Abnormality Detection System and Abnormality Detection Method

Patent metadata

Field	Value
Publication number	US-2022397894-A1
Application number	US-202217828714-A
Country	US
Kind code	A1
Filing date	May 31, 2022
Priority date	Jun 14, 2021
Publication date	Dec 15, 2022
Grant date	—

Abstract

Provided are an abnormality detection system and an abnormality detection method capable of performing more stable abnormality detection. An abnormality detection system that detects an abnormality of the target machine by a computer includes a communication unit configured to acquire first data from a first sensor attached to the target machine and second data from a second sensor attached to the target machine, an arithmetic unit, and a memory unit. The arithmetic unit includes an encoding unit trained to generate latent expressions including a predetermined latent expression that estimates the second data on the basis of the first data, a decoding unit trained to restore the first data from the latent expressions, and an abnormality detection unit configured to detect the abnormality of the target machine on the basis of a restoration error between the first data and the first data restored by the decoding unit.

First claim

What is claimed is:

1 . An abnormality detection system for detecting an abnormality of a target machine by a computer, wherein
the computer includes a communication unit configured to acquire first data from a first sensor attached to the target machine and second data from a second sensor attached to the target machine, an arithmetic unit, and a memory unit, and the arithmetic unit includes
an encoding unit trained to generate latent expressions including a predetermined latent expression that estimates the second data on the basis of the first data,
a decoding unit trained to restore the first data from the latent expressions, and
an abnormality detection unit configured to detect the abnormality of the target machine on the basis of a restoration error between the first data and the first data restored by the decoding unit.

2 . The abnormality detection system according to claim 1 , wherein
the second sensor is temporarily attached to the target machine in a training mode in order to obtain the predetermined latent expression, and the arithmetic unit further includes a state monitoring unit configured to monitor a state of the target machine on the basis of the first data and the second data estimated from the predetermined latent expression.

3 . The abnormality detection system according to claim 2 , wherein
the memory unit stores a sensor correspondence relation management unit in which the second sensor that outputs the second data that can be estimated from the first data is associated with each of first sensor candidates in advance, and the arithmetic unit further includes a first sensor selection unit configured to select any first sensor from the first sensor candidates on the basis of the sensor correspondence relation management unit.

4 . The abnormality detection system according to claim 1 , wherein
the abnormality detection unit is configured to detect the abnormality of the target machine on the basis of the restoration error between the first data observed by the first sensor and the first data restored by the decoding unit and an estimation error between the second data observed by the second sensor and the second data estimated from the predetermined latent expression.

5 . The abnormality detection system according to claim 1 , wherein
the memory unit stores a pre-trained model trained in advance on the basis of the first data from the first sensor and the second data from the second sensor attached to a machine of the same type as the target machine and having a different individual identification number, and the arithmetic unit is configured to use the pre-trained model as an initial model of a training model for adjusting a parameter of the encoding unit and a parameter of the decoding unit.

6 . The abnormality detection system according to claim 1 , wherein
the arithmetic unit further includes a data interpolation unit configured to interpolate the first data from the first sensor.

7 . An abnormality detection method for detecting an abnormality of a target machine by a computer, the abnormality detection method comprising:
the computer
acquiring first data from a first sensor attached to the target machine;
acquiring second data from a second sensor attached to the target machine;
training an encoding unit such that the encoding unit generates latent expressions including a predetermined latent expression that estimates the second data on the basis of the first data;
training a decoding unit such that the decoding unit restores the first data from the latent expressions;
calculating a restoration error between the first data and the first data restored by the decoding unit;
detecting the abnormality of the target machine on the basis of the calculated restoration error; and
monitoring a state of the target machine on the basis of the first data and the second data estimated from the predetermined latent expression.

8 . An abnormality detection system for detecting an abnormality of a target machine, the abnormality detection system comprising:
a feature data extraction unit configured to extract feature data on the basis of sensor data obtained from a sensor corresponding to the target machine; a model training unit configured to restore the feature data by inputting the extracted feature data to a machine learning model; a loss calculation unit configured to calculate a loss on the basis of the feature data restored by the model training unit, the feature data extracted by the feature data extraction unit, and a predetermined hyper parameter; an abnormality determination unit configured to determine the abnormality of the target machine on the basis of the loss calculated by the loss calculation unit; and a hyper parameter setting unit configured to select, on the basis of the feature data extracted by the feature data extraction unit, a hyper parameter corresponding to the feature data extracted from hyper parameters prepared in advance according to a type of the target machine as the predetermined hyper parameter and set the selected hyper parameter in the loss calculation unit.

Assignees

• Hitachi Ltd

Inventors

• Dohi Kota
• Purohit Harsh Pramodbhai
• Tanabe Ryo
• Yamamoto Masaaki
• Endo Takashi
• Kawaguchi Yohei

Classifications

• G05B23/024 Primary

  Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks · CPC title

• G05B23/0221

  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

• G05B23/0243 Primary

  model based detection method, e.g. first-principles knowledge model · CPC title

• G06N20/00

  Machine learning · CPC title