Title:
LEARNING CONTROL METHOD
Document Type and Number:
Japanese Patent JP3243870
Kind Code:
B2
Abstract:
PURPOSE: To automatically design a control algorithm at a higher speed by performing the double freedom degree control, i.e., the feedforward control and the feedback control in order to compensate the influences dependent nonlinearly on the high speed responsiveness to a target temperature and the state of this temperature and to keep a stable control state respectively.
CONSTITUTION: A feedforward manipulated valiable 14 is calculated by a neural network according to the target temperature distribution 11 of a wafer 31 together with a feedback manipulated variable 17 of deviation 16 between the target temperature at the measuring point of the distribution 11 and a measured temperature. Then, a feedback controlled variable 18 is calculated from the information on similarity to the past experiences acquired through the temperature control of the wafer 31. A ring of these three variables 14, 17 and 18 is defined as a manipulated variable 13 for temperature control of the wafer 31. Thus, it is possible to compensate the influences dependent nonlinearly on the high speed responsiveness to the target temperature and the state of the target temperature by the feedforward control and to keep a stable control state by the feedback control respectively. So that the design of a control algorithm is facilitated.
More Like This:
| JPH04131905 | FUZZY CONTROLLER |
| JPH07113859 | [Title of Invention] A method for controlling a robot arm |
| JP2775968 | [Title of Invention] Process Control Device |
Inventors:
Masao Adachi
Masaru Matsushima
Masaru Matsushima
Application Number:
JP3012093A
Publication Date:
January 07, 2002
Filing Date:
February 19, 1993
Export Citation:
Assignee:
株式会社日立製作所
International Classes:
G05B13/02; G06F9/06; G06F15/18; G06G7/60; G06N99/00; (IPC1-7): G05B13/02; G06G7/60; G06N1/00
Domestic Patent References:
| JP527806A |
Attorney, Agent or Firm:
Yasuo Sakuta