JP6890733B2 - Acquired data identification device, acquired data identification method and acquired data identification program - Google Patents

Description

The present invention relates to a technique for reducing the amount of data acquired from a monitored device performing remote monitoring.

Remote monitoring is carried out using various devices such as elevators, escalators, and railways as monitored devices. By performing remote monitoring, the state of the device can be grasped remotely, and the progress of deterioration of the device can be predicted. This makes it possible to plan for maintenance work to be carried out at the appropriate time.

Patent Document 1 describes that the diagnosis interval is optimized without degrading the diagnosis accuracy by adjusting the diagnosis interval based on the reliability of the predicted life.

Japanese Unexamined Patent Publication No. 2001-101235

At the stage of installing the monitored device, in order to improve the prediction accuracy, it may be designed to acquire the data with the highest accuracy of the sensor. With this design, if the number of monitored devices increases, a large amount of data will be acquired, and the amount of communication will increase. In addition, a large amount of storage is required to store a large amount of acquired data.
With the technique described in Patent Document 1, it may be possible to reduce the load of the diagnostic process by adjusting the diagnostic interval. However, the acquisition of data necessary for diagnosis is being carried out continuously, and a large amount of data is still acquired.
An object of the present invention is to reduce the amount of data acquired from a monitored device while maintaining prediction accuracy to some extent when performing remote monitoring.

The acquired data identification device according to the present invention is
Reduce the amount of data acquired for the multiple types of data so that the total amount of data acquired from the monitored device for remote monitoring during the reference period is less than the limit amount. , An evaluation data generation unit that generates evaluation data by reducing the amount of data of the plurality of types of data acquired in the past from the monitored device according to the acquisition pattern of the target.
For each of the plurality of acquisition patterns, the prediction accuracy when the state of the monitored device is predicted by giving the evaluation data generated by the evaluation data generation unit according to the target acquisition pattern to the prediction model is calculated. Precision calculation unit and
The accuracy calculation unit includes a pattern specifying unit that specifies an acquisition pattern with high prediction accuracy calculated by the accuracy calculation unit.

In the present invention, evaluation data is generated for each of a plurality of acquisition patterns that reduce the amount of data acquired so that the total amount of data is less than the limit amount, the prediction accuracy is calculated, and the acquisition pattern with high prediction accuracy is specified. As a result, it is possible to keep the amount of data acquired from the monitored device small while maintaining the prediction accuracy to some extent.

The block diagram of the acquisition data identification apparatus 10 which concerns on Embodiment 1. FIG. The explanatory view of the prediction model 32 which concerns on Embodiment 1. FIG. The explanatory view of the weight data 33 which concerns on Embodiment 1. FIG. The flowchart which shows the operation of the acquisition data identification apparatus 10 which concerns on Embodiment 1. FIG. The explanatory view of the acquisition pattern which concerns on Embodiment 1. FIG. The explanatory view of the accuracy calculation which concerns on Embodiment 1. FIG. An explanatory diagram for specifying an acquisition pattern according to the first embodiment. The block diagram of the acquisition data identification apparatus 10 which concerns on modification 3. The block diagram of the acquisition data identification apparatus 10 which concerns on Embodiment 2. FIG. The flowchart which shows the operation of the acquisition data identification apparatus 10 which concerns on Embodiment 2.

Embodiment 1.
*** Explanation of configuration ***
The configuration of the acquired data specifying device 10 according to the first embodiment will be described with reference to FIG.
The acquired data identification device 10 is a computer.
The acquired data identification device 10 is connected to a plurality of monitored devices 40 to be remotely monitored via a communication path 50. As the monitored device 40, various devices such as an elevator, an escalator, and a train are assumed. Here, the plurality of monitored devices 40 are one type of device. For example, the plurality of monitored devices 40 are all elevators. The communication path 50 is a mobile phone network or the like.

The acquisition data identification device 10 includes hardware for a processor 11, a memory 12, a storage 13, and a communication interface 14. The processor 11 is connected to other hardware via a signal line and controls these other hardware.

The processor 11 is an IC (Integrated Circuit) that performs processing. Specific examples of the processor 11 are a CPU (Central Processing Unit), a DSP (Digital Signal Processor), and a GPU (Graphics Processing Unit).

The memory 12 is a storage device that temporarily stores data. Specific examples of the memory 12 are SRAM (Static Random Access Memory) and DRAM (Dynamic Random Access Memory).

The storage 13 is a storage device for storing data. As a specific example, the storage 13 is an HDD (Hard Disk Drive). The storage 13 includes SD (registered trademark, Secure Digital) memory card, CF (CompactFlash, registered trademark), NAND flash, flexible disk, optical disk, compact disk, Blu-ray (registered trademark) disk, DVD (Digital Versaille Disk), and the like. It may be a portable recording medium.

The communication interface 14 is an interface for communicating with an external device such as the monitored device 40. As a specific example, the communication interface 14 is a port of Ethernet (registered trademark), USB (Universal Serial Bus), HDMI (registered trademark, High-Definition Multimedia Interface).

The acquisition data identification device 10 includes a pattern generation unit 21, an evaluation data generation unit 22, an accuracy calculation unit 23, a pattern identification unit 24, and an equipment control unit 25 as functional components. The functions of each functional component of the acquired data specifying device 10 are realized by software.
The storage 13 stores a program that realizes the functions of each functional component of the acquired data specifying device 10. This program is read into the memory 12 by the processor 11 and executed by the processor 11. As a result, the functions of each functional component of the acquired data specifying device 10 are realized.

The storage 13 stores a plurality of types of data 31, a plurality of prediction models 32, and weight data 33.

The plurality of types of data 31 are data acquired from sensors provided in each monitored device 40. Each monitored device 40 is provided with a plurality of sensors, and data of a type corresponding to the sensor is acquired from each monitored device 40.
For example, when the monitored device 40 is an elevator, the plurality of types of data 31 are types of data such as the input voltage of the hoist, the acceleration (vibration) of the car, and the speed of opening and closing the door.

As shown in FIG. 2, the plurality of prediction models 32 are models constructed corresponding to each of the plurality of failure modes indicating the target for failure prediction, and are models for performing state prediction according to the failure modes. The prediction model 32 is constructed by performing machine learning or the like. The failure mode is determined for each component such as component A and component B constituting the monitored device 40. Further, the failure mode may be set for each failure content such as a door opening / closing abnormality.

As shown in FIG. 3, the weight data 33 indicates the weight assigned to each failure mode. The weight for each failure mode is set by the user of the acquired data specifying device 10 or the like according to the magnitude of the influence of the failure mode. The magnitude of the influence of the failure mode is the degree of influence on the user of the monitored device 40 when a failure occurs in the target indicated by the failure mode, and the cost incurred when the target indicated by the failure mode fails. And so on.

In FIG. 1, only one processor 11 was shown. However, the number of processors 11 may be plural, and the plurality of processors 11 may execute programs that realize each function in cooperation with each other.

*** Explanation of operation ***
The operation of the acquired data specifying device 10 according to the first embodiment will be described with reference to FIGS. 4 to 7.
The operation of the acquired data specifying device 10 according to the first embodiment corresponds to the acquired data specifying method according to the first embodiment. Further, the operation of the acquired data specifying device 10 according to the first embodiment corresponds to the processing of the acquired data specifying program according to the first embodiment.

As a premise of operation, it is assumed that the storage 13 stores a plurality of types of data 31 obtained by operating each monitored device 40 in the past. The plurality of types of data 31 stored in the storage 13 are data acquired according to the initial settings of each monitored device 40. For example, the plurality of types of data 31 stored in the storage 13 are data acquired with the sensors provided in each monitored device 40 set to the highest accuracy.

(Step S1: Pattern generation process in FIG. 4)
The pattern generation unit 21 reduces the amount of data acquired for the plurality of types of data 31 so that the total amount of data of the plurality of types of data 31 acquired from one monitored device 40 during the reference period is less than the limit amount. Generate multiple acquisition patterns.
The reference period is a period set by the user of the acquired data specifying device 10, such as 1 day, 1 week, and January. The limit amount is the amount of data set by the user of the acquired data specifying device 10 based on the communication cost or the like allowed for one monitored device 40.

As shown in FIG. 5, the pattern generation unit 21 generates an acquisition pattern for each type of data 31 by changing the acquisition presence / absence, the acquisition cycle, the time resolution, and the signal resolution.
The presence / absence of acquisition means whether or not to acquire the data 31 of the target type. That is, the presence / absence of acquisition means whether or not the acquisition of the data 31 of the target type is stopped. The acquisition cycle means an interval for acquiring the data 31 of the target type. Time resolution means the interval at which data is sampled when collecting continuous data for a certain period of time. That is, the time resolution means how often the data is sampled at the timing indicated by the acquisition cycle. For example, suppose that the acquisition cycle is 10 minutes and data is collected once every 10 minutes for 5 seconds. In this case, time resolution means sampling data, for example, every 5 milliseconds or every 10 milliseconds. Signal resolution means the resolution of a value of data. Since the data collected by the sensor is often acquired as a voltage value or a current value, the signal resolution means the resolution of the voltage value or the resolution of the current value.

Specifically, the pattern generation unit 21 generates acquisition patterns of all combinations of acquisition presence / absence, acquisition cycle, time resolution, and signal resolution. At this time, the pattern generation unit 21 changes the acquisition cycle, the time resolution, and the signal resolution at each predetermined change interval. The pattern generation unit 21 calculates the total data amount of the plurality of types of data 31 in the reference period for each of the acquisition patterns of all combinations. Based on the calculated data amount, the pattern generation unit 21 extracts an acquisition pattern in which the total data amount of the plurality of types of data 31 in the reference period is smaller than the limit amount from the acquisition patterns of all combinations. As a result, the pattern generation unit 21 generates a plurality of acquisition patterns.
When there are many acquisition patterns in which the total data amount is less than the limit amount, the pattern generation unit 21 limits the total data amount among the acquisition patterns in which the total data amount is less than the limit amount. Only some acquisition patterns close to the quantity may be extracted. As a result, when the number of extracted acquisition patterns is large, it is possible to reduce the subsequent processing amount.

(Step S2 in FIG. 4: Evaluation data generation process)
The evaluation data generation unit 22 targets each of the plurality of acquisition patterns generated in step S1 by reducing the amount of data of the plurality of types of data 31 acquired in the past from the monitored device 40 according to the target acquisition pattern. Generates evaluation data for the acquisition pattern of. That is, the evaluation data generation unit 22 reduces the amount of data of the plurality of types of data 31 stored in the storage 13 according to the target acquisition pattern, and generates evaluation data for the target acquisition pattern.
Reducing the amount of data of the plurality of types of data 31 according to the acquisition pattern of the target means that all the data 31 of the target type is deleted when there is no acquisition of the target type. Further, it means that the data 31 of the target type is thinned out based on the acquisition cycle, the time resolution, and the signal resolution.

(Step S3 in FIG. 4: Accuracy calculation process)
The accuracy calculation unit 23 gives evaluation data about the target acquisition pattern generated in step S2 according to the target acquisition pattern to each prediction model 32 for each of the plurality of acquisition patterns generated in step S1 and monitors the target. The prediction accuracy when the state of the device 40 is predicted is calculated.
A specific description will be given with reference to FIG. The accuracy calculation unit 23 gives evaluation data about the acquisition pattern of the target to the target prediction model 32 for each prediction model 32, and predicts the state of the monitored device 40. The accuracy calculation unit 23 calculates the prediction accuracy of the target acquisition pattern with respect to the target prediction model 32 by comparing the predicted result with the actual state. Here, the evaluation data is generated from a plurality of types of data 31 that have been acquired in the past and stored in the storage 13. Therefore, the actual state of the monitored device 40 has already been specified. As a method of calculating the prediction accuracy by comparing the predicted result with the actual state, the prediction accuracy is calculated from the correct answer rate of the result of the excess judgment of the maintenance standard after a certain period of time when the evaluation data is obtained. A method and a method of calculating the prediction accuracy from the average of the differences between the predicted result for a certain period and the actual state can be considered.
The accuracy calculation unit 23 weights the prediction accuracy of the target acquisition pattern for each prediction model 32 with the weight indicated by the weight data 33, and is the total prediction accuracy for the entire plurality of failure modes of the target acquisition pattern. Calculate the accuracy. At this time, the accuracy calculation unit 23 weights with the weight corresponding to the same failure mode as the target prediction model 32. For example, the accuracy calculation unit 23 calculates the total accuracy by multiplying the prediction accuracy of the target acquisition pattern for each prediction model 32 by a weight and then summing the prediction accuracy.

(Step S4 in FIG. 4: Pattern identification process)
The pattern specifying unit 24 identifies an acquisition pattern having a high overall accuracy, which is the prediction accuracy for the entire plurality of failure modes calculated in step S3. As a result, an acquisition pattern in which the total amount of data acquired in the reference period is less than the limit amount and the overall accuracy is high is specified.
For example, as shown in FIG. 7, it is assumed that there are two acquisition patterns α and acquisition patterns β in which the total amount of data is substantially the same. In this case, the pattern specifying unit 24 identifies the acquisition pattern α having a high total accuracy, which is the sum of the prediction accuracy of each failure mode.

(Step S5 in FIG. 4: Device control process)
The device control unit 25 sets the acquisition pattern specified in step S4 in each monitored device 40. For example, the device control unit 25 updates the setting information file of each monitored device 40 with the information of the acquisition pattern.
As a result, the device control unit 25 causes each monitored device 40 to collect a plurality of types of data 31 according to the acquisition pattern specified in step S4. As a result, the total amount of data of the plurality of types of data 31 acquired from each monitored device 40 in the reference period thereafter becomes less than the limit amount.

The acquired data specifying device 10 or another device acquires a plurality of types of data 31 from each monitored device 40, gives the acquired plurality of types of data 31 to each prediction model 32, and provides each monitored device 40. Predict the state of. As a result, the timing of maintenance and inspection of each monitored device 40 is determined.

*** Effect of Embodiment 1 ***
As described above, the acquisition data identification device 10 according to the first embodiment specifies an acquisition pattern in which the total data amount of the plurality of types of data 31 acquired in the reference period is less than the limit amount and the overall accuracy is high. To do. As a result, it is possible to keep the amount of data acquired from the monitored device 40 small while maintaining the prediction accuracy to some extent.

*** Other configurations ***
<Modification example 1>
In the first embodiment, the acquisition pattern having the highest prediction accuracy is specified under the condition that the total amount of data is less than the limit amount. However, the acquisition pattern with the smallest total data amount may be specified under the condition that the prediction accuracy is higher than the limit accuracy.
In this case, in step S1 of FIG. 4, the acquisition data identification device 10 identifies an acquisition pattern whose prediction accuracy is higher than the limit accuracy, and in step S3 of FIG. 4, the acquisition data identification device 10 identifies each acquisition pattern. Calculate the total amount of data. Then, in step S4 of FIG. 4, the acquisition data identification device 10 identifies the acquisition pattern having the smallest total data amount.

<Modification 2>
In the first embodiment, the acquisition pattern is generated by changing the acquisition presence / absence, the acquisition cycle, the time resolution, and the signal resolution. This is because, when edge computing is not introduced, the processing settings for the data collected on the monitored device 40 side are often the degree of acquisition / non-acquisition, acquisition cycle, time resolution, and signal resolution. Is.
However, when data processing is possible on the monitored device 40 side, it is possible to specify an appropriate acquisition pattern by generating an acquisition pattern according to the data processing possible on the monitored device 40 side. Become.

<Modification example 3>
In the first embodiment, each functional component is realized by software. However, as a modification 3, each functional component may be realized by hardware. The difference between the third modification and the first embodiment will be described.

The configuration of the acquired data specifying device 10 according to the third modification will be described with reference to FIG.
When each functional component is realized by hardware, the acquisition data identification device 10 includes an electronic circuit 15 instead of the processor 11, the memory 12, and the storage 13. The electronic circuit 15 is a dedicated circuit that realizes the functions of each functional component, the memory 12, and the storage 13.

Examples of the electronic circuit 15 include a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, a logic IC, a GA (Gate Array), an ASIC (Application Specific Integrated Circuit), and an FPGA (Field-Programmable Gate Array). is assumed.
Each functional component may be realized by one electronic circuit 15, or each functional component may be distributed and realized by a plurality of electronic circuits 15.

<Modification example 4>
As a modification 4, some functional components may be realized by hardware, and other functional components may be realized by software.

The processor 11, the memory 12, the storage 13, and the electronic circuit 15 are referred to as processing circuits. That is, the function of each functional component is realized by the processing circuit.

Embodiment 2.
The second embodiment is different from the first embodiment in that the acquisition pattern is respecified. In the second embodiment, these different points will be described, and the same points will be omitted.

*** Explanation of configuration ***
The configuration of the acquired data specifying device 10 according to the second embodiment will be described with reference to FIG.
The acquired data specifying device 10 is different from the acquired data specifying device 10 shown in FIG. 1 in that it includes a trigger determination unit 26 as a functional component.

*** Explanation of operation ***
The operation of the acquired data specifying device 10 according to the second embodiment will be described with reference to FIG.
The operation of the acquired data specifying device 10 according to the second embodiment corresponds to the acquired data specifying method according to the second embodiment. Further, the operation of the acquired data specifying device 10 according to the second embodiment corresponds to the processing of the acquired data specifying program according to the second embodiment.

As described in the first embodiment, the processes of steps S1 to S4 are executed, and the acquisition pattern is specified.
Then, in step S5, the device control unit 25 sets the acquisition pattern for each monitored device 40. At this time, the device control unit 25 sets the acquisition pattern only for the remaining monitored devices 40 excluding some of the monitored devices 40 among the plurality of monitored devices 40, and for some of the monitored devices 40. Leave the default settings. As a result, the device control unit 25 causes only the remaining monitored devices 40 excluding some of the monitored devices 40 to collect the data 31 of a plurality of types according to the acquisition pattern specified in step S4.

(Step S6 of FIG. 10: Trigger determination process)
The trigger determination unit 26 determines whether or not the acquisition pattern is in a state of being respecified. As a specific example, the trigger determination unit 26 determines that the acquisition pattern is re-specified when a certain period of time has elapsed since the acquisition pattern was last specified. Alternatively, the trigger determination unit 26 adds a monitoring target device 40, a monitoring target item is added, a prediction model 32 is added, and the algorithm of the prediction model 32 is revised. It is determined that the acquisition pattern is respecified when an event such as that is revised occurs.
When the trigger determination unit 26 is in a state of respecifying the acquisition pattern, the trigger determination unit 26 returns the process to step S1. On the other hand, if the acquisition pattern is not in the state of being respecified, the trigger determination unit 26 determines whether or not the acquisition pattern is in the state of being respecified again after a certain period of time has elapsed.

When the process is returned to step S1, the processes of steps S1 to S4 are executed again, and the acquisition pattern is specified. At this time, in step S2, the evaluation data generation unit 22 generates evaluation data using the plurality of types of data 31 acquired from some of the monitored devices 40 for which the acquisition pattern was not set in step S5. This is because the plurality of types of data 31 acquired from the remaining monitored devices 40 are collected according to the acquisition pattern, and therefore appropriate evaluation data cannot be generated.
Then, in step S5, the device control unit 25 sets the acquisition pattern for each monitored device 40. At this time, the device control unit 25 sets the acquisition pattern only for the remaining monitored devices 40 excluding some of the monitored devices 40 among the plurality of monitored devices 40, and for some of the monitored devices 40. Leave the default settings.

*** Effect of Embodiment 2 ***
As described above, the acquisition data identification device 10 according to the second embodiment reidentifies the acquisition pattern. The prediction accuracy may change due to changes in the situation such as the accumulation status of the data 31, but by respecifying the acquisition pattern, it is possible to maintain a state in which the prediction accuracy is high.

*** Other configurations ***
<Modification 5>
In the failure prediction mechanism, the prediction model 32 may be replaced. If the algorithm of the prediction model 32 changes, the required data will change. When the algorithm of the prediction model 32 is evaluated by itself, even if it is superior to the previous algorithm, when the algorithm of the prediction model 32 is actually introduced, a large amount of data is required, and as a result, it may not be effective.
Therefore, in step S6 of FIG. 10, when the acquisition pattern is respecified triggered by the revision of the algorithm of the prediction model 32, the prediction model 32 before the revision and the prediction model after the revision Evaluation may be performed for both 32 and 32. Specifically, in steps S3 to S4 of FIG. 10, the acquired data specifying device 10 performs processing on both the prediction model 32 before the revision and the prediction model 32 after the revision. That is, in step S3 of FIG. 10, the accuracy calculation unit 23 calculates the prediction accuracy of both the prediction model 32 before the revision and the prediction model 32 after the revision, and calculates the total accuracy. In step S4 of FIG. 10, the pattern specifying unit 24 adopts the prediction model 32 before the revision and the prediction model 32 after the revision, whichever has the highest overall accuracy. Then, the pattern specifying unit 24 specifies an acquisition pattern with high overall accuracy for the adopted person.
As a result, it is possible to determine the effectiveness of the revised prediction model 32 after considering the required amount of data without actually introducing the revised prediction model 32.

<Modification 6>
Based on the same concept as in the modified example 5, it is possible to prepare a plurality of prediction models 32 and specify the most effective prediction model 32 in consideration of the amount of data. Specifically, in steps S3 to S4 of FIG. 10, the acquired data specifying device 10 performs processing for each of the plurality of prediction models 32. That is, in step S3 of FIG. 10, the accuracy calculation unit 23 calculates the prediction accuracy for each of the plurality of prediction models 32, and calculates the total accuracy. In step S4 of FIG. 10, the pattern specifying unit 24 adopts the prediction model 32 having the highest overall accuracy among the plurality of prediction models 32. Then, the pattern specifying unit 24 specifies an acquisition pattern with high overall accuracy for the adopted prediction model 32.

10 Acquisition data identification device, 11 Processor, 12 Memory, 13 Storage, 14 Communication interface, 15 Electronic circuit, 21 Pattern generation unit, 22 Evaluation data generation unit, 23 Accuracy calculation unit, 24 Pattern identification unit, 25 Equipment control unit, 26 Trigger judgment unit, 31 data, 32 prediction model, 33 weight data, 40 monitored equipment, 50 communication paths.

Claims (10)

Reduce the amount of data acquired for the multiple types of data so that the total amount of data acquired from the monitored device for remote monitoring during the reference period is less than the limit amount. , An evaluation data generation unit that generates evaluation data by reducing the amount of data of the plurality of types of data acquired in the past from the monitored device according to the acquisition pattern of the target.
For each of the plurality of acquisition patterns, the prediction accuracy when the state of the monitored device is predicted by giving the evaluation data generated by the evaluation data generation unit according to the target acquisition pattern to the prediction model is calculated. Precision calculation unit and
An acquisition data identification device including a pattern identification unit that identifies an acquisition pattern with high prediction accuracy calculated by the accuracy calculation unit. The accuracy calculation unit applies the evaluation data to the corresponding prediction model for each of the plurality of failure modes, calculates the prediction accuracy, and uses the prediction accuracy calculated for each of the plurality of failure modes to obtain the plurality of failure modes. Calculate the total accuracy, which is the prediction accuracy for the whole,
The acquisition data identification device according to claim 1, wherein the pattern identification unit identifies an acquisition pattern having high overall accuracy. The acquired data identification according to claim 2, wherein the accuracy calculation unit weights the predicted accuracy calculated for each of the plurality of failure modes by the weights assigned to each of the plurality of failure modes to calculate the total accuracy. apparatus. The acquired data identification device further
Claims 1 to 3 include a device control unit that sets the acquisition pattern specified by the pattern identification unit in the monitoring target device and causes the monitoring target device to collect the plurality of types of data according to the acquisition pattern. The acquired data identification device according to any one of the items. The evaluation data generation unit regenerates the evaluation data periodically or every time an event occurs.
When the evaluation data is regenerated, the accuracy calculation unit recalculates the prediction accuracy.
The acquisition data specifying device according to any one of claims 1 to 4, wherein the pattern specifying unit identifies a recalculated acquisition pattern having high prediction accuracy. There are a plurality of the monitored devices,
The acquired data identification device further
The acquisition pattern specified by the pattern specifying unit is set for the remaining monitoring target devices excluding some of the monitoring target devices among the plurality of monitoring target devices, and the remaining monitoring target devices are described according to the acquisition pattern. Equipped with a device control unit that collects multiple types of data
The acquisition data specifying device according to claim 5, wherein the evaluation data generation unit regenerates the evaluation data from the plurality of types of data collected from some of the monitored devices. The evaluation data generation unit regenerates the evaluation data when the prediction model is revised.
When the evaluation data is regenerated, the accuracy calculation unit calculates the prediction accuracy of the prediction model before the revision and the prediction model after the revision.
The pattern specifying unit adopts the prediction model before the revision and the prediction model after the revision, whichever has the highest prediction accuracy, and identifies the acquisition pattern having the highest prediction accuracy for the adopted prediction model. The acquired data specifying device according to any one of claims 1 to 6. The accuracy calculation unit calculates the prediction accuracy by giving the evaluation data to the target prediction model for each of the plurality of prediction models.
The pattern specifying unit adopts a prediction model having the highest prediction accuracy among the plurality of prediction models, and any of claims 1 to 6 for specifying the acquisition pattern having the highest prediction accuracy for the adopted prediction model. The acquired data identification device according to item 1. The evaluation data generation unit reduces the amount of data acquired for the multiple types of data so that the total amount of data of the multiple types of data acquired from the monitored device for remote monitoring during the reference period is less than the limit amount. For each acquisition pattern, the evaluation data is generated by reducing the amount of data of the plurality of types of data acquired in the past from the monitored device according to the target acquisition pattern.
The accuracy calculation unit calculates the prediction accuracy when the evaluation data is given to the prediction model according to the target acquisition pattern and the state of the monitored device is predicted for each of the plurality of acquisition patterns.
An acquisition data identification method in which the pattern identification unit identifies an acquisition pattern having high prediction accuracy. Reduce the amount of data acquired for the multiple types of data so that the total amount of data acquired from the monitored device for remote monitoring during the reference period is less than the limit amount. , Evaluation data generation processing that generates evaluation data by reducing the amount of data of the plurality of types of data acquired in the past from the monitored device according to the acquisition pattern of the target.
For each of the plurality of acquisition patterns, the prediction accuracy when the state of the monitored device is predicted by giving the evaluation data generated by the evaluation data generation process according to the target acquisition pattern to the prediction model is calculated. Precision calculation processing and
An acquisition data identification program that causes a computer to function as an acquisition data identification device that executes a pattern identification process for identifying an acquisition pattern with high prediction accuracy calculated by the accuracy calculation process.

Images