Abnormality determination device, learning device, and abnormality determination method

Patent metadata

Field	Value
Publication number	US-11526783-B2
Application number	US-202016788355-A
Country	US
Kind code	B2
Filing date	Feb 12, 2020
Priority date	Aug 26, 2019
Publication date	Dec 13, 2022
Grant date	Dec 13, 2022

Abstract

An abnormality determination device includes one or more processors. The processors input first input data to a first model to obtain first output data. The first output data is formed by restoring data with the reduced dimension to data with the same dimension as that of the first input data. The processors input second input data, which is a difference between the first input data and the first output data, to a second model, and obtain second output data. The second output data is formed by restoring data with the reduced dimension to data with the same dimension as that of the second input data. The processors obtain restored data that is a sum of the first output data and the second output data. The processors compare the first input data with the restored data and determine an abnormality in the first input data based on the comparison result.

First claim

What is claimed is:

1. An abnormality determination device comprising:
one or more processors configured to:
input data collected at a plant at current time as first input data to a first model, and obtain first output data output by the first model, the first output data being formed by restoring first data to a dimension identical to a dimension of the first input data, the first data being formed by reducing a dimension of the first input data;
input second input data to a second model, and obtain second output data output by the second model, the second input data being a difference between the first input data and the first output data, the second output data being formed by restoring second data to a dimension identical to a dimension of the second input data, the second data being formed by reducing a dimension of the second input data;
obtain restored data that is a sum of the first output data and the second output data;
compare the first input data with the restored data and determine an abnormality in the first input data based on a comparison result; and
identify an abnormal operating condition of the plant at the current time based on the determined abnormality in the first input data.

2. The abnormality determination device according to claim 1 , wherein the first input data is a plurality of pieces of time series data within a designated period.

3. The abnormality determination device according to claim 1 , wherein the first input data is a command value for a plant.

4. The abnormality determination device according to claim 1 , wherein the first input data is sensor data output from a sensor for use in a plant.

5. The abnormality determination device according to claim 1 , wherein the first input data is a computed value obtained by computation using the sensor data.

6. The abnormality determination device according to claim 1 , wherein the first input data is a command value for the plant.

7. The abnormality determination device according to claim 1 , wherein the first model and the second model are auto encoders.

8. The abnormality determination device according to claim 1 , wherein the one or more processors
learn the first model using data collected at the plant when it was in a normal operating state as a first input data; and
learn the second model using second input data that is a difference between the first input data collected at the plant when it was in a normal operating state and the first output data output by the first model in response to the first input data.

9. The abnormality determination device according to claim 1 , wherein the one or more processors control an output of the determined abnormality.

10. The abnormality determination device according to claim 1 , wherein the first model is a model that inputs the first input data that is data at a first time, and outputs the first output data that is data obtained by restoring the data at the first time.

11. An abnormality determination device including a learning device comprising:
one or more processors configured to:
learn a first model using first learning data representing data collected at a plant when it was in a normal operating state as first input data, wherein the first model inputs first input data and outputs first output data obtained by restoring first data to a dimension identical to a dimension of the first input data, the first data being formed by reducing a dimension of the first input data; and
learn a second model using second learning data representing a second input data that is a difference between the first input data collected at the plant when it was in a normal operating state and the first output data, and output second output data obtained by restoring second data to a dimension identical to a dimension of the second input data, the second data being formed by reducing a dimension of the second input data;
input data collected at the plant at current time as first input data to the first model, and obtain first output data output by the first model, the first output data being formed by restoring first data to a dimension identical to a dimension of the first input data, the first data being formed by reducing a dimension of the first input data;
input second input data to the second model, and obtain second output data output by the second model, the second input data being a difference between the first input data and the first output data, the second output data being formed by restoring second data to a dimension identical to a dimension of the second input data, the second data being formed by reducing a dimension of the second input data;
obtain restored data that is a sum of the first output data and the second output data;
compare the first input data with the restored data and determine an abnormality in the first input data based on a comparison result; and
identify an abnormal operating condition of the plant at the current time based on the determined abnormality in the first input data.

12. An abnormality determination method comprising:
inputting data collected at a plant at current time as first input data to a first model, and obtaining first output data output by the first model, the first output data being formed by restoring first data to a dimension identical to a dimension of the first input data, the first data being formed by reducing a dimension of the first input data;
inputting second input data to a second model, and obtaining second output data output by the second model, the second input data being a difference between the first input data and the first output data, the second output data being formed by restoring second data to a dimension identical to a dimension of the second input data, the second data being formed by reducing a dimension of the second input data;
obtaining restored data that is a sum of the first output data and the second output data;
comparing the first input data with the restored data and determining an abnormality in the first input data based on a comparison result; and
identifying an abnormal operating condition of the plant at the current time based on the determined abnormality in the first input data.

Assignees

• Toshiba Kk

Inventors

• Naito Susumu
• Nakata Kouta
• Taguchi Yasunori
• Kato Yuichi

Classifications

• G06N3/088 Primary

  Non-supervised learning, e.g. competitive learning · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N5/04 Primary

  Inference or reasoning models · CPC title

• G05B13/04

  involving the use of models or simulators · CPC title

• G06N20/00

  Machine learning · CPC title