JP2024066391A - GENERATION SYSTEM, PROGRAM, AND METHOD FOR GENERATION OF WAVEFORM EVALUATION MODEL - Google Patents

Abstract

A generation system, a program, and a method for generating a waveform evaluation model are provided.
[Solution] In a system 10 equipped with a computer functioning as a generation unit 13 including a waveform acquisition unit that acquires waveform data, an intention identification unit that identifies a user's intention, and a pseudo-waveform generation unit that generates pseudo-waveform data from the acquired waveform data by reflecting the identified user's intention, a waveform evaluation model generation method executed by the computer includes a waveform data acquisition stage that acquires waveform data, an intention identification stage that identifies the user's intention, a pseudo-waveform generation stage that generates pseudo-waveform data from the waveform data acquired in the waveform data acquisition stage in a manner that reflects the user's intention identified in the intention identification stage, and a learning execution stage that performs machine learning using the pseudo-waveform data generated in the pseudo-waveform generation stage to generate a waveform evaluation model that outputs the evaluation result of input waveform data.
[Selected figure] Figure 5

Description

The present invention relates to a generation system, a program, and a method for generating a waveform evaluation model.

Non-Patent Documents 1 and 2 describe a technique called data augmentation, which generates new data from existing data.
[Prior Art Literature]
[Non-Patent Literature]
[Non-Patent Document 1] Zhao, Zhengli, Zizhao, Zhang, Ting, Chen, Sameer, Singh, and Han, Zhang. "Image Augmentations for GAN Training." (2020).
[Non-Patent Document 2] Brian Kenji Iwana, and Seiichi Uchida. "An empirical survey of data augmentation for time series classification with neural networks". PLOS ONE 16, no.7 (2021): e0254841.

According to one embodiment of the present invention, a generation system is provided. The generation system may include a waveform acquisition unit that acquires waveform data. The generation system may include an intention identification unit that identifies a user's intention. The generation system may include a pseudo-waveform generation unit that generates pseudo-waveform data from the waveform data acquired by the waveform data acquisition unit in a manner that reflects the user's intention identified by the intention identification unit.

The generation system may include a similarity display control unit that controls display data indicating a similarity between the pseudo waveform data generated from the waveform data by the pseudo waveform generation unit and the waveform data from which the pseudo waveform data was generated, to be displayed to the user. The pseudo waveform generation unit may generate a plurality of pseudo waveform data from the waveform data by using at least one of a plurality of algorithms, a plurality of parameter settings, and a random number, and the similarity display control unit may control display data indicating a similarity between the plurality of pseudo waveform data generated from the waveform data by the pseudo waveform generation unit and the waveform data from which the plurality of pseudo waveform data was generated, to be displayed to the user. The intention identification unit may identify a plurality of combinations of algorithms and parameter settings according to an instruction from the user, and the pseudo-waveform generation unit may generate a pseudo-waveform data set including the plurality of pseudo-waveform data from the waveform data using the algorithm and the parameter setting for each of the plurality of combinations identified by the intention identification unit, and the similarity display control unit may control the display data indicating the similarity of each of the plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit to the waveform data from which the pseudo-waveform data set was generated, to be displayed to the user. The generation system may include a pseudo-waveform storage unit that stores the plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit by associating each of the plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit with recipe information capable of identifying a combination of the algorithm and the parameter setting used to generate the pseudo-waveform data set.

Any of the generation systems may include a pseudo-waveform storage unit that stores the multiple pseudo-waveform data sets generated by the pseudo-waveform generation unit by associating each of the multiple pseudo-waveform data sets with original waveform identification information that can identify the waveform data that was the source of generating the pseudo-waveform data set.

Any of the generation systems may include a range specification receiving unit that receives a specification by the user of a range of similarity for the display data displayed by the similarity display control unit, a range waveform selection unit that selects a plurality of pseudo waveform data corresponding to the range of similarity for which the range specification receiving unit has received the specification, and a pseudo waveform storage unit that stores a pseudo waveform data set including the plurality of pseudo waveform data selected by the range waveform selection unit.

In any of the generation systems, the similarity display control unit may control the display data indicating the similarity between the pseudo waveform data and the waveform data from which the pseudo waveform data was generated, and other display data indicating the similarity between the pseudo waveform data and other waveform data different from the waveform data, to be displayed to the user.

Any of the generation systems may include a preprocessing unit that performs at least one of a plurality of preprocessing processes on the waveform data, and a pseudo-waveform storage unit that stores a pseudo-waveform data set including a plurality of pseudo-waveform data generated by the pseudo-waveform generation unit from the waveform data that has been preprocessed by the preprocessing unit, in association with preprocessing identification information that can identify the preprocessing.

Any of the generation systems may include a preprocessing unit that performs at least one of a plurality of preprocessing processes on the waveform data, a learning execution unit that performs machine learning using the pseudo waveform data generated by the pseudo waveform generation unit from the waveform data that has been preprocessed by the preprocessing unit to generate a waveform evaluation model that outputs an evaluation result of input waveform data, and a waveform evaluation model storage unit that stores the waveform evaluation model by associating preprocessing identification information that can identify the preprocessing with the waveform evaluation model generated by the learning execution unit. The generation system may include a model acquisition unit that acquires the waveform evaluation model associated with the preprocessing identification information stored in the waveform evaluation model storage unit, a waveform input unit that performs preprocessing indicated by the preprocessing identification information on input waveform data and inputs the data to the waveform evaluation model, and an evaluation result output control unit that controls the output of the evaluation result of the input waveform data output from the waveform evaluation model.

Any of the generation systems may include a pseudo-waveform storage unit that stores the pseudo-waveform data generated by the pseudo-waveform generation unit by associating the pseudo-waveform data with pseudo-identification information indicating that the pseudo-waveform data is data generated from the waveform data, and a learning execution unit that performs machine learning using the pseudo-waveform data generated by the pseudo-waveform generation unit to generate a waveform evaluation model that outputs an evaluation result of input waveform data, and evaluates the waveform evaluation model without using the pseudo-waveform data.

According to one embodiment of the present invention, a program for causing a computer to function as the generation system is provided. According to one embodiment of the present invention, a computer-readable storage medium storing the program is provided.

According to one embodiment of the present invention, a method for generating a waveform evaluation model executed by a computer is provided. The generation method may include a waveform data acquisition step of acquiring waveform data. The generation method may include an intention identification step of identifying a user's intention. The generation method may include a pseudo-waveform generation step of generating pseudo-waveform data from the waveform data acquired in the waveform data acquisition step in a manner that reflects the user's intention identified in the intention identification step. The generation method may include a learning execution step of executing machine learning using the pseudo-waveform data generated in the pseudo-waveform generation step to generate a waveform evaluation model that outputs an evaluation result of input waveform data.

Note that the above summary of the invention does not list all of the necessary features of the present invention. Subcombinations of these features may also be inventions.

FIG. 2 is an explanatory diagram for explaining processing in the system 10. 13 shows an example of pseudo abnormal waveform data. 13 shows an example of a display related to a pseudo waveform confirmation process. 13 shows an example of a display related to a pseudo waveform confirmation process. 2 illustrates an example of a functional configuration of the system 10. 1 shows an example of a flow of a process for generating a waveform evaluation model by the system 10. 1 shows a schematic diagram of an example of management data 190. FIG. 11 is an explanatory diagram for explaining an expansion processing algorithm. FIG. 11 is an explanatory diagram for explaining an expansion/contraction processing algorithm. FIG. 11 is an explanatory diagram for explaining a scale random deformation algorithm. 10 shows an example of display data 400 displayed by the similarity display control unit 170. An example of the hardware configuration of the system 10 or a computer 1200 that functions as a part of the system 10 is shown in schematic form.

The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

FIG. 1 is an explanatory diagram for explaining the processing in system 10. System 10 may be realized by one device. System 10 may be realized by multiple devices. System 10 may be an example of a generation system.

The system 10 generates a pseudo waveform from an existing waveform. For example, the system 10 has a function for generating pseudo abnormal waveforms, which enables AI to learn even with a small amount of abnormal data. The system 10 has a pseudo abnormal waveform generation function that generates and outputs a pseudo abnormal waveform. The system 10 may have a classifier learning function that uses the pseudo abnormal waveform to train a classifier.

For example, abnormal data is often not measured sufficiently in the pre-launch process. In this situation, even if an AI model is generated to realize automation in the inspection process (automatically determining whether something is normal or abnormal), sufficient performance is often not obtained. This is because the model is prone to over-learning to a small amount of abnormal data, and the accuracy of the judgment is not sufficient.

In response to this, system 10 processes the abnormal waveform in a way that does not significantly impair its physical meaning, generating a pseudo-abnormal waveform. System 10 generates a large amount of pseudo-abnormal waveforms that have been processed in various ways from a small amount of abnormal waveforms.

For example, in a product manufacturing system, there are many situations where you want to inspect manufactured products for defects, but the manufacturing process has just been launched and there is only a small amount of abnormal data. To accurately determine whether a product is normal or abnormal, a certain amount of abnormal data is required, but if you continue to collect the required amount, in a manufacturing process with a process capacity of 3σ (probability of not exceeding specifications is approximately 0.3%), abnormal data will appear in 3 out of 1,000 products, and if you try to collect 100 pieces of abnormal data, you will have to manufacture 33,000 products. This means that it will take a long time to launch the AI for inspection.

According to the system 10 of this embodiment, a large amount of pseudo-abnormal data can be generated at a stage where a small amount of abnormal data has been collected, making it possible to generate an AI classifier with sufficient accuracy, thereby contributing to the early launch of an automated inspection process.

The system 10 may generate a pseudo-normal waveform from a normal waveform, not limited to an abnormal waveform. The system 10 may generate a pseudo-normal waveform from any waveform, not limited to an abnormal waveform or normal waveform.

The system 10 processes fixed-length waveform data acquired periodically, for example, by a motor that performs a repetitive operation. The system 10 processes data that has periodic waveforms and shows characteristics after conversion to the frequency domain, for example.

The system 10 includes a waveform storage unit 102. The waveform storage unit 102 stores waveform data that is the source of generating the pseudo waveform data. The waveform data may be time-series data. For example, the waveform storage unit 102 stores waveform data generated by observing or measuring the target system.

The waveform storage unit 102 may store abnormal waveform data generated by observing or measuring the target system when an abnormality occurs in the target system itself. The waveform storage unit 102 may store, for example, abnormal waveform data generated by observing or measuring the target system when an abnormality is deliberately caused in the target system. When the target system is a system that manufactures a product, the waveform storage unit 102 may store abnormal waveform data generated by observing or measuring the target system when an abnormality occurs in the product, or abnormal waveform data generated by observing or measuring the product.

The waveform storage unit 102 may store normal waveform data used to generate AI that distinguishes between normal and abnormal waveforms. The waveform storage unit 102 may store normal waveform data generated by observing and measuring the target system when the target system is operating normally. When the target system is a system that manufactures a product, the waveform storage unit 102 may store normal waveform data generated by observing and measuring the target system when the product is normal, or normal waveform data generated by observing and measuring the product.

The waveform storage unit 102 may store any waveform data, regardless of whether it is normal waveform data or abnormal waveform data.

The system 10 may perform an augmentation process 104 on the waveform data stored in the waveform storage unit 102. The augmentation process 104 may be performed by a pseudo-waveform generating unit included in the system 10.

The pseudo waveform generating unit generates various pseudo abnormal waveform data from the abnormal waveform data stored in the waveform storage unit 102, for example. The pseudo waveform generating unit generates pseudo abnormal waveform data by, for example, adding noise to the abnormal waveform data, expanding it in the vertical axis direction of the graph, shrinking it in the vertical axis direction of the graph, expanding it in the horizontal axis direction of the graph, shrinking it in the horizontal axis direction of the graph, etc. As a specific example, as shown in FIG. 2, the pseudo waveform generating unit generates pseudo abnormal waveform data 204 by expanding abnormal waveform data 202 in the horizontal axis direction of the graph, and generates pseudo abnormal waveform data 206 by expanding abnormal waveform data 202 in the vertical axis direction of the graph.

As a specific example, the pseudo-waveform generating unit generates approximately 100 pieces of pseudo-abnormal waveform data from approximately 10 pieces of abnormal waveform data.

The pseudo-waveform generating unit may similarly generate various pseudo-normal waveform data from normal waveform data. The pseudo-waveform generating unit may similarly generate various pseudo-waveform data from arbitrary waveform data.

The learning data storage unit 106 stores learning data for training the AI to distinguish between normal and abnormal waveforms. The learning data storage unit 106 may store the waveform data stored in the waveform storage unit 102 and the pseudo waveform data generated by the pseudo waveform generation unit as learning data. For example, the learning data storage unit 106 stores the normal waveform data and a small amount of abnormal waveform data stored in the waveform storage unit 102, and the pseudo abnormal waveform data generated by the pseudo waveform generation unit from the small amount of abnormal waveform data.

The system 10 may include an input receiving unit that receives user input 108 from a user of the system 10. The system 10 may also include an algorithm storage unit 110 that stores a plurality of algorithms for generating pseudo-waveform data from original waveform data.

The pseudo-waveform generating unit may execute a selection process 112 to select one or more algorithms from a plurality of algorithms stored in the algorithm storage unit 110 in response to the user input 108 received by the input receiving unit.

The pseudo-waveform generating unit may execute a parameter setting process 114 for setting parameters for generating pseudo-waveform data. The pseudo-waveform generating unit may set multiple parameters for one algorithm, such as a large parameter swing, a medium parameter swing, and a small parameter swing. The pseudo-waveform generating unit may execute the parameter setting process 114 in response to a user input 108 received by the input receiving unit. The pseudo-waveform generating unit may execute the parameter setting process 114 automatically. The pseudo-waveform generating unit executes the parameter setting process 114, for example, based on the difference between normal waveform data and abnormal waveform data.

The pseudo-waveform generation unit executes a generation process 116 in which the parameters set by the parameter setting process 114 are applied to the algorithm selected by the selection process 112 to generate pseudo-waveform data. The pseudo-waveform generation unit generates pseudo-waveform data, for example, by sequentially applying multiple parameters set by the parameter setting process 114 to one algorithm selected by the selection process 112. For example, for a noise-adding algorithm, the pseudo-waveform generation unit generates pseudo-waveform data by setting large amplitude parameters, generates pseudo-waveform data by setting medium amplitude parameters, and generates pseudo-waveform data by setting small amplitude parameters. The pseudo-waveform generation unit stores the generated pseudo-waveform data in the data storage unit 118.

As a specific example, the pseudo waveform generation unit generates pseudo waveform data by adding a random numerical value based on a normal distribution with a mean of 0 and a standard deviation of σ to the waveform data. As another specific example, the pseudo waveform generation unit generates pseudo waveform data by multiplying the waveform data by a random numerical value based on a normal distribution with a mean of 1 and a standard deviation of σ. The pseudo waveform generation unit may set the standard deviation σ by the parameter setting process 114. The pseudo waveform generation unit sets one or more standard deviations σ according to, for example, a user instruction. When generating pseudo abnormal waveform data, the pseudo waveform generation unit sets the standard deviation σ based on, for example, the difference between normal waveform data and abnormal waveform data. For example, the pseudo waveform generation unit increases the standard deviation σ the greater the difference between normal waveform data and abnormal waveform data.

The system 10 may use all of the pseudo-waveform data stored in the data storage unit 118 as learning data, or, depending on the wishes of the user of the system 10, may use only a portion of the pseudo-waveform data stored in the data storage unit 118 as learning data.

The system 10 may include a pseudo-waveform confirmation processing unit. The pseudo-waveform confirmation processing unit executes a user intention/reflection process 120 for the pseudo-waveform data stored in the data storage unit 118.

The pseudo-waveform confirmation processing unit may execute a distribution display process 122 that displays to the user a distribution that indicates the characteristics of the pseudo-waveform data generated for each combination of algorithm and parameter setting. For example, the pseudo-waveform confirmation processing unit displays to the user a histogram or graph that visually displays the degree of variation with which the pseudo-waveform data for each combination of algorithm and parameter setting is generated for the waveform data that was the source of generating the pseudo-waveform data. As a specific example, the pseudo-waveform confirmation processing unit displays, for one algorithm, a histogram showing the correlation between multiple pseudo-waveform data generated by setting a first parameter and the original waveform data, a histogram showing the correlation between multiple pseudo-waveform data generated by setting a second parameter and the original waveform data, and a histogram showing the correlation between multiple pseudo-waveform data generated by setting a third parameter and the original waveform data. The pseudo-waveform confirmation processing unit may similarly display histograms of multiple pseudo-waveform data generated by setting each of multiple parameters for other algorithms. By viewing the display, the user can consider which of multiple combinations of algorithms and parameter settings to adopt.

For example, the pseudo-waveform confirmation processing unit displays to the user a histogram or graph visually showing the variation of pseudo-abnormal waveform data for each combination of algorithm and parameter setting with respect to the abnormal waveform data from which the pseudo-abnormal waveform data was generated. The pseudo-waveform confirmation processing unit may display the average waveform of the abnormal waveform data together with the average waveform of the pseudo-abnormal waveform data for each combination of algorithm and parameter setting. The pseudo-waveform confirmation processing unit may display to the user a histogram or graph showing the statistical difference between the abnormal waveform data and the pseudo-abnormal waveform data for each combination of algorithm and parameter setting from various perspectives.

For example, the pseudo-waveform confirmation processing unit may display to the user a histogram or graph visually showing the variation of the abnormal waveform data and the pseudo-abnormal waveform data for each combination of algorithm and parameter setting relative to the normal waveform data. The pseudo-waveform confirmation processing unit may display the average waveform of the normal waveform data, the average waveform of the abnormal waveform data, and the average waveform of the pseudo-abnormal waveform data for each combination of algorithm and parameter setting. The pseudo-waveform confirmation processing unit may display to the user a histogram or graph showing the statistical difference between the normal waveform data, the abnormal waveform data, and the pseudo-abnormal waveform data for each combination of algorithm and parameter setting from various perspectives. As a specific example, the pseudo-waveform confirmation processing unit may display to the user a histogram 212 and a waveform 214 as illustrated in FIG. 3.

The pseudo-waveform confirmation processing unit may execute a selection process 124 that selects at least a portion of the pseudo-waveform data whose distribution is displayed by the distribution display process 122, in accordance with an instruction from the user. The pseudo-waveform confirmation processing unit may execute an individual display process 126 that individually displays the pseudo-waveform data selected by the selection process 124. By viewing the display, the user can individually check the pseudo-waveform data, which can be used as material for deciding which of multiple combinations of algorithms and parameter settings to adopt. As a specific example, the pseudo-waveform confirmation processing unit may provide the user with a display 220 as shown in FIG. 4.

The pseudo waveform confirmation processing unit executes a range selection process 128 that selects the range of the distribution to be adopted or the range of the distribution not to be adopted, according to instructions from a user who has viewed the display by the distribution display process 122 and the individual display process 126. The pseudo waveform confirmation processing unit regards the selection result by the range selection process 128 as the user's intention, and reflects the data of the adopted range as learning data. The pseudo waveform confirmation processing unit may store the data of the adopted range in the learning data storage unit 130.

The system 10 may include a learning execution unit 140 that generates an AI 142 that distinguishes between normal waveforms and abnormal waveforms through machine learning. The learning execution unit 140 may generate the AI 142 by executing machine learning using the learning data stored in the learning data storage unit 106. The learning execution unit 140 may execute machine learning using the normal waveform data, abnormal waveform data, and pseudo-abnormal waveform data stored in the learning data storage unit 106. The learning execution unit 140 may change the number of pseudo-abnormal waveform data used for learning in response to a user's instruction or the like. The learning execution unit 140 may execute machine learning that reduces the weight of the pseudo-abnormal waveform data for the abnormal waveform data in response to a user's instruction or the like.

The learning execution unit 140 may generate the AI 142 by performing machine learning using the learning data stored in the learning data storage unit 130. The learning execution unit 140 may perform machine learning using the normal waveform data, abnormal waveform data, and pseudo-abnormal waveform data stored in the learning data storage unit 106. The learning execution unit 140 may change the number of pseudo-abnormal waveform data used for learning in response to a user's instruction or the like. The learning execution unit 140 may perform machine learning that reduces the weight of the pseudo-abnormal waveform data for the abnormal waveform data in response to a user's instruction or the like.

The system 10 may include an algorithm storage unit 144 that stores multiple learning algorithms. The learning execution unit 140 may use one of the multiple learning algorithms stored in the algorithm storage unit 144 to train the AI 142. The learning execution unit 140 uses, for example, a learning algorithm corresponding to the algorithm used to generate the pseudo waveform data among the algorithms stored in the algorithm storage unit 110. The learning execution unit 140 uses, for example, a learning algorithm corresponding to the parameter settings used to generate the pseudo waveform data. The learning execution unit 140 uses, for example, a learning algorithm corresponding to the algorithm used to generate the pseudo waveform data among the algorithms stored in the algorithm storage unit 110 and the parameter settings used to generate the pseudo waveform data. The learning execution unit 140 may use a learning algorithm selected according to an instruction from a user of the system 10.

The app 150 having the AI 142 is provided to a device that performs a discrimination process to discriminate between normal and abnormal waveforms, for example, and the device executes the app 150 to detect anomalies in the target system. The app 150 having the AI 142 may also be provided to a device for any other purpose, such as a device that performs failure prediction.

As described above, the system 10 may include a pseudo-waveform generating unit that generates pseudo-waveform data from waveform data. The pseudo-waveform generating unit may generate pseudo-waveform data from waveform data in a manner that reflects the user's intention. The pseudo-waveform generating unit may generate pseudo-waveform data from waveform data using an algorithm corresponding to the user's intention among a plurality of algorithms. For example, the pseudo-waveform generating unit may generate pseudo-waveform data from waveform data using one or more algorithms selected by the user. The pseudo-waveform generating unit may generate pseudo-waveform data from waveform data according to parameter settings corresponding to the user's intention. For example, the pseudo-waveform generating unit may generate pseudo-waveform data from waveform data according to one or more parameter settings selected by the user. The pseudo-waveform generating unit may generate pseudo-abnormal waveform data from abnormal waveform data. The pseudo-waveform generating unit may process a small amount of abnormal waveform data without significantly impairing the physical meaning, to generate a large amount of abnormal waveform data.

As described above, the system 10 may include a pseudo-waveform confirmation processing unit. The pseudo-waveform confirmation processing unit may display to the user a distribution representing the characteristics of the pseudo-waveform data generated by the pseudo-waveform generating unit for each combination of the algorithm and the parameter setting. For example, the pseudo-waveform confirmation processing unit may display to the user a histogram or graph visually displaying the degree of variation with which the pseudo-waveform data for each combination of the algorithm and the parameter setting is generated with respect to the waveform data from which the pseudo-waveform data was generated. The pseudo-waveform confirmation processing unit may select, as data to be used for learning, the pseudo-waveform data corresponding to a range selected by the user from the histogram displayed to the user. The pseudo-waveform confirmation processing unit may display the pseudo-waveform data corresponding to the range selected by the user from the histogram displayed to the user and the original waveform data one by one so that they can be compared. The pseudo-waveform confirmation processing unit may display the variation of the pseudo-waveform data corresponding to the range selected by the user and the variation of the original waveform data side by side.

As described above, the system 10 may include a learning execution unit that performs machine learning using the pseudo waveform data generated by the pseudo waveform generation unit. The learning execution unit may generate an AI that distinguishes between normal waveforms and abnormal waveforms by machine learning using normal waveform data, abnormal waveform data, and pseudo-abnormal waveform data generated from the abnormal waveform data by the pseudo waveform generation unit.

Figure 5 shows an example of a schematic functional configuration of the system 10. The system 10 includes a memory unit 11, a registration unit 12, a generation unit 13, a pseudo-waveform confirmation processing unit 14, a learning unit 15, and an evaluation execution unit 16. Note that it is not essential that the system 10 includes all of these.

The memory unit 11 may include a waveform memory unit 102, an algorithm memory unit 110, and a pseudo waveform memory unit 168. The waveform memory unit 102 stores waveform data. The waveform data may be time series data. The waveform data may be in any format. For example, the waveform data may be data expressed by a series of numerical values. The waveform memory unit 102 may store waveform data that is the source for generating the pseudo waveform data. The waveform memory unit 102 may store evaluation waveform data for evaluating the waveform evaluation model.

The algorithm storage unit 110 stores an algorithm for generating pseudo-waveform data from waveform data. The algorithm storage unit 110 may store multiple algorithms.

The pseudo waveform storage unit 168 stores the pseudo waveform data generated by the generation unit 13. The pseudo waveform storage unit 168 may function as the data storage unit 118.

The registration unit 12 performs various registrations. For example, the registration unit 12 registers waveform data. The registration unit 12 stores the registered waveform data in the waveform storage unit 102.

The registration unit 12 may register waveform data that is the source of generating the pseudo waveform data. For example, the registration unit 12 registers so-called raw waveform data. For example, the registration unit 12 registers waveform data generated by observing or measuring an arbitrary target system.

The registration unit 12 may register abnormal waveform data generated by observing or measuring the target system when an abnormality occurs in the target system itself. The registration unit 12 may register, for example, abnormal waveform data generated by observing or measuring the target system when an abnormality is deliberately caused in the target system. When the target system is a system that manufactures a product, the registration unit 12 may register abnormal waveform data generated by observing or measuring the target system when an abnormality occurs in the product, or abnormal waveform data generated by observing or measuring the product.

The registration unit 12 may register normal waveform data generated by observing or measuring the target system when the target system is operating normally. When the target system is a system that manufactures a product, the registration unit 12 may register normal waveform data generated by observing or measuring the target system when the product is normal, or normal waveform data generated by observing or measuring the product.

The generation unit 13 generates pseudo-waveform data from the waveform data stored in the waveform storage unit 102. The generation unit 13 may generate one piece of pseudo-waveform data from one piece of waveform data. The generation unit 13 may generate multiple pieces of pseudo-waveform data from one piece of waveform data. The generation unit 13 may generate multiple pieces of pseudo-waveform data from multiple pieces of waveform data. The generation unit 13 may generate a larger number of pieces of pseudo-waveform data than the number of waveform data from multiple pieces of waveform data. The generation unit 13 may include a waveform acquisition unit 160, an intention identification unit 162, a preprocessing unit 164, and a pseudo-waveform generation unit 166.

The waveform acquisition unit 160 acquires waveform data that is the source of generating the pseudo waveform data. The waveform acquisition unit 160 may acquire the waveform data from the waveform storage unit 102.

The intention identification unit 162 identifies the intention of the user. The user may be a person who uses the system 10. The intention identification unit 162 may identify the intention of the user regarding generating pseudo-waveform data from the waveform data.

The intention identification unit 162, for example, identifies an algorithm to be used when generating pseudo-waveform data from waveform data as the user's intention. The intention identification unit 162 may identify one or more algorithms from among the multiple algorithms stored in the algorithm storage unit 110 in accordance with a user instruction. The intention identification unit 162 may identify multiple algorithms to be successively applied to the waveform data and the order in which they are applied in accordance with a user instruction.

The intention identification unit 162, for example, identifies parameter settings used when generating pseudo-waveform data from waveform data as the user's intention. The intention identification unit 162 may identify parameter settings in accordance with a user's instruction. As a specific example, the intention identification unit 162 identifies one or more parameter settings by accepting a user's selection from among multiple parameter settings. Also, as a specific example, the intention identification unit 162 identifies one or more parameter settings by accepting a user's input of a parameter setting.

The intention identification unit 162 identifies, for example, a combination of an algorithm and parameter settings used when generating pseudo-waveform data from waveform data as the user's intention. The intention identification unit 162 may identify one or more combinations of algorithms and parameter settings in accordance with an instruction from the user.

For example, when generating pseudo-waveform data from waveform data, the intention identification unit 162 identifies pre-processing to be applied to the waveform data as the user's intention. The pre-processing may be a process for extracting features from the waveform data. Examples of pre-processing include FFT (Fast Fourier Transform), normalization, filtering, slicing, sampling frequency conversion, and dimensional conversion. The intention identification unit 162 may identify pre-processing to be successively applied to the waveform data and the order in which to apply the pre-processing in accordance with a user's instruction.

The intention identification unit 162 identifies, for example, a combination of at least one of the algorithms and parameter settings used when generating pseudo-waveform data from waveform data and pre-processing applied to the waveform data as the user's intention.

The preprocessing unit 164 applies at least one of a plurality of preprocessing processes to the waveform data acquired by the waveform acquisition unit 160. The preprocessing unit 164 may apply the preprocessing process identified by the intention identification unit 162 to the waveform data acquired by the waveform acquisition unit 160. When the intention identification unit 162 identifies a plurality of preprocessing processes to be successively applied to the waveform data and the order in which they are to be applied, the preprocessing unit 164 may apply the plurality of preprocessing processes to the waveform data in order according to the order.

The pseudo waveform generating unit 166 generates pseudo waveform data from the waveform data acquired by the waveform acquiring unit 160. For example, the pseudo waveform generating unit 166 generates pseudo abnormal waveform data from abnormal waveform data. For example, the pseudo waveform generating unit 166 generates pseudo normal waveform data from normal waveform data.

The pseudo-waveform generating unit 166 may generate multiple pieces of pseudo-waveform data from one piece of waveform data. The pseudo-waveform generating unit 166 may generate multiple pieces of waveform data from one piece of waveform data by using at least one of multiple algorithms, multiple parameter settings, and random numbers. For example, the pseudo-waveform generating unit 166 generates multiple pieces of pseudo-waveform data by using multiple algorithms for one piece of waveform data. For example, the pseudo-waveform generating unit 166 generates multiple pieces of pseudo-waveform data by using multiple parameter settings for one piece of waveform data. For example, the pseudo-waveform generating unit 166 generates multiple pieces of pseudo-waveform data by using random numbers for one piece of waveform data. By using multiple algorithms, multiple parameter settings, random numbers, etc., it is possible to generate pseudo-waveform data of various variations from one piece of waveform data.

The pseudo-waveform generating unit 166 may generate pseudo-waveform data from the waveform data in a manner that reflects the user's intention identified by the intention identifying unit 162. The pseudo-waveform generating unit 166 generates pseudo-waveform data from the waveform data, for example, using one or more algorithms identified by the intention identifying unit 162. When the intention identifying unit 162 identifies multiple algorithms to be successively applied to the waveform data and the order in which they are applied, the pseudo-waveform generating unit 166 may generate pseudo-waveform data from the waveform data using the multiple algorithms in accordance with the order. The pseudo-waveform generating unit 166 generates pseudo-waveform data from the waveform data, for example, using one or more parameter settings identified by the intention identifying unit 162. The pseudo-waveform generating unit 166 generates pseudo-waveform data from the waveform data, for example, using the algorithm and the parameter setting for each of multiple combinations of the algorithm and the parameter setting identified by the intention identifying unit 162. The pseudo-waveform generating unit 166 generates pseudo-waveform data from the waveform data after the pre-processing unit 164 has performed one or more pre-processing steps identified by the intention identifying unit 162. It may be difficult to generate pseudo-waveform data suitable for learning from the waveform data, but by making it possible to generate multiple types of pseudo-waveform data using a method that reflects the user's intention, it is possible to increase the possibility of generating pseudo-waveform data suitable for learning.

The pseudo waveform generating section 166 may process the waveform data so as not to significantly impair the physical meaning, and generate pseudo waveform data. The pseudo waveform generating section 166 may generate pseudo waveform data by adding noise to the waveform data, processing the entire or partial section of the waveform data, expanding the waveform data in the vertical axis direction of the graph, shrinking the waveform data in the vertical axis direction of the graph, expanding the waveform data in the horizontal axis direction of the graph, shrinking the waveform data in the horizontal axis direction of the graph, or the like.

The pseudo waveform generating unit 166 generates pseudo waveform data, for example, using an algorithm (sometimes referred to as a noise addition processing algorithm) that adds random values based on a normal distribution with a mean of 0 and a standard deviation of σ to time-series waveform data. When using a noise addition processing algorithm, the pseudo waveform generating unit 166 may generate multiple pieces of pseudo waveform data from one waveform data by using multiple random values. When using a noise addition processing algorithm, the pseudo waveform generating unit 166 may generate multiple pieces of pseudo waveform data from one waveform data by using multiple parameter settings, such as using multiple standard deviations σ.

The pseudo waveform generating unit 166 generates pseudo waveform data, for example, by using an algorithm (sometimes referred to as a scale transformation algorithm) that multiplies random values based on a normal distribution with a mean of 1 and a standard deviation of σ for time-series waveform data. When using a scale transformation algorithm, the pseudo waveform generating unit 166 may generate multiple pieces of pseudo waveform data from one waveform data by using multiple random values. When using a scale transformation algorithm, the pseudo waveform generating unit 166 may generate multiple pieces of pseudo waveform data from one waveform data by using multiple parameter settings, such as using multiple standard deviations σ.

The pseudo waveform generating unit 166, for example, determines a reduce_ratio, determines a start point within the range of "time series data size × (1 - reduce_ratio)" from 0 of the time series waveform data, sets the end point to a point that is "time series data size × reduce_ratio" ahead of the determined start point, cuts out the section from the start point to the end point, and generates pseudo waveform data using an algorithm (sometimes referred to as an expansion processing algorithm) that expands the cut-out portion to the original size. When using the expansion processing algorithm, the pseudo waveform generating unit 166 may generate multiple pseudo waveform data from one waveform data by using multiple reduction ratios.

The pseudo waveform generating unit 166, for example, creates candidates for warp_scale, randomly determines one value from the candidates for warp_scale, determines the window_ratio, calculates "warp_size = window_ratio x time series data size", randomly reserves an area of warp_size in the range from 1 to "size in time series direction - warp_size - 1", and generates "waveform before the warp_size area", "waveform of the warp_size area", and "warp_ The pseudo waveform generator 166 determines three regions, namely, "waveform before the warp_size region", "waveform after the warp_size region", and "waveform after the warp_size region", and enlarges or reduces the region of the "warp_size region waveform" to a size of "warp_size x warp_scales", recombines the "waveform before the warp_size region", the enlarged or reduced "warp_size region waveform", and "waveform after the warp_size region", and generates pseudo waveform data using an algorithm (sometimes referred to as an expansion/contraction processing algorithm) that returns the data to their original size. When using an expansion/contraction processing algorithm, the pseudo waveform generator 166 may generate multiple pseudo waveform data from one waveform data by using multiple parameter settings, such as creating multiple warp_scale candidates and using multiple warp_scale candidates or using multiple window_ratios.

For example, the pseudo-waveform generating unit 166 determines the number of via points, determines random values based on a normal distribution with an average value of 1 and a standard deviation of σ for the number of via points determined (y-axis information), divides the size in the time series direction by the number of via points to arrange the via points at equal intervals (x-axis information), creates a spline curve using the y-axis information and x-axis information with the size in the time series direction of the waveform data as the x-axis, and generates pseudo-waveform data using an algorithm (sometimes referred to as a scale random deformation algorithm) that multiplies the created spline curve and the waveform data. When using the scale random deformation algorithm, the pseudo-waveform generating unit 166 may generate multiple pseudo-waveform data from one waveform data by using multiple random values. When using the scale random deformation algorithm, the pseudo-waveform generating unit 166 may generate multiple pseudo-waveform data from one waveform data by using multiple parameter settings, such as using the number of via points in multiple patterns or using multiple standard deviations σ.

The pseudo-waveform generating unit 166 may generate a pseudo-waveform data set including multiple pseudo-waveform data from one or multiple waveform data. The pseudo-waveform generating unit 166 may generate various pseudo-waveform data sets by using different algorithms, different parameter settings, or different pre-processing.

For example, the pseudo-waveform generating unit 166 generates a pseudo-abnormal waveform data set including multiple pseudo-abnormal waveform data. The pseudo-waveform generating unit 166 may generate multiple pseudo-abnormal waveform data sets. For example, the pseudo-waveform generating unit 166 generates a pseudo-normal waveform data set including multiple pseudo-normal waveform data.

The pseudo waveform storage unit 168 stores the pseudo waveform data generated by the pseudo waveform generation unit 166. The pseudo waveform storage unit 168 may store the pseudo waveform data in association with pseudo identification information indicating that the data is pseudo-generated. By associating the pseudo waveform data with the pseudo identification information, it becomes possible to distinguish the pseudo waveform data from raw waveform data.

The pseudo waveform storage unit 168 may store the pseudo identification information in association with the pseudo waveform data set. The pseudo waveform storage unit 168 may store the pseudo identification information in association with abnormal identification information indicating that the data is abnormal in the pseudo abnormal waveform data set. The pseudo waveform storage unit 168 may store the pseudo identification information in association with normal identification information indicating that the data is normal in the pseudo normal waveform data set.

The pseudo-waveform storage unit 168 may store a plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit 166 in association with recipe information that can identify the algorithm, data set, and pre-processing used to generate the pseudo-waveform data set. For example, when the pseudo-waveform generation unit 166 generates a pseudo-waveform data set from waveform data using an algorithm and parameter settings for each of a plurality of combinations of an algorithm and a parameter setting, the pseudo-waveform storage unit 168 stores a plurality of pseudo-waveform data sets in association with each of the plurality of pseudo-waveform data sets in association with recipe information that can identify the combination of the algorithm and parameter setting used to generate the pseudo-waveform data set. This makes it possible to understand under what conditions the pseudo-waveform data set was generated by referring to the recipe information. In addition, the recipe information can facilitate management of the pseudo-waveform data sets.

The pseudo-waveform storage unit 168 may store a plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit 166 by associating each of the plurality of pseudo-waveform data sets with original waveform identification information capable of identifying the waveform data from which the pseudo-waveform data set was generated. When the pseudo-waveform generation unit 166 generates a plurality of pseudo-waveform data sets from one waveform data, the pseudo-waveform storage unit 168 may store the plurality of pseudo-waveform data sets by associating the plurality of pseudo-waveform data sets with original waveform identification information capable of identifying the one waveform data. When the pseudo-waveform generation unit 166 generates a plurality of pseudo-waveform data sets from a plurality of waveform data, the pseudo-waveform storage unit 168 may store the plurality of pseudo-waveform data sets by associating the plurality of pseudo-waveform data sets with original waveform identification information capable of identifying the plurality of sets of waveform data. This makes it possible to easily identify the original waveform data when it is desired to compare a certain pseudo-waveform data set with the original waveform data or when it is desired to use the original waveform data together with a certain pseudo-waveform data set for learning.

When the pseudo waveform generating unit 166 generates a pseudo waveform data set from waveform data that has been preprocessed by the preprocessing unit 164, the pseudo waveform storage unit 168 may store the pseudo waveform data set by associating the pseudo waveform data set with preprocessing identification information that can identify the preprocessing. This makes it possible to easily understand what preprocessing was performed on the original waveform data before the pseudo waveform data set was generated.

The pseudo waveform confirmation processing unit 14 executes a process that allows the user to confirm the pseudo waveform data stored in the pseudo waveform storage unit 168. The pseudo waveform confirmation processing unit 14 may include a similarity display control unit 170, a range specification receiving unit 172, and a waveform within range selection unit 174.

The similarity display control unit 170 controls the display of display data to the user, which indicates the similarity between the pseudo waveform data generated from the waveform data by the pseudo waveform generating unit 166 and the waveform data from which the pseudo waveform data was generated. The similarity display control unit 170 causes the display data to be displayed on a display provided in the system 10, for example. The similarity display control unit 170 may also transmit the display data to an arbitrary device and cause the device to display the display data.

The similarity display control unit 170 may control to display to the user display data indicating the similarity between the multiple pseudo waveform data generated from the waveform data by the pseudo waveform generating unit 166 and the waveform data from which the multiple pseudo waveform data were generated. The similarity display control unit 170 may control to display to the user display data indicating the degree of variation with which the multiple pseudo waveform data are generated with respect to the waveform data from which the multiple pseudo waveform data were generated. The similarity display control unit 170, for example, calculates the similarity between each of the multiple pseudo waveform data and the waveform data from which the multiple pseudo waveform data were generated, and displays display data including at least one of a histogram and a graph showing the number of pseudo waveform data for each similarity. The similarity may be a value that indicates, for example, that the closer to 1 the value is, the more similar the data is to the original waveform data, and the closer to 0 the value is, the more the data is deviated from the original waveform data.

The similarity display control unit 170 may, for example, use the correlation coefficient between the waveform data and the pseudo waveform data as the similarity between the waveform data and the pseudo waveform data. The similarity display control unit 170 may calculate the similarity between the waveform data and the pseudo waveform data in any manner as long as it can represent the degree of similarity between the waveform data and the pseudo waveform data.

When multiple pseudo waveform data are generated from one waveform data, the similarity display control unit 170 may calculate the similarity between each of the multiple pseudo waveform data and the waveform data. When multiple pseudo waveform data are generated from multiple waveform data, the similarity display control unit 170 may calculate the similarity between the pseudo waveform data and each of the multiple waveform data for each of the multiple pseudo waveform data, and use the average as the similarity between the waveform data and the pseudo waveform data.

The similarity display control unit 170 may control the display data indicating the similarity between the pseudo waveform data and the waveform data from which the pseudo waveform data was generated, and the other display data indicating the similarity between the pseudo waveform data and other waveform data different from the waveform data from which the pseudo waveform data was generated, to be displayed to the user. For example, the similarity display control unit 170 controls the display data indicating the similarity between the pseudo abnormal waveform data generated from the abnormal waveform data and the abnormal waveform data, and the display data indicating the similarity between the pseudo abnormal waveform data and normal waveform data to be displayed to the user. This allows the user to grasp the relationship between the generated pseudo abnormal waveform data and the normal waveform data, in addition to the relationship between the generated pseudo abnormal waveform data and the original abnormal waveform data. By making it possible to grasp the relationship between the pseudo abnormal waveform data and the normal waveform data, for example, it is possible to identify pseudo abnormal waveform data that is less similar to the normal waveform data and is more likely to be useful for learning, among pseudo abnormal waveform data that have the same similarity to the original abnormal waveform data.

The similarity display control unit 170 may control the display data indicating the similarity of each of the multiple pseudo waveform data sets generated by the pseudo waveform generating unit 166 with the original waveform data from which the data was generated to be displayed to the user. For example, the similarity display control unit 170 calculates the similarity between the multiple pseudo waveform data included in the pseudo waveform data set and the original waveform data from which the data was generated for each of the multiple pseudo waveform data sets, and displays display data including at least one of a histogram and a graph showing the number of pseudo waveform data for each similarity. This makes it possible to present the user with, for example, histograms corresponding to each of the multiple pseudo waveform data sets, which can be used as material for considering which pseudo waveform data set to use for learning. Specifically, it is possible to select only pseudo waveform data sets that are highly similar to the original waveform data from the multiple pseudo waveform data sets, or to include pseudo waveform data sets that are not very similar to the original waveform data, and so on.

The range specification receiving unit 172 receives a user's specification of a range of similarity for the display data displayed by the similarity display control unit 170. For example, the range specification receiving unit 172 receives a specification of a range of similarity for a histogram or graph showing the number of pseudo waveform data for each similarity with the original waveform data displayed by the similarity display control unit 170. The range specification receiving unit 172 receives, for example, an input of an upper and lower limit of similarity by the user. The range specification receiving unit 172 receives, for example, an input by the user to select a range of similarity for a histogram or graph in the display data.

The within-range waveform selection unit 174 selects multiple pseudo waveform data corresponding to the range of similarity specified by the range specification reception unit 172. For example, the within-range waveform selection unit 174 selects pseudo waveform data corresponding to each similarity within the range of similarity specified by the range specification reception unit 172.

The pseudo-waveform storage unit 168 may store a pseudo-waveform data set including multiple pseudo-waveform data selected by the within-range waveform selection unit 174. The pseudo-waveform storage unit 168 may store a pseudo-waveform data set including multiple pseudo-waveform data selected by the within-range waveform selection unit 174 as a data set to be used for learning. This makes it easier for the user to consider, among the multiple pseudo-waveform data included in one or more pseudo-waveform data sets, to what extent pseudo-waveform data that is similar to the original waveform data should be adopted, and to what extent pseudo-waveform data that is not similar to the original waveform data should be rejected.

The similarity display control unit 170 may control to display to the user display data that allows a comparison of the pseudo waveform dataset and the original waveform data from which it was generated at the waveform level. For example, when a pseudo waveform dataset is generated from one waveform data, the similarity display control unit 170 displays display data in which the one waveform data and the multiple pseudo waveform data included in the pseudo waveform dataset are arranged on a graph. For example, when a pseudo waveform dataset is generated from multiple waveform data, the similarity display control unit 170 displays display data in which the multiple waveform data and the multiple pseudo waveform data included in the pseudo waveform dataset are arranged on a graph. For example, when a pseudo waveform dataset is generated from multiple waveform data, the similarity display control unit 170 displays display data in which the average waveform of the multiple waveform data and the multiple pseudo waveform data included in the pseudo waveform dataset are arranged on a graph.

The learning unit 15 performs machine learning using the pseudo-waveform data generated by the generation unit 13. The learning unit 15 may include a learning data storage unit 106, an algorithm storage unit 144, a learning execution unit 140, and a waveform evaluation model storage unit 146.

The learning data storage unit 106 stores learning data used for learning. The learning data storage unit 106 may acquire waveform data stored in the waveform storage unit 102 from the waveform storage unit 102 and store the data. The learning data storage unit 106 may acquire pseudo waveform data stored in the pseudo waveform storage unit 168 from the pseudo waveform storage unit 168 and store the data.

The learning data storage unit 106 may acquire and store pseudo waveform data selected as learning data from among the multiple pseudo waveform data stored in the pseudo waveform storage unit 168. The pseudo waveform storage unit 168 may acquire and store a pseudo waveform data set selected as a learning data set from among the multiple pseudo waveform data sets stored in the pseudo waveform storage unit 168, for example. The learning data storage unit 106 may acquire and store a pseudo waveform data set stored in the pseudo waveform storage unit 168 as a data set used for learning from the pseudo waveform storage unit 168. The learning data storage unit 106 may acquire and store information such as recipe information, original waveform identification information, pre-processing identification information, pseudo identification information, abnormality identification information, and normality identification information associated with the pseudo waveform data set, together with the pseudo waveform data set, from the pseudo waveform storage unit 168.

The algorithm storage unit 144 stores multiple learning algorithms. Examples of learning algorithms include, but are not limited to, Neural Network (NN), Visual Geometry Group (VGG), and Support Vector Machine (SVM), and any algorithm may be used.

The learning execution unit 140 generates a waveform evaluation model that outputs an evaluation result of input waveform data by executing machine learning using the learning data stored in the learning data storage unit 106. The learning execution unit 140 may generate the waveform evaluation model using a learning algorithm specified by the user from among multiple learning algorithms stored in the algorithm storage unit 144.

The learning execution unit 140 may evaluate the generated waveform evaluation model. The learning execution unit 140 may evaluate the waveform evaluation model using, for example, waveform data for evaluation that is registered in advance by the registration unit 12 and stored in the waveform storage unit 102. The learning execution unit 140 may evaluate the waveform evaluation model using waveform data for evaluation selected by the user from among multiple waveform data for evaluation.

The learning execution unit 140 may evaluate the waveform evaluation model without using pseudo waveform data. For example, when pseudo waveform data or a pseudo waveform data set is selected as the waveform data for evaluation, the learning execution unit 140 prevents the selected pseudo waveform data from being used in the evaluation. For example, when a user selects waveform data for evaluation, the learning execution unit 140 controls so that the pseudo waveform data and the pseudo waveform data set cannot be selected. This makes it possible to prevent the pseudo-generated waveform data from being used in the evaluation of the waveform evaluation model, thereby reducing the possibility that the reliability of the evaluation of the waveform evaluation model will be reduced.

The waveform evaluation model storage unit 146 stores the waveform evaluation model generated by the learning execution unit 140. When the learning execution unit 140 generates a waveform evaluation model by performing machine learning using pseudo waveform data associated with preprocessing identification information, the waveform evaluation model storage unit 146 may store the waveform evaluation model by associating the preprocessing identification information with the waveform evaluation model.

The evaluation execution unit 16 executes evaluation of the waveform data using the waveform evaluation model generated by the learning unit 15. The evaluation execution unit 16 may include a model acquisition unit 180, an input waveform acquisition unit 182, a waveform input unit 184, and an evaluation result output control unit 186.

The model acquisition unit 180 acquires the waveform evaluation model stored in the waveform evaluation model storage unit 146. When preprocessing identification information is associated with the waveform evaluation model, the model acquisition unit 180 acquires the waveform evaluation model and the preprocessing identification information.

The input waveform acquisition unit 182 acquires the input waveform data to be evaluated. The input waveform acquisition unit 182 may acquire input waveform data that is instructed to be input by the user.

The waveform input unit 184 inputs the input waveform data acquired by the input waveform acquisition unit 182 into the waveform evaluation model acquired by the model acquisition unit 180. When the model acquisition unit 180 acquires the waveform evaluation model and preprocessing identification information, the waveform input unit 184 may apply preprocessing indicated by the preprocessing identification information to the input waveform data acquired by the input waveform acquisition unit 182 and input the data to the waveform evaluation model. This can improve the accuracy of waveform evaluation compared to a case where the preprocessing indicated by the preprocessing identification information is not applied. In addition, during evaluation by the evaluation execution unit 16, it is possible to avoid the burden of separately managing what kind of preprocessing was applied to the pseudo waveform data to generate the waveform evaluation model acquired by the model acquisition unit 180.

The evaluation result output control unit 186 controls the output of the evaluation result of the input waveform data output from the waveform evaluation model. For example, the evaluation result output control unit 186 causes the evaluation result of the input waveform data to be displayed on a display of the system 10. The evaluation result output control unit 186 may also transmit the evaluation result of the input waveform data to an arbitrary device and cause the device to display the evaluation result.

The system 10 may be composed of one device. The system 10 may be composed of multiple devices. For example, the system 10 is composed of a device including a memory unit 11, a registration unit 12, a generation unit 13, a pseudo waveform confirmation processing unit 14, and a learning unit 15, and a device including an evaluation execution unit 16. In this case, the model acquisition unit 180 receives the waveform evaluation model stored in the waveform evaluation model storage unit 146 from the system 10. The device including the evaluation execution unit 16 may be realized by a program that causes an existing device to function as the evaluation execution unit 16. The program implements the evaluation execution unit 16 in the device, and the device acquires the waveform evaluation model stored in the waveform evaluation model storage unit 146, and can realize waveform evaluation using the waveform evaluation model. When preprocessing identification information is associated with a waveform evaluation model, the waveform input unit 184 automatically performs preprocessing indicated by the preprocessing identification information on the input waveform data acquired by the input waveform acquisition unit 182 and inputs the preprocessing to the waveform evaluation model. This eliminates the need for the user of the evaluation execution unit 16 to manage specific preprocessing when inputting input waveform data to the waveform evaluation model, and also prevents a decrease in evaluation accuracy caused by inputting input waveform data to the waveform evaluation model without performing specific preprocessing.

The system 10 may be configured with a device including a memory unit 11, a registration unit 12, a generation unit 13, and a pseudo-waveform confirmation processing unit 14, a device including a learning unit 15, and a device including an evaluation execution unit 16. The system 10 may be realized with device configurations other than these.

Figure 6 shows an example of the flow of a process for generating a waveform evaluation model by the system 10. Here, an example is described in which the system 10 generates a large amount of pseudo-abnormal waveform data using a small amount of abnormal waveform data stored in the waveform storage unit 102, and generates a normal/abnormal evaluation model that evaluates whether input waveform data is a normal waveform or an abnormal waveform using the pseudo-abnormal waveform data, abnormal waveform data, and normal waveform data.

In step (sometimes abbreviated as S) 102, the intention identification unit 162 identifies the user's intention. The intention identification unit 162 may identify multiple combinations of preprocessing, algorithms, and parameter settings according to the user's instructions.

In S104, the preprocessing unit 164 and the pseudo-waveform generating unit 166 perform preprocessing on a small amount of abnormal waveform data for one of the multiple combinations, and then generate a pseudo-abnormal waveform data set using an algorithm and parameter settings.

In S106, the pseudo waveform storage unit 168 stores the pseudo identification information, original waveform identification information, pre-processing identification information, and recipe information in association with the pseudo abnormal waveform data set generated in S104.

If the generation of pseudo-abnormal waveform data sets has not been completed for all of the multiple combinations identified in S102 (NO in S108), the process returns to S104, where the pre-processing unit 164 and the pseudo-waveform generating unit 166 generate pseudo-abnormal waveform data sets for the next combination. If it is determined that the generation is completed (YES in S108), the process proceeds to S110.

In S110, the similarity display control unit 170 controls the display data showing the similarity of each of the multiple pseudo abnormal waveform data sets to the small amount of abnormal waveform data from which they were generated, so that the display data is displayed to the user. In S112, the range designation receiving unit 172 and the waveform within range selection unit 174 select the pseudo abnormal waveform data to be used for learning. When the range designation receiving unit 172 and the waveform within range selection unit 174 receive a designation of a similarity range from the user, they select multiple pseudo abnormal waveform data corresponding to the similarity range that has been designated, as the pseudo abnormal waveform data to be used for learning. When all pseudo abnormal waveform data sets are selected by the user, the range designation receiving unit 172 and the waveform within range selection unit 174 may select all pseudo abnormal waveform data included in all pseudo abnormal waveform data sets as the pseudo abnormal waveform data to be used for learning.

In S114, the learning execution unit 140 generates a normal/abnormal evaluation model by executing machine learning using the pseudo-abnormal waveform data, normal waveform data, and abnormal waveform data selected by the range designation receiving unit 172 and the range waveform selection unit 174 in S112. Note that the process may proceed to S114 after performing S102 to S112 multiple times. In S116, the learning execution unit 140 evaluates the normal/abnormal evaluation model generated in S114 using the evaluation waveform data stored in the waveform storage unit 102.

If an instruction to register a normal/abnormal evaluation model is received from the user (YES in S118), the process proceeds to S120; if an instruction to register is not received (NO in S118), the process ends. In S120, the waveform evaluation model storage unit 146 stores the normal/abnormal evaluation model generated in S114 in association with the preprocessing identification information. Then, the process ends.

By executing the process shown in FIG. 6, the system 10 can generate a large amount of pseudo-abnormal waveform data that reflects the user's intentions, even if the amount of abnormal waveform data is small, and a normal/abnormal evaluation model can be generated using the pseudo-abnormal waveform data selected by the user while checking the displayed data, which can contribute to the generation of a normal/abnormal evaluation model with high evaluation accuracy.

Figure 7 shows an example of management data 190 for managing pseudo waveform data. System 10 may generate multiple projects and manage various data for each project. This makes it easier to manage multiple types of data when targeting multiple types of data, such as data for evaluating normality/abnormality of a ball screw and data for evaluating normality/abnormality of a motor.

For example, the registration unit 12 generates a project and registers waveform data according to a user's instruction. The registration unit 12 may register abnormal waveform data, normal waveform data, and waveform data for evaluation, etc. The waveform storage unit 102 stores the waveform data in association with the project. When the pseudo-waveform generation unit 166 generates a pseudo-waveform data set using waveform data corresponding to the project, the pseudo-waveform storage unit 168 stores the pseudo-waveform data set and related information in association with the project. Examples of related information include the number of data included in the pseudo-waveform data set, original waveform identification information, preprocessing identification information, and recipe information.

In FIG. 7, a data set called "ball screw abnormal data A" and a data set called "ball screw abnormal data B" are registered in a project called "ball screw pseudo waveform". The user can confirm from the management data 190 that the number of data items in "ball screw abnormal data A" is 80, that the original waveform is raw abnormal data A, that no preprocessing has been performed, and that the data was generated using a noise addition processing algorithm with a standard deviation of 0.0002, and that the number of data items in "ball screw abnormal data B" is 80, that the original waveform is raw abnormal data A, that downsampling with a sampling rate of 6 has been performed as preprocessing, and that the data was generated using an expansion/contraction processing algorithm with warp_scales set to [0.5, 2.0] and window_ratio set to 0.1. This can be used as material for considering what preprocessing, algorithm, and parameter settings to use to generate pseudo waveform data next.

Figure 8 is an explanatory diagram for explaining the expansion processing algorithm. Here, we will explain an example in which pseudo waveform data 308 is generated from waveform data 300 with reduce_ratio = 0.9.

The pseudo waveform generation unit 166 randomly determines a start point 303 within a range 302 of the data size 301 × (1-reduce_ratio) of the time series of the waveform data 300 from the start point of the time series of the waveform data 300. The pseudo waveform generation unit 166 sets the end point 304 to a point that is a distance forward from the start point 303 by the data size 301 × reduce_ratio of the time series. The pseudo waveform generation unit 166 cuts out a section 305 from the start point 303 to the end point 304. The pseudo waveform generation unit 166 expands the cut-out section 305 to the original size to generate pseudo waveform data 308.

Even if reduce_ratio is fixed, different pseudo waveform data 308 will be generated each time from the same waveform data 300 because the starting point 303 is randomly determined. By changing reduce_ratio, even more variations of pseudo waveform data 308 can be generated.

Figure 9 is an explanatory diagram for explaining the expansion/contraction processing algorithm. Here, we will explain an example in which pseudo waveform data is generated from waveform data 300 with warp_scales = [0.5, 2] and window_ratio = 0.1.

The pseudo waveform generation unit 166 randomly determines one value from warp_scales. The pseudo waveform generation unit 166 calculates warp_size = window_ratio x time series data size 301. The pseudo waveform generation unit 166 randomly secures an area of warp_size in the range 311 from 1 to "time series data size 301-warp_size-1", and determines three areas: a waveform 312 before the warp_size area, a waveform 313 of the warp_size area, and a waveform 314 after the warp_size area. The pseudo waveform generation unit 166 enlarges or reduces the area of waveform 313 to a size of "warp_size x warp_scales", recombines waveform 312, and the enlarged or reduced waveform 313 and waveform 314, and generates pseudo waveform data by returning them to their original size.

By randomly determining one value from warp_scale, different pseudo waveform data is generated each time from the same waveform data 300. By setting many candidates for warp_scale, even more variations of pseudo waveform data can be generated.

Figure 10 is an explanatory diagram for explaining the scale random transformation algorithm. Here, we will explain an example in which the number of via points is 5 and pseudo waveform data is generated from waveform data 300.

The pseudo-waveform generating unit 166 determines random values based on a normal distribution with a mean value of 1 and a standard deviation of σ for the number of waypoints (y-axis information), divides the size in the time series direction by the number of waypoints, and arranges waypoints 321 at equal intervals (x-axis information). Using the y-axis information and x-axis information, the pseudo-waveform generating unit 166 creates a spline curve 322 whose x-axis is the data size 301 of the time series of the waveform data. The pseudo-waveform generating unit 166 generates pseudo-waveform data by multiplying the created spline curve 322 with the waveform data.

Even if the number of via points is fixed, a random value is determined as the y-axis information, so that different pseudo waveform data is generated each time from the same waveform data 300. By changing the number of via points, even more variations of pseudo waveform data can be generated.

Figure 11 shows an example of display data 400 displayed by the similarity display control unit 170. In the display data 400, the horizontal axis indicates the degree of similarity between the waveform data from which the pseudo waveform data is generated and the pseudo waveform data, and the vertical axis indicates the degree, which is the number of pseudo waveform data.

FIG. 11 illustrates display data 400 including a graph 412 and a histogram 414 corresponding to a first pseudo waveform data set, a graph 422 and a histogram 424 corresponding to a second pseudo waveform data set, and a graph 432 and a histogram 424 corresponding to a third pseudo waveform data set of three pseudo waveform data sets.

The similarity display control unit 170 may switch the display of each of the graph 412, histogram 414, graph 422, histogram 424, graph 432, and histogram 434 on and off in accordance with a user instruction. The similarity display control unit 170 may apply semi-transparent display or the like to each of the graph 412, histogram 414, graph 422, histogram 424, graph 432, and histogram 434.

The display data 400 makes it easy to understand how similar each pseudo waveform data set is to the original waveform data, how much variation was present in the generation, etc.

The range specification receiving unit 172 may receive a specification of a similarity range for a pseudo waveform dataset selected by a user from among multiple pseudo waveform datasets. For example, when the graph 412 and the histogram 414 are selected by the user, the range specification receiving unit 172 receives a specification of a similarity range for the graph 412 and the histogram 414. The within-range waveform selecting unit 174 may select pseudo waveform data from the graph 412 and the histogram 414 that corresponds to the similarity range for which the specification was received. This makes it possible to first determine which pseudo waveform dataset to adopt from among multiple target pseudo waveform datasets, and then select the range of the pseudo waveform dataset to adopt.

The range specification receiving unit 172 may receive a specification of a similarity range for multiple pseudo waveform data sets. For example, the range specification receiving unit 172 receives a specification of a similarity range for the graph 412, histogram 414, graph 422, histogram 424, graph 432, and histogram 434. The within-range waveform selecting unit 174 may select pseudo waveform data corresponding to the specified similarity range from among the graph 412, histogram 414, graph 422, histogram 424, graph 432, and histogram 434.

In the above embodiment, an abnormal/normal evaluation model is given as an example of a waveform evaluation model, but this is not limiting. System 10 can generate various types of waveform evaluation models, such as a failure timing evaluation model that evaluates the failure timing of a target system.

System 10 is not limited to one-dimensional data such as waveform data, but may also be used for two-dimensional data such as images. For example, system 10 may generate pseudo two-dimensional data from two-dimensional data. System 10 may apply the pseudo waveform generation function described above to two-dimensional data, assuming that the horizontal axis of the two-dimensional data is time and the vertical axis is frequency. For example, system 10 applies the pseudo waveform generation function described above to frequency components extracted from each column (time) of the two-dimensional data. By performing the same process for all times and then combining them, the pseudo waveform generation function can be applied to two-dimensional data.

FIG. 12 shows an example of a hardware configuration of the system 10 or a computer 1200 functioning as a part of the system 10. The computer 1200 functioning as a part of the system 10 functions, for example, as a device including a storage unit 11, a registration unit 12, a generation unit 13, and a pseudo waveform confirmation processing unit 14. The computer 1200 functioning as a part of the system 10 functions, for example, as a device including a storage unit 11, a registration unit 12, a generation unit 13, a pseudo waveform confirmation processing unit 14, and a learning unit 15. The computer 1200 functioning as a part of the system 10 functions, for example, as a device including a learning unit 15. The computer 1200 functioning as a part of the system 10 functions, for example, as a device including an evaluation execution unit 16. A program installed in the computer 1200 can cause the computer 1200 to function as one or more "parts" of the device according to this embodiment, or cause the computer 1200 to execute an operation or one or more "parts" associated with the device according to this embodiment, and/or cause the computer 1200 to execute a process or a stage of the process according to this embodiment. Such programs may be executed by the CPU 1212 to cause the computer 1200 to perform specific operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

The computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, and a graphics controller 1216, which are connected to each other by a host controller 1210. The computer 1200 also includes input/output units such as a communication interface 1222, a storage device 1224, a DVD drive, and an IC card drive, which are connected to the host controller 1210 via an input/output controller 1220. The DVD drive may be a DVD-ROM drive, a DVD-RAM drive, etc. The storage device 1224 may be a hard disk drive, a solid state drive, etc. The computer 1200 also includes a ROM 1230 and a legacy input/output unit such as a keyboard, which are connected to the input/output controller 1220 via an input/output chip 1240.

The CPU 1212 operates according to the programs stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphics controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into itself, and causes the image data to be displayed on the display device 1218.

The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive reads programs or data from a DVD-ROM or the like and provides them to the storage device 1224. The IC card drive reads programs and data from an IC card and/or writes programs and data to an IC card.

ROM 1230 stores therein a boot program, etc., executed by computer 1200 upon activation, and/or a program that depends on the hardware of computer 1200. I/O chip 1240 may also connect various I/O units to I/O controller 1220 via USB ports, parallel ports, serial ports, keyboard ports, mouse ports, etc.

The programs are provided by a computer-readable storage medium such as a DVD-ROM or an IC card. The programs are read from the computer-readable storage medium, installed in storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable storage media, and executed by CPU 1212. The information processing described in these programs is read by computer 1200, and brings about cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be constructed by realizing the operation or processing of information according to the use of computer 1200.

For example, when communication is performed between computer 1200 and an external device, CPU 1212 may execute a communication program loaded into RAM 1214 and instruct communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of CPU 1212, communication interface 1222 reads transmission data stored in a transmission buffer area provided in RAM 1214, storage device 1224, a DVD-ROM, or a recording medium such as an IC card, and transmits the read transmission data to the network, or writes received data received from the network to a reception buffer area or the like provided on the recording medium.

The CPU 1212 may also cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, a DVD drive (DVD-ROM), an IC card, etc. to be read into the RAM 1214, and perform various types of processing on the data on the RAM 1214. The CPU 1212 may then write back the processed data to the external recording medium.

Various types of information, such as various types of programs, data, tables, and databases, may be stored in the recording medium and may undergo information processing. The CPU 1212 may perform various types of processing on the data read from the RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search/replacement, etc., as described throughout this disclosure and specified by the instruction sequence of the program, and writes back the results to the RAM 1214. The CPU 1212 may also search for information in a file, database, etc. in the recording medium. For example, when multiple entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored in the recording medium, the CPU 1212 may search for an entry whose attribute value of the first attribute matches a specified condition from among the multiple entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

The above-described programs or software modules may be stored in a computer-readable storage medium on the computer 1200 or in the vicinity of the computer 1200. In addition, a recording medium such as a hard disk or RAM provided in a server system connected to a dedicated communication network or the Internet can be used as a computer-readable storage medium, thereby providing the programs to the computer 1200 via the network.

The blocks in the flowcharts and block diagrams in this embodiment may represent stages of a process in which an operation is performed or "parts" of a device responsible for performing the operation. Particular stages and "parts" may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable storage medium, and/or a processor provided with computer-readable instructions stored on a computer-readable storage medium. The dedicated circuitry may include digital and/or analog hardware circuits and may include integrated circuits (ICs) and/or discrete circuits. The programmable circuitry may include reconfigurable hardware circuits including AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and memory elements, such as, for example, field programmable gate arrays (FPGAs) and programmable logic arrays (PLAs).

A computer-readable storage medium may include any tangible device capable of storing instructions that are executed by a suitable device, such that a computer-readable storage medium having instructions stored thereon comprises an article of manufacture that includes instructions that can be executed to create means for performing the operations specified in the flowchart or block diagram. Examples of computer-readable storage media may include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, and the like. More specific examples of computer-readable storage media may include floppy disks, diskettes, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), electrically erasable programmable read-only memories (EEPROMs), static random access memories (SRAMs), compact disk read-only memories (CD-ROMs), digital versatile disks (DVDs), Blu-ray disks, memory sticks, integrated circuit cards, and the like.

The computer readable instructions may include either assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk (registered trademark), JAVA (registered trademark), C++, etc., and conventional procedural programming languages such as the "C" programming language or similar programming languages.

The computer-readable instructions may be provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, or a programmable circuit, either locally or over a local area network (LAN), a wide area network (WAN), such as the Internet, etc., so that the processor of the general-purpose computer, special-purpose computer, or other programmable data processing apparatus, or the programmable circuit, executes the computer-readable instructions to generate means for performing the operations specified in the flowcharts or block diagrams. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

The present invention has been described above using an embodiment, but the technical scope of the present invention is not limited to the scope described in the above embodiment. It is clear to those skilled in the art that various modifications and improvements can be made to the above embodiment. It is clear from the claims that forms incorporating such modifications or improvements can also be included in the technical scope of the present invention.

The order of execution of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before" or "prior to," and it should be noted that the processes may be performed in any order, unless the output of a previous process is used in a later process. Even if the operational flow in the claims, specifications, and drawings is explained using "first," "next," etc. for convenience, it does not mean that it is essential to perform the processes in that order.

10 System, 11 Memory, 12 Registration, 13 Generation, 14 Pseudo-waveform confirmation processing unit, 15 Learning unit, 16 Evaluation execution unit, 102 Waveform storage unit, 104 Inflation processing, 106 Learning data storage unit, 108 User input, 110 Algorithm storage unit, 112 Selection processing, 114 Parameter setting processing, 116 Generation processing, 118 Data storage unit, 120 User intention/reflection processing, 122 Distribution display processing, 124 Selection processing, 126 Individual display processing, 128 Range selection processing, 130 Learning data storage unit, 140 Learning execution unit, 142 AI, 144 Algorithm storage unit, 146 Waveform evaluation model storage unit, 150 Application, 160 Waveform acquisition unit, 162 Intention identification unit, 164 Preprocessing unit, 166 Pseudo-waveform generation unit, 168 Pseudo-waveform storage unit, 170 Similarity display control unit, 172 range designation receiving unit, 174 range waveform selection unit, 180 model acquisition unit, 182 input waveform acquisition unit, 184 waveform input unit, 186 evaluation result output control unit, 190 management data, 202 abnormal waveform data, 204 pseudo abnormal waveform data, 206 pseudo abnormal waveform data, 212 histogram, 214 waveform, 220 display, 300 waveform data, 301 data size, 302 range, 303 start point, 304 end point, 305 section, 308 pseudo waveform data, 311 range, 312 waveform, 313 waveform, 314 waveform, 321 way point, 322 spline curve, 400 display data, 412 graph, 414 histogram, 422 graph, 424 histogram, 432 graph, 434 histogram, 1200 Computer, 1210 host controller, 1212 CPU, 1214 RAM, 1216 graphic controller, 1218 display device, 1220 input/output controller, 1222 communication interface, 1224 storage device, 1230 ROM, 1240 input/output chip

Claims (14)

A waveform acquisition unit that acquires waveform data;
An intention identification unit that identifies a user's intention;
and a pseudo-waveform generating unit that generates pseudo-waveform data from the waveform data acquired by the waveform acquiring unit in a manner that reflects the user's intention identified by the intention identifying unit. 2. The generation system according to claim 1, further comprising: a similarity display control unit that controls display data indicating a similarity between the pseudo-waveform data generated from the waveform data by the pseudo-waveform generation unit and the waveform data from which the pseudo-waveform data was generated, to be displayed to the user. the pseudo-waveform generating unit generates a plurality of pieces of pseudo-waveform data from the waveform data by using at least one of a plurality of algorithms, a plurality of parameter settings, and a random number;
The generation system according to claim 2, wherein the similarity display control unit controls the display data indicating the similarity between the plurality of pseudo waveform data generated from the waveform data by the pseudo waveform generation unit and the waveform data from which the plurality of pseudo waveform data were generated to be displayed to the user. the intention identification unit identifies a plurality of combinations of algorithms and parameter settings in accordance with an instruction from the user;
the pseudo-waveform generating unit generates, for each of the plurality of combinations identified by the intention identifying unit, a pseudo-waveform data set including the plurality of pseudo-waveform data from the waveform data by using the algorithm and the parameter setting;
The generation system according to claim 3, wherein the similarity display control unit controls the display data indicating the similarity of each of the multiple pseudo waveform data sets generated by the pseudo waveform generation unit to the waveform data from which the pseudo waveform data was generated to be displayed to the user. 5. The generation system according to claim 4, further comprising a pseudo-waveform storage unit configured to store the plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit in association with recipe information capable of identifying a combination of the algorithm and the parameter setting used in generating the pseudo-waveform data set, respectively. 5. The generation system according to claim 4, further comprising a pseudo-waveform storage unit that stores the plurality of pseudo-waveform data sets generated by the pseudo-waveform generation unit by associating each of the plurality of pseudo-waveform data sets with original waveform identification information capable of identifying the waveform data that was the source of generation of the pseudo-waveform data set. a range designation receiving unit that receives designation by the user of a range of similarity for the display data displayed by the similarity display control unit;
a range-specified waveform selection unit that selects a plurality of pieces of pseudo-waveform data corresponding to the range of similarity specified by the range specification receiving unit;
The generating system according to claim 3 or 4, further comprising: a pseudo-waveform storage unit configured to store a pseudo-waveform data set including the plurality of pseudo-waveform data selected by the in-range waveform selection unit. The generation system according to any one of claims 2 to 6, wherein the similarity display control unit controls to display to the user the display data indicating the similarity between the pseudo waveform data and the waveform data from which the pseudo waveform data was generated, and other display data indicating the similarity between the pseudo waveform data and other waveform data different from the waveform data. a preprocessing unit that performs at least one of a plurality of preprocessing steps on the waveform data;
a pseudo-waveform storage unit that stores a pseudo-waveform data set including a plurality of pseudo-waveform data generated by the pseudo-waveform generation unit from the waveform data that has been preprocessed by the preprocessing unit, in association with preprocessing identification information that can identify the preprocessing. a preprocessing unit that performs at least one of a plurality of preprocessing steps on the waveform data;
a learning execution unit that executes machine learning using the pseudo-waveform data generated by the pseudo-waveform generation unit from the waveform data that has been preprocessed by the preprocessing unit, thereby generating a waveform evaluation model that outputs an evaluation result of input waveform data; and
The generation system according to claim 1 , further comprising: a waveform evaluation model storage unit that stores the waveform evaluation model generated by the learning execution unit in association with preprocessing identification information capable of identifying the preprocessing. a model acquisition unit that acquires the waveform evaluation model associated with the preprocessing identification information stored in the waveform evaluation model storage unit;
a waveform input unit that performs preprocessing indicated by the preprocessing identification information on input waveform data and inputs the preprocessed data to the waveform evaluation model;
The generation system according to claim 10 , further comprising: an evaluation result output control unit configured to control the waveform evaluation model to output an evaluation result of the input waveform data output from the waveform evaluation model. a pseudo-waveform storage unit that stores the pseudo-waveform data generated by the pseudo-waveform generating unit in association with pseudo-identification information indicating that the pseudo-waveform data is data generated from the waveform data;
7. The generation system according to claim 1 , further comprising: a learning execution unit that performs machine learning using the pseudo-waveform data generated by the pseudo-waveform generation unit to generate a waveform evaluation model that outputs an evaluation result of input waveform data, and evaluates the waveform evaluation model without using the pseudo-waveform data. A program for causing a computer to function as the generation system according to any one of claims 1 to 6. 1. A computer-implemented method for generating a waveform evaluation model, comprising:
A waveform data acquisition stage for acquiring waveform data;
an intent identification stage for identifying a user's intent;
a pseudo-waveform generation step of generating pseudo-waveform data from the waveform data acquired in the waveform data acquisition step in a manner that reflects the user's intention identified in the intention identification step; and a learning execution step of executing machine learning using the pseudo-waveform data generated in the pseudo-waveform generation step to generate a waveform evaluation model that outputs an evaluation result of input waveform data.

Images