Title:
ANOMALY DETECTION METHOD FOR CONTROLLABLE MEDIUM-HIGH CALENDER BASED ON SPATIAL DOMAIN CONVERSION INDEPENDENT TREE
Document Type and Number:
WIPO Patent Application WO/2022/227240
Kind Code:
A1
Abstract:
An anomaly detection method for a controllable medium-high calender based on a spatial domain conversion independent tree. The method comprises: mapping running state data of a calender into a feature space by using a Gaussian kernel function, and converting a local anomaly into a global anomaly; and on this basis, constructing an independent tree model to perform anomaly detection on a nucleated data set. As for the problem of how to select an optimal division feature and division sample value, the present system provides a selection strategy for an optimal division feature and a division sample value. On the basis of maintaining a global anomaly detection capability, the detection accuracy of local outliers in running state data of a calender is also improved.
More Like This:
Inventors:
QU HONGCHUN (CN)
HU ANMING (CN)
LIU GUANGHUI (CN)
SONG CHUANDONG (CN)
CAO ZHIHAO (CN)
JIANG ZHENFENG (CN)
HU ANMING (CN)
LIU GUANGHUI (CN)
SONG CHUANDONG (CN)
CAO ZHIHAO (CN)
JIANG ZHENFENG (CN)
Application Number:
PCT/CN2021/100226
Publication Date:
November 03, 2022
Filing Date:
June 16, 2021
Export Citation:
Assignee:
UNIV ZAOZHUANG (CN)
SHANDONG MINGYUAN INTELLIGENT EQUIPMENT TECH CO LTD (CN)
SHANDONG MINGYUAN INTELLIGENT EQUIPMENT TECH CO LTD (CN)
International Classes:
G06K9/62; G01D21/02; G06F17/18; G06N20/00
Domestic Patent References:
| WO2019129977A1 | 2019-07-04 |
Foreign References:
| CN112070125A | 2020-12-11 | |||
| CN112631155A | 2021-04-09 | |||
| CN109506963A | 2019-03-22 |
Attorney, Agent or Firm:
JINAN SPRINGS PATENT&TRADEMARK OFFICE (CN)
Download PDF: