Title:
PHYSICAL MODEL IDENTIFICATION SYSTEM
Document Type and Number:
WIPO Patent Application WO/2021/229727
Kind Code:
A1
Abstract:
A physical model identification system according to the present invention first uses real data collected from a production line to learn a statistical model representing a control target on the production line. Next, the system creates quantity data by quantifying an input/output relationship between a registered input variable and an objective variable of the learned statistical model. Next, the system identifies a correction coefficient for a physical model in such a manner that the relationship between the registered input variable and the objective variable represented by the quantity data is maintained. The identified correction coefficient is reflected on the physical model implemented on a computer controlling the production line. This improves the accuracy of prediction of a preset value and a control value of the control target by a practical model, thus enabling stable operation and high-quality production on the production line.
Inventors:
IMANARI HIROYUKI (JP)
Application Number:
PCT/JP2020/019135
Publication Date:
November 18, 2021
Filing Date:
May 13, 2020
Export Citation:
Assignee:
TOSHIBA MITSUBISHI ELEC IND (JP)
International Classes:
G05B13/04
Domestic Patent References:
| WO2014006681A1 | 2014-01-09 |
Foreign References:
| JP2011141798A | 2011-07-21 | |||
| JP2008040660A | 2008-02-21 | |||
| JP2006107256A | 2006-04-20 | |||
| JP2019191836A | 2019-10-31 | |||
| JPH07191712A | 1995-07-28 | |||
| JPH05293516A | 1993-11-09 | |||
| CN108655186A | 2018-10-16 |
Attorney, Agent or Firm:
TAKADA, Mamoru et al. (JP)
Download PDF: