Symbolization apparatus and process control system and control support system using the same apparatus

• US 5,265,222 A
• Filed: 11/23/1990
• Issued: 11/23/1993
• Est. Priority Date: 11/27/1989
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Yoda, Mikio, Kurisu, Hiromitsu, Funabashi, Motohisa, Nishiya, Takushi, Kera, Kazuo
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Downs, Robert W.
• Application Number:
  US07/617,347
• Time in Patent Office: 1,096 Days
• Field of Search: 364/513, 364/807, 395/900, 395/3, 395/22
• US Class Current: 706/10
• CPC Class Codes: G06N 3/043 based on fuzzy logic, fuzzy membership or fuzzy inference, e.g. adaptive neuro-fuzzy inference systems [ANFIS]
• CPC Class Codes: G06N 7/023 Learning or tuning the parameters of a fuzzy system
• CPC Class Codes: Y10S 706/90 Fuzzy logic
• CPC Class Codes: Y10S 706/906 Process plant

Abstract

A symbolization apparatus which includes a membership function memory unit for storing a membership function used in fuzzy logic processing. In the symbolization apparatus a process condition change pattern is converted into an ambiguous symbol in order to build up a system capable of effecting an inference and decision approximate to a skilled operator against an object process. A process condition value memory is disclosed for holding a measurement of the process condition value and a process operation amount, a control history memory unit for storing operations and values converted into symbolic expressions, control units and a knowledge base for storing the information relating to the casual relations between the process control values and the process conditions. The process control system and a control support system comprise the function of condition identification approximate to that of a skilled operator by the use of the symbolization apparatus.

Claims

1. A symbolization apparatus for use in a process control system for producign symbols representing process conditions of a process system, said apparatus comprising:

means for detecting process conditions of said process system during every moment of operation of said process system;

memory means for storing said detected process conditions;

means for sampling said process conditions stored in said memory means and for performing a sum-of-product calculation using feature extraction filters stored in a filter memory unit to obtain values indicating polynomial expansion coefficients; and

means for transforming said values indicating piolynomial expansion coefficients in said process conditions into symbolic expressions in a form of fuzzy logic expressions indicating qualitative process conditions according to a membership function.

2. A process control system for detecting process conditions and producing control values based on said detected process conditions to thereby control a process system in target process conditions, comprising:

means for detecting process conditions of said process control system during every moment of operation of said process system;

memory means for storing said detected process conditions;

means for sampling said process conditions stored in said memory means and for performing a sum-of-product calculation using fature extraction filters stored in a filter memory unit to obtain values indicating polynomial expansion coefficients;

means for transforming said values indicating polynomial expansion coefficients in said process conditions into symbolic expressions in a form of fuzzy logic expressions indicating qualitative process conditions according to a membership function;

knowledge base means for storing causal relations between said process conditions and process control values;

combination network means for determining a combination of symbolic expressions according to said causal relations stored in said knowledge base means;

symbol conversion means for receiving said combination of symbolic expressions from said combination network means and obtaining a weighted average of said combination of symbolic expressions as process control values using coupling weighting factors; and

means for converting said process control values into control signals supplied to said process system.

3. A process control system according to claim 2, further comprising:

means for storing history of said symbolic expressions from said transforming means and said process control values from said symbol conversion means; and

parameter learning means for revising said membership function and said coupling weighing factors in said symbol conversion means according to said stored history of said symbolic expressions and said process control values by an error back propagation learning method.

4. A process control system according to claim 3, further comprising:

restoration means for restoring said symbolic expressions indicating qualitative process conditions to a time-series of data of the process in a form of polygonal line;

display means for displaying said time-series of data;

means for inputting a model which expresses the operation of said process system;

means for inputting a target condition of said process system;

simulation means for simulating change of conditions of said process system based on said model; and

means for determining optimum control values for achieving an optimum condition of said process system by means of a result of said simulation by said simulation means and inputted said target condition.

5. A process control system for detecting process conditions and producing control values based on said detected process conditions to thereby control a process system in target process conditions, comprising:

means for detecting process conditions of said process control system during every moement of operation of said process system;

memory means for storing said detected process conditions;

means for sampling said process conditions stored in said memory means and for performign a predetermined calculation using sampled process conditions to obtain values indicating levels and trends in said process conditions;

means for transforming said values indicating levels and trends in said process conditions into symbolic expressions indicating qualitative process conditions according to a membership function;

means for determining process control values by selecting process conditions close to desired conditions according to said symbolic expressions indicating qualitative process conditions;

means for converting said process control values into control signals supplied to said process system

wherein said determining means comprises;

means for storing said symbolic expressions from said transforming means,

knowledge base means for storing causal relations between said process conditions and said process control values,

combination network means for determining combinations of said symbolic expressions according to said causal relations stored in said knowledge base means, and

symbol conversion means for receiving said combinatinos of said symbolic expressions from said combination network means and for making weighted average of said symbolic expressions as said process control values using coupling weighing factors;

means for storing history of said symbolic expressions from said transforming means and said process cotnrol values from said determining means;

means for correcting said coupling weighing factors in said symbol conversion means according to said stored history of said symbolic expressions and said process control values by means of an error back propagation learning method;

restoration means for restoring said symbolic expressions indicating qualitative process conditions to a time-series of data of the process in a form of polygonal line;

display means for displaying said time-series of data; and

model generation means for generating and expressing a model of said process system in a form of connecting state of the combination network by means of said causal relations stored in said knowledge base means, said model generation means having a structure identical to that of said determining means.

6. A process control system for detecting process conditions and producing control values based on said detected process conditions to thereby control a process system in target process conditions, comprising:

means for detecting process conditions of said process control system during every moment of operation of said process system;

memory means for storing said detected process conditions;

means for sampling said process conditions stored in said memory means and for performing a predetermined calculation using sampled process conditions to obtain values indicating levels and trends in said process conditions;

means for transforming said values indicating levels and trends in said process conditions into symbolic expressions indicating qualitative process conditions according to a membership function;

means for determining process control values by selecting process conditions close to desired conditions according to said symbolic expressions indicating qualitative process conditions;

means for converting said process control values into control signals supplied to said process system

wherein said determining means comprises;

means for storing said symbolic expressions from said transforming means,

knowledge base means for storing causal relations between said process conditions and said process control values,

combination network means for determining combinations of said symbolic expressions according to said causal relations stored in said knowledge base means, and

symbol conversion means for receiving said combinations of said symbolic expressions from said combination network means and for making weighted average of said symbolic expressions as said process control values using coupling weighing factors;

means for storing history of said symbolic expressions from said transforming means and said process control values from said determining means;

means for correcting said coupling weighing factors in said symbol conversion means according to said stored history of said symbolic expressions and said process control values by means of an error back propagation learning method;

means for inputting process control values by an operator;

simulation means for simulating change of conditions of a simulation model of a process system;

comparator unit for comparing said procesas conditions of said history in said history storing means with the result of said simulatino by said simulation means; and

display means for displaying a comparison result from said comparator unit to display said result to said operator.

7. A process control system for detecting process conditions and producing control values based on said detected process conditions to thereby control a process system in target process conditions, comprising:

means for detecting process conditions of said process control system during every moment of operation of said process system;

memory means for storing said detected process conditions;

means for sampling said process conditions stored in said memory means and for performing a predetermined calculation using sampled process conditions to obtain values indicating levels and trends in said process conditions;

means for transforming said values indicating levels and trends in said process conditions into symbolic expressions indicating qualitative process conditions according to a membership function;

means for determining process control values by selecting process conditions close to desired conditions according to said symbolic expressions indicating qualitative process conditions;

means for converting said process control values into control signals supplied to said process system

wherein said determining means comprises;

means for storing said symbolic expressions from said transforming means,

knowledge base means for storing causal relations between said process conditions and said process control values,

combination network means for determining combinations of said symbolic expressions according to said causal relations stored in said knowledge base means, and

symbol conversion means for receiving said combinations of said symbolic expressions from said combination network means and for making weighted average of said symbolic expressions as said process control values using coupling weighing factors;

means for storing history of said symbolic expressions from said transforming means and said process control values from said determining means;

means for correcting said coupling weighing factors in said symbol conversion means according to said stored history of said symbolic expressions and said process control values by means of an error back propagation learning method;

searching means for searching for past symbolic expressions of said process conditions and said process control values in said history storing means to extract said symbolic expressions most similar to present symbolic expressions; and

restoration means for restoring said symbolic expressions received from said searching means to a time-series of data of the process in a form of polygonal line.