Control system having optimality decision means

• US 5,303,385 A
• Filed: 03/09/1990
• Issued: 04/12/1994
• Est. Priority Date: 03/17/1989
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Manchu, Tetsuo, Ueki, Shigeru, Saito, Yutaka, Matsumoto, Hiroshi, Morooka, Yasuo, Hattori, Satoshi, Katayama, Yasunori, Nakajima, Masaaki
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/490,848
• Time in Patent Office: 1,495 Days
• Field of Search: 364/148, 364/472, 395/22, 395/900, 395/904, 395/906, 395/61, 395/62, 395/11
• US Class Current: 700/50
• CPC Class Codes: B21B 37/32 by cooling, heating or lubricating the rolls
• CPC Class Codes: B21B 37/42 using a combination of roll bending and axial shifting of the rolls
• CPC Class Codes: G05B 13/0285 using neural networks and fuzzy logic
• CPC Class Codes: Y10S 706/90 Fuzzy logic
• CPC Class Codes: Y10S 706/906 Process plant

Abstract

A control system for controlling a plurality of control elements, which are effective to control of a given controlled system, is disclosed. In this control system, a characteristic value representing the characteristic of a given controlled system to be controlled is detected, and respective optimum manipulated variables of the plurality of control elements are derived on the basis of the detected characteristic value. The optimum manipulated variables are derived by setting a plurality of reference patterns representing different features of the characteristic value on the basis of a predicted value of the characteristic value, deriving degree of conviction that the detected characteristic value belongs to each of the reference patterns, presetting rules whereby a manipulated variable of each of the plurality of control elements is determined when the detected characteristic value belongs to each of the plurality of reference patterns, and deriving a manipulated variable of each control element on the basis of degree of conviction that the detected characteristic value belongs to each of the plurality of reference patterns and the manipulated variable determining rule preset with respect to each control element for each reference pattern.

Claims

1. A control system for controlling a plurality of control elements which are effective to control of a given controlled system, said control system comprising:

detection means for detecting a characteristic value representing the characteristic of a given controlled system to be controlled; and

an optimality decision unit for deriving respective manipulated variables of said plurality of control elements on the basis of said detected characteristic value, said optimality decision unit comprising;

means for setting a plurality of reference patterns representing different features of said characteristic value on the basis of a predicted value of said characteristic value;

means for deriving degree of conviction that said characteristic value detected by said detection means belongs to each of said reference patterns;

means for presetting rules whereby a manipulated variable of each of said plurality of control elements is determined when said detected characteristic value belongs to each of said plurality of reference patterns; and

means for deriving a manipulated variable of each control element on the basis of degree of conviction that said detected characteristic value belongs to each of said plurality of reference patterns and said manipulated variable determining rule preset with respect to each control element for each reference pattern.

2. A control system according to claim 1, wherein said manipulated variable determining rule determines relation between said each reference pattern and range of manipulated variable of said each control element.

3. A control system according to claim 1, wherein said characteristic value represents a pattern given by an arrangement of a plurality of values of a parameter representing the characteristic of said controlled system, and said reference patterns are a plurality of predicted representative patterns of said characteristic value.

4. A control system according to claim 1, wherein said plurality of reference patterns specifies a plurality of different ranges with respect to the magnitude of said characteristic value, and a plurality of ranges overlap each other.

5. A control system according to claim 1, wherein a predetermined characteristic of said controlled system is a sectional shape of a rolled material which is rolled by a rolling mill, and the characteristic value detected by said detection means is given in the form of a pattern representing the sectional shape of said rolled material.

6. A control system according to claim 1, wherein a predetermined characteristic of said controlled system is a sectional shape of the rolled material which is rolled by the rolling mill, and the characteristic value detected by said detection means is thickness of the rolled material detected substantially simultaneously at a plurality of positions along the section of said rolled material.

7. A control system for controlling a predetermined characteristic of a subject which is controlled by a plurality of actuators, said control system comprising:

a plurality of detection means for detecting a plurality of different data relating to a given parameter representing the predetermined characteristic of said subject and for producing output signals used for control of each of said actuators in response to the detected different data; and

optimality decision means for synthetically judging a state of said characteristic of said subject on the basis of the output signals of said plurality of detection means in accordance with a predetermined rule and for supplying a control command for each of said plurality of actuators on the basis of a result of said judging;

wherein said optimality decision means comprises;

means for storing a plurality of reference patterns predetermined with respect to combinations of output signals of said detection means and for storing manipulated variables of said actuators corresponding to said plurality of reference patterns; and

means for deriving the degree of conviction for said each reference pattern with respect to an output pattern given by a combination of output signals of said detection means and for determining manipulated variables of said plurality of actuators on the basis of said degree of conviction and manipulated variables of said actuators corresponding to the reference pattern.

8. A control system according to claim 7, wherein said optimality decision means comprises:

pattern recognition means for deriving degree of similarity between a pattern in combination of output signals supplied from said detection means and each of a plurality of reference patterns predetermined with respect to said output signals; and

command generation means for determining individual manipulated variables of a plurality of actuators on the basis of said degree of similarity and for converting the determined manipulated variables to command signals of said plurality of actuators.

9. A control system according to claim 8, wherein said pattern recognition means comprises an input layer for taking in the combination of output signals of said detection means and a series connection of a plurality of states of intermediate layers, and said intermediate layer of each stage comprises a plurality of nodes for receiving output signals of said input layer or an intermediate layer of the preceding stage, for multiplying the output signals by weighting coefficients and adding resultant products, for converting the resultant sum by using a predetermined function, and for outputting the result, and wherein the output of the intermediate layer of the last stage indicates the degree of similarity of said pattern.

10. A control system according to claim 9, wherein said command generation means determines the manipulated variables for said actuators by using a predetermined knowledge base and an inference device on the basis of the degree of similarity of said pattern.

11. A system for controlling a predetermined characteristic of a given controlled system by controlling a plurality of actuators, said system comprising:

detection means for detecting a plurality of values of a parameter representing said predetermined characteristic and for outputting the values as a plurality of detected signals;

pattern recognition means for receiving a combination of said plurality of detected signals as a pattern and for deriving degree of similarity of the combination of said detected signals to each of a plurality of predetermined reference patterns;

means for processing said degree of similarity by using fuzzy inference and for determining the manipulated variable of said each actuator; and

means for generating a command of said each actuator on the basis of said manipulated variable.

12. A control system according to claim 11, wherein said pattern recognition means comprises a neurocomputer of Rumelhart type.

13. A control system according to claim 11, wherein said pattern recognition means comprises a plurality of nodes corresponding to said plurality of reference patterns, and each node performs calculation by using said plurality of inputted detection signals and a predetermined function comprising a weighting coefficient and outputs the degree of similarity between the combination pattern of said detected signals and the corresponding reference pattern, and wherein said control system further comprises learning means for adjusting said weighting coefficient so that outputs of said plurality of nodes may indicate the correct degree of similarity to the corresponding reference pattern.

14. A control system according to claim 13, said learning means comprises:

input pattern generation means for storing said plurality of reference patterns and for inputting one selected out of said reference patterns to said pattern recognition means;

output pattern generation means, whereby a favorable pattern of output signals is outputted from said plurality of nodes when the same pattern as said one reference pattern is inputted to said recognition means;

comparison means for comparing a pattern of signals outputted from said plurality of nodes upon inputting of said one reference pattern to said pattern recognition means with the output pattern of said output pattern generation means and for thereby deriving the deviation;

means for adjusting said weighting coefficients of said plurality of nodes on the basis of the result of said comparison; and

means for controlling the operation of said input pattern generation means, said output pattern generation means and said weighting coefficient adjustment means.

15. A control system according to claim 11, further comprising:

memory means for storing said plurality of detected signals as time series data; and

means for selectively inputting said plurality of detected signals and the output signal of said memory means to said pattern recognition means.

16. An optimality decision unit comprising:

input means for receiving a combination of input signals representing a predetermined characteristic of a controlled system controlled by a plurality of actuators;

means for judging if the state of said predetermined characteristic is satisfactory on the basis of said input signals and for determining manipulated variables of said plurality of actuators as a result of the judging; and

output means for outputting control signals to said plurality of actuators on the basis of said determined manipulated variables; and

wherein said manipulated variable determining means comprises;

means for representing the state of said predetermined characteristic by a pattern obtained by a combination of said input signals;

means for storing a plurality of reference patterns representing a representative shape of said pattern;

means for deriving the degree of similarity between the pattern obtained by the combination of said input signals and each of said reference patterns; and

means for determining the manipulated variable of each of said actuators on the basis of said degree of similarity.