TW202121205A - Threshold value generation device, threshold value generation method, and threshold value generation program - Google Patents

Abstract

A threshold value generation device (20) has: a threshold value candidate group generation unit (21) for generating a first threshold value candidate group that includes one or more threshold value candidates (C1-C13), on the basis of a plurality of abnormality levels respectively assigned to a plurality of samples; a first determination accuracy calculation unit (24) for calculating first determination accuracies (TPR, TNR) respectively for the one or more threshold value candidates (C1-C13) that are included in the first threshold value candidate group; a restriction designation unit (22) for designating restrictions on the first determination accuracies; a first threshold value selection unit (23) for selecting the one or more threshold value candidates from the first threshold value candidate group on the basis of the restrictions and generating a second threshold value candidate group that includes the selected one or more threshold value candidates; a second determination accuracy calculation unit (26) for calculating second determination accuracies (TPR, TNR) of the one or more threshold value candidates (C1-C6) that are included in the second threshold value candidate group; and a second threshold value selection unit (25) for outputting, as a final threshold value, the threshold value candidate that is selected from the second threshold value candidate group on the basis of the second determination accuracies.

Description

Threshold value generating device, threshold value generating method, and recording medium on which threshold value generating program is recorded

The present invention relates to a threshold value generating device, a threshold value generating method, and a recording medium on which a threshold value generating program is recorded.

Various methods have been developed that use acoustic sensors or vibration sensors to measure the operating sound or vibration of the machine and analyze the waveform to estimate the soundness of the machine. One of the most representative methods in the MT (Mahalanobis Taguchi) legal system (see, for example, Non-Patent Document 1). In the MT method, the distribution formed by the collection of normal samples in the feature space is used as the reference space to learn in advance. When determining, the normal or abnormal identification is performed according to how much the observed feature vector deviates from the reference space. [Prior technical literature] [Patent Literature]

[Non-Patent Document 1] Tachibayashi Kazuo, "Introduction to the Taguchi Method", pp.167-185, Nikkei University Publishing House, 2004 [Non-Patent Document 2] Yuma Koizumi and 4 others, "Unsupervised Detection of Anomalous Sound based on Deep Learning and the Neyman-Pearson Lemma", IEEE/ACM Transactions on Audio Speech and Language Processing, 2019.

[Problems to be solved by the invention]

The system obtained by the MT method represents the horse-style distance of how much a certain sample deviates from the reference space. Mahalanobis’ distance means that if it is small, the sample is close to normal, and if it is large, the sample is close to abnormal. That is, the horse distance system can be interpreted as a value indicating the abnormality of the sample. However, the method of setting the threshold value for identifying normal/abnormality of the abnormality degree has not been established. In most cases, it is necessary to make trial and error in order to find an appropriate threshold value.

The problem of setting the threshold as shown above is not only the classical technique shown in the MT method, but also the modern technique like in-depth learning. For example, Non-Patent Document 2 describes a method of using a variational autoencoder to learn the potential distribution of the characteristics of a normal sample set, and determine whether it is normal or abnormal by the degree of deviation from the distribution. However, the system output from the final variational autoencoder is a value indicating the degree of abnormality similarly to the MT method, and Non-Patent Document 2 does not specifically describe how to set an appropriate threshold value for the degree of abnormality.

In order to automatically determine the appropriate threshold for the degree of abnormality, it is appropriate to solve the following issues.

The first is to establish a method for making the threshold value reflect the user's orientation. In most cases, if the abnormality of the normal sample set and the abnormal sample set is plotted on the number line, as shown in Figure 1, the abnormality of the normal sample (marked by ○) and the abnormal sample (marked by X) will be generated. The area where the abnormality of the collection is mixed. In the situation shown above, there is no threshold value that can fully identify normal/abnormal. It is advisable to take into account the misjudgment rate of normal samples or the miss rate of abnormal samples (trade off). Which of these benchmarks should be emphasized depends on the user. Therefore, it is appropriate to establish a method that allows the threshold value generation process to reflect the user's orientation.

The second system establishes a method for determining the most suitable threshold after reflecting the above-mentioned user's orientation, that is, the constraints specified by the user. For example, if the "missing rate of abnormal samples becomes a threshold of 10%" is generated, in a simple way, after considering various threshold candidates, the missed rate of abnormal samples selected from among them becomes 10%. The method of the person whose missed view rate is closest to 10%. However, in this method, as shown in Figure 2, the abnormality of the normal sample and the abnormal sample can be completely separated on the number line. If the missed rate of the abnormal sample is less than 10% without side effects, it will happen. Although it is obvious that there is a "better threshold", it has not been selected. The same problem also occurs when the area where the abnormality of the normal sample and the abnormality of the abnormal sample are mixed is small, or the number of samples that can be used to determine the threshold is small. Therefore, it is appropriate to establish a method that can determine an appropriate threshold under any conditions.

The purpose of the present invention is to obtain an appropriate threshold value based on the specified constraints. [Means to solve the problem]

A threshold value generating device of one aspect of the present invention is provided with: generating the threshold value of the first threshold value candidate group including one or more threshold value candidates based on the plural abnormalities assigned to the plural samples. Candidate group generation unit; a first judgment accuracy calculation unit that calculates the first judgment accuracy of each of the one or more threshold value candidates included in the first threshold value candidate group; specifies the restriction on the first judgment accuracy Restriction designation unit; from the aforementioned first threshold value candidate group, one or more threshold value candidates are selected based on the aforementioned restrictions, and a second threshold value candidate including one or more threshold value candidates selected above is generated The first threshold value selection part of the group; the second judgment accuracy calculation part that calculates the second judgment accuracy of each of the one or more threshold value candidates included in the second threshold value candidate group; and 2 Threshold value candidate group The threshold value candidate selected based on the aforementioned second determination accuracy outputs the second threshold value selection unit as the final threshold value.

Another aspect of the threshold value generation method of the present invention includes the steps of generating a first threshold value candidate group including one or more threshold value candidates based on the plural abnormalities assigned to the plural samples respectively; The step of the first determination accuracy of each of the one or more threshold value candidates included in the first threshold value candidate group; the step of specifying the restriction on the first determination accuracy; the step of specifying the restriction on the first threshold value candidate; Group, select one or more threshold value candidates based on the aforementioned constraints, and generate a second threshold value candidate group including one or more threshold value candidates selected above; calculate the aforementioned second threshold value candidate group The step of the second determination accuracy of each of the one or more threshold value candidates included; and the threshold value candidate output selected by the second threshold value candidate group based on the second determination accuracy as the final Threshold steps. [Effects of Invention]

According to the present invention, it is possible to obtain an appropriate threshold value based on the specified restriction.

The following describes the threshold value generating device, the threshold value generating method, and the computer-readable recording medium on which the threshold value generating program is recorded in the embodiment of the present invention with reference to the drawings. The following embodiments are only examples, and various changes can be made within the scope of the present invention.

The threshold value generating device of this embodiment generates the threshold value used when determining whether the machine is in a normal state or an abnormal state. The threshold value generation device of this embodiment is based on the result of analyzing the waveform detected by, for example, the operating sound or vibration of the machine using an acoustic sensor or a vibration sensor (i.e., a measuring device). Abnormality, to generate the threshold value used to determine whether the machine is in a normal state or an abnormal state. In the present embodiment, as a specific application example, the virtual threshold value generation device performs the quality inspection of the motor based on the vibration of the motor as the machine. The threshold value generation device of this embodiment detects vibrations that occur in a motor, and if the abnormality obtained as a result of the analysis is above the threshold value, the motor is regarded as a defective product.

"Export of anomaly" At the beginning, a specific example of the method of calculating the degree of abnormality based on the vibration waveform of the motor as the sample is described. First, the threshold value generating device of this embodiment performs wavelet transformation on the vibration waveform measured by the measuring device, thereby generating the spectrogram as shown in FIG. 3. Fig. 3 is a diagram showing an example of the frequency spectrum of the vibration of the motor. In FIG. 3, the vertical axis represents time, the horizontal axis represents frequency, and the degree of density represents the intensity of vibration.

Fig. 4 is a diagram schematically showing the frequency spectrum shown in Fig. 3 in rows and columns. In FIG. 4, the vertical axis represents time, and the horizontal axis represents frequency. In Fig. 4, one square as a quadrangle corresponds to one point on the frequency spectrum. In FIG. 4, it is assumed that the value of power is allocated to each of the plural squares. However, for the convenience of drawing, the time/frequency resolutions in Fig. 3 and Fig. 4 are different from each other.

FIG. 5 is a diagram showing an example in which the rows and columns shown in FIG. 4 are regarded as a feature vector in which the feature of the tone color at each time is digitized. In FIG. 5, the vertical axis represents time, and the horizontal axis represents frequency. The rows and columns shown in Fig. 4 are shown in Fig. 5, which can be regarded as a feature vector that digitizes the feature of tone color at each moment. In this way, the complex feature vector shown in FIG. 5 is generated from the vibration waveform of one motor.

Among them, the method of obtaining the feature vector from the vibration waveform is not limited to the method by the wavelet transformation as described above. In terms of obtaining the feature vector from the vibration waveform, for example, Fourier transform, Filter bank analysis, Cepstrum analysis, or LPC (Linear Predictive Coefficient) analysis can also be used. Wait. In addition, in the method of obtaining a feature vector from a vibration waveform, various acoustic feature quantities may be combined to form a feature vector. Various acoustic characteristic quantities, such as peak value, RMS (Root Mean Square, root mean square) value, and fundamental frequency, etc.

Fig. 6 shows an abnormal degree model learner 10 that receives a set of complex eigenvectors obtained from a sample of a normal motor (ie, a normal sample) and a sample of an abnormal motor (ie, an abnormal sample) to generate an abnormal degree model. Figure. As shown in FIGS. 3 to 5, a plurality of samples including a normal sample and an abnormal sample are used for analysis, and the set of feature vectors obtained as a result is input to the abnormality model learner 10. The abnormal degree model refers to the one that takes a feature vector as an input, and calculates a single abnormal degree by performing some transformations on the feature vector and outputs it. The abnormality model learner 10 constitutes the abnormality model shown above by the set of the given feature vectors. The following describes a case where the abnormality model learner 10 uses Linear Discriminant Analysis (LDA) as a specific example of the abnormality model learner 10.

LDA provides a collection of normal-level feature vectors and a collection of abnormal-level feature vectors as learning materials, and based on the distribution of these feature vectors, a method of finding the projective vector where the difference between normal and abnormal is most emphasized. The abnormal degree model learner 10 first obtains the average vector of all the feature vectors contained in both the normal level and the abnormal level: [Number 1] μ . Similarly, the abnormality model learner 10 calculates the standard deviation for each element of all feature vectors. At this time, the vector storing the standard deviation of each element of all feature vectors is set as: [Number 2] σ . These vectors all become N-dimensional when the feature vector is N-dimensional (N is a positive integer).

Next, the abnormality model learner 10 uses the above vector [Number 3] μ,　　σ , Normalize all feature vectors. Normalization refers to adjusting the vector so that the average of each element of the vector becomes 0 and the standard deviation becomes 1.

Specifically, normalization refers to the use of vectors: [Number 4] a , Find: [Number 5] (A-μ)÷σ . Among them, "÷" is the division of each element of the vector.

Next, the abnormality model learner 10 normalizes all the feature vectors as shown above, and obtains the average vector of the normal level and the average vector of the abnormal level respectively: [Number 6] μ ₁ , μ ₂ .

Similarly, the abnormal degree model learner 10 is based on the average vector of the normal level and the average vector of the abnormal level to obtain the covariance matrix: [number 7] Σ ₁ , Σ ₂ .

Finally, the abnormal degree model learner 10 obtains the projective vector by the following equations (1) and (2): [Number 8] w.

[Number 9]

A certain feature vector is input: [Number 10] x When it is converted into a single abnormality degree d, the formula system becomes as shown in the following formula (3). The formula (3) is equivalent to the abnormal degree model in LDA.

[Number 11]

However, the analysis method used by the abnormality model learner 10 is not limited to LDA. The abnormality model learner 10 can use, for example, a support vector machine, a neural network, or a mixed normal distribution model. In addition, the abnormality model learner 10 uses both the normal level and the abnormal level as learning materials, but it is also possible to use a method of learning using only a single class (class) data. At this time, the abnormal degree model learner 10 can use, for example, the MT method, principal component analysis (PCA), autoencoder, or level 1 support vector machine.

Fig. 7 is a diagram showing the multiple abnormality degree obtained by one motor as one sample in the example shown in Figs. 3 to 6. In the above abnormal degree model, a single feature vector is converted into a single abnormal degree. In addition, from the vibration waveform of one motor, as shown in FIG. 5, a complex feature vector is obtained. Therefore, at this time, one motor system obtains a complex abnormality degree as shown in FIG. 7.

Fig. 8 is a diagram showing a representative value calculated based on a complex number abnormality degree acquired by a motor as a sample, and this is assigned as the abnormality degree of the motor. As shown in Figure 8, in order to simplify the final judgment of normal or abnormal, the complex abnormality obtained by one motor (as shown in Figure 7) is totaled by some method, and the calculated single representative value is assigned as the The abnormality of the motor. The simplest method to calculate the representative value is to set the average value of the complex abnormality as the representative value. In terms of available representative values, any statistics such as maximum value, standard deviation, or mode can be used.

"Threshold value generation device 20" The following uses the method described above to explain that one sample, that is, one motor is assigned one abnormality degree. As shown in Figures 1 and 2, if the number of motors obtained by multiple motors is the same as the number of these motors At the time of abnormality degree, the abnormality degree is used to automatically generate a threshold value generating device and method that reflects the appropriate threshold value of the user's orientation.

FIG. 9 is a block diagram schematically showing the configuration of the threshold value generating device 20 of this embodiment. The threshold value generation device 20 is a device that can implement the threshold value generation method of this embodiment. The threshold value generation device 20 includes: a threshold value candidate group generation unit 21, a restriction designation unit 22, a first threshold value selection unit 23, a first determination accuracy calculation unit 24, a second threshold value selection unit 25, and The second determination accuracy calculation unit 26.

The threshold value candidate group generation unit 21 generates a first threshold value candidate group including one or more threshold value candidates based on the plural abnormality degrees to which the plural samples are respectively assigned. Here, the plural samples are plural motors. The first determination accuracy calculation unit 24 calculates the first determination accuracy of each of one or more threshold value candidates included in the first threshold value candidate group.

The restriction specifying unit 22 specifies the restriction on the accuracy of the first determination. The restriction specifying unit 22 accepts, for example, an input of a numerical value, and determines the restriction based on the numerical value. The input of the numerical value is performed by the user, for example. The first threshold value selection unit 23 selects one or more threshold value candidates based on the constraints from the first threshold value candidate group, and generates a second threshold value that includes the selected one or more threshold value candidates. Limit candidate group. For example, the first threshold value selection unit 23 selects threshold value candidates satisfying the restriction from the first threshold value candidate group, and generates second threshold value candidates including one or more selected threshold value candidates group.

The second determination accuracy calculation unit 26 calculates the second determination accuracy of each of one or more threshold value candidates included in the second threshold value candidate group. The second threshold value selection unit 25 outputs the threshold value candidates selected by the second threshold value candidate group based on the second determination accuracy as the final threshold value. For example, the second threshold value selection unit 25 selects the threshold value candidate whose second determination accuracy becomes the largest from the second threshold value candidate group and outputs it as the final threshold value.

"Regulation Designation Department 22" The restriction on the first judgment accuracy specified by the restriction specifying unit 22 is determined based on, for example, a numerical value input by the user. The constraints are, for example, the conditions shown below. The user selects one of the constraints (A1) ~ (A4) shown below, and can freely specify the value E among them. (A1) The omission of abnormal samples is set to E% or less. (A2) Set the erroneous detection of a normal sample to E% or less. (A3) Set the detection rate of abnormal samples to E% or more. (A4) Set the detection rate of the normal sample to E% or more.

Users who want to give priority to avoiding and missing abnormal samples should select restriction (A1) and set the value E at this time to a smaller value. In addition, a user who wants to give priority to avoiding false detection of a normal sample should select restriction (A2) and set the value E at this time to a small value.

Here, quantities called TPR (True Positive Rate) and TNR (True Negative Rate) are introduced. "Positive" means a positive that is not normal, that is, the object to be tested. Therefore, here, "Positive" corresponds to abnormal samples. In other words, the TPR system means "the proportion of abnormal samples determined by the system to be abnormal". TNR means "the proportion of normal samples determined by the system to be normal".

For the constraints specified above, the problem of choosing the most suitable threshold that meets them can be interpreted as providing the constraint of "setting one of TPR and TNR to an arbitrary value or more." and then selecting the most suitable threshold The problem. For example, the restriction (A1) of "setting the omission of abnormal samples to 10% or less." can be achieved by selecting the optimal threshold under the restriction of "setting TPR to 90% or more." Similarly, the above four types of constraints (A1) to (A4) can be replaced with the constraints (B1) to (B4) of "setting one of TPR and TNR to an arbitrary value or more" as shown below. In other words, the constraints (A1) to (A4) are equivalent to the constraints (B1) to (B4), respectively. (B1) Set TPR to (100-E)% or more. (B2) Set the TNR to (100-E)% or more. (B3) Set TPR to E% or more. (B4) Set TNR to E% or more.

The following is a description of the situation where the restriction of "Set the omission of abnormal samples to 20% or less" is specified when the restriction (A) is selected and E=20% is input. The restriction (A1) can be replaced with the restriction of TPR and TNR (B1), that is, "Set TPR to 80% or more."

"Threshold value candidate group generation unit 21" FIG. 10 is a diagram showing examples of threshold value candidates C1 to C13 included in the first threshold value candidate group generated by the threshold value candidate group generating unit 21. FIG. 11 is a diagram showing another example of the threshold value candidates C21 to C25 included in the first threshold value candidate group generated by the threshold value candidate group generating unit 21. The threshold value candidate group generating unit 21 generates a first threshold value candidate group including one or more threshold value candidates in order to obtain a threshold value that satisfies the specified restriction. There are various methods for generating threshold value candidates, but for one example, as shown in FIG. 10, a method of enumerating the middle of all adjacent anomalies plotted on a number line as threshold value candidates can be used. That is, if the abnormality of m samples is supplied, m-1 threshold value candidates are generated. m is a positive integer. In Fig. 10, the abnormality of 14 samples was given, and as a result, 13 threshold value candidates C1 to C13 were generated. The advantage of this method is that, as shown in FIG. 11, when the abnormality of the normal sample and the abnormality of the abnormal sample are greatly deviated, the margin at the time of recognizing the two is generated and becomes the maximum threshold value candidate C23. In this way, the generalization performance for unknown samples is improved.

"First Threshold Value Selection Unit 23 and First Judgment Accuracy Calculation Unit 24" FIG. 12 shows an example of the first determination accuracy calculated by the first determination accuracy calculation unit 24 and the constraints specified by the restriction designation unit 22 as shown in Table 1. As shown in FIG. In the first threshold value selection unit 23, for the first threshold value candidate group obtained as described above, the first determination accuracy calculation unit 24 is used to obtain the first determination accuracy for each threshold value candidate . Here, the "TPR and TNR set" is used as a specific example of the first determination accuracy. The example in FIG. 12 is a case where the set of TPR and TNR is obtained as the first determination accuracy for the threshold value candidates C1 to C13. Among these threshold value candidates, those that satisfy the aforementioned "Set TPR to 80% or more." are selected as the second threshold value candidate group and output.

FIG. 13 is a flowchart showing the operations of the first threshold value selection unit 23 and the first determination accuracy calculation unit 24. The first threshold value selection unit 23 selects a threshold value candidate that has not been selected from the first threshold value candidate group (step S11), and the first determination accuracy calculation unit 24 selects the selected threshold value For the candidates, the first determination accuracy is calculated (step S12).

Next, the first threshold value selection unit 23 determines whether the first determination accuracy satisfies the specified constraint, and if the constraint is satisfied (YES in step S13), the threshold value candidate that satisfies the constraint is added to the second threshold value For the candidate group (step S14), it is determined whether all threshold value candidates have been selected (step S15). The first threshold value selection unit 23 is to determine whether all threshold value candidates have been selected without adding threshold value candidates to the second threshold value candidate group if the first determination accuracy does not satisfy the specified constraints (NO in step S13) Threshold value candidates (step S15).

If the first threshold value selection unit 23 selects all threshold value candidates (YES in step S15), it outputs the second threshold value candidate group to the second threshold value selection unit 25 (step S16), if any For the unselected threshold value candidate (NO in step S15), the process returns to step S11.

"Second Threshold Value Selection Unit 25 and Second Judgment Accuracy Calculation Unit 26" FIG. 14 shows an example of the second determination accuracy calculated by the second determination accuracy calculation unit 26 as shown in Table 2. FIG. The second judgment accuracy calculation unit 26 obtains the second judgment accuracy for the threshold value candidates C1 to C6 shown in FIG. 14. In the second threshold value selection unit 25, since the final threshold value is unambiguously selected from the second threshold value candidate group, the first determination accuracy refers to evaluating the threshold values on different scales. This evaluation is performed by the second determination accuracy calculation unit 26. Here, "the smaller value of TPR and TNR" is used as a specific example of the second determination accuracy. In the example shown in FIG. 14, the threshold value candidate with the highest second determination accuracy is C6. Therefore, this threshold value candidate C6 is output as the final threshold value.

FIG. 15 is a flowchart showing the operations of the second threshold value selection unit 25 and the second determination accuracy calculation unit 26. The second threshold value selection unit 25 selects one threshold value candidate that has not yet been selected from the second threshold value candidate group (step S21), and the second determination accuracy calculation unit 26 selects the selected threshold value The value candidate calculates the second judgment accuracy (step S22).

Next, the second threshold value selection unit 25 determines whether the second determination accuracy is greater than the maximum value of the second determination accuracy stored in the memory (step S23), and if it is greater (YES in step S23), it is stored (also That is, update) the maximum value of the second judgment accuracy (step S24), and it is judged whether all threshold value candidates have been selected (step S25). The second threshold value selection unit 25 determines whether all threshold value candidates have been selected without updating the maximum value of the second determination accuracy if the second determination accuracy does not satisfy the specified restriction (NO in step S23) (Step S25).

If the second threshold value selection unit 25 has selected all threshold value candidates (YES in step S25), it outputs the threshold value candidate whose second determination accuracy is the largest as the final threshold value (step S26). There is an unselected threshold value candidate (NO in step S25), and the process returns to step S21.

However, the evaluation scale used as the first determination accuracy and the second determination accuracy may be other than those obtained by TPR or TNR. For example, the evaluation scale system can use any statistics such as the accuracy of the positive solution, the fitness rate, and the F-score (F-score or F-measure), or a combination thereof.

"effect" As shown in the above description, if the threshold value generating device 20 of this embodiment is used, the threshold value is made to reflect the orientation of the user in the form of a restriction on the first judgment accuracy, and the second judgment accuracy that satisfies the restriction is used. To select the appropriate threshold value, so that the appropriate threshold value can be selected while reflecting the user's orientation.

In addition, the method of specifying the range of the numerical value in which the determination accuracy can be obtained is intuitively understood by the user, and the user's labor for adjusting the threshold value is small. In addition, by adopting the form of so-called numerical range designation, there is room for the system to select a more appropriate threshold within the range. In this way, both the reflection of the user's orientation and the optimization of the threshold value can be considered.

In addition, only threshold value candidates satisfying the constraints specified by the user are selected as the second threshold value candidates, and therefore the final threshold value is one whose orientation of the user is reliably reflected.

In addition, the threshold value that is finally output is limited to one, so there is no need for the user to select the final threshold value from among a plurality of presented threshold value candidates, and additional tasks such as the user's labor can be reduced.

In addition, if the proportions of normal/abnormal samples in the entire data are significantly different, for example, the accuracy of judgments such as the accuracy of the correct solution or the F value will decrease the reliability. However, since TPR and TNR are not affected by the ratio of normal/abnormal samples, they can generate highly reliable thresholds under various conditions.

In addition, "the value of the smaller of TPR and TNR" as the second criterion is for "any sample that is input is judged as normal." or "any sample that is input is judged as abnormal.", etc. The alternate threshold value that is not helpful must become zero. Therefore, avoiding the selection of unhelpful threshold value candidates as shown above, it can be expected to generate practical threshold values in various situations.

In addition, if the degree of abnormality of the normal sample and the abnormal sample can be completely separated on the number line, the threshold value can be set anywhere between the normal sample closest to the abnormal sample and the abnormal sample closest to the normal sample , The judgment accuracy is 100%. In the case shown above, similar to the objective function used to optimize the support vector machine, by setting the threshold in the middle of the two, the generalization performance for the unknown sample can be maximized.

"Modifications" FIG. 16 is a diagram showing an example of the hardware configuration of the threshold value generating device 20 of this embodiment. As shown in FIG. 16, the threshold value generating device 20 has a memory 32 storing a program, and a processor 31 such as a CPU (Central Processing Unit) that executes the program. The program system may include a threshold value generation program for implementing the threshold value generation method of this embodiment. All or part of the functions of the threshold value generating device 20 shown in FIG. 9 can be realized by a processor 31 that executes a program. All or part of the functions of the threshold value generating device 20 shown in FIG. 16 may also be realized by a semiconductor integrated circuit. In addition, the threshold value generating device 20 may also have: a display 34 as an interface as a display means for the user to specify constraints on the accuracy of the determination; input elements 35 such as a mouse, a keyboard, and a touch panel; and as a memory The hard disk of the device 33.

10: Abnormal degree model learner 20: Threshold value generation device 21: Threshold value candidate group generation department 22: Restriction Designation Department 23: The first threshold selection part 24: The first judgment accuracy calculation unit 25: The second threshold selection part 26: The second judgment accuracy calculation unit 31: processor 32: memory 33: Hard Disk 34: display 35: input components

[Figure 1] is a diagram showing an example of plotting the abnormality of a normal sample and an abnormal sample on a number line. [Figure 2] is a diagram showing other examples where the abnormality of the normal sample and the abnormal sample is plotted on the number line. [Fig. 3] is a diagram showing an example of the frequency spectrum of the vibration of the motor. [Fig. 4] is a diagram showing the frequency spectrum shown in Fig. 3 formed into rows and columns and represented in a model. [Fig. 5] is a diagram showing an example of the row and column shown in Fig. 4 as the feature vector of the timbre at each time. [Figure 6] is a diagram showing an abnormality model learner that collects a set of complex eigenvectors based on a normal sample and an abnormal sample, and generates an abnormality model. [Fig. 7] is a graph showing the multiple abnormality degree obtained from one sample in the example shown in Figs. 3 to 6. [Fig. 8] is a graph showing that one representative value is calculated from the complex abnormality degree obtained from one sample, and one representative value is assigned as the abnormality degree of the sample. [Figure 9] is a functional block diagram schematically showing the configuration of the threshold value generating device of the embodiment of the present invention. [Fig. 10] is a diagram showing an example of threshold value candidates included in the first threshold value candidate group generated by the threshold value candidate group generating unit. [Fig. 11] is a diagram showing another example of threshold value candidates included in the second threshold value candidate group generated by the first threshold value selection unit. [FIG. 12] Table 1 shows an example of the first determination accuracy calculated by the first determination accuracy calculation unit and the constraints specified by the restriction designation unit. [Fig. 13] is a flowchart showing the operations of the first threshold value selection unit and the first determination accuracy calculation unit. [FIG. 14] is a graph shown in Table 2 with an example of the second determination accuracy calculated by the second determination accuracy calculation unit. [Fig. 15] is a flowchart showing the operation of the second threshold value selection unit and the second determination accuracy calculation unit. [Figure 16] is a diagram showing an example of the hardware configuration of the threshold value generation device of the embodiment.

20: Threshold value generation device

21: Threshold value candidate group generation department

22: Restriction Designation Department

23: The first threshold selection part

24: The first judgment accuracy calculation unit

25: The second threshold selection part

26: The second judgment accuracy calculation unit

Claims (11)

A threshold value generating device, which has: According to the plural abnormality degrees assigned to the plural samples, the threshold value candidate group generating unit for the first threshold value candidate group including one or more threshold value candidates is generated; A first determination accuracy calculation unit that calculates the first determination accuracy of each of the one or more threshold value candidates included in the first threshold value candidate group; Designation of the restriction designation unit that restricts the accuracy of the aforementioned first judgment; From the aforementioned first threshold value candidate group, one or more threshold value candidates are selected based on the aforementioned restrictions, and the first threshold value candidate group of the second threshold value candidate group including the one or more threshold value candidates selected above is generated. Limit selection department; A second determination accuracy calculation unit that calculates the second determination accuracy of each of the one or more threshold value candidates included in the second threshold value candidate group; and The second threshold value selection unit that outputs the threshold value candidate selected by the second threshold value candidate group based on the second determination accuracy as the final threshold value. For example, the threshold value generation device of claim 1, wherein the aforementioned restriction designation unit accepts the input of a numerical value, and determines the aforementioned restriction based on the aforementioned numerical value. For example, the threshold value generating device of claim 1 or 2, wherein the first threshold value selection unit selects the threshold value candidates satisfying the aforementioned restrictions from the first threshold value candidate group, and generates the threshold value candidates that include the selected ones. The above-mentioned second threshold value candidate group of one or more threshold value candidates that have been selected. For example, the threshold value generating device of claim 1 or 2, wherein the second threshold value selection unit selects the threshold value candidate whose second determination accuracy becomes the largest from the second threshold value candidate group, and outputs As the aforementioned final threshold. Such as the threshold value generating device of claim 1 or 2, wherein the aforementioned plural samples include: normal samples and abnormal samples, The first determination accuracy calculation unit outputs a set of TPR and TNR calculated based on the aforementioned complex abnormality degree as the first determination accuracy. Such as the threshold value generating device of claim 1 or 2, wherein the aforementioned plural samples include normal samples and abnormal samples, The second determination accuracy calculation unit outputs the smaller value of the TPR and TNR calculated based on the complex abnormality degree as the second determination accuracy. For example, the threshold value generation device of claim 5, wherein the second determination accuracy calculation unit outputs the smaller of the TPR and the TNR as the second determination accuracy. For example, the threshold value generating device of claim 1 or 2, wherein the threshold value candidate group generating unit is in the middle of adjacent abnormalities among the plurality of abnormalities arranged on a number line, and the first threshold is set Threshold value candidates included in the value candidate group. Such as the threshold value generating device of claim 1 or 2, wherein the aforementioned plural samples are plural machines, The aforementioned complex abnormality is the intensity of the sound or vibration emitted by the aforementioned complex machine. A threshold value generation method, which has: Steps of generating a first threshold value candidate group including one or more threshold value candidates according to the plural abnormality degrees allocated to the plural samples respectively; A step of calculating the first determination accuracy of each of the one or more threshold value candidates included in the first threshold value candidate group; Specify the step of restricting the accuracy of the aforementioned first judgment; The step of selecting one or more threshold value candidates from the aforementioned first threshold value candidate group based on the aforementioned restrictions, and generating a second threshold value candidate group including the aforementioned one or more threshold value candidates; A step of calculating the second determination accuracy of each of the one or more threshold value candidates included in the second threshold value candidate group; and The step of outputting the threshold value candidate selected by the second threshold value candidate group based on the second determination accuracy as the final threshold value. A recording medium on which a threshold value generation program is recorded. The threshold value generation program causes a computer to perform the following processing: According to the multiple abnormality degrees assigned to the multiple samples, the process of generating the first threshold value candidate group including one or more threshold value candidates; A process of calculating the first determination accuracy of each of the one or more threshold value candidates included in the first threshold value candidate group; Specify the processing to restrict the accuracy of the aforementioned first judgment; From the aforementioned first threshold value candidate group, one or more threshold value candidates are selected based on the aforementioned restrictions, and a second threshold value candidate group including the aforementioned one or more threshold value candidates is generated; The process of calculating the second determination accuracy of each of the one or more threshold value candidates included in the second threshold value candidate group; and The threshold value candidate selected by the second threshold value candidate group based on the second determination accuracy is output as the final threshold value.

Images