JP2020154849A - Controller, system, method, and program - Google Patents

Abstract

To provide a controller, a system, a method, and a program capable of obtaining a feature quantity based on which a state of an object can be highly precisely determined.SOLUTION: A controller 10 includes an acquisition unit 11 that acquires time-sequential numerical values representing states of an object 100 measured by a sensor 20, an extraction unit 12 that extracts a candidate for an extreme value from among the time-sequential numerical values, and a feature quantity extraction unit 13 that extracts, as a feature quantity, a time elapsing from a reference time until the candidate for the extreme value arises.SELECTED DRAWING: Figure 1

Description

The present invention relates to controllers, systems, methods and programs.

Conventionally, a controller has been used that calculates statistics such as the mean value and standard deviation of time-series numerical values measured by a sensor and determines the state of the target based on those values.

For example, in Patent Document 1 below, a feature amount is generated from data acquired from a controlled object, and the degree of deviation between the feature amount and an aggregate of feature amounts stored by a storage means is obtained. A control system that detects an abnormality that occurs in a controlled object based on a threshold value is described.

Further, Patent Document 2 below describes a prediction system in which a first calculation model and a second calculation model are applied to a measurement value to be predicted, and a plurality of obtained data are combined to obtain a prediction value. There is.

Further, in Patent Document 3 below, the characteristics of the transition of the values of the time-series data included in the time-series data group with the passage of time are calculated, and the similarity between the transition pattern corresponding to a specific label and the calculated transition characteristics. Is calculated, a specific label is recorded as the first feature information in association with the time series data group based on the similarity, and a plurality of first feature information is extracted to form a plurality of first feature information. A data processing system that generates the second feature information based on it is described.

Japanese Unexamined Patent Publication No. 2018-159981 JP-A-2018-163515 Patent No. 5678620

As described in Patent Documents 1 to 3, various attempts have been made conventionally to extract various feature quantities from the numerical values measured by the sensor and determine the state of the object based on the feature quantities. However, with the conventional feature amount, the state of the target may not be accurately determined.

Therefore, the present invention provides a controller, a system, a method, and a program capable of obtaining a feature amount capable of determining a target state with higher accuracy.

The controller according to one aspect of the present disclosure includes an acquisition unit that acquires a time-series numerical value indicating a target state measured by a sensor, an extraction unit that extracts extreme value candidates from the time-series numerical values, and a reference time point. It is provided with a feature amount extraction unit that extracts the time from to the generation of extreme value candidates as a feature amount.

According to this aspect, by extracting the time from the reference time point until the candidate for the extreme value is generated as the feature amount, it is possible to obtain the feature amount capable of determining the target state more accurately.

In the above aspect, the extraction unit may extract the central numerical value among the three numerical values as an extreme value candidate when three consecutive numerical values among the time series numerical values satisfy a predetermined magnitude relationship.

According to this aspect, extreme value candidates can be extracted by a process having a relatively light load, and feature quantities can be extracted at high speed.

In the above aspect, the state of the target is determined based on the selection unit that accepts the selection of one or more feature quantities from the plurality of types of feature quantities extracted by the feature quantity extraction unit and the one or more feature quantities. A generation unit that generates a determination model may be further provided.

According to this aspect, it is possible to select one or a plurality of feature quantities that are effective for determining the state of the target and generate a determination model capable of determining the state of the target with higher accuracy.

In the above aspect, an output unit may be further provided which outputs the relationship between the plurality of types of feature quantities and the state of the target in a mode in which the plurality of types of feature quantities can be compared.

According to this aspect, when selecting one or a plurality of feature quantities that are effective for determining the target state, the relationship between the plurality of types of feature quantities and the target state is compared, and a more appropriate feature quantity is compared. Will be able to be selected more efficiently.

In the above aspect, the reference time point may be the start time point of the cycle for controlling the object.

According to this aspect, it is possible to obtain a feature amount capable of more accurately determining the state of the target for each cycle of controlling the target.

A system according to another aspect of the present disclosure is a system including a sensor for measuring a time-series numerical value indicating a target state and a controller connected to the sensor, and the controller acquires a time-series numerical value. It has an acquisition unit, an extraction unit that extracts extreme value candidates from time-series numerical values, and a feature amount extraction unit that extracts the time from the reference time to the occurrence of extreme value candidates as a feature amount.

According to this aspect, by extracting the time from the reference time point until the candidate for the extreme value is generated as the feature amount, it is possible to obtain the feature amount capable of determining the target state more accurately.

A method according to another aspect of the present disclosure is to obtain a time-series numerical value indicating the state of an object measured by a sensor in a controller, and to extract an extreme value candidate from the time-series numerical value. The time from the reference time to the occurrence of extreme value candidates is extracted as a feature amount, and is executed.

According to this aspect, by extracting the time from the reference time point until the candidate for the extreme value is generated as the feature amount, it is possible to obtain the feature amount capable of determining the target state more accurately.

A program according to another aspect of the present disclosure allows the controller to acquire a time-series numerical value indicating the state of an object measured by a sensor, and to extract an extreme value candidate from the time-series numerical value. The time from the reference time to the occurrence of extreme value candidates is extracted as a feature amount, and is executed.

According to this aspect, by extracting the time from the reference time point until the candidate for the extreme value is generated as the feature amount, it is possible to obtain the feature amount capable of determining the target state more accurately.

According to the present invention, it is possible to provide a controller, a system, a method and a program capable of obtaining a feature amount capable of determining a target state with higher accuracy.

It is a figure which shows the outline of the system which concerns on embodiment of this invention. It is a figure which shows the functional block of the controller which concerns on this embodiment. It is a figure which shows the numerical value of the time series acquired by the controller which concerns on this embodiment. It is a figure which shows the plurality of types of feature quantities extracted by the controller which concerns on this embodiment. It is a figure which shows the state of a plurality of kinds of features and a target output by the controller which concerns on this embodiment. It is a flowchart of the judgment model generation process executed by the controller which concerns on this embodiment. It is a flowchart of the feature amount extraction process executed by the controller which concerns on this embodiment. It is a flowchart of the determination process executed by the controller which concerns on this embodiment.

Embodiments of the present invention will be described with reference to the accompanying drawings. In each figure, those having the same reference numerals have the same or similar configurations.

FIG. 1 is a diagram showing an outline of a system 1 according to an embodiment of the present invention. The system 1 according to the present embodiment includes a controller 10, a sensor 20, a computer 30, and a transfer device 50. The controller 10 is connected to the sensor 20, the computer 30, and the transport device 50, acquires a time-series numerical value indicating the state of the target 100 measured by the sensor 20, and determines the state of the target 100 based on the numerical value. judge. The determination result by the controller 10 is output to the computer 30 or used for controlling the transport device 50.

The controller 10 according to the present embodiment extracts a candidate for an extreme value from a time-series numerical value indicating the state of the target 100, and extracts as a feature amount the time from the reference time until the candidate for the extreme value is generated. A feature amount capable of determining the state of 100 more accurately is obtained.

The sensor 20 is a sensor that measures a predetermined physical quantity, for example, an optical sensor that measures the presence or absence of the target 100 based on the amount of received light, or a proximity sensor that measures the presence or absence of the target 100 based on a change in the magnetic field. It may be a distance measuring sensor that measures the distance to the target 100 using light, a sensor that measures the state of the transport device 50, a temperature sensor, or a pressure sensor.

The computer 30 is a general-purpose computer, and includes at least an arithmetic unit such as a CPU (Central Processing Unit), a memory, a display device, and an input device. The computer 30 displays the information received from the controller 10 and transmits the information input via the input device to the controller 10.

The transport device 50 is a device that transports the target 100, and may be, for example, a belt conveyor or a roller conveyor. The transport device 50 is an example of a device for operating the target 100, and the system 1 may include a device other than the transport device 50. For example, the system 1 may include a robot that operates the target 100 and a machine tool that manufactures the target 100. In that case, the robot or machine tool may be controlled by the controller 10.

FIG. 2 is a diagram showing a functional block of the controller 10 according to the present embodiment. The controller 10 includes an acquisition unit 11, an extraction unit 12, a feature amount extraction unit 13, a selection unit 14, a generation unit 15, a storage unit 16, a determination unit 17, and an output unit 18.

The acquisition unit 11 acquires a time-series numerical value indicating the state of the target 100 measured by the sensor 20. The time-series numerical value indicating the state of the object 100 may be acquired at a predetermined time interval or may be acquired irregularly. The acquired time-series numerical values may be stored in the storage unit 16.

The extraction unit 12 extracts extreme value candidates from the numerical values in the time series. As will be described in detail later, when three consecutive numerical values in the time series satisfy a predetermined magnitude relationship, the extraction unit 12 extracts the central numerical value among the three numerical values as an extreme value candidate. You can do it. As a result, extreme value candidates can be extracted by processing with a relatively light load, and feature quantities can be extracted at high speed.

The feature amount extraction unit 13 extracts the time from the reference time point until the candidate for the extreme value is generated as the feature amount. Here, the reference time point may be the start time point of the cycle for controlling the target 100. The cycle of controlling the target 100 may be, for example, a cycle of transporting the target 100 by a transport device 50 for a predetermined distance, or a cycle of operating the target 100 by a robot or a machine tool. By setting the start time of the cycle for controlling the target 100 as the reference time point, it is possible to obtain a feature amount capable of more accurately determining the state of the target 100 for each cycle for controlling the target 100.

The feature amount extraction unit 13 may extract the value of the extreme value candidate as a feature amount in addition to the time from the reference time to the occurrence of the extreme value candidate. In addition, the feature amount extraction unit 13 may extract statistics such as the average and variance of time-series numerical values in one cycle for controlling the target 100 as feature amounts.

The selection unit 14 accepts selection of one or a plurality of feature quantities from a plurality of types of feature quantities extracted by the feature quantity extraction unit 13. The selection of the feature amount may be performed via an input unit (not shown) provided in the controller 10, or may be performed via an input device of the computer 30.

The generation unit 15 generates a determination model 16a for determining the state of the target 100 based on the selected one or a plurality of feature quantities. The generation unit 15 may not only store the generated determination model 16a in the storage unit 16 but also output it to an external device such as a computer 30.

The storage unit 16 stores at least the determination model 16a. The determination model 16a may be, for example, a model that compares one or a plurality of feature quantities with a threshold value and determines the state of the target 100 based on the magnitude relationship between the one or a plurality of feature quantities and the threshold value. However, the determination model 16a may be a learning model such as a neural network that determines the state of the target 100 based on one or a plurality of feature quantities.

The determination unit 17 determines the state of the target 100 using the determination model 16a based on the feature amount extracted by the feature amount extraction unit 13. In the present embodiment, the object 100 whose state is determined by the determination unit 17 is illustrated as an object to be conveyed by the transfer device 50, but the object whose state is determined by the determination unit 17 is an arbitrary object. The target may be the transfer device 50, or may be a robot or a machine tool. The selection unit 14 and the generation unit 15 can select one or a plurality of feature quantities that are effective for determining the state of the target 100, and generate a determination model 16a that determines the state of the target 100 more accurately. it can.

The output unit 18 outputs the relationship between the plurality of types of feature quantities and the state of the target 100 in a manner in which the plurality of types of feature quantities can be compared. The output unit 18 outputs, for example, the relationship between the plurality of types of feature quantities and the state of the target 100 to the computer 30 in a manner capable of comparing the plurality of types of feature quantities. An output example by the output unit 18 will be described with reference to FIG. When the output unit 18 selects one or a plurality of feature quantities that are effective for determining the state of the target 100, the relationship between the plurality of types of feature quantities and the state of the target 100 is compared, and more appropriate features are compared. You will be able to select the amount more efficiently.

FIG. 3 is a diagram showing time-series numerical values acquired by the controller 10 according to the present embodiment. In the figure, the vertical axis shows the measured values of time-series numerical values, and the horizontal axis shows the elapsed time.

The time-series numerical values of this example include the first section P1, the second section P2, and the third section P3. The first section P1, the second section P2, and the third section P3 may be, for example, a section determined based on the control cycle of the transport device 50, and more generally, based on the control cycle for controlling the target 100. It may be a section defined by.

The controller 10 extracts the extreme value first candidate D1, the extreme value second candidate D2, and the extreme value third candidate D3 by the feature amount extraction process described later. Here, the first candidate D1 of the extremum is the maximum value, the second candidate D2 of the extremum is the minimum value, and the third candidate D3 of the extremum is the maximum value.

Further, the controller 10 has a time t1 from the start time of the first section P1 to the occurrence of the first candidate D1 of the extremum by the feature amount extraction process described later, and the second extremum from the start time of the first section P1. The time t2 until the candidate D2 is generated and the time t3 from the start time of the first section P1 until the extremum third candidate D3 is generated are extracted as feature quantities.

The controller 10 similarly extracts the feature amount for the sections other than the first section P1, such as the second section P2 and the third section P3.

FIG. 4 is a diagram showing a plurality of types of feature quantities extracted by the controller 10 according to the present embodiment. In the figure, as multiple types of features, "maximum value 1", "time_maximum value 1", "minimum value 1", "time_maximum value 1", "maximum value 2", and "time_maximum value" 2 ”is shown. Here, "maximum value 1" is the value of the first candidate D1 of the extreme value in FIG. 3, "time_maximum value 1" is time t1, and "minimum value 1" is the first candidate of the extreme value. The value of the two candidates D2, "time_maximum value 1" is the time t2, "maximum value 2" is the value of the third candidate D3 of the extreme value, and "time_maximum value 2" is. The time is t3. Further, in the figure, the inspection result of the target 100 is indicated by "good" or "bad" for each interval (separation) of the numerical value in the time series.

In the case of this example, for the first interval P1 (delimiter 1), the "maximum value 1" is "15.0", the "time_maximum value 1" is "10.0", and the "minimum value 1". Is "4.3", "time_maximum 1" is "40.0", "maximum 2" is "7.1", and "time_maximum 2" is "80. It is "0". Then, for the first section P1 (separation 1), the inspection result of the target 100 is "good".

Further, regarding the second section P2 (separation 2), the "maximum value 1" is "13.1", the "time_maximum value 1" is "10.0", and the "minimum value 1" is "4". "0.0", "time_maximum 1" is "40.0", "maximum 2" is "8.1", and "time_maximum 2" is "80.0". is there. Then, the inspection result of the target 100 is "good" for the second section P2 (separation 2).

Further, for the third section P3 (separation 3), the "maximum value 1" is "14.0", the "time_maximum value 1" is "10.0", and the "minimum value 1" is "4". .3 "," time_maximum 1 "is" 40.0 "," maximum value 2 "is" 6.2 ", and" time_maximum 2 "is" 83.0 ". is there. Then, with respect to the third section P3 (separation 3), the inspection result of the target 100 is "defective".

FIG. 5 is a diagram showing a plurality of types of feature quantities output by the controller 10 according to the present embodiment and the state of the target 100. In the figure, the first image DP1 showing the relationship between the feature amount of "maximum value 1" and the state of the object 100, the second image DP2 showing the relationship between the feature amount of "time_maximum value 1" and the state of the object 100, The third image DP3 showing the relationship between the feature amount of "minimum value 1" and the state of the object 100, the fourth image DP4 showing the relationship between the feature amount of "time_minimum value 1" and the state of the object 100, "maximum value 2" The fifth image DP5 showing the relationship between the feature amount of "" and the state of the object 100 and the sixth image DP6 showing the relationship between the feature amount of "time_maximum value 2" and the state of the object 100 are shown. The image of this may be displayed on the display device of the computer 30, but when the controller 10 includes a display unit, it may be displayed on the display unit of the controller 10.

From the first image DP1, it can be read that it is not possible to distinguish between good and bad of the target 100 depending on the magnitude of the feature amount of the "maximum value 1". Further, it can be read from the second image DP2 that it is not possible to distinguish between good and bad of the target 100 depending on the size of the feature amount of "time_maximum value 1".

Similarly, it can be read from the third image DP3 that it is not possible to distinguish between good and bad of the target 100 depending on the size of the feature amount of the "minimum value 1". Further, from the fourth image DP4, it can be read that it is not possible to distinguish between good and bad of the target 100 depending on the size of the feature amount of "time_minimum value 1".

On the other hand, from the fifth image DP5, it can be read that good or bad of the target 100 can be discriminated by the magnitude of the feature amount of the "maximum value 2". Further, from the sixth image DP6, it can be read that good or bad of the target 100 can be discriminated by the magnitude of the feature amount of "time_maximum value 2".

In this way, by outputting the relationship between the plurality of types of feature quantities and the state of the target 100 in a comparable manner for the plurality of types of feature quantities, one or a plurality of characteristics that are effective for determining the state of the target 100. Features can be selected more efficiently.

FIG. 6 is a flowchart of the determination model generation process executed by the controller 10 according to the present embodiment. First, the controller 10 acquires a time-series numerical value indicating the state of the target 100 (S10). The controller 10 may acquire a time-series numerical value directly from the sensor 20, or may acquire a time-series numerical value stored in the storage unit 16.

Next, the controller 10 extracts the feature amount from the numerical values in the time series (S11). The feature amount extraction process (S11) will be described in detail with reference to the following figure.

After that, the controller 10 determines whether or not the feature amount extraction process has been performed on the entire time-series numerical values (S12). When the processing is not completed for all the data (S12: NO), the controller 10 repeats the processing S10 and S11.

On the other hand, when the processing for all the data is completed (S12: YES), the controller 10 outputs the relationship between the state of the plurality of types of features and the state of the target 100 in a mode in which the plurality of types of features can be compared (S13). .. Then, the controller 10 accepts selection of one or a plurality of feature quantities from the plurality of types of feature quantities (S14).

After that, the controller 10 generates a determination model 16a for determining the state of the target 100 based on the selected one or a plurality of feature quantities. As a result, the determination model generation process is completed.

FIG. 7 is a flowchart of the feature amount extraction process executed by the controller 10 according to the present embodiment. FIG. 8 shows the details of the process (S11) for extracting the feature amount from the time-series numerical values of FIG.

First, the controller 10 determines whether the acquired data belongs to a delimiter different from the immediately acquired data (S111). When the acquired data belongs to a delimiter different from the data acquired immediately before (S111: YES), the controller 10 stores the time when the delimiter changes (S112). The time when the delimiter changes may be the time when the latest data was acquired, the time when the previous data was acquired, or the time between the time when the latest data was acquired and the time when the previous data was acquired. .. Further, the controller 10 resets the counter to N = 0 (S113).

After that, the controller 10 satisfies "two previous measurement values <one previous measurement value ≥ current measurement value", that is, the two previous measurement values are smaller than the previous measurement value and 1 It is determined whether the previous measured value is equal to or higher than the current measured value (S114). As a result, it is determined whether or not the time-series numerical value is a candidate for the maximum value. The controller 10 may determine whether or not "two previous measurement values ≤ one previous measurement value> current measurement value" is satisfied.

Further, the controller 10 determines whether or not "two previous measurement values ≥ one previous measurement value <current measurement value" is satisfied, or "two previous measurement values> one previous measurement value ≤ current measurement value". It may be determined whether the "measured value" is satisfied. As a result, it is determined whether or not the time-series numerical value is a candidate for the minimum value.

When the previous measurement value is smaller than the previous measurement value and the previous measurement value is greater than or equal to the current measurement value (S114: YES) (or the two previous measurement value is one previous measurement value) When it is larger than the measured value of and the previous measured value is equal to or less than the current measured value), the controller 10 increments the counter N (S115). Then, the controller 10 sets the previous measured value as a candidate for the Nth extreme value and extracts it as a feature amount (S116). Further, the controller 10 calculates the time when the candidate for the extreme value occurs and extracts it as a feature amount (S117).

Here, when calculating the time when the candidate for the maximum value occurs, the controller 10 satisfies the time when "two previous measurement values <one previous measurement value ≥ current measurement value" and "two previous measurements". The average of the times when "measured value ≤ one previous measured value> current measured value" may be used as the time when the candidate for the maximum value occurs. Similarly, when calculating the time when the candidate for the minimum value occurs, the controller 10 satisfies the time when "two previous measurement values ≥ one previous measurement value <current measurement value" and "two previous measurement values". The average of the times when "measured value> previous measured value ≤ current measured value" may be used as the time when the candidate for the minimum value occurs. As a result, the feature amount extraction process is completed.

FIG. 8 is a flowchart of the determination process executed by the controller 10 according to the present embodiment. First, the controller 10 acquires a time-series numerical value indicating the state of the target 100 (S20). The controller 10 may acquire a time series numerical value directly from the sensor 20.

Next, the controller 10 extracts the feature amount from the numerical values in the time series (S21). The process of extracting the feature amount (S21) may be the same as the process shown in FIG. 7.

After that, the controller 10 determines whether the extracted feature amount is the selected feature amount (S22). The selected feature amount is the feature amount selected in the process S14 shown in FIG. When the extracted feature amount is not the selected feature amount (S22: NO), is the controller 10 the acquisition of the time-series numerical value (S20), the extraction of the feature amount (S21), and the selected feature amount? The determination of whether or not (S22) is repeated.

When the extracted feature amount is the selected feature amount (S22: YES), the controller 10 determines the state of the target 100 by the determination model 16a based on the feature amount (S23). After that, the controller 10 outputs the determination result to the computer 30 and the like (S24).

Finally, the controller 10 determines whether or not to end the process (S25), and if the process is not completed (S25: NO), repeats the processes S20 to S24. On the other hand, when the process is terminated (S25: YES), the determination process is terminated.

The embodiments described above are for facilitating the understanding of the present invention, and are not for limiting and interpreting the present invention. Each element included in the embodiment and its arrangement, material, condition, shape, size, etc. are not limited to those exemplified, and can be changed as appropriate. In addition, the configurations shown in different embodiments can be partially replaced or combined.

Aspects of this embodiment include the following disclosures.

[Appendix 1]
An acquisition unit (11) that acquires a time-series numerical value indicating the state of the object measured by the sensor (20), and
An extraction unit (12) that extracts extreme value candidates from the time-series numerical values, and
A feature amount extraction unit (13) that extracts the time from the reference time to the occurrence of the extremum candidate as the feature amount, and
Controller (10).

[Appendix 2]
A system including a sensor (20) for measuring a time-series numerical value indicating a target state and a controller (10) connected to the sensor (20).
The controller (10)
The acquisition unit (11) for acquiring the time-series numerical values and
An extraction unit (12) that extracts extreme value candidates from the time-series numerical values, and
It has a feature amount extraction unit (13) that extracts the time from the reference time point until the candidate for the extreme value is generated as the feature amount.
system.

[Appendix 3]
On the controller (10)
Acquiring a time-series numerical value indicating the state of the object measured by the sensor (20),
Extracting extremum candidates from the time-series numerical values and
Extracting the time from the reference time to the occurrence of the extremum candidate as a feature amount, and
How to run.

[Appendix 4]
On the controller (10)
Acquiring a time-series numerical value indicating the state of the object measured by the sensor (20),
Extracting extremum candidates from the time-series numerical values and
Extracting the time from the reference time to the occurrence of the extremum candidate as a feature amount, and
A program that executes.

1 ... system, 10 ... controller, 11 ... acquisition unit, 12 ... extraction unit, 13 ... feature quantity extraction unit, 14 ... selection unit, 15 ... generation unit, 16 ... storage unit, 16a ... judgment model, 17 ... judgment unit, 18 ... Output unit, 20 ... Sensor, 30 ... Computer, 50 ... Conveyor device, 100 ... Target

Claims (8)

An acquisition unit that acquires time-series numerical values that indicate the state of the object measured by the sensor,
An extraction unit that extracts extreme value candidates from the time-series numerical values,
A feature amount extraction unit that extracts the time from the reference time to the occurrence of the extremum candidate as the feature amount,
Controller with. When three consecutive numerical values among the numerical values in the time series satisfy a predetermined magnitude relationship, the extraction unit extracts the central numerical value among the three numerical values as a candidate for the extreme value.
The controller according to claim 1. A selection unit that accepts selection of one or a plurality of feature quantities from a plurality of types of feature quantities extracted by the feature quantity extraction unit, and
A generation unit that generates a determination model for determining the state of the target based on the one or a plurality of feature quantities.
The controller according to claim 1 or 2, further comprising. The output unit further includes an output unit that outputs the relationship between the plurality of types of feature quantities and the state of the target in a manner comparable to the plurality of types of feature quantities.
The controller according to claim 3. The reference time point is the start time point of the cycle for controlling the object.
The controller according to any one of claims 1 to 4. A system including a sensor for measuring a time-series numerical value indicating a target state and a controller connected to the sensor.
The controller
An acquisition unit that acquires the numerical values of the time series,
An extraction unit that extracts extreme value candidates from the time-series numerical values,
It has a feature amount extraction unit that extracts the time from the reference time point until the candidate for the extreme value is generated as the feature amount.
system. To the controller
Acquiring a time-series numerical value indicating the state of the object measured by the sensor,
Extracting extremum candidates from the time-series numerical values and
Extracting the time from the reference time to the occurrence of the extremum candidate as a feature amount, and
How to run. To the controller
To acquire the time-series numerical value indicating the state of the object measured by the sensor,
Extracting extremum candidates from the time-series numerical values and
Extracting the time from the reference time to the occurrence of the extremum candidate as a feature amount, and
A program that executes.

Images