Information processing device and information processing method

Abstract

An information processing device (400) comprises: an extraction unit (401) that extracts, using an abnormality detection model of a facility, a fluctuation characteristic of abnormal-state generated data, which is generated by performing a simulation of an abnormality of the facility; and an update unit (402) that updates a parameter of the simulation on the basis of the fluctuation characteristic extracted by the extraction unit (401).

Description

Information processing device and information processing method

[0001] The present disclosure relates to an information processing device and an information processing method.

[0002] Many motors and gears are used in industrial facilities, industrial machinery, industrial robots, power generation facilities, etc., used in production at factories, etc. Abnormalities in equipment due to aging or wear and tear, as well as sudden equipment trouble, can lead to line stoppages, reducing productivity and causing accidents.

[0003] Therefore, there is a need for a method that enables efficient planned maintenance according to the state of the equipment by estimating the internal state from the characteristics of signals from sensors attached to the equipment using machine learning models, etc. In particular, in recent years, with the aim of fully automating maintenance work, there has been an increasing demand for technology that can not only estimate whether an equipment is normal or abnormal, but also estimate the location and / or cause of the abnormality.

[0004] A technical challenge when identifying not only the normality and abnormality of equipment, but also the location and / or cause of the abnormality, is the lack of data at the time of the abnormality.

[0005] When estimating only the normal and abnormal states of equipment, it is sufficient to learn only the characteristics of normal data, which can be acquired in large quantities, and evaluate changes from normal. On the other hand, to identify the abnormal location and / or the cause of the abnormality, abnormal data corresponding to each abnormal location is required. However, it is generally rare to be able to acquire a sufficient amount of abnormal data from on-site equipment for learning.

[0006] Some literature discloses methods for filling in the gaps by generating abnormal data using physical simulations.

[0007] However, it is generally difficult to accurately grasp physical parameters such as the dimensions and physical properties of the equipment to be diagnosed. Therefore, errors occur in abnormality data generated using inaccurate physical parameters. As a result, even if physical simulation is used to compensate for the lack of abnormality data, it does not contribute to improving the accuracy of diagnosis of the abnormality location and / or the abnormality cause. In order to improve the accuracy of diagnosis, it is necessary to improve the values ​​of the physical parameters.

[0008] For example, Patent Document 1 discloses a method for improving the value of a physical parameter by comparing actual measurement data under normal conditions with data generated under normal conditions and updating the parameters of a physical model to reduce the difference between them. Patent Document 2 discloses a method for improving the value of a physical parameter by comparing actual measurement data under abnormal conditions with data generated under abnormal conditions and updating the parameters of a physical model to reduce the difference between them.

[0009] JP 2018-190245 A International Publication No. 2016 / 195092

[0010] However, the method of Patent Document 1 has the potential to improve accuracy only for physical parameters necessary for generating normal data, and does not improve accuracy for physical parameters that only contribute to generating abnormal data. Therefore, the method of Patent Document 1 cannot avoid a decrease in accuracy of diagnosing the location and / or cause of the abnormality.

[0011] Furthermore, with the method of Patent Document 2, it is difficult to collect a sufficient amount of actual measurement data during an abnormality, so it is only possible to compare actual measurement data during an abnormality with generated data during an abnormality for only some of the abnormal locations or causes. Furthermore, unless the actual measurement data during an abnormality is labeled with the abnormal location or cause, an appropriate comparison cannot be made. However, accurately obtaining labels for the abnormal location or cause requires stopping the equipment and inspecting it. This leads to reduced productivity or increased costs. Therefore, even with the method of Patent Document 2, it is difficult to appropriately update the physical parameters, and a decrease in the accuracy of diagnosing the abnormal location and / or cause cannot be avoided.

[0012] As described above, it is difficult to generate a sufficient amount of diverse abnormality generation data in the methods disclosed in Patent Documents 1 and 2. For this reason, it is not possible to estimate the location and / or cause of the abnormality in the equipment with sufficient accuracy.

[0013] Therefore, the present disclosure provides an information processing device and an information processing method that can support the generation of various abnormality-related generation data.

[0014] An information processing device according to one aspect of the present disclosure includes an extraction unit that extracts fluctuation characteristics of abnormality-generated data generated by simulating an abnormality in equipment using an abnormality detection model of the equipment, and an update unit that updates parameters of the simulation based on the fluctuation characteristics extracted by the extraction unit.

[0015] In addition, an information processing method according to one aspect of the present disclosure includes the steps of extracting fluctuation characteristics of abnormality-generated data generated by simulating an abnormality in equipment using an abnormality detection model of the equipment, and updating parameters of the simulation based on the extracted fluctuation characteristics.

[0016] Furthermore, one aspect of the present disclosure can be realized as a program that causes a computer to execute the information processing method, or as a computer-readable non-transitory recording medium storing the program.

[0017] These comprehensive or specific aspects may be realized as a system, a method, an integrated circuit, a computer program, or a recording medium, or may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

[0018] According to one aspect of the present disclosure, it is possible to support the generation of a variety of abnormality generation data.

[0019] FIG. 1 is a block diagram showing the configuration of an equipment state estimation system according to an embodiment. FIG. 2 is a block diagram showing the configuration of an abnormality data generation unit according to an embodiment. FIG. 3 is a flowchart showing learning phase processing among the processing performed by an equipment state estimation device according to an embodiment. FIG. 4 is a flowchart showing diagnosis phase processing among the processing performed by an equipment state estimation device according to an embodiment. FIG. 5 is a diagram showing an example of a display screen displayed by an input / output unit of an equipment state estimation device according to an embodiment. FIG. 6 is a diagram for explaining the effect of an equipment state estimation device according to an embodiment. FIG. 7 is a block diagram showing the configuration of an information processing device according to a modified embodiment. FIG. 8 is a flowchart showing the operation of an information processing device according to a modified embodiment.

[0020] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings.

[0021] Note that the embodiments described below are examples, and all are comprehensive or specific examples. The numerical values, shapes, materials, components, component placement and connection configurations, steps, and step order shown in the following embodiments are examples and are not intended to limit the present disclosure. Furthermore, among the components related to the following embodiments, components that are not recited in independent claims are described as optional components. In other words, the present disclosure is not limited by the following embodiments.

[0022] Further advantages and benefits of certain aspects of the present disclosure will become apparent from the specification and drawings. Such advantages and / or benefits may be provided by some of the embodiments and features described in the specification and drawings, respectively, but not necessarily all of them may be provided to obtain one or more identical features.

[0023] Furthermore, each figure is a schematic diagram and is not necessarily an exact illustration. Therefore, for example, the scales of the figures do not necessarily match. Furthermore, in each figure, substantially the same components are given the same reference numerals, and redundant explanations are omitted or simplified.

[0024] Furthermore, in this specification, ordinal numbers such as "first" and "second" do not refer to the number or order of components unless otherwise specified, but are used for the purpose of avoiding confusion and distinguishing between components of the same type.

[0025] (Embodiment) [Configuration] First, an overview of the configuration and processing of an equipment state estimation system according to an embodiment will be described.

[0026] Fig. 1 is a block diagram showing the configuration of an equipment state estimation system 10 according to an embodiment of the present disclosure. As shown in Fig. 1, the equipment state estimation system 10 includes equipment 20, a sensor 30, and an equipment state estimation device 100. Note that the equipment state estimation system 10 does not necessarily have to include the equipment 20 and the sensor 30, and may instead acquire data indicating the operating status of equipment not included in the equipment state estimation system 10, and may detect an abnormality in the equipment or assist in diagnosing the cause of the abnormality based on the acquired data.

[0027] The equipment state estimation device 100 is realized by, for example, one or more computer devices. The equipment state estimation device 100 includes an information processing unit 101, a database (DB) 102, and an input / output unit 103. The information processing unit 101 also includes an anomaly detection model learning unit 110, an anomaly data generation unit 120, an anomaly diagnosis model learning unit 130, an anomaly detection unit 140, and an anomaly diagnosis unit 150. The information processing unit 101 or the equipment state estimation device 100 is an example of an information processing device in the present disclosure.

[0028] The equipment 20 is a machine whose state is to be estimated. The equipment 20 is, for example, a rotating machine such as a motor or a generator. Alternatively, the equipment 20 may be a mechanism in which multiple rotating machines are connected via a gearbox, a load, a chain, or the like. Alternatively, the equipment 20 may be a mechanism such as a robot arm or a mobile object incorporating such a mechanism.

[0029] The sensor 30 converts physical quantities such as vibrations, electromagnetic waves, and heat observed in the facility 20 into signals that can be processed electronically. In a learning phase, which will be described later, the sensor 30 outputs the converted signals to the anomaly detection model learning unit 110. In a diagnosis phase, which will be described later, the sensor 30 outputs the converted signals to the anomaly detection unit 140.

[0030] In this specification, the signal output by the sensor 30 is referred to as the "measured signal." The measured signal output from the sensor 30 when the equipment 20 is normal is referred to as the "normal measured signal." The measured signal output from the sensor 30 when the equipment 20 is abnormal is referred to as the "abnormal measured signal." The normal measured signal and the abnormal measured signal are each converted by the information processing unit 101 into data that can be processed by the anomaly detection model learning unit 110 or the anomaly detection unit 140. In this specification, data converted from the normal measured signal is referred to as the "normal measured data." Data converted from the abnormal measured signal is referred to as the "abnormal measured data." The measured signal and the measured data can be considered to be substantially the same.

[0031] The information processing unit 101 processes the setting information input from the input / output unit 103, the signal acquired from the sensor 30, and the information read from the database 102. The information processing unit 101 stores the information and / or data obtained by the processing in the database 102 or outputs the information and / or data to the input / output unit 103.

[0032] The information processing unit 101 is realized by one or more processors and one or more memories. The one or more processors perform predetermined processing by reading and executing programs stored in the one or more memories. The predetermined processing is processing performed by each of the functional components of the information processing unit 101, specifically, the anomaly detection model learning unit 110, the anomaly data generation unit 120, the anomaly diagnosis model learning unit 130, the anomaly detection unit 140, and the anomaly diagnosis unit 150. The one or more memories include a non-volatile memory in which the program is stored and a volatile memory that is a temporary storage area for executing the program. The program may be stored in the database 102.

[0033] The database 102 stores data received from the information processing unit 101 in accordance with the processing of the information processing unit 101. Alternatively, the database 102 outputs data necessary for the processing of the information processing unit 101 to the information processing unit 101. The database 102 is realized by a nonvolatile storage device such as a hard disk drive (HDD) or a solid state drive (SSD).

[0034] The input / output unit 103 receives inputs such as settings required for generating abnormality-time generated data from a user. The input / output unit 103 also visualizes the distribution of abnormality-time generated data generated by the information processing unit 101 and / or the system output such as the state of the equipment 20 that is the target of abnormality detection, and presents it to the user. The input / output unit 103 is an example of a display unit included in the information processing device of the present disclosure. Note that the user may be a manager of the equipment 20, a maintenance worker for the equipment 20, a user of the equipment state estimation system 10, or the like.

[0035] The input / output unit 103 is realized by input devices such as a touch panel, a keyboard, a mouse, a microphone, etc., and output devices such as a display, a speaker, etc. For example, the input / output unit 103 is a touch panel display that has a function of accepting operation input from a user and a function of displaying information to the user, but is not limited to this.

[0036] The anomaly detection model learning unit 110 learns an anomaly detection model using the measured signal input from the sensor 30. Here, the anomaly detection model is a model that learns normal characteristics of measured data using only normal measured data. For example, a machine learning model may be used to learn the normal characteristics. For machine learning, various known algorithms, such as an autoencoder using a neural network, a support vector machine, or a random forest, or ensemble learning that combines these, may be used. Alternatively, a rule-based method may be used for the anomaly detection model. In other words, a threshold value for a feature quantity of a certain signal determined from normal measured data may be used as the anomaly detection model. The learned anomaly detection model is output to the anomaly data generation unit 120 and the anomaly detection unit 140.

[0037] The abnormality data generation unit 120 generates abnormality generation data with an abnormality cause label using the abnormality detection model input from the abnormality detection model learning unit 110. The abnormality data generation unit 120 outputs the generated abnormality generation data to the abnormality diagnosis model learning unit 130. The specific functional configuration of the abnormality data generation unit 120 will be described later with reference to FIG. 2 .

[0038] In this embodiment, the term "cause of an abnormality" is used as a concept including an abnormality location and / or an abnormality cause. An abnormality location is information indicating the location where an abnormality occurred in the equipment 20 (for example, a component or part of a component of the equipment 20). An abnormality cause is information indicating the cause of an abnormality occurring in the equipment 20 (for example, a broken component or a component that is not operating normally). An abnormality location and an abnormality cause do not need to be distinguished from each other.

[0039] The anomaly diagnosis model learning unit 130 is an example of a generation unit according to the present disclosure, and generates an anomaly diagnosis model for diagnosing an anomaly factor of the equipment 20 based on a plurality of anomaly-generated data sets generated by performing a simulation or on the variation characteristics of each of the plurality of anomaly-generated data sets. Specifically, the anomaly diagnosis model learning unit 130 generates an anomaly diagnosis model based on the anomaly-generated data sets labeled with anomaly factors and provided by the anomaly data generation unit 120. The anomaly diagnosis model is a mathematical model that estimates an anomaly factor of the equipment from the anomaly-generated data sets. Machine learning techniques such as neural networks, support vector machines, or random forests, or ensemble learning that combines these, can be used to learn the anomaly diagnosis model. The anomaly diagnosis model learning unit 130 outputs the generated anomaly diagnosis model to the anomaly diagnosis unit 150.

[0040] In a diagnosis phase, which will be described later, the anomaly detection unit 140 determines whether the equipment 20 is normal or abnormal using the measured signal input from the sensor 30 and the anomaly detection model. When the anomaly detection unit 140 determines that the equipment 20 is abnormal, that is, when it detects an abnormality in the equipment 20, it outputs the measured signal in which the abnormality was detected, that is, the measured signal at the time of abnormality (specifically, measured data at the time of abnormality), to the anomaly diagnosis unit 150. In addition, the abnormality determination result is presented to the user via the input / output unit 103.

[0041] The abnormality diagnosis unit 150 is an example of a diagnosis unit according to the present disclosure, and when an abnormality is detected in the equipment 20 in a diagnosis phase described below, the abnormality diagnosis unit 150 diagnoses the cause of the abnormality in the equipment 20 using an abnormality diagnosis model. Specifically, the abnormality diagnosis unit 150 diagnoses the cause of the abnormality in the equipment 20 by evaluating the actual measurement data at the time of the abnormality output from the abnormality detection unit 140 using the abnormality location diagnosis model output from the abnormality diagnosis model learning unit 130. The diagnosis result of the abnormality cause is presented to the user via the input / output unit 103.

[0042] Next, a specific configuration of the abnormal data generating unit 120 according to this embodiment will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the configuration of the abnormal data generating unit 120 according to this embodiment.

[0043] As shown in FIG. 2, the abnormality data generating unit 120 includes a parameter determining unit 121 , a simulation executing unit 122 , an extracting unit 123 , a saving unit 124 , a calculating unit 125 , and a summarizing unit 126 .

[0044] The parameter determination unit 121 determines physical parameters to be used by the simulation execution unit 122. The physical parameters are an example of parameters for simulating an abnormality in the equipment 20.

[0045] In the present embodiment, the parameter determining unit 121 is an example of an updating unit according to the present disclosure, and updates the physical parameters based on the fluctuation features extracted by the extracting unit 123. The fluctuation features are information representing the features of the data generated under abnormal conditions, and more specifically, the amount of change in the data generated under abnormal conditions relative to the features of the actual measurement data under normal conditions that the anomaly detection model has learned.

[0046] For example, the parameter determination unit 121 updates the physical parameters based on the diversity calculated by the calculation unit 125. The diversity is the diversity of the fluctuation features extracted by the extraction unit 123. There is a positive correlation between the diversity of the fluctuation features and the diversity of the data generated under abnormal conditions. Specifically, the more diverse the fluctuation features, the more diverse the data generated under abnormal conditions. If the data generated under abnormal conditions is diverse, it becomes possible to estimate various abnormality factors.

[0047] Furthermore, the parameter determination unit 121 may update the parameters based on the summarization amount calculated by the summarization unit 126. The summarization amount is a summary amount of the distribution of the fluctuation features extracted by the extraction unit 123. The distribution of the fluctuation features has characteristics similar to the distribution of the abnormality-generated data. For example, if multiple fluctuation features are distributed evenly over a wide range, the multiple abnormality-generated data will also be distributed evenly over a wide range, i.e., the abnormality-generated data will be diverse.

[0048] The physical parameters may be determined from ranges set by the user on a display screen shown in Fig. 5, which will be described later. The values ​​of the physical parameters may be determined according to the distribution of the variation features output from the summarization unit 126 and the diversity of the variation features output from the calculation unit 125. The values ​​may be determined so as to increase the diversity of the variation features.

[0049] A method for determining physical parameters according to the diversity of the variable features and / or the distribution of the variable features may use black-box optimization such as Bayesian optimization or evolutionary strategy. In this case, the values ​​of the simulation parameters may be used as explanatory variables, and the diversity of the variable features and / or a summary of the distribution of the variable features may be used as the objective function. Furthermore, reinforcement learning such as policy gradient method and actor-critic may be used as a method for determining physical parameters. When using reinforcement learning, reinforcement learning may be performed using the parameter determination unit 121 as an agent, determining the value of the simulation parameter or determining the amount of change in the value of the simulation parameter as an action, the diversity of the variable features as a reward, the distribution of the variable features as a state, and the summary of the distribution of the variable features as an observation.

[0050] The simulation execution unit 122 uses a physical model based on the physical parameters determined or updated by the parameter determination unit 121 to simulate an abnormality in the equipment 20. The simulation execution unit 122 generates abnormality generation data by performing the simulation. The physical model may be an equivalent circuit model of the equipment 20 to be diagnosed, a more detailed FEM (Finite Element Method) model, or a proxy model that has learned the input / output relationships thereof.

[0051] The extraction unit 123 is an example of an extraction unit according to the present disclosure, and extracts fluctuation features of abnormality-generated data generated by simulating an abnormality in the equipment 20, using an anomaly detection model for the equipment 20. Specifically, the extraction unit 123 extracts fluctuation features of abnormality-generated data generated by the simulation execution unit 122, using the anomaly detection model. In other words, the extraction unit 123 extracts fluctuation features of the abnormality-generated data by evaluating the abnormality-generated data using the anomaly detection model. Here, the fluctuation feature is the amount of change in the abnormality-generated data relative to the features of normal-state actual measurement data learned by the anomaly detection model. For example, if a rule-based method is adopted as the anomaly detection model, in which a threshold is determined from normal-state actual measurement data for a frequency component associated with an abnormality factor, the frequency component associated with the abnormality factor may be extracted as the fluctuation feature. If an autoencoder is adopted as the anomaly detection model, the extraction unit 123 extracts a reconstruction error e defined by the following equation (1) using the reconstruction error of the autoencoder as the fluctuation feature: ｔ Vectors can be used.

[0052]

[0053] Here, f is an autoencoder that has learned the characteristics of the normal measured data, and is an anomaly detection model used for anomaly detection by the anomaly detection unit 140. ｔ is the abnormality generation data generated for the tth time.

[0054] The storage unit 124 stores the variation features output from the extraction unit 123. The storage unit 124 also outputs all of the variation features that have been previously stored to the calculation unit 125 and the summarization unit 126.

[0055] The calculation unit 125 is an example of a calculation unit according to the present disclosure, and calculates the diversity of the variation features extracted by the extraction unit 123. The calculation unit 125 can also be called a diversity evaluation unit. Specifically, the calculation unit 125 calculates the diversity of the variation features output from the storage unit 124, and outputs the calculated diversity to the parameter determination unit 121.

[0056] Diversity r ｔ can be defined as, for example, the average of the distances between the abnormality-generated data generated by performing a simulation and each of the multiple abnormality-generated data generated by performing simulations in the past before the abnormality-generated data. ｔ may be defined as the average of distances to a plurality of abnormality-generated data generated in the past, as in the following formula (2):

[0057]

[0058] Here, d is a function that represents the distance between vectors, such as cosine similarity.

[0059] Alternatively, diversity r ｔ can be defined as the average of the entropy of each dimension of the variation feature extracted by the extraction unit 123. Specifically, the diversity r ｔ may be defined as the average of the entropy of each dimension of the fluctuation features of a plurality of abnormality-generated data generated in the past, as in the following formula (3):

[0060]

[0061] where n is the number of dimensions of the variation feature. ｔ ｋ is the k-th element of the fluctuation feature of the abnormality-generated data generated at the tth time. h is a function that approximates the entropy from the data distribution.

[0062] The summarization unit 126 is an example of a summarization unit according to the present disclosure, and calculates a summary amount of the distribution of the fluctuation features extracted by the extraction unit 123. Specifically, the summarization unit 126 calculates a summary amount of the distribution of the fluctuation features of the abnormality-generated data stored in the storage unit 124, and outputs the calculated summary amount to the parameter determination unit 121.

[0063] The distribution summary quantity can be defined as a matrix including the mean and variance of each of the plurality of clusters in the variation feature space, for example, by classifying a plurality of abnormality-generated data generated by performing a simulation. Specifically, the distribution summary quantity is a matrix S ≡ ... ｔ It may be defined as:

[0064]

[0065] Here, n is the number of dimensions of the variation feature, m is the fixed number of clusters, μ indicates the center of gravity of the cluster, and σ indicates the variance of the cluster. In this case, the dimension of the summary quantity of the distribution does not depend on the amount of data generated up to that point, and it becomes possible to use this as an observation for reinforcement learning.

[0066] Alternatively, the distribution summary quantity can be defined as a matrix summarizing the frequency distribution of each dimension of the variation feature extracted by the extraction unit 123. Specifically, the distribution summary quantity is a matrix S summarizing the frequency distribution of each dimension of the variation feature shown in the following formula (5): ｔ It may be defined as:

[0067]

[0068] Here, l is the number of classes and w is the frequency. In this case, the dimension of the distribution summary does not depend on the amount of data generated so far, and it can be used as an observation for reinforcement learning.

[0069] Note that, if a small amount of actual measurement data under abnormal conditions collected from past failure cases or the like can be used in the learning phase, the evaluation using the small amount of actual measurement data under abnormal conditions may be added to the evaluation based on the diversity or distribution of the fluctuation features described above to update the simulation parameters. That is, the objective function of black-box optimization or the reward function of reinforcement learning may include a term for the similarity between the actual measurement data under abnormal conditions and the data generated under abnormal conditions, or a term for the diagnostic accuracy when a diagnostic model trained on the data generated under abnormal conditions is evaluated using the actual measurement data under abnormal conditions.

[0070] [Operation] Next, the operation of the equipment state estimation system 10 according to this embodiment will be described. First, an example of the learning phase processing among the processing performed by the equipment state estimation device 100 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the learning phase processing among the processing performed by the equipment state estimation device 100 according to this embodiment.

[0071] In step S10, the input / output unit 103 receives input of setting information necessary for processing by the information processing unit 101. The setting information includes, for example, parameter values ​​for a simulation of an abnormality in the equipment 20.

[0072] In step S11, the anomaly detection model learning unit 110 acquires physical quantities such as vibration and current of the equipment 20 as measured signals (measured data) using the sensors 30, and learns the features of the acquired measured data. Note that the learning phase is performed after confirming that the equipment 20 is operating normally. For this reason, all measured data acquired from the sensors 30 is considered to be normal measured data. In other words, the anomaly detection model learns normal features using normal measured data.

[0073] In step S12, the abnormality data generation unit 120 generates abnormality generation data. Specifically, the simulation execution unit 122 generates the abnormality generation data by simulating an abnormality in the equipment 20. The parameters of the simulation at this time are included in, for example, the setting information received in step S10.

[0074] In step S13, the extraction unit 123 extracts fluctuation features of the abnormality-generated data using the anomaly detection model. The extraction unit 123 stores the extracted fluctuation features in the storage unit 124.

[0075] In step S14, the calculation unit 125 calculates the diversity of the extracted variation feature. The information processing unit 101 outputs the diversity calculated by the calculation unit 125 to the input / output unit 103. The input / output unit 103 updates the diversity transition graph 207 (see FIG. 5 , which will be described later) based on the diversity obtained from the calculation unit 125.

[0076] In step S15, the input / output unit 103 updates the spatial distribution graph 206 (see FIG. 5, which will be described later) based on the fluctuation characteristics of the abnormality-generated data.

[0077] In step S16, the information processing unit 101 determines whether the termination condition is met. The termination condition is determined, for example, based on the number of abnormality-generated data items generated. For example, if the number of abnormality-generated data items generated exceeds a value input by the user as the termination condition via the input / output unit 103, the information processing unit 101 determines to terminate the learning phase. Alternatively, the termination condition may be determined based on a user instruction. For example, if the user presses the generation stop button 204 (see FIG. 5 described later), the information processing unit 101 may determine to terminate the learning phase. If the termination condition is not met, the process proceeds to step S17. If the termination condition is met, the process proceeds to step S18. Note that the termination condition is not limited to the example described above. For example, the termination condition may be determined based on the duration of the learning phase. If the duration of the learning phase exceeds a predetermined duration, the information processing unit 101 may determine to terminate the learning phase.

[0078] If the termination condition is not satisfied (No in S16), in step S17, the parameter determination unit 121 updates the simulation parameters based on the diversity and / or distribution of the variable features. If black-box optimization is used to update the parameters, a new value of the objective function is obtained in step S17. If reinforcement learning is used to update the parameters, new values ​​of the reward, state, and observation are obtained in step S17.

[0079] After updating the parameters, the process proceeds to step S12, where a simulation is performed to generate abnormality generation data. The simulation uses the parameters updated by the parameter determination unit 121. Thereafter, steps S12 to S17 are repeated until it is determined in step S16 that the termination condition is met.

[0080] If the termination condition is satisfied (Yes in S16), in step S18, the abnormality diagnosis model learning unit 130 uses the generated abnormality generation data to learn an abnormality diagnosis model.

[0081] In step S19, the abnormality diagnosis model learning unit 130 stores the generated abnormality diagnosis model in the database 102 or an internal memory of the abnormality diagnosis unit 150. This ends the learning phase and transitions to the diagnosis phase.

[0082] Next, an example of the processing in the diagnosis phase among the processing performed by the equipment state estimation device 100 according to the present embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing the processing in the diagnosis phase among the processing performed by the equipment state estimation device 100 according to the present embodiment.

[0083] In step S20, the information processing unit 101 acquires physical quantities such as vibration and current of the equipment 20 as measured signals (measured data) using the sensor 30. The anomaly detection unit 140 inputs the acquired measured data into an anomaly detection model. In the diagnosis phase, the measured data output from the sensor 30 is considered to be either "normal" or "abnormal."

[0084] In step S21, the anomaly detection unit 140 uses the anomaly detection model to determine whether the state of the equipment 20 is "normal" or "abnormal." If it is "abnormal," the process proceeds to step S22. If it is "normal," the process proceeds to step S23. At this time, the anomaly detection unit 140 also uses the anomaly detection model to extract fluctuation features of the actual measurement data. Note that the extraction of the fluctuation features may be performed by the extraction unit 123 of the anomaly data generation unit 120.

[0085] If the equipment 20 is determined to be "abnormal" (Yes in S21), in step S22, the abnormality diagnosis unit 150 inputs the actual measurement data or the fluctuation characteristics of the actual measurement data into the abnormality diagnosis model. Then, the abnormality diagnosis unit 150 identifies the cause of the abnormality of the equipment 20 using the abnormality diagnosis model.

[0086] After the cause of the abnormality has been identified, or if the equipment 20 has been determined to be "normal" (No in S21), in step S23, the input / output unit 103 displays the actual measurement data or the fluctuation characteristics of the actual measurement data together with the diagnosis results in the fluctuation characteristic space. Specifically, the input / output unit 103 displays the fluctuation characteristics of the actual measurement data in a spatial distribution graph 206 (see FIG. 5, which will be described later).

[0087] In step S24, the user determines whether regeneration of the abnormality-generated data is necessary based on the position of the actual measurement data in the spatial distribution graph 206. If the user determines that regeneration is necessary (Yes in S24), the input / output unit 103 accepts a regeneration instruction from the user. As a result, the equipment state estimation device 100 ends the diagnosis phase and moves to the learning phase. If regeneration is necessary, the process may be executed from step S12 in the flowchart shown in FIG. 3. If regeneration is not necessary (No in S24), the process moves to step S20, and the process of step S20 is performed on new actual measurement data. Thereafter, the processes of steps S20 to S24 are repeated until regeneration is necessary or until the diagnosis phase is ended.

[0088] [Display Example] Next, an example of a display screen displayed by the input / output unit 103 of the equipment state estimation device 100 according to the present embodiment will be described with reference to Fig. 5. Fig. 5 is a diagram showing an example of a display screen displayed by the input / output unit 103 of the equipment state estimation device 100 according to the present embodiment.

[0089] The parameter generation condition table 201 is an example of a GUI object for receiving input from a user of parameter generation conditions necessary for executing a simulation by the simulation execution unit 122. The parameter determination unit 121 determines or updates simulation parameters based on the conditions input in the parameter generation condition table 201.

[0090] 5, the parameter generation condition table 201 includes an input field 211 for inputting the type of parameter to be used in the simulation, and input fields 212, 213, and 214 for inputting the lower limit (minimum value), upper limit (maximum value), and initial value of the value to be used in generating the parameter. When there are multiple parameters, input fields 211, 212, 213, and 214 are provided for each parameter. Note that the input format for each input field may be such that text and numerical values ​​can be directly input, or that values ​​can be selected from multiple candidate values.

[0091] The execution condition table 202 is an example of a GUI object for receiving input of execution conditions for generating abnormality-occurrence data by simulation from a user. The execution condition table 202 includes an input field 221 for inputting the maximum number of pieces of generated data and an input field 222 for inputting the display interval.

[0092] The maximum number of generated data is the maximum number of abnormality-generated data to be generated, and can be used as one of the termination conditions for step S16 in Fig. 3. The display interval is the number of abnormality-generated data for updating the display of the spatial distribution graph 206 and the diversity transition graph 207. For example, in the example shown in Fig. 5, the display of the spatial distribution graph 206 and the diversity transition graph 207 is updated every time 10 abnormality-generated data are generated.

[0093] The generation start button 203 is an example of a GUI object for starting or resuming generation of abnormality-time generation data when selected by the user. For example, if it is determined in step S24 of FIG. 4 that regeneration is necessary, the user can select the generation start button 203.

[0094] The generation stop button 204 is an example of a GUI object for stopping or terminating the generation of abnormality-generated data when selected by the user. For example, this can be used as one of the termination conditions of step S16 in FIG.

[0095] Physical parameter distribution information 205 is information indicating the distribution of physical parameters used in generating abnormality generation data, i.e., in the simulation. Distributions 251, 252, and 253 correspond to parameter A, parameter B, and parameter C, respectively, entered in parameter generation condition table 201. The horizontal axis of each distribution represents the parameter value, and the vertical axis represents the frequency with which the corresponding value was used in the simulation.

[0096] The spatial distribution graph 206 is a graph of the variation feature space. Specifically, the spatial distribution graph 206 is a graph showing the spatial distribution of the variation features of the generated data and the measured data. Type 1, Type 2, and Type 3 correspond to the types of abnormality factors, respectively. By color-coding the markers for each Type, they can be easily visually distinguished. In the example shown in FIG. 5 , Type 1, Type 2, and Type 3 each form sets 261, 262, and 263, which have a certain degree of cohesion and spread, in the variation feature space.

[0097] Furthermore, the difference in the shape of the markers represents the difference between the measured data and the generated data. In the example shown in Fig. 5, the circle markers represent generated data, and the star marker 264 represents measured data. Note that if the variation features are four or more dimensions, they may be displayed as a three-dimensional graph using a dimensional reduction technique such as principal component analysis.

[0098] The diversity transition graph 207 is a graph showing the transition of the diversity of the fluctuation characteristics of the abnormal-state generated data. In the diversity transition graph 207, the vertical axis represents the diversity, and the horizontal axis represents the number of abnormal-state generated data generated.

[0099] As described above, in this embodiment, the input / output unit 103 displays the fluctuation feature space in the form of a graph (e.g., spatial distribution graph 206). The input / output unit 103 displays abnormality-occurring data generated by performing a simulation in the fluctuation feature space (e.g., sets 261, 262, and 263). The input / output unit 103 also displays actual measurement data detected as abnormal using the anomaly detection model in the fluctuation feature space (e.g., marker 264).

[0100] The input / output unit 103 also displays a graph of the change in diversity (e.g., a diversity transition graph 207). The input / output unit 103 also displays a first GUI object for accepting a user operation to stop and restart the execution of the simulation, i.e., to stop and restart the generation of abnormality generation data (e.g., a generation start button 203 or a generation stop button 204). The input / output unit 103 also displays a second GUI object for accepting a user operation to instruct the regeneration of the abnormality diagnosis model (e.g., a generation start button 203). Note that, although the case where the generation start button 203 is an example of each of the first GUI object and the second GUI object according to the present disclosure has been described here, the first GUI object and the second GUI object may each be provided separately.

[0101] Note that the display screen shown in FIG. 5 is merely an example and may be modified as appropriate. At least one of the elements of the parameter generation condition table 201, the execution condition table 202, the generation start button 203, the generation stop button 204, the distribution information 205, the spatial distribution graph 206, and the diversity transition graph 207 may not be included in the display screen. Also, although an example has been shown in which all elements (GUI objects) are displayed on a single display screen, this is not limiting. Multiple elements to be presented to the user may be distributed and displayed across multiple display screens or display windows, or may be displayed by scrolling or switching screens.

[0102] [Effects, etc.] Next, an example of the effects of the equipment state estimation device 100 according to the present embodiment will be described with reference to Fig. 6. Fig. 6 is a diagram for explaining the effects of the equipment state estimation device 100 according to the present embodiment.

[0103] Distribution information 301 indicates the distribution of physical parameters when the physical parameters are sampled from a uniform distribution within the range set in the parameter generation condition table 201 in Fig. 5. Distributions 311, 312, and 313 correspond to parameter A, parameter B, and parameter C, respectively.

[0104] The spatial distribution graph 303 shows the distribution of the fluctuation characteristics of the abnormality-generated data generated according to the physical parameters shown in the distribution information 301 .

[0105] Typically, physical models and anomaly detection models are nonlinear models with respect to the input. Therefore, the fluctuation features also have a nonlinear relationship with the physical parameters. Therefore, the distribution of the fluctuation features of the abnormality-generated data generated according to the physical parameters shown in the distribution information 301 is biased, resulting in low diversity. For example, as shown in FIG. 6 , the fluctuation features included in set 331 are concentrated in a narrow range within the fluctuation feature space. The same is true for sets 332 and 333. In other words, the generated abnormality-generated data contains a lot of data with similar features, and a wide variety of features cannot be fully generated.

[0106] Therefore, for example, the measured data input to the anomaly diagnosis model may have fluctuation characteristics that are different from any of the data generated under abnormal conditions that have been learned in the past. Specifically, as shown in Fig. 6, there is a possibility that the marker 334 representing the fluctuation characteristics of the measured data will not be included in the range of existence of any of the sets 331, 332, and 333. Therefore, it is conceivable that it may be difficult to determine which anomaly cause of the measured data should be classified as.

[0107] In contrast, distribution information 302 indicates the distribution of physical parameters when the physical parameters are sampled from the range set in the parameter generation condition table 201 so as to maximize the diversity of the variation features. Distributions 321, 322, and 323 correspond to the input parameters A, B, and C, respectively. Note that distributions 321, 322, and 323 are the same as distributions 251, 252, and 253 on the display screen in Figure 5, respectively.

[0108] The spatial distribution graph 304 shows the distribution of the fluctuation characteristics of the abnormality-generated data generated according to the physical parameters shown in the distribution information 302. The abnormality-generated data included in the spatial distribution graph 304 has more diverse fluctuation characteristics than the abnormality-generated data included in the spatial distribution graph 303. Specifically, sets 341, 342, and 343 all form larger sets than sets 331, 332, and 333. For example, the fluctuation characteristics included in set 341 exist in a wider range in the fluctuation characteristic space than the fluctuation characteristics included in set 331. Therefore, even when unknown actual measurement data is input to the abnormality diagnosis model, the abnormality factor can be diagnosed with high accuracy. For example, as shown in FIG. 6 , marker 344 representing the fluctuation characteristics of the actual measurement data is included in the range of existence of set 341. Therefore, it can be determined that the abnormality factor of the actual measurement data is the abnormality factor "Type 1" corresponding to set 341.

[0109] In this embodiment, the diversity of the variation characteristics from the normal measured data is evaluated, rather than the diversity of the abnormal-state generated data itself. This makes it possible to take into account the influence of variations within the normal measured data, and to generate abnormal-state generated data with diverse abnormal characteristics with a small number of generation attempts.

[0110] As described above, according to one aspect of the present disclosure, by extracting fluctuation features of data generated under abnormal conditions using an anomaly detection model that has learned the features of data measured under normal conditions, it is possible to appropriately improve physical parameters of a diagnosis target without comparing the data with data measured under abnormal conditions. This makes it possible to generate diverse data generated under abnormal conditions, which can contribute to realizing highly accurate estimation of the cause of an abnormality.

[0111] (Other) While various embodiments have been described above with reference to the drawings, it goes without saying that the present disclosure is not limited to such examples. It is clear that a person skilled in the art can conceive of various modifications or alterations within the scope of the claims, and it is understood that these also naturally fall within the technical scope of the present disclosure. Furthermore, the components of the above-described embodiments may be combined in any manner without departing from the spirit of the disclosure.

[0112] 1 may include the abnormality data generation unit 120, but may not include the abnormality detection model learning unit 110, the abnormality diagnosis model learning unit 130, the abnormality detection unit 140, and the abnormality diagnosis unit 150. Furthermore, the abnormality data generation unit 120 shown in FIG. 2 may include the extraction unit 123 and an update unit that updates simulation parameters, but may not include the simulation execution unit 122, the storage unit 124, the calculation unit 125, and the summarization unit 126. In other words, the information processing device according to the present disclosure may be realized with a simpler configuration than the configurations shown in the above-described embodiments.

[0113] 7 is a block diagram showing the configuration of an information processing device 400 according to a modified example of the embodiment. As shown in FIG. 7, the information processing device 400 includes an extraction unit 401 and an update unit 402. The information processing device 400 is also connected to an external simulation execution unit 410 via wired or wireless communication so as to enable input and output of information, data, signals, and the like to and from the external simulation execution unit 410.

[0114] The extraction unit 401 extracts fluctuation features of abnormality-generated data generated by simulating an equipment abnormality, using an equipment abnormality detection model. The simulation is performed by a simulation execution unit 410 different from the information processing device 400. The abnormality detection model is acquired from another device (not shown) different from the information processing device 400. The fluctuation features extracted by the extraction unit 401 are output to the simulation execution unit 410 and the other device (not shown) different from the information processing device 400.

[0115] The update unit 402 updates the simulation parameters based on the variation characteristics extracted by the extraction unit 401. The update unit 402 outputs the parameters updated based on the variation characteristics to the simulation execution unit 410.

[0116] 8 is a flowchart showing the operation of information processing device 400 according to a modified example of the embodiment. The operation of information processing device 400 is an example of an information processing method according to the present disclosure. As shown in FIG. 8 , the information processing method includes step S30 of extracting, using an equipment anomaly detection model, fluctuation features of abnormality-occurring data generated by simulating an equipment anomaly, and step S31 of updating simulation parameters based on the extracted fluctuation features.

[0117] In this way, according to the information processing device 400, similar to the above-described embodiment, by extracting fluctuation features of data generated under abnormal conditions using an anomaly detection model that has learned the features of data actually measured under normal conditions, it is possible to appropriately improve the physical parameters of the diagnosis target without comparing them with data actually measured under abnormal conditions. This makes it possible to support the generation of a variety of data generated under abnormal conditions and contribute to realizing highly accurate estimation of the cause of an abnormality.

[0118] Furthermore, in each of the above embodiments, the present disclosure has been described as an example configured using hardware, but the present disclosure can also be realized by software in cooperation with hardware.

[0119] Furthermore, each functional block used in the description of each of the above embodiments is typically realized as an LSI (Large Scale Integration), which is an integrated circuit. The integrated circuit controls each functional block used in the description of the above embodiments and may include an input section and an output section. These may be individually integrated into a single chip, or some or all of them may be integrated into a single chip. Here, the term LSI is used, but depending on the degree of integration, it may also be called an IC (Integrated Circuit), system LSI, super LSI, or ultra LSI.

[0120] Furthermore, the method of integration is not limited to LSI, and may be realized using a dedicated circuit or a general-purpose processor. It is also possible to use a field programmable gate array (FPGA) that can be programmed after LSI manufacturing, a reconfigurable processor that can reconfigure the connections or settings of circuit cells inside the LSI, or the like.

[0121] Furthermore, if an integrated circuit technology that can replace LSI emerges due to advances in semiconductor technology or other derivative technologies, it is natural that such technology may be used to integrate functional blocks. Possible applications include biotechnology and optical integrated circuits.

[0122] Furthermore, the general or specific aspects of the present disclosure may be realized as a system, an apparatus, a method, an integrated circuit, or a computer program. Alternatively, the general or specific aspects may be realized as a computer-readable non-transitory recording medium such as an optical disk, a HDD, or a semiconductor memory on which the computer program is stored. Alternatively, the general or specific aspects of the present disclosure may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

[0123] Furthermore, various modifications, substitutions, additions, omissions, etc. can be made to each of the above-described embodiments within the scope of the claims or their equivalents.

[0124] Summary of the present disclosure An information processing device according to a first aspect of the present disclosure includes an extraction unit that extracts fluctuation features of abnormality-generated data generated by simulating an abnormality in equipment using an abnormality detection model of the equipment, and an update unit that updates parameters of the simulation based on the fluctuation features extracted by the extraction unit.

[0125] In this way, the parameters are updated based on the fluctuation characteristics, which can support the generation of a variety of abnormality generation data.

[0126] An information processing device according to a second aspect of the present disclosure is an information processing device according to the first aspect, and is provided with a calculation unit that calculates the diversity of the variable features extracted by the extraction unit, and the update unit updates the parameters based on the diversity calculated by the calculation unit.

[0127] This makes it possible to update parameters so as to maximize diversity, for example, and to support the generation of diverse abnormality generation data.

[0128] An information processing device according to a third aspect of the present disclosure is an information processing device according to the second aspect, wherein the calculation unit calculates the average distance between the abnormality-generated data generated by performing the simulation and each of multiple abnormality-generated data generated by performing the simulation earlier than the abnormality-generated data, as the diversity.

[0129] This makes the diversity independent of the number of data sets, making it applicable to reinforcement learning.

[0130] An information processing device according to a fourth aspect of the present disclosure is the information processing device according to the second aspect, wherein the calculation unit calculates the average entropy of each dimension of the variation feature extracted by the extraction unit as the diversity.

[0131] This makes the diversity independent of the number of data sets, making it applicable to reinforcement learning.

[0132] An information processing device according to a fifth aspect of the present disclosure is an information processing device according to any one of the first to fourth aspects, wherein the anomaly detection model is an autoencoder, and the extraction unit extracts a reconstruction error of the autoencoder as the fluctuation feature.

[0133] This allows us to easily extract variation features using an autoencoder.

[0134] An information processing device according to a sixth aspect of the present disclosure is an information processing device according to any one of the first to fifth aspects, and includes a summarization unit that calculates a summary amount of the distribution of the fluctuation features extracted by the extraction unit, and the update unit updates the parameters based on the summary amount calculated by the summarization unit.

[0135] This allows the parameters to be updated so that the fluctuation characteristics are diversified based on the distribution of the fluctuation characteristics, thereby supporting the generation of diverse abnormality generation data.

[0136] An information processing device according to a seventh aspect of the present disclosure is an information processing device according to the sixth aspect, wherein the summarization unit classifies a plurality of abnormality-generated data generated by performing the simulation into a plurality of clusters in a variation feature space, and calculates a matrix including the mean and variance of each of the plurality of clusters as the summary quantity.

[0137] This makes the dimension of the distribution summary independent of the number of data, making it applicable to reinforcement learning.

[0138] An information processing device according to an eighth aspect of the present disclosure is an information processing device according to the sixth aspect, wherein the summarization unit calculates, as the summary quantity, a matrix that aggregates the frequency distribution of each dimension of the variation features extracted by the extraction unit.

[0139] This makes the dimension of the distribution summary independent of the number of data, making it applicable to reinforcement learning.

[0140] An information processing device according to a ninth aspect of the present disclosure is an information processing device according to any one of the first to eighth aspects, and includes a display unit that displays a variation feature space in a graph, and the display unit displays abnormality generation data generated by performing the simulation within the variation feature space.

[0141] This allows the user to visually grasp the fluctuation characteristics of the abnormality-generated data easily, making it easier to decide whether to adjust the parameters or to stop the generation of abnormality-generated data.

[0142] An information processing device according to a tenth aspect of the present disclosure is the information processing device according to any one of the second to fourth aspects, and includes a display unit that displays a graph of the change in the diversity.

[0143] This allows the user to monitor the diversity of the fluctuation characteristics of the abnormality-generated data being generated while the simulation is running. For example, it is possible to check how the diversity increases as the simulation is run, making it easier to decide whether to adjust the parameters or to stop generating the abnormality-generated data.

[0144] An information processing device according to an eleventh aspect of the present disclosure is an information processing device according to any one of the first to eighth aspects, wherein the display unit displays a first GUI object for accepting user operations to stop and restart the execution of the simulation.

[0145] This allows the user to instruct the stopping and starting of the simulation at a timing desired by the user. For example, if the user determines that a sufficient amount of diverse abnormality-generated data has been generated, the execution of the simulation can be stopped, which contributes to reducing the power consumption required for processing. Furthermore, if the user determines that the abnormality-generated data is insufficient, the execution of the simulation can be resumed to generate abnormality-generated data. This prevents the cause of the abnormality from being estimated based on inappropriate abnormality-generated data.

[0146] An information processing device according to a twelfth aspect of the present disclosure is an information processing device according to any one of the first to eighth aspects, and includes: a generation unit that generates an abnormality diagnosis model that diagnoses an abnormality factor of the equipment based on a plurality of abnormality-generated data generated by performing the simulation or on variation characteristics of each of the plurality of abnormality-generated data; and a diagnosis unit that diagnoses the abnormality factor of the equipment using the abnormality diagnosis model when an abnormality of the equipment is detected using the abnormality detection model.

[0147] This allows an abnormality diagnosis model to be generated based on data generated during various abnormal conditions, thereby improving the accuracy of estimating the cause of an abnormality.

[0148] An information processing device according to a thirteenth aspect of the present disclosure is the information processing device according to the twelfth aspect, and includes a display unit that displays a variation feature space in a graph, and the display unit displays, within the variation feature space, abnormality-generated data generated by performing the simulation and actual measurement data detected as an abnormality using the anomaly detection model.

[0149] This allows users to visually grasp the fluctuation characteristics of the measured data with ease, and also enables users to judge whether the diagnostic results are correct or not based on their own experience.

[0150] An information processing device according to a fourteenth aspect of the present disclosure is the information processing device according to the thirteenth aspect, wherein the display unit displays a second GUI object for accepting a user operation instructing regeneration of the abnormality diagnosis model.

[0151] This allows the user to regenerate the abnormality diagnosis model if the user determines that the diagnosis result is inappropriate, thereby preventing the cause of the abnormality from being estimated based on an inappropriate abnormality diagnosis model.

[0152] An information processing method according to a fifteenth aspect of the present disclosure includes the steps of extracting fluctuation characteristics of abnormality-generated data generated by simulating an abnormality in equipment using an abnormality detection model of the equipment, and updating parameters of the simulation based on the extracted fluctuation characteristics.

[0153] This makes it possible to support the generation of a variety of abnormality generation data, similar to the information processing devices according to the above-described aspects.

[0154] A program according to a sixteenth aspect of the present disclosure is a program that causes a computer to execute the information processing method according to the fifteenth aspect.

[0155] This makes it possible to support the generation of a variety of abnormality generation data, similar to the information processing devices according to the above-described aspects.

[0156] The present disclosure can be used in equipment abnormality estimation systems that detect abnormalities in equipment and estimate the causes of the abnormalities, and fault diagnosis systems that diagnose equipment failures.

[0157] REFERENCE SIGNS LIST 10 Equipment state estimation system 20 Equipment 30 Sensor 100 Equipment state estimation device 101 Information processing unit 102 Database 103 Input / output unit 110 Anomaly detection model learning unit 120 Anomaly data generation unit 121 Parameter determination unit 122, 410 Simulation execution unit 123, 401 Extraction unit 124 Storage unit 125 Calculation unit 126 Summarization unit 130 Anomaly diagnosis model learning unit 140 Anomaly detection unit 150 Anomaly diagnosis unit 201 Parameter generation condition table 202 Execution condition table 203 Generation start button 204 Generation stop button 205, 301, 302 Distribution information 206, 303, 304 Spatial distribution graph 207 Diversity transition graph 211, 212, 213, 214, 221, 222 Input field 251, 252, 253, 311, 312, 313, 321, 322, 323 Distribution 261, 262, 263, 331, 332, 333, 341, 342, 343 Set 264, 334, 344 Marker 400 Information processing device 402 Update unit

Claims

1. An extraction unit that extracts the fluctuation characteristics of abnormal data generated by simulating equipment malfunctions using the equipment malfunction detection model, The system includes an update unit that updates the simulation parameters based on the variation features extracted by the extraction unit, Information processing device.

2. The system includes a calculation unit that calculates the diversity of the variable features extracted by the extraction unit, The update unit displays the parameters based on the variability calculated by the calculation unit. The information processing apparatus according to claim 1.

3. The calculation unit calculates the diversity as the average distance between the abnormal time generation data generated by performing the simulation and each of the multiple abnormal time generation data generated by performing the simulation in the past prior to the said abnormal time generation data. The information processing apparatus according to claim 2.

4. The calculation unit calculates the average of the entropy of each dimension of the variation features extracted by the extraction unit as the diversity. The information processing apparatus according to claim 2.

5. The calculation unit adds the evaluation based on the actual measurement data during abnormal times to the diversity of the variation characteristics. The information processing apparatus according to claim 2.

6. The aforementioned anomaly detection model is an autoencoder, The extraction unit extracts the reconstruction error of the autoencoder as the variation feature. The information processing apparatus according to claim 1.

7. The system includes a summarization unit that calculates a summarization amount of the distribution of the variation features extracted by the extraction unit, The update unit updates the parameters based on the summarization amount calculated by the summarization unit. The information processing apparatus according to claim 1.

8. The summarization unit classifies the multiple anomaly-generated data generated by performing the simulation into multiple clusters within the variable feature space, and calculates a matrix containing the mean and variance of each of the multiple clusters as the summarization quantity. The information processing apparatus according to claim 7.

9. The summarization unit calculates a matrix, which aggregates the frequency distributions of each dimension of the variation features extracted by the extraction unit, as the summary quantity. The information processing apparatus according to claim 7.

10. It includes a display unit that graphs and displays the variable feature space, The display unit displays the abnormal time generation data generated by performing the simulation within the variable feature space. The information processing apparatus according to any one of claims 1 to 9.

11. The system includes a display unit that graphs and displays the changes in the aforementioned diversity. The information processing apparatus according to any one of claims 2 to 4.

12. The display unit displays a first GUI (Graphical User Interface) object for receiving user operations to stop and restart the execution of the simulation. The information processing apparatus according to claim 11.

13. A generation unit that generates multiple abnormal time generation data generated by performing the above simulation, or an abnormality diagnosis model that diagnoses the cause of the abnormality of the equipment based on the variation characteristics of each of the multiple abnormal time generation data, The system includes a diagnostic unit that, when an abnormality in the equipment is detected using the abnormality detection model, diagnoses the cause of the abnormality in the equipment using the abnormality diagnosis model. The information processing apparatus according to any one of claims 1 to 9.

14. It includes a display unit that graphs and displays the variable feature space, The display unit displays the abnormal time generation data generated by performing the simulation and the measured data detected as abnormal using the abnormality detection model within the variable feature space. The information processing apparatus according to claim 13.

15. The display unit displays a second GUI object for receiving user input to instruct the regeneration of the anomaly diagnosis model. The information processing apparatus according to claim 14.

16. The steps include: extracting the variation characteristics of the abnormal data generated by simulating equipment malfunctions using the equipment malfunction detection model; The step of updating the simulation parameters based on the extracted variation features includes: Information processing methods.

17. A program that causes a computer to execute the information processing method described in claim 16.