2, wherein the measured data further includes time series data of vibration of the power tool body and vibration of the fastened body. (Appendix 4) The data acquisition means includes an acceleration pickup; The axial force estimation device described in Appendix 3 is characterized in that the time series data of the vibration of the power tool body and the vibration of the fastened body are each acquired using the acceleration pickup as acceleration data in three directions: the axial direction (z direction) of the screw fastening power tool body, the left-right direction (x direction) perpendicular to the axial direction, and the depth direction (y direction). (Appendix 5) The data acquisition means further acquires attribute data relating to the power tool, 5. The axial force estimation device according to claim 1, wherein the axial force estimation means estimates the axial force during screw tightening by the electric tool according to a pre-learned criterion based on the attribute data and the actual measurement data acquired by the data acquisition means. (Appendix 6) 6. The axial force estimation device according to claim 5, wherein the attribute data includes a screw tightening torque, a rotation speed, and a screw tightening pressing force set in the power tool. <Screw fastening system> (Appendix 7) The present invention includes an electric power tool and an axial force estimation device, 7. A screw fastening system, comprising: the axial force estimation device according to claim 1, wherein the axial force estimation device is an axial force estimation device according to any one of claims 1 to 6. (Appendix 8) 8. The screw fastening system according to claim 7, which is applied to screws having a nominal length of 1 mm or more. <Axial force estimation method> (Appendix 9) A data acquisition step of acquiring actual measurement data relating to the power tool; an axial force estimation step of estimating an axial force at the time of fastening the screw by the electric tool based on the actual measurement data acquired in the data acquisition step and a previously learned judgment criterion; An axial force estimation method comprising: (Appendix 10) 10. The axial force estimation method according to claim 9, wherein the actual measurement data includes time-series data of a screw tightening torque and a screw tightening pressing force. (Appendix 11) The axial force estimation method according to claim 10, wherein the measured data further includes time series data of vibration of the power tool body and vibration of the fastened body. (Appendix 12) In the data acquisition step, The axial force estimation method described in Appendix 11, characterized in that time series data of the vibration of the power tool body and the vibration of the fastened body are obtained as acceleration data in three directions: the axial direction of the power tool body (z direction), the left-right direction perpendicular to the axial direction (x direction), and the depth direction (y direction). (Appendix 13) The data acquisition step further includes acquiring attribute data related to the power tool, 13. The axial force estimation method according to any one of appendices 9 to 12, wherein, in the axial force estimation step, the axial force at the time of fastening the screw by the electric tool is estimated according to a pre-learned judgment criterion based on the attribute data and the actual measurement data acquired in the data acquisition step. (Appendix 14) 14. The axial force estimation method according to claim 13, wherein the attribute data includes a screw tightening torque, a rotation speed, and a screw tightening pressing force set in the power tool. <Program> (Appendix 15) A program for causing a computer to execute the axial force estimation method according to any one of appendices 9 to 14. <Recording Media> (Appendix 16) A computer-readable recording medium having the program according to claim 15 recorded thereon. [Explanation of symbols]

[0079] 1 Axial force estimation device 2 Power tool body 3 Screws 11 Data Acquisition Methods 12 Axial force estimation means 12a, 12b Acceleration pickup 41 Object to be fastened 42 Fasteners

Claims

1. A data acquisition means for acquiring actual measurement data relating to the power tool; an axial force estimation means for estimating an axial force at the time of fastening the screw by the electric tool based on the actual measurement data acquired by the data acquisition means and a previously learned judgment criterion; An axial force estimation device comprising:

2. 2. The axial force estimation device according to claim 1, wherein the measured data includes time-series data of a screw tightening torque and a screw tightening pressing force.

3. 3. The axial force estimation device according to claim 2, wherein the actual measurement data further includes time series data of vibration of the power tool body and vibration of the fastened body.

4. The data acquisition means includes an acceleration pickup; The axial force estimation device of claim 3, characterized in that the time series data of the vibration of the power tool body and the vibration of the fastened body are each acquired using the acceleration pickup as acceleration data in three directions: the axial direction of the power tool body (z direction), the left-right direction perpendicular to the axial direction (x direction), and the depth direction (y direction).

5. The data acquisition means further acquires attribute data relating to the power tool, The axial force estimation device according to any one of claims 1 to 4, characterized in that the axial force estimation means estimates the axial force at the time of screw tightening by the electric tool according to a pre-learned judgment criterion based on the attribute data and the actual measurement data acquired by the data acquisition means.

6. 6. The axial force estimation device according to claim 5, wherein the attribute data includes a screw tightening torque, a rotation speed, and a screw tightening pressing force set in the power tool.

7. The present invention includes an electric power tool and an axial force estimation device, A screw fastening system, characterized in that the axial force estimation device is an axial force estimation device according to any one of claims 1 to 6.

8. 8. The screw fastening system according to claim 7, which is applied to screws having a nominal length of 1 mm or more.

9. A data acquisition step of acquiring actual measurement data relating to the power tool; an axial force estimation step of estimating an axial force at the time of fastening the screw by the electric tool based on the actual measurement data acquired in the data acquisition step and a previously learned judgment criterion; An axial force estimation method comprising:

10. A program for causing a computer to execute the axial force estimation method according to claim 9.

11. 11. A computer-readable recording medium having the program according to claim 10 recorded thereon.

Images