JP2020119267A - Information processing device, information processing program, and information processing method - Google Patents

Abstract

To provide an information processing device capable of finding a sign of potential failure that cannot be detected by human eyes.SOLUTION: An information processing device comprises an estimation unit that has a classification model generated by machine learning of historical data of one or more sensors provided in a rotary machine, inputs new time series data of the sensors to the classification model, and estimates whether the rotary machine has a sign of failure based on a trend of outputs from the classification model.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing device, an information processing program, and an information processing method.

2. Description of the Related Art Conventionally, an expert visually determines the presence or absence of a sign of an abnormality from the data of a sensor attached to mechanical equipment. However, even if the influencing factors are known, the degree of contribution that becomes abnormal depending on the combination is usually unknown, and even a skilled expert cannot find a sign of a potential failure by visual judgment. There is.

Patent Document 1 proposes an information processing unit that mechanically determines that there is a sign of abnormality when the average rate of change (rate of increase) of the suction temperature or the discharge temperature of the compressor in a predetermined period is equal to or greater than a predetermined value. ing.

Patent No. 5887217

According to the technique proposed in Patent Document 1, it becomes possible to determine the presence or absence of a sign of abnormality even without being a skilled expert, and it becomes possible to cope with a shortage of manpower. Since the judgment criteria and judgment methods used by experts are simply replaced by numerical values, it is not possible to find a sign of a potential failure that cannot be seen by human eyes.

The present invention has been made in consideration of the above points. An object of the present invention is to provide an information processing device, an information processing program, and an information processing method capable of finding a sign of a potential malfunction that cannot be seen by human eyes.

An information processing apparatus according to the first aspect of the present invention is
It has a classification model generated by machine learning of past data of one or more sensors related to the operating state of a rotating machine, and inputs new time-series data of the sensor into the classification model to perform the classification. An estimation unit for estimating whether or not the rotary machine has a sign of failure based on the trend of the output of the model.

According to this aspect, the classification model generated by machine learning the past data of the sensor is used to determine whether or not there is a sign of failure in the rotating machine from the new time series data of the sensor. In order to estimate, the judgment criteria and judgment method are unrelated to the knowledge and experience of specialists, and it is possible to find a sign of a potential failure that cannot be seen by human eyes.

An information processing apparatus according to a second aspect of the present invention is the information processing apparatus according to the first aspect,
The estimation unit estimates that the rotating machine has a sign of failure when the output trend of the classification model is a trend in which the probability of normal classification decreases with the passage of time.

An information processing apparatus according to a third aspect of the present invention is the information processing apparatus according to the first or second aspect,
Further comprising a storage unit that stores the type of failure in the maintenance record of the rotating machine in association with the type of classification by the classification model,
When the estimation unit determines that there is a sign of failure in the rotating machine, it identifies a classification in which the probability of being classified in the output trend of the classification model increases with the passage of time, The type of failure of the rotating machine is estimated based on the type of failure associated with the classification.

According to such a mode, not only a sign of a potential malfunction that cannot be seen by human eyes can be found, but also the kind of the malfunction can be estimated. As a result, it becomes possible to promptly respond to the type of defect.

An information processing apparatus according to a fourth aspect of the present invention is the information processing apparatus according to any one of the first to third aspects,
The sensor measures one or more of sound, vibration, temperature, current, voltage, frequency, pressure, flow rate, rotation speed, and AE (acoustic emission).

An information processing apparatus according to a fifth aspect of the present invention is the information processing apparatus according to any one of the first to fourth aspects,
The data is frequency, phase, and intensity data that is converted and extracted from the output of the sensor.

According to such an aspect, the amount of data can be significantly reduced, which makes it possible to speed up the processing.

An information processing apparatus according to a sixth aspect of the present invention is the information processing apparatus according to any one of the first to fifth aspects,
The rotary machine is any one of a pump, a blower, a refrigerator, a compressor, a dust remover, an agitator, a valve, a CMP device, a plating device, an electric motor, and a steam turbine.

An information processing apparatus according to a seventh aspect of the present invention is the information processing apparatus according to any one of the first to sixth aspects,
The system further includes a classification model generation unit that machine-learns past data of one or more sensors related to the operating state of the rotating machine to generate the classification model.

An information processing program according to an eighth aspect of the present invention is
Computer,
It has a classification model generated by machine learning of past data of one or more sensors related to the operating state of a rotating machine, and inputs new time-series data of the sensor into the classification model to perform the classification. Based on the trend of the output of the model, it functions as an estimation unit that estimates whether or not there is a sign of failure in the rotating machine.

An information processing method according to the ninth aspect of the present invention is
The time-series data of the sensor is input to the classification model in a classification model generated by machine learning of past data of one or more sensors related to the operating state of the rotating machine, and the output of the classification model is output. Based on the trend, it is estimated whether the rotary machine has a sign of failure.

According to the present invention, it is possible to find a symptom of a defect that cannot be seen by human eyes.

FIG. 1 is a block diagram showing a configuration of a system including an information processing device according to an embodiment. FIG. 2 is a schematic diagram showing the configuration of the rotary machine system according to the embodiment. FIG. 3 is a flowchart showing an example of an information processing method by the information processing device according to the embodiment. FIG. 4 is a conceptual diagram for explaining an example of processing of the classification model generation unit. FIG. 5 is a diagram showing an example of the defect association table. FIG. 6 is a conceptual diagram for explaining an example of processing of the estimation unit.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. Note that, in the following description and the drawings used in the following description, the same reference numerals are used for parts that may be configured the same, and duplicate description is omitted.

FIG. 1 is a block diagram showing the configuration of a system 100 including an information processing device 10 according to an embodiment.

As shown in FIG. 1, the system 100 according to the present embodiment includes a plurality of rotary machine systems 1 a to 1 c and an information processing device 10. Of these, the information processing device 10 includes rotating machine system control units 102a to 102c that control the rotating machine systems 1a to 1c, respectively, and an information processing unit 101 that can communicate with the rotating machine system control units 102a to 102c. ing. Hereinafter, the rotary machine system and the rotary machine system control unit forming a pair may be described by representing them with reference numerals 1 and 102, respectively.

FIG. 2 is a schematic diagram showing the configuration of the rotary machine system 1 according to one embodiment. Although a power generation system will be described below as an example of the rotary machine system 1, the power generation system is merely an example, and the rotary machine system according to the present embodiment is not limited to the power generation system. , Water treatment plant, sewage treatment plant, pumping plant, drainage plant, power plant (hydropower, thermal power, nuclear power, biomass, geothermal), petroleum/chemical plant, factory (shipbuilding/heavy machinery/steel/pulp pulp, etc.) ..

In the example shown in FIG. 2, the rotary machine system 1 is a power generation system, and includes a forced draft fan 2, a furnace 3, a boiler 4, a fan 5, a steam turbine 6, a boiler feed pump 7, and equipment. It has a cooling water pump 8.

Of these, the forced air blower 2 is a rotary machine, and the air in the atmosphere is sent to the furnace 3 by rotating the impeller. The furnace 3 burns combustible materials (coal, petroleum, LNG, wood, garbage, etc.) using the air supplied from the forced draft fan 2 to generate ash and exhaust gas.

The ventilator 5 is a rotating machine, and attracts the exhaust gas generated in the furnace 3 by rotating the impeller and discharges it from the chimney into the atmosphere. The boiler 4 utilizes the heat of the exhaust gas attracted from the furnace 3 to the blower 5 to heat the water supplied from the boiler feed pump 7 to generate steam.

The steam turbine 6 generates electricity and water by rotating the impeller using the energy of the steam generated by the boiler 4. The boiler feed water pump 7 is a rotary machine, and sends the water generated in the steam turbine 6 to the boiler 4 by rotating the impeller.

The equipment cooling water pump 8 is a rotating machine, and rotates the impeller to circulate the cooling water in each of the equipment 2 to 7.

Each rotary machine is provided with one or more sensors. One or more sensors can be selected from among sound, vibration, temperature, current, voltage, power frequency, pressure, flow rate, rotation speed, AE (acoustic emission), shaft displacement, shaft torque, equipment casing distortion, and frame distortion. One or two or more may be measured. In the illustrated example, the forced draft fan 2 is provided with a temperature sensor 2a, a current sensor 2b, a vibration sensor 2c, and an inlet air flow rate sensor 2d. Further, the furnace 3 is provided with a furnace pressure sensor 3a. Further, the fan 5 is provided with a temperature sensor 5a, a current sensor 5b, a vibration sensor 5c, and a stack exhaust gas flow rate sensor 5d. Further, the boiler water supply pump 7 is provided with a temperature sensor 7a, a current sensor 7b, a vibration sensor 7c, and a water supply flow rate discharge pressure sensor 7d. Further, the equipment cooling water pump 8 is provided with a temperature sensor 8a, a current sensor 8b, a vibration sensor 8c, and a cooling water temperature sensor 8d. The output of each sensor is transmitted to the information processing device 10.

Next, the configuration of the information processing device 10 will be described. In the following description, the information processing device 10 will be described as an example of the configuration in which the forced draft fan 2 is estimated based on the data of the sensor related to the operating state of the rotating machine to estimate whether or not there is a sign of failure in the rotating machine. A configuration for estimating whether or not there is a sign of a malfunction in the forced draft fan 2 based on the data of the sensors 2a to 2d provided will be described, but the configuration is not limited to this. It may be configured to estimate whether or not there is a symptom of failure in the induction draft fan 5 based on the data of the sensors 5a to 5d provided in the machine 5, or the sensor provided in the boiler feed water pump 7. It may be configured to estimate whether or not the boiler feed pump 7 has a failure sign based on the data of 7a to 7d, or to the data of the sensors 8a to 8d provided in the equipment cooling water pump 8. It may be configured to estimate whether or not the equipment cooling water pump 8 has a sign of failure based on the above.

As illustrated in FIG. 1, the information processing device 10 includes rotating machine system control units 102a to 102c that control the rotating machine systems 1a to 1c, and an information processing unit that can communicate with the rotating machine system control units 102a to 102c. And 101. At least a part of the information processing unit 101 and the rotary machine system control units 102a to 102c is realized by a computer.
The information processing unit 101 receives the process data of the rotary machine systems 1a to 1c from the rotary machine system control units 102a to 102c, and the information processing unit 101 outputs device alarm signals and failure signals to the rotary machine system control units 102a to 102c. It is output to 102c.

In the example illustrated in FIG. 1, the information processing unit 101 includes a control unit 12 and a storage unit 13, and the rotary machine system control units 102a to 102c respectively include a communication unit 11 and a device control unit 14. Have The respective units are connected to each other via a bus so that they can communicate with each other.

Of these, the communication unit 11 is a communication interface between the sensors 2a to 2d provided in the forced draft fan 2 (rotary machine) of the rotary machine systems 1a to 1c and the information processing device 10. The communication unit 11 transmits and receives information between the sensors 2a to 2d and the information processing device 10.

The storage unit 13 is, for example, a fixed data storage such as a hard disk. The storage unit 13 stores various data handled by the control unit 12. Further, the storage unit 13 stores the learned parameter 13a and the maintenance record association table 13b. The learned parameter 13a is data for realizing the classification model generated by the classification model generation unit 12a described later. In the maintenance record association table 13b, the type of classification by the classification model realized by the learned parameter 13a and the type of malfunction in the maintenance record of the forced draft fan 2 (rotating machine) are stored in association with each other.

FIG. 5 is a diagram showing an example of the maintenance record association table 13b. In the example shown in FIG. 5, of the four types of classifications C1 to C4, the state of “normal (no defect)” is associated with the classification C2, and the classification C2 is associated with the classification C2, based on the result of comparison with the maintenance record. , "Impeller deformation" is associated, category C3 is associated with the problem "adhesion of foreign matter on the impeller", and category C4 is associated with the problem "corrosion of the impeller". Is remembered.

The maintenance record association table 13b specifies, for example, in which classification in the classification model the data of the sensors 2a to 2d on the same date as (or before and after) the date when the maintenance record is created is classified. It is created by associating the identified classification with the type of failure recorded in the maintenance record. Specifically, for example, the maintenance record at time T0 records a defect called "deformation of impeller". On the other hand, as shown in FIG. 4, in the classification model, the data (white circles) of the sensors 2a to 2d at time T0 are recorded. ) Is classified into the classification C2, the malfunction of "deformation of the impeller" is associated and stored in the classification C2.

The control unit 12 is a control unit that performs various processes of the information processing device 10. As shown in FIG. 2, the control unit 12 has a classification model generation unit 12a and an estimation unit 12b. Each of these units may be realized by a processor in the information processing device 10 executing a predetermined program, or may be implemented by hardware.

The classification model generation unit 12a performs machine learning (unsupervised learning) on past data of each of the sensors 2a to 2d to generate a classification model (or a learned parameter 13a for realizing the classification model). As a generation algorithm of the classification model (or the learned parameter 13a for realizing the classification model), a hierarchical method such as a two-step method or a Ward method may be used, or a k-means method or a cohort Non-hierarchical methods such as methods may also be used.

FIG. 4 is a conceptual diagram for explaining an example of processing of the classification model generation unit 12a. In FIG. 4, each white circle indicates the data of the sensors 2a to 2d acquired at different dates and times (in the example shown in FIG. 4, only two axes are shown for the sake of easy understanding of the drawing). However, it goes without saying that the number of axes is not limited to two in the generation of the classification model according to the present embodiment). The classification model generation unit 12a mechanically classifies and classifies the past data (white circles) of the sensors 2a to 2d by performing machine learning (unsupervised learning) on the past data (white circles) of the sensors 2a to 2d. Generate a model. In the example shown in FIG. 4, the past data (white circles) of the sensors 2a to 2d are classified into four classifications C1 to C4 surrounded by broken lines. In FIG. 4, cross marks indicated by reference numerals A1 to A4 indicate central positions of the classifications C1 to C4.

The classification model generator 12a may machine-learn the outputs of the sensors 2a to 2d as past data of the sensors 2a to 2d to generate a classification model, or may perform frequency analysis from the outputs of the sensors 2a to 2d. The classification model may be generated by machine learning of the frequency, phase, and intensity data extracted by the machine learning.

According to the knowledge of the inventors of the present invention, the output itself of the sensors 2a to 2d has a data amount of about 30 megabytes per day, while the frequency, phase and The intensity data has a data amount of about 20 to 30 bytes per day. Therefore, the amount of data can be significantly reduced by using the data of the frequency, phase, and intensity that are frequency-analyzed and extracted from the outputs of the sensors 2a to 2d as the past data of the sensors 2a to 2d. As a result, the processing speed in the classification model generation unit 12a can be increased.

The estimation unit 12b has a classification model (classification model realized by the learned parameter 13a) generated by the classification model generation unit 12a, and sets new time series data of each of the sensors 2a to 2d as a classification model. input. From the classification model (classification model realized by the learned parameter 13a) generated by machine learning the past data of the sensors 2a to 2d, for each of the new time series data of the sensors 2a to 2c, The classification type is output in time series together with the information on the certainty. Specifically, for example, referring to FIG. 6, when the data of the sensors 2a to 2d at time T1 is input to the classification model, the probability that the data is classified into the classification C1 is 97% from the classification model. , The probability of being classified into the classification C2 is 0%, the probability of being classified into the classification C3 is 1%, and the probability of being classified into the classification C4 is 2%.

Then, the estimation unit 12b estimates whether or not there is a sign of failure in the forced draft fan 2 (rotating machine) based on the trend of the output of the classification model (the tendency of time-series fluctuations).

The estimation unit 12b estimates that there is a sign of failure in the forced draft fan 2 (rotating machine) when the trend of the output of the classification model is a trend in which the probability of being normally classified decreases with the passage of time. You may.

FIG. 6 is a conceptual diagram for explaining an example of the process of the estimation unit 12b. In the example shown in FIG. 6, each white circle corresponding to the data at each time T1 to T3 is gradually separated from the center position A1 of the classification C1 with the passage of time, that is, according to the classification model, the classification C1 (=“normal ( The probability of being classified as "No abnormality)") gradually decreases over time. Therefore, the estimation unit 12b estimates that the forced draft fan 2 (rotary machine) has a sign of failure based on the output trend of the classification model for the time-series data of times T1 to T3.

On the other hand, in the example shown in FIG. 6, the white circles corresponding to the data at the times T4 to T6 move away from or approach the center position A1 of the classification C1 as time passes (there is a certain tendency. That is, that is, the probability that the classification model classifies into the class C1 (=classification of “normal (no abnormality)”) decreases or increases with time (with a certain tendency). Not). Therefore, the estimation unit 12b estimates that there is no sign of failure in the forced draft fan 2 (rotary machine) based on the output trend of the classification model for the time series data from time T4 to T6.

When the estimation unit 12b estimates that the forced air blower 2 (rotary machine) has a sign of a failure, the estimation unit 12b classifies the classification model output probabilities such that the probability of classification increases with the passage of time. Then, the maintenance record linking table 13b stored in the storage unit 13 is referred to, and the type of malfunction of the forced draft fan 2 (rotating machine) is determined based on the type of malfunction linked to the classification. May be.

Specifically, for example, in the example shown in FIG. 6, each white circle corresponding to the data at each time T1 to T3 gradually approaches the central position A4 of the classification C4 as time passes, that is, depending on the classification model. The probability of being classified into the classification C4 gradually increases with the passage of time. Then, as shown in FIG. 5, in the maintenance record associating table 13b, a malfunction of "impeller corrosion" is associated and stored in the category C4. Therefore, the estimation unit 12b refers to the maintenance record association table 13b stored in the storage unit 13, and determines the type of failure based on the output trend of the classification model for the time series data of times T1 to T3. It may be presumed to be "impeller corrosion".

With reference to FIG. 1, the information processing device 10 may be located in the same location or in the same premises as the rotary machine systems 1a to 1c, may be located in another monitoring location, or may exist on a cloud server. Good.

In addition, the information processing apparatus 10 may share the configuration and functions in the same location and the same premises as the rotary machine systems 1a to 1c, the cloud server, and the monitoring location in a location different from the rotary machine systems 1a to 1c. Good.

For example, by installing the device control unit 14, the communication unit 11, and the estimation unit 12b in the same premises as the rotary machine systems 1a to 1c and at the same place and near the same, it is possible to estimate and utilize the control more quickly and with higher accuracy. Is possible.

Further, by installing the classification model generation unit 12a in a cloud server or a monitoring place and concentrating to generate a classification model (learned parameters for realizing the classification model), the specific rotating machine systems 1a to 1c have been heretofore used. The classification model for the event that has occurred in 1 can be utilized in the different rotary machine systems 1a to 1c, and many events and equipment samples can be obtained, so that the learning effect and estimation accuracy are improved.

Next, an example of an information processing method by the information processing device 10 having such a configuration will be described. FIG. 3 is a flowchart showing an example of the information processing method.

As shown in FIG. 3, first, the classification model generation unit 12a machine-learns past data of the sensors 2a to 2d to generate a classification model (learned parameter 13a for realizing the classification model) (step S11). ). In the present embodiment, the classification model generation unit 12a classifies the past data of each of the sensors 2a to 2d by machine learning of the data of the frequency, phase, and intensity extracted by the frequency analysis from the outputs of the sensors 2a to 2d. A model (learned parameter 13a for realizing the classification model) is generated. As a result, the data amount of the learning data can be compressed, as compared with the case where the outputs themselves of the sensors 2a to 2d are machine-learned, whereby the processing speed can be increased.

Of course, if the information processing apparatus 10 has a processing speed sufficient for machine learning the outputs themselves of the sensors 2a to 2d, the outputs themselves of the sensors 2a to 2d may be used as learning data without performing frequency analysis. Good. Alternatively, statistics such as an average value, a median value, a maximum value, a minimum value, and a standard deviation during a certain period (for example, one hour, one day, one week, one month, etc.) of the outputs of the sensors 2a to 2d are used as learning data. You may use.

Next, referring to FIG. 5, in the maintenance record association table 13b of the storage unit 13, the types C1 to C4 of classification by the classification model are associated with the types of defects in the maintenance record of the forced draft fan 2 (rotary machine). Are stored (step S12). These trouble phenomena occur due to the operating state changing over time.

Next, the estimation unit 12b inputs the new time-series data of the sensors 2a to 2d into the classification model (the classification model realized by the learned parameter 13a) (step S13). From the classification model, for each of the new time series data of the sensors 2a to 2c, the type of the classification is output in time series together with information on the certainty.

The estimation unit 12b estimates whether or not there is a sign of failure in the forced draft fan 2 (rotating machine) based on the output trend of the classification model (step S14). In the present embodiment, the estimation unit 12b uses the trend that the output trend of the classification model is such that the probability that the output of the classification model is classified into the classification C1 (=classification of “normal (no abnormality)”) decreases with time. It is determined whether or not there is, and if such a trend is present, it is presumed that the forced draft blower 2 (rotating machine) has a sign of failure.

When it is estimated that there is a sign of failure (step S15: YES), the estimation unit 12b refers to the maintenance record association table 13b, and based on the trend of the output of the classification model, the push blower 2 ( The type of malfunction of the rotating machine is estimated (step S16). More specifically, the estimation unit 12b identifies a classification in which the likelihood of being classified into the trend of the output of the classification model increases with the passage of time, and then the maintenance record association stored in the storage unit 13 is specified. With reference to the table 13b, the type of failure of the forced draft blower 2 (rotary machine) is determined based on the type of failure associated with the classification. After that, the control unit 12 determines whether or not the forced air blower 2 (rotary machine) should be stopped (step S17). When it is determined that the push blower 2 (rotary machine) should not be stopped (step S17: NO), the process returns to step S13 and the process is repeated for new data.

By the way, as mentioned in the section of the problem to be solved by the invention, according to the technique proposed in Patent Document 1, it becomes possible to determine the presence or absence of a sign of abnormality even if not a skilled expert. Although it will be possible to cope with the shortage, the judgment criteria and judgment method are simply replacement of the judgment criteria and judgment methods used by experts with numerical values, so there is a potential sign of potential failure that cannot be seen by human eyes. I couldn't find it.

On the other hand, according to the present embodiment, the estimation unit 12b uses the classification model generated by machine learning the past data of the sensors 2a to 2d to generate new time series of the sensors 2a to 2d. In order to estimate whether there is a sign of failure in the forced draft blower 2 (rotating machine) from the data of the above, the judgment criteria and judgment method have nothing to do with the knowledge and experience of specialists, and the latent It is possible to find signs of specific defects.

Further, according to the present embodiment, when the estimation unit 12b determines that the forced draft fan 2 (rotating machine) has a sign of a failure, the probability of being classified into the output trend of the classification model is time. The types of defects that increase with the passage of time are identified, and the types of defects of rotating machinery are estimated based on the types of defects associated with those categories, so potential warning signs that cannot be seen by the human eye are identified. Not only will you be able to find it, you will be able to estimate the type of failure. As a result, it becomes possible to promptly respond to the type of defect.

Further, according to the present embodiment, since the data of the sensors 2a to 2d are the data of the frequency, the phase and the intensity which are frequency-analyzed from the outputs of the sensors 2a to 2d, the data amount can be significantly reduced. As a result, the processing speed can be increased.

In addition, in the above-described embodiment, the forced blower 2 is described as an example of the rotary machine in which the sign of failure is estimated by the information processing device 10, but the rotary machine is limited to the blower. Instead, it may be a driven machine such as a pump, a refrigerator, a compressor, a dust remover, an agitator, a valve, a CMP (chemical mechanical polishing) apparatus, a plating apparatus, or a prime mover such as an electric motor or a steam turbine. May be

The sensor data and the process data of the entire rotary machine system used by the classification model generation unit 12a to generate the classification model may initially use all or many types of data, but model generation and learning are In some cases, the same phenomenon can be modeled with fewer types of sensor data and process data, and in such a case, faster processing becomes possible.

Although the embodiments and modifications of the present invention have been described above by way of example, the scope of the present invention is not limited to these and may be modified or modified according to the purpose within the scope of the claims. Is possible. Further, the respective embodiments and modified examples can be appropriately combined within a range in which the processing content is not inconsistent.

Further, the information processing device 10 according to the present embodiment may be configured by one or more computers, but a program for causing the one or more computers to realize the information processing device 10 and a recording medium recording the program. Is also covered by this matter.

By using this system, it becomes possible to grasp the status of the equipment in a timely manner, so it is possible to change from the conventional maintenance by threshold management and time preventive maintenance to the optimal maintenance such as status monitoring maintenance. It is possible to expect such effects as proper stocking and timely arrangement of replacement parts, optimal arrangement of maintenance personnel, and pre-planning of maintenance schedule.

1, 1a to 1c Rotating machine system 2 Push blower 3 Incinerator 4 Boiler 5 Induction fan 6 Steam turbine 7 Boiler water supply pump 8 Equipment cooling water pump 10 Information processing device 101 Information processing unit 102, 102a to 102b Rotating machine system control unit 11 communication unit 12 control unit 12a classification model generation unit 12b estimation unit 13 storage unit 13a learned parameter 13b maintenance record association table 14 device control unit

Claims (9)

It has a classification model generated by machine learning of past data of one or more sensors related to the operating state of a rotating machine, and inputs new time-series data of the sensor into the classification model to perform the classification. An information processing apparatus, comprising: an estimation unit that estimates whether or not there is a sign of a malfunction in the rotating machine based on a trend of an output of a model. The estimating unit estimates that the rotating machine has a symptom of failure when the output trend of the classification model is a trend in which the probability of being normally classified decreases with time.
The information processing apparatus according to claim 1, wherein: Further comprising a storage unit that stores the type of failure in the maintenance record of the rotating machine in association with the type of classification by the classification model,
When the estimation unit estimates that there is a sign of failure in the rotating machine, it identifies a classification in which the likelihood of being classified in the output trend of the classification model increases with the passage of time, Based on the type of failure associated with the classification, to estimate the type of failure of the rotating machine,
The information processing apparatus according to claim 1 or 2, characterized in that. The sensor measures one or more of sound, vibration, temperature, current, voltage, frequency, pressure, flow rate, rotation speed, AE (acoustic emission),
The information processing apparatus according to any one of claims 1 to 3, characterized in that: The data is frequency, phase and intensity data that is converted and extracted from the output of the sensor,
The information processing apparatus according to any one of claims 1 to 4, characterized in that: The rotating machine is any one of a pump, a blower, a refrigerator, a compressor, a dust remover, an agitator, a valve, a CMP device, a plating device, an electric motor, and a steam turbine.
The information processing apparatus according to any one of claims 1 to 5, characterized in that: 7. A classification model generation unit for machine learning the past data of one or a plurality of sensors related to an operating state of a rotating machine to generate the classification model, further comprising: The information processing device described. Computer,
It has a classification model generated by machine learning of past data of one or more sensors related to the operating state of a rotating machine, and inputs new time-series data of the sensor into the classification model to perform the classification. An information processing program, which is caused to function as an estimation unit for estimating whether or not there is a sign of a failure in the rotating machine based on a trend of an output of a model. A new time series data of the sensor is added to a classification model generated by machine learning of past data of one or a plurality of sensors provided in the rotating machine, and based on the output trend of the classification model, An information processing method, comprising estimating whether or not there is a sign of failure in the rotating machine.