Data processing: artificial intelligence – Fuzzy logic hardware – Digital fuzzy computer
Reexamination Certificate
1997-05-29
2002-03-19
Davis, George B. (Department: 2122)
Data processing: artificial intelligence
Fuzzy logic hardware
Digital fuzzy computer
C706S047000, C706S900000
Reexamination Certificate
active
06360212
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention concerns a process for designing a fuzzy regulator with at least one input parameter and at least one output parameter.
The principles of fuzzy regulators are described, for example, in the article “Fuzzy Control—Heuristic Regulation by Means of Fuzzy Logic” by Hans-Peter Preu&bgr;, published in “atp” (1992), no. 4, pages 176-183 and “atp” (1992), no. 5, pages 239-246. The fuzzy regulators described there are used for open- and closed-loop control processes, for optimization of processes and for heuristic strategies by means of fuzzy logic. In fuzzy regulation, the fuzzy regulator may essentially have any desired number of input and output parameters.
To design a fuzzy regulator in accordance with the present invention, first the numerical value ranges of the input and output parameters are characterized qualitatively by linguistic values such as “small,” “medium” or “large.” Each linguistic value is described by the membership function which quantifies the qualitative information provided by a linguistic value in such a manner that it yields a truth value for each numerical value occurring for a process parameter. Through this process, which is also known as fuzzification, the operating range of the process in question is divided into “fuzzy” sub-ranges. The number of sub-ranges corresponds to the number of linguistic values of an input parameter, or when there are several input parameters, the number of sub-ranges corresponds to the number of possible combinations of linguistic values of different input parameters. The control strategy is determined by IF-THEN rules for each of these sub-ranges or for several sub-ranges together. These rules are stored in a rule base. For each of these rules, a conclusion is given in the conditional part as a linguistic value whereby a combination of linguistic values of the input parameter may be provided, for example, by a linkage with the “AND” or “OR” operators.
To calculate the truth value of the conditional part, the truth values obtained from the membership functions of the individual inputs are linked through the appropriate operators used in the rules. In the operation identified as “inference” of calculating the conclusions for the individual rules, for example the membership function of the output parameters, named as the corresponding linguistic value in a rule, is limited to the truth value supplied by the conditional part of the rule. In the operation called “composition,” the effects of the rules concerning an output parameter are mutually superimposed, e.g. by forming the maximum of all membership functions of the output parameters. Finally, the calculation of the output parameter, referred to as “defuzzification,” follows. This is done, for example, by calculating the position of the center of gravity of the surface defined by all the limited membership functions over the value range of the output parameters.
A rule base is incomplete if a rule is not expressly assigned to each of the aforementioned sub-ranges of the process. On the other hand, each linguistic value of the output parameter does not have to be specified in a rule as a conclusion. It is disadvantageous that output parameters that are unfavorable for the control may arise in operating states not covered by a rule.
SUMMARY OF THE INVENTION
An object of the present invention is to provide a process for a fuzzy regulator allowing an incomplete rule base to be supplemented in a simple manner and thus providing a fuzzy regulator that does not cause instabilities in the control circuit.
In fact, the truth value of the conditional part of the implicit rule is determined from the truth values of the conditional parts of the explicit rules by selecting the maximum value and forming the complement to the value 1 by using the equation:
W
j
=1−max (
W
i
)
where
i=1, . . . , n, i≠j,
n=number of linguistic values of an output parameter and
w
j
=truth value of the implicit rule for the j-th linguistic value of the output parameter.
In addition, the truth value of the conditional part of the implicit rule is reduced by an evaluation factor. The truth value of the conditional part of the implicit rule may be linked to the truth value of the conditional part of another explicit rule.
Furthermore, the truth value of the conditional part of the implicit rule is linked to the truth value of the conditional part of one of the explicit rules, and the result of this linkage is used as its new truth value. With the new design process, a fuzzy regulator having favorable characteristics in the control circuit is available.
REFERENCES:
patent: 5179634 (1993-01-01), Matsunaga et al.
patent: 5208898 (1993-05-01), Funabashi et al.
patent: 5239620 (1993-08-01), Yamakawa et al.
patent: 5440672 (1995-08-01), Araki et al.
patent: 5487130 (1996-01-01), Ichimori et al.
patent: 0 513 689 (1992-11-01), None
IEEE Transactions On Systems, Man And Cybernetics, vol. 20, No. 5, 1990, New York, US, pp. 1115-1124, Liang-Jong Huang et al.: A Self-Paced Fuzzy Tracking Controller for Two-Dimensional Motion Control.
Linzenkirchner Edmund
Preuss Hans-Peter
Davis George B.
Kenyon & Kenyon
Siemens Aktiengesellschaft
LandOfFree
Process for designing a fuzzy regulator does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Process for designing a fuzzy regulator, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Process for designing a fuzzy regulator will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2862101