JP6700509B1 - Estimating apparatus, estimating method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an estimating device capable of accurately estimating a time when data to be monitored becomes a desired state. In an estimation device 10, a prediction interval data acquisition unit 120 uses time-series monitoring data stored in a monitoring data storage unit 110 to estimate a time transition of an n% prediction interval of a monitoring target 20. (Hereinafter, referred to as prediction interval data) is generated. The estimated time output unit 130 uses the prediction interval data and the reference value set for the monitoring target 20 to generate data indicating the time at which the reference value is estimated to enter the n% prediction interval, and outputs this. Output the data. [Selection diagram] Figure 1

Description

  The present invention relates to an estimation device, an estimation method, and a program.

  In order to monitor the monitoring target with high accuracy, it is important to analyze the data of the monitoring target with high accuracy. For example, Patent Document 1 discloses a technique relating to a maintenance index of a device that controls the flow rate of a fluid. Specifically, this control device keeps the flow rate of the fluid constant by controlling the opening degree of the valve. Then, using the current opening and an approximate expression that approximates the relationship between the valve opening and the fluid flow rate, and the current opening, the flow rate is calculated assuming that the valve opening is the upper limit. This flow rate is used as a maintenance judgment index.

JP, 2008-136842, A

  In order to monitor the monitoring target with high accuracy, it is important to accurately estimate the time when the data of the monitoring target is in a desired state. However, since the change status of the monitoring target varies, if the above-described estimation is performed using the approximate expression, the accuracy of the estimation result may not satisfy the required level.

  An object of the present invention is to accurately estimate the time when the data to be monitored is in a desired state.

According to the present invention, prediction section data, which is data generated using monitoring data including a monitoring result of a monitoring target and a monitoring time thereof, and which shows an estimation result of time transition of the n% prediction section of the monitoring target. A prediction interval data acquisition unit to be acquired,
Estimated time output for generating and outputting output data indicating a time when the reference value is estimated to enter the n% prediction interval using the prediction interval data and the reference value set for the monitoring target. Means and
An estimation device comprising:

According to the invention, the computer
Prediction section data that is data generated using monitoring data including the monitoring result of the monitoring target and the monitoring time, and that shows the estimation result of the time transition of the n% prediction section of the monitoring target,
An estimation method that uses the prediction interval data and a reference value set for the monitoring target to generate and output output data indicating a time when the reference value is estimated to enter the n% prediction interval. Will be provided.

According to the invention,
A function of acquiring prediction section data, which is data generated using monitoring data including a monitoring result of a monitoring target and a monitoring time thereof, and which shows an estimation result of time transition of the n% prediction section of the monitoring target.
A function of using the prediction interval data and a reference value set for the monitoring target to generate and output output data indicating a time when the reference value is estimated to enter the n% prediction interval;
A program that has

  According to the present invention, it is possible to accurately estimate the time when the data to be monitored is in a desired state.

It is a figure which shows the function structure of the estimation apparatus which concerns on 1st Embodiment with the usage environment of an estimation apparatus. It is a figure which shows an example of the hardware constitutions of an estimation apparatus. It is a flow chart which shows an example of processing which an estimating device performs. It is a graph for explaining the data output by the estimation time output unit. It is a figure for demonstrating the use environment of the estimation apparatus which concerns on 2nd Embodiment. It is a figure which shows the modification of FIG.

  Embodiments of the present invention will be described below with reference to the drawings. In all the drawings, the same constituents will be referred to with the same numerals, and the description thereof will not be repeated.

(First embodiment)
FIG. 1 is a diagram showing a functional configuration of an estimation device 10 according to the first embodiment together with a usage environment of the estimation device 10. The estimation device 10 is a device that estimates when the data of the monitoring target 20 is in a desired state. The estimation device 10 acquires monitoring data from the sensor 22 that monitors the monitoring target 20. The monitoring data includes the monitoring result of the monitoring target and the monitoring time.

  When there are a plurality of monitoring targets 20, the sensor 22 is provided for each of the plurality of monitoring targets 20. The monitoring data includes sensor identification information that identifies the sensor 22.

  As an example, the monitoring target 20 is a railroad track of a railroad or a rail joint. In this case, the sensor 22 is a sensor that measures a rail gauge or a sensor that measures a rail joint, and is mounted on, for example, a vehicle traveling on the rail. The sensor 22 stores the monitoring result, information specifying the measurement position (hereinafter, described as measurement position information), and information indicating the measurement date and time (hereinafter, described as measurement date and time information) in the storage device as monitoring data. The storage device is mounted in, for example, a vehicle. Then, the estimation device 10 acquires the data stored in this storage device. There are various acquisition methods. For example, the estimation device 10 may acquire the data from the storage device in a batch method, or may acquire the data from the storage device via the communication network in real time. Also, the monitoring data may be acquired from another device. The monitoring data may include at least one of past track displacement, facility information, number of tonnage, radius of curvature, weather, and temperature.

  The storage device described above may be a part of the estimation device 10 (that is, the monitoring data storage unit 110 described below). In this case, the sensor 22 transmits the monitoring data to the estimation device 10 via the communication device mounted on the vehicle and the communication network.

  Then, the estimation device 10 estimates the time when the rail gauge or the rail joint does not satisfy the criterion with a probability of a certain probability or more.

  As another example, the monitoring target 20 is the inventory of products in the vending machine or the number of sold products. In this case, the sensor 22 is built in the vending machine, and the monitoring result, that is, the number of products in stock or sold, information identifying the vending machine (hereinafter referred to as vending machine identification information), and measurement. The date and time information is stored in the storage device as monitoring data. Then, the estimation device 10 acquires the data stored in this storage device. There are various acquisition methods and may be batch or real-time, similar to the rail example. The monitoring result in this example may be the remaining number of products or the presence/absence of products. Then, the estimation device 10 estimates the time when the stock of the product falls below the standard with a probability of a certain probability or more, for example, the time when it becomes zero. The estimation device 10 may estimate the time when the number of sold products exceeds the standard.

  As another example, the monitoring target 20 is the remaining amount of cash in an ATM (Automatic Teller Machine). In this case, the sensor 22 is built in the ATM, and the storage device stores the monitoring result, that is, the remaining amount of cash for each currency, information identifying the ATM (hereinafter referred to as ATM identification information), and measurement date/time information as monitoring data. Remember. Then, the estimation device 10 acquires the data stored in this storage device. There are various acquisition methods, but it is preferable that the acquisition is in real time. Then, the estimation device 10 estimates, for each currency, the time when it is estimated that the reference value set for the remaining amount or payout amount of the currency falls within the n% prediction section. Note that the monitoring data may include the number of customers by hour, the surrounding environment, the day of the week, and the sequence of weekends and holidays.

  As another example, the monitoring target 20 may be at least one of the product balance of individual items and the number of products consumed (the number of products sold) in a store or electronic commerce. In this case, the monitoring data may be the same as the data in the vending machine. For example, the monitoring data preferably includes the past consumption number and the surrounding environment (at least one of weather, temperature, and location). Further, the monitoring data may include the presence/absence of peripheral events, shelving allocation, and product prices. Then, the estimation device 10 estimates the time when the reference value set for the product balance or the number of consumed products is estimated to be within the n% prediction section.

  As another example, the monitored object 20 may be a manufacturing machine. In this case, the monitoring data includes, for example, past failure definition variables, usage time, the number of passing products, temperature, and humidity. Then, the estimation device 10 estimates the time when it is estimated that the reference value set for the failure definition variable falls within the n% prediction interval. The fault definition variable is an allowable value set at the time of design.

  As another example, the monitoring target 20 may be a charge amount of a game used on a computer (for example, a portable terminal). In this case, the monitoring data is the past charge amount, the number of users, and the in-game action. Then, the estimation device 10 performs the estimation process of estimating the charge amount in the next measure, that is, the estimation process of the width of the estimation period of the charge amount.

  As another example, the monitoring target 20 may be a human health checkup result. In this case, the monitoring data is the past test result, age, sex, height, weight, sleep length, diet status, and presence/absence of exercise. Then, the estimation device 10 estimates the time when the reference value determined for a certain inspection item is estimated to fall within the n% prediction interval.

  As another example, the monitoring target 20 may be the required number of employees or the number of visitors in a restaurant. In this case, the monitoring data is the number of past visitors, the number of seats, business hours, surrounding environment, weather, and temperature. Then, the estimation device 10 estimates the time when it is estimated that the reference value related to the required number of employees or the number of visitors is within the n% prediction section.

  The estimation device 10 includes a monitoring data storage unit 110, a prediction section data acquisition unit 120, and an estimation time output unit 130.

  The monitoring data storage unit 110 stores the monitoring data generated by the sensor 22. Since the sensor 22 regularly measures and generates and transmits monitoring data, the monitoring data is, for example, time-series data. When the monitoring data includes sensor identification information, the sensor 22 may store information that specifies the monitoring target 20 corresponding to the sensor 22 in association with the monitoring data.

  The monitoring data storage unit 110 associates the monitoring data with data (hereinafter, referred to as environmental data) that the sensor 22 specifies the environment of the monitoring target 20. The monitoring data storage unit 110 may store the environmental data in association with the monitoring data, or may store the data specifying the storage location of the environmental data in association with the monitoring data. The environmental data includes, for example, meteorological data such as temperature and weather at the measurement location. When there are a plurality of sensors 22, the environmental data corresponding to each monitoring data is specified using, for example, sensor identification information.

  When the monitoring target 20 is a rail track or a rail joint, the environmental data includes accumulated tonnage at the measurement location and equipment (such as railroad crossings and stations) located within a reference distance (for example, 10 m) from the measurement location. It may include at least one of the data indicating the type. Further, the environment data may include data indicating the attribute of the rail. An example of this attribute is at least one of rail type and radius of curvature.

  Further, when the monitoring target 20 is the inventory or the number of products sold in the vending machine or the store, the environmental data indicates that the facility is located within a reference distance (for example, 100 m) from the vending machine (including other vending machines). May include data indicating the type of.

  The prediction interval data acquisition unit 120 uses the time-series monitoring data stored in the monitoring data storage unit 110 to show data indicating the estimation result of the time transition of the n% prediction interval of the monitoring target 20 (hereinafter, prediction interval data). Is described).

  At this time, the prediction section data acquisition unit 120 preferably uses a Gaussian process. In this case, the prediction section data acquisition unit 120 performs Gaussian process regression and generates prediction section data using this regression result. In the regression results, the horizontal axis represents time and the vertical axis represents the probability that the target event will occur. Then, the prediction section data acquisition unit 120 generates prediction section data using this regression result.

  When using the Gaussian process regression, the input data for generating the prediction interval data may include the environmental data stored in the monitoring data storage unit 110, in addition to the monitoring data described above.

  The value of n is input to the estimation device 10 by the user of the estimation device 10, for example, but may be set by default. When set by default, the value of n is, for example, 95(%), but is not limited to this. When there are multiple types of monitoring targets 20, the value of n may be different depending on the types of monitoring targets 20. For example, when one vending machine sells a plurality of types of products, the value of n may differ depending on the types of products. Increasing n means reducing the possibility that the monitoring target will be defective (for example, the possibility that the product will be out of stock in the vending machine). The frequency of replenishment) also increases. Whether or not defects are allowed is often determined by the type of monitoring target (for example, type of product). Therefore, if the value of n is changed depending on the type of the monitoring target, it is possible to prevent the frequency of maintenance from unnecessarily increasing.

  The estimation time output unit 130 uses the prediction interval data and the reference value set for the monitoring target 20 to indicate the time at which the reference value is estimated to enter the n% prediction interval (hereinafter referred to as output data). Description) is generated and this output data is output. The output data may include, for example, graph format data, but may be character string data indicating a time. The output destination of the output data is, for example, a display, but may be a printer, a stationary terminal, or a mobile terminal.

  FIG. 2 is a diagram illustrating an example of a hardware configuration of the estimation device 10. The main configuration of the estimation device 10 is realized by using an integrated circuit. This integrated circuit has a bus 402, a processor 404, a memory 406, a storage device 408, an input/output interface 410, and a network interface 412. The bus 402 is a data transmission path for the processor 404, the memory 406, the storage device 408, the input/output interface 410, and the network interface 412 to exchange data with each other. However, the method of connecting the processors 404 and the like to each other is not limited to bus connection. The processor 404 is an arithmetic processing unit realized by using a microprocessor or the like. The memory 406 is a memory realized by using a RAM (Random Access Memory) or the like. The storage device 408 is a storage device realized by using a ROM (Read Only Memory), a flash memory, or the like.

  The input/output interface 410 is an interface for connecting the estimation apparatus 10 to peripheral devices.

  The network interface 412 is an interface for connecting the estimation device 10 to an external device such as the sensor 22 via a communication network. The method for connecting the network interface 412 to the communication network may be wireless connection or wired connection.

  The storage device 408 stores a program module for realizing each functional element of the estimation device 10. The processor 404 realizes each function of the estimation device 10 by reading this program module into the memory 406 and executing it. The storage device 408 also functions as the monitoring data storage unit 110.

  FIG. 3 is a flowchart showing an example of processing performed by the estimation device 10. First, the user of the estimation device 10 inputs information (hereinafter, referred to as monitoring target identification information) that identifies the monitoring target 20 to be processed into the estimation device 10. In addition, the user of the estimation device 10 inputs the value of n in the n% prediction section and the reference value used by the estimation time output unit 130 (step S10). Note that the value of n and the reference value are set by default, and when the estimation device 10 uses this default value, the input of n and the reference value is omitted.

  Next, the prediction section data acquisition unit 120 of the estimation device 10 selects the monitoring data corresponding to the monitoring target specifying information input by the user (step S20).

  Here, the prediction section data acquisition unit 120 may also select the monitoring data of the monitoring target 20 whose environment is similar to the monitoring target 20 that is the processing target. Specifically, the prediction interval data acquisition unit 120 acquires the value of at least one item of the environmental data corresponding to the monitoring target specifying information input by the user regarding the monitoring target 20 that is the processing target. Then, the prediction interval data acquisition unit 120 defines the reference range for each item using the acquired item value. For example, the prediction interval data acquisition unit 120 provides a predetermined width (the upper width and the lower width may be the same or different) with respect to the value of the acquired environmental data as a reference. To set the reference range. Then, the monitoring target 20 (second monitoring target 20) whose item values satisfy the reference range (that is, the similarity of the environmental data satisfies the standard) is selected, and the monitoring data corresponding to the selected second monitoring target 20 is selected. select.

  For example, when the monitored object 20 is a rail gauge or a rail joint gap, the item used when selecting the second monitored object 20 preferably includes the cumulative passing tonnage.

  Next, the prediction section data acquisition unit 120 generates prediction section data (step S30). The estimation time output unit 130 generates output data using the prediction interval data generated in step S30 and the reference value input in step S10 (step S40), and outputs the generated output data (step S50). .

  FIG. 4 is a graph for explaining the data output by the estimated time output unit 130. In the example shown in the figure, the output data is the displacement of the monitored rail, that is, the time transition of the measured value of the displacement of the rail width, the predicted value of the displacement of the rail width, and the 95% prediction interval after the current date and time. Shows how it changes with time (that is, prediction interval data). Then, the estimated time output unit 130 outputs the date and time when the reference value enters the 95% prediction section. For example, when the reference value is -4.0 mm, the date and time output by the estimated time output unit 130 is around December 25, 2017.

  As described above, according to the present embodiment, the estimation device 10 generates prediction section data indicating the estimation result of the time transition of the n% prediction section of the monitoring target 20. Then, output data indicating the time when the reference value set for the monitoring target 20 is estimated to enter the n% prediction section, for example, the time when the monitoring target 20 should be maintained is generated. Therefore, it is possible to accurately estimate the time when the data of the monitoring target 20 is in a desired state.

(Second embodiment)
FIG. 5: is a figure for demonstrating the use environment of the estimation apparatus 10 which concerns on 2nd Embodiment. In this embodiment, the estimation device 10 does not have the monitoring data storage unit 110, and the prediction interval data acquisition unit 120 of the estimation device 10 does not generate prediction interval data. Instead, the data processing device 30 has a monitoring data storage unit 310 and a prediction section data generation unit 320. The data processing device 30 is, for example, a cloud type device. In this case, the estimation device 10 functions as a terminal. However, the data processing device 30 is not limited to this example.

  The monitoring data storage unit 310 has the same configuration as the monitoring data storage unit 110 in the first embodiment, and the prediction section data generation unit 320 has the prediction section data generation function of the prediction section data acquisition unit 120 in the first embodiment. have. Then, the prediction section data acquisition unit 120 acquires the prediction section data from the prediction section data generation unit 320 of the data processing device 30.

  Note that, as illustrated in FIG. 6, the estimation device 10 may include a prediction section data storage unit 140 that stores the prediction section data generated by the prediction section data generation unit 320. In this case, the prediction section data storage unit 140 preferably stores the prediction section data corresponding to the monitoring target specifying information in association with the monitoring target specifying information. Then, the prediction section data acquisition unit 120 reads the prediction section data corresponding to the monitoring target specifying information input by the user from the prediction section data storage unit 140.

  According to this embodiment, the same effect as that of the first embodiment can be obtained.

  Although the embodiments of the present invention have been described above with reference to the drawings, these are merely examples of the present invention, and various configurations other than the above may be adopted.

10 Estimating Device 20 Monitoring Target 22 Sensor 30 Data Processing Device 110 Monitoring Data Storage Unit 120 Prediction Section Data Acquisition Unit 130 Estimating Timing Output Unit 140 Prediction Section Data Storage Unit 310 Monitoring Data Storage Unit 320 Prediction Section Data Generation Unit

Claims (11)

Prediction interval data acquisition that acquires prediction interval data that is data generated using monitoring data including the monitoring result of the monitoring target and the monitoring time, and that shows the estimation result of the time transition of the n% prediction interval of the monitoring target Department,
Estimated time output for generating and outputting output data indicating a time when the reference value is estimated to enter the n% prediction interval using the prediction interval data and the reference value set for the monitoring target. Means and
Estimating device comprising. The estimation device according to claim 1,
The said prediction area data acquisition part is an estimation apparatus which produces|generates the said prediction area data using the said monitoring data. The estimation device according to claim 2,
The monitoring data is generated for a plurality of the monitoring targets,
The prediction section data acquisition unit, when generating the prediction section data on the first monitoring target, the monitoring data for the first monitoring target and the degree of similarity to the first monitoring target satisfy a criterion. A second estimating device using the monitoring data for the second monitoring target. The estimation device according to any one of claims 2 and 3,
The prediction interval data acquisition unit is an estimation device that generates the prediction interval data using a Gaussian process. The estimation device according to any one of claims 1 to 4,
The monitoring device is an estimation device that is a rail gauge or a gap between rail joints. The estimation apparatus according to claim 5,
The monitoring data is generated for the different monitoring targets of the same rail,
The environment of the first monitoring target and the environment of the second monitoring target satisfy the criteria of similarity,
The prediction interval data acquisition unit uses the monitoring data for the first monitoring target and the monitoring data for the second monitoring target when generating the prediction interval data for the first monitoring target. Estimator. The estimation device according to any one of claims 1 to 4,
The monitoring target is an estimation device that is a product inventory in a vending machine. The estimation device according to claim 7,
The vending machine sells multiple types of products,
An estimation device in which n is different depending on the type of the product. The estimation device according to any one of claims 1 to 4,
The monitoring target is an estimation device that is the remaining amount of cash in an ATM (Automatic Teller Machine). Computer
Prediction section data which is data generated using monitoring data including a monitoring result of a monitoring target and its monitoring time and which shows an estimation result of time transition of the n% prediction section of the monitoring target is acquired,
An estimation method that uses the prediction interval data and a reference value set for the monitoring target to generate and output output data indicating a time at which the reference value is estimated to enter the n% prediction interval. .. On the computer,
A function of acquiring prediction section data, which is data generated using monitoring data including a monitoring result of a monitoring target and its monitoring time, and which shows an estimation result of time transition of the n% prediction section of the monitoring target;
A function of using the prediction interval data and a reference value set for the monitoring target to generate and output output data indicating a time when the reference value is estimated to enter the n% prediction interval;
A program to have.

Images