Imitation data generating device, imitation data generating method, and program

Abstract

To generate imitation data that is not unnatural as data representing a resonance frequency of an object to be inspected.SOLUTION: A processor (11) of an imitation data generation device (10) executes the steps of: obtaining a measurement data group consisting of a plurality of pieces of measurement data, in which each piece of measurement data represents a resonance frequency of a first to nth (n is a natural number equal to or greater than 2) order of a sample; and generating an imitation data group consisting of a plurality of pieces of imitation data by imitating the measurement data group. In the step of generating the imitation data group, the processor (11) generates the imitation data group so that a distribution of the imitation data group reproduces a distribution of the measurement data group in a n-dimensional space of the first to nth resonance frequencies.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technique for simulating data representative of resonant frequencies. [Background technology]

[0002] Inspection of an object is carried out by checking the resonance frequency of the object to be inspected. For example, Patent Document 1 describes a device that checks the resonance frequency of a sealed container to determine whether the container is defective. In addition, in process compensation resonance testing (PCRT), the presence or absence of defects is determined by using the inherent resonance frequency and the distribution of its strength, which are generated depending on the internal conditions of the object, such as its structure, material, and the presence or absence of defects.

[0003] Patent Document 2 describes a system for estimating the damage state of a building after an earthquake, which acquires an actual response based on an earthquake using a vibration sensor and estimates the damage state using machine learning with the acquired actual response as an input. Patent Document 3 describes a method for automating the procedure in principal component analysis to further shorten the analysis work and suppress the occurrence of variations in the quality of the analysis.

[0004] To perform an inspection using resonant frequency, it is necessary to collect data that represents the resonant frequency of the target object, especially data that represents the resonant frequency of the target object that contains defects. Conventionally, data collection methods have been used, such as logging the resonant frequency and its strength using an actual part in PCRT, or calculating what resonant frequency will occur using a specified structural analysis software. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 58-9035 [Patent Document 2] JP 2020-128951 A [Patent Document 3] JP 2010-112887 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, it is difficult to collect data on defective products that occur unintentionally in the manufacturing industry. In addition, there is a problem that when data on defective products is simulated using structural analysis software, the data may be biased or data that deviates from data on defective products that actually occur may be generated. Even with the techniques described in Patent Documents 1 to 3, it was not possible to simulate data that does not deviate from data that represents actual measurement results as data that represents the resonant frequency of the sample to be inspected.

[0007] An aspect of the present invention aims to provide a technique for generating artificial data representing the resonant frequency of an object to be inspected, the artificial data not deviating from actually measured data. [Means for solving the problem]

[0008] In order to solve the above problems, an imitation data generating device according to one embodiment of the present invention includes one or more processors, and the processor executes the steps of acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth (n is a natural number equal to or greater than 2) resonant frequency of a sample, and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group, wherein in the step of generating the imitation data group, the processor generates the imitation data group such that a distribution of the imitation data group reproduces a distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies.

[0009] Moreover, an imitation data generating device according to one embodiment of the present invention includes one or more processors, and the processor executes the steps of acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth order (n is a natural number equal to or greater than 2) resonant frequency of a sample, and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group, wherein a distribution of the imitation data group in an n-dimensional space of the 1st to nth order resonant frequencies is substantially identical to a distribution of the measurement data group.

[0010] The imitation data generating device according to each aspect of the present invention may be realized by a computer. In this case, the program of the imitation data generating device that realizes the imitation data generating device on a computer by causing the computer to operate as each part (software element) of the imitation data generating device, and the computer-readable recording medium on which it is recorded, also fall within the scope of the present invention. Effect of the Invention

[0011] According to one aspect of the present invention, it is possible to generate artificial data representing the resonant frequency of an object to be inspected, the artificial data not deviating from the actually measured data. [Brief description of the drawings]

[0012] [Figure 1] 1 is a block diagram showing a configuration of an imitation data generating device according to a first embodiment of the present invention. [Diagram 2] 1 is a flowchart showing an imitation data generating method according to a first embodiment of the present invention. [Diagram 3] FIG. 2 is a diagram illustrating a measurement data group and an imitation data group according to the first embodiment of the present invention. [Figure 4] 1 is a block diagram showing a configuration of an imitation data generating system according to a first embodiment of the present invention. [Diagram 5] 4 is a flowchart illustrating a measurement data generating method according to the first embodiment of the present invention. [Figure 6]1 is a flowchart illustrating an imitation data generating method according to a first embodiment of the present invention. [Figure 7] 4 is a flowchart illustrating a method for generating a matrix product according to the first embodiment of the present invention. [Figure 8] FIG. 4 is a diagram for explaining a shift amount of measurement data according to the first embodiment of the present invention. [Figure 9] 11 is a flowchart illustrating an imitation data generating method according to a third embodiment of the present invention. [Figure 10] FIG. 11 is a diagram illustrating an example of an imitation data group according to the third embodiment of the present invention. [Figure 11] FIG. 11 is a diagram for explaining the amount of shift of measurement data according to the third embodiment of the present invention. [Figure 12] FIG. 11 is a diagram for explaining the amount of shift of measurement data according to the fourth embodiment of the present invention. [Figure 13] FIG. 11 is a diagram illustrating a measurement data group and an imitation data group according to the fourth embodiment of the present invention. [Figure 14] 13 is a flowchart illustrating an imitation data generating method according to a fifth embodiment of the present invention. [Figure 15] FIG. 13 is a diagram illustrating an example of an imitation data group according to the fifth embodiment of the present invention. [Figure 16] FIG. 13 is a diagram illustrating an example of an imitation data group according to the fifth embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] [Embodiment 1] An embodiment of the present invention will be described in detail below. Fig. 1 is a block diagram showing a configuration of an imitation data generating device 10 according to an embodiment of the present invention. The imitation data generating device 10 is a device that imitates measurement data that indicates the resonance frequency of an object to be inspected. The imitation data generating device 10 is, for example, a personal computer.

[0014] As an example, the object is a casting or a mold formed by casting. As an example, when inspecting the presence or absence of defects in a casting by using a resonant frequency, the imitation data generated by the imitation data generating device 10 is used as a comparison target for the inspection, or as training data when generating a trained model by machine learning using training data. Note that the object to be inspected is not limited to the above-mentioned example, and the inspection object may be another item. In addition, the use of the imitation data is not limited to the above-mentioned example, and the imitation data may be used for various purposes.

[0015] The imitation data generating device 10 includes one or more processors 11. The processor 11 executes an imitation data generating method M1. FIG. 2 is a flowchart showing the flow of the imitation data generating method M1 executed by the processor 11. The imitation data generating method M1 includes an acquiring step M11 and a generating step M12. The acquiring step M11 is a step of acquiring a measurement data group consisting of a plurality of measurement data. The measurement data is data representing the 1st to nth order (n is a natural number equal to or greater than 2) resonance frequencies of a sample, which is an object.

[0016] In the acquisition step M11, the processor 11 may acquire all or a part of the measurement data by reading it from a storage device built into the imitation data generation device 10 or from an external storage device. The processor 11 may also acquire all or a part of the measurement data from another device via a communication interface or an input / output interface.

[0017] The generating step M12 is a step of generating an imitation data group consisting of a plurality of imitation data imitation of the measurement data group. In the generating step M12, the processor 11 generates the imitation data group such that the distribution of the imitation data group reproduces the distribution of the measurement data group in an n-dimensional space of 1st to nth order resonance frequencies. As an example, the processor 11 generates a matrix representing the measurement data group, and generates the imitation data group by performing singular value decomposition or LU decomposition using the generated matrix.

[0018] The distribution of the artificial data group in n-dimensional space reproduces the distribution of the measured data group, meaning, for example, that the distribution pattern of the artificial data group in n-dimensional space is identical to or similar to the distribution pattern of the measured data group.

[0019] FIG. 3 is a diagram illustrating a measurement data group and an imitation data group that imitates the measurement data group. In the diagram, the measurement data group D11 is an example of a measurement data group acquired by the processor 11 in the acquisition step M11. The imitation data group D21 is an example of an imitation data group that imitates the measurement data group D11. In the diagram, the horizontal axis indicates the first resonance frequency [Hz], and the vertical axis indicates the second resonance frequency [Hz]. Each dot included in the diagram represents measurement data. In the example of FIG. 3, in order to facilitate understanding of the invention, a two-dimensional space represented by two resonance frequencies is illustrated, but the number of resonance frequencies represented by the measurement data group and the imitation data group is not limited to two and may be more than this.

[0020] As shown in Fig. 3, the distribution of the measurement data included in the measurement data group D11 in a two-dimensional space is similar to the distribution of the imitation data included in the imitation data group D21 in a two-dimensional space. In other words, the distribution of the imitation data group D21 in the n-dimensional space of the 1st to nth resonance frequencies is substantially the same as the distribution of the measurement data group D11. More specifically, in the example of Fig. 3, both the measurement data group D11 and the imitation data group D12 form a distribution that spreads obliquely from the lower left to the upper right in the two-dimensional space represented by the vertical and horizontal axes.

[0021] According to the above configuration, the imitation data generating device 10 generates an imitation data group so that the distribution of the imitation data group reproduces the distribution of the measurement data group in the n-dimensional space of the 1st to nth resonance frequencies. This makes it possible to generate imitation data that maintains the relationship between the multiple resonance frequencies of the actually measured sample and does not deviate from the data representing the actual measurement results.

[0022] [System Configuration] 4 is a block diagram showing a configuration of an imitation data generation system 1 according to this embodiment. The imitation data generation system 1 measures a resonance frequency of a sample 2 and imitates measurement data representing the measured resonance frequency. As an example, the sample 2 is a casting or a mold produced by casting. The imitation data generation system 1 includes a vibration generator 20 and a vibration receiver 30 in addition to an imitation data generation device 10.

[0023] (Vibration Generator) The vibration generator 20 is a device that vibrates the sample 2. The vibration generator 20 vibrates the sample 2 by, for example, an impulse response method or a sine wave sweep method. When the impulse response method is used, the vibration generator 20 is, for example, an impulse vibration device equipped with a hammer or the like. When the sine wave sweep method is used, the vibration generator 20 is, for example, a sine wave vibration device equipped with a piezoelectric element or the like.

[0024] (Vibration receiving device) The vibration receiving device 30 is a device that converts the vibration of the sample 2 into a signal. Examples of the vibration receiving device 30 include a microphone, a vibration meter, or a vibrator.

[0025] [Configuration of the Imitation Data Generating Device] The imitation data generating device 10 is realized by using a general-purpose computer. As shown in Fig. 1, the imitation data generating device 10 includes a processor 11, a primary memory 12, a secondary memory 13, an input / output IF 14, a communication IF 15, and a bus 16. The processor 11, the primary memory 12, the secondary memory 13, the input / output IF 14, and the communication IF 15 are connected to each other via the bus 16.

[0026] The secondary memory 13 stores an imitation data generation program P1. The processor 11 expands the imitation data generation program P1 stored in the secondary memory 13 onto the primary memory 12, and executes each step included in the imitation data generation method M1 according to instructions included in the imitation data generation program P1 expanded onto the primary memory 12. An example of a device that can be used as the processor 11 is a CPU (Central Processing Unit). An example of a device that can be used as the primary memory 12 is a semiconductor RAM (Random Access Memory). An example of a device that can be used as the secondary memory 13 is a flash memory.

[0027] An input device and / or an output device are connected to the input / output IF 14. An example of the input / output IF 14 is a Universal Serial Bus (USB). As an example, information acquired from the vibration receiving device 30 in the imitation data generation method M1 is input to the imitation data generation device 10 via the input / output IF 14. Also, as an example, the imitation data group generated in the imitation data generation method M1 is output from the imitation data generation device 10 via the input / output IF 14.

[0028] The communication IF 15 is an interface for communicating with other computers. The communication IF 15 may include an interface for communicating with other computers without a network, such as a Bluetooth (registered trademark) interface. The communication IF 15 may also include an interface for communicating with other computers via a LAN (Local Area Network), such as a Wi-Fi (registered trademark) interface.

[0029] In this embodiment, a configuration is adopted in which the imitation data generation method M1 is executed using a single processor (processor 11), but the present invention is not limited to this. That is, a configuration may be adopted in which the imitation data generation method M1 is executed using multiple processors. In this case, the multiple processors that execute the imitation data generation method M1 in cooperation with each other may be provided in a single computer and configured to be able to communicate with each other via a bus, or may be provided in a distributed manner in multiple computers and configured to be able to communicate with each other via a network. As an example, a processor built in a computer that constitutes a cloud server and a processor built in a computer owned by a user of the cloud server may execute the imitation data generation method M1 in cooperation with each other.

[0030] [Method of generating measurement data] 5 is a flowchart illustrating a measurement data generation method M20 performed by the imitation data generation system 1. Note that some steps may be performed in parallel or in a different order. In step S21, the measurer turns on the vibration generator 20 and the vibration receiver 30. In step S22, the measurer uses the vibration generator 20 to resonate the sample 2. In step S23, the vibration receiver 30 converts the vibration of the sample 2 into a signal.

[0031] In step S24, the processor 11 of the imitation data generating device 10 calculates a resonant frequency (natural frequency) from the signal converted by the vibration receiving device 30. When the impulse response method is used, the processor 11 calculates the resonant frequency by, for example, Fourier transform. When the sine wave sweep method is used, the processor 11 calculates the resonant frequency by, for example, differentiating the signal.

[0032] In step S25, the fake data generation device 10 records the measurement data representing the resonance frequency in the secondary memory 13. The measurement data is, for example, data representing 1st to nth order (n is a natural number equal to or greater than 2) resonance frequencies. The fake data generation device 10 performs the measurement data generation process shown in FIG. 5 for the multiple samples 2, whereby a measurement data group consisting of multiple pieces of measurement data is recorded in the secondary memory 13.

[0033] [Method of generating fake data] 6 is a flowchart illustrating an imitation data generation method M10 performed by the imitation data generation device 10. The processor 11 generates an imitation data group by imitating a measurement data group stored in the secondary memory 13. Note that some steps may be executed in parallel or in a different order.

[0034] In this operation example, the processor 11 generates an imitation data group so that the distribution of the imitation data group reproduces the distribution of the measurement data group in an n-dimensional space of 1st to nth order resonance frequencies. More specifically, the processor 11 generates an imitation data group so as to reproduce a part or all of the relative positional relationship between the measurement data in each of a plurality of axial directions representing the n-dimensional space.

[0035] (Steps S10 and S11) In step S10, the processor 11 acquires the measurement data group by reading the measurement data group from the secondary memory 13. In step S11, the processor 11 acquires a sample matrix f raw Generate.

[0036] (sample matrix) Sample matrix f raw is a matrix representing a group of measurement data, and is an m-row by n-column matrix in which each row represents the measurement data of each of m samples (m is a natural number of 2 or more). m is the number of measurement data, and n is the number of resonance frequencies included in each measurement data.

[0037] Sample matrix f rawAs an example, it is expressed by the following formula (1). In formula (1), the sample matrix f raw represents a group of measurement data d[i] (1≦i≦m). The measurement data d[i] are each the 1st to nth resonance frequencies f j_sample[i] (1≦j≦n).

number

[0038] (Step S12) In step S12, the processor 11 performs a process of standardizing or normalizing the resonance frequency represented by the measurement data. In this example, the processor 11 standardizes or normalizes the sample matrix f raw Standardization matrix f norm Calculate.

[0039] For example, the processor 11 calculates the sample matrix f by using the following formula (2) or (3). raw When equation (2) is used, in other words, the processor 11 standardizes or normalizes the sample matrix f raw The average value of each column of j_ave (0≦j≦n) is subtracted from each component to obtain the standardized matrix f norm When formula (3) is used, in other words, the processor 11 calculates the sample matrix f raw Standardized matrix f norm Calculate.

number

number

[0040] In equations (2) and (3), the matrix ave(f raw ) is expressed by the following formula (4). In formula (4), the average value f 1_ave f 1_sample[1] ~f 1_sample[m] Similarly, the average value f 2_ave f 2_sample[1] ~f2_sample[m] The same is true for the other columns, and the average value f j_ave (0≦j≦n) is f j_sample[1] ~f j_sample[m] is the average value.

number

[0041] In equation (3), the matrix std(f raw ) is expressed by the following formula (5). In formula (5), the standard deviation f 1_std f 1_sample[1] ~f 1_sample[m] Similarly, the standard deviation f 2_std f 2_sample[1] ~f 2_sample[m] The same is true for the other columns, with standard deviation f j_std (0≦j≦n) is f j_sample[1] ~f j_sample[m] is the standard deviation of

number

[0042] (Step S13) In step S13, the processor 11 calculates the standardized matrix f norm The variance-covariance matrix f cov For example, the processor 11 calculates the standardized matrix f norm In this example, since n resonant frequencies are handled, the resulting variance-covariance matrix f cov is a square matrix with n rows and n columns.

[0043] (Step S14) In step S14, the processor 11 calculates a random number matrix mtx based on a normal distribution. rand In this operation example, n resonant frequencies are handled and the number of samples is m, so the random number matrix mtx rand is a matrix with m rows and n columns. randThe values ​​of some of the components of mtx may be the same, or all of the components may be different. For example, the processor 11 may generate a random number matrix mtx in which the values ​​of multiple components included in each row are equal. rand Generate a random matrix mtx rand As an example, it is expressed by the following formula (6). In formula (6), the random number rand i (1≦i≦m) is, for example, a random number generated by a predetermined random function.

number

[0044] (Step S15) In step S15, the processor 11 calculates the variance-covariance matrix f cov Based on the matrix multiplication mtx pre For example, the processor 11 generates a variance-covariance matrix f cov By singular value decomposition or LU decomposition, pre Generates the matrix multiplication mtx. pre The process of generating the will be described later.

[0045] (Step S16) In step S16, the processor 11 performs the matrix multiplication mtx pre Random number matrix mtx rand The matrix mtx after multiplication and transposition post Generate the post-operation matrix mtx post is expressed by the following equation (7).

number

[0046] (Step S17) In step S17, the processor 11 calculates the post-operation matrix mtx post The pseudo-data group is represented by performing the inverse transformation of the standardization or normalization performed in step S12 on the pseudo-data group f generated As an example, in step S12, a sample matrix f rawis normalized by the above formula (2), the processor 11 performs the inverse transformation process by the following formula (8) in step S17. In other words, when formula (8) is used, the processor 11 performs the inverse transformation process by the following formula (8) in the following step S17. post The average value of each column of the matrix is ​​added to each element to create the imitation matrix f generated On the other hand, in step S12, the sample matrix f raw is standardized by the above formula (3), the processor 11 performs inverse transformation processing by the following formula (9) in step S17. In formula (9), the operator ◯ represents the Hadamard product.

number

number

[0047] Imitation matrix f generated is a matrix representing the fake data set, and the sample matrix f raw In other words, the pseudo-matrix f generated is an m-row × n-column matrix in which each row represents m pieces of imitation data. The imitation data generated by the imitation data generation device 10 is used as training data in a process of generating a trained model by machine learning using training data, for example.

[0048] FIG. 7 shows the matrix multiplication mtx performed by the processor 11. pre 7 is a diagram showing an example of a flow of a process of generating a variance-covariance matrix f cov n eigenvectors evac for each column of j (1≦j≦n) and n eigenvalues ​​eval j (1≦j≦n), and the matrix product mtx is calculated based on the calculated eigenvectors and eigenvalues. pre Generate.

[0049] In step S151, the processor 11 calculates the variance-covariance matrix f cov eigenvalues ​​of eval j(1≦j≦n) and eigenvector evec j (1≦j≦n) is calculated by singular value decomposition. In other words, the eigenvalue eval j and the eigenvector evec j Multiple pairs with are obtained.

[0050] In step S152, the processor 11 calculates a plurality of eigenvalues ​​eval j In addition to sorting in descending order, multiple eigenvectors evec j eigenvalue eval j In other words, the processor 11 rearranges the eigenvalues ​​eval j and the eigenvector evec j The pair of eigenvalues ​​eval j In other words, sort the eigenvalues ​​eval j and the eigenvector evec j The pairs of and do not change due to the sorting. j and the eigenvector evec j In the following, we will use the eigenvalue eval k (1≦k≦n) and eigenvector evec k (1≦k≦n).

[0051] In step S153, the processor 11 calculates the eigenvalue eval k Square root of s k Calculate the square root s k is expressed by the following equation (10).

number

[0052] In step S154, the processor 11 calculates the square root s k Diagonal matrix s diag Convert the diagonal matrix s diag is expressed by the following equation (11).

number

[0053] In step S155, the processor 11 calculates the eigenvector matrix evec and the diagonal matrix s diag The matrix product mtx is the inner product of pre Calculate the matrix product mtx pre is expressed by the following formula (12). In formula (12), the eigenvector matrix evec is k is a matrix whose elements are

number

[0054] Fig. 8 is a schematic diagram for explaining the shift amount of the measurement data according to the present embodiment. In the figure, the horizontal axis is the first resonance frequency [Hz], and the vertical axis is the second resonance frequency [Hz]. In the example of Fig. 8, the measurement data is shifted by an arrow A11 based on the normal distribution random number rand11 and the normal distribution random number rand12.

[0055] As described above, FIG. 3 is a diagram illustrating a measurement data group acquired by the imitation data generation device 10 in this embodiment and an imitation data group generated by the imitation data generation device 10. In the figure, the distribution of the imitation data group D21 in a two-dimensional space is similar to the distribution of the measurement data group D11. In other words, it can be said that the measurement data group D11 and the imitation data group D21 maintain a correlation between the resonant frequencies. In this way, according to this embodiment, the imitation data generation device 10 can generate a natural imitation data group that does not deviate from the measurement data group.

[0056] In this embodiment, in step S14 of FIG. 6, the processor 11 generates a random number matrix mtx rand This makes it possible to generate random numbers that show the same tendency across different frequency bands, for example, if the resonant frequency f1 is high in one sample, then the resonant frequency f2 is also high.

[0057] [Embodiment 2] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be given to members having the same functions as those described in the first embodiment, and the description thereof will not be repeated.

[0058] In the above-described first embodiment, the processor 11 calculates the matrix product mtx by singular value decomposition in step S15 of FIG. pre In this embodiment, the processor 11 generates the variance-covariance matrix f cov By LU decomposition, we obtain the matrix product mtx pre For example, the processor 11 generates a variance-covariance matrix f cov Matrix multiplication mtx by Cholesky decomposition pre Generate.

[0059] In this embodiment, the imitation data generating device 10 generates imitation data that maintains the correlation between the resonance frequencies of the measurement data, as in the above-described embodiment 1. In other words, according to this embodiment, the imitation data generating device 10 can generate a natural imitation data group that does not deviate from the actually measured data.

[0060] [Embodiment 3] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be given to members having the same functions as those described in the first and second embodiments above, and the description thereof will not be repeated.

[0061] Fig. 9 is a flowchart illustrating an imitation data generation method M10B performed in this embodiment by the processor 11. The flowchart shown in Fig. 9 includes processes of steps S12, S14B, and S17B instead of the processes of steps S12, S14, and S17 in the flowchart shown in Fig. 6.

[0062] In step S12B, the processor 11 calculates the sample matrix f raw Standardized matrix f norm Calculate.

[0063] In step S14B, the processor 11 calculates the random number matrix mtx rand Generate the random matrix mtx rand by a predetermined coefficient. The predetermined coefficient is, for example, 0.1 or 0.01. In this operation example, the processor 11 multiplies a random number matrix mtx rand Generate the random matrix mtx rand Multiply by a given coefficient. A random matrix mtx with multiple elements in each row having the same value. rand In other words, it indicates that the random numbers used in one sample are common.

[0064] In step S17B, the processor 11 calculates the post-operation matrix mtx post Then, we perform the inverse transformation of equation (3) to obtain the pseudo-matrix f generated As an example, the processor 11 generates the pseudo-matrix f generated Generate.

number

[0065] Fig. 10 is a diagram illustrating an example of an artificial data group D23 generated by performing the process shown in Fig. 9 on the measurement data group D11 of Fig. 3. In the diagram, the horizontal axis indicates the first resonance frequency [Hz], and the vertical axis indicates the second resonance frequency [Hz]. Comparing the measurement data group D11 of Fig. 3 with the artificial data group D23 of Fig. 10, the artificial data group D23 is a data group that is slightly shifted from the measurement data group D11, which is raw data.

[0066] FIG. 11 is a schematic diagram for explaining the shift amount of the measurement data according to this embodiment. In the figure, the horizontal axis is the first-order resonance frequency [Hz], and the vertical axis is the second-order resonance frequency [Hz]. In the example of FIG. 11, the measurement data is shifted by the amount of the arrow A21. In the figure, if the measurement data is shifted by the amount of the normal distribution, the measurement data may be shifted too much. Therefore, in this embodiment, the random number matrix mtx randBy multiplying by an appropriate coefficient (for example, 0.1), the measurement data is prevented from being shifted too much. In this embodiment, the value of the random number for calculating the shift amount is the same regardless of the frequency band (+1, +1 in FIG. 11). Therefore, the first resonance frequency and the second resonance frequency are shifted in the same way.

[0067] In the above-described first embodiment, the pseudo matrix f generated When generating the matrix ave(f raw ) to the post-operation matrix mtx post Add to create the imitation matrix f generated In contrast, in this embodiment, the processor 11 calculates the matrix ave(f raw ) but the raw sample matrix f raw After the operation, the matrix mtx post Add to create the imitation matrix f generated In this way, according to this embodiment, it is possible to generate imitation data that appears as if the measurement data has only fluctuated slightly.

[0068] [Embodiment 4] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be given to members having the same functions as those described in the first to third embodiments above, and the description thereof will not be repeated.

[0069] In the above-described third embodiment, in step S14B of FIG. 9, the processor 11 generates a random number matrix mtx rand In contrast, in this embodiment, the processor 11 randomly generates the values ​​of each component. rand The values ​​of some of the components of mtx may be the same, or all of the components may be different. For example, the processor 11 generates a random number matrix mtx by a predetermined random function. rand Generate each component of

[0070] FIG. 12 is a schematic diagram for explaining the shift amount of the measurement data. In the figure, the horizontal axis is the first-order resonance frequency [Hz], and the vertical axis is the second-order resonance frequency [Hz]. In the figure, if the measurement data is shifted by the normal distribution amount, the measurement data may be shifted too much. Therefore, in this embodiment, the random number matrix mtx rand By multiplying by an appropriate coefficient (for example, 0.1), the measurement data is prevented from being shifted too much. In this embodiment, the random number value for calculating the shift amount is −1 or +1 in the example of FIG. 12, which differs depending on the frequency band, so the first resonance frequency is shifted by the arrow A31 so that the first resonance frequency decreases and the second resonance frequency increases.

[0071] Fig. 13 is a diagram showing an example of the imitation data group D24 generated by the processor 11 in this embodiment. In the diagram, the horizontal axis indicates the first resonance frequency [Hz], and the vertical axis indicates the second resonance frequency [Hz]. Comparing the imitation data group D24 in Fig. 13 with the imitation data group D23 in Fig. 10, the imitation data group D24 is data in which the relationship between the resonance frequencies is slightly disturbed.

[0072] [Embodiment 5] Other embodiments of the present invention will be described below. For ease of explanation, the same reference numerals will be given to members having the same functions as those described in the first to fourth embodiments above, and the description thereof will not be repeated.

[0073] Fig. 14 is a flowchart showing an example of an imitation data generation method M10C performed by the processor 11 in this embodiment. The flowchart shown in Fig. 14 includes processes of steps S12C, S14C, and S17C instead of the processes of steps S12B, S14B, and S17B in the flowchart shown in Fig. 9.

[0074] In step S12C, the processor 11 calculates the sample matrix f raw The standard matrix f is obtained by subtracting the mean value of each column from each component. norm Calculate.

[0075] In step S14C, the processor 11 calculates the random number matrix mtx rand Generate the random matrix mtx rand by a predetermined coefficient. The predetermined coefficient is, for example, 0.1 or 0.01. In this operation example, the processor 11 multiplies a random number matrix mtx rand , or a random matrix mtx with different values ​​for each element rand Generate the random matrix mtx rand Multiply by a given coefficient. A random matrix mtx with multiple elements in each row having the same value. rand In other words, the random numbers used in one sample are the same. On the other hand, the random number matrix mtx rand In other words, different random numbers are used in one sample.

[0076] In step S17C, the processor 11 calculates the post-operation matrix mtx post In the sample matrix f raw Add to create the imitation matrix f generated As an example, the processor 11 generates the pseudo-matrix f generated Generate.

number

[0077] Figures 15 and 16 are diagrams showing an example of an imitation data group generated by performing the process shown in Figure 14 on the measurement data group D11 of Figure 3. The imitation data group D25 of Figure 15 is an imitation data group generated by the processor 11 when the same random number is used for one sample in step S14C of Figure 14. On the other hand, the imitation data group D26 of Figure 16 is an imitation data group generated by the processor 11 when different random numbers are used for one sample in step S14C of Figure 14.

[0078] Comparing the measurement data group D11 of Fig. 3 with the imitation data group D25 of Fig. 15, the imitation data group D25 has a larger variation in data because the standard deviation conversion is not performed. Also, comparing the measurement data group D11 with the imitation data group D26, the relationship between frequencies is slightly disturbed by the random number, and the variation is larger because the standard deviation conversion is not performed. In other words, the imitation data group D26 is a data group with a larger variation than the imitation data group D25.

[0079] [Additional Note 1] In each of the above-described embodiments, the processor 11 executes both the process of generating measurement data and the process of generating imitation data, but the measurement data may be generated by a device other than the imitation data generating device 10. In this case, the imitation data generating device 10 may acquire the measurement data generated by the other device via the input / output IF 14 or the communication IF 15.

[0080] In each of the above-described embodiments, the processor 11 calculates the random number matrix mtx rand (step S14 in FIG. 6, etc.), but a device other than the counterfeit data generating device 10 generates the random number matrix mtx rand In this case, the fake data generating device 10 may generate a random number matrix mtx generated by another device. rand is acquired via the input / output IF 14 or the communication IF 15.

[0081] 〔summary〕 The imitation data generation device according to the first aspect includes one or more processors, and the processor includes a step of acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth order (n is a natural number equal to or greater than 2) resonance frequency of a sample; and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group, wherein in the step of generating the imitation data group, the processor generates the imitation data group such that a distribution of the imitation data group reproduces a distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies.

[0082] According to the above configuration, the imitation data generating device can generate imitation data that represents the resonant frequency of the object to be inspected and that does not deviate from data that represents the actual measurement result.

[0083] The imitation data generating device according to aspect 2 has the following features in addition to the features of the imitation data generating device according to aspect 1. That is, in the appearance assessment device according to aspect 2, in the step of generating the imitation data group, the processor generates the imitation data group so as to reproduce the relative positional relationship between the measurement data in each of a plurality of axis directions representing the n-dimensional space.

[0084] According to the above configuration, the simulated data generating device can generate a group of simulated data that maintains the positional relationship in n-dimensional space of a plurality of resonant frequencies represented by the measurement data.

[0085] The imitation data generating device according to aspect 3 has the following features in addition to the features of the imitation data generating device according to aspect 1 or 2. That is, in the appearance assessment device according to aspect 3, the step of generating the imitation data group includes the steps of: calculating a standardized matrix by standardizing or normalizing a sample matrix, which is an m-row by n-column matrix in which each row represents measurement data of each of m samples (m is a natural number equal to or greater than 2), by the processor; calculating a variance-covariance matrix of the standardized matrix generated in the step of calculating the standardized matrix; generating a matrix product based on the variance-covariance matrix; multiplying the matrix product by a random number matrix and transposing it to generate a post-operation matrix; and performing an inverse transformation of the standardization or normalization on the post-operation matrix to generate an imitation matrix representing the imitation data group.

[0086] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0087] The imitation data generating device according to aspect 4 has the following features in addition to the features of the imitation data generating device according to aspect 3. That is, in the imitation data generating device according to aspect 4, in the step of generating the matrix product, the processor generates the matrix product by performing singular value decomposition or LU decomposition on the variance-covariance matrix.

[0088] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0089] The imitation data generating device according to aspect 5 has the following features in addition to the features of the imitation data generating device according to aspect 4. That is, in the imitation data generating device according to aspect 5, in the step of generating the matrix product, the processor calculates n eigenvectors and n eigenvalues ​​for each column of the variance-covariance matrix, and generates the matrix product based on the calculated eigenvectors and eigenvalues.

[0090] According to the above configuration, the imitation data generating device generates a group of imitation data in which part or all of the positional relationships in an n-dimensional space represented by a plurality of pairs of eigenvectors and eigenvalues ​​are maintained, thereby making it possible to generate imitation data that does not deviate from actually measured data as data representing the resonant frequency of an object.

[0091] The fake data generating device according to aspect 6 has the following features in addition to the features of the fake data generating device according to aspect 4. That is, in the fake data generating device according to aspect 6, in the step of generating the matrix product, the processor generates the matrix product by Cholesky decomposition.

[0092] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0093] The fake data generating device according to aspect 7 has the following features in addition to the features of the fake data generating device according to any one of aspects 3 to 6. That is, in the fake data generating device according to aspect 7, the processor further executes a step of generating the random number matrix based on a normal distribution.

[0094] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0095] The imitation data generating device according to aspect 8 has the following features in addition to the features of the imitation data generating device according to any one of aspects 3 to 7. That is, in the imitation data generating device according to aspect 8, in the step of calculating the standardized matrix, the processor calculates the standardized matrix by subtracting an average value of the components for each column of the sample matrix from each component, and in the step of generating the imitation matrix, the processor adds the average value of the components for each column of the post-operation matrix to each component to generate the imitation matrix.

[0096] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0097] The imitation data generating device according to aspect 9 has the following features in addition to the features of the imitation data generating device according to aspect 7. That is, in the imitation data generating device according to aspect 9, the processor, in the step of calculating the standardized matrix, calculates the standardized matrix by subtracting an average value of components for each column of the sample matrix from each component, in the step of generating the random number matrix, multiplies the generated random number matrix by a predetermined coefficient, and in the step of generating the imitation matrix, adds the sample matrix to the post-operation matrix to generate the imitation matrix.

[0098] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0099] The imitation data generating device according to aspect 10 has the following features in addition to the features of the imitation data generating device according to any one of aspects 7 to 9. That is, in the imitation data generating device according to aspect 10, in the step of generating the random number matrix, the processor generates, as the random number matrix, a matrix in which the values ​​of multiple components included in each row are equal.

[0100] According to the above configuration, the imitation data generating device can generate imitation data representing the resonant frequency of the object, the imitation data not deviating from the actually measured data.

[0101] The imitation data generation method of aspect 11 includes the steps of: one or more processors acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth (n is a natural number greater than or equal to 2) resonant frequency of a sample; and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group. In the step of generating the imitation data group, the processor generates the imitation data group such that the distribution of the imitation data group reproduces the distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies.

[0102] According to the above configuration, it is possible to generate pseudo data that does not deviate from actually measured data as data representing the resonance frequency of the object to be inspected.

[0103] The program of aspect 12 causes a computer to execute the steps of acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth (n is a natural number greater than or equal to 2) resonant frequency of a sample, and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group, wherein in the step of generating the imitation data group, the computer generates the imitation data group such that the distribution of the imitation data group reproduces the distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies.

[0104] According to the above configuration, it is possible to generate pseudo data that does not deviate from actually measured data as data representing the resonance frequency of the object to be inspected.

[0105] The imitation data generating device of aspect 13 includes one or more processors, and the processor executes the steps of acquiring a measurement data group consisting of a plurality of measurement data, each measurement data representing a 1st to nth (n is a natural number equal to or greater than 2) resonant frequency of a sample, and generating an imitation data group consisting of a plurality of imitation data that imitates the measurement data group, wherein the distribution of the imitation data group in an n-dimensional space of the 1st to nth resonant frequencies is approximately identical to the distribution of the measurement data group.

[0106] According to the above configuration, the imitation data generating device can generate imitation data that represents the resonant frequency of the object to be inspected and that does not deviate from data that represents the actual measurement result.

[0107] [Software implementation example] The functions of the imitation data generation device 10 (hereinafter referred to as the "device") can be realized by a program for causing a computer to function as the device, and a program for causing a computer to function as each control block of the device.

[0108] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program to realize each function described in each of the above embodiments.

[0109] The program may be non-transitory and may be recorded in one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be provided to the device via any wired or wireless transmission medium.

[0110] In addition, some or all of the functions of each of the control blocks can be realized by a logic circuit. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of each of the control blocks can be realized by, for example, a quantum computer.

[0111] Furthermore, each process described in each of the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI ​​may be executed by the control device or another device (for example, an edge computer or a cloud server).

[0112] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0113] 1. Fake Data Generation System 2. Sample 10. Fake Data Generator 11 Processors 12 Primary Memory 13 Secondary Memory 14 Input / Output Interface 15 Communication Interface 20 Vibration Generator 30 Vibration receiving device

Claims

1. One or more processors; The processor, acquiring a measurement data group consisting of a plurality of measurement data, each of which represents a 1st to nth order (n is a natural number equal to or greater than 2) resonant frequency of a sample; generating an imitation data group consisting of a plurality of imitation data by imitation of the measurement data group; In the step of generating the simulated data group, the processor generates the simulated data group such that a distribution of the simulated data group reproduces a distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies.

1. An imitation data generating device comprising:

2. In the step of generating the simulated data group, the processor generates the simulated data group so as to reproduce a part or all of a relative positional relationship between the measurement data in each of a plurality of axis directions representing the n-dimensional space.

2. The imitation data generating device according to claim 1.

3. The step of generating the simulated data group includes a step of calculating a standardization matrix by standardizing or normalizing a sample matrix, which is an m-row by n-column matrix in which each row represents measurement data of each of m samples (m is a natural number equal to or greater than 2), by the processor; calculating a variance-covariance matrix of the standardized matrix generated in the step of calculating the standardized matrix; generating a matrix product based on the variance-covariance matrix; generating a transposed post-operation matrix by multiplying the matrix product by a random number matrix; and performing an inverse transformation of the standardization or the normalization on the post-operation matrix to generate an imitation matrix representing the imitation data group.

3. The imitation data generating device according to claim 1 or 2.

4. In the step of generating the matrix product, the processor generates the matrix product by performing singular value decomposition or LU decomposition on the variance-covariance matrix.

4. The imitation data generating device according to claim 3.

5. In the step of generating the matrix product, the processor calculates n eigenvectors and n eigenvalues ​​for each column of the variance-covariance matrix, and generates the matrix product based on the calculated eigenvectors and eigenvalues.

5. The imitation data generating device according to claim 4.

6. In the step of generating the matrix product, the processor generates the matrix product by Cholesky decomposition.

5. The imitation data generating device according to claim 4.

7. The processor further performs the step of generating the random number matrix based on a normal distribution.

7. The imitation data generating device according to claim 3,

8. In the step of calculating the standardization matrix, the processor calculates the standardization matrix by subtracting an average value of elements in each column of the sample matrix from each element; In the step of generating the pseudo matrix, the average value of the components for each column of the post-operation matrix is ​​added to each component to generate the pseudo matrix.

8. The imitation data generating device according to claim 3, wherein the imitation data generating device is a device for generating imitation data.

9. In the step of calculating the standardization matrix, the processor calculates the standardization matrix by subtracting an average value of elements in each column of the sample matrix from each element; In the step of generating a random number matrix, the generated random number matrix is ​​multiplied by a predetermined coefficient; In the step of generating the pseudo matrix, the sample matrix is ​​added to the post-operation matrix to generate the pseudo matrix.

8. The imitation data generating device according to claim 7.

10. In the step of generating a random number matrix, the processor generates a matrix in which a plurality of components included in each row have equal values ​​as the random number matrix.

10. The imitation data generating device according to claim 7, wherein the imitation data generating device is a device for generating imitation data.

11. One or more processors acquiring a measurement data group consisting of a plurality of measurement data, each of which represents a 1st to nth order (n is a natural number equal to or greater than 2) resonant frequency of a sample; and generating an imitation data group consisting of a plurality of imitation data by imitation of the measurement data group, In the step of generating the simulated data group, the processor generates the simulated data group such that a distribution of the simulated data group reproduces a distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies. A method for generating fake data comprising the steps of:

12. On the computer, acquiring a measurement data group consisting of a plurality of measurement data, each of which represents a 1st to nth order (n is a natural number equal to or greater than 2) resonant frequency of a sample; generating an imitation data group consisting of a plurality of imitation data by imitation of the measurement data group; In the step of generating the simulated data group, the computer generates the simulated data group such that a distribution of the simulated data group reproduces a distribution of the measurement data group in an n-dimensional space of 1st to nth resonant frequencies. A program characterized by:

13. One or more processors; The processor, acquiring a measurement data group consisting of a plurality of measurement data, each of which represents a 1st to nth order (n is a natural number equal to or greater than 2) resonant frequency of a sample; generating an imitation data group consisting of a plurality of imitation data by imitation of the measurement data group; A distribution of the simulated data group in an n-dimensional space of 1st to nth resonant frequencies is substantially the same as a distribution of the measured data group.

1. An imitation data generating device comprising:

Images