Fuzzy inference-based digital control method and apparatus

• US 5,012,430 A
• Filed: 09/02/1987
• Issued: 04/30/1991
• Est. Priority Date: 09/03/1986
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Sakurai, Akito
• Primary Examiner(s): MacDonald, Allen R.
• Application Number:
  US07/092,137
• Time in Patent Office: 1,336 Days
• Field of Search: 364/148-151, 364/157-159, 364/162, 364/513, 364/131, 318/561
• US Class Current: 700/50
• CPC Class Codes: G05B 13/0275 using fuzzy logic only
• CPC Class Codes: G06N 7/04 Physical realisation
• CPC Class Codes: Y10S 706/90 Fuzzy logic
• CPC Class Codes: Y10S 706/903 Control

Abstract

A fuzzy inference-based digital control method and apparatus wherein state variables of the controlled object are adopted as inputs thereof, and fuzzy inference computations are executed for the input variables taken as fuzzy variables to derive the control output value to be supplied to the controlled object by referring to predetermined implication rules and membership functions corresponding thereto. The inference rule is divided into an inference rule group for deriving a value representing the state of the controlled object from the measured value obtained from the controlled object and another succeeding inference rule group for deriving the control output value from the derived state value. These inference rule groups are successively used to execute the inference computation in a multistep manner. An optimum control value for the controlled object is obtained with easy tuning.

Claims

1. In a digital control method including the steps of taking measurable state variable values of a controlled object as inputs or said state variable values and values computed from said state variables as inputs, computing values of fuzzy variables obtained from implication rules for fuzzy interference, expressed in the form of conditional statements, using said input variables as fuzzy variables, and controlling said controlled object according to the result of said computing, a fuzzy inference-based digital control method comprising, in said computing step:

a first step of computing fuzzy variables representing the state of the controlled object on the basis of selected ones of input measurable variable values in accordance with predetermined implication rules of a group regarding attributes of said controlled object; and

a second step of computing control variable values for the controlled object from inputs including at least the fuzzy variables determined by said first step, in accordance with predetermined implication rules of another group regarding a control output to be applied to the controlled object.

2. A fuzzy inference-based digital control method according to claim 1, wherein said implication rules of the first group are so prepared together with associated membership functions on the basis of an empirical law that unmeasurable variables representing attributes of the controlled object may be inferred from input state variable values, and wherein said implication rules of another group are so prepared together with associated membership functions on the basis of an empirical law that the output variable values sent to the controlled object may be derived from the attribute values of said controlled object.

3. A fuzzy inference-based digital control method according to claim 2, wherein each of the implication rules of said first group and said other group includes a plurality of IF-THEN statements.

4. A digital control method according to claim 1, wherein sets of inference rule groups and membership functions are stored in first and second storage means and include programs which are rewritable on the basis of control results supplied to the controlled object.

5. A method for applying digital control onto a controlled object on the basis of fuzzy inference, comprising:

a first step of inputting state variable values of the controlled object;

a second step of performing fuzzy inference computations to derive output variable values by making said input state variables into fuzzy variables and by referring to implication rules of a first group relating to attributes of the controlled object and corresponding membership functions associated with input variables;

a third step of using said output variable values as at least part of input fuzzy variable values of implication rules of the next group relating to a control output value to be applied to the controlled object and for deriving output variable values of said next group by performing fuzzy inference computations with respect to the implication rules of said next group and membership functions corresponding to input variables; and

a fourth step of deriving a control value to be applied to the controlled object on the basis of the output values of said first group and output values of said next group.

6. A digital control method according to claim 5, wherein the state variables of the controlled object include a plurality of kinds, and wherein said third step receives as another input at least a state variable of the controlled object which differs in kind from the input state variables of said second step.

7. A digital control method according to claim 6, further comprising the step of tuning said plurality of implication rules and membership functions to desired values on the basis of the result of the controlled object responding to the control output value derived at the fourth step.

8. A digital control method according to claim 5, wherein at least one of the second step and the third step further includes implication rules and membership functions of a plurality of subgroups, and cascade inference operations are executed for respective subgroups.

9. A digital control method according to claim 5, wherein the fuzzy inference rule group in the second step is so set beforehand as to derive the value representing the state of the controlled object, and the fuzzy inference rule group at the third step is so set beforehand as to derive the output value controlling the controlled object on the basis of the inferred value of the state value of the second step.

10. A digital control apparatus for controlling a controlled object on the basis of fuzzy inference, comprising:

input means for supplying measured values relating to the controlled object;

first storage means for storing therein inference rules of a first group preset in order to infer the states of the controlled object and membership functions associated with input variables;

first inference computation means for performing fuzzy inference computations with respect to the measured values supplied from said input means to derive state variables of the controlled object by referring to said inference rules and membership functions of the first group taken out from said first storage means;

second storage means for storing therein inference rules of a second group preset in order to infer control values for the controlled object on the basis of the state variables; and

second inference computation means, coupled to the output of said first inference computation means, for performing fuzzy inference computations with respect to at least the output values of said first inference computation means to derive control output values of the controlled object by referring to said inference rules and membership functions of the second group taken from said second storage means.

11. A digital control apparatus according to claim 10, wherein

said input means includes means for temporarily holding a plurality of kinds of measured values measured from the controlled object, and said input means supplies a measured value of a kind different from that of the measured value inputted to said first inference computation means to said second inference computation means.

state variable holding means for temporarily holding the outputs of said first inference computation means is coupled between said first inference computation means and said second inference computation means, and

output means connected to receive the outputs of said first variable holding means and the outputs of said second inference computation means for supplying said outputs as control outputs to the controlled object.