Search

BR-112024011588-B1 - NORMAL VECTOR RECORDING DEVICE, EQUIPMENT ABNORMALITY MONITORING SYSTEM AND EQUIPMENT ABNORMALITY MONITORING METHOD

BR112024011588B1BR 112024011588 B1BR112024011588 B1BR 112024011588B1BR-112024011588-B1

Abstract

NORMAL VECTOR RECORDING DEVICE, INSTALLATION ABNORMALITY MONITORING SYSTEM AND INSTALLATION ABNORMALITY MONITORING METHOD. A normal vector recording device according to the present invention includes: a time series signal clipping processing unit configured to clip a time series signal at a predetermined period of M (= 2) or more time series signal types indicating an operating state of an installation during normal installation operation; a normal vector recording processing unit configured to generate an M-dimensional vector including M types of variables at the same point in time from the time series signal clipped by the time series signal clipping processing unit, and record the generated M-dimensional vector at each point in time in a database as a normal vector; and a normal vector distribution density leveling unit configured to split a vector space of normal vectors, select a predetermined number of normal vectors from the normal vectors included in each split space in order to level a normal vector distribution density, and optimize a number of (...).

Inventors

  • Takehide Hirata
  • Masafumi Matsushita

Assignees

  • JFE STEEL CORPORATION

Dates

Publication Date
20260310
Application Date
20221219
Priority Date
20211222

Claims (8)

  1. 1. Normal vector recording device, CHARACTERIZED in that it comprises: a time series signal clipping processing unit (271) configured to clip a time series signal at a predetermined period of M (> 2) or more time series signal types indicating an operating state of a facility during normal facility operation; a normal vector recording processing unit (273) configured to generate an M-dimensional vector including M types of variables at the same point in time from the time series signal clipped by the time series signal clipping processing unit (271), and record the generated M-dimensional vector at each point in time in a database (5) as a normal vector; a normal vector distribution density leveling unit (277) configured to divide a vector space of normal vectors; select a predetermined number of normal vectors from the normal vectors included in each divided space and register the selected normal vectors in the database (5) as the normal vectors whose distribution density was leveled, so as to level a distribution density of normal vectors, and optimize a division number of the vector space of normal vectors so that a total number of selected normal vectors becomes a pre-designated number.
  2. 2. Normal vector registration device, according to claim 1, CHARACTERIZED in that the normal vector distribution density leveling unit (277) is configured to equally divide the vector space into a plurality of cells (SC), and select a predetermined number of normal vectors from each cell (SC), including normal vectors, and register the selected normal vectors in the database (5) as the normal vectors whose distribution density has been leveled, so as to level the normal vector distribution density, and optimize the cell (SC) sizes so that a total number of selected normal vectors becomes a pre-designated number.
  3. 3. Normal vector registration device, according to claim 1, CHARACTERIZED in that the normal vector distribution density leveling unit (277) is configured to define, within the vector space, a plurality of spheres (Si) whose central positions do not overlap and whose central positions are defined as randomly selected normal vectors; select a predetermined number of normal vectors from each sphere (Si), and register the selected normal vectors in the database (5) as the normal vectors whose distribution density has been leveled, so as to level the distribution density of the normal vectors, and optimize the sizes of the spheres (Si) so that the total number of selected normal vectors becomes a pre-designated number.
  4. 4. Normal vector registration device, according to any one of claims 1 to 3, CHARACTERIZED in that it further comprises a normal vector selection unit (275) configured to: calculate a distance between each normal vector registered in the database (5) and another normal vector registered in the database (5); and delete the normal vector from the database (5) according to the calculated distance.
  5. 5. Normal vector registration device, according to claim 4, CHARACTERIZED in that the normal vector selection unit (275) is configured to perform processing including: calculating a distance between each normal vector registered in the database (5) and another normal vector registered in the database (5); extracting a predetermined number of other normal vectors as neighboring vectors (NV) in ascending order of the calculated distance; and excluding the normal vector from the database (5) according to a distance to a centroid vector (VG) of the extracted neighboring vector.
  6. 6. Installation abnormality monitoring system, CHARACTERIZED in that it comprises an abnormality determination unit (379) configured to: clip a time series signal at a predetermined period of M (> 2) or more time series signal types indicating an operating state of an installation during an installation abnormality monitoring period; generate, from the clipped time series signal, an M-dimensional vector including M types of variables at the same point in time as an abnormality determination target vector; and conduct an installation abnormality determination based on a distance between the generated abnormality determination target vector and a normal vector recorded in a database (5) by the normal vector recording device, as defined in any one of claims 1 to 5.
  7. 7. Installation abnormality monitoring system, according to claim 6, CHARACTERIZED in that the abnormality determination unit (379) is configured to determine whether the installation needs to be repaired based on the number of times the installation has been determined to be abnormal.
  8. 8. Installation abnormality monitoring method, CHARACTERIZED in that it comprises an abnormality determination step including: clipping a time series signal over a predetermined period of M (> 2) or more time series signal types indicating an operating state of an installation during an installation abnormality monitoring period; generating, from the clipped time series signal, an M-dimensional vector including M types of variables at the same point in time as an abnormality determination target vector; and conducting an installation abnormality determination based on a distance between the generated abnormality determination target vector and a normal vector recorded in a database (5) by the normal vector recording device, as defined in any one of claims 1 to 5.

Description

Field [001] The present invention relates to a normal vector recording device, an anomaly monitoring system in an installation, and a method for monitoring anomalies in an installation. Background [002] Patent Literature 1 describes an anomaly monitoring system including: a normal data extraction unit that extracts a large number of normal data parts, with data from the same time defined as a normal data set, from time series data of a plurality of signals obtained from a facility during normal operation; a target monitoring data extraction unit that extracts data from the same time as a target monitoring data set from time series data of a plurality of signals obtained from the facility during operation; and a determination unit that represents the normal data and the target monitoring data as points in the same variable space, each of the plurality of signals being a variable, and determines the anomaly of the facility based on a degree of deviation of the target monitoring data from a normal data distribution. List of Citations Patent Literature [003] Patent Literature 1: JP 6003718 B Summary Technical problem [004] According to the anomaly monitoring system described in Patent Literature 1, it is possible to achieve universal and high-precision anomaly detection in an installation having a non-linear characteristic. However, the anomaly monitoring system described in Patent Literature 1 has problems including a problem of computation speed deterioration with an excessive number of normal data parts, which causes a failure to determine the anomaly within a predetermined time, and a problem of dramatically increased data capacity, which leads to an increase in cost. For this reason, it is necessary to reduce the number of normal data parts. However, reducing the number of normal data parts with uneven normal data distribution density would cause a problem of deterioration in anomaly detection accuracy. With this scenario, it has been desired to provide a technique for compressing the number of normal data parts without causing irregularity in the normal data distribution density. [005] The present invention was made in view of the above problems, and aims to provide a normal vector recording device, an anomaly monitoring system in installation, and a method for monitoring anomalies in installation capable of achieving high calculation speed and cost reduction through data capacity compression, while maintaining anomaly detection accuracy. Solution to the Problem [006] A normal vector recording device according to a first aspect of the present invention includes: a time series signal clipping processing unit configured to clip a time series signal at a predetermined period of M (> 2) or more types of time series signals indicating an operating state of a facility during normal facility operation; a normal vector recording processing unit configured to generate an M-dimensional vector including M types of variables at the same point in time from the time series signal clipped by the time series signal clipping processing unit, and record the generated M-dimensional vector at each point in time in a database as a normal vector; and a normal vector distribution density leveling unit configured to divide a vector space of normal vectors, select a predetermined number of normal vectors from the normal vectors included in each divided space so as to level a distribution density of normal vectors, and optimize a division number of the vector space of normal vectors, so that a total number of selected normal vectors becomes a pre-designated number. [007] In the normal vector registration device described above according to the first aspect of the present invention, the normal vector distribution density leveling unit is configured to divide the vector space equally into a plurality of cells and select a predetermined number of normal vectors from each cell, including normal vectors, so as to level the normal vector distribution density, and optimize the cell sizes, so that a total number of selected normal vectors becomes a pre-designated number. [008] In the normal vector registration device described above according to the first aspect of the present invention, the normal vector distribution density leveling unit is configured to define, within the vector space, a plurality of spheres whose central positions do not overlap each other and whose central positions are defined as randomly selected normal vectors, to select a predetermined number of normal vectors from each sphere in order to level the distribution density of the normal vectors and to optimize the sizes of the spheres so that the total number of selected normal vectors becomes a pre-designated number. [009] The normal vector registration device described above according to the first aspect of the present invention further includes a normal vector selection unit configured to: calculate a distance between each normal vector registered in the database