Emulation of a control system and control method for abnormality detection parameter verification

1. A control system, comprising:

a control device comprising a first processor that controls a control target and a database that stores the data associated with the control target in time series; and

an information processing device comprising a second processor and a connecting device that connects with and exchanges data with the control device,

wherein the first processor comprises:

executing control computation of a sequence control and a motion control related to the control target;

executing a process of collecting data associated with the control target; and

executing a process of monitoring a state of the control target, comprising:

executing a feature quantity generation process for generating a first feature quantity from the collected data; and

executing an abnormality detection process for detecting an abnormality occurring in the control target using the first feature quantity and an abnormality detection parameter suitable for detection of an abnormality occurring in the control target, wherein the abnormality detection parameter comprises a threshold value that is set for classifying the first feature quantity into a class of feature quantities suitable for detection of the abnormality occurring in the control target based on a result of machine learning,

wherein the first processor further comprises cyclically executing the process of collecting data and the control computation, and executing a cyclic execution process for causing the process of collecting data and the control computation to be executed with a higher priority than the monitoring process in the cyclic execution, and

the second processor comprises:

executing emulation of the monitoring process using the data associated with the control target from the control device,

emulating the cyclic execution process, wherein the emulation of the monitoring process is executed in the emulation of the cyclic execution process,

evaluating validity indicating whether or not the abnormality detection parameter that is held in the control device is suitable for detection of the abnormality occurring in the control target based on the result of the emulation of the cyclic execution process and a result of a comparison between a time taken for the emulation of the monitoring process in the emulation of the cyclic execution process and a length of a cycle of the cyclic execution process,

if evaluation of the validity indicates the abnormality detection parameter is not valid, generating a second feature quantity from time-series data of the data associated with the control target in the database of the control device, changing the abnormality detection parameter suitable for detection of the abnormality occurring in the control target by performing machine learning using the generated second feature quantity, and repeating the emulation of the monitoring process and the evaluating of the validity, and

if the evaluation of the validity indicates the abnormality detection parameter is valid, determining the abnormality detection parameter is suitable for detection of the abnormality.

2. The control system according to claim 1 ,

wherein the second processor comprises executing emulation of the feature quantity generation process, and

executing the emulation of the feature quantity generation process using the time-series data in the database to generate the second feature quantity from the time-series data.

3. The control system according to claim 1 ,

wherein the first processor holds the class of feature quantities suitable for detection of the abnormality occurring in the control target, and

the first processor updates the feature quantities of the class using the first feature quantity generated during execution of the monitoring process.

4. A control method that is executed by an information processing device comprising a second processor and a connecting device that connects with and exchanges data with a control device comprising a first processor that controls a control target and a database that stores the data associated with the control target in time series,

wherein the first processor comprises:

executing control computation of a sequence control and a motion control related to the control target,

executing a process of collecting data associated with the control target, and

executing a process of monitoring a state of the control target, comprising:

executing a feature quantity generation process for generating a first feature quantity from the collected data; and

executing an abnormality detection process for detecting an abnormality occurring in the control target using the first feature quantity and an abnormality detection parameter suitable for detection of an abnormality occurring in the control target, wherein the abnormality detection parameter comprises a threshold value that is set for classifying the first feature quantity into a class of feature quantities suitable for detection of the abnormality occurring in the control target based on a result of machine learning,

wherein the first processor further comprises cyclically executing the process of collecting data and the control computation, and executing a cyclic execution process for causing the process of collecting data and the control computation to be executed with a higher priority than the monitoring process in the cyclic execution, and

the control method comprises:

receiving the data associated with the control target from the control device; and

executing emulation of the monitoring process using the data associated with the control target received from the control device,

emulating the cyclic execution process, wherein the emulation of the monitoring process is executed in the emulation of the cyclic execution process,

evaluating validity indicating whether or not the abnormality detection parameter that is held in the control device is suitable for detection of the abnormality occurring in the control target based on the result of the emulation of the cyclic execution process and a result of a comparison between a time taken for the emulation of the monitoring process in the emulation of the cyclic execution process and a length of a cycle of the cyclic execution process,

if evaluation of the validity indicates the abnormality detection parameter is not valid, generating a second feature quantity from time-series data of the data associated with the control target in the database of the control device, changing the abnormality detection parameter suitable for detection of the abnormality occurring in the control target by performing machine learning using the generated second feature quantity, and repeating the emulation of the monitoring process and the evaluating of the validity, and

if the evaluation of the validity indicates the abnormality detection parameter is valid, determining the abnormality detection parameter is suitable for detection of the abnormality.