CN-116830055-B - Production facility monitoring method, production facility monitoring device, and program product

Abstract

The production facility monitoring method is a monitoring method for a production facility using a mahalanobis distance calculated from a plurality of variable data representing a state of the production facility, and includes an acquisition step of acquiring 1 st data which is the data of a past 1 st period up to a current time point, a prediction step of predicting 2 nd data which is the data of a2 nd period after the current time point, and a unit space creation step of creating a unit space which is a calculation basis for the mahalanobis distance from the 1 st data and the 2 nd data, wherein in the prediction step, the 2 nd data is predicted from 3rd data which is the data of a 3rd period in which the 1 st period is moved to the past by a predetermined period, and 4 th data and the 1 st data which are the data of a4 th period in which the 2 nd period is moved to the past by the predetermined period.

Inventors

• NAGANO ICHIRO
• SAITO MAYUMI
• KUNITAKE NOBUHIRO
• EGUCHI KOJI

Assignees

• 三菱重工业株式会社

Dates

Publication Date

20260505

Application Date

20220419

Priority Date

20210514

Claims (16)

1. 1. A production facility monitoring method which is a monitoring method of a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring method includes: An acquisition step of acquiring 1 st data, which is the data in the past 1 st period up to the current time point; A prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which becomes a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, In the predicting step, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time portion in the past, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time portion in the past, and the 1 st data, In the predicting step, the 2 nd data is predicted using a value representing the data change between the 3 rd period and the 4 th period or the data change between the 3 rd period and the 1 st period.
2. 2. A production facility monitoring method which is a monitoring method of a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring method includes: An acquisition step of acquiring 1 st data, which is the data in the past 1 st period up to the current time point; A prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which becomes a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, In the predicting step, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time portion in the past, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time portion in the past, and the 1 st data, In the predicting step, the 2 nd data is obtained by adding, for one variable among the plurality of variables, a value based on a difference between an average of the 4 th data and an average of the 3 rd data to the 1 st data.
3. 3. The production facility monitoring method of claim 2, wherein, Multiplying a value obtained by dividing the difference by the standard deviation of the 3 rd data by the standard deviation of the 1 st data, and adding the value obtained thereby to the 1 st data to obtain the 2 nd data.
4. 4. A production facility monitoring method which is a monitoring method of a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring method includes: An acquisition step of acquiring 1 st data, which is the data in the past 1 st period up to the current time point; A prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which becomes a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, In the predicting step, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time portion in the past, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time portion in the past, and the 1 st data, In the predicting step, the 2 nd data is obtained by adding, for one variable among the plurality of variables, a value based on a difference between an average of the 1 st data and an average of the 3 rd data to the 4 th data.
5. 5. The production facility monitoring method of claim 4, wherein, Multiplying a value obtained by dividing the difference by the standard deviation of the 3 rd data by the standard deviation of the 4 th data, and adding the value obtained thereby to the 4 th data to obtain the 2 nd data.
6. 6. The production facility monitoring method according to any one of claims 1 to 5, wherein, The 3 rd data is a data set of the variable during the 3 rd period in which the time share of the prescribed length set for each of the plurality of variables is moved to the past in the 1 st period, The 4 th data is a data set of the variable in the 4 th period, in which the time share of the predetermined length set for each of the plurality of variables is moved to the past in the 2 nd period, respectively.
7. 7. The production facility monitoring method according to any one of claims 1 to 5, wherein, The predetermined length of time set for at least one of the plurality of variables is 1 year.
8. 8. The production facility monitoring method of claim 7, wherein, The prescribed length of time set for at least another variable of the plurality of variables is a component replacement cycle of production facility equipment associated with the another variable.
9. 9. The production facility monitoring method according to any one of claims 1 to 5, wherein, The number of 1 st data constituting the unit space is greater than the number of 2 nd data constituting the unit space.
10. 10. The production facility monitoring method according to any one of claims 1 to 5, comprising: a selection step of randomly selecting data for creating the unit space among the 2 nd data, In the unit space creation step, the unit space is created using at least a part of the 1 st data and the data selected in the selection step.
11. 11. A production facility monitoring apparatus which is a monitoring apparatus for a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring device is provided with: An acquisition unit configured to acquire 1 st data, which is the data in the 1 st period in the past up to the current time point; a prediction unit configured to predict 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation unit configured to create a unit space that is a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, The prediction unit is configured to predict the 2 nd data based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, The prediction unit predicts the 2 nd data using a value indicating the data change between the 3 rd period and the 4 th period or the data change between the 3 rd period and the 1 st period.
12. 12. A production facility monitoring apparatus which is a monitoring apparatus for a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring device is provided with: An acquisition unit configured to acquire 1 st data, which is the data in the 1 st period in the past up to the current time point; a prediction unit configured to predict 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation unit configured to create a unit space that is a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, The prediction unit is configured to predict the 2 nd data based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, The prediction unit obtains the 2 nd data by adding, to the 1 st data, a value based on a difference between an average of the 4 th data and an average of the 3 rd data, for one variable among the plurality of variables.
13. 13. A production facility monitoring apparatus which is a monitoring apparatus for a production facility using a Marshall distance calculated from data representing a plurality of variables of a state of the production facility, wherein, The production facility monitoring device is provided with: An acquisition unit configured to acquire 1 st data, which is the data in the 1 st period in the past up to the current time point; a prediction unit configured to predict 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation unit configured to create a unit space that is a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, The prediction unit is configured to predict the 2 nd data based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, The prediction unit obtains the 2 nd data by adding, to the 4 th data, a value based on a difference between an average of the 1 st data and an average of the 3 rd data, for one variable among the plurality of variables.
14. 14. A production facility monitoring program product comprising a monitoring program for a production facility using mahalanobis distances calculated from data representing a plurality of variables of a state of the production facility, wherein, The monitor program causes a computer to execute: An acquisition step of acquiring 1 st data, which is the data of the 1 st period in the past up to the current time point; a prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which is a basis for calculating the mahalanobis distance from the 1 st data and the 2 nd data, In the step of predicting the 2 nd data, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, In the step of predicting the 2 nd data, the 2 nd data is predicted using a value indicating the data change between the 3 rd period and the 4 th period or the data change between the 3 rd period and the 1 st period.
15. 15. A production facility monitoring program product comprising a monitoring program for a production facility using mahalanobis distances calculated from data representing a plurality of variables of a state of the production facility, wherein, The monitor program causes a computer to execute: An acquisition step of acquiring 1 st data, which is the data of the 1 st period in the past up to the current time point; a prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which is a basis for calculating the mahalanobis distance from the 1 st data and the 2 nd data, In the step of predicting the 2 nd data, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, In the step of predicting the 2 nd data, the 2 nd data is obtained by adding the 1 st data to a value of a difference between the average of the 4 th data and the average of the 3 rd data, with respect to one variable among the plurality of variables.
16. 16. A production facility monitoring program product comprising a monitoring program for a production facility using mahalanobis distances calculated from data representing a plurality of variables of a state of the production facility, wherein, The monitor program causes a computer to execute: An acquisition step of acquiring 1 st data, which is the data of the 1 st period in the past up to the current time point; a prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which is a basis for calculating the mahalanobis distance from the 1 st data and the 2 nd data, In the step of predicting the 2 nd data, the 2 nd data is predicted based on 3 rd data which is the data of 3 rd period in which the 1 st period is moved by a predetermined length of time, 4 th data which is the data of 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data, In the step of predicting the 2 nd data, the 2 nd data is obtained by adding, to the 4 th data, a value based on a difference between an average of the 1 st data and an average of the 3 rd data, for one variable among the plurality of variables.

Description

Production facility monitoring method, production facility monitoring device, and program product Technical Field The present invention relates to a production facility monitoring method, a production facility monitoring apparatus, and a production facility monitoring program. The present application claims priority from 2021-08212, filed by the japanese patent office on the basis of month 14 of 2021, and the contents thereof are incorporated herein. Background The production facility is sometimes monitored using a mahalanobis distance (Mahalanobis distance) that represents the divergence between a standard dataset representing a variable representing the state of the production facility (state quantity available for sensor acquisition, etc.) and the measured data for that variable. Patent document 1 describes a method for monitoring a production facility using a mahalanobis distance, in which the mahalanobis distance is calculated using a plurality of unit spaces set from the operation period. Here, the unit space is a data set that is a reference for determining whether or not the operation state of the production facility is normal. More specifically, in patent document 1, a mahalanobis distance for data acquired during a start-up operation of a production facility is calculated using a unit space created from a state quantity of the production facility during the start-up operation of the production facility, and a mahalanobis distance for data acquired during a load operation of the production facility is calculated using a unit space created from a state quantity of the production facility during the load operation of the production facility. Technical literature of the prior art Patent literature Patent document 1 Japanese patent No. 5031088 Disclosure of Invention Technical problem to be solved by the invention The unit space that becomes the basis of the calculation of the mahalanobis distance is generally constituted by data (reference data) acquired by a sensor during the past. When the mahalanobis distance of the measurement data (evaluation target data) is calculated for a time point (for example, a current time point or a recent time point) in a period subsequent to the past period (period in which the reference data is acquired) using such a unit space, there is a case where the data trend in the past period in which the reference data is acquired does not coincide with the data trend in the period in which the evaluation target data is acquired. In this case, the accuracy of detecting the abnormality of the production facility based on the mahalanobis distance calculated for the evaluation target data may be poor. In view of the above, an object of at least one embodiment of the present invention is to provide a production facility monitoring method, a production facility monitoring apparatus, and a production facility monitoring program that can detect abnormality of a production facility with high accuracy. Means for solving the technical problems A production facility monitoring method according to at least one embodiment of the present invention is a monitoring method of a production facility using a mahalanobis distance calculated from a plurality of variable data representing a state of the production facility, The production facility monitoring method includes: An acquisition step of acquiring 1 st data, which is the data in the past 1 st period up to the current time point; A prediction step of predicting the 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation step of creating a unit space which becomes a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, In the predicting step, the 2 nd data is predicted based on the 3 rd data which is the data of the 3 rd period in which the 1 st period is moved by a predetermined length of time, the 4 th data which is the data of the 4 th period in which the 2 nd period is moved by the predetermined length of time, and the 1 st data. Further, a production facility monitoring device according to at least one embodiment of the present invention is a monitoring device for a production facility using a mahalanobis distance calculated from a plurality of variable data indicating a state of the production facility, The production facility monitoring device is provided with: An acquisition unit configured to acquire 1 st data, which is the data in the 1 st period in the past up to the current time point; a prediction unit configured to predict 2 nd data, which is the data of the 2 nd period after the current time point, and A unit space creation unit configured to create a unit space that is a basis for calculating the mahalanobis distance based on the 1 st data and the 2 nd data, The prediction unit is configured to predict the 2 nd data based on the 3 rd data, which is the data of the 3 rd period in which the 1 st period is moved by a predetermined length of time,