Data processing apparatus, method, and program

Patent metadata

Field	Value
Publication number	US-12124251-B2
Application number	US-202217942976-A
Country	US
Kind code	B2
Filing date	Sep 12, 2022
Priority date	Nov 1, 2021
Publication date	Oct 22, 2024
Grant date	Oct 22, 2024

Abstract

According to one embodiment, a data processing apparatus includes a processor. The processor calculates, from the first measurement data, a first differential value set that is a set of first differential values in a time direction at a time included in the first period of the measurement values of the sensor of interest. The processor calculates, from the second measurement data, a second differential value set that is a set of second differential values in a time direction at a time included in the second period of the measurement values of the sensor of interest. The processor generates a first differential value distribution and a second differential value distribution using the second differential value set.

First claim

What is claimed is:

1. A data processing apparatus for detecting a drift regarding a time derivative, the apparatus comprising a processor configured to:
acquire first measurement data that is a set of measurement values in a first period among time series data of measurement values of a sensor of interest among a plurality of sensors included in a system;
acquire second measurement data that is a set of measurement values in a second period different from the first period among the time series data of the measurement values of the sensor of interest;
calculate, from the first measurement data, a first differential value in a time direction at a first time included in the first period of the measurement values of the sensor of interest, wherein the first differential value represents a time derivative in the first period of the measurement values of the sensor of interest, wherein the processor calculates a first differential value set that is a set of first differential values calculated by changing the first time in the first period;
calculate, from the second measurement data, a second differential value in a time direction at a second time included in the second period of the measurement values of the sensor of interest, wherein the second differential value represents a time derivative in the second period of the measurement values of the sensor of interest, wherein the processor calculates a second differential value set that is a set of second differential values calculated by changing the second time in the second period;
generate a first differential value distribution using the first differential value set;
generate a second differential value distribution using the second differential value set;
detect whether a drift of the time derivative of the measurement value of the sensor of interest has occurred between the first period and the second period based on the generated first differential value distribution and the generated second differential value distribution, and generate a drift detection signal indicating whether the drift of the time derivative has been detected; and
output the first differential value distribution, the second differential value distribution, and drift information regarding detection of the drift of the time derivative based on the generated drift detection signal,
wherein the measurement values of the plurality of sensors are monitored by a machine learning model to detect an anomaly in the measurement values,
wherein the machine learning model is trained by training data in which a data period is the first period, and the second period is a period after the first period;
wherein the processor executes processing of detecting a drift of the time derivative for each of measurement values of a plurality of the sensors of interest,
wherein the processor detects whether a number of sensors for which the drift of the time derivative is detected is larger than or equal to a threshold value;
wherein the processor generates and outputs the drift information to include information prompting update of the machine learning model when it is detected that the number of sensors for which the drift of the time derivative is detected is larger than or equal to a threshold value.

2. The apparatus according to claim 1 , wherein the processor calculates each of the first differential values by calculating a difference between a measurement value at a first attention time in the first period and a measurement value at a first calculation target time separated from the first attention time by a predetermined time, and
calculates each of the second differential values by calculating a difference between a measurement value at a second attention time in the second period and a measurement value at a second calculation target time separated from the second attention time by the predetermined time.

3. The apparatus according to claim 2 , wherein the predetermined time is shorter than a time window length of part of time series data to be input to the machine learning model applied to the time series data to detect whether an anomaly has occurred.

4. The apparatus according to claim 2 , wherein the predetermined time is a time obtained by subtracting one unit time from a time corresponding to a time window length of part of time series data to be an input to the machine learning model applied to the time series data of measurement values of the sensor of interest.

5. The apparatus according to claim 2 , wherein the predetermined time is a time set by a user instruction.

6. The apparatus according to claim 1 , wherein the processor displays at least one of a first trend graph indicating a temporal transition of each of the first differential values in the first period and a second trend graph indicating a temporal transition of each of the second differential values in the second period.

7. The apparatus according to claim 1 , wherein the processor is further configured to calculate a score that is similarity or dissimilarity between the first differential value distribution and the second differential value distribution.

8. The apparatus according to claim 7 , wherein the processor is further configured to perform detection by regarding that the drift of the time derivative of the measurement value of the sensor of interest has occurred between the first period and the second period in a case where the score is less than a threshold when the score represents the similarity, and in a case where the score is greater than the threshold when the score represents the dissimilarity.

9. The apparatus according to claim 8 , wherein the processor outputs list information of one or more sensors in which the drift of the time derivative is detected.

10. The apparatus according to claim 8 , wherein the processor outputs at least one of information regarding a number of sensors in which the drift of the time derivative is detected and information regarding a temporal change in the number of sensors.

11. The apparatus according to claim 1 , wherein the processor outputs information regarding at least one of a maximum value, a minimum value, an average value, and a median value of the measurement values in the first period and/or at least one of a maximum value, a minimum value, an average value, and a median value of the measurement values in the second period.

12. The apparatus according to claim 1 , wherein the processor further configured to create a scatter diagram in which the first measurement data is plotted in a first display mode and the second measurement data is plotted in a second display mode different from the first display mode in a two-dimensional area, the two dimensional area including a first axis representing time of interest and a second axis representing time that is a predetermined time away from the time of interest, and
outputs the scatter diagram.

13. The apparatus according to claim 1 , wherein the processor is further configured to:
acquire each piece of time series data of measurement values from one or more sensors;
acquire the machine learning model trained by training data in which a data period is the first period;
input the each piece of time series data into the trained machine learning model, and generate a predicted value; and
output information regarding a trend graph of the measurement values and the predicted value.

14. The apparatus according to claim 13 , wherein the processor is further configured to determine that there is an anomaly in the sensor that has output the measurement values when a difference between each of the measurement values and the predicted value is larger than or

Assignees

• Toshiba Kk
• Toshiba Energy Systems & Solutions Corp

Inventors

• Taguchi Yasunori
• Nakata Kouta
• Naito Susumu
• Kato Yuichi
• Tominaga Shinya
• Nagura Isaku
• Miyake Ryota
• Terakado Yusuke
• Aoki Toshio

Classifications

• G05B19/418

  Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] · CPC title

• G06N20/00

  Machine learning · CPC title

• G05B23/0221 Primary

  Preprocessing measurements, e.g. data collection rate adjustment; Standardization of measurements; Time series or signal analysis, e.g. frequency analysis or wavelets; Trustworthiness of measurements; Indexes therefor; Measurements using easily measured parameters to estimate parameters difficult to measure; Virtual sensor creation; De-noising; Sensor fusion; Unconventional preprocessing inherently present in specific fault detection methods like PCA-based methods · CPC title